New And Enduring Themes In Development Economics 9789812839428, 9789812839411

Citation preview

New and Enduring Themes in Development Economics

Statistical Science and Interdisciplinary Research Series Editor: Sankar K. Pal (Indian Statistical Institute) Description: In conjunction with the Platinum Jubilee celebrations of the Indian Statistical Institute, a series of books will be produced to cover various topics, such as Statistics and Mathematics, Computer Science, Machine Intelligence, Econometrics, other Physical Sciences, and Social and Natural Sciences. This series of edited volumes in the mentioned disciplines culminate mostly out of significant events — conferences, workshops and lectures — held at the ten branches and centers of ISI to commemorate the long history of the institute.

Vol. 1

Mathematical Programming and Game Theory for Decision Making edited by S. K. Neogy, R. B. Bapat, A. K. Das & T. Parthasarathy (Indian Statistical Institute, India)

Vol. 2

Advances in Intelligent Information Processing: Tools and Applications edited by B. Chandra & C. A. Murthy (Indian Statistical Institute, India)

Vol. 3

Algorithms, Architectures and Information Systems Security edited by Bhargab B. Bhattacharya, Susmita Sur-Kolay, Subhas C. Nandy & Aditya Bagchi (Indian Statistical Institute, India)

Vol. 4

Advances in Multivariate Statistical Methods edited by A. SenGupta (Indian Statistical Institute, India)

Vol. 5

New and Enduring Themes in Development Economics edited by B. Dutta, T. Ray & E. Somanathan (Indian Statistical Institute, India)

Vol. 6

Modeling, Computation and Optimization edited by S. K. Neogy, A. K. Das and R. B. Bapat (Indian Statistical Institute, India)

Sandhya - New and Enduring.pmd

2

5/8/2009, 11:27 AM

Platinum Jubilee Series

Statistical Science and Interdisciplinary Research — Vol. 5

New and Enduring Themes in Development Economics Editors

Bhaskar Dutta Indian Statistical Institute, India & Warwick University, UK

Tridip Ray Indian Statistical Institute, India

E. Somanathan Indian Statistical Institute, India

Series Editor: Sankar K. Pal

World Scientific NEW JERSEY

•

LONDON

•

SINGAPORE

•

BEIJING

•

SHANGHAI

•

HONG KONG

•

TA I P E I

•

CHENNAI

A-PDF Merger DEMO : Purchase from www.A-PDF.com to remove the watermark

Published by World Scientific Publishing Co. Pte. Ltd. 5 Toh Tuck Link, Singapore 596224 USA office: 27 Warren Street, Suite 401-402, Hackensack, NJ 07601 UK office: 57 Shelton Street, Covent Garden, London WC2H 9HE

British Library Cataloguing-in-Publication Data A catalogue record for this book is available from the British Library.

NEW AND ENDURING THEMES IN DEVELOPMENT ECONOMICS Statistical Science and Interdisciplinary Research — Vol. 5 Copyright © 2009 by World Scientific Publishing Co. Pte. Ltd. All rights reserved. This book, or parts thereof, may not be reproduced in any form or by any means, electronic or mechanical, including photocopying, recording or any information storage and retrieval system now known or to be invented, without written permission from the Publisher.

For photocopying of material in this volume, please pay a copying fee through the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA. In this case permission to photocopy is not required from the publisher.

ISBN-13 978-981-283-941-1 ISBN-10 981-283-941-0

Printed in Singapore.

Sandhya - New and Enduring.pmd

1

5/8/2009, 11:27 AM

FOREWORD The Indian Statistical Institute (ISI) was established on 17th December, 1931 by a great visionary Prof. Prasanta Chandra Mahalanobis to promote research in the theory and applications of statistics as a new scientific discipline in India. In 1959, Pandit Jawaharlal Nehru, the then Prime Minister of India introduced the ISI Act in the parliament and designated it as an Institution of National Importance because of its remarkable achievements in statistical work as well as its contribution to economic planning. Today, the Indian Statistical Institute occupies a prestigious position in the academic firmament. It has been a haven for bright and talented academics working in a number of disciplines. Its research faculty has done India proud in the arenas of Statistics, Mathematics, Economics, Computer Science, among others. Over seventy five years, it has grown into a massive banyan tree, like the institute emblem. The Institute now serves the nation as a unified and monolithic organization from different places, namely Kolkata, the Head Quarters, Delhi, Bangalore and Chennai, three centers, a network of five SQC-OR Units located at Mumbai, Pune, Baroda, Hyderabad and Coimbatore, and a branch (field station) at Giridih. The platinum jubilee celebrations of ISI have been launched by Honorable Prime Minister Prof. Manmohan Singh on December 24, 2006, and the Govt. of India has declared 29th June as the “Statistics Day” to commemorate the birthday of Prof. Mahalanobis nationally. Prof. Mahalanobis, was a great believer in interdisciplinary research, because he thought that this will promote the development of not only Statistics, but also the other natural and social sciences. To promote interdisciplinary research, major strides were made in the areas of computer science, statistical quality control, economics, biological and social sciences, physical and earth sciences. The Institute’s motto of ‘unity in diversity’ has been the guiding principle of all its activities since its inception. It highlights the unifying role of statistics in relation to various scientific activities. In tune with this hallowed tradition, a comprehensive academic programme, involving Nobel Laureates, Fellows of the Royal Society, Abel prize winner and other dignitaries, has been implemented throughout the Platinum Jubilee year, v

vi

Foreword

highlighting the emerging areas of ongoing frontline research in its various scientific divisions, centers, and outlying units. It includes international and national-level seminars, symposia, conferences and workshops, as well as series of special lectures. As an outcome of these events, the Institute is bringing out a series of comprehensive volumes in different subjects under the title Statistical Science and Interdisciplinary Research, published by the World Scientific Press, Singapore. The present volume titled “New and Enduring Themes in Development Economics” is the fifth one in the series. The volume consists of twenty chapters, written by eminent economists and experienced researchers from different parts of the world. The chapters deal with different aspects of development economics, e.g., inequality and well-being, interface between law and economics, political economy, labour economics, agriculture, macroeconomics and public finance. They represent a fine balance between theoretical and empirical studies addressing different challenging problems. Although some of the investigations are based on Indian data, the observations are of significance to other third world countries. I believe the state-of-the art studies presented in this book will be very useful to both researchers as well as policy makers. Thanks to the contributors for their excellent research contributions, and to the volume editors Profs. Bhaskar Dutta, Tridip Ray and E. Somanathan for their sincere effort in bringing out the volume nicely in time. Initial design of the cover by Mr. Indranil Dutta is acknowledged. Sincere efforts by Prof. Dilip Saha and Dr. Barun Mukhopadhyay for editorial assistance are appreciated. Thanks are also due to World Scientific for their initiative in publishing the series and being a part of the Platinum Jubilee endeavor of the Institute.

August 2008 Kolkata

Sankar K. Pal Series Editor and Director

PREFACE This volume is a selection of papers by participants at the ISI Platinum Jubilee conference on Comparative Development held at the ISI, Delhi in December 2007. All papers have been peer reviewed. The set of papers in the volume span an exciting set of both new and wellestablished topics in development economics, thus justifying the title of the volume. A brief discussion of a small selection of the papers indicates the range of topics covered. Rethinking aspects of well-being and happiness is one of the exciting new areas in development economics. John Helliwell, Haifang Huang and Anthony Harris examine the determinants of life satisfaction as measured by surveys of individuals. They show that income, measures of deprivation, and other personal and national circumstances are together good predictors of life satisfaction. In another paper in the section on inequality and wellbeing, T. Lakshmanaswami explores the intergenerational links in the transmission of economic advantage under assortative marital sorting. The interface between law and economics is another exciting area of research, at least partly because it is of great topical interest. For instance, public outrage arising from witness subversion and perversion of justice has reversed verdicts on appeal in several recent cases in India. Brendan O’Flaherty and Rajiv Sethi examine the role of public outrage in delivering justice when judicial effectiveness is poor. Some of the papers are in political economy, another very popular area of research. Abhirup Sarkar conducts a theoretical analysis of why in developing countries like India, political parties often call general strikes that are unlike strikes in factories – the former are often held as general protests without any specific economic goals. Suman Majumdar and Sharun Mukand examine the role of policy interventions in bringing about institutional change in a framework where existing political incentives determine a country’s economic institutions. T.N. Srinivasan’s paper goes back to an issue which has been hotly debated by development economists in recent times. What is the cause of the agrarian crisis in India in since the 1990s? Has it been caused by the opening up of the Indian economy to external competition? Or is it the result of India vii

viii

Preface

implementing the Agreement on Agriculture of the Uruguay Round of Multilateral Trade Negotiations? Or has it been caused by the decline of public investment in agriculture in response to rising fiscal deficits? Srinivasan argues that the principal reason for the agrarian crisis has been the faulty government policies prior to 1990 – in particular a development policy which placed far to much emphasis on import substituting industrialization with emphasis on heavy industry and insulation from the world economy. The volume also includes several papers on what can be broadly categorized as Labour economics. In a particularly interesting paper, Kaushik Basu, Gary Fields and Shub Dasgupta construct a theoretical model to analyze the effects of labour legislation which make it costly for firms to dismiss or retrench workers. They show that such laws can cause wages and employment to rise or fall, depending on the parametric conditions prevailing in the market. They then isolate conditions under which an anti-retrenchment law raises wages and employment. Other papers in the volume discuss several contemporary issues in areas such as macroeconomics and public finance. Several of the empirical papers are based on Indian data. But, of course, the lessons learnt from them are relevant for all developing countries. We are grateful to all those who helped make the conference and this volume a success, particularly our doctoral students Ashokankur Datta, Mridu Prabal Goswami, Namrata Gulati, Ridhima Gupta, Dushyant Kumar, Debdatta Saha, and Soumendu Sarkar who helped with the editing. Bhaskar Dutta Tridip Ray E.Somanathan

CONTENTS Foreword Preface

Inequality and Wellbeing International Differences in the Determinants of Life Satisfaction John F. Helliwell, Haifang Huang and Anthony Harris

3

Unequal Chances: The Intergenerational Transmission of Economic Advantage under Marital Sorting T. Lakshmanasamy

41

Addressing Equity Issues in Watershed Development Projects in Bhil Adivasi Areas of Western Madhya Pradesh Rahul Banerjee

57

Politics, Law and Economics A Theory of the Corrupt Keynesian Toke Aidt and Jayasri Dutta

93

On Effecting Institutional Change Sumon Majumdar and Sharun Mukand

113

On the Political Economy of General Strikes Abhirup Sarkar

133

Public Outrage and Criminal Justice: Lessons from the Jessica Lal Case Brendan O’Flaherty and Rajiv Sethi

145

ix

x

Contents

Lobbying for Trade Regime and Tariff Settings Katsuzo Yamamoto

165

Labour Labor Retrenchment Laws and Their Effect on Wages and Employment: A Theoretical Investigation Kaushik Basu, Gary S. Fields and Shub Debgupta

181

Work Migration in and Investment in Origin Communities Ghazala Mansuri

207

Labour Market Reform and Poverty: The Role of Informal Sector Sugata Marjit, Saibal Kar and Dibyendu Sundar Maiti

229

Measuring Harm due to Child Work and Child Labour: Patterns and Determinants for India Diganta Mukherjee and Saswati Das

241

Agriculture Development Strategy: The State and Agriculture Since Independence T.N. Srinivasan

267

Land-Use Changes and Agricultural Growth in India, Pakistan, and Bangladesh, 1901-2004 Takashi Kurosaki

303

Macro and Public Finance The Natural Interest Rate in Emerging Markets Ashima Goyal

333

Intergovernmental Transfer Rules, State Fiscal Policy and Performance in India Poulomi Roy and Ajitava Raychaudhuri

369

Contents

xi

Tax Evasion and Administrative Costs Rohit Prasad

401

Inequality, Public Investment and Deficits in India Errol D’Souza

429

Environment Climate Change and the Kyoto Protocol Parkash Chander

447

Finance Recent Trends in Microfinance Institutions: Some Theoretical Implications Suman Ghosh and Eric Van Tassel

467

This page intentionally left blank

,QHTXDOLW\DQG:HOOEHLQJ

This page intentionally left blank

INTERNATIONAL DIFFERENCES IN THE DETERMINANTS OF LIFE SATISFACTION JOHN F. HELLIWELL, HAIFANG HUANG AND ANTHONY HARRIS Canadian Institute for Advanced Research and Department of Economics University of British Columbia This paper uses measures of life satisfaction, drawn from the first wave of well-being data now being collected annually by the Gallup World Poll, to investigate differences across countries, cultures and regions, in the factors linked to differences in life satisfaction. We first examine two-level global and regional equations for life satisfaction covering 105 countries, paying special attention to several key variables: income, age, attainment of basic needs (as represented by presence of running water, and having sufficient food), having someone to count on, perceived corruption levels, and perceived sense of freedom. We then estimate and compare the coefficients of cross-sectional life satisfaction equations for each of the 105 countries. While many of the key variables affecting life satisfaction have quite comparable effects in almost all countries, we are nonetheless able to identify some interesting regional and other cross-country differences.

1. Introduction A previous paper (Helliwell14) argued that direct measures of life satisfaction provide a useful way to assess the quality of development within and across communities and nations. A case was made that previous doubts about the usefulness of comparing measures of subjective well-being across cultures and over time are being resolved in favour of subjective measuresa.

a

First, earlier claims that each person has a psychological set-point for subjective well-being to which he or she invariably returns (Brickman and Campbell2, Brickman et al3, Lucas et al19) have been replaced by research showing that adaptation to most changes in life circumstances is partial in nature (Lucas, Diener and Scollon6, Lucas18). Second, experimental evidence that retrospective assessments of well-being differ from Bentham-like (Kahneman, Wakker and Sarin17) integrals of momentary assessments (Kahneman15, Frederickson and Kahneman8, Kahneman and Riis16) was held not to threaten the usefulness of retrospective evaluations of satisfaction, especially as the latter, are what govern future decisions (Wirtz et al28). Third, in response to suggestions that freedom and capabilities, which were held to be of fundamental value to well-being (Sen25, 26), would be left out of account by measures of life satisfaction, it was shown that measures of life satisfaction appear to differ from assessments of positive and negative affect in just the ways that make life satisfaction an appropriate measure. Indeed, in this paper we show that a sense of personal freedom is highly significant as a support for higher measures of life satisfaction. 3

4

John F. Helliwell, Haifang Huang and Anthony Harris

The earlier paper compared results from the Gallup World Poll and the World Values Survey, focusing on modeling differences among nations in average scores from different measures of the quality of life. That paper also estimated two-level regressions based on individual data, and argued that most of the cross-country variance in survey measures of life satisfaction can be explained by measurable differences in life circumstances in those countries, under the assumption that people all over the world have similar basic preferences, and answer life satisfaction questions in roughly comparable ways. In this paper we dig further into the data to see to what extent the assumption of common preferences is justified. More particularly, we shall concentrate on using Gallup World Poll data on the quality of life to estimate cross-sectional life satisfaction equations in each of 105 countries. We shall then see to what extent the results on a country-by-country basis are consistent with the use of two-level analysis in which coefficients are assumed to be the same for residents in all countries. To set the stage for this analysis of the nature of and possible reasons for cross-country differences in the determinants of life satisfaction, we shall first present some results for global estimates of several different equations, and then regional estimates of the same equation that we shall then estimate separately for each country. 2. Two-Level Global and Regional Estimates Assuming Similar Preferences Our key dependent variable comprises individual answers, from roughly 1,000 respondents in each of 105 countriesb, to a question asking respondents to evaluate their lives at present using a ladder with steps numbered from zero at the bottom to 10 at the top, with 0 representing the worst possible life and 10 the best possible life. This is the Cantril ladder form for measuring the quality of life. Comparisons in Helliwell14 between results from the Gallup data, using this question, and from the World values Survey asking respondents to assess satisfaction with their lives as a whole on a scale of 1 to 10 suggest that the different ways of asking the question and framing the scale make some difference to the results, perhaps contributing to the greater correlation between income and life satisfaction in the Gallup data. However, even with the slightly b

The wave of Gallup World poll we use covers 129 countries/regions. Twenty of them however do not report household income. In addition we were not able to find per capita GDP for Myanmar. Three additional countries/regions are excluded from the sample because of other missing information.

International Differences in the Determinants of Life Satisfaction

5

different question forms, different sample selection and administration techniques, different survey years, and different structures and ordering of questions, the two bodies of data gave strikingly consistent views of the factors contributing to life satisfaction around the world. For 75 countries in both surveys, the simple cross-country correlation of average life satisfaction scores was 0.8, suggesting that there are consistent cross-country differences in life satisfaction that are stable through time and robust to alternative survey techniques. Analysis in the earlier paper also showed that the high cross-country correlation of life satisfaction scores was not just due to both being highly correlated with average per capita incomes, because the residual cross-country variance of life satisfaction beyond that explained by income was also significantly correlated between the surveys. The basic observations are at the individual level, and we are interested in estimating the extent to which individual life satisfaction depends on circumstances and events at the individual, household, community and national levels. We have developed three inter-related ways of unravelling the data. The first, which we emphasize in this paper, is to use the individual-level data in equations that are separate for each country. A second is to measure and account for international differences in life satisfaction using national average data, and a third is to use multi-level analysis to explore individual-level and higher-level correlates simultaneously. The second and third approaches were the focus of a previous paper (~elliwell'~), while in this paper we concentrate on the estimation and interpretation of separate equations for each country. Before proceeding to the analysis of country-by-country coefficients, we shall first use the two-level model to estimate data pooled either globally or by geographic region. The basic estimation form for the two-level analysis of the ordered life satisfaction responses is:

where LSij is life satisfaction for respondent i in country j, measured on a scale of 0 to 10, yij is the level of household income of the respondent, the Xj are other individual or household-level variables, and the Zj are national-level variables, with the same value being used for all individual observations in country j. We use the log form for both household and national average income, to reflect standard economic assumptions and many empirical results suggesting that less affluent agents derive greater utility from extra income. In general, we employ national-level variables for which we also have household-level

John F. Helliwell, Hagang Huang and Anthony Harris

observations, in which case the y coefficients represent contextual effects, or, in other terms, the extent of positive or negative externalitiesc. In all equations robust standard errors are estimated assuming errors to be clustered by country. When we calculate compensating differentials for non-financial determinants of life satisfaction, we take into account the functional form of equation (1). Thus in our theoretically and empirically preferred case where income is in log form and X is in linear form, P= p/ 6 will be the log change in income that has for the average respondent the same life satisfaction effect as a change in the non-financial life characteristic X. Table 1 shows life satisfaction equations using individual-level data from samples ranging fkom about 70,000 to 83,000 respondents in 105 countries, with the smaller sample sizes resulting from missing observations for some variablesd. Table 1 starts with the simplest equation in the first column, where life satisfaction is determined by gender, age (in quadratic form), marital status, and the logarithm of household income. We then sequentially add two measures of basic needs (running water and enough money for adequate food), a measure of social connectedness (having someone to count on), a measure of trust in institutions (each individual's assessments of corruption in business and government), and a measure of the individual's sense of freedom to choose. The exact wording of each question is shown in Table 5.

*

We do not include national level values for gender, the age variables, and marital status. Although there are some differences among countries and regions in population age structure and marital status, experiments adding the national averages to equation 1 do not reveal significant effects or materially alter the sizes of other coefficients. The lack of response to the question of household income is responsible for most of the reduction in sample size.

Add Count on help -0.084 [0.022]** -0.035 [0.005]** 0.031 [0.005]** 0.074 [0.047] -0.18 [0.054]** 0.441 [0.027]** 0.234 [0.052]** -0.618 [0.034]** 0.544

Adjust for sample† -0.094 [0.025]** -0.038 [0.005]** 0.034 [0.005]** 0.083 [0.048] -0.157 [0.057]** 0.434 [0.026]** 0.203 [0.056]** -0.606 [0.036]** 0.55

Add Corruption -0.097 [0.025]** -0.037 [0.005]** 0.034 [0.005]** 0.079 [0.048] -0.156 [0.057]** 0.432 [0.026]** 0.208 [0.056]** -0.601 [0.036]** 0.55

Add Freedom -0.1 [0.025]** -0.036 [0.005]** 0.032 [0.005]** 0.072 [0.048] -0.156 [0.058]** 0.426 [0.026]** 0.195 [0.053]** -0.579 [0.037]** 0.524

[0.036]**

[0.035]**

[0.035]**

[0.033]**

-0.233 [0.042]**

-0.194 [0.043]** 0.39 [0.037]** -0.122 [0.089] 0.5 [0.298]

Perception of corruption Freedom to choose Log of GDP per capita, PPP Average: Running water

-0.254 [0.089]** 0.642 [0.282]*

-0.152 [0.091] 0.405 [0.297]

-0.135 [0.091] 0.421 [0.296]

-0.131 [0.089] 0.475 [0.296]

-0.13 [0.089] 0.467 [0.296]

International Differences in the Determinants of Life Satisfaction

Table 1: Determinants of well-being, global sample Regression method: weighted survey linear regression with countries as PSU Demographics Add Basic and income need -0.095 -0.093 Male [0.023]** [0.023]** -0.041 -0.038 Age [0.005]** [0.005]** 0.037 0.034 Age squared/100 [0.005]** [0.005]** 0.053 0.074 Married or as if married [0.047] [0.047] -0.221 -0.19 Separated, divorced [0.054]** [0.054]** or widowed 0.564 0.461 Log of household income†† [0.027]** [0.027]** 0.252 Home has running water [0.052]** -0.67 Not enough money for [0.036]** food in last 12 months Has someone to count on

7

8 John F. Helliwell, Haifang Huang and Anthony Harris

-0.739 -0.074 -0.137 -0.076 -0.082 -0.093 [0.367]* [0.358] [0.357] [0.342] [0.342] [0.342] 1.884 1.879 1.332 1.297 1.295 1.302 [0.630]** [0.626]** [0.631]* [0.615]* [0.615]* [0.619]* -0.974 -0.983 -0.982 -1.051 -0.819 -0.862 [0.333]** [0.330]** [0.328]** [0.324]** [0.326]* [0.327]** 1.676 1.682 1.686 1.713 1.713 1.305 [0.434]** [0.433]** [0.432]** [0.434]** [0.434]** [0.442]** 4.79 4.688 4.598 4.678 4.676 4.682 Constant [0.780]** [0.779]** [0.778]** [0.773]** [0.773]** [0.775]** 83219 83219 83219 69801 69801 68210 Observations 0.3 0.32 0.32 0.32 0.32 0.33 R-squared 105 105 105 105 105 105 Number of nations Standard errors in brackets, * significant at 5%; ** significant at 1% †: A sizable portion of respondents have missing values in perception of corruption regarding business, or regarding government or both. The adjustment of sample size simply restraint the regressions on those who have valid response to both perception questions. ††: The individual household incomes in the Gallup data are divided by their country means to get relative incomes within each country. These figures are then converted into common level form by adding the resulted relative income to the average GDP per capita in 2003 measured at Purchasing Power Parity (from Penn World table 6.2) †††: The following countries are in the Gallup data we use but not included in the regression due to missing information: Egypt; Iran; Morocco; Lebanon; Saudi Arabia; Jordan; Turkey; Pakistan; China (Peoples Republic); West Bank & Gaza (Palestine); Philippines; Laos; Myanmar(Burma); Belarus; Kyrgyzstan; Moldova; Ukraine; Croatia; Cuba; Iraq; Kuwait; Macedonia (Republic Of); United Arab Emirates; Average: Not enough money for food Average: Has someone to count on Average: Perception of corruption Average: Freedom to choose

International Differences in the Determinants of Life Satisfaction

9

Table 5: Exact wordings of key questions in Gallup World Poll WP16: Life today; 0-10 point scale Please imagine a ladder/mountain with steps numbered from zero at the bottom to ten at the top. Suppose we say that the top of the ladder/mountain represents the best possible life for you and the bottom of the ladder/mountain represents the worst possible life for you. If the top step is 10 and the bottom step is 0, on which step of the ladder/mountain do you feel you personally stand at the present time? WP27: Count on help; yes/no binary response If you were in trouble, do you have relatives or friends you can count on to help you whenever you need them, or not? WP134: Freedom to choose; yes/no binary response In this country, are you satisfied or dissatisfied with your freedom to choose what you do with your life? Corrupt: Perception of corruption; Corrupt=(WP145+ WP146)/2 WP145: Is corruption widespread within businesses located in this country, or not? yes/no

WP146: Is corruption widespread throughout the government in this country, or not? yes/no

WP40: Not enough money for food; yes/no binary response Have there been times in the past twelve months when you did not have enough money to buy food that you or your family needed? WP33: Running water; yes/no binary response Does your home or the place you live have running water?

Age effects are estimated by a quadratic form in age; in all cases there is a general U-shape, with some variation among country groupings. Marital status is divided into three categories: married or equivalent, single, and a combination of divorced, separated and widowed, with single being treated as the base case in estimation. The log of household income is a very strong correlate of individual life satisfaction in the Gallup equations, even when the equation adds, as in column 2, responses to other life-circumstance questions determined at least partly by income: running water and having enough money for food. As shown in Table 2 that divides the sample by regions, the income coefficient is if anything higher in the richer countries (as previously noted by Deaton4) and shows no obvious tendency to drop as individual income rises, beyond the non-linearity implied by the logarithmic form for income. As shown in Helliwell14 the correlation between income and life satisfaction is higher with the Gallup World Poll data

10

John F. Helliwell, Haifang Huang and Anthony Harris

than with World Values Survey (WVS) data. It is possible that the Gallup life satisfaction question, taking the form of a Cantril ladder, invites respondents to think in relative terms more than when they are simply asked, as in the World Values Survey, to assess their life satisfaction on a scale running from 1 to 10. The current version of the Gallup survey is asking the question in both forms, to help answer this question. Attempts are also being made to expand the number of income categories, and thereby to give more income variation in the group of higher income earners. The WVS captures this already, by choosing income categories to match income deciles, and the next rounds of the Gallup survey should be able to provide at least this amount of income detail. Table 2: Determinants of well-being, regressions by regions Regression method: weighted survey linear regression with countries as PSU Latin Western America Eastern Europe, and Europe N. America, Caribbean Asia & FSU Australia & N.Z. Male -0.215 -0.053 -0.221 -0.092 [0.036]** [0.029] [0.061]** [0.079] Age -0.039 -0.058 -0.061 -0.042 [0.009]** [0.010]** [0.009]** [0.009]** Age squared/100 0.042 0.047 0.049 0.044 [0.009]** [0.009]** [0.009]** [0.009]** Married or as if 0.183 0.106 0.139 0.138 married [0.049]** [0.087] [0.079] [0.112] Separated, -0.165 -0.012 -0.107 -0.096 divorced or widowed [0.058]* [0.097] [0.130] [0.118] Log of household 0.555 0.482 0.47 0.428 income†† [0.059]** [0.053]** [0.047]** [0.041]** Home has running 0.221 0.167 0.294 0.22 water [0.311] [0.077]* [0.113]* [0.119] Not enough -0.628 -0.709 -0.626 -0.714 money for food in last 12 [0.107]** [0.062]** [0.082]** [0.041]** months Has someone to 0.886 0.582 0.616 0.443 count on [0.098]** [0.063]** [0.097]** [0.080]**

Africa -0.012 [0.038] -0.001 [0.009] -0.001 [0.011] -0.027 [0.065] -0.148 [0.100] 0.29 [0.038]** 0.28 [0.089]** -0.385 [0.057]** 0.428 [0.045]**

International Differences in the Determinants of Life Satisfaction Perception of corruption Freedom to choose Log of GDP per capita, PPP Average: Running water Average: Not enough money for food Average: Has someone to count on Average: Perception of corruption Average: Freedom to choose Constant Observations R-squared Number of nations

11

-0.267

-0.301

-0.149

-0.063

-0.094

[0.057]**

[0.102]**

[0.091]

[0.085]

[0.112]

0.525

0.523

0.372

0.221

0.289

[0.087]**

[0.058]**

[0.084]**

[0.069]**

[0.074]**

0.231

-0.429

-0.287

0.16

-0.257

[0.449]

[0.184]*

[0.182]

[0.134]

[0.110]*

16.2

-0.641

1.999

-0.41

0.463

[13.879]

[0.874]

[0.746]*

[0.594]

[0.525]

1.289

-2.052

-0.607

-0.887

-0.249

[0.894]

[1.118]

[1.455]

[1.031]

[0.645]

3.346

-0.782

-0.416

-1.045

1.648

[3.517]

[1.878]

[1.983]

[1.898]

[0.569]**

-1.004

1.81

-0.058

0.678

-0.376

[0.437]*

[1.035]

[1.227]

[0.848]

[0.768]

0.811

1.81

3.058

1.89

0.896

[2.434] -12.713 [14.022] 13388 0.19

[1.006] 5.213 [1.364]** 12907 0.21

[1.146]* 3.541 [2.176] 11870 0.16

[1.602] 6.193 [1.334]** 11192 0.25

[0.643] 2.737 [1.293]* 18304 0.13

21

20

22

15

26

Standard errors in brackets, * significant at 5%; ** significant at 1% ††: The individual household incomes in the Gallup data are divided by their country means to get relative incomes within each country. These figures are then converted into common level form by adding the resulted relative income to the average GDP per capita in 2003 measured at Purchasing Power Parity (from Penn World table 6.2)

As might be expected, the coefficient of log income drops when the basic needs are included. The drop of the income coefficients suggests some form of additional non-linearity of the income effect, with the coefficient on log income in the equation including basic needs reflecting the value of income for aspects of life beyond running water and the presence of enough money for food.

John F. Helliwell, Haifang Huang and Anthony Harris

12

The idea that income should matter more for those whose basic needs have not been met can be tested alternatively by dividing the sample in two: the richer part including those with enough money for adequate food, and the poorer part, including all those who have sometime in the past year not had enough money for food. The results are shown in Table 3. Somewhat surprisingly, the coefficient on log income is slightly lower in the group whose basic food needs have sometimes not been met. To some extent this is probably due to the higher incidence of food poverty in Africa, where the estimated income coefficient is systematically below that in other regions. This in turn may be due to the greater prevalence of subsistence living in Africa, and the lesser relevance of money income. Money income may well be a less adequate measure of full income in Africa than in other regions. Table 3: Global sample split according to whether basic needs for food have been met Regression method: weighted survey linear regression with countries as PSU Male Age Age squared/100 Married or as if married Separated, divorced or widowed Log of household income Home has running water Has someone to count on Perception of corruption Freedom to choose Log of GDP per capita, PPP Average: Running water Average: Not enough money for food

Basic needs met† -0.133 [0.025]** -0.044 [0.005]** 0.04 [0.005]** 0.164 [0.043]** -0.094 [0.049] 0.442 [0.028]** 0.235 [0.067]** 0.513 [0.045]** -0.229 [0.040]** 0.37 [0.040]** -0.161 [0.093] 0.65 [0.306]* -0.082 [0.359]

Basic needs not met† -0.042 [0.044] -0.021 [0.009]* 0.015 [0.009] -0.081 [0.067] -0.214 [0.085]* 0.391 [0.035]** 0.162 [0.065]* 0.551 [0.044]** -0.094 [0.087] 0.417 [0.048]** -0.179 [0.100] 0.425 [0.351] -0.355 [0.462]

International Differences in the Determinants of Life Satisfaction Average: Has someone to count on Average: Perception of corruption Average: Freedom to choose Constant Observations R-squared Number of nations

1.466 [0.850] -0.751 [0.295]* 1.652 [0.520]** 4.203 [0.892]** 47618 0.29 105

13

1.197 [0.540]* -0.796 [0.582] 0.679 [0.521] 4.143 [1.017]** 20592 0.17 105

Standard errors in brackets, * significant at 5%; ** significant at 1% †: The two columns in the table are split-sample estimations according to respondents' answer to whether there have been times in the past twelve months when they did not have enough money to buy food that they or their family needed.

There are other interesting differences between the equations for the two groups of respondents divided according to food adequacy. First, there is a significant female life satisfaction advantage among those reporting adequate food, but this is absent for those reporting inadequate food. This suggests that women are more likely than men to bear the psychological brunt of the consequences of having inadequate food for the family. Second, for those with inadequate food, the life satisfaction benefit of marriage is less, and the negative consequences of separation, divorce or widowhood are greater, in each case relative to being unmarried. Part of the gains from marriage probably flow from having greater capacity to provide for basic needs.e Third, those reporting lack of enough money for food are also much less likely to report having family or friends they can count on in times of trouble, to a much greater extent than can be explained by inter-country differences in this measure of social connectedness.f This suggests the lack of social connection is associated with a lack of food adequacy regardless of what countries the respondents live in. One possible explanation of the correlation is that the absence of social networks strong enough to provide assistance in times of need is likely to lead to a greater chance of basic needs going unmet. Additional support for this possibility is e

f

These patterns in coefficient differences persist within most regions when the sample is split by food adequacy. Therefore the difference in regional weightings in the global sample split by food adequacy are not responsible for the coefficient differences. Thus the average values of having someone to count on are 0.89 for those with food adequacy and 0.74 for those without enough money for food. By comparison, the country averages of having someone to count on for the same two sub samples are 0.86 and 0.80, respectively. These data are from the summary statistics following the Table.

14

John F. Helliwell, Haifang Huang and Anthony Harris

shown by the fact that the coefficients of having someone to count on and of the country averages of having someone to count on are both more significant in the smaller low-food-adequacy sample. Likewise, events such as warfare might occur that simultaneously destroy social capital and reduce access to basic needs. Returning to the discussion of basic needs, the second and subsequent columns in Table 1 and all equations in Table 2 show strong life satisfaction effects from the two measures of basic needs. While both running water and lack of enough money for food attract roughly similar coefficients in the separate equations for respondents in different regions, the summary statistics following Table 2 show that the actual prevalence of these measures of poverty is far higher in Africa, and far lower in Western Europe, North America, Australia and New Zealand, grouped together as region 1. For example, running water is found in the homes of 99% of the region 1 respondents and 73% of Asian respondents, in contrast to only 19% of those in Africa. By comparison, lack of enough money for food was reported by 9% of the region 1 respondents, compared to 21% in Asia and 55% in Africa. The next variable added, in the third and subsequent columns of Table 1, and all of the regional equations in Table 2, is a measure of positive social connections, as represented by whether the respondent has relatives or friends they can count on to help them whenever help is needed. In all parts of the world, most respondents report that they have family or friends they can count on, ranging from 75% in Africa to 81% in Asia and 91% in the countries of region 1. And in all regions this social support is tightly linked to life satisfaction, with a coefficient that exceeds that on log income in every region except Asia, where the two coefficients are roughly equal. This implies incomeequivalent life satisfaction for social connections that are very high indeed. It would appear that respondents in Western Europe, North America, Australia and New Zealand are richer in social as well as economic terms than those living elsewhere, and attach even higher values to such social support. The coefficient on having someone to count on is 0.88 (t = 9.1), twice as high as in Asia and in Africa (approximately 0.44 in both regions, t = 5.5, 9.5 in Asia and Africa, respectively). Some other variables indicative of personal or community-level social capital are available for only a subset of the Gallup respondents. But they all show the high values attached to mutually supportive social connections. For example, as shown in Helliwell14 those who think that their lost wallets would be retuned by a neighbour or the police are more satisfied with their lives (by

International Differences in the Determinants of Life Satisfaction

15

0.15 and 0.22 points), as are those who express confidence in the police (0.22). Respondents appear to value not only the support they get from others, but their own support for others. For instance, those who in the last month had donated money or time to an organization, or aided a stranger needing help were systematically more satisfied with their lives, especially for donations (0.30) and helping a stranger (0.16), as shown by the second equation in Table 4a of Helliwell14. The 4th column in Table 1 reduces the sample size to that for which observations are available for the variables added in columns 5 and 6. No significant coefficient changes result from using the smaller sample size, and the number of countries remains 105. Column 5 adds the average of each individual’s binary assessments of whether corruption in their country is widespread in business and government. For the global sample on average, an individual who thinks that corruption is widespread in business and government has life satisfaction that is lower by 0.23 points, about half the size of the coefficients on income and having family or friends to rely on. Table 2 shows that the estimated effects are largest for those living in the transition countries (Russia and Eastern Europe) and countries of region 1, and lower in Latin America, Asia and Africa. The perceived prevalence of corruption is highest in the transition countries (0.88) and lowest in region 1 (0.50). Regional averages for Asia, Latin America and Africa range from 0.77 to 0.80. There is a large variation among countries in the level of perceived corruption, both within and across regions, with Russia at .94, New Zealand at 0.22 and Singapore at 0.20. Table 4 provides a more precise way of testing for inter-regional differences in coefficients. The equation is estimated using region 1 as the base case, and tests for coefficients in other regions that are different from those in region 1. The key significant differences, as was already suggested in Table 2, and will appear also in the Figures 1 - 6 are: income coefficients are lower in Africa; effects of social connection are lower in Asia and Africa and higher in region 1 than in any other region; the effects of corruption are less in Asia and Africa, especially Asia (and, as with social connections, are greater in region 1 than in any other region), and the effects of a sense of personal freedom are also lower in Asia and Africa, and highest in region 1. Thus for social connections, corruption and a sense of personal freedom, the effects are smallest in Africa and Asia and largest in region 1.

16

John F. Helliwell, Haifang Huang and Anthony Harris Table 4: Differences in determinants of well-being across regions Regression method: weighted survey linear regression with countries as PSU Global Sample Male -0.102 [0.024]** Age -0.036 [0.005]** Age squared/100 0.032 [0.005]** Married or as if married 0.101 [0.039]* Separated, divorced or widowed -0.123 [0.050]* Log of household income 0.54 [0.063]** Home has running water 0.25 [0.048]** Not enough money for food in last 12 months -0.614 [0.093]** Has someone to count on 0.91 [0.103]** Perception of corruption -0.382 [0.096]** Freedom to choose 0.586 [0.090]** Average: Running water 0.159 [0.303] Average: Not enough money for food -0.452 [0.431] Average: Has someone to count on 0.754 [0.573] Average: Perception of corruption -0.368 [0.312] Average: Freedom to choose 0.813 [0.380]* Interaction: Hshld income x Region 2† -0.181 [0.078]* Interaction: Hshld income x Region 3 -0.036 [0.079] Interaction: Hshld income x Region 4 -0.148 [0.080] Interaction: Hshld income x Region 5 -0.281 [0.073]** Interaction: Enough food x Region 2 -0.149 [0.127] Interaction: Enough food x Region 3 -0.067 [0.129]

International Differences in the Determinants of Life Satisfaction Interaction: Enough food x Region 4 Interaction: Enough food x Region 5 Interaction: Someone to count on x Region 2 Interaction: Someone to count on x Region 3 Interaction: Someone to count on x Region 4 Interaction: Someone to count on x Region 5 Interaction: Corrupt x Region 2 Interaction: Corrupt x Region 3 Interaction: Corrupt x Region 4 Interaction: Corrupt x Region 5 Interaction: Freedom to choose x Region 2 Interaction: Freedom to choose x Region 3 Interaction: Freedom to choose x Region 4 Interaction: Freedom to choose x Region 5 Region 2 (E. Europe & FSU) dummy Region 3 (L. America & Caribbean) dummy Region 4 (Asia) dummy Region 5 (Africa) dummy Constant Observations R-squared Number of nations

-0.07 [0.107] 0.217 [0.118] -0.306 [0.121]* -0.233 [0.150] -0.525 [0.121]** -0.464 [0.120]** 0.128 [0.142] 0.196 [0.158] 0.501 [0.153]** 0.289 [0.143]* -0.112 [0.116] -0.14 [0.134] -0.357 [0.115]** -0.296 [0.118]* -0.362 [0.223] 0.387 [0.259] -0.118 [0.271] -0.471 [0.285] 5.516 [0.679]** 68210 0.34 105

Standard errors in brackets, * significant at 5%; ** significant at 1% †: Variables starting with "Interaction" are interactive terms with regional dummies. In this table, Western Europe, US, Canada, Australia & NZ are used as the comparator group. Region dummies: 1: W. Europe, N. America, Australia & N.Z.; 2: E. Europe & FSU; 3: Latin America and Caribbean; 4: Asia; 5: Africa

17

John F. Helliwell, Haifang Huang and Anthony Harris

18

0

Coefficients of log income .5 1

1.5

Figure 1: International difference in determinants of well-being: log of household income

1

2

Regions

3

4

Regions

Mean

Std. Dev.

1

0.56

0.24

2

0.54

0.22

3

0.53

0.32

4

0.46

0.17

5

0.28

0.16

1: Western Europe, U.S. Canada, Australia & Nz 2: Eastern Europe & FSU 3: Latin America and Caribbean 4: Asia 5: Africa

5

International Differences in the Determinants of Life Satisfaction

19

-2

Coefficients of basic need for food -1.5 -1 -.5 0

.5

Figure 2: International difference in determinants of well-being: not enough money for food

1

2

Regions

3

4

Regions

Mean

Std. Dev.

1

-0.66

0.39

2

-0.74

0.26

3

-0.60

0.41

4

-0.62

0.30

5

-0.37

0.25

1: Western Europe, U.S. Canada, Australia & Nz 2: Eastern Europe & FSU 3: Latin America and Caribbean 4: Asia 5: Africa

5

John F. Helliwell, Haifang Huang and Anthony Harris

20

0

Coefficients of count on help .5 1 1.5

2

Figure 3: International difference in determinants of well-being: count on help

1

2

Regions

3

4

Regions

Mean

Std. Dev.

1

0.77

0.37

2

0.62

0.32

3

0.73

0.57

4

0.39

0.26

5

0.41

0.22

1: Western Europe, U.S. Canada, Australia & Nz 2: Eastern Europe & FSU 3: Latin America and Caribbean 4: Asia 5: Africa

5

International Differences in the Determinants of Life Satisfaction

21

-1

Coefficients of corruption perception -.5 0 .5

1

Figure 4: International difference in determinants of well-being: perception of corruption

1

2

3

Regions

Mean

Std. Dev.

1

-0.31

0.22

2

-0.28

0.36

3

-0.25

0.41

4

-0.11

0.29

5

-0.17

0.44

Regions

1: Western Europe, U.S. Canada, Australia & Nz 2: Eastern Europe & FSU 3: Latin America and Caribbean 4: Asia 5: Africa

4

5

John F. Helliwell, Haifang Huang and Anthony Harris

22

-.5

Coefficients of freedom to choose 0 .5 1

1.5

Figure 5: International difference in determinants of well-being: freedom to choose

1

2

Regions

3

4

Regions

Mean

Std. Dev.

1

0.55

0.30

2

0.52

0.24

3

0.34

0.36

4

0.25

0.28

5

0.19

0.28

1: Western Europe, U.S. Canada, Australia & Nz 2: Eastern Europe & FSU 3: Latin America and Caribbean 4: Asia 5: Africa

5

International Differences in the Determinants of Life Satisfaction

23

20

Coefficients of freedom to choose 40 60 80 100

120

Figure 6: International difference in determinants of well-being: turning point of the age-wellbeing curve†

1

2

3

Regions

Mean

Std. Dev.

1

45.95

19.02

2

52.29

12.11

3

61.95

20.34

4

47.02

11.42

5

44.93

16.20

Regions

4

5

1: Western Europe, U.S. Canada, Australia & Nz 2: Eastern Europe & FSU 3: Latin America and Caribbean 4: Asia 5: Africa †: Figure 6 and its summary statistics exclude countries that have turning points being negative or greater than 125. Six countries are excluded as the result.

24

John F. Helliwell, Haifang Huang and Anthony Harris

To convert any of these effects into an income-equivalent value requires division by the estimated income coefficient. The smallest compensating differentials for non-financial aspects of life are obtained by using the income coefficient obtained if the other income-related variables (food and water) are removed from the equation. This gives an income coefficient of just over 0.5, slightly smaller than the coefficient on having someone to rely on, as estimated in that same equation. Thus having someone to rely on has a life satisfaction effect roughly ten times larger than a 5% change in income (i.e. 102) negative coefficients in 81 of the 105 individual country regressions, while the running water variable, which has much more of its variance between countries, has significant positive coefficients in only 26 of the 105 regressions. The social support variable has significant positive coefficients in 69 of the regressions, and the corruption variable in 35. The quadratic pattern of age effects is almost universal, with 89 countries having coefficients that are negative on age and positive on age squared. The gender effect for males is negative in 78 of the 105 countries, although significantly so in only 23. The other demographic variables are also fairly weakly defined in the national samples, reflecting the small sample sizes and the variety of individual experiences. The log of household income is positive in 103 of the 105 country regressions, and significantly so in 91 cases. This is so even though the equations contain two other income-dependent variables: adequacy of money for food, and running water in the home. For all variables the means of the country coefficients are very close to the values estimated in Table 1, as would be expected if the national samples were drawn from a global population with broadly similar responses to these variables. Figure 13 provides a graphic example of the cross-national consistency of parameters estimates by showing the individual coefficients for India, juxtaposed with those for Asia as a whole, and for the global sample.j Finally, it is necessary to address more directly the experimental (e.g. Heine and Norenzayen11) and other evidence (e.g. Kahneman15, Diener and Suh5, Kahneman and Riis16) that cross-national comparisons of retrospective assessments of subjective well-being are rendered difficult or possibly uninformative by cultural differences in the ways in which questions are interpreted, scales are used, values are determined and answers are framed (Heine et al12, Schmidt and Bullinger27). What is meant by culture in this context? Matsumoto21 defines culture as “a dynamic system of rules – explicit and implicit – established by groups in order to ensure their survival, including j

The global coefficients are from the final equation in Table 1, the Asia coefficients from Table 2, and the India coefficients from the single-country equation for India, drawn from the group of 105 for which the coefficients are shown in Figures 1 and 2.

John F. Helliwell, Haifang Huang and Anthony Harris

36

attitudes, values, beliefs, norms, and behaviours …communicated across generations, relatively stable but with the potential to change across time.” This bears striking similarities to the OECD22 definition of social capital (Putnam23, 24, Halpern10 ) as “networks together with shared norms, values and understandings that facilitate co-operation within and among groups”. In international research into the well-being consequences of differences in the quality of social capital (Helliwell and Putnam13), it is presumed that key aspects of social norms (e.g. trust) can be meaningfully measured and compared across cultures and over time. The use of pooled international samples with constant coefficients implies also that the well-being consequences of different levels of trust, for example, are comparable across cultures. Figure 13: Cross-national comparisons of determinants of well-being

0.9 0.7 0.5 0.3 India

0.1

Asia Global

-0.1 -0.3 -0.5 -0.7 -0.9 Log of household income

Home has running water

Not enough money for food in last 12 months

Has Perception of Freedom to someone to corruption choose count on

International Differences in the Determinants of Life Satisfaction

37

Figure 14: Well-being: measured and predicted

9 8 7 6 5 4 3 2 1 0 Russia

Canada

Life today

Denmark

United States

Thailand

India

Predicted life today

Coefficients

India

Asia

Global

Log of household income

0.33

0.428

0.426

Home has running water

0.35

0.22

0.195

-0.699

-0.714

-0.579

Not enough money for food in last 12 months Has someone to count on

0.643

0.443

0.524

Perception of corruption

0.035

-0.063

-0.194

Freedom to choose

0.291

0.221

0.39

Standard errors Log of household income

0.057

0.041

0.026

Home has running water

0.106

0.119

0.053

Not enough money for food in last 12 months

0.128

0.041

0.037

Has someone to count on

0.118

0.08

0.033

Perception of corruption

0.166

0.085

0.043

Freedom to choose

0.136

0.069

0.037

Log of household income

5.789474

10.43902

16.38462

Home has running water

3.301887

1.848739

3.679245

Not enough money for food in last 12 months

-5.46094

-17.4146

-15.6486

Has someone to count on

5.449153

5.5375

15.87879

t-statistics

Perception of corruption

0.210843

-0.74118

-4.51163

Freedom to choose

2.139706

3.202899

10.54054

38

John F. Helliwell, Haifang Huang and Anthony Harris

Our research and results suggest that some of the key inter-cultural differences in norms and values emphasized in the literature are supported in the subjective well-being data of the Gallup World Poll. For example, the wellbeing costs of living in a society with high perceived levels of corruption in business and government appear to be slightly less in countries where corruption is a long established feature of the status quo. Similarly, the wellbeing value attached to a sense of personal freedom is slightly higher in societies classed as individualistic rather than collectivist. But while these differences qualitatively confirm some key experimental cross-cultural findings, what appears to us remarkable is that application of the same well-being equation to 105 different national societies shows the same factors coming into play in much the same way and to much the same degree. This is illustrated by Figure 14, which shows actual and predicted values of life satisfaction obtained by applying the same model, with coefficients restricted to be the same for all countriesk. Thus the international differences in predicted values are entirely due to differences in their underlying circumstances. 4. Conclusion We have estimated the same life satisfaction equation for 105 countries. The results are strikingly consistent among countries, cultures and regions. Thus it would appear that the large international differences in life satisfaction are not due to differences in underlying preferences but to identifiable differences in life circumstances. Since these results are based on a single cross-section within each of the countries, with samples averaging less than 1,000, new rounds of survey evidence should enable more precise estimates within each country, testing of a broader range of underlying models, and first attempts to disentangle some of the complex dynamics that underlie the cross-sectional correlations reported here. Acknowledgments This paper is part of the ‘Social Interactions, Identity and Well-Being’ research program of the Canadian Institute for Advanced Research, and is also supported by grants from the Social Sciences and Humanities Research Council of Canada.

k

The equation is that shown in the right-hand column of Table 1.

International Differences in the Determinants of Life Satisfaction

39

References 1. C.P. Barrington-Leigh, and J.F. Helliwell, Empathy and Emulation: Subjective Well-Being and the Geography of Comparison Groups.” Paper presented at the Annual Meetings of the Canadian Economics Association, Halifax (June 2007). 2. P. Brickman and D.T. Campbell, Hedonic Relativism and Planning the Good Society.” In M.H. Appley, ed., Adaptation Level Theory: A Symposium. (New York: Academic Press) 287-302 (1971). 3. P. Brickman, D. Coates and R. Janoff-Bullman, Journal of Personality and Social Psychology, 36, 917 (1978). 4. Angus Deaton, Income, Aging, Health and Well-Being around the World: Evidence from the Gallup World Poll, NBER Working Paper No. 13317 (2007). 5. E. Diener and E.M. Suh, eds., Culture and Subjective Well-Being. (Cambridge MA: MIT Press) (2000). 6. E. Diener, R.E. Lucas and C.N. Scollon, American Psychologist, 61(4), 305 (2006). 7. Richard Easterlin, Does Economic Growth Improve the Human Lot? Some Empirical Evidence, In P.A. David and M. Reder, eds., Nations and Households in Economic Growth (New York: Academic Press), 89-125 (1974). 8. B.L. Frederickson and D. Kahneman, Journal of Personality and Social Psychology, 65, 45 (2003). 9. Gallup Organization, The State of Global Well-Being (New York: Gallup Press) (2007). 10. David Halpern, Social Capital (Cambridge: Polity Press) (2005). 11. S. Heine and A. Norenzayan, Perspectives on Psychological Science, 1(3), 251 (2006). 12. S. Heine et al., Journal of Personality and Social Pyschology, 82(6), 903 (2002). 13. John F. Helliwell and Robert D. Putnam, Phil Trans R. Soc Lon. B, 359, 1435 (2004). Reprinted in F.A. Huppert, B. Kaverne and N. Baylis, eds., The Science of Well-Being. (London: Oxford University Press, 2005, 43559). 14. John F. Helliwell, Life Satisfaction and Quality of Development, (2008). 15. D. Kahneman, Objective Happiness, In E. Diener, N. Schwarz and D. Kahneman, eds., Well-Being: The Foundations of Hedonic Psychology (New York: Russell Sage) (1999). 16. D. Kahneman and Jason Riis, Living, and Thinking About it: Two Perspectives on Life, In F.A. Huppert, N. Baylis and B. Keverne, eds., The Science of Well-Being (Oxford: Oxford University Press) (2005).

40

John F. Helliwell, Haifang Huang and Anthony Harris

17. D. Kahneman, P.P. Wakker, and R. Sarin, Quarterly Journal of Economics 112, 375 (1997). 18. Richard E. Lucas, Journal of Personality and Social Psychology, 92(4), 717 (2007). 19. R.E. Lucas, A.E. Clark, Y. Georgellis and E. Diener, Journal of Personality and Social Psychology, 84(3), 527 (2003). 20. Erzo F.P. Luttmer, Quarterly Journal of Economics, 963 (January, 2005). 21. D. Matsumoto, Culture and Psychology. (2nd ed.,Pacific Grove: Brooks Cole) (2000). 22. Organization for Economic Co-operation and Development, The WellBeing of Nations: The Role of Human and Social Capital. (Paris) (2001). 23. Robert D. Putnam, Making Democracy Work (Princeton: Princeton University Press) (1993). 24. Robert D. Putnam, Bowling Alone: The Collapse and Revival of American Community (New York: Simon and Schuster) (2000). 25. A. Sen, Development as Capability Expansion, In K. Griffin and J. Knight, eds., Human Development and the International Development Strategy for the 1990s. (London: Macmillan) (1990). 26. A. Sen, Development as Freedom. (New York: Knopf Press) (1999). 27. S. Schmidt, and M. Bullinger, Cross-cultural Quality of Life Assessment Approaches and Experiences from the Health Care Field, In Gough and McGregor, eds. Well-Being in Developing Countries: from Theory to Research. (Cambridge, Cambridge University Press) 219-41 (2007). 28. D. Wirtz, J. Kruger, C.N. Scollon, and E. Diener, Psychological Science, 14(5), 520 (2003).

UNEQUAL CHANCES: THE INTERGENERATIONAL TRANSMISSION OF ECONOMIC ADVANTAGE UNDER MARITAL SORTING *

T. LAKSHMANASAMY

Inequality perpetuates in the absence of intergenerational mobility. The intergenerational transmission of socioeconomic advantage negates the norm of equality of opportunity. This inequality in opportunities or unequal chances is further strengthened under marital sorting, as assortative marriage matches the affluent with the rich and vice versa. Parental investments in the human capital of children are in part motivated by the prospects of attracting a better partner for their children. Parents care about the earnings capacity or the combined income of the married partners while deciding the marriage of their children. The status of his/her own parents as well as that of the parents of his/her partner influences the economic status of children. This paper explores the intergenerational links in the transmission of economic advantage under such marital sorting. We estimate the intergenerational transmission of education and income between parents and children, and between parents-in-law and daughters-in-law and sons-in-law. The estimated intergenerational transmission is quite substantial and more pronounced in the case of daughters. The elasticity of children’s income is strongly positive with respect to in-laws income. The empirical results for education also reveal strong positive assortative mating pattern. Spousal earnings is just as elastic as the children’s own earnings with respect to the parents and in-laws education and income

1. Introduction Public expenditure on education, it is generally argued, is to increase equality of opportunity. One of the benefits of education is the spillover effects on the later generations; having more educated citizens will have longer run effects by improving the outcomes of the children. However, there is little causal evidence to suggest this is true. The literature on intergenerational mobility tries to explain the transmission of the socioeconomic status of parents to those of children. It particularly analyses the effects of parental education and income on the children’s economic outcomes as adults. Most commonly, this is measured as the association of incomes across generations. Sometimes the relationship may be a mere selection: the type of parents who have more education earn higher income, and have children who will do so as well. Alternatively, the *

Professor, Department of Econometrics, and Director i/c, Centre for Population Studies, University of Madras, Chennai. Email: [email protected]. 41

42

T. Lakshmanasamy

relationship may be more causation: more education makes the parent a different type of person, and thus leads to his children having higher educational and earnings outcomes. Goldberger (1989) is quite explicit in emphasizing the distinction between endowments (genes) and investments (environments) in explaining intergenerational mobility. Data show that children from richer families enjoy more human capital investments and earnings. Becker and Tomes (1979; 1986) present an economic model of intergenerational transmission that takes into account nonlinearities and credit constraints, which has been used and extended to derive several important predictions. Mulligan (1999) finds a strong positive correlation between child’s earnings and parental income after controlling for measures of human investment. The strong association between incomes across generations indicates weak income mobility, and is often regarded as a violation of the norms of equality of opportunity. If an individual’s income is strongly related to his or her parent’s income, then a child from poor family background has limited opportunities to escape his or her own start in life (Blanden, 2005). Under equal opportunity conditions, the expected earnings of children are independent of parental earnings (income). Ermisch et al. (2006) find that about 40–50 percent of the covariance in income between generations attributable to assortative mating in Germany and Britain, driven by the strong spousal correlations in human capital. Thus, marital sorting plays a crucial role in that parents invest on children with expectation of better future marital partner. Consequently, the continuing unequal chances perpetuates under intergenerational transmission of economic advantage. The persistence of income inequality across generations also leads to the unequal distribution of educational attainments. Thus, intergenerational persistence is expected if income-earning endowments have an inherited component. Additional intergenerational persistence will occur if capital markets are imperfect and there are greater returns to human capital and marital matching. 2. Empirics of Intergenerational Mobility The early literature on income mobility has observed an elasticity of son’s earnings with respect to father’s earnings around 0.2, or even less (Becker and Tomes, 1986). More recent studies use long-run measures from longitudinal survey data. These studies suggest that the elasticity between the permanent components of son’s and father’s earnings is about 0.4 (Solon, 1999). Bjorkland and Jantti (1997) and Bjorkland et al. (2002) find greater income mobility in US and Scandinavian countries, more so in the latter. The evidence also suggest that

Intergenerational Transmission of Economic Advantage under Marital Sorting

43

the more compressed is the income distribution, the smaller the correlation between parental and child outcomes (Black et al. 2003). Bratberg et al. (2005), exploring the relationship between income inequality and mobility in Norway, argue that, over time, the compression of the earnings structure has increased intergenerational earnings mobility. Similar studies on intergenerational transmission of education are rather scanty. Studies by Deardon et al. (1997) and Mulligan (1999) suggest intergenerational education elasticity between 0.20 and 0.45. Raaum et al. (2001) also find similar results for Norway. Eide and Showalter (1999) adopt a quantile regression approach to investigate the role of education as an earnings transmission mechanism across generations. Recently, Hirvonen (2007) examined the issue from gender perspective by taking into consideration the extent to which education influence the transmission of earnings between parents, and daughters and sons. Using quantile regression, the possible non-linearity in the transmission of economic advantage from one generation to another is also examined. While Eide and Showalter (1999) find that education is more valuable at the bottom and tends to compress the son’s conditional income distribution, the Swedish data of Hirvonen (2007) show that education is more valuable at the upper end than at the bottom tail of both daughters’ and sons’ conditional income distribution. In general, education explains just about one third of the intergenerational income correlation. More recent research in this area attempts to distinguish causation from mere correlation in ability across generations. Generally, three broad approaches have been used. Behrman and Rosenzweig (2002) use data on pairs of identical twins to difference out any correlation attributable to genetics. The OLS estimates suggest a positive and significant relationship (0.13) between mother’s and children’s schooling. Plug (2002) and Sacerdote (2002) use data on adopted children to investigate the causal relationship and finds a positive effect of father’s education on child education, but no significant effect of the mother’s. The third approach is to use instrumental variables. Chevalier (2003) uses a change in the compulsory schooling law in Britain in 1957 to identify the effect of parental education on their offspring education and finds a positive effect of mother’s education on her child’s education. Similarly, Black et al. (2003) use the 1959 Norwegian reform in school education that increased the number of compulsory schooling from 7 to 9 years as an instrument for parental education. Despite significant OLS relationships, the 2SLS estimates provide little evidence of a causal relationship between parent’s education and children’s education. This leads them to conclude that the high correlation between parental and children’s education are due primarily to selection rather than

44

T. Lakshmanasamy

causation. Holmlund (2007) uses the education reform in Sweden in the 1950s and 1960s which extended compulsory education from 7 to 9 years. The differences-in-differences and sibling-difference estimates indicate significant intergenerational income mobility. The bulk of the research has estimated the average transmission of earnings across generations, basically by applying OLS or IV in the regression of son’s earnings on the conditional mean of the father’s earning. The observed nonlinearities in intergenerational earnings are often explained by credit constraints, as parents are constrained by the possibility to finance education of their children (Becker and Tomes, 1986). Recent analyses by quantile regression method show that intergenerational effects to be somewhat higher for lower earnings, implying the intergenerational mobility is lower at the lower end of the earnings distribution than at the upper end, for sons as well as for daughters (Bratberg et al. 2007). While the UK, US and some other European countries appear to have high levels on intergenerational income transmission, the Scandinavian countries and Canada appear rather mobile by comparison (Corak, 2004). Similar results are also reported by Raaum et al. 2007), who focusing on the role of gender and marital status, confirms that earnings mobility in Nordic countries is typically greater than in the US and in the UK. However, for married women mobility is uniform across countries for women’s own earnings, while the usual differences hold for family earnings. Bratberg et al. (2007) show that the pattern of intergenerational earnings mobility in the Nordic countries as highly nonlinear, while in the US and UK the relationship is much close to linear. Blanden et al (2007) explain the UK intergenerational income persistence in terms of noncognitive traits, education and labour market attachment Unfortunately, most of the studies on intergenerational income mobility neglect the mobility of daughters and the influence of mother’s earnings on daughters. Though the intergeneration transmission studies on education do consider the influence of mother’s education, they ignore the impact of income. Presumably, such neglects have stemmed from the view that, in societies in which married women’s labour force participation rates are lower than men’s, women’s earnings (and incomes) are likely to be an unreliable measure of their status. Further, the studies on the role of spousal education and earnings in intergenerational mobility of daughters are rather scanty. However, studies on labour market participation of married women show a high degree of spousal correlations and that the education and earnings predict the status women have in the society. A recent work by Chadwick and Solon (2002) considers daughter’s intergenerational income mobility using family income (rather than

Intergenerational Transmission of Economic Advantage under Marital Sorting

45

just earnings) and husband’s earnings, thus providing a link between assortative marriage and intergenerational mobility. Also Blanden (2005) finds for Canada that the intergenerational correlation for sons with partners is 0.185 compared with 0.168 for daughters with partners. This new direction has implications for marital relations and for the unpaid work of women in the household sector. Recently, Fernandez et al. (2004) finds evidence for a direct positive relationship between a mother-in law’s working during her son’s childhood and the probability of daughter-in-law working. In line with the significant intergenerational mobility in Nordic countries, Holmlund (2007) also finds that in Sweden the effect of marital sorting on intergenerational income mobility as rather negligible. Thus, marital sorting seems to play a key role in shaping intergenerational family income persistence (Raaum et al. 2007). Evidences show that assortative mating is equally important for men and women for understanding how family earnings are transmitted across generations. None of these studies incorporate the significant relationship between the income and education of in-laws and the children’s education and earnings. This relationship is so important in an environment of assortative marriage market, in which the partners are matched with complementary characteristics. A considerable body of research show that there is systematic positive sorting of partners with respect to socioeconomic backgrounds and market characteristics, non-wage incomes, and possibly wages (Becker, 1991; Lam and Schoeni, 1994). Since parents invest substantially on children’s human and financial capital, the prospective life partners of children are to be matched to a large extent by the prospective in-laws background. Those sons and daughters raised by own parents eventually become someone’s spouse, and the way in which this matching occurs has important consequences for their own socioeconomic position. Accordingly, how inequality evolves over generations depends on who marries whom. On an empirical observation in Brazil, Lam and Schoeni (1994) interpret the greater effect of father-in-law’s schooling than that of father’s schooling on the wages of male workers as an indication of the high degree of assortative mating in the marriage market. The empirical evidences show that the in-laws relationship is strong and in no case is the parent to son-inlaw/daughter-in-law’s elasticity substantially below the parent to son or daughter elasticity (Blanden, 2005). Thus, both parents and parents-in law shape their offsprings’ status and hence their intergenerational mobility. This paper examines the extent to which assortative mating influences intergenerational transmission and presents evidence on the role of in-laws characteristics on the intergenerational mobility

46

T. Lakshmanasamy

of daughters and daughters in-law in India. We also consider the role of spousal background along with in-laws background for sons and daughters as well as sons-in-law and daughters-in-law in the intergenerational mobility. We find that the intergenerational transmission of education status is stronger than that of income, but the latter is also quite substantial. We also find that assortative mating is an important element in the intergenerational transmission process. This has important implications for the provision of equality of opportunity through public education. 3. A Model of Intergenerational Transmission Following Ermisch et al. (2006), let the parents (and in-laws) care about the expected joint income of their adult offspring, which is the expected sum of their child’s and his/her future partner’s incomes E(yt + ytp), besides their own consumption Ct–1.Then the utility function is given by U = φ ln[E(yt + ytp)] + (1 – φ)ln(Ct–1)

φ∈(0,1)

(1)

where y’s represent income, t indicates the generation, p the partner, φ measuring the relative preference for child’s future family income as against parent’s own consumption. Incomes are assumed to increase with human capital: yt = γ01 + γ1Ht + et

(2)

ytp = γ02 + γ2Htp + etp

(3)

where γ1 and γ2 are non-negative parameters. These income equations allow the returns to human capital to differ between the sexes. The matching function under assortative marriage market can be specified as Htp = α0 + α1Ht + vtp

(4)

which links own and potential partner’s human capital. Parents choose Ht to maximize their utility subject to equations 2 to 4 and their own budget constraint yt–1 = Ct–1 + λHHt

(5)

where λH is the relative unit price of child’s human capital. Solving this problem implies that the optimal level of child’s human capital is a linear function of parent’s income. Then the child’s income equation is

Intergenerational Transmission of Economic Advantage under Marital Sorting

47

yt = β0 + β1yt–1 + u1t

(6)

where β1 = φγ1/λH and the partner’s income is given by ytp = δ0 + δ1yt–1 + u2t

(7)

where δ1 = α1φγ2/λH. From the definitions of these parameters, it follows that δ1/β1 = α1 γ2/γ1. If the income returns are to human capital are the same for men and women (γ1 = γ2), then the ratio δ1/β1 identifies α1, the degree of assortative mating on human capital in (4). The model has implications for the relationship between child’s family (joint) income (yt + ytp), and that of his/her parents, whereby cov(yt + ytp,yt–1) = cov(yt,yt–1) + cov(ytp,yt–1) = (β1 + δ1)var(yt–1)

(8)

The contribution that assortative mating makes to the intergenerational mobility is taken to be μ = cov(ytp, yt–1)/cov(yt + ytp, yt–1)

(9)

μ = [δ1/(δ1 + β1)] = [α1γ2/(γ1 + α1γ2)]

(10)

which is given by

It is straightforward to see that μ decreases with β1 and increases with δ1, and (10) implies that α1 = (γ1/γ2)[μ/(1 – μ)]. 4. Empirical Analysis The intergenerational persistence of economic status is the result of the assortative mating process in which the ‘likes’ marry the ‘likes’ (Becker, 1991). It is a character-specific mate selection which would not have occurred by chance or random process. Most of the models closely follow the intergenerational mobility model of Lam and Schoeni (1993, 1994) and Chadwick and Solon (2002). While Ermisch et al. (2006) and Chadwick and Solon (2002) are concerned with the intergenerational mobility by exploiting the relationship between the son’s earnings and his father’s and father-in-law’s education, Lam and Schoeni (1993, 1994) explores the effects of father’s education on wages as representing the impact of inherited characteristics and the effect of father-in-law’s education as the correlation with the uninherited attributes through assortative mating. Let the intergenerational determination of a child’s earnings be

48

T. Lakshmanasamy

lnyi = β0 + βiyj + εI i = son(s), daughter(d), son-in-law(sl), daughter-in-law(dl) j = father(f), mother(m), father-in-law(fl), mother-in-law(ml)

(11)

where yi denotes the earnings of ith child, yj denotes the earnings of jth parent, and the slope coefficient βi is the intergenerational elasticity of ith child’s earnings (income) with respect to jth parent’s earnings (income). This is positive with assortative mating. The assortative mating can be summarized by a correlation ρ between the earnings of g and k: ρ = Corr(lnyg,lnyk) g = son or daughter; k = daughter-in-law or son-in-law

(12)

Then, the βk, representing the elasticity is βk = βgρ√[var(ln yk)/var(ln yg)

(13)

Thus, if there is no assortative mating on earnings (ρ = 0), this elasticity is zero. With positive assortative mating on earnings, this elasticity is positive (Altonji and Dunn, 1991). Further, this has implications for the connection between the two families. Let ω denote the k’s share of the family’s combined earnings (where family income is comprised entirely of k and g (spousal earnings), then the elasticity of g’s family (joint) income with respect to that of his/her parents is β = ωβk + (1 – ω)βg

(14)

the share-weighted average of the separate elasticities of the g’s own earnings and his/her spouse’s. If there is no assortative mating, so that βk = 0, and if the male earnings are greater than the female earnings, then the daughter’s (son’s) family income is much less (more) elastic with respect to her parent’s (his inlaws) income than her (his) own earnings are. Suppose that assortative mating is very positive, and βk is just as large as βg. Then in the typical family, in which ω is much more than half, the association between the daughter’s (g’s) family income and that of her (his) parents is mostly accounted for by her husband’s (wife’s) earnings. Similar arguments apply for the case of education also. The assortative mating process implies that matching is made to suit the spousal career advancement. In terms of the intergenerational transmission model, better educated parents tend to provide more education to their offspring and search

Intergenerational Transmission of Economic Advantage under Marital Sorting

49

for similar educated life partners. Thus the matching process is not random, but selective. Estimation of the model parameters requires data that provide information on the socioeconomic position of individuals, their partners, and their parents and in-laws. As such type of data is not readily available, in the empirical analysis on intergenerational mobility we use a primary survey data collected during September 1996 - March 1997 in Tamil Nadu as part of a larger project on the marriage and determinants of age at marriage of females. A total of 1014 respondents have been selected from both urban and rural areas, of which 566 (55.8 percent) are married and 448 (44.2 percent) are unmarried females. A special questionnaire has been designed separately for married women and for unmarried females. Both the questionnaires seek information about age, age at marriage, education, occupation and household characteristics of the respondents as well the parents. The married female questionnaire further seeks information on husband, in-laws, marriage related aspects and post–marital behaviour of the household. The unmarried female questionnaire seeks information on marital search and marriage related expectations. Table-1 presents the distribution of the educational background of the sample households. Most females are secondary educated, followed by degree course; compared to the children’s educational level, parents are less educated; a sizable number of female parents are illiterate. Compared to them, more male parents are either primary or secondary educated. Very few parents as well as children are higher educated. The educational backgrounds of male child are better compared to their female counterparts. A sizable male children are college educated. Thus, though parents are less educated, their children are better educated. This shows the educational mobility of children. The descriptive statistics presented in Table-2 shows that the mean educational level of daughters is 12.39 years, compared to the 8 years of their mothers and 9.4 years of education of father. The male children education of 14.42 years is also higher than their parental education. The spousal education of children is thus higher than their in-laws, again supporting intergenerational mobility. A similar picture also holds for children’s annual income compared to their parental and in-laws household income. Daughter’s earnings are higher than mother and mother-in-law’s earnings and male children earnings are also higher than parental or in-laws earnings. Moreover, male earnings are higher than their spousal earnings in all cases. These results reinforce the educational background results that the mating process is assortative and that there is an upward intergenerational educational and income mobility with respect to both parental and in laws backgrounds.

T. Lakshmanasamy

50

Table-1. Distribution of Educational Status Education Status

Total Sample

Married Sample

Daughter Father Mother Daughter Husband Father Mother

Father in law

Mother in law

Illiterate

-

3.2

27

-

-

3.9

36.5

4.7

38.8

Primary

7.7

27.9

35.2

11.1

2.3

35.7

34.3

36.5

39.7

Secondary

44.6

51.9

31

48.4

35.3

44.2

24.9

50.7

19.1

Diploma/

4.2

4.2

2.7

3.4

3.2

2.1

1.7

1.8

0.7

UG

32.2

10.6

3.3

24

27.7

6.9

1.7

4.7

1.1

PG

10.4

1.6

0.6

12

28.8

1.1

0.9

1.6

0.6

Certificate

Research

0.8

0.6

0.1

1.1

2.7

0.5

-

0.2

-

Total

100

100

100

100

100

100

100

100

100

Table-2. Descriptive Statistics of the Variables Mean

S.D.

Daughter’s education (yrs)

Variable

12.29

3.46

Husband’s education (yrs)

14.42

3.14

Father’s education (yrs)

9.40

3.34

Mother’s education (yrs)

7.99

3.17

Father in law’s education (yrs)

9.11

2.98

Mother in law’s education (yrs)

7.15

2.97

Daughter’s earnings (Rs./day)

229.10

155.97

Husband’s earnings (Rs./day)

341.65

201.08

Father’s earnings (Rs./day)

261.38

195.53

Mother’s earnings (Rs./day)

214.97

136.59

Father in law’s earnings (Rs./day)

246.87

178.25

Mother in law’s earnings (Rs./day)

185.11

143.42

Combined spousal earnings (Rs./day)

557.69

297.30

Daughter’s parental income (Rs./annum) (x100)

502.52

381.82

Husband’s parental income (Rs./annum) (x 00)

588.06

480.42

Table-3 presents the correlation among the educational and income backgrounds of parents and in laws and children. All the background characteristics of parents and in laws are positively correlated with children education and earnings. The influence of in laws characteristics on daughter and son in laws characteristics are also strong and statistically significant just like

Intergenerational Transmission of Economic Advantage under Marital Sorting

51

that of parental characteristics. Similarly, the correlation between spousal earnings and education is significantly and strongly positive. These results again indicate the positive assortative mating and intergenerational correlation of economic status among families that are matched through marriage. Table-3. Correlation between Education and Earnings of Child and Parents and in-Laws Hus. Edu

Father Mother Fa. in Edu. Edu. law Edu.

Dau. Edu.

0.818* 0.546

Hus. Edu.

1.00

Father Edu. Mo. Edu. Father in law Edu. Mo. in law Edu. Dau. Earn. Hus. Earn. Father Earn. Mo. Ear Fa. in law Ear Mo. in law Ear

Mo. in Dau. Earn. Law Edu.

Hus. Earn.

Fa. Mo. Earn. Earn.

Fa. in Law Earn.

Mo. in Combined Earn. Law Earn.

0.224

0.535

0.527

0.633

0.537

0.264 0.644

0.165

0.402

0.597

0.712* 0.410

0.678

0.556

0.543

0.620

0.306 0.797* 0.273

0.419

0.594

0.862* 0.700* 0.784* 0.722* 0.874* 0.576 0.775* 0.637

0.601

0.817*

1.00

0.685

1.00

0.575* 0.804* 0.609* 0.729* 0.715 0.676

0.530

0.574

1.00

0.860* 0.335

0.488

0.242 0.639

0.353

0.700* 0.421

1.00

0.530

0.617

0.437 0.730* 0.398

0.818* 0.587

1.00

0.914* 0.565 0.714* 0.684 1.00

0.476

0.978*

0.488 0.733* 0.805* 0.612

0.979*

1.00

0.538

0.561 1.00

0.105

0.005

0.346

0.480

0.740

1.00

0.618

0.762

1.00

.556

Table-4 presents the regression coefficients of intergenerational educational transmission. All the coefficients are positive and statistically highly significant. The effect of father’s education on children’s education is stronger than that of mother’s education. Similar results hold good for in-laws educational background. The influence of mother-in-law’s education on the daughter-inlaw’s education is slightly lower than the effect of mother’s education on the education of daughters. However, in the case of son-in-law the mother-in-law’s education effect much stronger than the influence of mother’s education on

T. Lakshmanasamy

52

son’s education. In contrast, the effect of father in-law’s education on the education of both daughter-in-law and son-in-law is lower compared to that of father. However, father-in-law’s education has no significant effect on daughterin-law’s education. Further, with respect to spousal education the case for assortative mating has been strongly supported. Overall, the intergenerational transmission of education is stronger for both daughters and sons. The strong influence of in-laws also suggests significant intergenerational transmission of education via marriage market. Table-4. Intergenerational Elasticity of Educational Mobility Dependent Variable: Education of Child Independent Variable

Total Sample

Mother’s Education

Father in Law Education

0.245* (10.27) 0.253* (9.55) -

Mother in Law Education

-

Spousal Education

-

Father’s Education

Married Sample Daughter 0.251* 0.210* (10.64) (4.45) 0.256* 0.085* (9.80) (2.94) 0.017 (0.72) 0.101* (3.78) 0.055* 0.613* (5.37) (17.02)

Husband 0.318* (6.35) 1.064* (28.16) -

0.361* (6.88) 1.048* (27.50) -

-

-

-

0.117* (2.65)

0.418* (8.07) 1.003* (26.66) 0.146* (3.15) 0.168* (3.94) 0.418* (8.07)

*Significant at 1 per cent level.

Table-5 presents the regression coefficients of intergenerational elasticity of income mobility. In all specifications the dependent variable is the logarithm of children’s earnings, with various choice of independent variables. The results clearly demonstrate the existence of intergenerational transmission in income. All the explanatory variables are positive and statistically significant at 1 percent level. As shown in the first column of Table-5, the estimated income elasticity of daughter’s earnings with parental income is 0.39 and that of male children is 0.38. Similarly, in the case of married daughters, the elasticity of daughter’s earnings with respect to in-laws income is much greater (0.53) than the elasticity of son’s earnings (0.39) with respect to in-laws earnings. The income elasticity of married daughter with respect to in-laws family income (0.37) is similar to that of with respect to parental income, where as in the case of son’s income elasticity, it is much higher (0.30) than the own parental income

Intergenerational Transmission of Economic Advantage under Marital Sorting

53

(0.22). These results again reinforce the earlier observation that the family backgrounds are positively matched in the marriage market. Table-5. Intergenerational Elasticity of Income Mobility Dependent Variable: Ln (Earnings of Child) Independent Variable

Total Married Sample Sample Daughter

ln(parental Income)

0.390* 0.501* (13.58) (11.46)

ln(in-laws Income)

-

Ln(Spousal Income)

-

Husband 0.372* (7.60)

0.142* (3.26)

0.383* (10.81)

-

0.525* 0.367* (10.23) (6.70)

0.212* (4.46)

-

-

-

0.784* (11.84)

-

0.224* (9.37)

0.051 (1.57)

0.394* 0.304* (13.58) (9.37)

0.149* (4.52)

-

0.410* (11.84)

-

One of the main objectives of this paper is to explore the role of assortative mating in the intergenerational mobility of married children. Towards this end, we reestimated the model with married daughters only. As shown in the second column of Table-5, the intergenerational elasticity in family income for married daughters increases to 0.50. The intergenerational elasticity with spousal income has been 0.78 for daughters and 0.41 for male children. Thus, for both men and women nearly 40-80 percent of the covariance between the spousal earnings can be attributable to sorting in the marriage market, a result similar to that of Ermisch et al. (2006) for Germany. With this almost 50 percent of one’s social position attributable to the process of who marries whom, assortative mating is a major factor in the intergenerational transmission of inequality. Further, this also indicates that there are larger income gains from assortative marriage. 71.2 All these results suggest a considerable degree of intergenerational transmission, especially among the daughters and daughters-in-law. Further, assortative mating appears to play a crucial role. For married couples, a major factor in the intergenerational transmission of income status is that the elasticity of spousal income with respect to the in-laws income is more than the elasticity of their own parental income. 5. Conclusion This paper has estimated the intergenerational transmission of education and income, not only among daughters and sons, but also among daughters-in law and sons-in-law. The estimated intergenerational transmission is quite substantial, more pronounced in the case of daughters. The elasticity of

54

T. Lakshmanasamy

children’s income is also strongly positive with respect to in-laws income and education. The empirical results on spousal backgrounds also reveal strong positive assortative mating pattern. Among married children, spousal earnings appear just as elastic as the children’s own earnings with respect to the parental income and education. This effect is driven by the strong spousal correlation of human capital. The intergenerational transmission of socioeconomic advantage appears to be much stronger for daughters compared to that for sons. The combined strength of socioeconomic advantage and assortative matching thus fosters inequality and produces a radically unequal chance for intergenerational mobility. The empirical findings indicate intergenerational transmission increases with parental investments, especially on human capital and marriage. Policies that reduce the importance of family background for individual’s failure or success are appropriate; two such policies might be public investments in human capital and bequest References 1. Altonji, J.A. and T. Dunn (1992) “Relationships among the Family Incomes and Labor Market Outcomes of Relatives”, Research in Labor Economics, 12, 269–310. 2. Becker, G.S. (1991) A Treatise on the Family, Cambridge: Harvard University Press. 3. Becker, G.S. and N. Tomes (1979) “An Equilibrium Theory of the Distribution of Income and Intergenerational Mobility”, Journal of Political Economy, 87, 6, 1153–1189. 4. Becker, Gary S. and Nigel Tomes (1986) “Human Capital and the Rise and Fall of Families”, Journal of Labor Economics, 4, 3, Part 2, S1–S39. 5. Behrman, J.R. and M.R..Rosenzweig (2002) “Does Increasing Women’s Schooling Raise the Schooling of the Next Generation?”, American Economic Review, 91, 1, 323–334. 6. Black, S.E., P.J. Devereux and K.G. Salvanes (2003) “Why the Apple Doesn’t Fall Far: Understanding the Intergenerational Transmission of Human Capital”, NBER Working Paper No. 10066. 7. Bjorkland, A. and M. Jantti (1997) “Intergenerational Income Mobility in Sweden Compared to the United States”, American Economic Review, 87, 5, 1009–1018. 8. Bjorkland, A., T.Eriksson, M. Jantti, E. Osterbacka and O.Raaum (2002) “Brother Correlations in Earnings in Denmark, Finland, Norway and Sweden Compared to the United States”, Journal of Population Economics, 4, 757–772.

Intergenerational Transmission of Economic Advantage under Marital Sorting

55

9. Blanden, J. (2005) “Essays on Intergenerational Mobility and Its Variation over Time, Place and Family Structure”, Unpublished Ph.D. Thesis, University college, London. 10. Blanden, J., P. Gregg and L. Macmillan (2007) “Accounting for Intergenerational Income Persistence: Noncognitive Skills, Ability and Education”, Economic Journal, 117, C43–C60. 11. Bratberg, E., O. Nilsen and K.G. Vaage (2002) “Assessing Changes in Intergenerational Earnings Mobility”, University of Bergen, mimeo. 12. Bratberg, E. O.A. Nilsen and K.G. Vaage (2005) “Intergenerational Earnings Mobility in Norway: Levels and Trends”, Scandinavian Journal of Economics, 107, 3, 419–435. 13. Bratberg, B., K. Roed, O. Raaum, R. Naylor, M. Jantti, T. Eriksson and E. Osterbacka (2007) “Nonlinearities in Intergenerational Earnings Mobility: Consequences for Cross-Country Comparisons”, Economic Journal, 117, C72–C92. 14. Chadwick, L. and G. Solon (2002) “Intergenerational Income Mobility among Daughters”, American Economic Review, 92, 1, 335–344. 15. Chevalier, A. (2003) “Parental Education and Child’s Education: A Natural Experiment”, University College Dublin, mimeo. 16. Corak, M. (2004) “Do Poor Children Become Poor Adults? Lessons for Public Policy from a Cross Country Comparison of Generational Earnings Mobility”, in M.Corak (ed.): Generational Income Mobility in North America and Europe, Cambridge, Mass.: Cambridge University Press. 17. Couch, K.A. and T.A. Dunn (1997) “Intergenerational Correlations in Labor Market Status: A Comparison of the United States and Germany”, Journal of Human Resources, 32, 1, 210–232. 18. Deardon, L., S.Machin and H.Reed (1997) “Intergenerational Mobility in Britain”, Economic Journal, 107, 47–66. 19. Eide, E.R. and M.H. Showalter (1999) “Factors Affecting the Transmission of Income Across Generations: A Quantile Regression Approach”, Journal of Human Resources, 34, 2, 253–276. 20. Ermisch, J., M. Francesconi and T. Siedler (2006) “Intergenerational Mobility and Marital Sorting”, Economic Journal, 116, 659–679. 21. Fernandez, R., A. Fogli and C. Olivetti (2004) “Mothers and Sons: Preference Formation and Female Labor Force Dynamics”, Quarterly journal of Economics, 116, 1249–1299. 22. Grawe, N. and C.B. Mulligan (2002) “Economic Interpretations of Intergenerational Correlations”, Journal of Economic Perspectives, 16, 3, 45–58. 23. Goldberger, A.S. (1989) “Economic and Mechanical Models of Intergenerational Transmission”, American Economic Review, 79, 3, 504513.

56

T. Lakshmanasamy

24. Holmlund, H. (2007) “Intergenerational Mobility and Assortative Mating: Effects of an Educational Reform”, London School of Economics, mimeo. 25. Horvonen, L. (2007) “Intergenerational Income Mobility: The Role of Education as an Income Transmission Mechanism”, Swedish Institute for Social Research, Stockholm University, mimeo. 26. Lam, D. and R.F. Schoeni (1993) “Effects of Family Background on Earnings and Returns to Schooling: Evidence from Brazil”, Journal of Political Economy, 101, 4, 710–740. 27. Lam, D. and R.F. Schoeni (1994) “Family Ties and Labor Markets in the United States and Brazil”, Journal of Human Resources, 29, 4, 1235–1258. 28. Mulligan, C.B. (1999) “Galton vs The Human Capital Approach to Inheritance”, Journal of Political Economy, 107, 184–224. 29. Plug, E. (2002) “How Do Parents Raise the Educational Attainment of Future Generations?”, IZA Discussion Paper No. 52. 30. Raaum, O., K.G. Salvanes and E.O. Sorensen (2006) “The Neighborhood is Not What It Used to Be”, Economic Journal, 116, 278–300. 31. Raaum, O., B. Bratberg, K. Roed, E. Osterbacka, T. Eriksson, M. Jantti and R. Naylor (2007) “Marital Sorting, Household Labor Supply, and Intergenerational Earnings Mobility across Countries”, IZA Discussion Paper No.3037, Bonn. 32. Sacerdote, B. (2002) “The Nature and Nurture of Economic Outcomes”, American Economic Review, 92, 2, 344–348. 33. Solon, G. (1999) “Intergenerational Mobility in the Labor Market”, in O. Ashenfelter and D. Card (eds.): Handbook of Labor Economics, Volume 3A, Amsterdam: North-Holland, 1761–800.

ADDRESSING EQUITY ISSUES IN WATERSHED DEVELOPMENT PROJECTS IN BHIL ADIVASI AREAS OF WESTERN MADHYA PRADESH RAHUL BANERJEE 74, Krishnodayanagar Khandwa Naka, Indore Madhya Pradesh - 452001 Email: [email protected] The western Madhya Pradesh region of India, which is largely populated by Bhil adivasi (indigenous people) peasants, is typical of other such adivasi regions of the country in that fragmentation of landholdings coupled with the neglect of dryland agriculture has severely jeopardised the livelihoods of the people and forced them to further mine their immediate environment for subsistence needs. In such a scenario systematic work to bring about equitable and sustainable development is hindered by the fact that common property resources are most often privatised and people who are in control do not want to let go of them. This paper details how two NGOs, SAMPARK and SAMAJ PRAGATI SAHAYOG have creatively overcome this through inspiring communitarian problem solving and the building up of social capital. However the paper concludes by underlining the fact that widespread equitable and sustainable development through the replication of such one off successes is prevented due to the lack of political capital needed on a larger scale to challenge the dominant paradigms of governance and development.

1. Introduction The western Madhya Pradesh region in which the Bhil adivasis reside stretches across the seven districts of Jhabua, Dhar, Barwani, West Nimar, East Nimar, Dewas, Indore and Ratlam. Geographically this area begins roughly from the Malwa Plateau in the north descending the escarpment of the Vindhya hills into the Nimar plains flanking the lower Narmada river valley and ends with the Satpura ranges to the south. While Jhabua, West Nimar, Dhar and Barwani have overwhelming majorities of Bhils the other districts have lesser concentrations with Indore having the least. The average percentage of scheduled tribe population is about 43% of the total population. This region has three distinct agro-ecological zones - Malwa Plateau, Nimar Plains and Jhabua Hills. The first two have considerable areas of deep clayey soils which are fertile while the last has mostly thin unfertile lateritic soils and the terrain is hilly. Unfortunately the demographically numerous Bhils mostly hold title to marginal plots of unfertile 57

58

Rahul Banerjee

and unirrigated land situated in the upper watersheds of the first two regions and the hilly terrain of the last region. Following on the national policy in this regard the concentration of government finances after independence on the promotion of green revolution agriculture on the more fertile lands belonging to non-adivasi farmers in the first two agro-ecological zones to the neglect of the much wider dryland areas of the Bhils in the upper watersheds has skewed the resource access pattern of the region against them. The benefits of the green revolution were cornered by the traders, who traded in the inputs and the increased output. The large farmers too benefited immensely by earning huge surpluses from low production costs due to state subsidised supply of inputs and the use of their extra-economic powers over the adivasis to keep wage levels depressed (Banerjee, 2003). This weakening of the primary agricultural base of the Bhils combined with ill conceived and even more badly implemented poverty alleviation schemes of the government to provide supplementary incomes which have invariably failed has meant that the adivasis have remained in the clutches of sahukars who dominate the rural markets of the region exploiting the former through unremunerative prices for their produce, exorbitant prices for the agricultural inputs and usurious interest rates on loans advanced to them (Aurora, 1972). Consequently most of the Bhil adivasi peasants have to rely on migration either permanent or seasonal to make ends meet (Mosse et al, 2002). It has now become fairly well established from qualitative analysis of tribal development policies within the larger area of scholarship of the adivasi predicament in India that the institutions set up under the provisions of the Constitution of India and the various laws enacted from time to time for the protection of the adivasis have not functioned properly primarily due to the wrong development policies adopted by the state which have tended to strengthen rather than weaken the inequality in political and economic power in favour of the non-adivasis vis-a-vis the adivasis (Sharma, 2001). The state has also failed to provide good and adequate education services which has resulted in the adivasis remaining unequipped to negotiate the complexities of the modern centralised system of governance into which they have been forcefully integrated (Rahul & Subhadra, 2001). The poverty induced by these development policies has adversely affected the nutritional levels of the food intake of the adivasis and combined with the lack of good and cheap health services has led to a decline in their general health. Moreover, the even greater lack of both education and health services for the women due to the inherent patriarchy in Bhil society has meant that they have not been able to smash age-

Addressing Equity Issues in Watershed Development Projects in Bhil Adivasi Areas

59

old patriarchal structures and their consequent lack of reproductive rights has led to a population explosion putting further pressure on already scarce resources (Rahul, 1999). Combined with the agricultural development policies described earlier these have produced a scenario wherein adivasis are continually suffering from the imperfections of the modern market system, which has increasingly penetrated into their subsistence lifestyles forcing them to live on the edge and mine rather than conserve environmental resources vital to their survival. The common property resources (CPRs) have become so scarce that there is tremendous competition to privatise and denude them. Thus in the western Madhya Pradesh region some of the land under the forest department and most of the other cultivable common lands, have been encroached on by adivasi peasants for cultivation. These systemic problems have led to the following obstacles to sustainable development being sought to be implemented through watershed programmes conducted by NGOs in the region – 1. Fragmentation of land holdings has resulted in most adivasis being left with unviable holdings of 1 ha or less in area. 2. Village common lands have been progressively encroached upon for agriculture and so are not easily available for restorative treatment. 3. The stranglehold of the sahukars over the lives of the adivasis has resulted in the latter being unable to make any savings and investments. 4. The decay of the traditional community gram sabha leading to an escalation in disputes and a degeneration of customs of communitarian labour and sharing of costs of social functions. 5. Lack of adequate education and health services which adversely affected the capacity of the people to work and participate meaningfully in development and governance. 6. Women have been marginalised from development programmes and have instead had to bear disproportionately the costs of maldevelopment. There was a rethinking in the beginning of the decade of the 1990s all over the country with regard to the implementation of watershed development leading to the popularisation of the “ridge to valley” approach as opposed to the treatment of isolated areas and the active involvement of the beneficiaries in planning, implementation and post project maintenance of the created structures (Shah, 1993, GOI, 1994). The Government of Madhya Pradesh initiated the

60

Rahul Banerjee

ambitious Rajiv Gandhi Watershed Development Mission (RGWM) in 1994 incorporating these new ideas by pooling all the funds being made available to it by the Government of India for poverty alleviation and treatment of drought prone areas under various schemes. This increased stress on watershed development on the part of the Government of Madhya Pradesh came from the realisation that since the state is situated across a drainage divide involving as many as six river basins, the terrain is undulating and water storage in the natural system is low. Moreover the state has only a limited share in the river waters since the state lies on the upper catchment. Thus conventional dam centred water resources development adopted thus far had proved costly and inefficient (RGWM/TARU 2001). Apart from the government many NGOs too began to implement watershed development programmes along these lines. The efforts made by two such NGOs, one SAMPARK situated on the intersection of the Malwa Plateau and Jhabua Hills agro-ecological region and the other SAMAJ PRAGATI SAHAYOG situated in the Nimar Plains to try and address the problems noted above and overcome the inequalities being faced by the adivasis and especially women are described and analysed here. 2. Work of SAMPARK SAMPARK (SAMPARK, 1995) has laid stress on first resolving socioeconomic problems before embarking in a big way on implementing physical watershed development work. The important thing was to mobilise the people to battle the control of the sahukars and then use this unity to overcome internecine squabbles among the adivasis over the control of CPRs, which was preventing the implementation of sustainable watershed development. Thus the traditional gram sabhas were activated once again and initiatives were launched in the fields of micro-finance, cheap resolution of disputes, revival of traditional labour pooling customs, reduction of the cost of social functions such as weddings and rites of passage, primary education and primary health and physical watershed development works were continued only in a low key manner. It has been established through practice that successful community based micro-finance is one of the more important desiderata for broadbased rural development (NABARD, 1999). Only after these efforts began to bear fruit in the form of greater social cohesion, a gradual liberation from the clutches of the sahukars and a better social status for the women was largescale watershed development launched again from 1998 onwards. Work has been done in five different watersheds inhabited by about 8000 Bhil and Patelia adivasis in sixteen villages spread over

Addressing Equity Issues in Watershed Development Projects in Bhil Adivasi Areas

61

an area of 7580 hectares in Petlawad tehsil of Jhabua district in Madhya Pradesh with funds provided by CASA and Action Aid funding agencies. In addition to this SAMPARK is also responsible for the social mobilisation of the beneficiaries in the DANIDA funded Comprehensive Watershed Development Project (CWDP) in the tehsil in which the physical works are carried out by agencies of the GOMP. We shall first give details of the district and tehsil characteristics and then describe how the organisation has dealt with the problem of inequalities in watershed development with the example of one watershed. 2.1. District and Tehsil Characteristics The Roopapada village in the Kalighati Panchayat is situated in Petlawad tehsil of Jhabua district. The district lies in the Vindhya ranges at the edge of the Malwa plateau and the land is hilly without much tree cover and prone to heavy erosion. Petlawad tehsil is drained by the Mahi river which forms the northern boundary of the district. Geologically five rock formations are found in the district. These are Deccan Trap, Alluvium, Cretaceous-Lameta, Aravalis and Banded Gneissic Complex. The whole of Petlawad tehsil has the Deccan Trap formation which is also known as the Malwa Trap. It falls in the Malwa plateau agroclimatic zone with medium to black medium soils of medium levels of the three main nutrients of Nitrogen, Phosphorus and Potassium. The better quality lands in the tehsil are held mostly by the non-adivasis while the 75 % majority adivasis have the lower quality lands which are mostly unirrigated and lie in the upper watershed regions. Thus the break up of crop production for the whole petlawad tehsil shown in Fig. 1 does not adequately reflect the crop mix of the adivasis which has a higher proportion of cereals and pulses and less of cotton, oilseeds and fodder. Unfortunately disaggregated data showing this difference is not available in collated form with the government. The yields of crops in the tehsil are shown in Fig. 2 and as is evident these are below the national average. Some Human Development Indicators for Jhabua district along with its rank among the forty five districts of Madhya Pradesh are given below in Table 1. Clearly the district is very backward. The poverty ranking is not that bad because the people migrate to nearby developed areas and earn supplementary incomes. The per capita food production ranking too is high because there are no big towns and cities in the district, which can push up the percentage of the non-food producing population.

62

Rahul Banerjee

Figure 1: Cropping Pattern of Petlawad Tehsil in Percentage (Source: District Statistical Handbook 2000, Department of Economics and Statistics, GOMP)

Figure 2: Crop Yields in Petlawad Tehsil in Qu./Ha Source: District Statistical Handbook 2000, Department of Economics and Statistics, GOMP

Addressing Equity Issues in Watershed Development Projects in Bhil Adivasi Areas

63

Table 1: Selected Human Development Indicators for Jhabua District Source: Third Human Development Report Madhya Pradesh 2002, GOMP No. 1. 2. 3. 4. 5. 6. 7. 8.

Indicator Human Development Index Gender Development Index Population Dependent on Agriculture (%) Infant Mortality Rate Life Expectancy (2001) Total Fertility Rate Below Poverty Line (%) Annual Per Capita Food Production (kgs)

Value 0.372 0.450 90.6 130 55.8 7.0 31.2 268.22

Rank Among 45 Districts 45 43 2 42 30 45 20 21

2.2. Roopapada Watershed The land use pattern of the watershed is as shown in Table 2 below. In reality after accounting for the encroachments for habitation and cultivation there are only about 45 ha of village common land available for protection and development work. There are two tanks in the village built by the government, which are used mainly for recharging purposes and sometimes for occasional protective irrigation for a cotton crop. Table 2: Landuse Pattern in Roopapada Watershed(ha) Source: Sampark Records Total Area

Forest Area

239.08

19.10

Irrigated Area 20.16

Unirrigated Area 129.23

Uncultivable Wasteland 26.05

Wasteland 44.54

There are 67 families with a population of 402 of whom 207 are males and 195 females all engaged in agriculture. Fifteen of these belong to the Patelia tribe and the rest to the Bhil tribe. The Patelias consider themselves to be socially superior to the Bhils and practise untouchability with them. The Patelias are slightly better off economically than the Bhils but the biggest landholder who is also the Patel or headman of the village has only about 3 ha of land and is a Bhil. Thus they are more or less all below the poverty line. The patel has been successful in acquiring a tractor through a loan given by the government. The land near the ridgeline belongs to the forest department. There is one patch of common wasteland just below the ridgeline of about 19 ha and two patches of 15 and 10 ha lower down in the watershed. There is a small hamlet of seven Bhil families on the ridgeline and these were in control of the uppermost wasteland. Another hamlet of 8 Bhil families were in control of the lowermost

64

Rahul Banerjee

patch of 10 ha. These 15 families all owed allegiance to the Patel Rama. The 15 ha patch in between was free for all. Given the severe shortage of agricultural land there is a tendency throughout western Madhya Pradesh of bringing common wastelands under the plough. Thus some of the land in the uppermost patch was being cultivated by some of the families in the nearby hamlet and it was being used for cattle grazing by some thirty families. Grazing was being done on the forestland nearby also. Watershed work could not be started as long as the ridgeline CPR remained under private control. The problem thus was that due to the heavy deprivation being suffered by all of them some of the people in the village who were slightly more powerful in social and economic terms were appropriating the use of the CPRs to the further detriment of the ecology and economic viability of the watershed as a whole. This is a typical situation of conflict between gainers and losers in watersheds which vitiates proper watershed development that has been noted in many places (Kerr, 2002). The strategy adopted by SAMPARK when it first entered the village in 1996 was to form a self-help group with twenty Bhil families. The successful running of this SHG resulted in 29 more Bhil families deciding to form another SHG in 1997. All these families then began reviving their traditional customs of labour pooling and community dispute resolution. There are at present two women’s and one men’s SHGs with a combined membership of 72 and savings of Rs 1,74,783 and freedom from the debt of sahukars. A group of people from these SHGs were then taken on an exposure tour to see the work of Tarun Bharat Sangh in Alwar in Rajasthan and that of Anna Hazare in Ralegan Shiddhi in Maharashtra. These people then came back and related to the whole group their experiences and they then began to together exert pressure on the Patel and his kinsmen to stop grazing the ridgeline patch so that it could be treated and planted. Seeing that the Patel was not amenable to the social fencing and protection of the common lands the rest of the villagers decided to go ahead without his consent and protect the land from free grazing. The Patel refused to agree with this and sent his daughter to graze his animals in the protected area. When she was stopped by the chowkidar Babu then the Patel’s son went and beat him up. The villagers responded by calling a traditional panchayat meeting in which the Patel also came. A fine of Rs 500 was imposed on the Patel’s son for his misdemeanour but the Patel refused to pay up. Then the villagers decided to boycott the Patel’s family as is the traditional custom on non-compliance with panchayat rulings. Initially the Patel refused to be cowed down but eventually he found it extremely difficult as no one would come to his house or to work on his field. Then he called a meeting

Addressing Equity Issues in Watershed Development Projects in Bhil Adivasi Areas

65

of the panchayat and paid up the fine. Finally good sense prevailed and a compromise was reached that the seven families on the ridgeline would stop free grazing on the wasteland but they would retain their small plots of agricultural land within it. After this in 1998 contour trenching, gully plugging and pasture development work was started on public land as also field bunding work on some of the private agricultural land. Plantation work was done through community contribution of labour and social fencing was employed for protection. The fodder was cut by the whole village and distributed equally among all villagers irrespective of whether they were members of the SHGs or not at a nominal price of Rs 2 per bundle of grass. As the work progressed and the benefits of the treatment works became manifest it became clear that these accrued differentially to the upper and lower watershed inhabitants in terms of water availability. While the wells in the lower watershed and the two tanks there began to have more water those in the upper watershed did not show similar recharge.

Figure 3. Difference between Heading and Sunken Water Retaining Structures Table 3: Physical Work done in Roopapada Watershed (Sampark Records) Year 1998 1999 2000 2001 2002 2003

Work Done Field bunding on private farm land Contour trenching and gully plugging on public land Gabian Structures on nullahs Gully plugs and dugouts on public land Gully plugs and dugouts on public land Gully plugs and dugouts on public land

Area (ha) 9 20

Beneficiary Households 12 38

3 13 34 3

25 29 40 19

66

Rahul Banerjee

This problem was solved by borrowing from the expertise of the DANIDA CWDP, which, as mentioned earlier, was being implemented in another part of the tehsil. The DANIDA project had considered this problem at the design stage itself and come up with a novel solution for it. Most watershed structures are “heading” ones with larger surface water area in which the evaporation loss is high and the percolation low. To reverse this the DANIDA project had designed a new “sunken” structure called the dugout in which the soil and rock are dugout to a depth and so the surface and head available for percolation is much more than that for evaporation (Fig. 3). Thus good technology too can help when the basic social mobilisation is good. The added advantage is that the percolating water immediately recharges the nearby wells (GOMP, 1997). The details of the physical works done are given in Table 3 above. While the work done upto 2001 was funded by CASA the work done in 2002 and 2003 was done totally through shramdan or community labour. In 2002, 399 person days of labour were contributed while in 2003, 180 person days of labour were contributed. This is what is crucial as for sunken structures it is absolutely a must that they be cleaned each year if they are to retain their efficacy as water rechargers and this is possible only if the user group is well organised and conscious about its responsibilities of maintenance. In addition to this as many as 30,000 saplings have been planted over the years with a survival rate upwards of 90%. The villagers have refurbished their traditional simple stone monument to their village God on the top of the ridge in the regenerated forest and they celebrate their monsoon Diwasa festival, which is a thanksgiving to God for having provided them with the means of livelihood, with fanfare there. Finally the villagers have also revived their traditional custom of “adjipadji” under which they pool their labour together for the labour intensive agricultural operation of weeding. 600 person days of community labour were thus generated in 2003. Thus a conflict situation in which the more powerful group of adivasis were preventing the sustainable use of the CPRs by privatising them was amicably resolved basically through the revival of the traditional gram sabha which in turn gained its strength from the successful battle against the control of the sahukars through the means of the SHGs. The visible benefits in terms of fodder availability and groundwater recharge then enthused the other hamlet of adivasis lower down in the watershed to agree to give up control of the patch of 10 ha that they had with them and here too watershed development and plantation work has been carried out. The 15 ha patch in the middle has also been treated but plantation work has not been carried out in it. A situation in which conflicts

Addressing Equity Issues in Watershed Development Projects in Bhil Adivasi Areas

67

had arisen between the poor hailing from the same socio-economic background but residing in the upper and lower reaches of the watershed because of deprivation caused by negative external forces had been resolved by reducing this deprivation, primarily through the creation of what has come to be called “social capital” (D’Silva & Pai, 2003). Moreover since stress was laid from the beginning on the mobilisation of women through SHGs they have played a major role in all the activities of the community apart from addressing their own issues especially in the field of reproductive health and rights which is normally a neglected sphere of social work. In collaboration with the NGO, CEHAT, SAMPARK has initiated a women's community health programme anchored by a trained woman from the village itself to take care of general health as well as reproductive health issues in the village itself. Finally in rural areas in India considerable antagonism exists between the “local state” of petty government officials and the common people who have to approach them as supplicants for services which are their legitimate due (Corbridge et al, 2003). The history of Bhil adivasi deprivation is laced with many tales of woe generated by the inhospitable behaviour of the local state officials within the larger narrative of the unjust policies of the colonial, national and global states. This contradiction has given rise to a different kind of conflict situation in Roopapada of late. The highest reaches of the watershed are controlled by the forest department as mentioned earlier. The people have begun to dig dugouts in this area through shramdan seeing the tremendous importance in terms of immediate recharge benefits. This has led to objections being raised by the local forest guard. At present the dispute is of a muted nature but if the villagers decide to go ahead with more dugouts in future then the conflict might escalate. For the time being the villagers have told the forest guard that he should go and tell his superiors that they should arrange for funds for the treatment of this land as otherwise they would do this work through shramdan. 2.3.

Scaling up

Experience has shown that mobilising people in a single watershed alone is not enough to ensure sustainability of the gains from watershed development without building up wider networks and institutions that can create a positive counter-culture of change that can challenge the negative attitudes of those ranged against the emancipation of the poor at various levels (Yugandhar, 1999). The Gram Sabha in Roopapada also has health and education committees that take care of primary health and night school education with the help of

68

Rahul Banerjee

SAMPARK and also try and see that government services in these spheres improve. There are many other such all purpose gram sabhas and these now total ninety in number. These Gram Sabhas are organised in clusters of five each and then federated together in the form of a mass organisation – Lok Jagriti Manch. This federation has covered villages that have not had watershed work done in them but have SHGs and other communitarian activities. Thus the ground has been prepared for the implementation of watershed programmes on a much larger scale than is being done at present by solving the social problems that normally prevent the equal and positive impact of such programmes. This federation has tackled common issues of importance as follows – 1. The centralised wholesale sourcing of seeds, fertilisers and pesticides and their subsequent distribution all through voluntary work leading to massive savings on the retail prices of these inputs. In 2001- 2002 the federation bought agricultural inputs worth Rs 7,95,813.00 effecting a saving of Rs 1,46,811.00 on the retail prices of these inputs. Apart from this the saving on the interest that they would have had to pay if they had loaned the inputs from the sahukars is Rs 2,00,000.00. 2. The fixing of rates for bride price and other donations that the bridegroom has to give to the bride’s family at the time of marriage. The expenses had reached over Rs fifty thousand and were a major caused of indebtedness. They have now been brought down to around Rs fifteen thousand. Similarly the expenses for the Rakhi festival and rites of passage too have been reduced considerably. 3. Advocacy both at the mass level and at the policy level for better utilisation of natural resources and also for the provision of greater food and employment security. The LJM is an active participant in the national campaigns for food and employment security as well as the campaign for sustainable use of water resources. Given the fact that there is tremendous pressure on land in western Madhya Pradesh and especially among the adivasis in Petlawad, work done on revenue land alone cannot ensure sustainable livelihoods to people. Thus it is most essential that the management of forestland too should be handed over to the people. There have been only desultory efforts in this direction. Unless more positive steps are taken a conflict situation is bound to arise and so the LJM has begun advocacy on this issue also.

Addressing Equity Issues in Watershed Development Projects in Bhil Adivasi Areas

4.

69

The conservation and promotion of traditional agricultural practices and seeds that are more suited to dry land areas. This is in line with the findings of state of the art agricultural research, which has established that investment on research, extension and inputs for the strengthening of dryland agriculture are now higher than that on green revolution agriculture (Fan & Hazell, 1997).

SAMPARK has thus graduated from a narrow technical approach to natural resource management to a more holistic socio-political approach. It has facilitated the setting up of civil society institutions that can not only amicably resolve disputes over natural resource management by ensuring a better distribution of costs and benefits but which can gradually aspire to challenge the hegemony of traders in the market, government officials in the bureaucracy, the politicians in the higher level democratic institutions of the state and the patriarchal social structure of traditional Bhil society. Most importantly it has realised the importance of positive participation in the market not only as a sine qua non for the long-term sustenance of successfully implemented watershed development projects but for the overall future development of the adivasis. Given the fact that today markets have become truly universal from the village to the global level and have become more influential than the state it is imperative that adivasis learn to take advantage of the market in what is probably the biggest civilisational change ever faced by any community (Nathan & Kelkar, 2003). 3. Work of SAMAJ PRAGATI SAHAYOG Often even after good implementation of watershed development programmes at the micro level the negative socio-economic and institutional inequalities resulting from macro level policies which have been enumerated in the introduction have meant that by and large sustainability of agriculture and biomass production and security of livelihoods has not been achieved in the Western Madhya Pradesh region. This is illustrated here with the example of a watershed in the Bagli tehsil of Dewas district that has been completely treated by the NGO – SAMAJ PRAGATI SAHAYOG, first in an adhoc manner from 1993 onwards and later more systematically through comprehensive watershed development under the state government's Rajeev Gandhi Watershed Development Mission (RGWM) from 1998 onwards. Data regarding the socioeconomic and natural resource profiles are first provided followed by the description of the inequalities and their partial resolution through the actions of the NGO and an analysis of the reasons for this failure.

Rahul Banerjee

70

3.1. Socio-economic Profile 3.1.1. Identification Tables 4 and 5 below give details on the location and accessibility of the study area. Table 4 1.1

District : Dewas

1.2

Tehsil : Bagli

1.3

Janpad : Bagli

1.4

Gram Panchayat : Bhikupura

1.5

Village: Neemkhera

1.5.1

Hamlets : Neemkhera, Neemkheri, Padhav

1.6

Headquarters of Grampanchayat : Bhikupura

3.1.2. Accessibility Table 5 3.1

Distance from Tehsil Headquarters (Kms)

20

3.2

Distance to Nearest Town (Kms)

20

3.3

Type of approach Road

Macadamised but in bad condition

3.4

Nearest Railway Station

80

3.1.3. Agro-Ecological Zone The watershed is situated in the Nimar Plains agro-ecological zone and drains into the Narmada river eventually. The land is a little more undulating and the soils of poorer quality than the flatter black soil lands prevailing lower down the Narmada valley where the Nimar plains broaden out to sixty kms on each side of the river. The normal average annual rainfall is about 850 mm. 3.1.4. Demography Tables 6 and 7 below provide details on demographics and land-ownership in the study area. Table 6 No. of Households Population

*

177 Total

Male

Female

ST

SC

1045 (100) 532 (50.9) 513 (49.1) 923 (88.3) 14 (1.3) BPL Households *

Others 108 (10.3)

105

The no. of households include those from the nearby villages of Ratatalai and Bhikupura who own agricultural land within the watershed and village boundary but who reside in their respective villages.

Addressing Equity Issues in Watershed Development Projects in Bhil Adivasi Areas

71

3.1.5. Households and Occupation Table 7 Total HHs (Nos)

Gross Sown Area (ha)

No. of Households Landless

Marginal (< 1 ha.)

Small (1–2ha.)

Semimedium

Medium

Large

(5-10 ha.)

(> 10 ha)

28

3

(2–5ha.) 177

540

10

45

50

41

Six people are engaged as teachers with the Government, two as policemen and one as a clerk in the Public Works Department. Three people earn supplementary incomes as carpenters. One person runs a grocery shop and five people work for the NGO. So given the extreme fragmentation of landholdings and the low productivity of agriculture a majority of the people have to depend on supplementary labour either inside or outside the watershed. Thus there is a large surplus of unutilised labour as shown in Table 10 below. This excess labour availability is a feature of the whole region and it naturally depresses the wages that are paid for agricultural labour in and around the watershed. There are three non-adivasi families who are the large landholders and along with other non-adivasi landholders in the nearby village of Bhikupura they exerted considerable political influence over the adivasis. 3.1.6. Estimate of Surplus Labour in the Watershed (Persondays/year) Table 8 1.

Total Working Population

511

2.

Available No. of Working days per worker per year

280

3.

Total required persondays of employment per year (1*2)

143080

4.

Average employment in agriculture persondays per ha. per year

100

5.

Gross Sown Area

540

6.

Total employment generated in agriculture persondays per year (4*5)

54000

7.

Employment generated NTFP collection persondays per year

10000

8.

Other Employement like grazing, fuelwood collection etc

15000

9.

Total Employment persondays per year (6+7+8)

79000

10.

Surplus labour in the Watershed persondays per year (3-9)

64080*

* This is only an indicator of the level of surplus labour that is available and not its absolute value.

72

Rahul Banerjee

3.1.7. Migration Members of about forty households regularly migrate during the harvesting of soyabean and wheat to the Malwa region for a period of fifteen to thirty days for employment. Apart from this youth of about twentyfive families have migrated permanently to the city of Indore to work in various factories. There has not been any reduction in the migration levels on a sustained basis. There was some impact of the watershed works but once these were completed the people had to begin migrating once again. 3.1.8. Infrastructure Facilities Information on infrastructure is given in Table 9 below. Table 9 Within

Outside - distance

Anganwadi Centre

1

Primary School (I to V)

3

Middle School (VI to VIII)

-

Poojapura 5kms

Secondary School (IX to X)

-

-do-

Higher Secondary School (XI to XII)

-

-do-

Primary Health Centre

-

Bagli 20kms

Telephone Exchange ( CDMA)

-

-do-

Electric Grid Station

-

Udainagar 15kms

Bank

-

Poojapura 5 kms

Agricultural Coop. Society

-

-do-

Fair Price Shop

-

-do-

3.1.9. Local Organisation The NGO has been instrumental in setting up a water-users group, a forest-users group, an appropriate agricultural technology group, a women's group and four self-help microfinance groups which are all operating at various levels of efficiency. The women's groups and SHGs are the most active and have been instrumental in mobilising the women to take active part in the affairs of the community and also challenge the patriarchal structure of the traditional Bhil society. Apart from this there is a village watershed development committee that supervised the implementation of the watershed works and looks after the

Addressing Equity Issues in Watershed Development Projects in Bhil Adivasi Areas

73

maintenance of the structures created. The panchayat involves itself only in building of roads and community buildings and does not take any interest whatsoever in natural resource management mainly because it is controlled by the non-adivasis of Bhikhupura who are inimical to the development and progress of the adivasis as will become clear by and by. 3.1.10. Village Industries The NGO conducts training of women in tailoring and garment making and the products are marketed by it. The NGO also arranges for the collection of Neem seeds from all over the tehsil which are then pressed in an oil mill set up in the village. The oil and cake are sold for use as pesticides and fertilisers. 3.1.11. Market Access The nearest weekly market is at Poojapura five kilometers away. The Mandi set up by the government Agricultural Marketing Board in Bagli is non-functional because the traders do not want it to operate. Over the past two years the Consumer Foods division of the Indian Tobacco Company has begun sourcing the soyabean produce directly from farmers for its edible oil plants and it has set up a buying office in Dakachya village about seventy kms from the village. Small ruminants are marketed at Poojapura but the nearest market for large ruminants is at Haat Pipliya about 40 kms away. 3.2. Natural Resource Profile The natural resource profile has been segmented into land, agricuture, water, livestock and energy and described separately below. 3.2.1.

Land Area, Value and Type

Fertile black soil flat land sells at rupees eighty thousand to one lakh per acre while poorer quality land can sell at values as low as rupees ten thousand per acre. Oustees from the Indira Sagar dam being built on the river Narmada have been seeking land to resettle on and have pushed up prices somewhat in the region. There has not been any significant change in ownership of land in the watershed after treatment but the three large landholder families in the village after having lost their erstwhile power are keen to sell and resettle somewhere else where they are not in a minority. But since they want to sell enblock they have not yet got a buyer who can pay the rupees fifty lakhs or so that they are demanding.

Rahul Banerjee

74

The soil is quite deep along the nallah banks and valleys where the land is relatively flat, it is also quite deep and well drained along the gentle slopes but shallow on higher slopes and absent on steeper slopes which have been reduced to murram through erosion. The major soil types are as follows – 1. Deep and shallow black soil with higher water retaining potential 2. Deep and medium red soils with low water holding potential 3. Mixed red and yellow soils with slightly better water retaining potential 4. Gravelly and pebbly murram soils with very low water retaining potential The underlying geological formations are primarily sedimentary and metamorphic in nature formed from Vindhyan shale and limestone and Deccan Trap basalts. 3.2.2. Landuse Pattern The land use pattern is as given below in Table 10 Sl. No.

Category

Area (ha.)

%age

1.

Agricultural Land

480

40

2.

Culturable waste land

120

10

3.

Encroached government land

15

1.3

4.

Pasture land

95

7.9

5.

Waste land

155

12.9

6.

Habitations and roads

15

1.3

7.

Forest land with the Revenue Department

120

10

8.

Forest land with the Forest Department

200

16.6

Total

1200

100

There are considerable amount of common lands in the watershed but apart from the forests under the Forest Department most of this land has been degraded due to the pressure of grazing animals and deforestation due to cutting for timber and other needs. The forests are of a tropical dry deciduous type with mainly teak trees with about 40% crown cover. There are in addition species

Addressing Equity Issues in Watershed Development Projects in Bhil Adivasi Areas

75

like anjan, tendu, neem, beeja, dhavra, khair while the shisham and tinach which are excellent timber trees are slowly disappearing. 3.2.3. Cropping Pattern The cropping pattern on agricultural land is given in Table 10 below. Table 11. Cropping Pattern Major Crops

Rainfed

Irrigated

Area (ha.) Yield* By-product (Kgs/ha)

Area (ha.) Yield* By-product (Kgs/ha)

Kharif Jowar

120

860

crop residue used as fodder

Redgram

65

800

crop residue used for fencing and fuel

Maize

10

900

crop residue used as fodder

Other Pulses

10

700

crop residue used as fodder

Hybrid Cotton

180

600

crop residue used as fuel

Indigenous Cotton

30

400

crop residue used as fuel

Soyabean

65

900

crop residue used as fodder

Total

480

-

Rabi Wheat

45

2000

crop residue used as fodder

Gram

15

600

crop residue used as fodder

Total

60

Summer Vegetables**

10

-

* The yields are for the best quality soils in the watershed as those for the lower quality soils is much less. ** Household production mainly so no data on yields are available.

76

Rahul Banerjee

There has been a slight increase in rabi cultivation after wateshed treatment due to greater availability of water but since this is constrained by various reasons to be discussed later this has not been very significant. There has been a shift away from jowar towards soyabean as part of a long term trend in this region and this cannot be attributed to watershed development. There has also been a shift from traditional varieties of jowar and maize to hybrid varieties which are more high yielding and shorter maturing. However cultivation of cotton has increased somewhat due to the greater availability of water. According to an estimate made by the NGO the erosion of soil before treatment of the watershed used to be as high as 21,000 tonnes per year. There has been a considerable reduction in this in the post treatment period. 3.2.4. Water Resources The rain falls mostly within the months of June through September in short bursts of a total of 45 days or so with a high degree of variability leading to frequent droughts. The undulating terrain, low water retentivity and low percolation of most of the soils means that runoff quotients are high and as much as 20 percent of the total precipitation is lost through this. Thus apart from the months of July, August and September the water balance in the other months is negative due to the effects of evapo-transpiration and severely so in the summer months. The rock formations underlying the northern part of the watershed are amenable both to recharge and storing of rain water but most of the village has hard shale and limestone strata which do not favour either. Consequently the shallow aquifer which is tapped by the dug wells is not able to support drawals for agriculture in most of the farms of the village and these go dry very soon as does the nala flowing through the watershed. There are as many as thirty wells but only about ten of these provide good water for irrigation. Nine tubewells have also been sunk but only six supply any reliable amounts of water the others having gone dry. There are four handpumps operating in the village providing water for drinking for people and livestock. In the summer months these handpumps become the lifeline of the village as most other sources go dry. There has been a hundred percent increase in the irrigated area since watershed development due to greater availablility of subsoil water and also investment by the NGO in deepening and construction of walls for wells and the private investment by some farmers in boring tubewells. However, since initially only 6 percent of the land was being irrigated this increase has been minuscule in absolute terms.

Addressing Equity Issues in Watershed Development Projects in Bhil Adivasi Areas

77

3.2.5. Livestock There are only local breeds of cows in the watershed numbering 250 big and small and no crossbreds. The farmers rely mostly on bullocks for agricultural operations and there are as many as 450 of these big and small. Interestingly there aren't any tractors in the watershed as people prefer to hire them when required from the nearby villages of Bhikupura and Ratatalai. While the cows are reared for milk for household consumption and also as progenitors of male calves to be honed into bullocks later buffaloes are reared for milk production on a commercial scale. Some of the milk is made into ghee but most of it is supplied to restaurants in Poojapura or to the cooperative dairy. The poor availability of water and the inefficient operation of the dairy severely restrict the production of milk in the watershed. Following the norm in adivasi households most of them have between five and ten goats which provide a reliable source of supplementary income. The sources of fodder are mainly the waste and forest lands and crop residues. Three households with better irrigation facilities grow some fodder crops. 3.2.6. Energy Sources Traditionally the people used to rely on fuelwood gathered from the forest and crop residues for household energy needs. Irrigation energy needs were met by animal power. With the coming of electricity the irrigation needs were met by this or by diesel powered pumps. The NGO has introduced household bio-gas units which have been successful and there are as many as twenty households who have these running and meet their fuel needs for cooking from this. This programme has been more successful in the nearby village of Sobliyapura where too the NGO has conducted watershed treatment works. Many people in nearby villages have now begun adopting these units after being impressed with the results. 3.3. Pre-Project Inequalities The two main inequalities preventing the implementation of a successful development programme in the watershed are the extra-economic powers of the non-adivasis and the government bureaucracy. The problems created by the two in the initial phases prior to watershed development are delineated below.

78

Rahul Banerjee

3.3.1. Non-adivasi Opposition The non-adivasis in the watershed and in the surrounding villages feared that the provision of work by by the NGO at statutory minimum wages for watershed development would soak up the surplus labour and push up wages for agricultural operations. The non-adivasis also thought that once the adivasis became organised they would rebel against their political hegemony and reorient the development policies and programmes in the region away from the interests of the non-adivasis. Consequently the non-adivasis vehemently opposed the work of the NGO and instigated the adivasi leaders of the area including elected members of various Panchayat institutions to prevent the NGO from striking roots among the people. This opposition also got strengthened with support from the local bureaucracy in the revenue and forest departments, which too was averse to the generation of any awareness among the adivasis that would reduce their extra-legal powers. Thus work had to be started not from the ridge which was under forest department control or even from the higher cultivable lands which were in the control of the non-adivasis but in the lowest part of the watershed in the Neemkheri hamlet whose adivasi inhabitants showed more gumption and decided to go along with the NGO when no one else would. The people of Neemkheri were even prepared to confront the adivasis of the nearby villages who were being instigated by the non-adivasis to oppose the work of the NGO. This show of toughness on their part eventually paved the way for the start of the soil and water conservation work. 3.3.2. Local Bureaucratic Opposition The Government of Madhya Pradesh website page on its forest department states that adivasis are the single largest destroyers of forest wealth and their organisation into mass movements by activists has led to problems for the Forest Department (www.mp.nic.in) This sums up the attitude of the bureaucracy not only in the forest department but across the board throughout the administration towards the organisations of adivasis that demand their rights. So it is not surprising that many administrative officials opposed the NGO and its mass mobilisational work. A complaint was rigged by the non-adivasis that the NGO was misusing the funds at its disposal instead of spending them in development works for the adivasis. The Subdivisional Magistrate in Bagli then took this opportunity to start an enquiry against the NGO and eventhough this did not reveal any malfeasance the official kept on harassing the organisation.

Addressing Equity Issues in Watershed Development Projects in Bhil Adivasi Areas

79

3.4. Resolution of Pre-Project Inequalities This nexus between the powerful non-adivasis and the government officials proved to be so frustrating that the NGO had to use lobbying tactics with higher level bureaucrats to ease the situation. Since the NGO was implementing the works in the first phase of development with funds from the Council for Advancement of People's Action and Rural Technology (CAPART), a central government development funding agency, it got it to recommend it to the district collector. The district collector took a personal interest in the work of the NGO and at their suggestion decided to hold a special camp in Neemkhera village in January 1995 to update all land records, which had not been done for a few generations. Thus on paper the landholdings were quite large when in fact they had been fragmented. This was depriving the adivasi farmers from the benefits due to them under several government schemes for marginal adivasi farmers. The cooperative credit system in Madhya Pradesh financed by the National Bank for Agriculture and Rural Development (NABARD) too is open to only those farmers who are registered landholders and have title to agricultural land in their own name. Thus the adivasis of the region were being deprived of this cheap source of credit also. There was an overwhelming response to this camp, which had to be extended to three days from the initial one day. The success of this huge event established the credentials of the NGO in the whole region and considerably reduced the power and enthusiasm of the non-adivasis. Simultaneously the payment of statutory minimum wages and the generation of employment in the watershed development works bolstered the incomes of the adivasis and increased their ability to withstand the might of the non-adivasis. Gopal, who was the first person to associate with the NGO, recounts with a smile how the other villagers had warned him not to hob-nob with these dubious outsiders as they would cause all kinds of problems. There was a rumour at that time that these outsiders had come to cheat the adivasis out of their lands on the false pretext of improving them. There were also doubts as to whether the NGO would indeed be able to pay the relatively high wages that it was promising on a sustained basis. However, when these payments were indeed made on a regular basis and when the collector held the revenue camp in Neemkhera through the aegis of the NGO all doubts were set to rest and the people began to come forward enthusiastically. Another important intervention by the NGO was the initiation of microcredit groups from 1995 onwards. These groups were linked to banks to increase disbursal amounts using the savings of the members as the margin. This

80

Rahul Banerjee

combined with the greater accessibility to credit from the cooperative society after the updating of land records considerably reduced the dependence of adivasis on the extortionate moneylenders. Over time these groups have become so efficient that they have completely freed its members from the clutches of moneylenders and even helped them to invest in their land. More importantly the women's groups have helped the women to become more vocal in the affairs of the community and also challenge traditional patriarchal structures in society. Finally an agreement was thrashed out with the forest department for the protection and treatment of the forest land on the ridge as this was crucial to soil and water conservation in the watershed. The villagers from Bhikupura too were convinced not to graze their animals in the forest in Neemkhera and the protection and treatment of the forest began. 3.5. Post-Project Inequalities and their Resolution The post-project inequalities arose over the use and management of the two natural resources of water and forests. While the disputes over water were resolved after some initial hiccups the dispute over forests, which is in a way related to water also has not been resolved thus severely constraining the possibility of sustainable development of the watershed. These inequalities and their resolution or otherwise are detailed below. 3.5.1. Inequality in Water use and its Resolution The recharge of water due to soil and water conservation works carried out in the ridge line forest area had led to the augmentation of the flow in the main nala flowing through the watershed from north to south and it had water in it throughout the rabi cropping season. The non-adivasi farmers who had their lands in the north on both sides of the nala began to use this excess water to bring more acreage under rabi cultivation. To do this they dug the bed of the nala and inserted concrete pipes to guide the water into their wells on the side of the nala. These were called "naarda"s. The adivasi farmers on the other hand sought a more equitable distribution. They argued with the help of the NGO that since the farmers in the southern part of the watershed were much poorer economically and had less access to ground water due to the peculiar underlying rock formations there they should have priority of use of the surface water of the nala and the farmers in the north should depend on the groundwater that they could tap through their wells. Thus, contrary to other watersheds where normally the upstream inhabitants lose out due to watershed development (Kerr,2002), here was a situation in which upstream farmers were in a better

Addressing Equity Issues in Watershed Development Projects in Bhil Adivasi Areas

81

position with regard to the use of the benefits of watershed development as compared to downstream farmers due to the underlying geology of the watershed. Moreover the upstream farmers were also economically and politically more powerful and so able to exploit this opportunity to the hilt. Sustained efforts at reversing this inequality and bringing about a more equitable use of the augmented nala flow finally yielded results as a formal agreement was signed by ninetynine percent of the co-riparians in October 1995 as follows – 1. All farmers would shut their naardas drawing water from the main naala. 2. All farmers who own wells would not use motorised pumps to draw water from the naala. 3. Only those farmers who did not have wells could take water from the naala for irrigation and these drawals would by turns in limited quantities according to commonly decided rules. 4. Once the flow in the naala stopped no drawal of water would be allowed and the remaining water would be exclusively for the livestock of the village. Following this all six naardas in the naala were packed with concrete and twenty farmers in the upstream area withdrew their pumps and diesel engines from the nala. As a result for the first time fifteen downstream coriparians were able to grow irrigated wheat and gram. This agreement was resisted by the nonadivasi farmers and they tried to make it inoperative with the help of people from outside the village but then the villagers sat in a dharna and imposed a social boycott on the former. This worked and eventually they had to come round to respecting the agreement, which is working very well under the supervision of the Village Watershed Committee. Thus as in many other cases of upstream-downstream conflict in this case too an amicable solution was found through the strengthening of communitarian organisations of the inhabitants of the watershed. 3.5.2. Inequalities in Forest Management and their Resolution Forest management, however, proved to be a stormier cup of tea. The local forest department officials had never liked the idea of the forest on the ridgeline being managed by the adivasis through a forest protection committee. They had to go along initially because the Divisional Forest Officer had to agree and sign the document that sanctioned the watershed development work in the village. However, they soon instigated the non-adivasis and the adivasis in Bhikupura

82

Rahul Banerjee

village to start grazing their animals in the Neemkhera forest. Complaints made by the Neemkhera forest protection committee to the district administration too did not result in any action against the trespassers. Soon it became difficult to prevent anyone, whether from Neemkhera or outside, from grazing their animals in the forest. So from 1996 onwards the forest has remained unprotected from grazing. This conflict arising from the intransigent attitude of the forest department has become even more acute because the people have come up with yet another scheme. Seeing that there was a limit to the development of groundwater irrigation in the watershed given its underlying rock structure which could not assure a large amount of storage the villagers hit upon the idea of building a large earthen dam on the main nala in the upstream part of the watershed where both storage and recharge were possible given the lie of the land. In this way it would be possible to ensure irrigation by gravity through canals to the whole of the watershed and especially the southern part. But this would involve submerging one hundred hectares of forest land. Naturally the forest department has rejected this proposal outright. The people proposed that they would fully protect the remaining hundred hectares of forest all the year round and plant saplings in the degraded revenue wastelands, which are not being used for anything other than grazing and restock them to compensate for the loss due to the reservoir but this plea has fallen on deaf years. Consequently the inequality arising out of the adverse control over the forest lands being exercised by the forest department has not been resolved and so the potential for more intensive development of this land and its use for the wider spread of irrigation in the watershed as a whole remain unrealised. 3.6. Implications Activists of the NGO have written up their experiences in the Neemkhera watershed as a comprehensive book on the development of dryland areas inhabited by adivasis (Shah et al, 1998). The plan for achieving sustainability of natural resource management and livelihood practices in the watershed within ten years that they had chalked out in that book has a few serious flaws which are the bane of watershed development generally in this country. To tackle the obvious and ubiquitous opposition of the local bureaucracy it has been proposed that the state at a higher level should evolve policies and programmes that are pro-poor and involve NGOs and Panchayati Raj Institutions (PRI) for their proper implementation. Eventhough their own experiences with the PRIs hadn't been very encouraging they felt that the new

Addressing Equity Issues in Watershed Development Projects in Bhil Adivasi Areas

83

Panchayati Raj Extension to Scheduled Areas Act (PESA) would provide an enabling framework for a better involvement of the adivasis in local development processes. Needless to say their expectations have been hopelessly belied. The character of the state is uniformly anti-people right from the lowest rung upto the topmost levels. The bureaucracy and the politicians are more interested in lining their pockets and not in providing good governance. In fact the adivasis in the whole of the Bagli tehsil of which Neemkhera is a part had organised themselves into a strong mass organisation Adivasi Morcha Sangathan (AMS) to put a stop to this corruption of the administration and also wrest control of the forests from the forest department to better husband these crucial resources in accordance with the provisions of PESA. The result was that in 2001 in a sudden crackdown sanctioned by the Government of Madhya Pradesh at the highest level the members of the AMS were severely repressed and the organisation smashed. This anti-people nature of the state in Madhya Pradesh has proved to be the single most perverse obstacle in the path of sustainable development of the adivasis in particular and the rural poor in general. This is an inequality that has no easy way of being internalised. The strengthening of PRIs provide a way out. The Hariyali Guidelines of the Government of India state that the government watershed development programmes should be implemented through PRIs. However, to prevent the misappropriation of these funds it is necessary for the Gram Sabhas or the general bodies of the Panchayats to be active and involved in local governance and this can only come about through the proliferation of mass organisations of the poor. Lobbying with higher level officials may serve the limited purpose of establishing the credentials of the NGO as in the present case but it cannot bring about any longterm and far reaching changes in policy without mass mobilisation on the concerned issues. Watershed development alone cannot bring about sustainable development. There may be various local factors, like the underlying geology in Neemkhera, which cannot be tackled through watershed development alone. Thus, sometimes, bigger water storage structures will have to be contemplated and these have major policy and legal implications. For instance the building of the dam in Neemkhera will be possible only if permission is granted by the Ministry of Environment and Forests at the Centre as it involves the submergence of a hundred hectares of standing forests. The NGO has not addressed this major question at all and benignly assumed that the state will turn pro-poor and take care of such anomalies suo moto! There is thus a need to build in more flexibility in natural resource management policies and rules.

84

Rahul Banerjee

Finally the NGO has not seriously addressed the question of sourcing of inputs and marketing of agricultural produce like the other NGO Sampark in Petlawad has done. Though it has discussed and done some experimental work in the introduction of newer less water intensive crop varieties, these have not taken root among the farmers. More attention needs to be given to the optimisation of cropping on marginal sized farms so as to make them both nutritionally and commercially viable. So while the power of local exploiters can be readily neutralised through the organisation of the beneficiaries of a watershed development programme the de facto opposition of the state and its local functionaries flying in the face of de jure policies for the development of the poor is a much tougher nut to crack as is the control of input and output markets by traders and manufacturers.

4. Conclusions The performance of SAMPARK and SAMAJ PRAGATI SAHAYOG can be compared to that of the RGWM as a whole under implementation contemporaneously all over Madhya Pradesh so as to highlight the crucial parameters of successful implementation of watershed programmes faced with local and structural inequalities. A critical independent review of the work of the RGWM has brought out the following major deficiencies of the project – 1. The increase in wage employment in the agricultural sector wasn’t significant enough to neutralise the accompanying growth in workforce. A tendency to search for more remunerative and stable employment in urban areas was marked, this made a decline in migration levels minimal. 2. Increases in cropped area and crop-mix changes had differentially benefited medium and large farmers. While only 25 percent of marginal farmers had reported crop-mix changes, small landholding sizes and limited capacity to invest in water extraction technologies had constrained their potential benefits. 3. There is a need for special attention towards SHGs and WTCGs given (i) requirement of full-time specialised professional input and (ii) complex issues relating to repayment, adequacy of capital support, procurement and marketing linkages

Addressing Equity Issues in Watershed Development Projects in Bhil Adivasi Areas

4.

5.

85

and profit sharing arrangements that demand serious attention both at the operational and strategic levels. Lack of an initially agreed maintenance strategy which makes the fate of public water harvesting structures contingent on expectation of negotiated settlement later on Inadequately structured initial mobilisation that (i) reaches out only to select individuals and ignores the silent majority and risks perpetuating traditional power structures along with their less desirable traits and (ii) does not lay upfront the terms of engagement in terms of responsibilities and obligations of various village level groups. (RGWM/TARU op cit) This review clearly reveals how important it is for equitable and sustainable distribution of benefits in a watershed project to – i. organise the beneficiaries of the watershed project first before embarking on the physical implementation itself, ii. generate alternative sources of income to tackle the problem of chronic unemployment which leads to migration and iii. address the problems related to capital support and market linkages.

Given these problems that surface in micro watershed impementation the World Environment Conference in Rio-de-Janeiro in 1992 gave rise to the concepts of River Basin Management and Integrated Water Resource Management to make possible more holistic planning and a Global Water Partnership was established to push these ideas for adoption in the water sector on a large scale (GWP, 2000). However, scholars and practitioners have criticised this concept too for being narrowly underpinned by neo-liberal principles, dominated by technical and managerial concerns and concentrating on water conservation while neglecting the vital resource of human labour and questioned its applicability in third world countries like India (Mollinga, 2006). Subsequently there is at present an ongoing attempt to redefine and reformulate the concept to make it suitable to the management of natural resources and promotion of sustainable livelihoods in Indian conditions (Shah & Prakash, 2007). Thus there is a distinct need to look beyond watersheds and increase the size of natural resource planning units.

86

Rahul Banerjee

Additionally according to the World Bank deeply biased credit systems and the inefficient and non-transparent functioning of regulatory and support institutions result in powerful players in rural markets hurting the poor with their operations (World Bank, 2001). Thus SAMPARK and SAMAJ PRAGATI SAHAYOG by building up people's institutions that can effectively tackle these problems has clearly scored decisively over the RGWM. The fundamental difference lies in the approach to watershed development. The Government of Madhya Pradesh treats decentralised watershed management and dryland agriculture as adjuncts to modern flood irrigated green revolution agriculture and not as central programmes that should replace it. Moreover the building up of people's institutions that can challenge established power structures within and without the government is beyond it. The methodology developed by SAMPARK and SAMAJ PRAGATI SAHAYOG is eminently suited to adivasi regions where there is still extant a traditional communitarian culture even if under threat of disintegration. It can be easily replicated with success in other adivasi areas. Indeed SAMPARK has gone one step further in being able to harness the market forces to the advantage of the adivasis. The adivasis in Petlawad have been ingenious enough to make the best of both worlds. The increase in crop yields and livestock earnings has in many cases obviated the need to migrate for work any more. Nevertheless the adivasis in Roopapada do migrate for short spells when there is a lull in activity on their own fields in the kharif season. In this way they earn enough money to defray the expenses of the sowing of the rabi crop. Thus they can hold on to the excess kharif crop instead of selling it immediately and so get a better price for it. Thus nowadays instead of the adivasis going to the traders, it is the latter who come to the village vying with each other to get the adivasis to sell to them. What more eloquent ode can there be to the virtues of well-organised and efficiently implemented comprehensive watershed development as a viable means of bringing about equitable and sustainable development. Also of note is the fact that both the NGOs have included women's empowerment through micro-credit groups in their project design itself and followed this up with robust implementation and this has resulted in the empowerment of women and their equal participation in community affairs as also in the challenging of patriarchal social structures. This is very much in line with current thinking on women's empowerment through micro-credit (Burra et al, 2005). However, a caveat has to be entered that there are still some serious obstacles which cannot be dealt with at the micro-watershed level. Consequently

Addressing Equity Issues in Watershed Development Projects in Bhil Adivasi Areas

87

the concept of social capital which is relevant to some extent at the local level has come to be criticised because it is inadequate when it comes to the design of strategies to counter the larger political economy of modern development (Harris, 2001). The local state and the local power centres may be successfully neutralised through the formation of social capital in one small area but unless such isolated successes are replicated on a larger scale across dryland adivasi areas there cannot be any widespread change in the developmental situation of the adivasis. Thus as mentioned earlier the many NGOs and mass organisations of the western Madhya Pradesh region tried to utilise the empowering provisions of the Panchayat Extension to Scheduled Areas Act 1996 (PESA) to institutionalise such an equitable and sustainable developmental model as developed by SAMPARK and SAMAJ PRAGATI SAHAYOG all over the region by empowering the panchayats. However, this movement for increasing autonomy of the adivasis was not tolerated by the higher level state and political parties and in a grossly repressive and illegal crackdown this mobilisation was squashed with the killing of four adivasis in police firing in Mehendikhera village in Dewas district in 2001 (Rahul, 2001). So the creation of social capital is just one necessary condition of equitable sustainable development and for sufficiency it must be complemented by the other necessary condition of the creation of political capital of the adivasis on a wider scale to be able to influence development policies in their favour. The National Rural Employment Guarantee Act 2005 (NREGA) and the The Right to Information Act 2005 (RTIA) now provide further legal instruments to make this possible. However, because of the lack of political empowerment of the adivasis these provisions too are not being availed of to their advantage by them. Unfortunately there is a lack of understanding in official circles of this dire need for political empowerment of the poor and especially of the adivasis on a massive scale as a sine qua non for the proper implementation of watershed development programmes in the face of inequalities in power at the macro level as evidenced in the latest report on watershed development submitted recently to the government (GOI, 2006). References 1. Aurora, G. S. Tribe-Caste-Class Encounters: Some Aspects of Folk-Urban Relations in Alirajpur Tehsil, Administrative Staff College, Hyderabad. (1972). 2. Banerjee, R. Status of Informal Rural Financial Markets in Adivasi Dominated Regions of Western Madhya Pradesh, Working Paper No. 2,

88

3. 4.

5. 6. 7. 8. 9. 10. 11. 12. 13. 14.

15. 16.

Rahul Banerjee

Department of Economic Analysis and Research, NABARD, Mumbai.(2003) Burra, N., Deshmukh-Ranadive, J. & Murthy, R K. Microcredit, Poverty and Empowerment: Linking the Triad, Sage Publications, Delhi.(2005) Corbridge, S., Williams, G., Veron, R. & Srivastav, V. Making Social Science Matter: How the Local State Works in Rural Bihar,Jharkhand and West Bengal, Economic & Political Weekly (EPW), 38 : 24 & 25, Mumbai.(2003) D’Silva, E. & Pai, S. Social Capital and Development Action: Development Outcomes in Forest Protection and Watershed Development, EPW, 38 : 14, Mumbai. (2003). Fan, S. & Hazell, P. Should India Invest More in Less Favoured Areas, Environment and Production Technology Division Discussion Paper No 25, International Food Policy Research Institute, Washington, DC. (1997). Govt. of India, Ministry of Rural Development. Report of the Technical Committee on Drought Prone Areas Programme and Desert Development Programme, Delhi. (1994). GOI.From Hariyali to Neeranchal: Report of the Technical Committee on Watershed Programmes in India, Department of Land Resources, Ministry of Rural Development, Government of India, Delhi. (2006). Government of Madhya Pradesh, Department of Agriculture. Danida Comprehensive Watershed Development Project, Petlawad. (1997) Harris, J. Depoliticising Development: The World Bank and Social Capital, LeftWord, Delhi. (2001) GWP. Integrated Water Resources Management, Technical Advisory Committee Background Paper No. 4, Global Water Partnership, Stockholm.(2000) Kerr, J. Watershed Development, Environmental Services and Poverty Alleviation in India, World Development Vol 30, pp 1387-1400. (2002) Mollinga, P. IWRM in South Asia: Global Theory, Emerging Practice and Local Needs, Sage, Delhi. (2006) Mosse, D., Gupta, S., Mehta, M., Shah, V., Rees., J and KribP Project Team. Brokered Livelihoods: Debt, Migration and Development in Tribal Western India, The Journal of Development Studies, Vol, 38 No.5, London. (2002) National Bank for Agriculture and Rural Development. Report of the Task Force on Supportive Policy and Regulatory Framework for Microfinance in India, Mumbai.(1999) Nathan, D. & Kelkar G. Civilisational Change: Markets and Privatisation among Indigenous Peoples, Economic & Political Weekly, Vol, 38 No 20. (2003).

Addressing Equity Issues in Watershed Development Projects in Bhil Adivasi Areas

89

17. Rahul. The Bottle that does not Cheer: Bhil Women’s Fight Against Male Oppression and Alcoholism, Manushi 113, New Delhi. (1999). 18. Rahul. The Bhils : A People Under Threat, Humanscape, September, Mumbai. (2001) 19. Rahul & Subhadra. Schooling of Tribals in Madhya Pradesh: Problems and Prospects, Journal of Educational Planning and Administration Vol. XV No. 1. (2001). 20. RGWM/TARU.Evaluation of RGWM Watersheds in Madhya PradeshFinal Report for UNICEF, New Delhi-Hyderabad, TARU Leading Edge. 21. SAMPARK. (1995). Looking Backward to Leap Forward: Annual Report 1994-95, SAMPARK, Peltlawad.( 2001). 22. Shah, A. & Prakash, A. IWRM in India: From Critique to Constructive Engagement, Mimeo. (2007). 23. Shah, M., Banerji, D., Vijayshankar, P. S. & Ambasta, P. India's Drylands: Tribal Societies and Development through Environmental Regeneration, Oxford University Press, Delhi. (1998). 24. Shah, P. Participatory Watershed Management Programmes in India: Reversing Our Roles and Revising Our Theories in Rural People’s Knowledge, Agricultural Research and Extension Practice, IIED Research Series, Vol 1 (3), IIED, London. (1993). 25. Sharma, B D.Tribal Affairs in India: The Crucial Transition, Sahayog Pustak Kutir Trust, Delhi. (2001). 26. World Bank. World Development Report 2002: Building Institutions for Markets, Oxford University Press, London. (2001). 27. Yugandhar, B. N. Watershed Based Development in Arid and Semi-Arid Areas of Andhra Pradesh, Journal of Rural Development, Vol 18 No 3, Hyderabad. (1999). 28. Government of Madhya Pradesh - http://www.mp.gov.in/default.htm

This page intentionally left blank

Politics. Law and Economics

This page intentionally left blank

April 20, 2009

14:2

WSPC - Proceedings Trim Size: 9in x 6in

CK5

A THEORY OF THE CORRUPT KEYNESIAN TOKE S. AIDT Faculty of Economics and Jesus College, University of Cambridge Cambridge CB3 9DD United Kingdom E-mail: [email protected]. JAYASRI DUTTA Department of Economics, University of Birmingham E-mail: [email protected]. We evaluate the impact of real business cycle shocks on corruption and economic policy in a model of entry regulation in a representative democracy. We find that corruption is pro-cyclical and regulation policy is counter-cyclical. Corrupt politicians engage in excessive stabilization of aggregate fluctuations and behave as if they were Keynesian. We also find that business cycle shocks can induce political instability with politicians losing office in recessions. JEL classification: D72; K42; O41. Keywords: Corruption; entry regulation; performance voting; business cycles.

1. Introduction Do politicians collect larger bribes in booms than in recessions? Do they introduce excessive entry restrictions to create the artificial scarcity needed to collect those extra bribes? Do corrupt politicians engage in excessive stabilization of aggregate fluctuations? Are corrupt politicians Keynesian? Can business cycle shocks induce political instability? The aim of this paper is to provide some answers to these questions. We evaluate the impact of real business cycle shocks on corruption and economic policy in a model of entry regulation in a representative democracy.∗ A leading example of the type of entry regulation that we have in mind comes from India. It takes the form of comprehensive systems of industrial licensing that “ ..sought to regulate domestic entry and import competition, ... to penalize unauthorized expansion of capacity, ... and indeed to define and delineate virtually all aspects of investment and production through a maze of Kafkaesque controls”(Bhagwati1 pp. 49-50). Elected ∗ The

model is similar to the one developed in Aidt and Dutta (in Press). In that paper, we use the model to study the relationship between growth and corruption. 93

April 20, 2009

14:2

94

WSPC - Proceedings Trim Size: 9in x 6in

CK5

Toke S. Aidt and Jayasri Dutta

governments constructed that maze from 1950 onwards and it started to be dismantled in the 1990s, in efforts initiated by yet other elected governments. Similar systems developed in other countries in the region, such as Bangladesh and Pakistan (Srinivasan2 ). The cost of complying with multiple legal requirements and red tape is another example of the type of entry restrictions we want to capture. This phenomenon is extensively documented by De Soto3 in his seminal study of the legal obstacles that a would-be entrepreneur has to go through to operate a firm legally in Peru. He shows that it would take more than 300 days of work at a cost of 32 times the monthly minimum wage to get the permits and approvals needed to set up a small two-sewing machine clothing factory in a Lima shanty town. No wonder that many would-be entrepreneurs prefer to stay informal or are tempted to pay bribes to get the paperwork done faster. The corruption potential in economies with excessive entry regulation is enormous, and it is not surprising that empirical studies find that corruption levels and measures of entry regulation are strongly correlated: excessive entry regulation and corruption go hand in hand (e.g., Treisman;4 Djankov et al.;5 Paldam6). It is also interesting to note that after the licensing system was dismantled, India’s score on Transparency International’s corruption perception index improved from around 2.7 in the mid-1990s to 3.5 in 2007. At the same time, Sharma7 reports that industrial de-regulation during the 1980s led to a significant rise in firm productivity. The tight connection between regulation of economic activity, allocative inefficiency, and corruption forms the cornerstone of our model: entry restrictions are implemented and maintained by corruptible politicians because of their corruption potential. In the model, governments can regulate entry into the production sector by issuing production licenses. Output and wages increase, and profits decline with the number of licenses, or the degree of liberalization. This sets the stage for social conflict. Workers earn wages, and would like to see the licensing system abolished. Entrepreneurs would like a license for themselves, as it allows them to earn super-normal profits. Politicians are elected by majority rule. Once in office, they can restrict the number of licenses and charge for the ones they issue. This is the source of corruption.† Their bribe income depends on having the licensing † This concept of corruption is similar to the “grabbing hand view” of government advocated by Shleifer and Vishny.8,9 For an overview of the vast literature on corruption, see Bardhan,10 Rose-Ackerman,11 and Aidt12

April 20, 2009

14:2

WSPC - Proceedings Trim Size: 9in x 6in

A Theory of the Corrupt Keynesian

CK5

95

system in place. The majority of the population are workers, and they lose out. They attempt to control politicians by holding them accountable for their actions while in office. To this end, they set performance standards, and vote a politician out of office if he is too corruption and his performance fails to comply with the standard, as in Ferejohn,13 Coate and Morris14 and Persson et al.15 Importantly, the economy is subject to (real) business cycle shocks. These impact directly on wages, profits and output and are propagated by the licensing policy. We study the cyclical properties of economic policy (industrial licensing) and corruption in this environment. It matters greatly for outcomes whether shocks are observed by voters or not. If voters can make their performance standards contingent on observed business cycle conditions, it is constrained efficient to induce politicians to behave as if they were Keynesian. To get reelected, they must restrict entry into the economy in a boom and allow entry in a recession. Economic policy entails excessive stabilization of aggregate fluctuations in a corrupt democracy and, as a consequence, corruption is pro-cyclical. Politicians collect bribes in a boom, less so in a recession. In contrast, when shocks are unobserved, they can induce political instability, as voters may rationally vote politicians out of office in recessions in order to discipline them in booms. This makes entry regulation pro-cyclical and corruption counter-cyclical. It is well-documented empirically that corruption depends on economic factors such as the level of GDP, the growth rate of output, inflation etc. (e.g., Treisman;4 Paldam6 ). We are, however, not aware of any studies that evaluate corruption at the business cycle frequency. The existing theoretical literature studies the link between economic development (economic growth) and corruption.‡ The focus is on the long run rather than on short run implications of corruption. The main contribution of this paper is to make a beginning at closing this gap. We do so by proposing a theory of corruption and business cycle shocks. Before we present the theory, however, it is instructive to look at some data on the cyclical properties of the industrial licensing system in India. Table 1 reports the correlation between the number of industrial licenses issued (or the number of factories) and the Solow residual (in the previous year) for the pre-liberalization period (1975-1989) and the post-liberalization period (1990-2003), respectively. The correlations are conditional on unobserved state fixed effects ‡ See,

for example, Murphy et al.,16 Parente and Prescott,17 Krusell and Rios-Rull,18 Blackburn et al19 and Aidt et al20

April 20, 2009

14:2

WSPC - Proceedings Trim Size: 9in x 6in

CK5

Toke S. Aidt and Jayasri Dutta

96

and the number of firms in the previous year. We observe, firstly, that the licensing policy is statistically significantly affected by business cycle conditions. Insofar as industrial licenses are a major source of corruption, this is indirect evidence that corruption has a cyclic component. Secondly, we notice that the licensing policy is pro-cyclical in the period 1975-1989, but the number of firms has been counter-cyclical since deregulation – we pick 1990 arbitrarily: deregulation of licenses started in the late 1980’s and was essentially completed by 1993. In other words, the cyclical properties of the number of firms seems to have changed dramatically after the liberalization initiated in the 1990s.§ Table 1. The Relationship between the number of industrial enterprises and the Solow residual in India, 1975-2003. Dependent variable Period Constant Log(factories)t−1 Solow Residualt−1 Number of states Number of observations

Log(factories)t 1975-1989

Log(factories)t 1990-2003

3.32 (7.27) 0.66 (15.06) 0.14 16 240

0.99 (3.10) 0.87 (24.07) -0.04 16 208

Note: The regressions include state fixed effects. t-statistics in brackets.

The rest of the paper is organized as follows. In Section 2, we set out the economic model. In Section 3, we describe the political system. In Section 4, we study regulation policy and corruption in an economy that is subject to (real) business cycle shocks. In Section 5, we conclude. 2. The Economy We consider an economy with a continuum of individuals, indexed by j, with measure 1.¶ The size of the population is constant. Time is discrete, indexed by t = 0, 1, 2, · · · . Each individual has one unit of labor each period. A homogeneous consumption good, y, is produced every period. Individuals live for ever, consume their net income each period, and derive no utility § The source is Annual Survey of Industries data for 1975 -2003. Solow residuals calculated in the usual way. ¶ The specification of the economy is inspired by Lucas.21

April 20, 2009

14:2

WSPC - Proceedings Trim Size: 9in x 6in

CK5

A Theory of the Corrupt Keynesian

97

from leisure. Utility is linear in consumption. The discount factor is β ∈ (0, 1). At any point in time, an individual can either be a worker or an entrepreneur. Workers supply labor to a competitive labor market. Entrepreneurs run firms and supervise workers. The firm owned by entrepreneur j produces with the following production technology: α yjt = At s1−α jt jt ,

0 < α < 1,

(1)

where jt denotes the labor input hired by entrepreneur j; sjt denotes the time spend on supervision by entrepreneur j; and At is the level of technology, common to all firms. Profits are retained by the entrepreneur who runs the firm. A would-be entrepreneur needs to obtain a license to operate a firm from the government. The politician running the government can choose the number of licenses and determine who gets them. A license confers the right, but not the obligation to operate a firm for one period. License holder j chooses how much time to spend on supervision, sjt ∈ [0, 1], and supplies the remaining part of her time endowment to the labor market. Non-license holders have no choice of occupation. They work full time for a firm and earn the real wage, wt . The real wage adjusts to clear the labor market each period. Let λt ∈ [0, 1] be the number of licences issued in period t. We lose nothing by assuming that licenses are held by individuals j ∈ [0, λt ]. The state of the economy at time t is summarized by et = (At , λt ). In our analysis, the stochastic process that drives At is exogenous, while λt is endogenously determined. Let nt ≤λt be the number of firms operating n in period t. National income is Yt = 0 t yjt dj. For any sequence of states {e0 , · · · , et , · · · }, with et ≥ 0, an equilibrium of the economy is a sequence {· · · , (nt , Yt , wt ), · · · } such that all individuals and firms optimize, and the labor market clears each period. We write πjt = yjt − wt jt as the equilibrium profit level of firm j at time t. At a symmetric equilibrium, πjt = πt . Proposition 2.1 establishes that the equilibrium is stationary: the number of firms, employment, and incomes depend only on the current state of the economy. Proposition 2.1. Let et = (At , λt ) be the state of the economy at time t. An equilibrium exists whenever et > 0. Let λH = (1 − α). Then equilibrium quantities and incomes are functions of the current state of the economy only n(et ) = min[λt , λH ];

Y (et ) = At n(et )1−α (1 − n(et ))α ;

April 20, 2009

14:2

WSPC - Proceedings Trim Size: 9in x 6in

CK5

Toke S. Aidt and Jayasri Dutta

98

w(et ) = α

Y (et ) ; 1 − n(et )

π(et ) = (1 − α)

Y (et ) . n(et )

Furthermore, π(et ) = w(et ) if and only if λt ≥ λH ; otherwise π(et ) > w(et ). For all et , the number of workers is greater than or equal to α. National income, Yt , is maximized at nt = λH . Wages increase and profits decrease with λt whenever λt < λH . National income, wages, and profits increase with At for all λt ∈ (0, 1]. Proof. See Appendix.

When the number of licenses issued is less than λH , all licenses are fully utilized and they carry a scarcity rent, i.e., πt > wt . The number of firms is nt = λt and the licensing system imposes a binding constraint on entry and output: the economy is allocative inefficient. When the number of licenses is greater than (or equal to) λH , the economy is fully liberalized. Licenses are no longer scarce and some are not utilized in equilibrium. The number of firms is nt = λH and each license holder is indifferent between being a full time entrepreneur or a full time worker, i.e., πt = wt . Liberalization achieves allocative efficiency and maximum national income. Workers welcome this, while entrepreneurs do not, as they see profits decline. This distributional impact is central to our analysis. Positive productivity shocks increases national income, wages and profits proportionally. Negative shocks has the opposite effect. In an economy with λH firms, these fluctuations are efficient. 3. A Representative Democracy We wish to study the determination of entry regulation and corruption in societies with representative democracy. In a representative democracy, voters delegate decisions to elected politicians, who once in office, are free to design the licensing system as they see fit. Voters can respond after the fact and hold the politician accountable for past decisions, as in Ferejohn.13 Proposition 2.1 shows that the fraction of workers is at least α. We assume that α > 1/2 and so the majority of the population are workers. For simplicity, we refer to the worker-voters as the voters. Formally, the incumbent politician runs against a challenger in the election held at the  Although

entrepreneurs can also vote, it is without loss of generality that we focus exclusively on the voting behavior of workers.

April 20, 2009

14:2

WSPC - Proceedings Trim Size: 9in x 6in

CK5

A Theory of the Corrupt Keynesian

99

end of each period. He is reelected for another term if he gains a majority. At the beginning of his tenure, voters announce an election rule, ηt (.), specifying the probability of reelection as a function of observable indicators of the politician’s performance.∗∗ We restrict attention to threshold election rules that specify a performance standard that the politician has to satisfy to get reelection. That is, ηt (.) = 1 if the standard is satisfied and zero otherwise. The fact that a license to run a firm can have economic value suggests that it can be sold at a price. The incumbent politician has a temporary monopoly on the sale of licences and is tempted to sell government property for personal gain.†† Each period, the incumbent chooses λt , and the price, bt , at which he sells each license. Accordingly, the politician’s bribe income is: Bt = λt bt .

(2)

Lemma 3.1 evaluates the bribe function, relating the number of licenses to the maximum surplus that can be extracted. Lemma 3.1. The incumbent politician prices each license at bt where    α 1−α  λ 1−λ −α , 0]. (3) bt = max[At (1 − α) λ 1−λ The politician’s bribe income, Bt (λt , At ) = λt bt , is maximized at λt = λL ≡

 1 (2 − α − (4 − 3α)α) 2

(4)

with 0 < λL < λH . λL is independent of At while the maximized bribe income is proportional to At . Proof. See Appendix. In the absence of elections, the politician extracts the maximum bribe, B(λL , At ), every period by setting λt = λL . Since λL < λH , the bribe maximizing policy imposes excessive regulation. The intuition follows from Proposition 2.1. A license is valuable only if it is scarce. Liberalization ∗∗ The

constrained efficient performance standard may be specified in terms of the number of licenses or in terms of utility levels depending on circumstances and on the information available to voters. †† This is the definition of corruption given by Shleifer and Vishny.8

April 20, 2009

14:2

100

WSPC - Proceedings Trim Size: 9in x 6in

CK5

Toke S. Aidt and Jayasri Dutta

reduces scarcity and the price each license commands. We note that politicians can, ceteris paribus, for a given license policy extract more rent in a booming economy than during a recession. Politicians care about holding public office for many reasons. One of them is that power allows them to make money, because they can sell government property and earn Bt . We assume that the payoff of the politician in office at time t is upt = Bt .

(5)

We normalize the payoff of politicians out of office to zero. We assume that there is an unlimited supply of potential politicians willing to serve. Politicians apply the same discount factor as citizens. We can now define the game between politicians, workers, and would-be entrepreneurs, as it unfolds over time. Workers earn the market wage and get utility uw t = wt . Entrepreneurs have to pay the bribe, bt , to obtain their license. Lemma 3.1 implies that entrepreneurs get per-period utility uet = πt − bt = wt . The timing of events is as follows. At the beginning of each period, a politician is already in office. Voters announce a performance standard. Next, the politician chooses how many licenses to issue and at what price. Would-be entrepreneurs can accept or reject the offer of a license at the announced price.‡‡ Once bribes and licenses have been exchanged, production takes place. Finally, at the end of each period, an election is held. The outcome of the election is determined by the policy implemented by the incumbent relative to the standard. After that, the sequence of events repeats itself. With regard to the business cycle shock, we shall consider two scenarios. In one scenario, the shock is realized at the beginning of each period and observed by everyone. In the other, we assume that voters cannot observe business cycle conditions directly, nor can they infer them from observing their wage income. This effectively means that we assume that voters cannot observe policy directly. We continue, however, to assume that the politicians can observe the shock and tailor his policy to it. We require that voters, given the information they hold about within-period events, set the performance standard such that their life-time utility is maximized subject to the sequence of incentive compatibility constraints and subject to equilibrium in the private sector. ‡‡ We could assume that the surplus is being split more evenly between the politician and the entrepreneurs. This would bring out the underlying conflict of interest between workers and entrepreneurs more clearly. However, since this is not important for the results, we focus on the simpler case where the politician has all bargaining power.

April 20, 2009

14:2

WSPC - Proceedings Trim Size: 9in x 6in

A Theory of the Corrupt Keynesian

CK5

101

4. Corruption and the Business Cycle From Proposition 2.1, we know that the level of technology together with the licensing policy determine all variables of economic interest at each t. Outcomes, hence, depend critically on the sequence of technology levels. An implication, then, is that corruption varies with the business cycle. Business cycle shocks are propagated by the licensing system which is the only propagation mechanism operating in the model. Since the allocation of resources is efficient in the absence of a licensing system, any stabilization of aggregate fluctuations introduced by the system is inefficient and excessive. It matters greatly for the nature of these inefficiencies, however, whether shocks are observed by voters or not. As mentioned above, we consider two cases. In the first case, voters observe the state of the business cycle before they announce their performance standard. In the second case, voters neither observe the state of the business cycle, nor the policy choice (or the level of corruption). The politician, on the other hand, observes the shock before setting the licensing policy. 4.1. The Corrupt Keynesian To keep it simple, suppose that the stochastic process for technology shocks is given by  1 + μ with probability p ≥ 0 , (6) At = 1 with probability 1 − p and that the shocks are independent over time. The economy is in a boom if At = 1 + μ > 1 and, else, in a recession. We interpret μ as a measure of the amplitude of the cycle. The probability p, on the other hand, can be interpreted as a crude measure of persistence. If, for example, p is close to one booms are almost permanent in the sense that At is almost always 1 + μ. Voters observe the state of the business cycle before they announce the performance standard for the period. In this case, it is without loss of generality that we specify the election rule as a function of the observed policy directly, i.e.,  1 iffλt ≥ λs (At ) s . (7) ηt (λt ; λ (At )) = 0 otherwise Since business cycle conditions are known at the time when the standard is set, it is optimal to tailor the performance standard to business cycle

April 20, 2009

14:2

102

WSPC - Proceedings Trim Size: 9in x 6in

CK5

Toke S. Aidt and Jayasri Dutta

conditions. In particular, let  λ(At ) =

λB if At = 1 + μ λR if At = 1

(8)

be the state dependent performance standard used by voters. Proposition 4.1. Define λB = max{ λ| (1 + μ)B(λB ) = (1 − β + μ (1 − pβ)) B(λL )} λR = max{ λ| B(λR ) = (1 − β − μpβ) B(λL )}.

(9) (10)

The constrained efficient licensing policy is (1) λt = λB if At = 1 + μ (2) λt = λR if At = 1 with λH > λR > λB > λL . Proof. Let voters announce the performance standard given in equation (8). If period t is a boom, the value function of the politician is vtB = (1 + μ)B (λB ) + β max vt+1

(11)

and if period t is a recession, the value function is given by vtR = B (λR ) + β max vt+1 .

(12)

B R We note that vt+1 = pvt+1 + (1 − p) vt+1 . In either case, if the politician chooses a policy below the standard, he is replaced by the challenger at the next election and his continuation payoff is zero. Alternatively, he can choose a policy at or above the standard and be reelected. The payoffs associated with these two options are denoted viD (.) and viC (.), respectively, for i ∈ {R, B}. The politician chooses λt = λs (At ) if and only if the following conditions are satisfied

v(λs (At )) = max viC (λt ),

(13)

v(λs (At )) ≥ viD (λL )

(14)

λt

where i = B if A = 1 + μ and i = R if A = 1. The first condition is satisfied whenever λs (At ) > λL because B  (.) ≤ 0 for λt ≥ λL . The second condition

April 20, 2009

14:2

WSPC - Proceedings Trim Size: 9in x 6in

CK5

A Theory of the Corrupt Keynesian

103

– the incentive compatibility condition – requires that vtB ≥ (1 + μ)B(λL ) and vtR ≥ B(λL ), respectively. Solving equations (11) and (12), we get B (λR ) β(1 − p) + (1 + μ) B (λB ) (1 − β (1 − p)) ; (1 − β) B (λR ) (1 − βp) + B (λB ) βp(1 + μ) vR = . 1−β

vB =

(15) (16)

The constrained efficient performance standard solves v B = (1 + μ)B(λL ) and v R = B(λL ). A simple calculation yields the expressions given in equa2 tions (9) and (10). Notice that (1−β+μ(1−pβ)) − (1 − β − μpβ) = βμ+pβμ > 1+μ 1+μ 0. Since B  < 0, we conclude that λR > λB . We note that λR < λH and λB > λL Corollary 4.1 (The corrupt Keynesian). Corruption is pro-cyclical and economic policy is counter-cyclical, i.e., entry regulation is lax in a recession and strict in a boom. Proposition 4.1 shows that economic policy is more inefficient during booms than during recessions. Since inefficient economic policy by itself reduces output this phenomena can be interpreted as active Keynesian stabilization policy driven by the desire of corrupt politicians to collect bribes. The other side of the coin, then, is that corruption is pro-cyclical. A booming economy presents greater temptations, and politicians stand to gain more from selling favors. As a consequence, societies must concede more to dishonest politics. The intuition is straightforward. An increase in national income raises the stakes because politicians can potentially extract much larger bribes. They are, therefore, more likely to defect from a given standard. Realizing this, voters are willing to accept more entry restrictions and higher levels of corruption during a boom than during a recession. An alternative intuition is that politicians want to get reelected in recessions so that they can be around to collect large bribes in booms. This makes it easier for voters to discipline politicians in a recession. The distortion in economic policy is increasing in the amplitude of the cycle (μ). This is simply because larger fluctuations in output enhance the temptation to collect bribes when the economy is booming. A high degree of persistence (as captured by a larger p) makes it harder for voters to reduce corruption and to promote efficient licensing policies both in booms and recessions. The reason is that a high p makes the temptation to collect bribes almost permanent. This makes it harder for voters to discipline their

April 20, 2009

14:2

104

WSPC - Proceedings Trim Size: 9in x 6in

CK5

Toke S. Aidt and Jayasri Dutta

politicians and corruption is as a result high most of the time. An implication of the analysis, then, is that corruption tends to be high, on average, in societies with large, but relatively persistent movements in technology. 4.2. Unobserved Shocks and Political Instability Unobserved and unanticipated productivity shocks may result not only in cyclical movements in economic policy, but also in political instability. In fact, it may be constrained efficient for voters to set performance standards that politicians cannot comply to in some states of the world. To see this, suppose that politicians have an informational advantage over workers. While politicians can observe the state of the business cycle before deciding on the licensing policy for the period, workers cannot observe neither At nor λt . They only observe their wage income wt = At w (λt ). Workers must therefore specify the performance standard in terms of utility levels (incomes), rather than in terms of policy outcomes. We continue to assume that the stochastic process for At is given by equation (6) and that the shocks are independent over time. We restrict attention to values of μ that satisfy the following condition:

w(λH ) H) where λB ∈ (λL , λH ) is Assumption 4.1. μ ∈ w(λ w(λB ) − 1, w(λL ) − 1 defined in equation (9). The assumption ensures two things. It guarantees that workers can demand higher utility levels in booms than in recessions. A sufficient condition H) for this is that the amplitude of the cycle, μ, is larger than w(λ w(λB ) −1. On the other hand, if the the amplitude is too large, workers can in effect design a utility standard that replicates the constrained efficient, state dependent solution characterized in Proposition 4.1. In particular, this is possible if it is impossible in a recession for a politician who liberalizes the economy completely to deliver the lowest possible wage income that can be delivered in a boom ((1 + μ) w (λL )). In this case, workers simply ask for (1+μ)w(λB ) if w > (1 + μ) w (λL ) and for w (λR ) if not where λB and λR are defined in Proposition 4.1. Politicians will comply to this, and outcomes are as if workers could observe the cycle directly. To rule this possibility out, we H) assume that μ < w(λ w(λL ) − 1. This implies that observing wage income is not This rules out the possibility that workers can use any information that they learn about the state of the business cycle in one period to predict what the state might be in the following period.

April 20, 2009

14:2

WSPC - Proceedings Trim Size: 9in x 6in

CK5

A Theory of the Corrupt Keynesian

105

sufficient to deduce if wages are high because the economy is booming or because of liberalization. Since workers can neither observe nor, under Assumption 4.1, deduce the state of the business cycle, the performance standard must be state independent, i.e.,  1 iffwt ≥ ws s , (17) ηt (wt ; w ) = 0 otherwise where ws is the utility threshold required for reelection. Faced with the performance standard ws politicians must implement a state dependent licensing policy in order to be reelected. Denoting the best response to the standard ws in state i by λi (ws ) for i = B, R, we can write the incentive compatibility constraints in the two states as IC B : vtB = (1 + μ) B (λB (ws )) + βvt+1 ≥ (1 + μ) B (λL ) IC R : vtR = B (λR (ws )) + βvt+1 ≥ B (λL ) B R where vt+1 = pvt+1 + (1 − p) vt+1 . We can make the following preliminary s observation. Since w is the same in a boom as in a recession, if the politician tries to comply, he must issue more licenses in a recession than in a boom. Importantly, this implies that the politician may not always be willing to comply in a recession. In particular, we can show the following result.

Lemma 4.1. Let Assumption 4.1 be satisfied. If workers set ws such that the politician is just willing to seek reelection in a boom (IC B binds), then the politician will not seek reelection in a recession (IC R fails). Proof. Assumption 4.1 implies that workers can demand higher utility in return for reelection in a boom than in a recession. Let the highest utility standard that politicians will comply to in a boom be wB and let the corresponding utility standard in a recession be wR . The corresponding number of licences issued are λB (wB ) and λR (wR ). Suppose that reelection requires delivery of at least wB at all times. In a recession, the politician must issue licences λR to satisfy wB = w(λR ) where λR > λR (wR ). It follows immediately that the politician will not comply to this. Whenever ws is set to make IC B bind, he accordingly deviates and sets λ = λL in a recession

Lemma 4.1 basically shows that workers face a trade off. If they want the politician to implement the best possible licensing policy in a boom,

April 20, 2009

14:2

106

WSPC - Proceedings Trim Size: 9in x 6in

CK5

Toke S. Aidt and Jayasri Dutta

they cannot get him to comply in a recession. They may, therefore, consider setting a standard that the politician will, in fact, satisfy in both states of the cycle. In this case, we can show the following result. Lemma 4.2. Let Assumption 4.1 be satisfied. If workers set ws such that the politician is just willing to seek reelection in a recession (IC R binds), then the politician will also seek reelection in a boom (IC R is non-binding). Proof. Suppose that ws = wR is such that IC R binds. In a boom, the politician can satisfy this standard by issuing licenses up to the point where wR = (1 + μ) w (λB ). Assumption 4.1 implies that λB < λB (wR ) and it follows immediately that the politician will comply

The two lemmas allow us to restrict attention to two types of performance standards. Performance standard PS1 is such that the politician only complies in a boom, while performance standard PS2 is such that he always complies, and receives an additional rent in a boom. We begin by characterizing the performance standard of type PS1 that maximizes the lifetime utility of workers. Suppose for this purpose that workers set a standard wP S1 that makes IC B bind. Lemma 4.1 implies that the politician does not comply in a recession, and the value function associated with that state of the cycle is vtR = B(λL ). Given that, we can write the value function in a boom as  B vtB = (1 + μ) B λB (wP S1 ) + βpvt+1 + β (1 − p) B(λL ). We can solve this equation to get  (1 + μ) B λB (wP S1 ) + β (1 − p) B(λL ) B v = . 1 − βp

(18)

(19)

(20)

To maximize their wage income in a boom, workers set wP S1 such that v B = (1 + μ) B(λL ). To satisfy this standard, the politician would have to S1 S1 ∈ (λL , λH ) licences in a boom where λP is the largest λ that issue λP B B solves

B (λ) =

1 − β + μ (1 − βp) B(λL ). 1+μ

(21)

April 20, 2009

14:2

WSPC - Proceedings Trim Size: 9in x 6in

CK5

A Theory of the Corrupt Keynesian

107

S1 We noticethat λ P = λB . The optimal reelection threshold is then wP S1 = B P S1 (1 + μ) w λB . The expected lifetime utility of a worker is  S1 + (1 − p) w (λL ) p (1 + μ) w λP B P S1 . (22) U = 1−β

In a similar fashion, we can characterize the performance standard of S2 detype PS2 that maximizes workers’ lifetime utility. Let wP S2 and λP R note the utility threshold required for reelection and the licensing policy required in a recession to achieve the threshold, respectively. Lemma 4.2 implies that the politician will also want to satisfy the utility threshold

P S2  −1 w P S2 in a boom and can do so by setting λB = w w 1+μ . Given that, we can write the value functions associated with compliance in the two states as S2 B R vtR = B(λP R ) + βpvt+1 + β (1 − p) vt+1

(23)

B R vtB = (1 + μ) B(λB ) + βpvt+1 + β (1 − p) vt+1 . B

Solving for v and substituting the result into the expression for v rearranging yields vR =

(1 − βp) βp (1 + μ) S2 B(λP B(λB ). R )+ 1−β 1−β

(24) R

and

(25)

To maximize their wage income in a recession subject to compliance, workers set wP S2 such that v R = B(λL ). To achieve this, the politician issues S2 S2 licenses where λP is either the largest λ that solves λP R R 

B (λ) =

(1 − β)B(λL ) − βp (1 + μ) B(λB ) 1 − βp 

(26)



or λH if (1 − β)B(λL ) − βp

(1 + μ) B(λB ) < 0. λB is implicitly defined  P S2  by w λR = (1 + μ) w λB . The optimal utility threshold of type PS2 S2 accordingly is ws = w(λP R ) ≤ w (λH ). The lifetime utility of workers is  S2 w λP R P S2 U . (27) = 1−β

Comparing the maximized lifetime utility associated with the two types of performance standards, we get the main result of the analysis. Proposition 4.2. Suppose that Assumption 4.1 holds. There exists a p ∈ (0, 1) such that for p > p, it is constrained efficient for voters to use performance standard PS1.

April 20, 2009

14:2

108

WSPC - Proceedings Trim Size: 9in x 6in

CK5

Toke S. Aidt and Jayasri Dutta

Proof. A direct comparison of equations (22) and (27) yields that performance standard PS1 is better than performance standard PS2 iff  S1  S2 p (1 + μ) w λP + (1 − p) w (λL ) > w λP . (28) B R  P S2  P S1 > w (λH ) We notice w  λR ≤ w (λH ). It follows that (1 + μ) w λB S1 S2 > w λP . Assumption 4.1 ensures that this is the ⇒ (1 + μ) w λP B R case. It then follows that for p close enough to 1, inequality (28) must hold Corollary 4.2. (Political instability) Unobserved real business cycle shocks induce political instability. Politicians lose office during recessions and are reelected in booms. The proposition and the corollary establish that it can be constrained efficient for workers to induce political instability. This happens when recessions are unlikely and the amplitude of the cycle is moderately large. The intuition is that workers want politicians to deliver as efficient a policy in a boom as possible but is unable to tell when the economy is booming. Unfortunately, politicians are unwilling to replicate this in a recession. They then forgo reelection and collect the maximum bribe. Workers are willing to accept this inefficiency when recessions are rare. We observe that in contrast to the case where the performance standard can be tailored to the cycle, performance standard PS2 magnifies rather than dampens aggregate fluctuations. That is, more licenses are issued in a boom than in a recession. Moreover, measured by bribe income relative to GDP, corruption is counter-cyclical. Lots of bribes are collected in a recession because politicians make no attempt to get reelected, while in a boom they must pander to voters to stay in office. This result is consistent with evidence from numerous studies of vote and popularity functions. This literature shows that incumbent politicians are much more likely to be reelected when economic conditions are benign than when they are not (see, e.g., Nannestad and Paldam, 1994). This empirical regularity is usually interpreted as evidence that politicians are rewarded for good performance. Our analysis, however, suggests an alternative interpretation: it is rational for voters to ask too much of their politicians in recessions and that is why they only get reelected in booms. 5. Conclusion In a corrupt democracy, corruption levels and regulation of economic activity fluctuate systematically with the business cycle. We show that corrupt

April 20, 2009

14:2

WSPC - Proceedings Trim Size: 9in x 6in

CK5

A Theory of the Corrupt Keynesian

109

politicians behave like Keynesians when voters can tailor their performance standards to business cycle conditions. Economic policy entails excessive stabilization of aggregate fluctuations in a corrupt democracy and, as a consequence, corruption levels are pro-cyclical. Politicians collect bribes in a boom and hold back in a recession. In contrast, when voters cannot observe business cycle conditions, it may be constrained efficient to induce political instability with turnover of politicians in recessions. In this case, entry regulation magnifies aggregate fluctuations and corruption becomes counter-cyclical. Acknowledgments We would like to thank the ESRC for research support (grant no. L138251006). We also thank the participants in the International Conference on Comparative Development at the Indian Statistical Institute in December 2007 for constructive comments. Part of this paper was written when Toke Aidt visited the Indira Gandhi Institute of Development Research (IGIDR) in Mumbai in December 2007. He is grateful for the hospitality of the IGIDR and thanks the Cambridge Bombay Society for financial support.

Appendix Proof of Proposition 2.1. For each λ > 0, individuals j ≤ λ are license holders, and have the right to choose sj > 0 and employ workers in their firm. Suppose sj (e) > 0. Profit maximization implies  1  αA 1−α j (e, w) = sj w and yj = As1−α α j ≡ sj y(w); j

πj = (1 − α)yj ≡ sj π(w).

A license holder earns π(w)sj + w(1 − sj ) which is maximized at sj = 1 whenever π(w) > w. In this case, all licences are used, i.e., n(e) = λ and the total supply of labor is 1 − λ. Labor market clearing requires that λj (e, w) = 1 − λ. Therefore, equilibrium national income, the wage rate, and profit per firm satisfy Y (e) = Aλ1−α (1 − λ)α ;

w(e) = α

Y (e) ; 1−λ

π(e) = (1 − α)

Y (e) . λ

April 20, 2009

14:2

WSPC - Proceedings Trim Size: 9in x 6in

CK5

Toke S. Aidt and Jayasri Dutta

110

From these, we obtain the condition π(e) > w(e) ⇒ λ < (1 − α) ≡ λH . Suppose λ ≥ λH . Let n ≤ λ. Firms maximize profits and all labor is employed. Equilibrium national income, the wage rate, and profit per firm satisfy Y (A, n) Y (A, n) ; π(A, n) = (1−α) . 1−n n Note that n > 0 ⇒ π(A, n) ≥ w(A, n) from the occupational choice of individuals j ≤ λ; that n = λH is the unique solution to π(A, n) = w(A, n); and that π(A, n) < w(A, n) whenever n > λH . This establishes that π(e) = w(e) ⇔ λ ≥ λH and that n(e) = λH for λ ≥ λH . We see that 1 − n(e) ≥ α for all e. Finally, write

Y (A, n) = An1−α (1−n)α ;

w(A, n) = α

Y (e) = An(e)1−α (1 − n(e))α  w(e) = αA

n(e) 1 − n(e)

with

n(e) = min[λ, λH ]; 

1−α ;

π(e) = (1 − α)A

1 − n(e) n(e)

α .

We note that Y, w and π are monotonically increasing in A; that π and w1 decrease with n; and that Y attains its maximum at n = λH . Lemma 3.1. A license is valid for one period. Its “price”, bt , cannot exceed its value to the holder, i.e., bt ≤ π(λt , At ) − w(λt , At ).

(A.1)

The politician extracts the entire surplus and so, condition (A.1) is binding. The total bribe is B(λt , At ) = λt (π(λt , At ) − w(λt , At )) .

(A.2)

The bribe function is concave and differentiable, with B(0, At ) = 0 = B(λH , At ), limλ→0 B  (0, At ) = ∞, and B  (λH , At ) ≤ 0. Hence, the total bribe income is maximized at some λL ∈ (0, λH ). Note that λL is stationary, and independent of productivity At . Thus, we can write B(λL , At ) = At B(λt ). References 1. J. N. Bhagwati, India in Transition – Freeing the Economy (Oxford: Clarendon Press, 1993). 2. T. N. Srinivasan, Economic reforms in south asia, in Economic Policy Reform – The Second Stage, ed. A. Krueger (University of Chicago Press, Chicago, 2000)

April 20, 2009

14:2

WSPC - Proceedings Trim Size: 9in x 6in

A Theory of the Corrupt Keynesian

CK5

111

3. H. D. Soto, The Other Path: The Invisible Revolution in the Third World (Harper, New York, 1990). 4. D. Treisman, Journal of Public Economics 76, 399(June 2000). 5. S. Djankov, R. L. Porta, F. Lopez-De-Silanes and A. Shleifer, The Quarterly Journal of Economics 117, 1(February 2002). 6. M. Paldam, European Journal of Political Economy 18, 215(June 2002). 7. G. Sharma, Competing or collaborating siblings? industrial and trade policies in india, Unpublished working paper, Department of Economics, University of Missouri, (2007). 8. A. Shleifer and R. W. Vishny, Quarterly Journal of Economics 108, 599(August 1993). 9. A. Shleifer and R. W. Vishny, Journal of Economic Perspectives 8, 165(Spring 1994). 10. P. Bardhan, Journal of Economic Literature 35, 1320(September 1997). 11. S. Rose-Ackerman, Corruption and Government – Causes, Consequences and Reform (Cambridge University Press, Cambridge, 1999). 12. T. S. Aidt, Economic Journal 113, F633(November 2003). 13. J. Ferejohn, Public Choice 50, 5 (1986). 14. S. Coate and S. Morris, American Economic Review 89, 1327(December 1999). 15. T. Persson, G. Roland and G. Tabellini, Quarterly Journal of Economics 112, 1163(November 1997). 16. K. M. Murphy, A. Shleifer and R. W. Vishny, Quarterly Journal of Economics 106, 503(May 1991). 17. S. Parente and E. C. Prescott, Barriers to Riches (MIT press, Cambridge MA, 2000). 18. P. Krusell and J.-V. Rios-Rull, Review of Economic Studies 63, 301(April 1996). 19. K. Blackburn, N. Bose and M. E. Haque, Journal of Economic Dynamics and Control 30, 2447 (2006). 20. T. S. Aidt, J. Dutta and V. Sena, Journal of Comparative Economics 36, 195(November 2008). 21. R. E. Lucas, Bell Journal of Economics 9, 508(Autumn 1978).

This page intentionally left blank

April 20, 2009

14:17

WSPC - Proceedings Trim Size: 9in x 6in

isid07˙instnchange3

ON EFFECTING INSTITUTIONAL CHANGE SUMON MAJUMDAR Department of Economics, Queens University Kingston, Ontario E-mail: [email protected] SHARUN MUKAND Department of Economics, University of Warwick and Tufts University E-mail: [email protected] In this paper, we examine the role of policy intervention in engineering institutional change. More specifically, in a framework where the underlying political incentives determines the quality of a region’s economic institutions, we examine the role of broad developmental policy in improving institutions and therbey welfare. Two effects emerge. First, by increasing political accountability, such policies encourage democratic governments to invest in good institutions – the incentive effect. However, such developmental policies also increase the incentive of the rentier elite to tighten their grip on political institutions. In some cases, this latter political control effect can outweigh the former incentive effect, and result in an overall deterioration of institutional quality. However, it may also indirectly encourage the elite to modernize. Taking this possibility into account, the framework has the potential to explain a diverse set of possibilites as a result of developmental policy.

1. Introduction Institutions matter. There is growing consensus that the quality of a country’s institutions matter for its growth and development.a An important issue that faces a policymaker is how does one build better institutions? Of course attempts at institutional reform are not new. However, the most striking aspect of the institution building experience across the globe is that institutional reform is not easy. Most attempts at reform end up in failure as the experience of countries such as Venezuela, Bolivia and Kenya attest. In this paper we describe a simple framework to examine the relationship between the quality of political and economic institutions, and consequently the impact of policy interventions on them. We ask whether development a Recent

contributions to this empirical literature include Hall and Jones,1 Rodrik, Subramaniam and Trebbi2 and Acemoglu, Robinson and Johnson,3 to name a few. 113

April 20, 2009

14:17

114

WSPC - Proceedings Trim Size: 9in x 6in

isid07˙instnchange3

Sumon Majumdar and Sharun Mukand

policy can be a successful catalyst for helping bring about improvement in both sets of institutions? Addressing this classic question allows us to also throw light on the various degrees of success or failure at institution building that different policies may achieve. As is now commonly recognized, institutional quality is not an exogenously given structure but is impacted by the interaction of competing economic and political interests. Accordingly, our framework follows Acemoglu and Robinson4 in describing institutions as the product of the interests of competing groups which control political power. Consider a country which has two groups. One group is in a minority but form an economic “ elite” and can be considered to be primarily rentier-landlords; members of the other majority group are primarily wage-earners and are poorer. The quality of the region’s economic institutions is a function of the government’s policy and effort at improving such basic fundamentals as law and order, protection of property rights, contract enforcement etc. The trouble is that the current (backward) institutional structure is conducive to earning rents by the elite in a non-competitive environment; any change/improvement to the existing institutional set-up that may encourage other entrepreneurs to invest is likely to adversely affect the rents earned by the elite. The elite’s motivation to control the levers of government is driven by a desire to avoid the adverse distributional outcomes that institutional reform may entail. Whether in fact the elite can do so depends of course on the nature of the country’s political institutions. There are two countervailing pressures from the two competing groups at work in determining the government’s incentives. We show that democratic elections (i.e. “ power” in the hands of the majority) need not automatically result in good economic institutions if political incentives are poor. For a region plagued with weak economic fundamentals, elections do not provide enough of a reward (to good governance) for a democratic government to escape the clutches of influence by the elite. The democratic process, in effect, remains captured by the elite, and consequently economic institutions remain of poor quality. Indeed this picture has been observed in Colombia, Venezuela and other Latin American countries, where the introduction of democracy did not change the economic status-quo. Similarly, as noted by Acemoglu and Robinson4 despite the abolition of slavery with the Civil War in the U.S., relative underdevelopment and low wages persisted in the South well into the twentieth century. Can intervention by a policymaker who is external to the region or country provide the prospect of institutional change and economic improvement?

April 20, 2009

14:17

WSPC - Proceedings Trim Size: 9in x 6in

On Effecting Institutional Change

isid07˙instnchange3

115

This formulation captures a number of plausible scenarios. For instance, this “ external” policymaker may be a country such as the U.S. or an international agency such as the U.N. confronting the task of transforming institutions in say, Afghanistan or East Timor. Alternatively, this external intervenor may be the central/federal government attempting to improve both the quality of democracy and the quality of economic institutions in a backward province.b In either case, we consider the effectiveness of broad developmental policy which encourages investment in a region, be it through investment in infrastructure (thereby reducing the cost of doing business there), or by tax-breaks and subsidies for those whose invest in the region. We identify two channels through which such a policy can impact both political and economic institutions in the region. The first channel is what we call the incentive effect of development policy. By raising accountability and rewarding good governance, such a policy encourages the government to strengthen economic institutions and improve property rights. Indeed by doing so it also simultaneously improves the strength of the underlying political institutions. However, there is a second effect at work. By encouraging investment, development policy gives rise to the spectre of a large loss in economic rents by the elite. This prospect of an erosion in economic rents gives the elite an incentive to tighten its grip and deploy additional resources to control the levers of government. Through this channel of a political control effect, development policy can also have the adverse effect of potentially undermining political institutions.c In Mexico, Fox7 cites the case of development policy in the backward Chiapas province; this increased the “voice” of the endogenous people in the region, but at the same time, cases of election malpractices also went up. This double-edged aspect of policy intervention is worth emphasizing. Under some conditions, the incentive effect is strong enough to ensure that development policy results in not just better protection of property rights, but also transforms democracy by freeing government from the elite’s grip. When the political control effect outweighs

b Of

course, in many instances we have change being spurred by individual leaders who spot a ‘window of opporunity’ for change and persuade the populace to make the costly investments to transform the status-quo. In Majumdar and Mukand5 we elaborate on such a model of the role of leadership in catalyzing institutional change. c One of the main advantages of democracy as a political institution is that it promotes political selection of good quality candidates (for a discussion see Besley6 ). As will be evident later in the context of our framework, greater political control by the elite prevents elections from effectively sorting the good from the bad politicians.

April 20, 2009

14:17

116

WSPC - Proceedings Trim Size: 9in x 6in

isid07˙instnchange3

Sumon Majumdar and Sharun Mukand

the incentive effect, a benign development policy can result in an overall deterioration of the country’s institutions. It implies that there could be a non-monotonic relationship between resources allocated towards development and its effectiveness as a tool to build institutions. When either the resources allocated are too few or too much, developmental policy is ineffective, the former due to a too weak incentive effect, while the latter case is due to a too overwhelming political control effect. It is an intermediate range of resource allocation that makes development policy most effective. This result thus provides an important cautionary note in the use of development policy as a tool to transform institutions.d However, influencing government policy is costly for the elite, either directly in terms of monetary costs or in terms of yielding on other noneconomic issues. As development policy raises an incumbent government’s rewards from electoral accountability, it also increases the amount of resources the elite need to devote to successfully influence economic policy. We argue however that developmental policy can also have the secondary effect of prompting the elite to change their technology closer to the frontier so as to be less dependent on an insular institutional setup for their profits. In doing so, they diminish their own incentives to control government policy on improving institutions in the region, and thereby eliminate their own de facto political control. Overall this can lead to sharp improvements on both the political and economic fronts. If however the elites are completely entrenched, in that their traditional technology is very far from the technological frontier and/or the costs of reorganization for them are too large, development policy is unlikely to erode their “political control” of government. This raises the question as to whether the use of coercive instruments (such as forcible seizure and redistribution of the elite’s assets) by the external policymaker may be more effective in such a situation. At a broader level our analysis suggests that even in a democracy with regular elections, sometimes the use of coercive technology may be the only way to improve democratic functioning and have security of property rights. Relatedly, whether the instrument being used is developmental or coercive, our framework suggests the importance of the policymaker’s ability in recognizing the underlying situation. In particular, leaders with a good

d For instance, Sachs8 (and Bono) has been arguing that greater resource allocation towards developmental policy “ will make poverty history” by transforming (among other things) institutions and governance. This notion has been adopted in the 2005 G8 summit for increasing resource allocations for developmental policy towards the developing world.

April 20, 2009

14:17

WSPC - Proceedings Trim Size: 9in x 6in

On Effecting Institutional Change

isid07˙instnchange3

117

knowledge of local conditions and a talent for recognizing the right ‘window’ of opportunity are the most effective at ensuring permanent and dramatic institutional change (see Majumdar and Mukand5 ). The recent literature on the evolution of political institutions (see Acemoglu and Robinson4 ) is clearly related to this paper. Exploring rationales for voluntary extension of the franchise, this work emphasizes the threat of revolution by the disenfranchised majority (Acemoglu and Robinson4 ), the elites’ aim of improving welfare by reducing the space for narrow redistributive political competition (Lizzeri and Persico9), and the role of economic cleavages and group formation within the elite (Llavador and Oxoby10 ). This positive analysis of voluntary internally-influenced democratization is clearly important in enhancing our understanding of the sources of the spread of democracy. However, especially since World War II, there have been many instances where the spur to democracy has been from direct and indirect forms of external influence. Such projects of institutional engineering has had mixed results. On the one end we have successes such as Japan, Germany and East Timor while on other end we have notable failures such as Somalia and Haiti. Attempts at spreading democratization and better institutions in backward regions of countries such as Brazil and Mexico have also had limited success. Similarly, the initiation of five-year plans during the fifties in India was with the explicit objective of politically empowering the effectively disenfranchised in many parts of India along with promoting better economic institutions in some states in India; again the results have been mixed. In this paper we take a first step in exploring the effects of policies aimed at bringing about comprehensive institutional change. Thus at a broader level, our contribution is related to the literature on institutional change initiated by North.11,12 However unlike us, much of this literature focuses on internal mechanisms which engender gradual institutional change, such as shifts in technology and resources (Greif 13 ) or a stochastic shock (Roland14 ). On the political economy front, our paper is also related to the literature examining the relationship between institutional structure and political accountability. This literature explores the effect of different institutional setups (e.g. democracies versus autocracies (Persson and Tabellini15 ), the basis of political power – broad or narrow (de Mesquita et. al.16 )) on political accountability, corruption and related phenomena. While related to this literature, our contribution explores the effect of political accountability on the institutional structure itself and how changes in one can (or cannot) bring about changes in the other. Further, we emphasize that the elite un-

April 20, 2009

14:17

118

WSPC - Proceedings Trim Size: 9in x 6in

isid07˙instnchange3

Sumon Majumdar and Sharun Mukand

dermine the quality of democracy by interfering with the process of political selection that is intrinsic to the successful functioning of a democracy. In line with recent work by Jones and Olken17 and Besley,6 our framework emphasizes the importance of political selection and leadership for good governance. We emphasize that in imperfect democracies, political selection is constrained and good leaders may be prevented from emerging, despite free and fair elections. The rest of the paper is organized as follows. The next section describes the basic framework connecting the political process with the (endogenous) nature of institutions. Analyzing the various effects at play in determining the equilibrium of this interaction allows us to also identify the effect of developmental policies on both political and economic institutions, which we do in section 2.1. Section 2.2 considers the possibility of the elite to modernize, and how the incorporation of this possibility leads to a rich set of possibilities in terms of institutional change. Section 3 briefly concludes. 2. Changes Due to External Policy: A Framework As argued in the introduction, some degree of institutional change could be the result of externally imposed policy. Although many instances of external intervention comprise the use of force to directly change the political landscape, we will focus here on the effects of much less coercive policies. More specifically, we wish to discuss the effects of broad developmental policies on the political and institutional structure in a society. While coercive external intervention attracts much popular attention, the scope for it is often narrow and usually involves too many constraints on various dimensions to make it a viable option except in extreme cases. Thus policies based on economic incentives are often a more relevant option. This is specially so when discussing institutional change in a particular province or region within a given country. Constitutional constraints may mean that even for a well-meaning federal government, recourse to forceful means to improve institutions in an economically-backward province is not an option. Instead other policies need to be considered. In this paper, we wish to sketch a sparse political economy framework to discuss the various impacts that developmental policies can have on the political structure in a region and consequently on the quality of the institutions there. Economic Structure Our aim is to first consider a simple structure to analyze the relationship between politics and the underlying institutions

April 20, 2009

14:17

WSPC - Proceedings Trim Size: 9in x 6in

On Effecting Institutional Change

isid07˙instnchange3

119

in the region, and to consider the effects of externally imposed policies on this relationship. To that end, let us consider the economy as consisting of two groups with a conflict of interest — landowning elites and labor. The elites use a labor-intensive technology to produce output from land; being dependent on labor, elites would like to maintain low wages so as to reap maximum profits. On the other hand, for the majority of the population, their only endowment is labor. They of course would wish wages to be as high as possible. If there is an entry of other investors who also have a demand for labor, then wages rise and the general population gains from it. The elites’ interests are of course the opposite: their profits diminish when wages increase; thus they would like to restrict the entry of investors who will raise the demand for labor. Although stylized, this structure captures a simple conflict of interest between two groups in the economy. There are many alternate formulations of the same basic structure. For example, one could also consider the elites as having monopoly access to a particular sector of the economy. Since the entry of other investors would erode their monopoly rents, the elites would be opposed to such entry. On the other hand, the population at large would gain from such competition. We assume that there are several potential investment opportunities in the province. However, investors into this province fear that their output or returns from the investment may get appropriated or stolen. Thus, the effectiveness of property rights and the local law and order situation is crucial to their decision on whether or not to invest. If these institutions function well in the region, investors are easily attracted there and consequently it is beneficial for the majority of the population. Our interest here is to understand what factors affect the quality of these institutions which form the backbone of economic transactions. To develop each investment opportunity requires the investment of some capital and the use of local labor. The magnitude of these costs of course play a crucial role in the decision of potential investors to invest in this region or not. Infrastructure is usually an important element that affects these investment costs. When basic infrastructure is better, lesser amount of capital investment is presumably required as not all facilities need to be privately provided. For example, if the power supply is unreliable, separate private electricity generators may need to be set up, or if public roads are in a poor condition, the costs of transportation are higher and need to be taken into account. In such cases, attracting investors maybe difficult. This is an area where external policy can be of help. For example, develop-

April 20, 2009

14:17

120

WSPC - Proceedings Trim Size: 9in x 6in

isid07˙instnchange3

Sumon Majumdar and Sharun Mukand

mental projects can improve the infrastructure and reduce the amount of capital that needs to be brought in from outside to invest in this province Alternately, changes in the taxes or subsidies to investment into this region can also reduce the cost of doing business here. Our objective is to analyze the impact of such changes on institutions in the province and on overall welfare. Policies and the Political Structure As described above, the effectiveness of institutions that facilitate economic transactions is crucial in potential investors’ decisions on whether or not to invest in this province. By virtue of their monopoly over the legal use of force, governments have a big role to play in determining the climate of contract enforcement, property rights and the law and order situation in the region. The effectiveness of such institutions is thus not an exogenous factor and depends very crucially on action (or inaction) by the local government in their implementation. Our aim in this framework is to endogenize the quality of institutions in the province as a function of the underlying political structure and the incentives there. Ensuring good institutions and an investment-friendly climate is a multifaceted task requiring both skill, initiative and experience of the government. We assume that the level of protection in the province is a function of the government’s ability, resources and experience in matters of effective governance. Ability and experience are required to determine which are the particular sectors that need the most attention, which bureaucrats are the best-suited to assign to such tasks that promote investor protection and how to inspire the government machinery towards prioritizing such tasks. At the same time, initiative, effort and resources are required to implement policies and stand up to elements that maybe disruptive to a healthy business climate. Without such initiative and resources, even talented governments will probably not find much success in improving the quality of institutions in the region. However, doing all of these is costly and requires time and resources to be diverted from other activities (or from self-gratification). What determines the government’s incentives for such tasks? Our aim is to analyze the various countervailing forces at play in determining the government’s policies. Although it maybe a region with poorly developed property rights, let us assume that this province is part of a larger nation in which the basic structure of democracy, namely regular elections, gets implemented. As is often observed, while the central/federal government may not be able to

April 20, 2009

14:17

WSPC - Proceedings Trim Size: 9in x 6in

On Effecting Institutional Change

isid07˙instnchange3

121

directly yield influence over the day to day activities of provincial governments, even in developing countries it may at least be forceful enough to uphold the conduct of regular elections. Similarly even at a national level, international pressure is often great enough to ensure that elections get held at regular intervals. We will thus assume that at the end of every period the incumbent government comes up for re-election, at which stage it faces a randomly drawn challenger in an election and the electorate may decide to retain it or choose a new government into power. How do these electoral incentives impact the government? The electorate (the majority of whom are labor-suppliers) here consists of identical agents whose objective is to choose the government that is most likely to gain them the maximum welfare. We assume that while potential investors in the region can observe the level of protection and the investment climate in the province, ordinary citizens are unable to judge the nitty-gritty details of the overall level of security. However, by observing whether or not investors have decided to put down their capital in the province, citizens can infer the level of property rights protection and hence the ability and policies adopted by the incumbent government. Such models of imperfect observability of the government’s ability and policies and the resulting impact on the incumbent’s incentives have been widely used in the political economy literature (Majumdar, Mani and Mukand18 is a particular example of such a model; Persson and Tabellini15 contain other examples). By raising wages, investment benefits the majority. Thus the population would prefer to have in office a government of high ability, which is more likely to lead to good institutions and thereby attract investors. To focus purely on the countervailing pressures exerted by the two groups in the economy, we assume that governments care only about the overall rents that they accumulate. These rents could be those from remaining in office, or from the payoffs that interested agents may pay the government in order to influence its policies. Traditional Elite As described before, while investment in the province improves employment opportunities, and thus welfare of the majority of citizens, traditional rents of the landowning elites are imperiled. In a more general framework, elites may fear their monopoly rents getting eroded in the face of competition. These provincial elite, either by virtue of their local informational or enforcement advantage do not require governmental protection to operate, and would thus like to maintain the current status-quo of a low level of institutional protection which dissuades outside investors from investing in the province.

April 20, 2009

14:17

122

WSPC - Proceedings Trim Size: 9in x 6in

isid07˙instnchange3

Sumon Majumdar and Sharun Mukand

The traditional elite would thus like to influence the government to not devote resources into property right protection, thereby enabling them to maintain their monopoly hold on labor and keep wages low. We assume that the elites are organized into a lobby group that offers a bribe to the government in exchange for not enforcing a regime of good property rights protection and keeping institutional quality poor. Thus the political structure here is a very simple version of the Grossman-Helpman19 model of special interest politics where lobbies offer contributions to the government in return for implementing their preferred policies. The difference is that here we have only one lobby actively at work; the majority’s influence on the government comes from their power at the ballot. A second aspect of the setup also deserves comment. While the elites here seek to influence only government policy as regards institutional quality, one may ask whether elites can influence policy even more deeply. For example, by allowing in outside investors, and getting the government to tax their profits and transfer these proceeds to the landowners, can the elites gain even more? While a pertinent issue, one should note that gains to investment accrue not only to the investors, but much of the gains in fact go to the majority of the population in the form of increased wages. Thus, unless the government can completely tax these wage-gains from the general population and then transfer them to the landowners, the elites will be opposed to outside investment even in this case. In reality, the ability of governments (especially in developing countries) of such complete taxation of widely dispersed gains, is usually rather limited.

2.1. Equilibrium effects The main focus of our analysis here is the equilibrium quality of institutions. Note that institutions which uphold law and order and property rights protection here are welfare-improving for the majority of the population as it attracts investment into the province, thus raising wages. The question is whether the government has enough incentives in this framework to be induced into investing resources and implementing policies that promote such institutional quality. There are two countervailing forces at work. The electoral system provides one set of incentives. Good governance, if realized by the electorate, is rewarded by re-election for a second term. The question then is how does the electorate recognize a talented government? Note that in this framework, good governance in the form of good institutions is not directly observed by the majority, but indirectly through

April 20, 2009

14:17

WSPC - Proceedings Trim Size: 9in x 6in

On Effecting Institutional Change

isid07˙instnchange3

123

whether or not investment is attracted to the province as a result. When this inference process is sharper, the electoral system provides better incentives to the incumbent government to devote resources into improving institutional quality. On the other hand, if the government puts in resources into property rights protection, then it is likely that outside investors will be attracted, wages for labor will rise and the elite will lose its monopoly rents this period. Thus, the landowning elites’ lobby will seek to directly influence governmental decision through the offer of bribes in exchange for the government implementing their preferred outcome, namely that of a low level of property rights protection. There is a further effect at work in their calculus. If investment gets attracted this period, then realizing that a talented government is in power, the electorate is likely to retain such an incumbent. Having already attained experience in matters of good governance, the reelected government is then even more likely to maintain a good investment climate in the future thus further imperiling the elites’ rents. This second effect thus reinforces the elites’ desire to bribe the incumbent government’s policy as regards (low) institutional quality. The government, in making its decision of whether or not to put in effort into good governance weighs the potential benefits that the two groups have to offer. In considering an equilibrium of this game, the relevant question in our context is then when does the equilibrium involve (a) the government adopting and investing in a policy of good institutions thereby resulting in an improvement in welfare for the majority, versus (b) the government succumbing to the influence of the elite and maintaining the status-quo climate of poor institutions under which no investment is attracted and the elite retain their monopoly rents. The answer to this question depends on how the incentives provided by the electoral system compare with what the elite are willing to pay to influence the government. If the former dominates, then the equilibrium is of type (a) and the outcome of this interaction involves a triumph of the democratic system in which the result is good governance, free and secure economic transactions and an improvement in welfare. On the other hand, if the latter is the dominant factor, then even with democratic elections, an atmosphere of dysfunctional institutions, insecure property rights and poor law and order gets maintained, resulting in very little investment into the region and the elites’ monopoly rents are perpetuated. Although a democratically elected government exists and there are regular elections, the incentives provided by the threat of being voted out of power due to

April 20, 2009

14:17

124

WSPC - Proceedings Trim Size: 9in x 6in

isid07˙instnchange3

Sumon Majumdar and Sharun Mukand

non-performance is too weak to overcome the influence offered by the elite. As a result, governmental policy is effectively captured by the elite. The problem stems from the fact that the government’s actions are observed (and rewarded) only indirectly by the electorate. Attempts at good governance are recognized only through its effect on attracting investment into the province. When this channel is weak, incentives for good governance are muted and the elite find it relatively easy to influence the government to do their bidding. Democracy effectively gets captured. This is the situation highlighted by the 2003 Latinobarometro’s finding that 71% of the respondents in Latin America felt that democracy had been captured by special interests. What factors affect which equilibrium occurs? Majumdar and Mukand20 analyze a formal model along these lines and characterize the stationary equilibrium of the dynamic interaction between a short lived agent, namely the incumbent government, playing against a long-lived opponent, the infinitely-lived elite. Their analysis identifies the probability of investment occurring in the province in the presence of property rights protection (which they denote by qinv ) as an important factor in determining this equilibrium. The main proposition in the paper is summarized by figure 1. In the region where the bribe offered by the elite is higher than the electoral incentives, the government is effectively captured by the elite and puts in effort e = 0 at good governance; on the other hand when electoral incentives dominate, the government devotes resources e = 1 into upholding good institutional quality. From the figure, it can be seen that the equilibrium is characterized by e = 0 either when the probability of investment qinv is too small or too high. For intermediate values of qinv , the stationary equilibrium involves e = 1. Effect of Policy The question then arises that faced with such a region, what can be done to improve property rights, investment and overall welfare of the majority there? Consider the effects of developmental policy or an investment-promoting policy for this region. This could come about due to a direct subsidy on such investment or indirectly through bettering of the infrastructure facilities such as roads, communication and power in the province. When local infrastructure is better, less private capital is required to start an investment project and thus risks to investors are reduced. The effect of such policies is a rise in qinv , the probability that investment occurs when there is protection for property rights in the province. How does this impact the government’s incentives? Firstly, it rewards good

April 20, 2009

14:17

WSPC - Proceedings Trim Size: 9in x 6in

isid07˙instnchange3

On Effecting Institutional Change

Fig. 1.

125

The region where bribing is effective

governance. As qinv rises, the probability of investment in the presence of effective property rights increases; thus by making governmental effort on this front more visible, it rewards good governance and thereby increases the incumbent government’s incentive of putting in effort e = 1. This is the incentive effect, and serves to reduce the moral hazard problem inherent in the political set-up. At the same time however, the elites’ fears of their monopoly rents getting eroded also rises due to the increased possibility of investment occurring. Furthermore, by raising the chances of a government of high ability being re-elected, an increase in qinv serves to also raise the efficacy of the political system in re-electing able governments. However, high ability governments, once re-elected, are more difficult to influence during their second term in office owing to their experience factor at good governance. As qinv increases, this fear of the increased chances of re-election of a high ability uninfluenceable government causes the elite to seek to prevent the political game from proceeding to the second period, where it would be beyond the sphere of their influence. This is the political control effect, and serves to further raise the bribe the elite are willing to pay. Which of the two effects dominate? As the above figure shows, the incentive effect always dominates for low values of qinv . In other words, take a province in which the chances of attracting investors is low, say due to it having very high capital requirements. Any policy that lowers this cost

April 20, 2009

14:17

126

WSPC - Proceedings Trim Size: 9in x 6in

isid07˙instnchange3

Sumon Majumdar and Sharun Mukand

and/or increases the gains from investment i.e. by raising qinv , can serve to improve matters by changing the equilibrium from one with persistent elite influence and poor institutions to one where the dominant incentive for the government is electoral and it makes a concerted effort at good governance. When the infrastructure facilities are very poor to begin with i.e. when qinv is very low, the visibility of government policies towards protecting the rights of investors is extremely limited and this sharply limits the government’s incentive at expending effort towards such policies. In such situations, policies designed to attract outside investment raises qinv and by raising the government’s incentives, has a positive effect. On the other hand, for provinces with a moderately high level of qinv , any further rise in it can sometimes have an adverse effect on a previously well-functioning system. In this case, a decrease in the capital requirements for investment can cause a change from an equilibrium of good governance to one where governments are captured by the elite. While increases in qinv increases the incentives of governments to put in effort here too, at the same time it also raises the elites’ fear that high ability governments beyond their sphere of influence are more likely to get recognized and thus re-elected by the electorate. This raises more than proportionately the bribe that the elite are willing to pay to prevent the recognition of such governments; at such ranges, the political control effect dominates. Thus beyond a certain point, any policy that raises qinv can change the equilibrium from one where governments are uninfluenced and put in effort at good governance to one where the elites are willing to pay a high enough bribe to get the government to put in zero effort into property rights protection. Overall this corollary suggests a note of caution for the use of developmental policy as a tool to effect institutional change and the importance of underlying initial conditions in the process. In some cases, a policy of encouraging outside investment can have the unintended consequence of destabilizing previously well-functioning institutions. However, even in cases where an increase in qinv may not result in reducing directly the chance of government capture by improving electoral incentives enough, it does lead to an increase in the minimum amount of bribe that is required to influence the incumbent government. Thus the costs to the elite of controlling the government increase. Recall that it is the elites’ dependence on a labor intensive technology that leads to them fearing a rise in wages, and is the driving force in their desire to prevent outside investment into the province. Suppose there exists an alternative technology which requires less labor for its use; its adoption would clearly

April 20, 2009

14:17

WSPC - Proceedings Trim Size: 9in x 6in

On Effecting Institutional Change

isid07˙instnchange3

127

make the elite less sensitive to increases in the wage. However, changing to such a technology may involve substantial reorganization of the entire production process and is likely to be costly. Thus if the elite were assured that labor wages would remain low, they would be averse to incurring the expense of such reorganization. If however the costs of trying to control the government and ensuring low wages are high enough, would they be willing to incur the required reorganization cost to modernize their technology? In answering this question, it also allows a discussion of the various possibilities that may arise in the process of institutional change. 2.2. Modernization by the Elite? Suppose there exist alternative technologies that require less labor per plot of land to produce output. Moving to a new technology is however costly and may involve the cost of actual purchase of machinery etc. as well as the cost of reorganization of the entire production-process. Thus in the absence of any other incentives, the elites will be reluctant to shift from their current labor-intensive technology. Allowing for the possibility of modernization by the elite, let us now consider the impact of developmental policies to attract investment into this province. As discussed before, such policies could either be in the form of an improvement in infrastructure and other facilities that reduce the private capital requirements for investment or it could be in the form of a direct subsidy to the capital costs of investors. Before proceeding further, it is perhaps useful to recall that our presumption for this region is that in the absence of any form of policy intervention, political incentives are under the grip of the minority elite and consequently economic institutions are poor. We now delineate the effects that the introduction of various forms of policymaking may have on the region in question under different scenarios. Our focus is on the impact of alternative (and possibly complementary) forms of policy intervention on the region trapped with inefficient institutions rather than on the policymaker. Although in practice the reception of policies initiated by an outside policymaker are often affected by the policymaker’s perceived (self) interests, in the present paper we ignore such issues. Case I: Consider first the simplest case where the introduction of a policy to reduce investment costs pushes the minimum bribe required to successfully

April 20, 2009

14:17

128

WSPC - Proceedings Trim Size: 9in x 6in

isid07˙instnchange3

Sumon Majumdar and Sharun Mukand

influence the government beyond what the elite are willing to pay. Such a policy improves the incentives provided through the electoral system enough for the incumbent government to not succumb to influence by the elite. Governments now devote resources into improving institutions, property rights get protected, the political process remains free of elite capture, and income of the majority goes up. In this case, the elite realize that reelection is a powerful enough tool to influence the government into exerting effort ensuring a good institutional structure and there is a high probability that investment will get attracted and consequently wages will rise. Two cases arise. IA: Institutional Improvement with Democratic Consolidation. If the expected rise in future wages is big enough, the elite willingly incur the costs of reorganization and move to a labor-saving technology. Once they have thus reorganized, their incentives to influence the government’s institutional improvement efforts disappear. As a result, democracy succeeds in that governments perform with little interference on their good governance objectives, institutions to support economic activities improve, wages and overall welfare rises. Further more, such institutional improvement is consolidated in the sense that even if the external policy to reduce investment costs were to be (unexpectedly) withdrawn, institutions would continue to remain good as following their modernization, elites would no longer have the incentive to influence government policies in this regard. Thus, in this case, even short-term policy interventions can have long-term effects. Of course, the short-term policy of support to investment in the province should not be viewed as “short term” by the elites. If so, then their decision to modernize or not would depend on their expectations about how long they expect this policy to last. Thus this model points out in a simple manner the dangers of an explicit deadline for interventionist policy, an issue which has seen wide discussion in nation-building exercises (Dobbins21 ) IB: Institutional Improvement with Democratic Fragility. On the other hand, if the costs of reorganization are too high, the elite remain traditional, but the introduction of a policy to encourage investment pushes the government’s incentives out of the clutches of the elite. However, their costs of reorganization are too high relative to the benefits of a lower future wage-bill and the elite are not induced to modernize their technology.

April 20, 2009

14:17

WSPC - Proceedings Trim Size: 9in x 6in

On Effecting Institutional Change

isid07˙instnchange3

129

As a result, even though there is institutional improvement in the short-run, democracy is still fragile in the sense that if this policy were withdrawn, the situation would go back to the case where the electoral incentives are too weak and therefore the prevalence of elite capture and bad institutions. When confronted with a country that is close to this threshold, an external policymaker can bring about permanent institutional change at a relatively low cost perhaps by subsidizing the elites’ costs of reorganization and giving them easier access to modern technology. While such a policy does not directly improve infrastructure or other economic conditions in the region, it works indirectly by removing the impediments to good governance by elected governments. By getting the elite to modernize, it helps align (or rather lessens the divergence of) interests between the elite and the majority. The most ambitious (and successful) such experiment in recent times has been the incorporation of the nascent democracies of Portugal and Spain into the European Union. Developmental expenditure from the EU towards building infrastructure, privatizing an inefficient public sector and retraining the labor force with upwards of 120 billion dollars in aid was a crucial element of this strategy. More recently, consider the ongoing nation building experiment in Afghanistan. The traditional landowning elite obtains its revenue from opium production and smuggling. Not surprisingly, this group has little interest in improving institutions. Aware of this, much of recent developmental efforts have been aimed at giving these landowners and opium producers incentives to switch production to other crops and engage in other economic activity (Goodson22 ). Case II : Consider the alternate case where even under a policy of encouraging investment into the province, electoral incentives do not improve enough and democracy still remains potentially captured by the elite. In this case, while the elite ensure a low level of property rights in the province (thereby de facto keeping out investment) by using a bribe to influence governmental policy, any marginal improvement in the visibility of good governance efforts mean that the minimum bribe required to influence the government goes up. Thus the elites’ cost of controlling the government increases. Thereby the indirect political cost they incur of continuing with a labor-intensive technology goes up. What if they instead adopted a technology that was less dependent on labor? Again two cases arise depending on how big this increase in influence cost is.

April 20, 2009

14:17

130

WSPC - Proceedings Trim Size: 9in x 6in

isid07˙instnchange3

Sumon Majumdar and Sharun Mukand

IIA: Indirect Institutional Improvement with Democratic Consolidation. This is the case where the immediate electoral incentives for the government does not improve enough and the elites’ still maintain de facto control over the government’s policy process. However their cost of wielding control increases by enough so as to encourage the elite to reduce dependence on governmental protection and thus provides them with an indirect motivation to modernize. In this circumstance, the elite modernize their technology and adapt such that their profits are less a function of low wages. Therefore, they feel less threatened by an improvement in overall institutional quality and the higher wages that will then result. Thus in this case, changes, when they take place are dramatic and happen on many fronts. It involves modernization by the elite as well as improvements in the underlying institutions that encourage economic transactions, investment takes place, wages rise and thus welfare of the general population improves. IIB: Elite Entrenchment, Institutional Persistence and Democratic Failure. From a policymaker’s perspective, the most difficult scenario is when the region’s institutional equilibrium has weak political accountability and the elite find it relatively easy to retain political control. Furthermore, the elite are too far away from the technology frontier and are reluctant to modernize. This is the situation which is likely to see the most persistence in traditional inefficient institutions. Here, nothing changes either in terms of who holds actual control or in terms of economic outcomes. It is also the situation which is perhaps the most difficult to rectify. If the basic infrastructure in the province is very poor (thus creating extremely weak incentives for governments at good governance), developmental policy in the sense of marginal changes in investor attraction are unlikely to change the political incentive and in the underlying governance institutions. Similarly, marginal changes in the cost of adopting modern technology is unlikely to make a difference. In such a situation, it appears that forcible modernization of the elite or removing their source of monopoly rents is necessary for democracy to work. In practice, this would require the external policymaker to use some kind of coercive policy which results in a large scale redistribution of land and other assets. The necessity of such coercive policy is clear in many instances of nation building - from postwar Germany to Bosnia, Kosovo and East Timor (Dobbins et. al.21 ). Perhaps the classic instance where the use of coercive technology was necessary and successful is postwar Japan. In particular,

April 20, 2009

14:17

WSPC - Proceedings Trim Size: 9in x 6in

On Effecting Institutional Change

isid07˙instnchange3

131

the military defeat of Japan had diminished the ability of the political and economic elites to block institutional change (Kawagoe23). Accordingly General MacArthur (and policymakers at SCAP) attempted to take advantage of this temporary weakness of the traditional elite. In an array of policy measures, they attempted to restructure society so as to diminish the influence of the traditional sources of power. First, they attempted to breakup the hold of the traditional zaibatsu holding companies - “eighty three of the leading zaibatsu were broken up into their component parts and antimonopoly laws were passed to prevent their reestablishment”(Dobbins et. al.21 ). Further, labor was given the right to organize into unions, to bargain collectively and to strike. Contemporaneously, MacArthur helped push through the most sweeping land reform bill through the Japanese Diet and oversaw its implementation. Clearly to General MacArthur, establishment of a vibrant democracy meant tackling the economic and political roots of the traditional elite. 3. Conclusion In this paper, we examine the role of policy intervention in effecting institutional change. We identify two effects of developmental policies. First, by increasing political accountability, such policies may encourage nascent democratic governments to invest in good institutions – the incentive effect. However, we argue that such developmental policies may also increase the incentive of the rentier elite to tighten their grip on political institutions. In some cases, this latter political control effect may outweigh the former incentive effect, and result in an overall deterioration of institutional quality. However such policies also impact the elites’ incentives to modernize. In some cases, developmental policy increases the cost of political control to such an extent that the elite find it infeasible to maintain political control. They modernize and in essence bring their interests closer in line with those of the majority. In such cases, developmental policy can result in dramatic improvement in institutional quality and welfare. In other cases the policy maybe rather ineffective, and more coercive means of change maybe called for. Overall for policy intervention to be successful, it suggests good knowledge of local conditions and an appropriately tailored policy. Acknowledgements We thank seminar participants at Queen’s University, Carleton University, and at the 2007 ISI Delhi conference on Comparative Development for

April 20, 2009

14:17

132

WSPC - Proceedings Trim Size: 9in x 6in

isid07˙instnchange3

Sumon Majumdar and Sharun Mukand

useful comments. Majumdar’s research is supported in part by funding from the Social Sciences and Humanities Research Council of Canada. The usual caveat applies. References 1. R. E. Hall and C. I. Jones, Quarterly Journal of Economics 114, 83 (1999). 2. D. Rodrik, A. Subramaniam and F. Trebbi, Journal of Economic Growth 9, 131 (2004). 3. D. Acemoglu, S. Johnson and J. Robinson, Institutions as the fundamental cause of long-run economic growth, in Handbook of Economic Growth, eds. P. Aghion and S. Durlauf, (A) (North Holland, 2005) pp. 386–472. 4. D. Acemoglu and J. Robinson, Economic Origins of Dictatorship and Democracy (Cambridge University Press, 2006). 5. S. Majumdar and S. Mukand, The leader as catalyst: On leadership and the mechanics of institutional change QED Working Paper 1128, (2007). 6. T. Besley, Journal of Economic Perspectives 19, 43 (2005). 7. J. Fox, World Politics 46 (1994). 8. J. Sachs, The End of Poverty: Economic Possibilities of Our Time (Penguin, New York, 2005). 9. A. Lizzeri and N. Persico, Quarterly Journal of Economics 119, 705 (2004). 10. H. Llavador and R. Oxoby, Quarterly Journal of Economics 120, 1155 (2005). 11. D. North, Institutions, Institutional Change and Economic Performance (Cambridge University Press, 1990). 12. D. North, Structure and Change in Economic History (Norton, 1981). 13. A. Greif, Institutions and the Path to the Modern Economy: Lessons from Medieval Trade (Cambridge University Press, 2005). 14. G. Roland, Economics of Transition (MIT Press, 2001). 15. T. Persson and G. Tabellini, Political Economics: Explaining Economic Policy (MIT Press, 2000). Zeuthen Lecture Book Series. 16. B. B. de Mesquita, A. Smith, R. M. Siverson and J. D. Morrow, The Logic of Political Survival (MIT Press, 2003). 17. B. Jones and B. Olken, Quarterly Journal of Economics 120, 835 (2005). 18. S. Majumdar, A. Mani and S. Mukand, Journal of Development Economics 75, 137 (2004). 19. G. M. Grossman and E. Helpman, Special Interest Politics (MIT Press, 2001). 20. S. Majumdar and S. Mukand, On building institutions Mimeo., (2007). 21. J. Dobbins, The Beginner’s Guide to Nation-Building (RAND Corporation, 2007). 22. L. Goodson, Asian Survey 45 (2005). 23. T. Kawagoe, Deregulation and protectionism in japanese agriculture, in The Economic Development of Modern Japan, 1945 - 1995: From Occupation to the Bubble Economy, ed. S. Tolliday, Elger Reference Collection, Vol. 2 (Cheltenham, U.K., 2001) pp. 516–41.

ON THE POLITICAL ECONOMY OF GENERAL STRIKES ABHIRUP SARKAR Indian Statistical Institute Calcutta April, 2008 Email: [email protected] The paper seeks to explain why in LDCs like India political parties call general strikes which, unlike strikes in a factory, are often held as general protests without any specific economic goal to achieve. We argue that political parties call general strikes to signal their strength, which determines their probabilities of winning elections as perceived by the vulnerable. The vulnerable are crucially dependent on political favours for their survival, in turn, are more likely to join the party which can signal more strength and a higher probability of winning. We show that an increase in the relative size of the vulnerable increases the frequency of general strikes.

1. Introduction Economic theory of strikes is mainly concerned with situations of bargaining taking place between a labour union and a firm where the union, in order to squeeze more out of the bargaining process, resort to a strike. The question, lying at the heart of the phenomenon of strikes, is why do the parties get involved in a costly strike when they can agree to a mutually beneficial outcome in advance? The question was implicit in Hicks’s (1963) theoretical discussion on strikes and is called the Hicks Paradox. Subsequently several attempts, including Ashenfelter and Johnson (1969), Cross (1969), Kennan (1980), Fudenberg, Levine and Ruud (1983) and Myerson (1984), (see Kennan (1986) for a survey) have been made to explain the phenomenon of strikes and the Hicks Paradox. The early attempts tried to explain strikes by assuming that at least one party does not behave optimally. The more recent literature highlights incomplete information as the core factor explaining costly strikes. More specifically, one or both parties are assumed to be incompletely informed either about the size of the pie or about the utility function of the rival and a strike turns out to be a costly way of revealing information. In contrast with the existing literature, the present paper is concerned with a different kind of strikes. These strikes, often named general strikes, are called by political parties or by organizations affiliated to them. They involve total

133

134

Abhirup Sarkar

cessation of all normal activities, including work, in a city or a region usually for a day but sometimes even for longer periods. The immediate economic gains from calling these strikes are not obvious. They are called against the government or the ruling party on issues which may be varied and multiple. Thus strikes may entail protests against inflation, demand for increasing the rate of interest paid on government provident funds, disapproval of government inaction in dealing with riots or terrorist attacks and even registering one’s complain regarding national policies favouring globalization and free markets. Some strikes are called as instantaneous reactions to current political events while others are planned months ahead. Two things are noteworthy about these general strikes. First, more often than not the issues on which strikes are called become secondary and unimportant and sometimes it is completely forgotten as to why the strike was called. Second, sometimes different political parties call strikes on the same issue but on different dates. The two taken together would imply that these general strikes are organized not to achieve any specific economic gain, but to signal general political power. The occurrence of frequent general strikes is mostly observed in less developed regions with large informal sectors. In these sectors, property rights are not well defined, legal institutions are also quite weak. As a result public goods are often used for private purposes. Examples range from use of city pavements for selling one’s merchandise to illegal encroachment of public land for building shanties. For all this one needs political protection. Again, since the legal system is inefficient, time-consuming and expensive, disputes are often settled by the locally dominant political party the association with whom certainly helps to reach a smoother settlement. Even in the agricultural sector, political favour helps one to get small benefits like subsidized credit, seeds, irrigation water, housing and so on through the local government which is dominated by a political party. Thus the poor and the underprivileged, who mostly work in the informal sector including agriculture, seek political protection for their livelihood. On the other hand, there is a competition among political parties to provide protection for the vulnerable, because the more people a party can bring under its protective umbrella the higher is its political support. One possible way to attract potential protection seekers is by signaling political power. We argue in this paper that signaling of political power can be accomplished through successful general strikes. The main purpose of the paper is to show that an

On the Political Economy of General Strikes

135

increase in the relative size of the informal sector increases the possibility of a general strike. The model that we develop for the purpose of demonstrating this is an extension of Sarkar (2007). It is different from the standard political economy models discussed, for example, in Persson and Tabellini (2000) as far as freedom of choice with regard to voting is concerned. While in the standard approach each citizen if is allowed to vote freely according to his preferences, in the present model a section of voters, specifically those belonging to the informal sector, are constrained to vote for the party which gives them protection. In other words, the informal sector voters sell their votes in exchange of protection and hence make explicit economic calculations to determine their political preferences. In what follows, the economic and political environment is described in section 2. The nature of the equilibrium outcomes and the main result of the paper are discussed in section 3. Section 4 concludes the paper. 2. The Environment 2.1. The Economic Environment We consider a less developed economy with two sectors, one formal and one informal. In the formal sector, labour and capital produce output through a smooth, well-behaved neo-classical production function. Both the real wage rate and the stock of capital are given in the formal sector.a The given stock of capital fixes the exact position of the marginal productivity of labour schedule and given the marginal productivity of labour schedule, the fixed wage rate determines the level of employment. In the informal sector, output is produced by labour alone and in particular, output is assumed to be proportional to the labour employed. We assume that the economy is small and open. By suitable choice of units, prices of the formal and the informal sector outputs are taken to be unity. Let w and v be the wage rates in the formal and informal sectors respectively.b We assume that w > v. Therefore, a worker always prefers to work in the formal sector. Those who cannot find employment in the formal sector are residually employed in the informal sector. At any period, two a b

The real wage rate is fixed by unions and the capital stock is given by past accumulation. In each sector, the real wage rate is equal to the marginal productivity of labour. In the informal sector, marginal productivity is equal to average productivity as well because output is proportional to labour employed.

Abhirup Sarkar

136

generations, young and old, coexist in the economy. Population size of each generation is normalized to unity. We assume that the formal sector employs only young people. Let n be the number of young people employed in the formal sector. Then the remaining (1 – n) young people are employed in the informal sector. Of the old population, n workers worked in the formal sector when they were young. By assumption, they do not work when they are old.c The remaining (1- n), who had worked in the informal sector when they were young, continue doing so when they are old.d Thus to summarize, at any period, n young workers work in the formal sector, where n is determined by the given capital stock and the real wage rate in the formal sector. Again n old people, who are erstwhile formal workers, do not work at all. Finally, (1-n) young workers and (1-n) old workers work in the informal sector. 2.2. The Political Environment There are two political parties in the region denoted by A and B. At any period, the economy can be in one of the two states, X A , X B , where X i denotes the state where party i is in power. Let us define p ≡ prob[ X t +1 = X A | X t = X A ] 1 − p ≡ prob[ X t +1 = X B | X t = X A ]

(1)

as the respective conditional probabilities of party A and party B coming to power in the next period given that party A is in power in the current period. Similarly we define q = prob[ X t +1 = X B | X t = X B ] 1 − q = prob[ X t +1 = X A | X t = X B ]

(2)

as the respective conditional probabilities of party B and A coming to power in the next period given that party B is in power in the current period. We confine our analysis to cases where the transition probabilities remain unchanged over time. However, the equilibrium levels of p and q are to be determined endogenously. c d

They get a pension from their employers and might also depend on their past savings. There is no pension in the informal sector. Neither is the informal income high enough to allow adequate savings which can sustain a worker in his old age. So he has no choice but to work when old.

On the Political Economy of General Strikes

137

We identify the informal sector as the political sector and the formal sector as the non-political sector. In the formal sector, production is carried on without any political patronage or support. More importantly, property rights are well defined in the formal sector. In the informal sector, on the other hand, political patronage is essential for production. There are, in fact, two common features of the people earning their bread in the informal sector. First and foremost, all of them are extremely vulnerable. They lack the security of formal sector jobs. Second, they do not always live by formal laws and norms. Some of them might be living on illegally encroached government or railway land. Others might illegally occupy pavements of city streets to sell their ware. They become more vulnerable because they do not have any well-defined property rights within the formal legal framework. These people depend in a fundamental way on political parties for their livelihood. It is their vulnerability which compels them to do so. A political party gives them protection and in return gets their support at the time of elections. We assume that an individual working in the informal sector has to get patronage from one of the two political parties. In exchange he castes his vote in favour of his patron at the time of the elections. In other words, an informal sector worker sells his vote, apart from his labour, for his survival. 2.3. The Sequence of Events The sequence of events taking place in the economy is as follows. Given the stock of capital in the formal sector, n young workers are employed there. The residual labour force 1 – n is absorbed by the informal sector. Then in each sector output is produced. The relative price between these outputs is given by the rest of the world and we choose units in such a way that this relative price is unity. Also, choosing output as the numeraire the absolute price is also taken to be equal to one. After output is produced in the formal sector, the young employed in that sector consume and save, the old who have retired from the formal sector consume their previous period's savings. In the informal sector the young and the old are engaged in production and each individual gets an income v. Moreover, the old in the informal sector get a rent in addition to their wages if the party they had supported when they were young is in power. This is basically a reward for their political support. It may be obtained through securing an important position in the government or the party which helps one to extract rents from the formal sector or in terms of getting exclusive favours from the government, e.g. habitable land from the government at subsidized prices. There is a fixed amount of favour to be

Abhirup Sarkar

138

distributed which, we assume, is equally divided between old party loyalists in the informal sector. After they receive their income the young and the old in the informal sector consume. We further assume that the old in both sectors die after they consume and before they vote. This means that only the young are allowed to vote in our model. This assumption saves us from a lot of unnecessary algebraic complications. Finally, elections are held and a party is chosen for the next period. 3. Equilibrium 3.1. Determination of Equilibrium Total employment in the formal sector is denoted by n and that of young individuals in the informal sector is denoted by (1-n). Out of the (1-n) young people employed in the informal sector, suppose a fraction α joins party A and the remaining (1- α ) joins party B when party A is in power. Similarly, let a fraction β of young informal workers join party B and the remaining (1- β ) join party A when party B is in power. Suppose party A is in power and assume 0 < α < 1. The equilibrium distribution of young informal workers is determined in such a way that on the margin a worker is indifferent between joining party A and party B. This would mean that v + p[v +

R

α (1 − n)

] + (1 − p )v = v + (1 − p )[v +

R ] + pv (1 − α )(1 − n)

(3)

Equation (3) may be interpreted in the following way. The left-hand side represents expected income from joining party A over the two period life span of an informal worker. In the current period he earns v with certainty. In the future period when he gets old, he earns v plus a rent if his party is in power. The probability of this is p. We have assumed that the total amount of rent to be distributed among old party loyalists is constant and each gets an equal share. In the event his party is not in power, the probability of which is (1-p), the worker earns just v when old. The right hand side represents the payoff of a young informal worker if he joins party B when party A is in power. Equilibrium allocation of informal workers between the two parties requires that the two sides of equation (5) are equal so that no worker has an incentive to change political loyalty. To keep the analysis tractable we assume that the total rent R is the same under party A as under party B. After some simplification, equation (5) reduces to

On the Political Economy of General Strikes

139

p= α

(4)

Considering the distribution of informal sector workers between the two parties when party B is in power, we get, exactly in a similar way q=β

(5)

Equations (4) and (5) have the intuitive explanation that the number of people willing to trade their votes in exchange of political favours from the presently ruling party would depend directly on the probability of its reelection. Indeed, given our simplifying assumptions, the two are equal. The model is closed by bringing in the voting process. The voting process expresses probability p as a function of α . This function, along with equation (4) determines the equilibrium value of p. Determination of q is similar. We shall assume that there are two types of voters. First, there are voters in the formal sector that are free to vote for any party. Second, there are opportunistic voters in the informal sector who trade their votes for a particular party in exchange of political support which is essential for their survival. We further assume that the formal sector voters are random voters so that π A , the vote share of party A in the formal sector, is a random variable with a distribution function F (.) . We have, from the definition of p p ≡ Prob [party A wins | party A is in power] = Prob [π A n + α (1 − n) > (1 − π A )n + (1 − α )(1 − n)]

(6)

Equation (6) may be further simplified as p = 1 − Prob [π A ≤

1 1 − 2α ] + 2 2n

(7)

so that we may write

p = 1 − F ( x), where x ≡ α + ≡ p (α )

1 − 2α 2n

It is straightforward to verify that ∂p 1 = − f ( x )(1 − ) > 0 ∂α n

(8)

140

Abhirup Sarkar

∂2p 1 = − f ′( x )(1 − ) ∂α 2 n

where f (x) is the probability density function. We assume that f (x) is single peaked and is symmetric around 1 2 . Hence f ′(x) is greater than, equal to or less than zero according as x is less than, equal to or greater than 1 2 . This allows us to draw the upward rising function p(α ) as convex up to α = 1 2 and concave thereafter, the point α = 1 2 being a point of inflection. In equilibrium, the perceived probability of winning, as obtained from equation (4), must be equal to the actual probability of winning, i.e. p(α ) = α

(9)

A few comments on equilibrium are now in order. First, since f (x) is symmetric around 1/2, at α = 1 2 we have p(α ) = α which guarantees that an equilibrium always exists. Second, depending on the shape of the p(α ) curve, there could be multiple equilibrium (as shown in figure 1A and 1B) or a unique equilibrium (as shown in figure 2). Third, if n > 1 2 , i.e. the proportion of random and independent voters, all belonging to the formal sector, is greater than the proportion of deterministic and politically dependent voters, all belonging to the informal sector, then the outcome of the voting process will be unambiguously random. To see why, note that if n > 1 2 , then p (0) > 0 and p(1) < 1 which guarantees that in equilibrium 0 < p < 1. If, on the other hand, n < 1 2 , then p (0) = 0 and p(1) = 1 so that there are three equilibria at p = 0, p = 1 2, p = 1. It is easily seen that similar equilibria exist when party B is in power. Indeed, the two parties are symmetric in every respect. In what follows, we shall confine ourselves to the case where there are multiple equilibria. We have already pointed out that for n < 1 2 , there are three equilibria as shown in figure 1B. When n > 1 2 , from figure 1A it is clear that we have multiple equilibrium if and only if at α = 1 2 , the slope p ′(α ) > 1 . This requires that n 1 f( )> 2 1− n

(10)

On the Political Economy of General Strikes

141

It can be easily verified that when n < 1 2 , (10) is always satisfiede. In what follows, we shall assume that condition (12) is satisfied, so that the equilibrium is not unique. Given that the p (α ) function has a single point of inflection, there are three equilibria as shown in figures IA and IB. The question arises as to which equilibrium, out of the three possible equilibria, is to be chosen and how? 3.2. General Strikes and Equilibrium Selection

In equilibrium, expectations are fulfilled. We call p *(α *) an equilibrium when a perceived probability p* by the informal sector workers induces a fraction α * of them to join party A which, in turn, leads to an actual f probability of winning p*. With multiple equilibriums, therefore, it is the perceived probability which becomes crucial in equilibrium selection. In the present context, we focus on general strikes as a device of signaling political power which becomes crucial in the formation of the perceived probability of winning and hence equilibrium selection. Suppose at the beginning of each period, a political event e takes place. The event e is a random variable distributed over some support [emin , emax]. The opposition party observes the event e and decides whether to call a general strike on the basis of the event. The general strike, if called, is successful with probability θ (e) with θ ′(e) > 0 and a failure with probability 1 − θ (e) . Thus a higher value of e refers to a more favourable political event from the point of view of the opposition party. The outcome of the general strike determines the probability of winning for the party in power in the next election as perceived by informal sector workers. Suppose party A is in power. We assume that if a successful general strike takes place, then the perceived probability of winning is greater than 1/2 and if the general strike is a failure the perceived probability is less than half. Finally, if the opposition party chooses not to call a strike, the perceived probability of winning for the incumbent is greater than 1/2, that is, inaction does not signal political power for the opposition.

e

f

In this case, the right hand side of (10) is less than unity so that violation of condition (10) implies that f (1/2) < 1. This is not possible because f (x) reaches a maximum at x = ½ and the area under f (x) is unity. We are looking at long run equilibria where the actual probability of winning is interpreted as the proportion of cases where party A has won when α = α*.

142

Abhirup Sarkar

Now, by assumption, each informal sector worker knows the structure of the model. Hence, a successful strike fixes the equilibrium at E1 and an unsuccessful strike or a no strike at E3 (in figure IA or IB). From our assumptions it follows that equilibrium does not take place at E2. Let B be the benefit of winning an election and let C be the cost of calling a general strike for either party. The party in opposition will call a strike if the benefit from such an action is at least as much as the cost, i.e.

θ (e)(1 − p l ) B + (1 − θ (e))(1 − p h ) B − C ≥ (1 − p h ) B

(11)

where p l and p h are the probabilities associated with equilibrium at E1 and E3 respectively with p l < p h . Condition (13) may be simplified to

θ (e)( p h − p l ) B ≥ C

(12)

Let e satisfy the above condition with strict inequality, i.e.

θ (e )( p h − p l ) B = C

(13)

Then ∀e ≥ e , a general strike is called. We now come to the main proposition of the paper. Suppose n < 1/2, so that we have 0 < p l < p h < 1 . Now let us consider a rise in the relative size of the informal sector, i.e. a fall in n. It is straightforward to show that ∂p (α , n) 1 ≤ 0 if α ≥ 2 ∂n

and

∂p (α , n) 1 ≥ 0 if α ≤ 2 ∂n

1 with a fall in n. This is 2 shown in figure 3. From figure 3 it is clear that a fall in n shifts equilibrium E1 to E1′ and equilibrium E 3 to E ′3 . As a result, p l goes down and p h goes up,

This means that the p(α ) curve rotates around α =

leading to a fall in ( p h − p l ) . Therefore, from equation (13) e goes down and strikes become more likely. The following proposition summarizes our finding:

On the Political Economy of General Strikes

143

Proposition. A rise in the relative size of the informal sector increases the expected benefits from calling a strike and therefore makes general strikes more likely.

It is not difficult to understand the intuition behind the result. A successful general strike sends a positive signal to informal sector workers. An increase in the relative size of the informal sector makes this signal more valuable and hence increases the expected benefit of calling a strike which, in turn, makes general strikes more likely. 4. Conclusion

The paper tries to explain the phenomenon of general strikes which is different from strikes arising out of industrial disputes between workers and their employers. While the existing literature contains extensive analysis of the latter type of strikes, the former type, which are often observed in less developed regions of the world, needs to be understood. Especially since the general strikes are often called without any possibility of immediate and obvious economic gains to the party calling the strike. We argue in this paper that general strikes are called to signal political power. In less developed regions there are large informal sectors where property rights are not well defined. Political protection and patronage are necessary in these sectors for earning a living. A political party, which can signal power by calling a successful strike, is likely to get more followers who would seek its protection and patronage. A larger following, in turn, increases the probability of winning elections. The main conclusion of the paper is that a larger informal sector increases the probability of general strikes. Now, if we make the reasonable supposition that general strikes deteriorate the industrial climate and increase the relative size of the informal sector, we are led to infer that underdevelopment, characterized by the existence o large informal sectors, has a tendency to persist ⎯ large informal sectors leading to more frequent strikes and more frequent strikes leading to large informal sectors. References

1. A. Orley and G. Johnson, American Economic Review. 59, 35 (1969). 2. C. John, The economics of bargaining, New York: Basic Books (1969). 3. F. Drew, L. David and R. Paul, Strike activity and wage settlements, NBER Conference on the Economics of Trade Unions (1983). 4. H. John, The theory of wages, 2nd Edition, London: Macmillan (1963).

144

Abhirup Sarkar

5. K. John, Journal of Labour Research, 1,77 (1980) 6. K. John, Handbook of Labour Economics, Vol II, Elsevier Science Publishers (1986). 7. M. Roger, Econometrica, 52(2) 461 (1984). 8. Torsten and G. Tabellini, Political Economics, Cambridge, Massachusetts, MIT Press (2000). 9. Abhirup, On the political economy of a backward region, Unpublished Manuscript (2007).

April 20, 2009

14:44

WSPC - Proceedings Trim Size: 9in x 6in

JessicaLal

PUBLIC OUTRAGE AND CRIMINAL JUSTICE: LESSONS FROM THE JESSICA LAL CASE BRENDAN O’FLAHERTY Department of Economics, Columbia University Email: [email protected] RAJIV SETHI Department of Economics, Barnard College, Columbia University Email: [email protected] Witness tampering and public outrage have combined to affect judicial outcomes in a series of high-profile criminal cases in India. We study how these phenomena operate together in a country with extremes of wealth and poverty, but with functioning judicial and political systems. Bribes and threats are intricately linked in the strategic interaction between offenders and witnesses. Not only do bribes provide a direct incentive that can suppress testimony, they also signal a greater likelihood of retaliation and hence serve as implicit threats to witnesses. The possibility of public outrage turns out to be an effective constraint on witness tampering. In many situations, greater media effectiveness can improve the administration of justice, even when more obvious improvements in judicial effectiveness cannot.

1. Introduction During the early hours of April 30, 1999, a thirty-four year old model named Jessica Lal1 was shot and killed at a private party in a South Delhi restaurant, allegedly for refusing to serve liquor to a guest after the bar had closed. The man in question was identified by three eye witnesses as Manu Sharma, the son of a senior Congress Party politician and former Union minister. After several days in hiding, Sharma surrendered to authorities in Chandigarh. In an interview with police that was subsequently broadcast on national television, Sharma confessed to the murder. This confession was later retracted, and a plea of non-guilty entered at trial. During the trial the three critical eye witnesses recanted earlier statements made to the police, and twenty-nine witnesses of lesser importance did the same. One of the eye witnesses, Shyan Munshi, changed his testimony so completely that his revised statement was used as exculpatory evidence by the defence.

145

April 20, 2009

14:44

146

WSPC - Proceedings Trim Size: 9in x 6in

JessicaLal

Brendan O’Flaherty and Rajiv Sethi

Sharma and eight other codefendants were acquitted of all charges in February 2006, resulting in a public outcry against what was perceived to be a gross miscarriage of justice. During the weeks that followed the verdict, petitions were circulated, protest marches organized, and candlelight vigils held. Prime Minister Mammohan Singh publicly expressed concern at the general phenomenon of witnesses changing their testimony, in an oblique reference to the Jessica Lal case. President Abdul Kalam received a petition of 200,000 names collected by journalists at NDTV, and promised action. The decision was appealed to the Delhi High Court by the prosecution in March, and in December the lower court ruling was reversed with respect to three of the defendants. Manu Sharma was convicted of murder and sentenced to life in prison, and two other defendants were convicted for conspiracy and destruction of evidence. (The verdict has been appealed to the Supreme Court, but Sharma remains in prison.) Shyan Munshi and other prosecution witnesses who turned hostile during the trial currently face possible charges of perjury.a The Jessica Lal case is one of several recent high-profile criminal cases that illustrate in dramatic fashion the manner in which witness tampering and public outrage can interact to determine judicial outcomes in India (four others are discussed below). In countries with high levels of inequality but with legal systems that function well enough that affluent individuals fear what poor people will say about them in court, there are incentives and opportunities for powerful criminal defendants to bribe or threaten witnesses. There are costs to witnesses from accepting bribes, but also significant risks associated with testifying against powerful interests in countries with limited witness protection programs. The relative costs of different actions depend, in expectation, on the extent to which public outrage over miscarriages of justice can be channeled through the press. Freedom of the press can therefore be an important constraint on the miscarriage of justice in developing countries. The mechanism through which public sentiment affects the functioning of the criminal justice system is an example of what Albert Hirschman (1970)3 has called voice. Hirschman argued that voice could serve as an alternative to competition in enhancing efficiency within organizations. Since

a For a collection of contemporaneous news reports from the earliest stages of the case to the present day, see http://www.rediff.com/news/jessica.html. The Delhi High Court decision is available online as Criminal Appeal No. 193 of 2006 at http://delhihighcourt.nic.in/.2

April 20, 2009

14:44

WSPC - Proceedings Trim Size: 9in x 6in

Public Outrage and Criminal Justice

JessicaLal

147

judicial systems, unlike firms, are not subject to the pressures exerted by competitive markets, voice takes on even more importance in determining their level of functionality. This paper explores the manner in which voice interacts with more familiar incentives in determining the incidence of witness tampering. We begin with the observation that the offer of a bribe can affect witness incentives in two quite distinct ways. There is a direct monetary inducement to remain silent (or give false testimony), and there is a more subtle effect on witness beliefs regarding the likelihood of retaliation. If the offer of a bribe is a signal that the offender is more likely to retaliate, bribes can be far more effective in reducing testimony than the monetary inducements alone would predict. We show how such effects can arise in equilibrium, and how the possibility of public outrage in response to a miscarriage of justice affects beliefs and behavior. Our model is characterized by unobserved heterogeneity among both witnesses and offenders. Witnesses differ in the satisfaction they experience from testifying, and offenders differ in the costs they would incur to harm witnesses who testify, and the cost of giving bribes. Some offenders have powerful revenge motives and gain from attacking witnesses who have testified against them, even after the fact, and even when those witnesses have accepted no bribes. This motive is amplified if a witness testifies even after accepting a bribe. We identify the existence of equilibria in which offenders with strong revenge motives are disproportionately more likely to offer bribes. Bribes therefore serve as a (noisy) signal of the viciousness of the offender: on average, offenders who offer bribes are more likely to attack witnesses than those who do not. In addition, the acceptance of a bribe changes offender payoffs in such a manner as to make retaliation more likely. Both of these effects result in an increased subjective evaluation on the part of witnesses of the likelihood that testimony will be met with retaliation. Bribes are not simply bribes; they are also veiled threats. Greater judicial effectiveness and greater media effectiveness have very different equilibrium effects in this model. If the likelihood and severity of punishment faced by offenders (conditional on truthful witness testimony) is increased, more offenders offer bribes, and these are perceived by witnesses to be less threatening on average. The rate of testimony conditional on a bribe offer therefore increases, although overall levels of testimony may decline since attempted bribery is more common. In contrast, an increased likelihood of public outrage (for instance due to greater media effective-

April 20, 2009

14:44

148

WSPC - Proceedings Trim Size: 9in x 6in

JessicaLal

Brendan O’Flaherty and Rajiv Sethi

ness) results in fewer offers of bribes but a greater perception on the part of witnesses that testimony will be met with retaliation. Bribes become less frequent but more threatening. The likelihood of testimony conditional on a bribe offer may decline, but the overall rate of testimony could nevertheless rise since in incidence of attempted bribery declines. Interestingly, greater media effectiveness increases the rate of testimony whenever judicial effectiveness fails to do so. Finally, more witness tampering occurs when wealth differences between offenders and witnesses are great, and favor offenders. Lowering inequality can therefore reduce the incidence of witness tampering. This paper is a theoretical contribution to a growing body of literature that finds real effects flowing from the organization of communications media. For instance Besley and Burgess (2002)4 show how newspaper circulation affects the distribution of food aid and calamity relief in India, Gentzkow, Glaeser, and Goldin (2006)5 relate the reduction in corruption in US politics between 1870 and 1920 to changes in the newspaper business, and Djankov et al. (2003)6 relate state media ownership to various measures of poor government performance. Our result that greater press freedom can be effective in securing convictions even when greater judicial effectiveness cannot do so may perhaps be viewed as a theoretical corroboration of the empirical finding of McMillan and Zoido (2004)7 that Fujimori’s government in Peru was willing to pay vastly more in bribes to influence a television station than to influence a judge. Note, however, that notional freedom of the press need not imply substantive freedom, and effective control of the media by governments can arise endogenously without formal censorship; see Besley and Prat (2006)4 for theoretical insights into this process. Before proceeding to the formal model and analysis, we survey four additional cases involving prominent defendants, witness reversals, significant media attention, and public outrage. 2. Other Cases 2.1. The Priyadarshini Mattoo Case Priyadarshini Mattoo was a twenty-five year old law student when she was raped, brutally beaten and strangled to death at her New Delhi residence on January 23, 1996. Although there was no eye witness to the murder, physical and circumstantial evidence pointed immediately to Santosh Kumar Singh, the son of a senior police officer then posted in Pondicherry. Singh had been

April 20, 2009

14:44

WSPC - Proceedings Trim Size: 9in x 6in

Public Outrage and Criminal Justice

JessicaLal

149

stalking Mattoo for over a year at the time of the murder, with multiple instances of harassment having been reported to the police. He was seen outside her residence by a neighbor immediately prior to the attack, and blood stained pieces of his motorcycle helmet vizor were found beside the body. DNA tests confirmed the presence of his semen on her clothes and her blood on his helmet. What seemed like an open and shut case, however, ended in an acquittal in 1999. Defence claims that the physical evidence had been tampered with while in police custody were given enough credence by trial judge to allow for reasonable doubt. Suspicions were raised by the judge regarding deliberate police misconduct, including false depositions, traced to the influence of the father of the accused: by the time of the trial Singh’s father was among the most senior police officers in Delhi. The acquittal triggered massive public outrage, and the case was appealed to the Delhi High Court in 2000. Little action was taken for several years, until the 2006 acquittal of Manu Sharma for the Jessica Lal murder led to renewed scrutiny and a sense of urgency in the part of the Court. The verdict of the lower court was reversed in October 2006, with the High Court finding that the “circumstantial evidence in the case is absolutely inconsistent... with the innocence of the respondent.” Justices RS Sodhi and PK Bhasin observed that the acquittal by the trial judge had “shocked the judicial conscience” of their Court. Singh was sentenced a few days later to death by hanging.

2.2. The Nitish Katara Case While studying at the Institute of Management Technology in Ghaziabad in 1998, Nitish Katara became romantically involved with a classmate, Bharti Yadav. Bharti was the daughter of D.P. Yadav, a major force in Uttar Pradesh politics, and the sister of Vikas Yadav, who was subsequently convicted for destruction of evidence in the Jessica Lal case. Her family disapproved of the relationship and Katara received multiple threats over the course of their relationship. On the night of February 2, 2002, Katara attended a wedding at which several members of Bharti’s family were present. Four witnesses observed him leaving in the company of three men, including Vikas Yadav and a cousin, Vishal Yadav. Katara’s remains, charred and battered beyond recognition, were found on a roadside the next morning. Vikas Yadav and a cousin, Vishal Yadav, went into hiding but were arrested a few days later. A

April 20, 2009

14:44

150

WSPC - Proceedings Trim Size: 9in x 6in

JessicaLal

Brendan O’Flaherty and Rajiv Sethi

detailed confession, admitting to the abduction and murder, was recorded by UP police and aired on national television. Vikas admitted to having killed Nitish with a hammer blow to the head and setting his lifeless body on fire. He subsequently led police to the spot where the body had been dumped and the murder weapon concealed. Once the trail began, however, one witness after another “turned hostile”, including all four witnesses who had earlier reported having seen Katara depart with Vikas and Vishal Yadav. In testimony before the court, Bharti Yadav denied a romantic relationship with Katara, admitting only to a vague friendship. There remains a single witness, a passer-by whose scooter broke down on the road taken by the accused, and who has testified to seeing Katara in the vehicle. This witness has reported having received threats against his life, and is currently under police protection. The case remains unresolved, and under intense public scrutiny.

2.3. The BMW Hit-and-Run Case At around 4am on January 10, 1999, a speeding black BMW crashed through a police checkpoint in Delhi, killing four people on the spot. Two others subsequently succumbed to injuries, leaving just one survivor, Manoj Malik. Three of the dead were police constables. A passer by who witnessed the crash, Sunil Kulkarni, came forward a few days later. According to his initial report to police, three individuals stepped out of the car, briefly inspected the damage, then fled from the scene. The damage was so extreme that the speed of the vehicle upon impact was estimated to be 140 kmph (about 90 mph). The car was alleged to have been driven by Sanjeev Nanda, son of businessman and arms dealer Suresh Nanda, and grandson of Navy Admiral S.M. Nanda. Also present in the vehicle were his friends Siddharth Gupta and Manik Kapoor. They were returning to Delhi from a party in Gurgaon, and Sanjeev was found to have elevated levels of alcohol in his blood several hours after the incident. The BMW was traced by police following oil leaks from the scene of the crash to the Gupta residence, where it was determined to have been cleaned of blood and human remains. All three were charged with culpable homicide and destruction of evidence. The trial is still in progress, but there have already been extraordinary changes in witness testimony. Claiming that his initial statement was made under police pressure, Kulkarni testified in October 1999 that it had been a truck rather than a BMW that had ploughed through the police check-

April 20, 2009

14:44

WSPC - Proceedings Trim Size: 9in x 6in

Public Outrage and Criminal Justice

JessicaLal

151

point.b The prosecution dropped him as a witness, considering him to be unreliable, but he was subsequently reinstated. In May 2007, Kulkarni identified Nanda as one of the three occupants of the vehicle, but then retracted this testimony two months later, saying that he had been unable to see any of them clearly.c Nanda now maintains that he was not in the vehicle at the time and had nothing to do with the accident. A sting operation by NDTV turned up evidence of attempted bribery, and it now appears that the witness spent eighteen months residing at a farmhouse owned by the lead lawyer for the defence. Nanda is free on bail while the proceedings continue. 2.4. The Best Bakery Case The Best Bakery was a Muslim owned and operated business in Vadodara, Gujarat that was set on fire by a Hindu mob during communal riots on March 1, 2002.10 Fourteen people were killed in the attack, including the owner Habibullah Shaikh and eight other members of his family. Among the witnesses was Zahira Shaikh, the owner’s eighteen year old daughter, who identified several individuals in the mob in a March 2 statement to police. The accused were also identified by Zahira in a statement before the National Human Rights Commission (NHRC) of India three weeks later. Twenty-one individuals were charged in the murder. During the trial in May 2003, Zahira and other surviving members of her family retracted earlier statements made to the police, resulting in the acquittal of all the accused on June 27. Shortly thereafter, on July 11, Zahira claimed in a sworn statement before the NHRC that she and other family members had been threatened and forced to retract their statement. In doing this, she was assisted by Teesta Setalvad, secretary of the NGO Citizens for Justice and Peace. The NHRC petitioned the Supreme Court of India to order a retrial in a state other than Gujarat, and the Court did so on April 12, 2004. The retrial began in October 2004, before a Special Court of Sessions in Mumbai. At a dramatic November 3 press conference one day before she was due in court, Zahira changed her testimony yet again, claiming that she had been abducted by Setalvad and her organization, threatened and confined for months, and forced to file false statements against the accused. A sting b “BMW

witness takes U-turn; says it was truck”, Indian Express, October 2, 1999.8 Witness Identifies Nanda in BMW Case”, The Hindu, May 18, 2007;“BMW Case: Another U-Turn by Kulkarni”, The Times of India, July 20, 2007.9

c “Key

April 20, 2009

14:44

152

WSPC - Proceedings Trim Size: 9in x 6in

JessicaLal

Brendan O’Flaherty and Rajiv Sethi

operation by Tehelka magazine in December revealed that a substantial bribe of 1.8 million rupees (about $40,000) was paid to Zahira and her family in order to retract her testimony against the accused. On February 24 2006, nine of the original accused were convicted and eight acquitted by the Court. In separate proceedings, Zahira was convicted of perjury and contempt of court. 2.5. Overview Two common themes run through these cases: dramatic changes in witness testimony in response to threats or bribes, and significant media and public involvement. In the Lal, Mattoo, and Best Bakery cases, public action played a major role in the reversal of earlier verdicts. In the Best Bakery case, such action was mediated through an NGO, Citizens for Justice and Peace. Media outlets such as Tehelka magazine and NDTV aired confessions which clearly galvanized public opinion.d We explore these common themes more formally below, using a model of witness intimidation and public outrage which highlights strategic aspects of the interaction between offenders and witnesses. 3. A Model of Witness Tampering Several of the cases discussed above involve not only charges of witness intimidation but also allegations of bribery. Why might threats to witnesses be supplemented with bribes? The most obvious explanation is that they offer witnesses an additional direct inducement to turn hostile. But there is also a more subtle reason: bribes may affect witness beliefs regarding the consequences of ignoring threats, and hence make testimony more risky. If bribes can lower the rate of witness testimony to a greater extent than threats alone, they may be worthwhile for offenders despite the greater costs entailed. We explore these interactions in a model that allows for bribery, retaliation, and the possibility that public outrage can lead to the reversal of acquittals and the punishment of witnesses who accept bribes.e d Even

some of the main protagonists made an appearance in multiple cases: Delhi High Court Justices RS Sodhi and PK Bhasin handled the cases of both Lal and Mattoo, and Vikas Yadav was among the accused in the cases of Lal and Katara. e The model developed here builds on our earlier work on witness intimidation (O’Flaherty and Sethi, 2007),11 which was motivated by interactions between offenders and witnesses in the United States. In that work we considered explicit threats but not bribes, and did not take into account the possibility that public outrage could result

April 20, 2009

14:44

WSPC - Proceedings Trim Size: 9in x 6in

Public Outrage and Criminal Justice

JessicaLal

153

3.1. Preliminaries Consider a crime involving one offender and one witness. The testimony of the witness is crucial to the case against the offender. If the witness does not testify, the offender will be acquitted. If the witness testifies truthfully, the offender suffers expected punishment ρ > 0. The magnitude ρ reflects both the probability that the offender will be convicted conditional on the witnesses testimony, and the severity of the punishment that will ensue upon conviction. Thus ρ could increase either because of changes in the way prosecutors and judges operate that increase the probability of conviction contingent on witness testimony, or because of harsher sentences. We will call changes in ρ changes in judicial effectiveness. In order to deter the witness from testifying (or more generally to induce false testimony) the offender may offer the witness a bribe, which the witness may either reject or accept. Acceptance of the bribe has value α > 0 to the witness. Following this decision, the witness may either testify or decide against doing so. If the witness testifies, the criminal may retaliate by harming the witness. In this case the witness suffers damages δ, and the offender incurs a cost which is assumed to be private information and, for reasons discussed below, contingent on whether or not a bribe was accepted prior to the testimony. An offender who does not attempt to bribe the witness, or whose bribe is rejected, pays a cost γ to harm a witness who testifies. This cost is drawn from a continuous distribution F (γ) with support [γmin , ∞), where γmin < 0. Hence there are some offenders who take pleasure in harming witnesses who have testified against them, even if the witnesses have rejected a bribe or were never offered one. We refer to γ as the offender type. The cost of harming a witness who has accepted a bribe is dependent on the offender type, and denoted c(γ), assumed to be continuous and strictly increasing with c(γ) ≤ γ, with strict inequality holding for all γ > γmin . In other words, the cost of harming a witness who testifies after accepting a bribe is lower than the cost of harming a witness who has not accepted a bribe. This is motivated by the idea that testimony following the acceptance of a bribe is a form of betrayal, or a violation of an implicit contract, which further inflames the revenge motive for harming witnesses who testify. in the reversal of an acquittal. Akerlof and Yellen (1994)12 have also explored the manner in which a fear of reprisals affects community cooperation with law enforcement and the equilibrium level of crime, but without considering the strategic effects of bribery or voice.

April 20, 2009

14:44

154

WSPC - Proceedings Trim Size: 9in x 6in

JessicaLal

Brendan O’Flaherty and Rajiv Sethi

Since c(0) < 0, there is some subset of offenders who are prepared to harm witnesses who testify after having accepted a bribe, even if such offenders would not harm witnesses who either reject the bribe or were not offered one. Let γ¯ > 0 denote the offender type for which c(¯ γ ) = 0. Then the proportion of offenders who are prepared to harm witnesses who testify after accepting a bribe is simply F (¯ γ ); we refer to these offenders as violent. Since γ is private information, a witness cannot be certain that an offer of a bribe comes from a violent offender. The cost of bribing a witness is also dependent on the offender type, and given by a non-negative, continuous, and strictly increasing function b(γ).Hence bribing a witness is cheapest for those offenders for whom harming the witness is also cheap. Witnesses are also heterogeneous, and vary in the payoff they obtain from testifying against offenders. A witness who testifies receives a (privately observed) payoff β > 0, drawn from a continuous distribution G(β) also having support [0, ∞). We refer to β as the witness type. This payoff may be interpreted as the satisfaction obtained from an act of retribution, or simply from performing one’s civic duty. Which of these motives drives witnesses to testify depends on the context; retribution may have been particularly important in the Best Bakery case. Since offenders cannot observe β, they cannot know whether or not the offer of a bribe will induce a witness to refrain from testifying. We introduce voice into this simple model of witness tampering by allowing for the possibility that public outrage can result in the reversal of an acquittal that is perceived to be unjust. Specifically, if the witness accepts a bribe and turns hostile, there is a probability η that the acquittal will be reversed upon appeal. In this case the offender receives the same punishment, ρ, that would have followed an immediate conviction. In addition, the witness receives punishment σ. This makes it more costly for witnesses to accept bribes, and makes bribes less effective in securing acquittals. We assume that α > ησ, otherwise bribes would never be accepted even by witnesses who refuse to testify.f The parameters α, δ, ρ, η and σ, the functions b(γ) and c(γ), and the distributions F and G are all assumed to be commonly known. None of our results are sensitive to the manner in which ties are broken, f One

way to interpret the condition α > ησ is that the size of the bribe is chosen by offenders to be large enough to outweigh the expected penalty from accepting it. A fully endogenous treatment of α beyond the scope of this paper.

April 20, 2009

14:44

WSPC - Proceedings Trim Size: 9in x 6in

Public Outrage and Criminal Justice

JessicaLal

155

since the set of witnesses and offenders who are indifferent in equilibrium is of zero measure. We adopt the convention that ties are broken in favor of legality: witnesses indifferent between acceptance and rejection of a bribe will reject, witnesses indifferent between testifying and not doing so will testify, offenders indifferent between offering a bribe and not doing so will not do so, and offenders indifferent between harming a witness and not doing so will not do so. We focus below on six key probabilities, all endogenously determined in equilibrium. For a witness who has not been offered a bribe, let qn denote the likelihood of testifying, and pn the likelihood of being harmed conditional on testifying. For a witness who has been offered a bribe, let qr denote the likelihood of rejecting it and testifying, and pr the likelihood of being harmed conditional on having done so. Finally, let qa denote the likelihood of accepting and testifying, and pa the likelihood of being harmed conditional on this. 3.2. Witness Behavior A witness who has not been offered a bribe can either testify and obtain a payoff β − pn δ, or remain silent and get payoff 0. Hence the likelihood that the witness will testify is given by qn = 1 − G(pn δ).

(1)

A witness who has been offered a bribe has four options: accept and testify (AT ), reject and testify (RT ), accept and not testify (AN ), and reject and not testify (RN ). The expected payoffs to the witness from the first two options depend on the endogenous probabilities of harm as follows: πAT = α + β − pa δ, πRT = β − pr δ.

(2)

Note that there is no public outrage as long as the witness testifies, even if he does so after accepting a bribe. In this case the offender is convicted and the offer and acceptance of the bribe remain undetected. Of the two options in which the witness does not testify, rejection of the bribe yields πRN = 0 and is therefore strictly dominated by acceptance of the bribe, yielding πAN = α − ησ, which we have assumed to be strictly positive.

April 20, 2009

14:44

WSPC - Proceedings Trim Size: 9in x 6in

JessicaLal

Brendan O’Flaherty and Rajiv Sethi

156

A witness will accept the bribe and refuse to testify if and only if β < min{pa δ − ησ, pr δ + α − ησ}. Hence the likelihood of testimony conditional on a bribe being offered is qa + qr = 1 − G(min{pa δ − ησ, pr δ + α − ησ}).

(3)

What if β ≥ min{pa δ − ησ, pr δ + α − ησ}? In this case the witness will reject the bribe and testify if α ≤ δ (pa − pr ) ,

(4)

and accept the bribe and testify if this inequality does not hold. Since (4) is independent of β, it will never be the case that both AT and RT are selected with positive probability. Hence either qa = 0 or qr = 0 in equilibrium. 3.3. Offender Behavior Since witnesses who are not offered a bribe testify with probability qn , the expected payoff to criminals who do not offer a bribe is πn = −qn (ρ + min{0, γ})

(5)

The expected payoff to an offender who does offer a bribe is πb = −qa (ρ + b(γ) + min{0, c(γ)})−qr (ρ + min{0, γ})−(1−qa −qr ) (b(γ) + ηρ) , (6) Where the last term reflects that fact the conviction will result with probability η even if a bribe is accepted and the witness remains silent. 4. Equilibrium Since the costs of bribery and retaliation both rise with γ, it may be conjectured that there exists an equilibrium with a partitional structure, in which offenders with high costs choose not to bribe and those with low costs bribe. The bribe serves also as a threat since it signals a higher likelihood of retaliation. Witnesses are therefore less likely to testify when offered the bribe. We identify conditions under which such an equilibrium does indeed exist.g We focus on equilibria with the property that there exists some γ˜ > γ¯ such that all offenders with γ < γ˜ offer a bribe in the equilibrium and g Without any restrictions on out-of-equilibrium beliefs, there always exists an equilibrium in which no intimidation occurs and witnesses who are offered a bribe (off the equilibrium path) believe that the offer is from a nonviolent offender. Such equilibria are discussed in the appendix, together with the restrictions on out-of-equilibrium beliefs that are sufficient to rule out their occurrence.

April 20, 2009

14:44

WSPC - Proceedings Trim Size: 9in x 6in

Public Outrage and Criminal Justice

JessicaLal

157

those with γ ≥ γ˜ do not. In this case, equilibrium probabilities of offender violence conditional on witness testimony satisfy pn = 0, F (0) , pr = F (˜ γ) F (¯ γ) pa = , F (˜ γ)

(7) (8)

so pa > pr > pn . Those who are offered a bribe face a greater likelihood of violence conditional on testimony, and this likelihood is highest if they accept the bribe before testifying. We show below that the following conditions are sufficient for the existence of such an equilibrium: b(0) + ηρ < G(F (0)δ) (ρ + γmin ) , b(¯ γ ) + ηρ < ρG(F (0)δ).

(9) (10)

These two conditions require that the costs of bribing witness are not too large relative to the other specifications of the model, in the case of offenders with γ = 0 and γ = γ¯ . (Note that since γmin < 0, the second condition is not implied by the first.) Condition (9) ensures that offenders are not so strongly retaliatory that they want witnesses to testify against them simply to get the satisfaction of harming them, while (10) ensures that the threshold γ˜ exceeds γ¯ , so all potentially violent offenders offer bribes. In addition, we assume the following: α < δ (F (¯ γ ) − F (0)) .

(11)

We show below that this condition ensures that (4) holds in equilibrium, so witnesses who testify after being offered a bribe do so after rejection rather than acceptance. As a consequence, 0 = qa < qr < qn = 1 in equilibrium. We then have (see Appendix for proof): Proposition 1. If (9–11) hold, then there exists an equilibrium with the following properties: (i) all violent offenders and some nonviolent offenders offer bribes, (ii) all witnesses who do not receive bribe offers testify and remain unharmed, and (iii) all witnesses who do receive bribe offers either reject the offers and testify, or accept them and do not testify.

April 20, 2009

14:44

WSPC - Proceedings Trim Size: 9in x 6in

JessicaLal

Brendan O’Flaherty and Rajiv Sethi

158

5. Comparative Statics 5.1. Greater Judicial Effectiveness Note that the marginal offender type γ˜ must be indifferent between bribing the witness and not doing so: πb = πn for this type. Since qn = 0, πn = −ρ. And since γ˜ > γ¯ > 0 and qa = 0, (6) simplifies to πb = −qr ρ − (1 − qr ) (b(˜ γ ) + ηρ) . Hence the indifference condition for the marginal type is simply b(˜ γ ) = (1 − η) ρ.

(12)

An increase in the penalty ρ that offenders face if the witness testifies truthfully therefore implies an increase in the threshold γ˜ , and hence and increase in F (˜ γ ), which is the proportion of offenders who choose to offer bribes. Assuming that the conditions (9–11) continue to hold, witnesses will either reject the bribe and testify, or accept it and refrain from testifying. The likelihood of violence conditional on testifying decreases, since pr =

F (0) . F (˜ γ)

The probability of testifying conditional on receiving an offer of a bribe is qr = 1 − G(pr δ + α − ησ), which rises as a result. Proposition 2. Greater judicial effectiveness results in an increased incidence of attempted bribery, but a decline in the proportion of bribes that are accepted, and a decline in the likelihood that witnesses who testify will be harmed. The effect on the overall rate of testimony, however, is ambiguous. If qr is initially low and changes little in response to the decline in pr , the increased incidence of attempted bribery may result in a decline in testimony and convictions. On the other hand if qr is initially high (so bribes are frequently rejected) and is sensitive to changes in pr , the overall rate of testimony and offender conviction will rise. To be precise, the overall rate of testimony (and hence conviction) is   δF (0) + α − ησ . (13) t = qr F (˜ γ ) + (1 − F (˜ γ )) = 1 − F (˜ γ) G F (˜ γ)

April 20, 2009

14:44

WSPC - Proceedings Trim Size: 9in x 6in

Public Outrage and Criminal Justice

JessicaLal

159

An increase in F (˜ γ ) , arising for instance from harsher penalties for convicted offenders, raises t if and only if g F (˜ γ) 1 > = , G δF (0) pr δ

(14)

where g is the density function derived from G, and g/G is evaluated at the initial equilibrium. Thus harsher penalties are effective when the elasticity g/G is high and the danger of turning down a bribe pr δ is great. 5.2. Greater Voice From the indifference condition for the marginal offender (12), the effect of an increase in the voice parameter η is to lower the threshold γ˜ and hence the incidence F (˜ γ ) of bribery, but with an increase in the likelihood of violence conditional on testimony: pr =

F (0) . F (˜ γ)

But this need not result in a lower rate of testimony conditional on being bribed, since qr = 1 − G(pr δ + α − ησ). Proposition 3. Public outrage results in a decreased incidence of attempted bribery, but an increase in the likelihood that witnesses who testify will be harmed. The effect on the rate of testimony is again ambiguous. Testifying is less attractive because the expected cost of reprisals from the offender is greater, but not testifying is also less attractive because the expected cost of reprisals from public outrage is also greater. Since greater media effectiveness reduces F (˜ γ ) while harsher penalties increase it, greater media effectiveness always increases testimony when (14) fails and greater judicial effectiveness reduce testimony. But the direct effect of greater media effectiveness increases testimony holding F (˜ γ) constant. So there are instances when both greater media effectiveness and greater judicial effectiveness raise testimony, and there are no instances when neither does so. Hence public outrage and greater judicial effectiveness have exactly opposite effects on the incidence of bribery and the likelihood of violence.

April 20, 2009

14:44

160

WSPC - Proceedings Trim Size: 9in x 6in

JessicaLal

Brendan O’Flaherty and Rajiv Sethi

Even though both may lead to greater convictions and overall testimony, the channels through which they operate are very different. Voice is not simply another form of deterrence, it leads to qualitatively different outcomes. 5.3. Wealth and Inequality In most of the cases that motivate this paper, the offender was rich and powerful, and the witnesses were not. Does this inequality matter? Rich offenders have lower b (γ) schedules than poor offenders. In (12) a shift downward in the b(γ) schedule raises the bribe threshold γ˜ . Richer offenders are more likely to offer bribes. Thus comparing rich and poor offenders, rich offenders act as if they face harsher penalties: they attempt more bribes, but fewer bribes are accepted, and they are less likely to harm witnesses who testify. Richer witnesses would value the non-pecuniary benefits of testimony β relatively more than the pecuniary benefits of bribery α. This is reflected in a rightward shift of the G distribution. Richer witnesses do not affect the propensity of offenders to bribe (12) but they are more likely to testify after they have rejected a bribe: qr is higher. Overall, then, the rate of testimony (13) rises as the wealth of witnesses rises. These comparative statics results presume that conditions (9–11) continue to hold. But conditions (9) and (10) both have b on the left hand side and G on the right. As offenders get poorer and witnesses get richer, the left hand side rises and the right hand side falls. Thus the existence of the witness-tampering equilibrium we have described depends on inequality between offenders and witnesses. Growing inequality can give rise to witness tampering even without any changes in legal institutions. 6. Conclusion In many ways the witness tampering cases that have motivated our analysis reveal the strength of the Indian judicial system. Civilians were bribed and threatened, not prosecutors or judges. Poor people were bribed because it was feared that their testimony would be credible. And the media exposed it. A lot has to be going right for witness tampering to be publicly acknowledged problem. We have shown, moreover, that effective legal institutions do not eliminate witness tampering and may even exacerbate it. A major open question that remains is the following: what determines the willingness of the media to expose witness tampering and the willingness

April 20, 2009

14:44

WSPC - Proceedings Trim Size: 9in x 6in

JessicaLal

Public Outrage and Criminal Justice

161

of ordinary people to testify in the face of threats, specific or amorphous? Public outrage is seldom entirely spontaneous; it is often intentionally provoked. This may be done by media outlets to increase circulation or ratings, or by politicians for electoral advantage. Understanding this process, and therefore understanding how the contentiousness of Indian politics gives rise to such cases is an important missing piece of our analysis. The central message of this paper is that legal institutions alone are not enough to produce justice. In terms of Albert Hirschman’s memorable formulation, effective judicial systems require a balance of exit and voice. Open competition in politics and the media facilitates and amplifies the expression of voice, keeping blatant miscarriages of justice at least occasionally in check. Acknowledgements We thank participants at the ISI Conference on Comparative Devlopment, especially Bhaskar Dutta, Parikshit Ghosh, Arunava Sen, and E. Somanathan, for suggesting that our earlier work on witness intimidation, suitably modified and extended, could be relevant to the Indian context. We also thank Kai Friese for helpful comments and suggestions, and Om Prakash Shokeen for assistance with sources. This material is based upon work supported by the Behavioral Sciences Program at the Santa Fe Institute. Appendix Proof of Proposition 1. Consider a candidate equilibrium such that, for some γ˜ > γ¯ all offenders with γ < γ˜ offer a bribe and those with γ ≥ γ˜ do not. At any such equilibrium, pn = 0 and qn = 1. Assume for the moment that (4) holds, so qa = 0. (We show below that this is consistent with equilibrium). From (7) and (8), and the fact that F (˜ γ ) < 1, we have pr > F (0), and pa > pr . Hence from (3) and the assumption that (4) holds and qa = 0, qr = 1 − G(pr δ + α − ησ) < 1 − G(F (0)δ)

(15)

April 20, 2009

14:44

WSPC - Proceedings Trim Size: 9in x 6in

JessicaLal

Brendan O’Flaherty and Rajiv Sethi

162

Now for any γ ≤ 0, πn = − (ρ + γ) and, since qa = 0, πb = −qr (ρ + γ) − (1 − qr ) (b(γ) + ηρ) ≥ − (1 − G(F (0)δ)) (ρ + γ) − (1 − qr ) (b(γ) + ηρ) ≥ − (1 − G(F (0)δ)) (ρ + γ) − (b(γ) + ηρ) ≥ − (ρ + γ) + G(F (0)δ) (ρ + γmin ) − (b(γ) + ηρ) . Hence an offender with γ < 0 will choose to bribe if b(0) + ηρ < G(F (0)δ) (ρ + γmin ) , which is (9). Now consider offenders with γ ≥ 0. For any such offender, πn = −ρ and πb is decreasing, with limγ→∞ πb = −∞. So there exists a unique γ˜ for which πn = πb . (This follows from the continuity of πb in γ at γ = 0.) To verify that γ˜ > γ¯ , we need to show that the type γ¯ strictly prefers to offer a bribe. Note that for this type, using (15) and qa = 0, we have πb = −qr ρ − (1 − qr ) (b(¯ γ ) + ηρ) ≥ − (1 − G(F (0)δ)) ρ − (1 − qr ) (b(¯ γ ) + ηρ) ≥ − (1 − G(F (0)δ)) ρ − (b(¯ γ ) + ηρ) So type γ¯ prefers to offer a bribe if b(¯ γ ) + ηρ < ρG(F (0)δ), which is (10). Suppose this holds, so γ¯ < γ˜ . The likelihood that the witnesses will accept the bribe and refrain from testifying is given by (3). Witnesses who testify will do so after rejection or the bribe if F (¯ γ ) − F (0) α ≤ pa − pr = . δ F (˜ γ)

(16)

Since F (˜ γ ) ≤ 1 a sufficient condition for this is (11). Hence (4) holds and qa = 0 in equilibrium.  Equilibria without Intimidation. We have focused in the text on equilibria in which violent offenders intimidate witnesses. Without any restrictions on out-of-equilibrium beliefs, however, there exists a trivial (and implausible) equilibrium in which no

April 20, 2009

14:44

WSPC - Proceedings Trim Size: 9in x 6in

JessicaLal

Public Outrage and Criminal Justice

163

offender type offers a bribe, and a witness who received a bribe (off the equilibrium path) believes that it comes from an offender with γ > γ¯ . In this case the witness will simply accept the bribe and testify, believing that pa = 0. Knowing this, no offender will offer a bribe. While this can be an equilibrium, the beliefs that sustain it are entirely implausible: since b(γ) is increasing, it would be far more reasonable to suppose that the offender making the offer had a very low value of γ. We make the conservative assumption that a witness receiving an offer of a bribe off the equilibrium path believes that the offender type is drawn from the distribution F (γ), which corresponds exactly to the prior belief of the witness. In other words, the witness believes that the offer was made in error, and all types are equally likely to make such an error. This is sufficient to rule out the equilibrium in which no bribes are offered. To see why, note that in any such equilibrium pn = F (0) and hence qn = 1 − G(δF (0)) from (1). Hence the offender payoff, conditional on no offer of a bribe, is given by πn = − (1 − G(δF (0))) (ρ + min{0, γ}) . If a bribe is offered, witness beliefs are pa = F (¯ γ ) and pr = F (0). Hence from (3), qa + qr = 1 − G(min{δF (¯ γ ), δF (0) + α}). Now consider an offender with γ = γmin . If δF (0) + α < δF (¯ γ ), then qa = 0 and this offender’s payoffs satisfy πb − πn = (G(δF (0) + α) − G(δF (0))) (ρ + γmin ) > 0, so this offender type prefers to offer a bribe. Similarly, if δF (¯ γ ) ≤ δF (0)+α, then qr = 0 and this offender’s payoffs satisfy πb − πn = (G(δF (¯ γ )) − G(δF (0))) (ρ + γmin ) > 0. Again this offender prefers to offer a bribe, contradicting the hypothesis that no bribes are offered in equilibrium. References 1. The Jessica Lal murder case rediff.com(December, 2007). 2. State v. Sidhartha Vashisht Delhi High Court. Criminal Appeal No. 193 of 2006(December, 2006). 3. A. Hirschman, Exit, Voice, and Loyalty: Responses to Decline in Firms, Organizations, and States (Cambridge: Harvard University Press, 1970).

April 20, 2009

14:44

164

WSPC - Proceedings Trim Size: 9in x 6in

JessicaLal

Brendan O’Flaherty and Rajiv Sethi

4. T. Besley and A. Prat, American Economic Review 96, 720 (2006). 5. E. G. Gentzkow M. and C. Goldin, The Rise of the Fourth Estate: How Newspapers Became Informative and Why it Mattered, in Corruption and Reform: Lessons from America’s Economic History, eds. E. Glaeser and C. Goldin (Chicago: University of Chicago, 2006), pp. 187–230. 6. T. N. Djankov S., McLeish C. and A. Shleifer, Journal of Law and Economics 46, 341 (2003). 7. J. McMillan and P. Zoido, Journal of Economic Perspectives 18, 69 (2004). 8. BMW case: Another u-turn by Kulkarni The Times of India(July, 2007). 9. Key witness identifies Nanda in BMW case The Hindu(May, 2007). 10. Best Bakery case: Dateline(December 2007). 11. B. O’Flaherty and R. Sethi, Witness Intimidation, Tech. Rep. 0708-07, Department of Economics, Columbia University (2007). 12. G. Akerlof and J. Yellen, Gang behavior, law enforcement, and community values, in Values and Public Policy, eds. T. M. Henry Aaron and T. Taylor (Washington, D.C.: Brookings, 1994), pp. 173–209.

April 20, 2009

14:54

WSPC - Proceedings Trim Size: 9in x 6in

yama˙proc

LOBBYING FOR TRADE REGIME AND TARIFF SETTINGS KATSUZO YAMAMOTO Kobe University

1. Introduction There has clearly been a dominant stream toward trade liberalization in the last one or two decades. Such stream can be divided into two visible trends: bilateralism (Free Trade Agreements or Regional Trade Agreements) and multilateralism (negotiations within the GATT/WTO framework). However, in recent years, the main trend of trade policy cooperation between countries is notably FTAs and the negotiations toward multilateral free trade tend to reach a standoff. The aim of this paper is to clarify with a political economic model the reason why bilateral free trade negotiations instead of multilateral negotiations have been actively concluded around the world.a Earlier studies about FTAs in the political economic approach have been conducted with both perfect and imperfect competitive markets. Grossman and Helpman (1995)5 explore FTA formations in perfect competitive markets using Grossman and Helpman (1994)6 political contributions approach.b On the other hand, Krishna (1998)8 and Ornelas (2005a)9 analyze FTA formations in oligopolistic markets. Krishna (1998)8 focuses on firms’ profits, which mainly affect government’s decision of concluding an FTA through their lobbying activities, and shows that the more firms an external country has, the more frequently a bilateral free trade agreement is realized. This is because the trade diversion associated with a bilateral agreement becomes larger. Here, the trade diversion means the effect of the shift from the consumptions supplied by an external country to those supplied by ina Literature

reviews on political economic approaches can be found in detailed surveys conducted by Hillman (1989),1 Magee et al. (89),2 Rodrik (1995)3 and Helpman (1996)4 b Grossman and Helpman (1994)6 analyze the relation between lobbying activities and actual trade policy decisions using the common agency model of Bernheim and Whinston (1986),7 and succeed in explaining why industries, by offering political contributions, can make incumbent officeholders practice the trade policy of protection that benefits these industries. 165

April 20, 2009

14:54

166

WSPC - Proceedings Trim Size: 9in x 6in

yama˙proc

Katsuzo Yamamoto

ternal countries. Ornelas (2005)9 furthermore develops Krishna’s (1998)8 model so as to consider firms’ lobbying activities more explicitly and their effects on government’s tariff setting, and finds that an FTA induces member governments to lower their external tariffs and enhance trade volumes even between member and non-member countries by doing so. He calls his effect as trade creation effect and shows that this effect is greater when FTA member governments are more politically motivated. Additionally, it is also suggested that governments will endorse only welfare-improving FTAs and welfare-improving effect by concluding FTA is larger when its government has more its political bias. Our analysis extends Ornelas (2005a)9 works and explores the point the earlier studies have not clarified satisfactorily. We focus on the effects of firm’s lobbying activity over government’s action not only on tariff setting but also on trade regime decision. Our main results can be summarized in the following two points. First, a government prefers to participate in an FTA if its partner country’s market size is large enough. However, if the market size of the rest of the world is large enough compared to that of the partner country, then the government prefers to carry out complete free trade. Second, in the increase of the domestic government’s political bias, it tends to choose an unilateral or an FTA regime rather than a multilateral regime. The remainder of this paper is organized as follows. Section 2 presents the basic formulation of the model. In Section 3, we solve the equilibrium tariff level under a given trade regime. Section 4 shows what trade regime the government chooses in equilibrium and clarifies the causes of the government’s trade regime decision. Section 5 summarizes our main conclusions and presents some future problems.

2. The Model The model below deals with an oligopolistic competition environment, in which each market is segmented. This model is an extension from the duopolistic competition model of Brander and Krugman (1983).10 We consider three countries i, j, k ∈ {A, B, C} and two sectors Z, N in this model. Here, zji is the quantity of good Z supplied by a single firm from country j to country i’s market, and P i is the equilibrium price of good Z in country i’s market. The total sales of the oligopolistically supplied good Z in coun try i are denoted by z i ; therefore, z i = zii + j=i zji . There exists only one firm supplying non-numeraire good Z in each country.

April 20, 2009

14:54

WSPC - Proceedings Trim Size: 9in x 6in

Lobbying for Trade Regime and Tariff Settings

yama˙proc

167

Aggregate utility of the consumers in country i is assumed to take the form, 1 (1) U i (z i , mi ) = αi z i − (z i )2 + ni for all i ∈ {A, B, C}, 2 where ni denotes the consumption of the competitively produced numeraire good N . Because of the quasi-linear form of the utility function, the inverse demand function takes the linear form, P i = αi − z i

for all i ∈ {A, B, C},

(2)

where αi represents country i’s market size. To simplify, we assume that each country’s marginal costs of production of good Z in terms of good N is constant and equal to c. Additionally, we denote ai ≡ αi − c > 0. Then, there is a positive correlation of ai with country i’s market size. We consider the following three trade regimes R = {u, b, m}. u means unilateral, where each country imposes import tariffs on all imported goods ; b means bilateral, where country A and B sign an FTA and abolish import tariffs against each other, but they continuously impose tariffs on the outside country C ; and m means multilateral, where all countries A, B, and C realize complete free trade. Our purpose in this paper is to clarify the effect of firms’ lobbying activity on each country’s decision of trade policies. We consider a three stage noncooperative game with the timing of events as follows: (1) Country A’s firm proposes his contribution schedule to country A’s government contingent to his decision of trade policies CA (r, tA (r)), where ti (r) are the specific tariffs imposed by country i under the trade regime r ∈ R. Here, tA (b) is imposed on imports from country C but is not imposed on those from country B. On the other hand, tA (u) is imposed on both imports from country B and C. Likewise, tA (m) = 0; that is, all imports are not imposed any tariffs on. (2) Country A’s government chooses its trade regime r ∈ R. (3) Country A’s government decides its external tariff level tA (r). We add the following two assumptions about countries B and C: they obey country A’s decision on a trade regime; and there are no lobbying activities of any special interest groups in their countries. We need these assumptions in order to spot on particular country’s political elements and policymakings and to investigate the effect of firm’s lobbying activity in this country for the domestic governments trade regime decision. Additionally, it is postulated that the unilateral regime is adopted and country A’s firm

April 20, 2009

14:54

WSPC - Proceedings Trim Size: 9in x 6in

yama˙proc

Katsuzo Yamamoto

168

doesn’t any lobbying activities in the beginning of the game: that is, country A’s government try to mazimize its domestic welfare under unilateralism in status quo. We solve the equilibrium of this game by backward inductions using the concept of Subgame-Perfect Nash Equilibrium. 3. Tariff Setting 3.1. Unilateral Regime First, we consider the situation where unilateralism is accomplished as a trade regime. Each country’s import tariff is simply added to the marginal costs of firms, so country j’s firm’s effective marginal costs of exports to country i’s markets become c + ti (u). Hence, we can denote the profits of each country’s firm in country i’s markets by πii (u, ti (u)) = [pi (u, ti (u)) − c]zii (u, ti (u)) = [ai − z i (u, ti (u))]zii (u, ti (u)) for all i ∈ {A, B, C},

(3)

πji (u, ti (u)) = [pi (u, ti (u)) − c − ti (u)]zji (u, ti (u)) = [ai − z i (u, ti (u)) − ti (u)]zji (u, ti (u)) for all i, j ∈ {A, B, C} and i = j.

(4)

We can solve the Cournot-Nash equilibrium output level under unilateral regime from these equations as follows: ai + 2ti (u) for all i ∈ {A, B, C}, 4 ai − 2ti (u) for all i, j ∈ {A, B, C} and i = j. zji (u, ti (u)) = 4 zii (u, ti (u)) =

(5) (6)

Substituting them into equations (3) and (4), we can easily get πji (u, ti (u)) = [zji (u, ti (u))]2 . We assume that all tariff revenues are equally distributed to domestic consumers, then we can define each country’s welfare to be composed of the consumer surplus, firm’s profits and tariff revenues; therefore, country i’s welfare can be represented as follows:  zji (r, ti (r)) W i (r, ti (r)) = CSi (r, ti (r)) + πi (r, ti (r)) + ti (r) j=i

for all i, j ∈ {A, B, C} and r ∈ R,

(7)

April 20, 2009

14:54

WSPC - Proceedings Trim Size: 9in x 6in

yama˙proc

Lobbying for Trade Regime and Tariff Settings

169

where CSi (r, ti (r)) is the consumer surplus of country i and πi (r, ti (r)) ≡  πii (r, ti (r)) + j=i πij (r, ti (r)) is the profit by country i’s firm. By the above equation, we can calculate the welfare of country i under unilateralism as follows: W i (u, ti (u)) =

11ai 2 + 12ti (u) · ai − 20[ti (u)]2  j + πi (u, tj (u)) 32 j=i

for all i, j ∈ {A, B, C}.

(8)

Here, the last term, which represents the profit of country i’s firm, is independent of ti (u). When organized lobbies pay campaign contributions to politicians, country A’s government has the following two concerns: (1) social welfare and (2) gathering support for the purpose of holding their office. Therefore, the government’s objective function can be represented as follows:c GA (r, tA (r)) = W A (r, tA (r)) + β · CA (r, tA (r))

β ≥ 0 for all r ∈ R. (9)

Meanwhile, consider the situation where there are no lobbying activities of country A’s firm. Maximizing country i’s welfare with respect to ti (u), we can solve country i’s optimal tariff under unilateralism as follows: ti∗ (u) =

3 ai 10

for all i ∈ {A, B, C}.

(10)

By substituting each country’s optimal tariff level into equation (8), country A’s welfare is represented by  40ai 2 + j=i aj 2 i∗ i A∗ W (u) ≡ W (u, t (u)) = for all i, j ∈ {B, C}. (11) 100 Likewise, the profits of country i’s firm can be solved as follows:  16ai 2 + j=i aj 2 ∗ A∗ πi (u) ≡ πi (u, t (u)) = for all i, j ∈ {B, C}. 100

(12)

In the situation where country A’s firm pay contributions to country A’s government, country A’s optimal tariff can be solved by substituting CA (u, tA (u)) = πA (u, tA (u)) − V into equation (9) and maximizing c Such a form of the objective function is called the political support function. Here, β means the weight the government attaches to campaign contributions compared to the aggregate welfare. See Hillman (1982) and Grossman and Helpman (1994).6

.

April 20, 2009

14:54

WSPC - Proceedings Trim Size: 9in x 6in

yama˙proc

Katsuzo Yamamoto

170

GA (u, ti (u)). Here, V represents country A’s firm’s reservation payoffs. We call such tariff level the political tariff and represent it by tAP (u). Then, tAP (u) =

3 + 2β aA . 10 − 4β

(13) 1 2 in order AP Since ∂t ∂β(u) =

Here, we restrict the weight for campaign contributions to β
0, an increase in β induces a higher political tariff level.

Using country A’s political tariff level and the other countries’ optimal tariff level, we can solve country A’s welfare and the profits of country A’s firm as follows:   11 1  2 AP A AP + Φ aA 2 + aj W (u) ≡ W (u, t (u)) = 32 100 j=A

for all j ∈ {B, C}, P πA (u) ≡ πA (u, tAP (u)) = ΨaA 2 +

(14)

1  2 aj 100 j=A

for all j ∈ {B, C}, where Φ ≡

(3+2β)(15−22β) , 8(10−4β)2

Ψ≡

(15)

4 (5−2β)2 .

3.2. Bilateral Regime In this subsection, we consider the negotiation of an FTA between country A and B under the existence of three countries; A, B, C. Countries signing an FTA abolish the import tariff against each other, but they continue to impose the prior tariff level ti (b) on the outside country. Here, country C is interpreted as the rest of the world. The profits made by each country’s firm in country i’s market participated into FTA are represented by πji (b, ti (b)) = [pi (b, ti (b)) − c]zji (b, ti (b)) = [ai − z i (b, ti (b))]zji (b, ti (b)) for all i, j ∈ {A, B},

(16)

i i (b, ti (b)) = [pi (b, ti (b)) − c − ti (b)]zC (b, ti (b)) πC i (b, ti (b)) for all i ∈ {A, B}. (17) = [ai − z i (b, ti (b)) − ti (b)]zC

April 20, 2009

14:54

WSPC - Proceedings Trim Size: 9in x 6in

yama˙proc

Lobbying for Trade Regime and Tariff Settings

171

We can solve the Cournot-Nash equilibrium output level from these equations as follows:d ai + ti (b) for all i, j ∈ {A, B}, 4 ai − 3ti (b) i for all i ∈ {A, B}. (b, ti (b)) = zC 4

zji (b, ti (b)) =

(18) (19)

Then, the profits made by country j’s firm in country i’s market are i i (b, ti (b)) = [zC (b, ti (b))]2 ; Therefore, country πji (b, ti (b)) = [zji (b, ti (b))]2 , πC i’s welfare under bilateralism is represented by W i (b, ti (b)) =

11ai 2 + 6ti (b) · ai − 21(ti (b))2 + πij (b, tj (b)) + πiC (u, tC∗ (u)) 32 for all i, j ∈ {A, B}. (20)

If there are no special interest groups, each country that participates in the FTA sets the optimal tariff so as to maximize its own welfare. Maximizing country i’s welfare with respect to ti (b), we can solve country i’s optimal tariff level under FTA as follows: 1 ti∗ (b) = ai 7

for all i ∈ {A, B}.

(21)

Then, country i’s welfare is calculated by substituting each country’s optimal tariff level into equation (20), W i∗ (b) ≡ W i (b, ti∗ (b)) =

1750ai2 + 400aj 2 + 49aC 2 4900

for all i, j ∈ {A, B}. (22)

Additionally, we can solve the profits of country i’s firm as follows: πi∗ (b) ≡ πi (b, ti∗ (b)) =

400ai 2 + 400aj 2 + 49aC 2 4900

for all i, j ∈ {A, B}, (23)

where πi (b, ti (b)) ≡

 j

πij (b, tj (b)) + πiC (u, tC∗ (u)) for all i, j ∈ {A, B}.

d In the situation where country A and B conclude an FTA with each other, the equilibrium production quantities from each country’s firm against country C’s market are the same as those under unilateralism. Hence, country C’s optimal tariff under bilateralism is also the same as that under unilateralism.

April 20, 2009

14:54

WSPC - Proceedings Trim Size: 9in x 6in

yama˙proc

Katsuzo Yamamoto

172

In the situation where country A’s firm pay contributions to country A’s government, country A’s political tariff level can be solved by substituting CA (b, tA (b)) = πA (b, tA (b)) − V into equation (9) and maximizing GA (b, ti (b)). Therefore, tAP (b) =

3 + 2β aA 21 − 2β

for all i ∈ {A, B}.

(24)

AP

48 Since ∂t∂β = (21−2β) 2 > 0, an increase in β induces a higher political tariff level against the external country C. Using country A’s political tariff level and the other countries’ optimal tariff level, we can calculate Country A’s welfare and the profits of country A’s firm as follows:   11 4 1 AP A AP + Θ aA 2 + aB 2 + aC 2 , (25) W (b) ≡ W (b, t (b)) = 32 49 100 P (b) ≡ πA (b, tAP (b)) = ΩaA 2 + πA

where Θ ≡

(3+2β)(63−54β) 32(21−2β)2 ,

Ω≡

4 1 aB 2 + aC 2 , 49 100

(26)

36 (21−2β)2 .

3.3. Multilateral Regime The profits made by each country’s firm in country i’s market under multilateral are represented by πji (m) = (pi − c)zji (m) = (ai − z i (m))zji (m) for all i, j ∈ {A, B, C}. (27) We can solve the Cournot-Nash equilibrium output level from these equations as follows: zji (m) =

ai 4

for all i, j ∈ {A, B, C}.

(28)

Then, the profits made by country j’s firm in country i’s market are πji (m) = [zji (m)]2 . The country i’s welfare and the profits of country i’s firm are represented by  11ai 2 + 2 j=i aj 2 1  2 , πi∗ (m) = aj W i∗ (m) = 32 16 j for all i, j ∈ {A, B, C}.

(29)

April 20, 2009

14:54

WSPC - Proceedings Trim Size: 9in x 6in

yama˙proc

Lobbying for Trade Regime and Tariff Settings

173

Under multilateralism, all countries practice complete free trade. Thus, each country’s welfare is independent on the amount of political contributions contingent to tariff level, that is, W i∗ (m) = W iP (m). Likewise, we can easily derive that πi∗ (m) = πiP (m). 4. Trade Regime Decision Our main concern in this section is to clarify what trade regime can be chosen by country A, which has the initiative in carrying out it. The equilibrium trade regime rP in the second stage of the trade-policy-decision game in our model is as follows; rP = u

if GA (u, tAP (u)) > GA (s, tAP (s)) for all s ∈ {b, m},

rP = b

if GA (b, tAP (b)) > GA (s, tAP (s)) for all s ∈ {u, m},

rP = m

if GA (m) > GA (s, tAP (s)) for all s ∈ {u, b}.

As in Ornelas (2005a),9 we formulate that the contribution schedule of country A’s firm depends on its bargaining power in a lobbying negotiation between the government and the firm. If the government does not possess bargaining power at all, then the contribution schedule of the firm is customized so that contributions leave the government with its reservation payoff G0 = W A∗ (u); therefore, min (r, tAP (r)) = CA

W A∗ (u) − W AP (r) β

for all r ∈ R.

(30)

If the firm has no bargaining power, then the contribution schedule of the firm is customized so that contributions leave the firm with its reservation payoff V0 = π A∗ (u); therefore, max P ∗ CA (r, tAP (r)) = πA (r) − πA (u).

(31)

Generally, the equilibrium contribution schedule of the firm lies between these two values. Letting γ and 1 − γ denote respectively the government’s and the firm’s bargaining power, for any γ ∈ (0, 1], the equilibrium contribution schedule of country A’s firm under unilateralism can be represented as min max (r, tAP (r)) + γ · CA (r, tAP (r)). CA (r, tAP (r)) = (1 − γ)CA

(32)

April 20, 2009

14:54

WSPC - Proceedings Trim Size: 9in x 6in

yama˙proc

Katsuzo Yamamoto

174

Substituting it into equation (9), we can derive the payoff of country A’s government under each trade regime as follows:e GAP (r) ≡ GA (r, tAP (r)) = γW AP (r) + (1 − γ)W A∗ (u) P ∗ + β · γ[πA (r) − πA (u)] for all r ∈ R.

(33)

By solving the equilibrium trade regime in the second stage of the upper game, the following proposition can be derived.f Proposition 4.1. If the partner country’s market size is enough large, then the domestic government prefers to participate in an FTA with its country. However, if the market size of the rest of the world is large enough compared to that of the partner country, then the government prefers to carry out complete free trade.

aC 2 aA 2

Υ

rP = m rP = u

rP = b

Ξ

0 Fig. 1.

45◦

Λ

aB 2 aA 2

The condition of each equilibrium trade regime

Figure 1 shows the relationship between country B and C’s market size relative to country A’s market size and depicts the condition that each trade e Here,

if γ = 0, then the politicians’ payoff is equal to the domestic welfare under unilateralism, that is, GAP (r) = W A∗ (u). Then, the government is indifferent to trade regimes. Hence, we eliminate the case where country A’s government has no bargaining power against the domestic firm. f About the detailed derivation of the following propositions, see Appendix.

April 20, 2009

14:54

WSPC - Proceedings Trim Size: 9in x 6in

yama˙proc

Lobbying for Trade Regime and Tariff Settings

175

regime is realized in equilibrium with respect to each country’s relative market size. Here, the trade regime decision of the government is independent on the government’s bargaining power γ. Furthermore, we can confirm that such a condition depends on the scale of the domestic government’s political bias which distort the government’s trade policy decision. The following proposition summarizes the effect of shifting this bias β on the government’s trade regime decision. Proposition 4.2. As a parameter β reflecting the government’s taste for campaign contributions relative to its sensitivity to the average voter’s wellbeing increases, the government tends to choose unilateralism rather than bilateralism or multilateralism, or bilateralism rather than multilateralism. Figure 2 depicts how these conditions change according to an increase of β. This proposition can be easily shown by this figure. The reason why the government prefers to keep unilateral situation or conclude an FTA rather than choose complete free trade is because the government can make the firm pay more contributions by setting the tariff level most preferred by the firm under an unilateral or a bilateral regime than under a multilateral regime. Ornelas (2005)9 shows that an FTA improves the welfare of a participating country by more, the grater its government’s political bias. This result implies that the net welfare benefit of an FTA is greater when the trade policy distortions arisen from political bias are larger. Our investigations can show that if the domestic government’s political bias becomes aC 2 aA 2

Υ

rP = m rP = u

rP = b

Ξ

0 Fig. 2.

Λ

aB 2 aA 2

The effect of an increases in β on the equilibrium trade regime

April 20, 2009

14:54

WSPC - Proceedings Trim Size: 9in x 6in

yama˙proc

Katsuzo Yamamoto

176

aC 2 aA 2

rP = b

rP = m

rP = u

rP = b

45◦

0 Fig. 3.

aB 2 aA 2

The equilibrium trade regime when country A chooses an FTA partner

larger, an equilibrium trade regime tends to be more protectionism in spite of a profitable FTA. As shown in Figure 3, our propositions are still realized even if country A’s government can choose the partner country from either B or C in determining their trade regime.

5. Concluding Remarks In this paper, we analyze the effect of a firm’s lobbying activity over the trade regime decision of a government. Our main results are summarized as follows. First, if the partner country has an enough market size, then the government prefers to participate in an FTA with its country. However, if the market size of the rest of the world is large enough compared to that of the partner country, then the government prefers to carry out complete free trade. Second, in the increase of the weight the government attaches to campaign contributions compared to the aggregate welfare (that is, if the government is more politically motivated), they tend to keep protectionism instead of choosing an concluding an FTA or realizing multilateral free trade. Here, the values of both the government’s bargaining power and the weight of campaign contributions compared to the aggregate welfare are exogenously given in this paper. One of our future problems is to analyze the effects of firms’ lobbying activity with endogenous these values.

April 20, 2009

14:54

WSPC - Proceedings Trim Size: 9in x 6in

Lobbying for Trade Regime and Tariff Settings

yama˙proc

177

Acknowledgments I would like to thank Satya P. Das, Taiji Furusawa, Kaoru Ishiguro, Seiichi Katayama, Toru Kikuchi, Noritsugu Nakanishi, Akira Okada, Yoshikatsu Tatamitani and the participants of the numerous conferences and seminars for their helpful advices.

Appendix In this appendix, we show the detailed calculation that is required in order to derive our propositions. From equation (33), we can rewrite the equilibrium trade regime as follows: P P (u) − πA (s)] > 0 for all s ∈ {b, m}, rP = u if W AP (u) − W AP (s) + β[πA P P (b) − πA (s)] > 0 for all s ∈ {u, m}, rP = b if W AP (b) − W AP (s) + β[πA

rP = m

P P if W AP (m) − W AP (s) + β[πA (m) − πA (s)] > 0 for all s ∈ {u, b}.

Solving the condition that each trade regime is realized in equilibrium, we can get aB 2 > Λ, (34) aA 2 aC 2 aB 2 P P (m) − πA (u)] > 0 ⇔ > − + Υ, (35) W AP (m) − W AP (u) + β[πA aA 2 aA 2 aC 2 125 aB 2 P P · W AP (m) − W AP (b) + β[πA (m) − πA (b)] > 0 ⇔ > + Ξ, aA 2 343 aA 2 (36) P P W AP (b) − W AP (u) + β[πA (b) − πA (u)] > 0 ⇔

1 1 400[Φ−β ( 16 −Ψ)] 400[Θ−β ( 16 −Ω)] where Λ ≡ 4900[Φ−Θ+β(Ψ−Ω)] , Υ≡ , Ξ≡ . 351(1+β) 21(1+β) 21(1+β) Figure 1 depicts these conditions with respect to each country’s relative market size. In this figure, we can show that what trade regime the government choose in equilibrium. Next, we consider how an increase of the weight the government attaches to campaign contributions compared to the aggregate welfare affect ∂Υ ∂Ξ these conditions. Since ∂Λ ∂β > 0, ∂β > 0, ∂β > 0, such an effect can be depicted by Figure 2. Here, we can confirm that both areas where unilateral and bilateral regimes can be realized in equilibrium are expanded with the increase of β. On the contrary, a multilateral regime is more difficult to be realized in equilibrium.

April 20, 2009

14:54

178

WSPC - Proceedings Trim Size: 9in x 6in

yama˙proc

Katsuzo Yamamoto

References 1. A. L. Hillman, The Political Economy of Protection (Harwood Academic Publishers, Chur, 1989). 2. W. A. Magee S., W. A. Brock and L. Young, Black Hole Tariffs and Endogenous Policy Theory (Cambridge University Press, Cambridge, 1989). 3. D. Rodrik, Political Economy of Trade Policy, in Handbook of International Economics, eds. G. M. Grossman and K. Rogoff (North-Holland, Amsterdam, 1995), pp. 1457–1494. 4. E. Helpman, Politics and Trade Policy, in Advances in Economics and Econometrics: Theory and Applications, eds. D. M. Kreps and K. F. Wallis (Cambridge University Press, Cambridge, 1996), pp. 19–45. 5. G. M. Grossman and E. Helpman, American Economic Review 105, 667 (1995). 6. G. M. Grossman and E. Helpman, American Economic Review 84, 833 (1994). 7. B. D. Bernheim and M. D. Whinston, Quarterly Journal of Economics 101, 1 (1986). 8. P. Krishna, Quarterly Journal of Economics 113, 227 (1998). 9. E. Ornelas, Journal of International Economics 67, 471 (2005). 10. J. Brander and P. Krugman, Journal of International Economics 15, 313 (1983).

Labour

This page intentionally left blank

LABOR RETRENCHMENT LAWS AND THEIR EFFECT ON WAGES AND EMPLOYMENT: A THEORETICAL INVESTIGATION KAUSHIK BASU Department of Economics, Cornell University, Ithaca, New York 14853. Email: [email protected] GARY S. FIELDS School of Industrial and Labor Relations, Cornell University, Ithaca, New York 14853. Email: [email protected] SHUB DEBGUPTA Corporate Executive Board, 2000 Pennsylvania Avenue, NW, Washington, DC 2006. Email: [email protected] Many countries have legislation which make it costly for firms to dismiss or retrench workers. In the case of India, the Industrial Disputes Act, 1947, requires firms that employ 50 or more workers to pay compensation to any worker who is to be retrenched. This paper builds a theoretical model to analyze the effects of such anti-retrenchment laws. Our model reveals that an anti-retrenchment law can cause wages and employment to rise or fall, depending on the parametric conditions prevailing in the market. We then use this simple model to isolate conditions under which an anti-retrenchment law raises wages and employment. In a subsequent section we assume that the law specifies exogenously the amount of compensation, s, a firm has to pay each worker who is being dismissed. It is then shown that as s rises, starting from zero, equilibrium wages fall. However beyond a certain point, further rises in s cause wages to rise. In other words, the relation between the exogenously specified cost to the firm of dismissing a worker and the equilibrium wage is V-shaped.

1. Introduction Many countries legislate guidelines or procedures to be followed, including compensation to be paid to the workers, in the event of their being dismissed or retrenched. At times, these laws virtually prohibit the dismissal of laborers, disregarding any contract that a worker and his or her employer may have signed

181

182

Kaushik Basu, Gary S. Fields and Shub Debgupta

at the time of employment. At times, the legislation includes some form of predetermined severance compensation. This tendency to legislate conditions for compensation which override voluntary contracts, has been witnessed in many countries, with varying degrees of stringency. Some countries, such as India and France, are known for their strict anti-retrenchment regulations. In the case of India, the Industrial Disputes Act, 1947, requires firms that employ 50 or more workers to pay a compensation, which is specified in the Act, to any worker who is retrenched. In addition, an amendment to the Act, which became effective in 1984, requires firms which employ more than 100 workers to actually seek prior permission from the government before retrenching workers. And, as Datta Chaudhuri6 has noted, the government seldom gives permission and, in general, places a lot of a priori restrictions on the terms for hiring and firing workers (see, also Mathur,16; Edgren9; Papola18)a. What is interesting about such laws is how lay opinion on them is at divergence from the opinion of economists. The popular wisdom on this issue is that these anti-retrenchment laws help labor, but hurt the development process, as they force firms to maintain huge workforces that reduce their ability to make profit. One problem with the conventional wisdom is that it fails to capture the fact that anti-retrenchment laws raise the effective cost of employing labor and, as a result, firms may hire fewer workers. Additionally, it is conceivable that, given the presence of such laws, some firms may not enter into production in the first place. Hence, the economists' view of this is often the opposite of the lay opinion: By burdening firms with the risk that they may not be able to fire their workers or that they will have to pay very large compensations in order to do so, the anti-retrenchment laws cause a decline in the demand for labor and thereby cause a lowering of wages and so ultimately hurt workers. This paper builds a theoretical model to analyze whether anti-retrenchment laws can be expected to help workers or not. The model is a useful tool of analysis for nations like India, France, Italy and Zimbabwe, which have stringent anti-retrenchment laws that have come in for criticism and debate and, in some cases, calls for repeal. It can also be useful for countries, like eastern European ones and China, which have weak anti-retrenchment laws and have debated strengthening these. a

It is worth noting that the state governments of India are allowed to amend this legislation of the central government, within certain limits, since labor laws belong to what are called ‘concurrent list’ policies in India. Hence, there are some inter-state variations that have appeared over the years (Besley and Burgess4).

Labor Retrenchment Laws and Their Effect on Wages and Employment

183

Our theoretical investigation reveals that the truth is more complicated than either of the two views mentioned above. In section 2 we first build a very simple benchmark model, in which there is no industry-wide risk but individual firms face stochastic shocks in each period which may cause fluctuations in the number of workers that they would ideally want to employ. We then consider two alternative legal regimes – one in which an employer cannot retrench workers once they are employed and another regime in which the employer can freely hire and fire. We show that even in such a simple model there is no unique effect on wages: an anti-retrenchment law can cause wages to rise or fall, and aggregate employment to rise or fall, depending on the parametric conditions prevailing in the market. We then use this simple model to isolate conditions under which an anti-retrenchment law raises wages and employment and conditions under which it lowers wages and employment. In subsequent sections we progressively relax the assumptions of the benchmark model. Section 3 endogenizes industry size. Section 4 introduces an exogenously determined amount of compensation, s, that a firm has to pay each worker who is being dismissed. Then, s = 0 and s = ∞ correspond to the two regimes in the benchmark model. In Section 4 we also model the dynamic optimization problem that each firm faces. By using a combination of algebra and geometry we demonstrate how the labor market equilibrium can be characterized very simply. The model is then used to demonstrate a paradoxical result: As s rises, starting from zero, equilibrium wages fall (as expected). However beyond a certain point, further rises in s cause wages to rise. In other words, the relation between the exogenously specified cost to the firm of dismissing a worker and the equilibrium wage that prevails in the market is Vshaped. We also use the model to comment on the possibility of frictional unemployment and the consequence of industry size being endogenously determined. Among notable studies on this topic we would draw attention to papers by Lucas15, Bentolila and Bertola5, Hopenhayn and Rogerson13, Fallon and Lucas10, Anderson and Meyer1, Besley and Burgess4, and Basu3. Bentolila and Bertola5 use aggregate data from France, Germany, Italy and UK to calibrate their continuous-time stochastic model to analyze the labor demand of a single monopolist in the face of changing hiring and firing costs. Their results, based on a partial equilibrium analysis, were that dismissal costs actually raised long run employment. Their results were based on a specific calibrated model and are therefore applicable primarily to the four countries studied.

184

Kaushik Basu, Gary S. Fields and Shub Debgupta

Anderson and Meyer1 using data on the United States, showed that if unemployment insurance taxes are seen as an adjustment cost for changes in labor size, then an increase in these taxes will decrease demand for labor in a high demand state, but will increase employment in a low demand state. In our paper we provide a theoretical validation for this observation and incorporate it into the anti-retrenchment setting. Fallon and Lucas10 in an empirical study on the anti-retrenchment laws in Zimbabwe and India, showed that as the laws were strengthened (making it more difficult to layoff a worker), long-run demand for employees fell by 25.2 percent and 17.5 percent, respectively. They could not determine any significant reduction in wages as a result of the laws. In the case of India the point has been made time and again that rigid labor market legislation may have hurt India's overall growth and efficiency. The claim that is being forwarded in this paper is different. It is being argued that the legislation may have hurt the very constituency that it was meant to protect, to wit, labor. Hence, Kannan's14 observation that wages in the eighties have not kept pace with labor productivity and Ghose's12 finding that employment per unit of gross value added in manufacturing fell monotonically throughout the eighties (see, also, Dev7) sit very well with the theoretical findings of this paper and the fact that India's labor market legislation was made more rigid in the eighties.b In fact, data from Annual Survey of Industries, Government of India, show that, between 1982-83 and 1990-91, the number of workers employed in firms employing 100 to 199 workers fell by 28.5 percent and the number of workers employed in firms employing over 200 workers fell by 43%. It should be recalled that in 1984 new laws were enacted making retrenchment by firms employing over 100 workers especially difficult. It is not surprising that firms would respond to this by cutting down on labor intensity or switching over to contract labor. In this paper, we abstract away from specific institutional aspects of job security legislation and of the industrial labor markets in different countries, and use a parsimonious version, which we believe capture pertinent aspects of the economic environment. We show that the special case, which is nested within a general model, produces theoretically ambiguous labor market effects. Thus, b

One matter that we do not go into explicitly but is important for assessing worker welfare in developing and transition economies is the problem of employment in public sector concerns. There are interesting parallel questions that arise in that context concerning over-staffing and retrenchment. Some of the practical problems arising in that context are discussed in Rama19. Commenting on state-owned enterprises in China, Meng17 has drawn our attention to the important fact that restrictions on labor policy, such as controls on retrenchment, arise not just from the law but often from the prevalent political and social culture.

Labor Retrenchment Laws and Their Effect on Wages and Employment

185

the seemingly contradictory results obtained in the literature, where in some cases more restrictive anti-retrenchment laws cause a fall in employment versus the cases where they result in a rise in employment, can be obtained from the single model presented in this paper for different parameter values. It is hoped that these results will not only provoke further empirical work, but give us pause for thought in our effort to draft policy and legislation to improve labor standards and, more generally, labor welfare. 2. Labor Market Equilibria With or Without Retrenchment In this section we model the likely effects of anti-retrenchment laws. We shall, in particular, consider two alternative legal regimes: one in which employers are free to retrench workers at will and another in which no retrenchment, whatsoever, is allowed. These will be referred to as the "free retrenchment (F) regime" and "no retrenchment (N) regime", respectively. Reality is, of course, more complicated, where retrenchment laws take shades of gray instead of being black and white. However, formalizing the two polar cases helps us fix our ideas and gives us some benchmarks to use. Once we have formalized these we shall go on in Section 4 to formulate an intermediate case with severance compensation. Our aim in this section is to model the kinds of equilibria that arise in the labor market under the F and N regimes. In particular, we want to study the impact of alternative legal regimes on wages and employment. Since antiretrenchment laws are enacted with the aim of enhancing the welfare of labor it is worth checking formally whether this actually happens, once everyone has had time to adjust to the new laws. As stated in intuitive terms earlier, it turns out that anti-retrenchment laws do not always help workers. The formal model below illustrates how, given certain parametric configurations, a switch from regime F to regime N can actually lower the wage and aggregate employment. In other words, if such parametric configurations occur, the anti-retrenchment legislation would, paradoxically, work to the laborers' detriment. It is shown that there are also parametric configurations where the non-paradoxical result occurs, that is, workers' wages rise. As a base case, let us suppose there are n identical firms, each endowed with a production function as follows: x = ϕ f ( L), f '( L) > 0, f ''( L) < 0

(1)

186

Kaushik Basu, Gary S. Fields and Shub Debgupta

L is the amount of labor used by the firm, x the firm's output, and ϕ a stochastic variable which takes values of 1 and 0 with probability p and (1-p) respectively. To keep the algebra simple we model this stochastic shock as being independent across firms and across time, thereby ruling out industrywide fluctuations and intertemporally shocks.c Then, appealing to the law of large numbers, we assume that in each period a randomly selected pn firms have ϕ =1 and (1-p)n firms have ϕ = 0.d We ignore the integer problem here by assuming that pn is an integer. It is also assumed, for the time being, that in each period or year each firm gets to see ϕ before making its hiring decision. If for a certain firm ϕ = 1, we shall describe that as a good year for the firm. A bad year is one in which ϕ = 0. There are different ways of interpreting φ. It could represent input and technology shocks that firms receive. In a bad year φ = 0 and the firm is unable to produce. Alternatively, as suggested in footnote c, we could think of φ as denoting the price of the product. The firms are scattered in different geographical locations. In each location demand for the product can be high or low. When demand is high, the price of the product in one and demand is low, it is zero. Let us first model the F regime. Each firm is free to hire and lay off workers as it sees fit. In a good year, a firm's demand for labor is given by solving the following problem:

Maximize f ( L) − wL L

The first order condition is given by: f '( L) = w.

c

(2)

This is in contrast to models involving seasonality, in which aggregate demand for labor fluctuates from lean season to peak season. While such models are quite common in the development literature (see, for instance, Bardhan2; Dreze and Mukherjee8; and Ray20), they pertain typically to the agricultural sector. In a model of the manufacturing sector, such as ours, it is not unreasonable to think of fluctuations in demand for the product of particular firms caused by shifts in demand from one segment of the industry to another. Our model may be thought of as a stylized description of this. It is not very difficult to incorporate correlated shocks in the model, but given the limited objective of the present paper, we have preferred to keep this simple. A more sophisticated model would allow for idiosyncratic and correlated shocks. d It is worth noting that we could, instead of assuming n firms, have assumed that there is a continuum of firms, spanning, say, the interval [0,1]. In that case we would get the same result without having to use the law of large numbers.

Labor Retrenchment Laws and Their Effect on Wages and Employment

187

Since f"(L) < 0, the function f'(L) can be inverted and written as g(w). Hence, in a good year, with market wage w, a firm's demand for labor is given by L = g ( w).

(3)

In a bad year since ϕ = 0, a firm's demand for labor is obviously 0. Hence in the F regime, the aggregate demand for labor is given by png(w). Let the aggregate supply function of labor be given by s(w) such that s'(w) ≥ 0. Assuming that the usual equilibrating forces are free to operate in the F regime, the market equilibrium is one in which aggregate labor demand equals aggregate labor supply. Clearly wF is the equilibrium wage in the F-regime if and only if

png ( w F ) = s ( w F ).

(4)

Figure 1, below, depicts this equilibrium. The equilibrium amount of labor demanded and supplied is denoted by LF. Wage

s(w)

EF

wF

png(w)

LF

Labor Figure 1.

188

Kaushik Basu, Gary S. Fields and Shub Debgupta

Let us now depict the equilibrium in an N-regime. We are therefore considering an economy in which there is a law requiring that once a firm hires a laborer it cannot retrench him. Since we already know what the aggregate supply curve is, namely, s(w), all we have to do is to work out the demand for labor in this regime and then look for the wage that equates demand and supply. A short-cut method for working out the demand for labor in an N-regime is to assume that each employer must decide how much labor to hire before ϕ is revealed. What we are claiming is that, in an N-regime, this is a good approximation of the case where the firms have to choose how many workers to hire, after seeing the value of φ. The reason why this method works is that apart from the first period, the employer will (in an N-regime) be effectively stuck with a certain amount of labor in all periods without knowing each period's realization of ϕ. So unless the future is too heavily discounted, the fact that ϕ is known in one period, namely, the first period is of negligible importance. Hence, our method causes an error, but of a sufficiently small order, so that it may be ignored. This is demonstrated in Appendix A. For a risk-neutral firm that does not know the value of ϕ in advance, in the N-regime, the firm's problem is to maximize expected profits Maximize f ( L) − wL L

Hence, by the first-order condition we get f '( L) = w / p.

(5)

Hence, the firm's demand for labor is given by: ⎛ w⎞ L = g⎜ ⎟ ⎝ p⎠

(6)

where g(.) is, as before, the inverse of f '(L). Because each firm is identical ex ante, the aggregate demand for labor is ng(w/p). So wN is an equilibrium wage in an N-regime if the amount of labor demanded equals the amount supplied at that wage: ng ( w N / p ) = s ( w N ).

(7)

On inspecting (4) and (7) it appears possible that wN may exceed wF, be less than wF, or be the same as wF depending on the value of p and the shape of the

Labor Retrenchment Laws and Their Effect on Wages and Employment

189

g(.) function. This is easily proved by considering the following special case of the above model. Let the production function, f (L), be the following quadratic one. f ( L) = ( A / B) L − L2 / 2 B

(8)

with A, B > 0. As long as L is less than A (and we shall throughout confine attention to such cases), (8) is a reasonable production function, satisfying f'(L) > 0 and f"(L) < 0. It is easy to verify that (8) implies that the firm has a linear labor demand function g(w) = A – Bw.

(9)

Using (9) the aggregate demand for labor in the economy is given by LF = pn[A – Bw]

(10)

LN = n[A – Bw / p]

(11)

in the F-regime, and by

in the N-regime. These two labor demand curves are illustrated in Figure 2. One sees that they necessarily cross. The point of intersection is denoted by E. Wage Demand in F-Regime

A/B

S

pA/B

S'

E

N

EF

pA/(1+p)B

EN′

S EF′ S' Figure 2.

190

Kaushik Basu, Gary S. Fields and Shub Debgupta

Suppose we begin with a world in which retrenchments are freely allowed. Hence, the aggregate demand curve of labor is given by the steeper demand curve in Figure 2, marked 'Demand in F-regime'. If the laws of the land are changed and firms are not allowed to retrench labor, the effect on wages depends on where the initial equilibrium was. Suppose first that the labor supply curve was like the curve marked SS in Figure 2, so that the equilibrium lay to the left of E, at the point labeled EF in the figure. Then a switch to the Nregime would move the equilibrium from EF to EN, thereby lowering wages and employment, and hurting workers. On the other hand, if the labor supply curve is to the right of E, as for curve S'S', and the equilibrium was at EF', a switch to the N-regime moves the equilibrium to EN', raising wages and employment, and benefiting workers. The special case of p = ½ is analyzed in the following footnote.e Thus far we have assumed that the labor supply function, s(w), remains unaffected by regime switches between no-retrenchment (N) and freeretrenchment (F) regimes. At one level this is as it should be, since, in the aggregate, there is no involuntary unemployment in our model. But, if we assume that workers mind the transaction cost of switching jobs, the same wage w is more attractive to the worker when it comes with a no-retrenchment clause. Extending our assumption that a better offer elicits a greater labor supply, we may assume we have two supply functions, sF(.) and sN(.), for the two regimes, and that, for all w,f sF(w) ≤ sN(w).

This complicates the analysis in Figure 2. If the original equilibrium were at EF, then a switch in regime to no-retrenchment causes an even more precipitous fall in the wage, since SS shifts right as the regime changes. However the effect on employment is now ambiguous. If SN is sufficiently to e

This is in contrast to models involving seasonality, in which aggregate demand for labor fluctuates from lean season to peak season. While such models are quite common in the development literature (see, for instance, Bardhan2; Dreze and Mukherjee8 and Ray20), they pertain typically to the agricultural sector. In a model of the manufacturing sector, such as ours, it is not unreasonable to think of fluctuations in demand for the product of particular firms caused by shifts in demand from one segment of the industry to another. Our model may be thought of as a stylized description of this. It is not very difficult to incorporate correlated shocks in the model, but given the limited objective of the present paper, we have preferred to keep this simple. A more sophisticated model would allow for idiosyncratic and correlated shocks. f In case we were working with a Lewis-type perfectly-elastic labor-supply function, the worker's preference for avoiding transactions cost would imply that the supply curve in the N-regime is below the supply curve in the F-regime. The analysis thereafter proceeds in the usual manner.

Labor Retrenchment Laws and Their Effect on Wages and Employment

191

the right of SF, employment may actually rise. Moreover, the welfare-effect of the switch in regime is also ambiguous, since the same wage in an N-regime gives greater welfare to the worker (by virtue of a lower transactions cost). Hence a lower wage in an N-regime may or may not give lower total welfare. Similarly if the original equilibrium were at EF', a change of regime to noretrenchment will now exert a smaller upward pressure on wages. Indeed wages may now fall. This will happen if the rightward shift of S'S' is sufficiently large. So far we have shown that the special case of the simple model is theoretically ambiguous. Our aim now is to generalize this model in stages. 3. Endogenizing Industry Size

In the above exercise we assumed industry size to be fixed at n. Let us now suppose, as is often assumed in models of perfect competition with free entry, that there is a very large number of firms and, in equilibrium, industry size gets determined endogenously by using a zero expected profit condition. To avoid some trivial equilibria with zero production, we shall make a small modification to the function f (L). We assume that there exists some positive number, K, such that, for all L ≤ K , f ( L) = 0 , and for all L > K, f '(L) > 0, f"(L) < 0. In other words, the total product (TP) curve, f (L), the marginal product (MP) curve, f'(L), and the average product (AP) curve, f (L)/L, looks as in Figure 3 below. ƒ(L) wL MP TP

w AP

K

Labor

Labor Figure 3.

192

Kaushik Basu, Gary S. Fields and Shub Debgupta

Now define, as before, g ( w) ≡ arg max[ f ( L) − wL]. ( L)

How do the equilibria in the F and N regimes compare? Maintaining the assumption of free entry of firms, consider first a freeretrenchment (F) regime. If the wage is w, in the F-regime each firm earns a profit of f(g(w)) – wg(w). New firms will keep entering the industry as long as the above expression is greater than zero. Likewise if the expected profit is less than zero we do not have an equilibrium since firms will be exiting from the industry. In other words, in an F-regime, an equilibrium wage is w , where this is defined implicitly by, f(g( w )) – w g( w ) = 0. w is illustrated in Figures 3 and 4. Now consider the no-retrenchment (N) regime. A firm in an N-regime facing wage w, expects to earn

pf(L) – wL if it hires L workers. As Appendix A shows, this can be justified in a limiting sense even in a dynamic model. Define A( w) ≡ arg max[ pf ( L) − wL]. L

It should be evident from the previous section that this function is the same as g(w/p). If the wage is w in an N-regime each firm expects to earn a profit of: pf (A( w)) − wA( w).

Hence, using a justification as above, we find that in an N-regime, is an equilibrium wage if: pf (A( w)) − wA( w) = 0.

Labor Retrenchment Laws and Their Effect on Wages and Employment

193

We now have the following: Claim: w < w and A( w) = L( w) To prove this claim observe that to define w as opposed to w we simply have to pretend that the production function f(L) is instead pf(L). Whatever we do with f(L) to get to w , we have to do with pf(L) to get to w . It is obvious from Figure 3 that

w = max[f (L)/L] Hence,

w = max[ pf ( L) / L]. Since p < 1, it follows that w < w . Clearly, the value of L that maximizes

f (L)/L also maximizes pf (L)/L. Hence, A( w) = g ( w) . This establishes the above claim. In Figure 4, let w and w be as shown and let s(w) be the aggregate supply curve of labor. Hence aggregate employment in the F-regime will be LF and in the N-regime LN as shown.

Figure 4.

194

Kaushik Basu, Gary S. Fields and Shub Debgupta

What happens to industry size as we switch regimes from N to F? The number of firms in the N and F regimes are given by, respectively, LN / A( w) and LF /g (w) . Since by the above claim A( w) = g ( w) and given LF > LN , it follows that LN LF > . g ( w) A( w)

Therefore, as expected, the industry is larger under the F-regime, and so too are wages and employment. Therefore, in a country in which, for reasons of law, institutions or technology, there is free entry into industries, a noretrenchment regime is unequivocally worse for the workers. 4. Model with Worker Compensation and Optimization over Time 4.1. The Case of No Frictional Unemployment

In this section we generalize the preceding model by assuming that, as in reality, the government does not ban retrenchments nor make them totally free, but instead insists that when a firm lays off a worker, it makes a severance payment of s to the worker. In this model we also make the inter-temporal decision of the firm explicit. Suppose a firm, after one or more good years, finds itself in a bad year. It can then retrench workers by paying them s or it can hold on to the workers in the hope that a good year will come up soon and thereby save the cost of the severance payment. This is the decision problem that the firm has to solve through dynamic optimization. As before, we shall assume that there are n identical firms each facing a production function as in equation (1). As before, a 'good year' is one in which ϕ = 1 and a 'bad year' one in which ϕ = 0. Let us denote the representative firm's discount factor by δ = 1/(1 + r), where 0 < δ < 1. Now, let L and M be the number of workers that a firm employs in a good year and in a bad year, respectively. Denoting the present value of a firm's profits, starting from a good year, by G and the present value of a firm's profits, starting from a bad year, by B, we get the following two equations G = f ( L) − wL + δ [ pG + (1 − p){− s ( L − M ) + B}]

(12)

B = − wM + δ [ pG + (1 − p ) B]

(13)

Labor Retrenchment Laws and Their Effect on Wages and Employment

195

To understand G, note that if we start with a good year, then the firm's profit in that year is f(L) – wL. In the following year, if the firm is lucky and has another good year (which occurs with probability p) then from that year on, the firm's expected present value is G. This explains the third term on the righthand side of equation (12), δ pG. Next note that if the next year is bad, then the firm spends s(L-M) in severance payments and thereafter expects the payoff of B. Hence the last term, δ (1-p){-s(L – M) + B}. Equation (13) is likewise easy to interpret, remembering that profit in a bad year is –wM, because the firm produces nothing, but must pay for the wages.g Solving (13) for B we have, B = [δ pG − wM ]/[1 − δ (1 − p)].

Inserting this in equation (12) and rearranging terms, we get

δ (1 − p ) ⎡1 − δ (1 − p) ⎤ G=⎢ [ f ( L) − wL − δ (1 − p ) sL] + [ s (1 − δ (1 − p )) − w]M (14) ⎥ 1−δ ⎣ 1− δ ⎦ Let us assume that a firm comes into existence in a good year.h Hence, the firm's problem is to choose L and M to maximize G subject to 0 ≤ M ≤ L . From equation (14), it is clear that the relation between G and M is affine. Hence, it is obvious that the firm will set M = 0, if s(1-δ (1-p)) < w and it will set M = L, if s(1-δ (1-p)) > w. Suppose now s(1-δ(1-p)) < w. From the above paragraph it is clear that the firm will choose M = 0. Inserting this in (14), we can work out the firm's first∂G order condition for maximizing G, = 0 , to be as follows: ∂L f '( L) = w + δ (1 − p ) s.

g

(15)

In deriving these expressions it is important to treat the firm's decision to employ L and M workers in, respectively, the good and bad years as given. Otherwise, if the very first year is a bad one and M is positive the firm will be tempted to not employ any one in that year and then onwards employ L and M depending on if it is a good year or a bad year. If we want to give the firm this leeway, then we will need to distinguish between two kinds of B, namely, B': the present value of the firm's profits starting from a bad year, which is however not the first year in the firm's life and B": the present value of the firm's profits starting from a bad year, which is also the first year in the firm's life. B" looks like B in equation (13) but without the –wM term. Since the B that appears in (12) will have to be B', our analysis here remains unaffected by introducing this additional complication. h We are making this assumption purely for reasons of simplicity. It is easy to alter this and check that the essential results remain unchanged if we abandon this assumption.

196

Kaushik Basu, Gary S. Fields and Shub Debgupta

Next suppose s(1-δ (1-p)) > w. Then the firm sets M = L. Inserting this in ∂G (14) and working out the first-order condition, = 0, we get ∂L f '( L) =

w 1 − δ (1 − p)

(16)

Finally, when s(1-δ (1-p)) = w, M can be any value between 0 and L and ∂G = 0 turns out to be the same as (15). ∂L To sum up, the quantities of labor demanded by firms in good and bad years (that is, respectively, L and M) are given as follows. [s(1-δ (1-p)) < w] ⇒ [M = 0 and L is chosen so that f '(L) = w + δ (1-p)s]

(17)

[s(1-δ (1-p)) > w] ⇒ [M = L and L is chosen so that f '(L) = w/{1 - δ (1-p)}] (18) [s(1-δ (1-p)) = w] ⇒ [M ∈ [0,L] and L is chosen so that f '(L) = w + δ (1-p)s] (19) Equations (17), (18) and (19) tell us what each firm's demand for labor in a good and bad year will be, given the wage, w, and severance payment, s. Our next task is to solve for w and check the effect of varying s on the equilibrium wage and employment.. Let us write the solution from equations (17) – (19) as L(w,s) and M(w,s). We will treat these as functions though we know that in some non-generic special cases M(w,s) is non-unique. The aggregate demand for labor in any single period is

pnL(w,s) + (1-p)nM(w,s). For simplicity, let us take labor supply to be perfectly inelastic at N. That is, no matter what the wage, the supply of labor equals N. In equilibrium w must be such that aggregate labor demand equals aggregate labor supply:

pnL(w,s) + (1-p)nM(w,s)= N.

(20)

We can now solve (20) to get the equilibrium wage as a function of the severance payment:

w = w(s), the properties of which are derived below.

Labor Retrenchment Laws and Their Effect on Wages and Employment

197

In Figure 5, let ABC be the marginal product of labor curve, f '(L). MC

Z

A w/[1-δ(1-p)]

w + δ(1-p)s s

Y

B

δ(1-p)s

C Wage, w

s[1-δ(1-p)]

0

Figure 5

The marginal cost of labor, MC, is given by the last terms in (17) – (19): ⎧ w + δ (1 − p ) s, MC = ⎨ ⎩ w /{1 − δ (1 − p )},

if [1 − δ (1 − p)]s ≤ w otherwise

Given the marginal product function, in a good year a representative firm demands labor up to the point where marginal product equals marginal cost. Hence, the good year demand for labor, L, by a representative firm as a function of MC is given by the line ABC. To analyze the firm's demand for labor in a bad year, first mark the severance payment, s, on the y-axis. This is the amount the firm will have to pay each worker it lays off. Note that, if s < MC, it is cheaper to pay the severance payment than employ existing workers, and therefore M = 0. This is shown as the dotted line sA. However, if s > MC, the workers will be retained

198

Kaushik Basu, Gary S. Fields and Shub Debgupta

as it is more expensive to layoff workers than hiring them this period, therefore M = L. Hence the bad year demand for labor, M, as a function of MC is given by the line AsBC. The left-hand panel in the same figure shows how the wage, w, translates into MC. Given a wage, w, the MC may be read off by moving up from w to the line marked ZYO and then going horizontally to the y-axis. To see why MC = s where the two curves intersect, note that at the point of intersection w = w + δ (1 − p ) s. 1 − δ (1 − p) It is easy to juggle this equation and show that when this equation holds, s must be equal to the left-hand term (and therefore also the right-hand term). Hence s = MC. It is also worth taking note the Z-curve is a vertical shift of the 45o line from the origin and as s increases the Z-curve moves up. The aggregate demand curve for labor is easily constructed from the righthand panel of Figure 5 and is shown in the right-hand panel of Figure 6. In Figure 6, ab is the pn-times horizontal blow-up of the line ABC in Figure 5, and ac is the n times horizontal blow-up of ABC. The aggregate demand depends on the size of the severance payment. If it is s , then the aggregate labor demand is adec. It it is s then it is aghc. The lines marked X and Z are the same as the two lines in the left-hand panel of Figure 5. Recall that the line Z moves down as s becomes smaller and coincides with the 45o ray when s = 0 Superimpose the aggregate supply curve in this figure. This is shown as the vertical line through N. The relation between s and w is now easily read off Figure 6. Clearly if s = s , the equilibrium wage is w . If s = s , the

equilibrium wage is w . As s rises from s to s , the wage rises from w to w . To understand this observe that the aggregate demand curve for labor makes a discontinuous jump from the ab curve to the ac curve at MC = s. Hence, as long as s happens to be between s and s , supply equals demand exactly at MC = s. Since the X and Z intersect at a height of s and X is unchanged as s changes, for all s between s and s , equilibrium wage must be between w and w . If s rises above s , the wage remains fixed at w since supply equals demand at MC = w . If s falls below s , the equilibrium MC clearly remains at s , but the equilibrium wage rises. Thus if s = 0, equilibrium occurs at g, MC = s and the wage = wo. It follows that the w(s) curve looks as in Figure 7.

Labor Retrenchment Laws and Their Effect on Wages and Employment

199

Figure 6.

Wage,w

wo

w

w

s

s Figure 7.

s

200

Kaushik Basu, Gary S. Fields and Shub Debgupta

Whether wo exceeds w or not depends on the parameters of the model. But as Figure 7 shows, if the severance payment is zero the wage is quite high. As s rises, w falls. But after a certain point as s rises, w rises. Hence, the response of the equilibrium wage, w, to the statutory severance payment, s, is Vshaped or, more vividly and depending a bit on your hand-writing, lower-caseb-shaped. Intuitively, what is happening is this. When s is very low, firms retrench workers and make the severance payment whenever they do not need them. So, if s rises, w must fall to ensure that aggregate labor demand remains unchanged. Aggregate labor demand has to be unchanged in equilibrium because labor supply is perfectly inelastic. As s continues to rise, beyond a point, firms become indifferent between retrenching and holding on to their workers. For further rises in s, w must rise to keep firms indifferent between retrenching and not retrenching workers. Finally, when s becomes sufficiently large, there is no retrenchment; so further increases in s have no consequence; and w remains unchanged. 4.2. The Case of Frictional Unemployment

Up to now we did not discuss who gets the severance payment. We could either assume that government collects it as a kind of tax or the worker gets it as compensation. If it is the latter, then being dismissed is always desirable from a worker's point of view, since equilibrium always being at full employment, the worker gets immediately absorbed by the labor market and, in addition, he gets the severance payment. To avoid such a phenomenon, let us introduce some frictional unemployment in the model. Assume that once a worker is laid-off, he needs to spend one period searching before he finds a new job. Let E be the number of people employed in each period. Given the assumption that frictional unemployment lasts one period, it follows that if (1-p) workers get retrenched in each period, then (1-p)E people are unemployed (and searching for new jobs) in each period. Since N is the total labor force, it must be that N = E + (1-p)E, from which we have E = N /(2 − p).

(21)

Labor Retrenchment Laws and Their Effect on Wages and Employment

201

Therefore, if s < w/{1-δ(1-p)}, from (17), we know that the fraction (1-p) of the labor force gets retrenched in each period. Hence, (21) holds in that case. To see what the equilibrium looks like in that case, suppose s = s (in Fig. 6 and 8). By (21), E = N/(2-p). Draw a vertical line at N/(2-p) as shown in Figure 8. MC

w/[1-δ(1-p)] MC*

k

g

s

Wage,w

N/(2-p)

w*

N

Figure 8.

Equilibrium, in the figure, occurs at k. Hence the equilibrium wage is w* and the total unemployment in the economy is N – N/(2-p) = N(1-p)/(2-p). If s is lowered below s , unemployment will remain unchanged, but the equilibrium wage will rise. The reader may also note that once we allow for frictional unemployment of the kind just described we may have a model with multiple equilibria: one with low wage, full employment, and another with high wage, frictional unemployment and retrenchment.

202

Kaushik Basu, Gary S. Fields and Shub Debgupta

5. Conclusion

In the context of the recent protests in France, against its government’s effort to make hiring and firing workers easier—or adopting what is often, euphemistically, called ‘flexible’ labor laws—and in the context of the ongoing debate in India about repealing some of its strongly protective labor laws—it has been argued in popular discourses that flexible labor laws could help the economy but at the cost of worker welfare, especially the welfare of organized labor. What this paper tries to show is that flexible labor laws can be in the interest of all laborers—that is, easier firing rules can increase employment and wages of all workers. The word ‘can’ is however important, because this need not always be so. There are contexts where the adoption of more flexible laws can hurt workers. This theoretical finding shifts our main task to determining under what conditions workers are better off with less protection. A large part of this paper was devoted to isolating the conditions. It was shown, for instance, that all workers would be better off with more flexible labor laws if entry of new firms into industry happened to be easy. Our model also demonstrates that in a legal regime where the government exogenously specifies the compensation to be paid to a worker who is being retrenched, the relation between the size of this compensation and the equilibrium wage rate is V-shaped. Thus, paradoxically, beyond a certain level, further increases in the exogenously specified level of compensation increases labor welfare. This theoretical investigation paves the way for empirical work and also further analytical research. Among extensions that will be interesting to pursue are the incorporation of stochastic shocks which are correlated across firms, and so have industry-wide effects; strategic interactions between governments, firms and unions; and the effects of partial unionization. In deciding whether a particular country will be better off relaxing or making more stringent it’s rules of firing workers, it will be important to do empirical work to decide if the nation falls into categories, identified in the paper, where we know how the law could affect wages and employment and, hence, labor welfare. Acknowledgements

The authors would like to thank Richard Freeman, Garance Genicot, Jeff Hammer, Costas Meghir, Yew-Kwang Ng, Martin Rama, and Tridip Ray for helpful comments and suggestions. The paper also benefited from seminar presentations at the World Bank, Washington, D.C.; ILO, Geneva; Presidency College, Calcutta; and WISE, Xiamen University, Xiamen. A part of the work

Labor Retrenchment Laws and Their Effect on Wages and Employment

203

on this paper was done when Kaushik Basu was visiting the Indian Statistical Institute, Appendix A

This appendix demonstrates the claim made in Section 2 concerning aggregate labor demand in an N-regime with an infinite horizon. Suppose a firm takes the decision concerning how much labor to employ only after the value of ϕ has been revealed. Hence, there are two alternative scenarios for which we have to solve the firm's problem: ϕ = 1 or ϕ = 0. Let the firm's discount factor be a constant δ. If the firm employs L units of labor in each period, the present value of its profits will be [ϕ f ( L) + δ pf ( L) + δ 2 pf ( L) + ...] − [ wL + δ wL + δ 2 wL + ...] = ϕ f ( L) − pf ( L) + pf ( L) /(1 − δ ) − wL /(1 − δ ).

Maximizing this with respect to the choice of L, we get the following first order condition f '( L) = w /[ϕ − p )(1 − δ ) + p ].

(A.1)

If we wanted to work out the N-regime case in this manner we would have to replace (5) with (A.1) and proceed with the exercise in Section 2 with this replacement. Note however that a law which prevents retrenchment makes significant difference only if the firm's discount factor, δ, is large. If the next period is unimportant to the firm (i.e. δ = 0) then clearly the firm's demand for labor is unaffected by whether regime F or N is in place. To study the effect of a regime switch in its extremity we therefore need to consider large δ. Let us therefore consider the case where δ → 1. In that case the denominator in (A.1) goes to p: (ϕ - p)(1-δ) + p → p, as δ → 1. Therefore in the limit (as δ → 1), (A.1) becomes f '( L) = w / p.

204

Kaushik Basu, Gary S. Fields and Shub Debgupta

But this is exactly equation (5). Hence (6) can be thought of as an expression for labor demand in the N-regime, if there were virtually no discounting of the future. References

1. Patricia M..Anderson, and Bruce D. Meyer, 'The Effects of Unemployment Insurance Taxes and Benefits on Layoffs Using Firm and Individual Data', National Bureau of Economic Research Working Paper, no. 4960(1994). 2. Pranab, Bardhan, Quarterly Journal of Economics, Vol. 98, no. 3, 501(1983) 3. Kaushik Basu, ‘Labor Laws and Labor Welfare in the Context of the Indian Experience,’ in A. de Janvry and R. Kanbur (eds.), Poverty, Inequality and Development: Essays in Honor of Erik Thorbecke, New York: Springer(2006). 4. Tim Besley, and Robin Burgess, Quarterly Journal of Economics, 119 (1), 91(2004). 5. Samuel Bentolila, and Guiseppe Bertola, Review of Economic Studies, Vol. 57, 61(1990). 6. Mrinal Datta Chaudhuri, 'Labour Markets as Social Institutions in India', CDE Working Paper No. 16, Delhi School of Economics (1994). 7. Mahendra S Dev, Economic and Political Weekly, vol. 35, Jan 8 and 15 (2000). 8. Jean Dreze, and Anindita Mukherjee, 'Labor Contracts in Rural India: Theories and Evidence, in: S. Chakravarty ed., The Balance Between Industry and Agriculture in Economic Development: Proceedings of the Eighth World Congress of the International Economic Association, Delhi, India, Vol. 3 St. Martin's Press, New York(1989). 9. Gus Edgren, ed., Restructuring Employment and Industrial Relations: Adjustment Issues in Asian Enterprises, I.L.O-ARTEP, New Delhi (1989). 10. Peter R Fallon, and Robert E.B. Lucas, The World Bank Economic Review, Vol. 5, no. 3, 395 (1991). 11. Peter R.Fallon, and Robert E.B. Lucas, Journal of Development Economics, Vol. 40, 241(1993),. 12. Ajit Ghose, Indian Journal of Labor Economics, vol. 37 (1994),. 13. Hugo Hopenhayn, and Richard Rogerson, Journal of Political Economy, Vol. 101, no. 5, 915(1993). 14. K.P Kannan,., Economic and Political Weekly, Vol. 29, July 23 (1994). 15. Robert E.B.Lucas, , 'India's Industrial Policy', in: Robert E.B. Lucas and Gustav F. Papanek, eds., The Indian Economy, Westview Press, Boulder, Colorado (1988).

Labor Retrenchment Laws and Their Effect on Wages and Employment

205

16. Ajeet N.Mathur, 'The Effects of Legal and Contractual Regulations in Employment in Indian Industry', in: Gus Edgren, ed., Restructuring Employment and Industrial Relations: Adjustment Issues in Asian Enterprises, ILO – ARTEP, New Delhi (1989), 17. Xin Meng, Labour Market Reform in China, Cambridge University Press, Cambridge, U.K(2000). 18. T.S. Papola, Indian Journal of Labor Economics, vol. 37(1994). 19. Martin Rama, World Bank Economic Review, vol. 13, 1(1999). 20. Debraj Ray, Development Economics, Princeton University Press, Princeton, New Jersey(1998).

This page intentionally left blank

April 20, 2009

15:11

WSPC - Proceedings Trim Size: 9in x 6in

gmansuri

WORK MIGRATION AND INVESTMENT IN ORIGIN COMMUNITIES∗ GHAZALA MANSURI Development Research Group, The World Bank 1818 H St. NW, Washington DC 20433 Email: [email protected]

1. Introduction Recent debates on “feasible globalization” have focused on the importance of opening up international labor markets to low skill guest workers from developing countries. It is argued that income gains from such a liberalization of labor markets would be large and could contribute significantly to a reduction in inequalities of wealth and opportunity, both within and across countries. Key to this is the expectation that migration will fuel private investments in both physical and human capital in origin communities. This chapter focuses on the impact of work migration on investment in the human capital of children in sending communities.a The data come from rural Pakistan where temporary work migration is substantial, with more than one in four rural households reporting at least one migrant. Migrants, almost exclusively young males, retain close ties to their origin households by sending substantial and regular remittances and returning home frequently.b Migration is also undertaken mainly by relatively poor and low skill households who would otherwise tend to be credit constrained and have considerable exposure to uninsured income risk which could dampen incentives to invest in the human capital of children. Indeed, most human development indicators point to low investments in children. Almost a third of rural children suffer from severe stunting, school enrollment and retention rates are low and one-third of children are active in the labor market. ∗ The views expressed in this paper are those of the author and should not necessarily be attributed to the World Bank, its executive directors, or the countries they represent. a The results presented in this chapter draw from two papers, Mansuri (2007a)1 and Mansuri (2007b).2 b According to the 1998 census records, some 10 million individuals (almost 8% of the country’s population) were either domestic or international migrants.

207

April 20, 2009

15:11

208

WSPC - Proceedings Trim Size: 9in x 6in

gmansuri

Ghazala Mansuri

At the same time, migration can itself create new constraints which could dampen this potential enhanced investment. For one, migration changes family structure, potentially increase the work burden of those left behind, including older children, and may also induce additional psychological stresses. In rural Pakistan, male absence is likely to create more stringent constraints given prevailing female seclusion practices and low female mobility. On the other hand, there is a good bit of evidence which suggests that gender differences in preferences over the welfare of children can be important. In particular, investments in girls can increase in contexts where mothers exercise greater control over the use of household resources. For human capital investments, therefore, male absence could be a boon since it would tend to augment any direct resource effect of migration by changing the balance of household preferences over child schooling and health. While female headship is relatively uncommon in rural Pakistan, this is not the case with migrant households. Almost one in five such households has no resident adult male and, given the relative youth of most migrants, this is most often the case at a time in the family life cycle when crucial investments in child health and schooling need to be made. In sum, the direct resource effect of migration is likely to be mediated through the changes it induces in the structure of migrant households and their communities. Since these changes can either augment or dampen household incentives to invest, the net effect of migration is likely to depend on the relative strength of these competing forces. Since individuals and households choose to migrate, the main econometric challenge in assessing the impact of migration on child schooling or healthlies in dealing with the endogeneity of the migration decision. One way to address this potential endogeneity problem is to identify selection in the migration decision by using instrumental variables. A number of recent papers have used migration prevalence rates in origin communities as an instrument for the opportunity to migrate. However, any number of unobserved community characteristics, such as local labor market conditions or the availability or quality of health care services, could affect both household investment decisions and the propensity to migrate. What is really needed therefore is an instrument which influences a household’s opportunity to migrate, is uncorrelated with the outcomes of interest and varies across households in a community. The papers summarized in this chapter develop such an instrument. Sections 3 and 4, respectively, present evidence on the impact of migration on household investments in schooling and early childhood growth.

April 20, 2009

15:11

WSPC - Proceedings Trim Size: 9in x 6in

gmansuri

Work Migration and Investment in Origin Communities

209

Migration has a positive and extremely significant impact on height-forage, particularly for girls. Moreover, when the sample is split by age group, this height advantage is sustained in the older age group underscoring the long term salutary benefits of averting nutritional and other health shocks in early childhood. Estimation using child weight-for-age z-scores yields similar results. The results also suggest that, as expected, selection into migration as well as community level unobservables are quite important. Migrant households also make substantially larger investments in education. Children in migrant households are not only more likely to attend school, they are also more likely to stay in school and accumulate more years of schooling in comparison to their counterparts in non-migrant households in the same village. They are also less likely to be involved in economic work and report working for substantially fewer hours. I also find support for large gender differentials in the gains from migration, with relative gains for girls outstripping those for boys by a good margin, resulting in a substantial net reduction in gender inequalities in access to education. 2. Data and Context 2.1. Data The studies discussed in this chapter use data from the Pakistan Rural Household Survey (PRHS) 2001-02, a nationally representative multipurpose household survey. The survey contains a separate migration module which collected information on the migration history of each household member.c For migrants, data was also collected on the year and duration of migration, migration destination, remittances, and social networks accessed prior to and post migration. Migrants were interviewed directly when possible. Otherwise, the individual designated as the male head of the household reported migration and other information for each migrant. Information on remittances received by the household over the 12 months preceding the survey was collected in a separate module on transfers to and from the household. In addition, the survey contains detailed household and individual characteristics, including demographics, occupation, health, education, investment in farm and non-farm assets, agricultural production and household expenditure. Complete data is availc In

the PRHS, individuals who were away from the household at the time of the survey, were classified as households members, provided they were regarded as members of the household before they left and had not set up a permanant home elsewhere. This enabled collection of all relevant data on current migrants.

April 20, 2009

15:11

210

WSPC - Proceedings Trim Size: 9in x 6in

gmansuri

Ghazala Mansuri

able for 2531 rural households in 143 villages in 16 districts across all 4 provinces. For purposes of the analysis attention is confined to male migrants age 18 or older who migrated for economic reasons.d Using this definition, 977 men (about 15% of all men in this age range) are classified as migrants. There is detailed data on 816 of these for whom the migration section was completed. Of these, 32% were back from a migration episode in the survey year, the rest were current migrants. At the household level, 691 households (25% of all households) had at least one male migrant and in about 40% of cases the migrant had returned and was in the household for much of the survey recall period. Information on migration was also collected at the village level by conducting a brief census in all 143 sample villages. The census asked each household whether any household member was currently living and working outside the village. This definition thus includes both domestic and international migrants. Using this definition, 14% of households report a current migrant.e

2.2. Migrant Characteristics, Migration Destination and Return Migration Since the focus of this work is on the impact of migration on the sending community, permanent out-migrants will tend to be excluded from the analysis by definition. This is partly due to the inability of a household survey in the sending community to trace families who had permanently left the village by the time of the survey. Second, if individual members of a household had left the village and set up permanent homes elsewhere, they were not considered household members, and are therefore also excluded from the survey. Since the main channel through which such households or individuals could impact the outcomes of interest would be transfers to

d There

is virtually no migration among children under 18. The few who do not live at home move to join a family member or to attend school in a neighboring rural area. Women also typically migrate to join family members, most often a spouse. While 8% of reported migrants are women, over 82% report migrating to join a family member. Only 13 women (1% of the sample of migrants) report migrating for any economic reason. e Note that this is significantly below the migration incidence we get from the household survey since the latter is not restricted to current migrants. It is worth noting that the number of households in the sample with a current migrant is just above 13%, as we would expect from the census.

April 20, 2009

15:11

WSPC - Proceedings Trim Size: 9in x 6in

gmansuri

Work Migration and Investment in Origin Communities

211

their extended families who remain behind, it is important to note that such transfers are in fact negligible.f The median age at first migration in the sample is 22 and over twothirds of migrants in the sample are married and have their spouses and/or children living in the village. In fact, the typical migrant is either a household head (38%) or an older son of the head (54%). This is suggestive of the extent to which migrants are attached to their families of origin in the villages. Using survey data on per capita expenditures or a measure of wealth based on an asset index, migrants households are significantly wealthier than non-migrants. They also have almost 3 additional grades of schooling than non-migrants. However, they are significantly less likely to own a non-farm business (11% compared to 14% for non-migrant households) or be engaged in agricultural production (53% as compared to 60% for nonmigrants). They also own significantly less land than non-migrants. The only available measure of wealth that is not correlated with migration is a household’s inherited land wealth. This indicates that migrants are drawn largely from the lower end of the wealth distribution. This is particularly so for international migrants. However, they are also much more likely to be active in the market for land.g On average, migrants leave their communities thrice and stay away for a period of 5 years each time.h Close to a third had migrated to international destinations, primarily in the middle east. Of the rest, the majority (over 80%) had moved to an urban destination within the country. The rest migrated to another rural area. The importance of networks in determining the migration decision is evident in the sharp regional differences observed in the destination of migration as well as the variation in the extent of migration experience across the country. Migrants from the North-West Frontier Province (NWFP) and Baluchistan, for example, are much more likely to migrate abroad. However, f See

footnote 14 on this. are, however, significantly more active in the land market. Interestingly, most of this is due to land purchase by international migrants. Domestic migrants are about as active in the land market as non-migrant households. h Using data on completed migration episodes only. If we look at the elapsed duration of ongoing migration episodes instead, the median (elapsed) duration is 5 years and the average (elapsed) duration is closer to 8 years.In the survey, those who reported having migrated more than once (a third of the sample), were asked to provide details of their longest migration episode. All others, including the seasonal migrants (4%), were asked to report the details of their first migration. g Migrants

April 20, 2009

15:11

212

WSPC - Proceedings Trim Size: 9in x 6in

gmansuri

Ghazala Mansuri

while most international migrants from NWFP go to Saudi Arabia (the principal destination for migrants in the sample), their counterparts from Baluchistan are more likely to migrate to United Arab Emirates and Oman, which are also closer to the coast of Baluchistan. For domestic migrants, the port city of Karachi is by far the main urban destination, however, domestic rural migrants tend to migrate largely within their own regions of origin (appendix Table A1). At the provincial level, NWFP is, by far, the region with the most migration experience. The average proportion of migrant households per village is more than one-third, as compared to the national average of 14%. However, there is significant variation in migration experience, at the village level, within all four provinces In NWFP, for example, the percent of migrant households range from a low of 9% to a high of 78%. The range in Punjab is from 0 to 62%. This variation makes it possible to identify differences in migration opportunities at the village level and below. The importance of migrant networks at destination is also clearly evident in the sample. More than two-thirds of migrants, report knowing someone in their place of migration before they left. Almost 70% also report living with someone they knew when they first arrived at their migration destination. Most migration also appears to be legal and a substantial number of migrants report moving with jobs in hand. Over one-half report that their first job was arranged by a family member, a friend, or a member of their clan (biradari). In contrast, only 7% acquired their first job through a recruiting agent. Almost one-half of all international migrants, and almost three-fourths of internal migrants, also report starting work almost immediately upon arrival in the host country/region and almost all (98%) succeeded in finding a job soon after migration. While the data on migration costs is noisy, it suggests a very wide variation in costs by migration destination. International migrants report spending almost 9 times as much as domestic migrants. Average reported migration costs were about Rs. 70,000 (US$1,250) i for international migrants, but only Rs. 8,122 (US$145) for domestic migrants. It is unclear, however, whether migration destination is a function of pre-migration household wealth. International migrants are, not surprisingly, much wealthier than domestic migrants, using either per capita expenditures or an asset index. However, there is no difference in the inherited land holdings of domestic and international migrants. There is also no difference in the average i Using

an exchange rate of Rs.56 for US$1.

April 20, 2009

15:11

WSPC - Proceedings Trim Size: 9in x 6in

gmansuri

Work Migration and Investment in Origin Communities

213

educational level of international and domestic migrants. j Migration destination also does not appear to have a significant effect on the duration of a migration episode. Focusing on completed migration spells, international migrants were away almost 2 more years, on average, but the difference is only weakly significant (p-value < 0.10). More than two-thirds of migrants report sending remittances to their families in the village and three-fourths of those who sent remittances did so on a regular basis. k Migrants are more likely to remit if they are household heads, are married and have children. They are also more likely to remit if they migrate abroad and are more educated.l On average, remittance receiving households get Rs. 46,000 (about $840) annually, of which about 63% is sent regularly. Not surprisingly, international migrants send much larger remittances than domestic migrants ( (Rs. 84,000 ($1500) as compared to Rs.21,000 ($382) on average).m

2.3. Community Migration Prevalence Rates and Selection into Migration The main econometric challenge is examining the impact of migration on household behavior in an origin community lies in dealing with the endogeneity of the migration decision. Individuals and households choose to migrate and many of the same characteristics which influence the decision to migrate are likely to also affect other household decisions such as how much capital investment to undertake or how much to invest in the education or health of children. One way to address this potential endogeneity problem is to identify selection in the migration decision by using instrumental variables. A number of recent papers (including the three discussed in this chapter) have used migration prevalence rates in the origin communities as an instrument for the opportunity to migrate. j However, households with international migrants have about 1.6 grades more of schooling overall. k Remittances from international migrants constitute the single largest source of foreign exchange earnings for the country. According to one estimate, US$2.4 billion (or 4% of the country’s GNP) is currently remitted annually by international migrants (see Gazdar (2003)). l One-half of all remittance senders are located abroad. m It is worth noting that remittance income, at the household level, comes almost entirely (95%) from family members working outside the village.

April 20, 2009

15:11

214

WSPC - Proceedings Trim Size: 9in x 6in

gmansuri

Ghazala Mansuri

There is a large literature in labor economics that has documented the role of informal networks in securing employment (see, for example, the review of this literature by Montgomery (1991)). For workers seeking employment in unknown and distant labor markets, information problems are likely to be of even greater significance, increasing the value of such informal networks (Borjas (1992),3 Boyd (1989),4 Menjivar (1995)5 ). Migrant networks are seen as reducing the costs of migration for potential migrants via two channels. First, they constitute an information network which can educate potential migrants about conditions in specific migration destinations as well as potential hazards and costs, both at home and in the migration destination (Massey et al (1993),6 Orrenius (1999)7 ). Second, they serve to relax credit constraints (Genicot and Senesky (2004)8 ). A number of studies have also shown that networks increase the economic returns to migration (see, for example, Munshi (2003)9 ). All of this implies that the probability of migration should be higher for households residing in communities with significant migration experience and this is indeed borne out in empirical studies.n However, any number of unobserved community characteristics, such as local labor market conditions or the availability or quality of health care services, could affect both household investment decisions and the propensity to migrate.o What is really needed therefore is an instrument that varies across households within a community so that the effect of community level unobservables can be removed. The studies discussed in this chapter obtain within village variation in the instrument in two ways. First, data from the village census on household land ownership is used to construct migration prevalence rates within land holding groups in the village. Access to migrant networks is expected to vary significantly across landholding groups within a village since inequalities in land ownership are pervasive (the mean land gini at the village level is 0.75), n Winters et. al. (2001)10 show that the probability of migration to the United States is higher for households living in Mexican communities which have greater experience with migration. Banerjee (1991)11 and Caces(1986) have shown the importance of networks in the rural-urban migration decision in the Indian and Philippine context, respectively. Ilahi (1999), has shown the importance of extended family networks in financing migration costs in Pakistan. o For example, Alderman and Garcia (1994)12 among others have shown that the estimated impact of income and education on child health status will be biased if community level unobservables are ignored. More significantly, they find that it is the quality of health care services that matters the most, as opposed to the more commonly measured health service availability or distance to a health care facility.

April 20, 2009

15:11

WSPC - Proceedings Trim Size: 9in x 6in

gmansuri

Work Migration and Investment in Origin Communities

215

and access to land is an important marker of social and political status. This is also evident in village settlement patterns. Land is also acquired primarily via inheritance, or a gift from parents during their lifetime, so a substantial change in land distribution due to migration is not likely.p Second, I use a feature of migration that is particular to the context under study to obtain household level variation in this instrument. Mobility and seclusion restrictions on women typically require the presence of an adult male in the household.q Households with a single adult male are therefore much less likely to undertake migration, and among households with more than one male, young men are more likely to migrate if the male they would leave behind is an older man.r Sections 3 and 4 describe the instruments used in each case in detail. The basic identification argument is that the size of the migrant network within land holding groups in the village, interacted with the number of adult males in the household, should affect a household’s opportunity to send a migrant but is unlikely to be correlated with unobservable household or child attributes that affect the costs or returns to investment in physical or human capital. In all cases, village fixed effects are used to clean out the potential impact of any time invariant unobserved community characteristics that could be correlated with the migration prevalence rate. All specifications also include a set of exogenous household characteristics, including in particular, the household’s own inherited land holdings and a set of variables that capture household composition. Of course, migration could also affect other aspects of household composition, including decisions regarding fertility and coresidence in extended families.s If so, it may not be correct to think of family size and composition as exogenous in examining the impact of male migration on investment in productive assets or in the health and education of children. It is therefore useful to verify if there is need to be concerned about the endogeneity of any household demographic variables before turning to the main results. p In

the PRHS 2001-02, for example, 83% of landowning households report that all of their land is inherited. However, migrants are significantly more likely to purchase land, as we show below. q All males age 18 and up are classified as adult. r In the papers, I show that, conditional on appropriate household demographic characteristics, the number of adult males exercises no additional influence on household decisions related to any child outcomes of interest. s If, for example, households where males intend to migrate are more likely to form extended families, or household fertility choices affect the health and education outcomes of children.

April 20, 2009

15:11

216

WSPC - Proceedings Trim Size: 9in x 6in

gmansuri

Ghazala Mansuri

2.4. Migration and Household Structure Table 1 presents the results for five measures of household structure: size, coresidence of more than one nuclear family, the household’s dependency ratio, the number of adult males and female headship. The first row presents the OLS coefficient of the migration effect. Two variables which capture the household’s pre-migration wealth, its inherited land holdings and the maximum years of schooling of adults in the household, are included as controls. Regional differences in family structure that may be correlated with migration prevalence rates are cleaned out using tehsil t fixed effects. Ignoring selection (columns 1, 3, 5, 7 and 9), migrant households are significantly larger, have more adult males, a lower dependency ratio, and are much more likely to have a female head. After instrumenting for migration, however, the only discernible difference between migrants and non-migrants is in the prevalence of female headship, where large differences remain. The instruments for migration are the overall migration prevalence rate in the village and the coefficient of variation of village rainfall. This, along with the earlier discussion of the determinants of migration and return migration, suggests that it is the opportunity to migrate which is influenced by the availability of adult males in the household, and not the other way around, so that households with low male endowments, who live in high migration prevalence areas, are more likely to have female heads. It is useful to note that there are also no discernible effects of migration destination on household structure, once tehsil fixed effects are included. International migrants have only slightly larger households and more adult males, but the difference is not significant. There are also no differences in dependency ratios, the likelihood of a female head, or an extended family structure. International migrants are somewhat more likely to be heads of their households (60% as opposed to 52% for domestic migrants), have a spouse in the household (89% as compared to 71% for domestic migrants) and to have children in the household (83% as compared to 67%), but the differences are insignificant once tehsil fixed effects are included. 3. The Impact of Migration on Child Schooling and Labor Market Activity Low educational attainment in many developing countries has been viewed as arising, at least partly, from barriers to private investment, due to incomt Administrative

unit below district. There are usually 2 to 4 tehsils in a district.

April 20, 2009

15:11

WSPC - Proceedings Trim Size: 9in x 6in

gmansuri

Work Migration and Investment in Origin Communities

217

plete or absent credit markets. Migration, in so far as it releases such resource constraints, is expected to increase human capital investment among the poor. Moreover, where gender differences in educational attainment are large a reduction in resource constraints may also imply a reduction in gender inequalities in access to schooling since investment in the education of girls is more sensitive to household wealth (see figures 1-6 in the Appendix). However, in a context where there is evidence of strong boy preference, as is the case in rural Pakistan, evidence of differentially better outcomes for girls due to migration may be capturing the effect of a rather different process: Male migration often leaves women effectively in charge of their households. If women have more benevolent preferences towards their children, and particularly their daughters, then women’s ability to make decisions regarding the disposal of such resources due to the absence of males may well provide mothers with an opportunity to more easily realize their preferences with regard to investments in their children’s education.u The observe effect, is that economic migration can also be disruptive of family life in any number of ways which could adversely affect school attainment. The absence of adult male labor may, for example, result in greater pressure on children to assist with housework, household production or childcare. In environments where female seclusion practices are important, migration could, as discussed above, generate additional constraints on female mobility which could adversely impact incentives to send girls to school. If these competing effects are important, female heads could face a potential conflict between their desire to invest in their children, and the pressures generated by “male absence” with possible adverse consequences, particularly for the gender allocation of labor and schooling among children. Mansuri (2006a)1 examines the effect of migration on schooling and child labor, focusing specifically on gender differentials in school attainment and labor market activity. The impact of female headship is examined by confining attention to migrant households and asking whether schooling and labor market outcomes vary significantly by female headship and whether there are discernible gender differentials in outcomes.

uA

substantial body of research has identified important gender differences in preferences over the welfare of children and has shown, in particular, that investments in child education increase significantly in contexts where mothers exercise greater control over the use of household resources.

April 20, 2009

15:11

218

WSPC - Proceedings Trim Size: 9in x 6in

gmansuri

Ghazala Mansuri

The impact of migration on school attainment and labor market activity is estimated using the following regression function Sijv = β1 Mijv + β2 Bijv + β3 Bijv Mijv + γ1 Cijv + γ2 Xjv + ηjv + εijv (1) Where Sijv is a measure of school attainment for child i, in household j and village v. Mjv is an indicator of whether the household has a migrant, Bijv is the child’s gender, and Cijv and Xjv are vectors of exogenous child and household characteristics. The mean zero error term ηjv captures the effects of unobserved factors common to a given village and household. The child-specific error term εijv reflects measurement error in the schooling variables and, potentially, unobserved attributes of the child, including innate ability or parental preferences which vary by child gender. The key difference between ηjv and εijv is that while the latter is not likely to be correlated with the migration decision, the former could influence both the decision to migrate and investments in human capital formation. At the village level, ηjv may, for example, include unobserved variation in local labor market conditions or in school quality, while at the household level, it could include preferences over human capital accumulation, access to credit or insurance markets or costs that affect schooling but are not observed in the data. Since the instrument set used (as discussed in section 2.3) varies at the household level, any time invariant characteristics of the village can be differenced out so that attention can be confined to differences in the educational attainment of children in migrant and non-migrant households within each village. Differencing equation 1 across households within a village, yields Sij = β1 Mj + β2 Bij + β3 Bij Mj + γ1 Cij + γ2 Xj + ζij

(2)

where ζij = ηj + εij . As noted in section 3, this village fixed effects estimator is not robust to correlation between Mj and ζij , since the latter contains not just measurement error but also unobserved attributes of the household which could be correlated with household decisions regarding child schooling and labor market activity as well as the decision to migrate. That is, in general, E [ζij |Mj = 1] = 0, and OLS estimates of β1 will still be biased. As before, this potential endogeneity is dealt with by instrumenting for Mj . The principal instrument for Mj in equation - is the relevant within land group migration prevalence rate for each household in a given village (V Mlv ), the number of adult males in the household (N Aj ) and V M Llv interacted

April 20, 2009

15:11

WSPC - Proceedings Trim Size: 9in x 6in

gmansuri

Work Migration and Investment in Origin Communities

219

with N Aj . Additional instruments are constructed by interacting N Aj with the village land gini (V G).v These two instruments are interacted with the child’s gender in order to estimate β3 . The set of exogenous child characteristics, Cij , include the child’s gender, age and age squared, mother’s and father’s level of education (in completed grades), the number of other school age siblings (age 6-17); the number of siblings age 5 or younger, also interacted with the child’s gender in order to to ascertain any gender differences in child care responsibilities, and the presence of an older brother or sister under age 18. The rationale for this last set of variables is that children who have older siblings may be less likely to be removed from school in the event of an unexpected income loss or have fewer household production related responsibilities. The set of exogenous household level characteristics, Xj , includes a further set of demographic controls, specifically, an indicator for whether there is more than one married male with coresident spouse and/or children in the household and the household dependency ratio. It also includes the household’s inherited land holdings (in acres) the main control for household wealth. 3.1. Schooling Most studies that have looked at the impact of migration on child schooling have focused on accumulated schooling as measured in completed grades, regardless of whether the child ever attended school. Where enrollment rates are low, there is substantial school withdrawal and children are actively engaged in the labor market, accumulated schooling is likely to reflect the combined effects of several distinct schooling decisions. This makes it important to disaggregate the impact of migration on accumulated schooling by examining schooling outcomes at both the intensive and extensive margin. The survey contains data on schooling outcomes for 7181 children age 5 to 17 who belong to 2126 households. Of these, 29% belong to migrant households. There is wide variation, in practice, in the age at which children start school. However, very few begin school after age 9. In examining the school enrollment decision and accumulated schooling, therefore, the sample is restricted to children age 10-17 in order to ensure that estimates of enrollment and accumulated schooling are robust to potential late entry. v The

direct effect of any village level observables has already been removed using the village fixed effect and we control for the household’s own land holdings directly in the second stage.

April 20, 2009

15:11

220

WSPC - Proceedings Trim Size: 9in x 6in

gmansuri

Ghazala Mansuri

In looking at retention rates, however, the sample includes only children 10-15, since this is the age group which is most at risk for dropping out of school during the transition point from primary to middle school. Overall, 58% of children age 10-17 report having enrolled in school at some point. Of these, 38% had dropped out of school by the survey year. The bulk of dropouts, over 85%, had dropped out either before or at the end of primary school (grade 5). While these overall rates of enrollment and retention are quite poor, they conceal very large gender differences. In the sample, 58% of girls age 10-17 had never been to school as compared to only 26% for boys in this age range. The picture only worsens when one looks at school retention rates. While only 25% of enrolled boys in the sample had dropped out by the survey year without completing high school, 44% of enrolled girls were no longer in school. Not surprisingly, boys also have significantly more years of schooling, completing an additional half grade more than girls on average (p-val 0.000). The first four rows of table 5 present the estimation results for four measures of school attainment. The first specification in all cases (columns 1-3) presents the OLS estimates of the migration coefficient and its interaction with child gender under the assumption that the migration decision is uncorrelated with unobserved village and household attributes. The second specification (columns 4-6) instruments for migration using the instrument set described above. All specifications include the full set of child and household controls. The migration effect is positive and significant for all schooling outcomes. Children in migrant households are not only more likely to attend school, they are also more likely to stay in school in the age range when school dropout rates peak, and have higher completed grades in their age cohort. Accounting for selection into migration only strengthens these results indicating that the selection bias is negative. The evidence also supports migration induced gender differentials in enrollment rates, school retention and accumulated schooling. Enrollment rates increase by 54% for girls in migrant households (from 0.35 to 0.54). In comparison, being in a migrant household raises enrollment rates for boys by only 7% (from 0.73 to 0.78). As a result, the gender gap in enrollment rates in migrant households is quite a bit smaller (at 0.24 as compared to .39 among non-migrant households). Of course since enrollment rates for boys are much higher to begin with, one might argue that raising the enrollment rates for girls should be easier. Even in this respect, though, the gains for

April 20, 2009

15:11

WSPC - Proceedings Trim Size: 9in x 6in

gmansuri

Work Migration and Investment in Origin Communities

221

girls are larger. The increase in enrollment closes the gap relative to 100% enrollment by 29% for girls, compared to 19% for boys. The decline in dropout rates is also substantially larger for girls. The dropout rate for girls falls by 55% (from 0.56 to 0.25), while it declines by 44% for boys (from 0.25 to 0.14). As a consequence the gender gap in dropout rates falls sharply (from 0.31 to 0.11). Once again this implies that for girls the gap from 0 dropout rates is closed by 70% while for boys this gap is closed by 44%. In contrast, the gender differential in accumulated schooling is rather small. Girls in migrant households have about 1.5 more years of schooling compared to their counterparts in non-migrant households in the same village, while boys have about a grade more. Significantly though, prospects of migration in the future do not appear to exercise any effect on years of schooling for either boys or girls. Once the sample is restricted to children who enrolled in school at some point we get larger differences in accumulated schooling for girls and boys. In fact, comparing children within migrant households, girls actually exceed boys in absolute terms, completing about a fifth of a grade more. The increase in accumulated schooling for girls is also larger than for boys if children in migrant households are compared with children in non-migrant households. While girls in migrant households complete almost 2 grades more than girls in non-migrants households, boys complete a little less than a grade more than boys in non-migrant households. The net effect is that in non-migrant households, boys are almost a full grade ahead of girls in their age cohort, while in migrant households, girls more than make up the gap, exceeding boys in their age cohort by 0.2 grades. The IV results, though strong, could well underestimate the impact of migration on education since schooling investments are time sensitive. Mansuri (2006a)1 compares siblings within a household before and after migration to check this. The results indicate that children who attain school age after the household’s first migration episode do much better than those who are too old to benefit. Since there is no gender difference in the average age in first grade, the evidence suggests either a disproportionate benefit to girls from improved household capacity to bear income risk or a higher incidence of migration generated burdens on boys. We turn to this issue next.

April 20, 2009

15:11

222

WSPC - Proceedings Trim Size: 9in x 6in

gmansuri

Ghazala Mansuri

3.2. Child Labor As one might expect, there is a strong negative correlation between school enrollment, school retention and labor market activity. Children who are enrolled in school are significantly less likely to report work and work fewer days than those who are never enrolled in school and the differences are large, particularly for children between the ages of 10 and 17. Children who drop out of school, are also significantly more likely to report some work and work more days than children who remain in school. This suggests that there could be a trade-off between child labor activity and time spent in school, at least for some households, The survey contains labor market participation information for 5780 children age 7 to 17 who belong to 1992 sample households. There is data on five major categories of work. Work on the family farm, agricultural wage work, work on a family enterprise or home based productive activity, nonfarm wage work and care of livestock. For children up to age 13 there is also information on time spent on fetching firewood and water. Two definitions of work can be constructed using this data. The more restrictive definition (I) , includes only directly income generating activities. It therefore excludes livestock care, and the fetching of firewood and water. The less restrictive definition (II) includes all work. Using the more restrictive definition, 18% of all children in the age group 7-17 report doing some work and among children age 15 and up, more than a third report some work activity. Interestingly, there appears to be little difference in reported work activity by gender. Using an eight hour work day and 30 days of work per month, the median days worked by children who report some labor market activity is 1.3 months over a one year period. The average number of days worked is substantially higher at 2 months since there is a strong positive correlation between age and labor market participation. Again, there are no discernible gender differences in days worked. While boys work 12 more days, on average, per year, this difference arises entirely from children 16 and older where boys work for 24 more days per year than girls (p-value < 0.00). In all other age groups the difference is small and insignificant. If the broader definition of work is used, the labor activity of children age 7-17 rises to 29%. The median number of days worked are close to 2 months and the average number of days worked rises to about 3 months. The principal work activity of children is unpaid work on the family farm. 63% of working children report working on their own family farms.

April 20, 2009

15:11

WSPC - Proceedings Trim Size: 9in x 6in

gmansuri

Work Migration and Investment in Origin Communities

223

However, more than a third (35%) also report working as agricultural wage labor and 13% report working on a family enterprise. In contrast, only 8% report working as non-farm wage labor. There are strong gender differences in the type of work undertaken by male and female children. Boys are much more likely to work on the family farm (68% of boys and 58% of girls: p-value < 0.000), but girls are more likely to work on the household’s non-farm enterprise (16% of girls and 10% of boys: p-value < 0.00). Girls are also more likely to work as agricultural wage workers (51% of girls as compared to only 20% of boys: p-value < 0.0000), while boys are more likely to do non-agricultural wage work (14% of boys but only 2% of girls: p-value < 0.0000). These differences are consistent with other work that has highlighted mobility constraints for girls, particularly after adolescence.w Non-agricultural wage work typically requires travel outside the village, and is often undertaken individually. In contrast, agricultural wage work is typically undertaken jointly with other family members, particularly other adult females. There is also a pronounced age Both measures of labor market activity use the more restrictive definition of work pattern. Wage labor is largely undertaken by children 14 and older in the sample, and this is particularly evident for non-farm wage work, which is almost exclusively done by boys 14 and up. Figures 7-10 (Appendix) depict this age pattern. Girls work more than boys until about age 13 but their labor market activity falls off somewhat after that. Girls also bear a substantially larger burden of work in poorer households. Their participation in labor market activity starts out significantly higher than that of boys, but falls sharply with wealth, and is significantly below that for boys in the highest wealth quintile (see figures 11-14). The last two rows of table 5 present the estimation results for two measures of labor market activity: (1) an indicator variable which equals 1 if the child reported any labor market activity over the survey period, and (2) the number of reported days of work. Both measures of labor market activity use the more restrictive definition of work. The first specification (columns 1-3), as before, presents the OLS estimates of the migration coefficient and its interaction with child gender under the assumption that the migration decision is uncorrelated with unobserved village and household attributes. The second specification (columns 4-6) instruments for migration using the instrument set described above. All specifications include the full set of child and household controls. w See

the Pakistan Country Gender Assessment, The World Bank (2005)

April 20, 2009

15:11

224

WSPC - Proceedings Trim Size: 9in x 6in

gmansuri

Ghazala Mansuri

Migration has a strong dampening effect on child labor market participation, regardless of whether the indicator is any work or reported days worked over the survey year. Overall, days worked fall by about 66% (from 27 to 10 for boys and 27 to 9 for girls). Unlike schooling, however, there are no gender differentials in labor market activity. Interestingly, the presence of older siblings is significant and reduces labor market activity for both boys and girls. Finally, accounting for selection into migration serves to strengthen the migration effect substantially, suggesting, again, that there is significant negative selection into migration. While these results are consistent with the results on schooling, given the strong negative correlation between school attainment and labor market activity, they do not suggest that the poorer performance of boys in school retention or accumulated schooling can be ascribed to increased labor market activity.

4. Migration and Child Health There is considerable epidemiological evidence that very young children are particularly vulnerable to shocks that lead to growth faltering, with substantial long-term health consequences. The main child growth measures used in the paper are weight-for-age (WAZ) and height-for-age (HAZ) zscores. Child height, in particular, is a good indicator of underlying health status and studies have shown that children experiencing slow height growth are found to perform less well in school, score poorly on tests of cognitive function, and have poorer psychomotor skills and fine motor skills. They also tend to have lower activity levels, interact less frequently in their environments and fail to acquire skills at normal rates. Studies have also shown that taller women experience fewer complications during child birth, have children with higher birth weights and face lower risks of child and maternal mortality. Growth faltering in young girls may therefore be of particular significance, given its inter-generational consequences. Gender differentials in the impact of migration on child nutritional status, are assessed by estimating a version of equation 2, where Sij is child’s i’s WAZ or HAZ z-score in household j. The instrument set for migration is as described earlier, and, as before, village fixed effects remove any unobserved community characteristics, such as the availability or quality of health care facilities. The set of exogenous child characteristics, Cij , include the child’s gender, age and age squared, mother’s and father’s level of education (in com-

April 20, 2009

15:11

WSPC - Proceedings Trim Size: 9in x 6in

gmansuri

Work Migration and Investment in Origin Communities

225

pleted grades), mother and father’s age, mother’s height, the total number of siblings, the presence of an older brother or sister under age 18 and the presence of a grandmother and grandfather in the house. The rationale for this last set of variables is that young children who have grandparents in the house may get better supervision and child care and possibly also some additional nutritional resources. The set of exogenous household level characteristics, Xj , is as described earlier. The final sample includes all children, age 7 months to 10 years, whose z-scores are in an acceptable range.x This leaves a total sample of 4731 children, for whom HAZ scores are in the acceptable range (of which 1157 are under age 3) and 4906 children with WAZ scores in this range. About 27% of these children belong to migrant households. The mean HAZ is -1.92. More than half the sample (54%) has HAZ scores which are below -2 SD a level that signifies some stunting and about a third have scores below -3 indicating severe stunting.y The mean WAZ in this sample is -1.82. Close to 44% of the sample has WAZ scores below -2 SD and 23% have scores below -3 indicating severe malnourishment. In contrast, 29% have HAZ scores in the normal range and 31% have WAZ scores in the normal range. The first panel in Table 7 presents the estimation results for HAZ. The second and third panels disaggregate the results for HAZ by age.since early childhood height deficits can have permanent effects.z The last panel presents the results for WAZ. The first three columns present the estimates of the migration coefficient under the assumption that the migration decision is uncorrelated with unobserved village and household attributes. The last three columns present the coefficients after instrumenting. All specifications control for the full set of household and child variables described above.

x We

use -6 to 6 as the acceptable range. Scores outside this range are typically indicative of measurement error in either height, weight or age. In doing this we follow the convention used for dealing with extreme z-score values. See for example, Hoddinott and Kinsey (2001).13 y This is somewhat higher than the 48.9% stunting rate reported for the country as a whole in the Pakistan Demographic and Health Survey (1990) This is not surprising given that the PRHS sample includes only rural children. z The main disadvantages of using anthropometic measures is that body size is likely to be affected by inputs other than consumption and is a stock variable. Neither of these is a concern for us since we are not looking at changes in body size for the same child over time.

April 20, 2009

15:11

226

WSPC - Proceedings Trim Size: 9in x 6in

gmansuri

Ghazala Mansuri

OLS estimates indicate that migration has a large positive effect on both measures of child growth, however, there do not appear to be any significant differences by gender. Instrumenting increases the coefficient estimates (columns 4-6), indicating a negative selection bias, and gender difference are now significant. In particular, girls in migrant households obtain larger gains than boys. Focusing first on child height, young girls in migrant households are indeed taller than girls in non-migrant households and the difference is significant at the 5% level. While girls in non-migrant households are about 0.25 SD below boys, migration increases the HAZ score of girls by 1.5 SD in comparison to 0.82 SD for boys. As a consequence, girls in migrant households actually do better than boys in such households in absolute terms. Once the sample is split by age, the HAZ score for younger girls increases by almost 1.8 SD, while the increase in only 0.36 SD for boys, with the consequence that the HAZ score for girls and boys in this age group is a full SD apart and this difference is significant at the 5% level. This effect is almost fully sustained among older girls. The coefficient on gender differentiated effects is now significant at the 1% level. Older girls gain about 1.6 SD in HAZ scores while boys gain only 0.7 SD. This large impact on girls and the relatively smaller effect on boys suggests that girls in rural Pakistan might bear a higher degree of uninsured household risk. Further, given the epidemiological evidence on the impact of height by age 3 on adult height, the fact that the impact of migration on the height of young girls is carried over as girls grow, provides support to the contention that the ability to avert growth shocks at a young age can have persistent positive growth effects. Turning next to the evidence on WAZ scores, the results suggest that girls from migrant households have better weight for age z-scores than do boys. While boys are better off by about 0.21 SD in non migrant households, girls in migrant households more than make up this loss, gaining 1.21 SD against a gain of about 0.6 SD for boys. Since weight for age is a short term measure, subject to substantial fluctuation near the time of measurement, there is no advantage in splitting the sample by age group, however.

5. Conclusions This chapter examines the impact of temporary economic migration on private investments in human and physical capital among migrant households

April 20, 2009

15:11

WSPC - Proceedings Trim Size: 9in x 6in

gmansuri

Work Migration and Investment in Origin Communities

227

in origin communities in rural Pakistan. It draws on several papers written by the author which examine the ways in which migration induced changes in household structure shape the sorts of investments migrant households undertake with their increased resource flows. The results discussed in the chapter focus on the impact of migration on child schooling, labor market participation and health. The data come from a representative survey conducted in rural Pakistan in 2001-2002. The incidence of migration is quite high, with one in four rural households reporting at least one migrant. Since individuals and households choose to migrate, many of the same characteristics which influence the decision to migrate are likely to also affect other household decisions such as how much capital investment to undertake, or how much to invest in the education or health of children. The papers discussed in this chapter develop an instrument for the migration decision which varies at the household level. This is important since the results suggest that there is negative selection into migration and that the ability to clean out community level unobservables is likely to be quite important. The results highlight two key findings. First, migration has a positive and extremely significant impact on both child schooling and health and there are important gender differences in this effect, with relative gains for girls outstripping those for boys by a good margin, resulting in a substantial net reduction in gender inequalities in access to education and health. Second, children in migrant households are less likely to be involved in economic work and report working for substantially fewer hours. References 1. G. Mansuri, Migration, School Attainment and Child Labor in a Rural Economy, Tech. Rep. 3945, World Bank Policy Research Working Paper (2007). 2. G. Mansuri, Migration, Sex Bias and Child Growth in Rural Pakistan, Tech. Rep. 3946, World Bank Policy Research Working Paper (2007). 3. G. Borjas, Quarterly Journal of Economics 107, 123 (1992). 4. M. Boyd, International Migration Review 23(3), 638 (1989). 5. C. Menjivar, International Journal of Comparative Sociology 36, 219 (1995). 6. D. J. Massey, G. Hugo, A. Kouaouci, A. Pellegrino and J. E. Taylor, Population and Development Review 20, 699 (1993). 7. P. Orrenius, The Role of Family Networks, Coyote Prices and the Rural Economy in Migration from Western Mexico:1965-1994, Tech. Rep. 9910, Federal Reserve Bank of Dallas Working Paper (1999). 8. G. Genicot and S. Senesky, Determinants of Migration and “Coyote” Use among Undocumented Mexicans in the United States, tech. rep., Georgetown University (2004).

April 20, 2009

15:11

228

WSPC - Proceedings Trim Size: 9in x 6in

gmansuri

Ghazala Mansuri

9. K. Munshi, The Quarterly Journal of Economics 118, 549 (2003). 10. A. d. J. Winters, Paul and E. Sadoulet, Journal of Human Resources 36, 159 (2001). 11. Banerjee, Journal of Development Economics 36(2), 337 (1991). 12. A. H. and M. Garcia, Economic Development and Cultural Change 42, 485 (1994). 13. H. J. and B. Kinsey, Oxford Bulletin of Economics and Statistics 63, 409 (2001).

LABOR MARKET REFORM AND POVERTY – THE ROLE OF INFORMAL SECTOR SUGATA MARJIT Centre for Studies in Social Sciences, Kolkata Email: [email protected] SAIBAL KAR Centre for Studies in Social Sciences, Kolkata HWWI, Hamburg, Germany DIBYENDU SUNDAR MAITI Centre for Studies in Social Sciences, Kolkata Recent papers, discussing the impact of economic reform in India, argue that the positive effect of reform is more significant in states, which are not ‘labour friendly’. Also labour market reforms seem to be a pre-condition for success of liberal policies as far as their impact on poverty is concerned. We argue that the exact mechanism behind such a link is yet to be clarified. We try to provide such a mechanism in terms of a general equilibrium model involving formal and informal workers. Our framework is capable of providing such a link and shows that there are occasions when such link is violated. Key words: Informal labour, capital mobility, labour market reform JEL Code: F13, F16, J21, J31, O17

1. Introduction Economic reform and poverty in India has emerged as a topic of great interest among economists ever since India started liberalizing its economic policies in the early 1990s. High rates of GDP growth in the recent years have encouraged economists and policy analysts to explore whether such growth has contributed to the reduction in poverty across states. Although rates of poverty in urban and rural areas have shown declining trends in general, the outcome varies considerably across states. Topalova (2005), for example, argues that tariff reduction on importable commodities has not been effective in reducing the incidence and depth of poverty across districts in India with concentration of import-competing activities. Using a specific factor model of trade the study

229

230

Sugata Marjit, Saibal Kar and Dibyendu Sundar Maiti

shows that in the presence of limited factor mobility, trade liberalization caused to increase the extent of rural poverty in India. In a similar vein (also considering product and labor market deregulations) and in connection with the effect of trade on poverty in India, Hasan, Mitra and Ural (2006) provide contradictory evidence showing that the impact of trade reform on poverty is actually more visible in states with relatively ‘flexible’ labour market conditions. Moreover, this is consistent with the position of Besley and Burgess (2004). Flexible labour market characteristics, as exemplified and quantified by Besley and Burgess (2004) do however, have some exceptions. According to their results, Maharastra and Gujarat despite being labor friendly in terms of the conditions set out in this paper have shown impressive improvements. The present study intends to trace the exact link between labour market flexibility and poverty in the presence of a huge informal labour market, as would be the case with most developing countries, including India. Labour market rigidities usually lead to the hiring of informal workers who are hired at a wage rate lower than the one prevailing in the formal sector. More specifically, in India more than 90% of the workforce is absorbed in the informal segment if agriculture is included in the estimate. On the other hand, in the presence of less aggressive labor unions pro-employer governments may help to reduce hiring and firing costs of the organized workers. Hence, in those states more people are likely to find jobs in the formal sector. Greater employment should consequently have a negative impact on poverty. Here we confine ourselves only to the definition of income poverty, such that people are poor if they earn a low wage as is common among the informal sector workers in many poor countries. The workers do not have to be necessarily unemployed in order to be considered poor; they may have a job and may still be living in stark poverty due to the prevalence of very low market determined wages. It is also likely that a more employer friendly policy will lead to a rise in the informal wage since increased labour demand in the formal sector will subsequently draw from the pool of informal workers. Thus, if informal workers are poor to start with, flexible labour market conditions should increase their wages and hence reduce poverty. Therefore, there are two distinct effects of labor market reforms on poverty, as reflected in rising wage in the informal segment and more employment in the formal sector. This is however, a fairly naive and incomplete argument. The realities certainly demand consideration of more intricate relationships. Suppose, we consider capital to be freely mobile between the formal and the informal sectors. Then, as flexible labour market conditions increase return to capital in the formal sector, capital is drawn into the formal sector and away from the informal

Labor Market Reform and Poverty – The Role of Informal Sector

231

sector. A pure supply side response will then be a cut back in the existing wage in the informal sector, hurting the left-out informal workers. Moreover, if formal sector does not expand sufficiently despite increased inflow of capital, poverty must go up. Therefore, depending on the degree of capital reallocation between the formal and the informal sector two countervailing possibilities can be distinctly captured through a competitive general equilibrium framework. In situations where capital faces a more restrictive mobility condition, stringent labour market regulations are harmful for informal workers since low employment in the formal segment leads to greater crowding into the informal sector lowering the wage rate. However, if capital could move freely, aggressive trade unions in the organized sector will cause to push capital away into the informal segment thereby raising the informal wage despite substantial absorption of workers. Hence, the interest of the organized and unorganized workers will converge. Thus whether labor market reforms help the informal workers depend on the behaviour of capital flows. The problems with some of the recent papers written on the impact of reform in India and its relationship with labour market flexibility are quite a few. First, the exact theoretical or testable hypothesis regarding the relationship between labour market reform and poverty is not properly analyzed. Second, in the presence of a vast informal labour market, the focus should have been much more on this segment rather than on the organized sector. Third, it is evident from the various rounds of NSSO that real informal wage has increased substantially in all the states in the post-reform era with absolutely no revolutionary changes in labour market conditions anywhere. This fact has been hardly taken into account. However, in a recent paper Marjit and Kar (2007) show that the effect of trade reform on the real informal wage in various states in India is positive and such improvement in the informal wage can be significantly explained by the accumulation of real fixed assets and gross value added in the so-called urban Non-Directory Manufacturing Enterprises (NDMEs, employing up to five workers, as per NSSO definitions). Furthermore, they show by constructing a panel for the states and over the years (1984-85 to 2000-01 with five year intervals in the data series) that the incidence of urban poverty (BPL percentage) in a given period is negatively and significantly affected by a rise in the informal wage in the previous period. With this backdrop the present paper draws on the growing literature on informal labour market in the developing economies and builds up a model where product market reform and labour market are simultaneously implemented in a general equilibrium framework allowing for some degree of

232

Sugata Marjit, Saibal Kar and Dibyendu Sundar Maiti

capital mobility. In this set up product market reform and labour market reform have conflicting outcomes depending upon the degree of inter-sectoral capital mobility. The second section describes the working of the basic model drawn from Carruth and Oswald (1981), Agenor and Montiel (1996), Marjit (2003), Marjit and Maiti (2006) and Marjit, Kar and Beladi (2007). It also discusses the impact of both types of reform and derives conditions under which both will improve the informal wage. The informal wage in this paper is used as a proxy for poor people’s income and may be interpreted as the minimum requirement for being on the margin. Any drop in wages would push the individuals below the poverty line. Section 3 concludes. 2. The Model We have a two sector economy producing X and Y with labour and capital. X is produced in the formal sector with workers paid a fixed wage w . If workers do not find a job in the formal sector, they go the informal sector where everyone gets a job and earns a market determined wage w. It is assumed that w > w . Note that, w though exogenous in the framework, can be endogenously determined either through the action of an optimizing union (Carruth and Oswald, 1981), Dasgupta and Marjit (2006)a or through a model of ‘effort observability’ as developed by Esfahani and Saleh-Isfahani (1989). Agenor and Montiel (1996) make extensive use of this framework in analysing development policies in a macroeconomic context. The fixity of w is assumed because the crucial focus of the analysis rests elsewhere as we shall describe and one can treat changes in w as changes in effective hiring cost. Thus lowering of w is synonymous with more flexible labour market conditions. X and Y both are traded goods with prices exogenously determined in the rest-of-the-world. This is the case of a small open economy. We shall discuss the implications of relaxing this assumption later. However, the fixity of prices is an artefact to focus on the pure supply side responses. One can provide a more profound justification behind such assumption. In a very interesting paper, Foster and Rosenzweig (2004) argue that whenever there has been a productivity increase in the Indian agriculture, the consequent higher rural wage has discouraged rural industrialization. Thus the supply side effect could not be compensated by greater demand for local goods through the increased income effect. Therefore, the importance of supply side a See appendix for a brief derivation on endogenous wage formation in the presence of labor unions in the formal sector.

Labor Market Reform and Poverty – The Role of Informal Sector

233

effect must not be ignored even if there might be some demand side repercussions. In our model, the production functions exhibit CRS and diminishing returns and the markets are competitive. Capital is imperfectly mobile between the formal and the informal sectors. Absolute immobility of capital at one end gives us the specific-factor model while the perfect mobility yields a 2×2 HeckscherOhlin-Samuelson framework. These are two special cases in our model. Competitive price conditions imply wa LX + ra KX = PX (1 + t )

(1)

wa LY + Ra KY = PY

(2)

aij s are input-output coefficients derived by factor price ratios ‘t’ denotes a measure of “protection”/artificial subsidy/protective regulation which protects market and effectively increases the price. Workers try to find a job in the high wage sector. The unsuccessful ones are absorbed in the informal sector. a LX X + a LY Y = L

(3)

Full employment of capital implies K X + KY = K

(4)

a KX X = K X

(5)

a KY Y = K Y

(6)

However, Kx and Ky once allocated act as imperfect substitutes. In other words, there is a mobility cost. KX r = f ( ), f ′ > 0 KY R

(7)

Kˆ X − Kˆ Y = μ (rˆ − Rˆ )

(7a)

One can show that

Where ‘٨’ denotes proportional change and μ ∈ [0, ∞) denotes the mobility elasticity with μ = 0 , it is a standard specific factor model. With μ → ∞ , we have perfect mobility of capital.

234

Sugata Marjit, Saibal Kar and Dibyendu Sundar Maiti

(3), (4), (5) and (6) can be combined as, a LX ( K − K Y ) + aa LY .K Y = L a KX KY

Given PX (1 + t ) and w , (1) determines r. Hence,

(8)

a LX gets determined. Then a LY

(2), (7) and (8) determine w, R , Ky. In this framework, product market reform implies a decline in t and labor market reform is synonymous with a decline in w . From equations (1) and (2) it is perfectly possible to pre-empt the isolated implications of product market reform and labor market reform in the economy. We would nevertheless derive a general condition in the appendix in order to emphasize on the potential impact of a simultaneous occurrence of both, which also leads to proposition I we present below. Intuitively, a product market reform only, i.e. a decline in t, with full mobility of capital should indicate a decline in the sectoral rates of return to capital and hence an improvement in the wage received by the workers in the informal sector. On the other hand, a labor market reform, where the workers in the formal sector now suffer due to a fall in the negotiated wage, would cause to draw in capital from the other sector given the initial differential in the rates and subsequently lower the return to the informal workers as well. The argument may be summarized as the following claim. Claim I. a. Perfect mobility of capital implies that the labor market reform hurts the informal workers while the product market reform is beneficial for them. b. Immobility of capital implies exactly the opposite of (a) When both the product market reform and the labor market reform are undertaken simultaneously the implications are countervailing and therefore an improvement in the informal wage is only conditionally feasible. And yet, there is a possible case that both can lead to beneficial impact on the informal wage (see appendix for proof).

A ⎞ ⎛ A Proposition I: ∃μ , μ ∈ ⎜ 1 , 1 ⎟ such that both types of reform undertaken ⎝ B1C B 2C ⎠ simultaneously will improve w.

Labor Market Reform and Poverty – The Role of Informal Sector

235

Proof: A fall in both t and w increase the informal wage under ‘some’ capital mobility if and only if, the income-share of labor used in the production of commodity X is positive and less than 1. Since this is always true except for the special case where X is produced by labor only, which is not relevant here, there always exists a possibility of wage gain for the informal workers (detailed proof in appendix). Proposition I implies that although the success of both types of reform depends on the extent of capital movement and in a way conflicting in nature, there are certain degrees of capital mobility as defined in the above range, when the positive impact of tariff reduction outweighs the negative impact of labor market reform. This is not a trivial result since this is tantamount to identifying the critical degree, or at least the critical zone of capital mobility which can ensure a rise in informal wage despite presence of labor market reforms. This zone may certainly be treated as important information when considering capital mobility as a policy variable for improving the conditions of the poor informal workers is a target. 2.1. Aggressive labor

Aggressive labor force may negotiate a higher formal wage compared to a more submissive labor force. Another way of characterizing labor aggressiveness should be as follows. No matter whether it is the formal or the informal sector, a region is said to be more aggressive if perceived labor cost is higher than in another region with the same ( w, w ) . This is justified by the observation that it might be more expensive to maintain the same level of productivity in two regions. Relatively aggressive workforce might imply bad work culture, loss of actual time of work etc. Even though for the organized workforce it may not get reflected in the nationally negotiated wage rate, it will be reflected in the local informal wage rate. We capture this effect by a factor α > 1 attached to the labor coefficient in the competitive price conditions. What we show next is that the Besley and Burgess (2004) proposition is an outcome of our general equilibrium framework. Once we use the wider interpretation of the phenomenon of labor aggression, the competitive price conditions change to wα a LX + ra KX = PX (1 + t ) wα a LY + ra KY = PY

With α > 1 implying more aggressive labor force.

(9) (10)

236

Sugata Marjit, Saibal Kar and Dibyendu Sundar Maiti

Differentiating we get, wˆθ LY = −θ KY rˆ − θ LY αˆ ⎛ Tˆ − θ LX αˆ ⎞ = −θ KY ⎜ ⎟ − θ LY αˆ ⎝ θ KX ⎠

=−

θ KY ˆ ⎡θ θ ⎤ T + αˆ ⎢ KY LX − θ LY ⎥ θ KY θ ⎣ KX ⎦

(11)

Note that even if we do not bring in the policy reform into the picture, greater value of α will reduce informal wage if X is relatively capital intensive, a reasonable assumption we suppose. More aggressive labor will affect the informal workers because the effect of a unit cost increase will be felt more on a sector which uses greater amount of labor per unit. For the same reason the positive effect of trade reform on w, a drop in T, will be dampened. 3. Concluding Remarks

Product market and labor market reforms should have different impact on informal wage, a benchmark of poor people’s income in a developing country. The role of capital mobility becomes quite crucial in the context. While more flexible capital movement between the formal and informal segments helps in improving the informal wage in the context of product market reform, the same may hurt informal workers when hiring (or firing) costs go down in the formal sector. This implies that labor-friendly states will have high informal wage when capital does not move much. Movement of capital can itself be a time dependent phenomenon. We propose to examine the theoretical outcomes with the help of the data available through the NSS. Our earlier empirical analysis strongly suggests that the capital formation in informal sector pushes up the informal wage and the rise of wage has significant negative impact on urban poverty between 1989 and 2000. Acknowledgments

We are indebted to seminar participants at the ISI, Calcutta, ISI, Delhi and Viswa Bharati University for helpful comments. Saibal Kar acknowledges the Humboldt Fellowship during which part of this research was carried out. We also thank an anonymous referee for helpful suggestions. The usual disclaimer applies.

Labor Market Reform and Poverty – The Role of Informal Sector

237

Appendix

Endogenous determination of w The labor union is concerned only with the wage setting in the formal sector, i.e. sector X, given the sectoral stock of capital (we will derive two variations – one with the sector-specific capital and the other with fully mobile capital). The Union’s utility function is given by: U = U ( w, L X ( w))

where, L X = a LX X =

(A.1)

a LX w K X = φ ( )K X a KX r = φ(

w )K X f ( PX , w)

(A.2)

since from competitive price conditions in equation (1), r is determined by ( PX , w) . Now, given K X , it is easy to show from (A.2) that

δ LX < 0, as φ ′ < 0 . δw

From (A.1),

δU δU δU δ L X =0 ⇒ + =0 δw δ w δ LX δ w Let us assume sufficient restrictions on U, such that,

(A.3)

δ 2U < 0 . From (A.3), δ w2

consider an equilibrium value of formal wage solving the relation: w * We have now set the framework for capturing the labor market reforms. Consider a slight modification of equations (A.1) and (A.2), as follows, U = U ( wλ , L X ( wλ ))

and

where,

LX = φ(

wλ )K X f ( PX , λ w)

γ > 1 ⇒ Pro - labor regulations γ < 1 ⇒ Anti - labor regulations

(A.1.1)

(A.2.1)

238

Sugata Marjit, Saibal Kar and Dibyendu Sundar Maiti

It also implies that any rise in γ would be a move towards pro-labor regulations and vice versa. Given (γ , K X , PX ) and with some restrictions on the functions, U (.), φ (.) and f (.) we can derive,

w* = ϕ (γ , K X , PX )

(A.3)

We reinstate this optimal value of w * in equation (A.3) and differentiate totally with respect to γ , such that the relationship between union-determined dw * wage rate and the labor market reform turns out to be negative, i.e., 0, i = 1, 2,3

(A.4)

Now using equations (1) and (2), we can re-write the equations of change (with fixed commodity prices) as:

and

ˆ LX wˆθ LX + rˆθ KX = −γθ

(A.5)

wˆθ LY + Rˆθ KY = 0

(A.6)

ˆ LX + rˆθ KX = 0 Using (A.5) and (A.6), (−α 1 + 1)γθ Here, as long as, (−α 1 + 1) > 0 , a rise in γˆ will lead to a fall in rˆ , and the rest of the results hold. In other words, a move towards pro-poor labor regulations would unambiguously reduce the return to capital accruing to that sector. The result would be indifferent even if capital were fully mobile between the two sectors earning the same return in both places. The added implication would have been a rise in the unorganized wage as well due to a pro-labor market reform. In fact, we have argued in the main text that places where the reform is labor friendly in nature, the informal sector can register an increase in wages with palpable impact on the level of poverty, provided capital is relatively free to move.

General Condition for Claim I.

We are interested in the impact of a decline in t and w on w, the informal wage. We have to solve for wˆ as a function of wˆ and (1 + tˆ ) = Tˆ . We follow Jones

Labor Market Reform and Poverty – The Role of Informal Sector

239

(1971) and Marjit and Kar (2007) closely to derive the following, by differentiating equations (8) and using (7a). −λ LX

σX ˆ ˆ λ ( w − T ) − μλ KX (λ LY − λ LX KY )(rˆ − Rˆ ) σ KX λ KX

(A.7)

Where λ s are allocation shares of labor and capital in each sector, σ s are elasticity of factor substitution and θ s are the cost-shares. Substituting for rˆ, Rˆ etc. by differentiating competitive price equations we get,

[ − A1 + μ B1C ] + wˆ [ A1 − μ B 2C ] wˆ = Tˆ − D1 − μ D 2C − D1 − μ D 2C

(A.8)

Where, A1 =

λ LX σ X λ λ θ λ λ , B1 = KX , B 2 = KX LX , C = λ LY − LY KY θ KX θ KX θ KX λ KY

D1 =

λ LY σ Y θ , D2 = λ KY LY θ KY θ KY

(A.8) helps us in framing Claim I. Proof of Claim I (a): When μ → α , from (A.8) B B wˆ = Tˆ. 1 + wˆ 2 − D2 D2 Therefore, wˆ > 0 if Tˆ < 0 and wˆ < 0 if wˆ < 0 Proof of (b): When μ → 0 , from (A.8) A A wˆ = Tˆ. + wˆ . 1 D1 − D1

Therefore, wˆ < 0 if Tˆ < 0 and wˆ > 0 if wˆ < 0. QED Proof of Proposition I:

From (A.8) wˆ =

Tˆ ( A1 − μ B1C ) ˆ ( μ B 2C − A1 ) +w D1 + μ D2C D 1 + μ D 2C

(A.9)

240

Sugata Marjit, Saibal Kar and Dibyendu Sundar Maiti

It is easy to check that wˆ wˆ < 0 iff A1 < μ B1C and < 0 iff μ B 2C < A1 wˆ Tˆ

Therefore, for both types of reform to have a positive impact on w one must have, A1 A 1 1 > 1 Or, > B 2C B1C B 2 B1 Or, B 2 < B1 which always holds as 0 < θ LX < 1 . References

1. P.R. Agenor and P.J.Montiel, Development Macroeconomics, NJ: Princeton University Press (1996). 2. T.Besley, and R.Burgess Quarterly Journal of Economics, 19, 91 (2004). 3. A.Carruth and A. J.Oswald European Economic Review, 16, 285 (1981). 4. Central Statistical Organisation (CSO). Annual Survey of Industries, Ministry of Statistics and Programme Implementation, New Delhi, Government of India, various years. 5. I.Dasgupta and S. Marjit Evasive reform: informalisation in a liberalised economy with wage-setting unions, in B. Guha-Khasnobis and R. Kanbur, (Eds.) Informal Labor Markets and Development, NY: Palgrave-MacMillan (2006). 6. H.S.Esfahani, and D.Salehi-Isfahani, Economic Journal, 99, 818 (1989). 7. Andrew Foster and M. R.Rosenzweig, Economic Development and Cultural Change, 52, 509 (2004). 8. R.Hasan, , D Mitra and B. P.Ural India Policy Forum 3, 71 (2006). 9. R. W.Jones The specific-factor model in trade, theory and history, in J. N. Bhagwati et al. (Eds.), Trade, Balance of Payments and Growth, Amsterdam: North Holland (1971). 10. S. Marjit, Journal of Development Economics, 72, 371 (2003). 11. S. Marjit, and D.Maiti Globalisation, economic reform and informal labor, in: B. Guha-Khasnobis and R. Kanbur, (Eds.), Informal Labor Markets and Development, NY: Palgrave- MacMillan (2006). 12. S. Marjit, S. Kar, and H.Beladi Review of Development Economics, 11, 313 (2007). 13. S. Marjit, and S.Kar The urban informal sector and poverty – effects of trade reform and capital mobility in India”, MPIA-PEP-Network working Paper # 2007-09, www.pep-net.org, (2007). 14. Petia Topalova. NBER Working Paper No. 11614, MA: NBER (2005).

MEASURING HARM DUE TO CHILD WORK AND CHILD LABOUR: PATTERNS AND DETERMINANTS FOR INDIA DIGANTA MUKHERJEE ICFAI Business School, Kolkata, India Email: [email protected] SASWATI DAS Indian Statistical Institute, Kolkata, India This paper uses household level data from National Sample Survey Organization (NSSO) of India, the 55th round (1999 - 2000), to study the pattern of child labour and child work from the perspective of potential harm hence caused to the children. We first comment on the relative magnitude of the usual incidence measures and the harm adjusted measures put forth by us. We have considered structured light work as skill improving and hence beneficial for the children. This gives rise to the incidence of negative harm (or positive net benefit) to some children due to work. Secondly, we study the possible determinants or correlates of such activity and consequent harm among education, income and social status related variables. We find that the parents' level of education plays an important role in reducing harm due to activity by the child; thus establishing the linkage between social and human capital outcomes in the family. The child’s own education is also seen as being important in determining this extent.

1. Introduction There exists a difference between child work and child labour. The former is more generic and relates to children who are engaged in work, economic or noneconomic, paid or unpaid which are performed at or outside homes. The length of such involvement is also important given that every involvement of child in work does not affect their growth. Analytical clarity in this regard is not merely an academic exercise. It is closely related to the pattern of solutions to any specific problem. Child labour is an aberration that is to be eliminated forthwith. Scrambling all forms of deprived childhood into one category of ‘child labour’ is compounding confusion and disagreement rather than being helpful. Although the issue and determinants of child labour has been studied quite extensively

241

242

Diganta Mukherjee and Saswati Das

(see Das and Mukherjee 2 and the references cited therein), that of child work, as distinct from child labour, has received relatively less attention in the literature. It is a question to consider whether in the macro statistics of child labour, such fine-tuning is feasible at all. Many activities which children may be undertaking, like weeding the fields, looking after cattle, doing household chores or collecting firewood, may wrongly have been recorded as ‘labour’, even if children are less than 14 years old. What do we know of the threshold where the work of children becomes child labour and what of the threshold where child labour becomes intolerable child labour? When are daily practices associated with the process of primary socialization not beneficial any longer but rather harmful to the child? For an answer to those questions one shall have to go down to the micro setting and study the matrix of daily practices. One could try to construct the different permissible and impermissible combinations by developing a matrix. In the matrix, a crucial distinction could be made between child labour, child work and idleness; between engagements in the labour market; working within the family enterprise and household work; and between proper schooling, partial schooling and no schooling at all. Depending on the age and the hours involved in each of the activities one could decide on whether it is a matter of child labour or not. Following Alec Fyfe3, we may argue that a child’s work as physical and mental involvement in a family or social activity can be a gradual initiation into adulthood and a positive element in the child’s development. Light work, properly structured and phased, is not child labour. Work, which does not detract from other essential activities of children, namely leisure, play and education, is not child labour. Child labour is work which impairs the health and development of the children. The International Labour Organisation 4 accepts that millions of young people legitimately undertake work, paid or unpaid, that is appropriate for their age and level of maturity. By doing so, they learn to take responsibility, gain skills and add to their own and the family’s well-being. Hence a distinction should be made between a) child-friendly forms of socialization, including light work, b) child labour at specific ages and up to specific degrees of strain but not interfering with school and a healthy childhood, c) non-enrolment in school, even if not labouring, d) child labour interfering with school, and e) the worst and intolerable forms of child exploitation, even amounting to child bondedness. The engagement of children in labour processes is not only a phenomenon of the developing countries. Neither is it a new development. In ancient societies, the difference between the daily occupations of adults and children

Measuring Harm Due to Child Work and Child Labour

243

were gradual; learning the skills, customs and tricks socialized children so that by the time they turned adults, they had learned the tricks of the trade. The modern job based on modern technology and the associated complex division of labour requires at least some general education to cope with it. There exists a visible trade off between early entry in the labour market without training and that with gradual entry with some training that involves cost. Thus, in the west, work in productive activity (outside the family, against a monetary compensation) not exceeding two hours a day is not considered to be detrimental to the development of a child, after (s)he attains a certain age. There are several levels of activity by children that are to be considered. At the worst, we have the category of child labour who is economically active as their principal activity. But all the children who are economically active are not child labour. Again, some of the children who are economically inactive may also belong to the category of child labour, although they do not perform productive labour. We consider productive and unproductive (family or social) labour. Children engaged in domestic chores within their own households are not considered as economically active. Statistical Information and Monitoring Programme on Child Labour (SIMPOC) excludes from the child labour category many instances of work done in and around the household: Child labour does not include activities such as helping out, after school is over and schoolwork has been done, with light household or garden chores, childcare or other light work. Although claiming such light activities as proper work only trivializes the genuine deprivation of childhood faced by the millions of children involved in child labour that must be effectively abolished, sometimes these activities may take up a substantial amount of time and hence may be detrimental to the future wellbeing of the child. Thus, in this paper we consider this category also as a form of child work and based on intensity, identify it as harmful or otherwise. We need to specify a minimum duration of work carried out regularly or usually by the child that attracts a return in cash or kind or both to the child or to the parents/guardians or both to define a child worker. The ILO allows for 14 hours per week as the cut-off point for light work from 10 years onwards. Light work by children aged 10 to 14 is work which is not hazardous in nature and which does not exceed 14 hours per week. ILO Convention No. 33 and findings of research on the impact of child labour on school attendance and performance support the chosen cut-off point. One hour of work per day during the reference week is sufficient for classifying a person as at work in economic activity during that week. The minimum age for employment or work should normally not be less than 15

244

Diganta Mukherjee and Saswati Das

years, but developing countries may fix it at 14, and a number of countries have fixed it at 16. We used the age of 15 as a cut-off point for all countries in our global estimates. Therefore, "child labour" as estimated in this document consists of all children under 15 years of age who are economically active excluding (i) those who are under five years old and (ii) those between 10 - 14 years old who spend less than 14 hours a week on their jobs, unless their activities or occupations are hazardous by nature or circumstance. Table 1 illustrates the classification of forms of work according to the above principles. Table 1. Child labour as defined for the purpose of global estimates Age Group

Form of Work Non-Hazardous work (in non-hazardous industries/occupations and < 43 hours/week) Light work (< 14 hours/week)

Regular Work (14 to 42 hours/week)

5–9 10 – 14

An additional issue that we address in this paper is to explain the effect of parental human capital and occupational characteristics on the extent of child work. As documented in Das and Mukherjee2 this linkage is important for incidence and nature of child labour. The rest of the paper is organized as follows. Section 2 provides an overview of the general pattern of economic activity of children, child labour and hazardous work across the globe in 2000. We also provide information on recent trend in work participation in India. Section 3 describes the data set from the National Sample Survey Organisation (NSSO) of India that we have used for our analysis, the definitions and model specifications for measuring harm as well as for investigating the linkage between harm and the factors like human capital (parental and self) and occupational characteristics. The descriptive and regression results are discussed in Section 4. Finally section 5 concludes. 2. General patterns of child work and labour To initiate our discussion on child labour and work, we first look at some global estimates of economic activity by children in Table 2 below. It shows that around 17.6% of the world’s children (aged 5 – 14) are economically active with almost equal rates for the boys and the girls.

Measuring Harm Due to Child Work and Child Labour

245

Table 2. Global estimates of economically active children ages 5 to 17 in 2000, by gender and age group Gender and age group Boys

5-9

12.3

307900

70900

23

616400

109000

17.7

15-17

170200

75100

44.1

786600

184100

23.4

5-9

291800

35000

12

10-14

291300

66800

22.9

5-14

583100

101800

17.5

15-17

161800

65800

40.7

744900

167600

22.5

5-9

600200

73100

12.2

10-14

599200

137700

23

5-14

1199400

210800

17.6

332100

140900

42.4

1531100

351700

23

15-17 Total

38100

Work ratio (%)

5-14

Total girls Both gender

308500

Number at work (‘000)

10-14

Total boys Girls

Total population (‘000)

Source: International Programme on the Elimination of Child Labour (IPEC) 2002:17-18

The methodology used in the data collection for the 2000 analysis has a number of interesting departures. The categorization of the various categories of children at work in itself provides transparency. The figure of 210 million of total economically active children has the following components. All the children below the age of 12 have been counted as child labourers if they worked. This verdict is based on ILO resolution 138 which, in developing countries with a weak economy and insufficiently developed educational infrastructure, allows children only from the age of 12 onwards to be involved in light work (less than 14 hours a week) only. The number of such children was calculated to be 109.7 million. In addition, of the 101.1 ‘economically active children’ in the 12 – 14 age categories, 76.6 million were considered as working in activities than did not qualify as ‘light work’. To get a break down of these numbers in to those who are actually child labour and those who are not, we further report this classification in Table 3 below. The figure of approximately 250 million of children who can be termed as child labour, which the ILO projected in the mid-1990s, soon developed into

Diganta Mukherjee and Saswati Das

246

an icon. The accuracy of the statistics was not the first concern. The high accounts have been useful in the debate on social clauses in international trade agreements. Table 3. Children in economic activity, child labour, and hazardous work (by gender and age group), 2000 Gender and age group

Economic ally active children (EAC) (‘000s)

Child labour (‘000s)

Child labour as per cent of EAC

Children in hazardous work (CHW) (‘000s)

CHW as percent of EAC

CHW as per cent of child labour

5-11 Total

109700

109700

100

60500

55.2

55.2

Boys

56300

56300

100

30700

54.5

54.5

Girls

53400

53400

100

29800

55.8

55.8

12-14 Total

101100

76600

75.8

50800

50.2

66.3

Boys

52700

41500

78.7

30600

58.1

73.7

Girls

48400

35100

72.5

20200

41.7

57.5

5-14 Total

210800

186300

88.4

111300

52.8

59.7

Boys

109000

97800

89.7

61300

56.2

62.7

Girls

101800

88500

86.9

50000

49.1

56.5

15-17 Total

140900

59200

42.0

59200

42.0

100

Boys

75100

34400

45.8

34400

45.8

100

Girls

65800

24800

37.7

24800

37.7

100

All Total

351700

245500

69.8

170500

48.5

69.5

Boys

184100

132200

71.8

95700

52.0

72.4

Girls

167600

113300

67.6

74800

44.6

66.0

Source: IPEC 2002: 17-18

A breakdown of the child labour number across geographical regions is provided in Table 4 below which highlights the high global regional disparities in this respect.

Measuring Harm Due to Child Work and Child Labour

247

Table 4. Child labour across the globe Number of Children (‘000s)

Type of Economies

Work Ratio

Age 5-9

Age 10-14

Age 5-9

Age 10-14

Developed Economies

800

1,700

1.4

2.8

Transition Economies

900

1,500

1.3

4.2

Asia and the Pacific

40,000

87,300

12.3

26.5

Latin America & Caribbean

5,800

11,600

10.6

21.5

Sub-Saharan Africa

20,900

27,100

23.6

34.7

Middle East and North Africa

4,800

8,600

10.8

19.6

Source: IPEC 2002: 17-18

The National Sample Survey (NSS) data in India suggest that India had 22 million child labourers in 1983, 17 million in 1987, 13 million in 1991 and 10 million in 2000 (Kannan 5, D. P. Chaudhri et al. 1, Lieten 6). So one could conclude that the trend apparently is downward. Whereas the participation rate of rural boys and girls in 1983 was respectively 13.5% and 12.5%, in 2000 it was 4.7% and 4.9% respectively. In urban India it had come down to 2.7% and 1.9% respectively. Table 5 provides recent information on age specific work participation rates in India. Table 5. Age-specific work participation rates in India, 1983-2000 Location and gender Rural boys

Age 5 – 9

Age 10 – 14

1983

1987

1993

2000

1983

1987

1993

2000

2.5

2.4

1.1

0.6

25.3

19.0

13.9

9.1

Rural girls

2.6

2.3

1.4

0.7

24.0

18.0

14.2

9.6

Urban boys

0.8

0.5

0.5

0.3

11.3

8.5

6.6

4.9

Urban girls

0.7

0.5

0.5

0.2

7.0

6.5

4.5

3.6

Source: NSS data in Varma and Satpathy (2004)

The economic consequences of harm due to child labour have been documented in Das and Mukherjee8. They estimate a wage premium equation on education for the adults, with significant positive coefficient, which acts as the benchmark for foregone future income when a child is forced to neglect studies in order to work.

248

Diganta Mukherjee and Saswati Das

3. Data and methodology As mentioned earlier we use the household level data collected and made available by National Sample Survey Organization (NSSO) for the large sample round (55th) conducted during 1999-2000 on employment and unemployment situation in India. Some important concepts and definitions followed in this study are described below in the subsequent paragraphs. 3.1. The sample One salient feature of the 1999-2000 survey was that the rotation-sampling scheme was adopted. The survey period was divided into four sub-rounds, each with duration of three months. Under rotation sampling scheme, 50 per cent of the sample first stage units (fsus) of each sub-round was revisited in the subsequent sub-round. fsu's are urban frame survey blocks for the urban sector. The ultimate stage units are households at the subsequent stage. A sample of 10,400 fsus (rural and urban combined) were surveyed at all-India level during the survey period. Out of 10,400 fsus, a total of 3,900 fsus (1,300 each from sub-rounds 1, 2 and 3) were revisited in the subsequent quarters. NSSO makes available both types of data file, one, including the fsus visited only once during the period and another type including the revisited fsus also. In the present analysis only first type of data files were used to avoid the repetition. 3.2. Constructing the variables Activity status: In 1999-2000 survey, NSSO used three approaches for classification of the activity statuses of the person surveyed. These are: 1. number of persons usually employed - usually employed in the principal status and all workers taking into account the employed according to both the principal and subsidiary statuses, 2. the average number of persons employed in a week based on the current weekly status and 3. the average number of persons- days employed per day. Of the three approaches, the usual principal status approach is best suited as a measure of the economic activity in an economy with seasonal fluctuations in the employment. This is because in this approach the criterion used is the pattern of activities followed by the person for a relatively long period of time (NSSO, 2001). We considered only the child population aged 5-14 for the purpose of our estimates. In our present study we considered only those children usually

Measuring Harm Due to Child Work and Child Labour

249

employed in the principal status and termed them as labourers and those aged 10 – 14 and working for an average of less than 14 hours/week as economically active. Father’s/mother’s education: Adult education has been categorized as below: (a) not literate 0; (b) literate but below primary 1; (c) primary and middle 2; (d) secondary 3; (e) higher secondary 4; (f) graduate and above 5. Father’s occupation: Only two categories of occupation have been considered. One category represents those who work in household enterprise (self employed) or own account worker, employer or work as regular salaried/wage employee. An own account enterprise is an undertaking run by household labour, usually without any hired worker employed on a ‘fairly regular basis’. By ‘fairly regular basis’ it is meant that the major part of the period of operation(s) of the enterprise during the last 365 days (NSS, 2001). Another category is if other than these specified cases. Child education: Child education has been categorized as below, (a) not literate 0; (b) literate but below primary 1; (c) primary 2; (d) above primary 3. Child labour incidence: In NSS data relationships between family members can only be identified using the information regarding ‘relation to head’. Due to incomplete information child labour incidence for only following two cases could be considered: 1) head of the household is father with living spouse; and 2) head of the household is grand father with only one son or one daughter with his or her spouse alive. It is worth mentioning here that filtering through these conditions not only reduces the sample size but may bias our results also. The family composition may be related to child labour decision. Other variables: Other variables used in our analysis are average monthly per capita expenditure as proxy for per capita income and dummies for caste (general and scheduled) and religion (Hindu and Islam).

250

Diganta Mukherjee and Saswati Das

3.3. Defining harm due to child work The topic of discussion in this paper is a quantification of benefit / harm to the child due to (excess) activity in the young age. This we set out to do in the following way. Following the ILO’s benchmark rule that any work in excess of 2 hours / day (14 hours / week) is considered to be detrimental to the child’s development but light work (up to this extent) may be beneficial in the sense of learning, we formulate an index of harm due to work for any child. To get a workable definition for harm, we categorise activity of the children into the following four groupsa. L: child labour (days, considering 8 hrs/day). Those who are usually employed in the principal status. For those who are not usually employed but active (economically or otherwise), we have three classifications. E: child work (non labour), not attending school (in hours). ES: child work (labour) with attending school (in hours). EES: child work (non labour) with attending school (in hours). The two traditional ways of measuring child labour and economic activity of children, as proposed in the ILO, IPEC report, represents two polar ways of looking at economic activity by the children. The incidence of child labour only considers the state L as harmful and ignores the other states E and EES completely. On the other hand, the incidence of economic activity of children treats both L and ES at par in terms of harmfulness to the children, and thus ignoring the beneficial initiation aspects of economic activity. To incorporate this possibility, we propose the following formulation as a measure of harm, H. For age (5, 9): H = L + βES + αE + σ EES For age (10, 14): H = L + β(ES – 0.25) + α(E – 0.25) + σ (EES-0.25) if EES, ES, E > 0.125 H = L – βES – αE - σ EES a

if EES, ES, E ≤ 0.125

For details on the classification of activity status of children, see the appendix.

(1)

Measuring Harm Due to Child Work and Child Labour

251

where 0 < α, β, σ < 1, EES, ES and E are measured in terms of man days (taking key from the NSSO, we consider a 8 hours work day). Thus, the cut-off is at 0.25 (= 2 hours / day) and 0.125 corresponds to 1 hour of work. Note that according to this specification a suitably low amount of light work would be assigned a negative level of harm, hence indicating the beneficial aspects of such work. 3.4. Alternative specifications The parameters α, σ and β reflects the researcher’s value judgement regarding the overall importance of subsidiary work on the child’s well-being. Depending on our judgement regarding the relative importance of different kinds of activities in causing harm, we consider alternative orderings for the parameters. One possibility we may consider is that the potential for loss is greater for children who are also going to school rather than for those who are not attending school. For this case, we take β, σ > α. An alternative theory could be that the children who are in status E are being prevented from going to school because of child work. In that case, we should consider the potential for loss to be greater in status E than in ES and EES. This may be modelled by taking β, σ < α. Also, the average extent of harm caused by this kind of work may be given a larger or smaller importance compared to the status L. This is achieved by selecting values near 0 as well as near 1 for α, σ and β. A small value indicates that the harm potential of light work is considered to be quite small, whereas a value close to 1 gives it a higher weightage. The choice of any particular value is entirely a matter of value judgement and should be informed by local conditions. In our final analysis, we have considered eight alternative specifications for (α, σ, β) covering the range discussed above, namely (0.2, 0.3, 0.35), (0.4, 0.5, 0.55), (0.6, 0.7, 0.75), (0.75, 0.85, 0.9), (0.7, 0.6, 0.65), (0.3, 0.2, 0.25), (0.5, 0.4, 0.45) and (0.85, 0.75, 0.8). 3.5. Model Specification and Hypotheses It is interesting to ask whether, apart from the affluence of the household, the educational background and occupational characteristics of the parents influence the work status of the child. We consider the parents’ education and father’s occupation (as very few mothers are reported as working on a job) as potentially important explanatory variables apart from MPCE, caste and religion dummies. Thus, we consider the following hypotheses.

Diganta Mukherjee and Saswati Das

252

H0A: parental education levels do not influence the decision to engage the child in work to the extent that it causes harm. H0B: father’s occupational characteristics do not influence the decision to engage the child in work to the extent that it causes harm. H0C: the child’s education levels do not influence the decision to engage him/her in work to the extent that it causes harm. We now discuss the testing of hypothesis H0A, H0B and H0C. For the limited dependent variable harm (H), we use the following transformed model: H + 0.125τ + 0.001 1.001 - H = γ 0 + γ 1MPCE + γ 2GC + γ 3 SC

lhm = log

+γ 4 Hindu + γ 5 Islam + δ 1 fgedu +δ 2 mgedu + δ 3 focu + δ 4 sedu + ε

(2)

Where

τ = max{α , β ,σ } MPCE = average monthly per capita expenditure; GC = indicator or dummy variable for general caste; SC = indicator or dummy variable for schedule caste; Hindu = indicator or dummy variable for “religion = Hindu” Islam = indicator or dummy variable for “religion = Islam” fgedu = father’s education; mgedu = mother’s education; focu = father’s occupation; sedu = child’s education;

γ ' s and δ 's are the parameters of the model and ε is the random noise term. The transformation used on H to arrive at lhm is to ensure that the range of values becomes a large interval symmetric around zero so that normality may be safely assumed. The choice of transformation is dependent on the value of τ. For + 0.0885 τ = 0.7, the transformation is lhm = log ( H1.001 and for τ = 0.2, it is −H ) H + 0.026 lhm = log ( 1.001 − H ) etc.

Measuring Harm Due to Child Work and Child Labour

253

The model (2) uses observations including all active children (economically or otherwise). Hypothesis H0A now becomes H ' 0 A : δ 1 = δ 2 = 0 . Similarly, H0B and H0C becomes H ' 0 B : δ 3 = 0 and H ' 0C : δ 4 = 0 respectively. 4. The Results We do our analysis separately for four subsets of the child population, the male and the female in rural and urban areas. The basic descriptive statistics are presented in the table below. As discussed earlier, we considered eight alternative parametric specifications but as the results are very robust, we present the harm indices for three representative cases. The salient features of the descriptive statistics table are as follows. Very few children aged 5 – 9 are in status ES (both attending school and economically active). As expected, girl children are mostly classified as nonlabour where as boy children are mostly classified as labour. This would be due to the predominance of girl children doing extensive household work but who are not paid wages. The average figures for harm for different choices of values for α, σ and β are very similar for the L and ES status, implying that the quantification of harm that we attempt here is quite robust with respect to the choice of parameter values. But the values for status E and EES vary quite a lot. We discuss these in more detail below. For the 10 – 14 age group, MPCE is highest for the EES category most of the time followed by the ES, E category and finally for the L category. For the 5 – 9 age group, again E and EES category has higher average MPCE followed by the L category. The averages for the ES category are not discussed as they are based on too few observations. Father’s, mother’s and child’s education and father’s occupation also show similar pattern as MPCE. These observations prompt us to expect a negative relationship between harm and these independent variables (MPCE, fgedu, mgedu, sedu and focu) in our subsequent regression analysis. 4.1. Studying the incidence results To study the pattern of harm due to child labour or work, we look more closely at the rates of incidence of such events. We compute the rates of incidence of child labour, activity by children and also the harm adjusted measure of incidence for the alternative choices of harm potential ordering and parametric configurations that are considered here (section 3). These are denoted by I (L), I (EAC), I (H, 0.2, 0.3, 0.35), I (H, 0.6, 0.7, 0.75) and I (H, 0.85, 0.75, 0.8)

254

Diganta Mukherjee and Saswati Das

respectively. We compute these for the two regions, two genders and two age groups separately, giving us eight categories in all. I (L) and I (EAC) are computed in the usual way. The harm adjusted measures are computed for each of the eight categories by considering a (frequency) weighted average of harm for each of the four working statuses L, E, ES & EES. The results are presented in Table 7. The first thing to note from the table is that the rate of incidence of child labour is roughly half (2%) of that of EAC (≈ 4%) in our analysis. This is broadly consistent with ILO findings reported in Section 2. The harm adjusted figures are more interesting. Although these figures take into account the additional effect of harm due to excess activity (all activity by children aged 5 – 9 and above 2 hours per day for those aged 10 – 14), they are in some cases (for boys aged 10 – 14 in both the rural and urban sector for the low values of α, σ and β) less than the simple child labour incidence rates. This is because the simple I (L) measure counts each child labour as 1 but the I (H, ., ., .) measures explicitly consider the intensity of activity and hence the average man-days’ consideration is incorporated in the analysis. The average intensity in the L-status for boys aged 10 – 14 are 0.90 in the rural sector and 0.92 in the urban sector. Also, the incidence of status E and ES are also much lower than the status L for this age group. Thus, the aggregate turns out to be lower than I (L). For the other cases, the natural ordering of I (L) < (H, 0.2, 0.3, 0.35) < I (H, 0.6, 0.7, 0.75) < I (H, 0.85, 0.75, 0.8) < I (EAC) is preserved. Similar discrepancies do not occur in case of girls as the incidence of E & ES status is much more predominant among them. This outweighs the weightage considerations even though the average intensity in the L status is not much higher for the girls than boys. The value of I (H,.,.) always turns out to be between I (L) and I (EAC). The incidence rates (according to any measure) are uniformly lower for the younger cohort than the older one. This is partially good in the sense that it indicates that the youngest children at least are getting some respite from harmful extent of activity. The range of values for I (L) and I (EAC) are also illuminating. While the range for I (L) is 0.24% to 4.49%, that of I (EAC) is much larger (0.56% to 12.18%). This highlights the role of economic activity in bringing out the discrepancies among the different genders, age groups and locations.

Age Work Location Total individuals Group status in sample Rural 1938 5-9 L Male ES

EES 10-14 L ES E EES Rural Female

3256

5-9

L ES E EES

10-14 L ES

255

E

Measuring Harm Due to Child Work and Child Labour

E

Table 6. Descriptive statistics Frequency Incidence Harm* Harm* Harm* MPCE* Fgedu* Mgedu* Focu* Sedu* No. (0.2,0.3,0.35) (0.6,0.7,0.75) (0.85,0.75,0.8) (%age) 78 4.02 0.94 0.94 0.94 356.00 0.42 0.01 0.65 0.28 0.16 0.16 0.16 136.78 0.83 0.11 0.48 0.56 3 0.15 0.19 0.19 0.19 399.33 1.00 0.00 0.67 1.00 0.17 0.17 0.17 98.29 1.00 0.00 0.58 0.00 69 3.56 0.20 0.40 0.85 429.80 0.88 0.49 0.46 0.46 0.00 0.00 0.00 220.44 1.16 1.04 0.50 0.61 0.26 0.26 0.26 460.86 1.52 1.18 0.76 1.12 90 4.64 0.20 0.20 0.20 171.02 1.11 0.93 0.43 0.39 1292 66.67 0.90 0.90 0.90 381.67 0.45 0.15 0.57 0.80 0.20 0.20 0.20 157.69 0.80 0.52 0.49 0.97 64 3.30 0.05 0.07 0.08 484.48 1.52 0.64 0.80 2.13 0.13 0.14 0.17 178.20 1.31 0.97 0.41 0.68 186 9.60 0.20 0.40 0.85 358.16 0.60 0.25 0.49 0.70 0.00 0.00 0.00 132.75 1.09 0.76 0.50 0.94 0.22 0.38 0.56 497.53 1.36 1.01 0.81 1.68 156 8.05 0.00 0.03 0.00 162.92 1.04 0.94 0.39 0.72 76 2.33 0.93 0.93 0.93 396.36 0.54 0.20 0.57 0.14 0.17 0.17 0.17 172.17 0.99 0.61 0.50 0.42 1 0.03 0.00 0.00 0.00 923.00 5.00 3.00 1.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 145 4.45 0.20 0.40 0.85 361.27 0.59 0.26 0.48 0.21 0.00 0.00 0.00 178.01 0.96 0.72 0.50 0.43 0.27 0.27 0.27 478.96 1.53 1.17 0.83 1.19 83 2.55 0.18 0.18 0.18 149.08 1.23 0.99 0.38 0.50 983 30.19 0.87 0.87 0.87 366.87 0.51 0.12 0.57 0.57 0.22 0.22 0.22 162.49 0.84 0.45 0.49 0.87 0.03 0.04 0.04 475.95 1.84 0.66 0.67 1.70 43 1.32 0.11 0.12 0.15 178.85 1.23 0.10 0.47 0.86 1767 54.27 0.20 0.40 0.85 386.65 0.67 0.16 0.57 0.67 0.02 0.00 0.01 186.80 1.00 0.53 0.50 0.95

Urban Male

595

5-9

4.82

L

36

6.05

ES

0

0.00

E

33

5.55

EES

23

3.87

403

67.73

ES

9

1.51

E

54

9.08

EES

36

6.05

L

29

3.17

ES

1

0.11

E

47

5.14

EES

24

2.62

172

18.80

9

0.98

594

64.92

39

4.26

10-14 L

Urban Female

915

5-9

10-14 L ES E EES

0.22 0.04 0.92 0.10 0.00 0.00 0.20 0.00 0.15 0.05 0.92 0.15 0.02 0.06 0.20 0.00 0.22 0.00 0.88 0.14 0.00 0.00 0.20 0.00 0.21 0.10 0.91 0.17 0.02 0.06 0.20 0.00 0.22 0.00

0.38 0.01 0.92 0.10 0.00 0.00 0.40 0.00 0.15 0.05 0.92 0.15 0.03 0.09 0.40 0.00 0.38 0.00 0.88 0.14 0.00 0.00 0.40 0.00 0.21 0.10 0.91 0.17 0.04 0.08 0.40 0.00 0.38 0.01

0.55 0.01 0.92 0.10 0.00 0.00 0.85 0.00 0.15 0.05 0.92 0.15 0.04 0.13 0.85 0.00 0.56 0.00 0.88 0.14 0.00 0.00 0.85 0.00 0.21 0.10 0.91 0.17 0.06 0.12 0.85 0.00 0.56 0.01

461.10 167.52 563.33 370.69 0.000 0.000 632.24 436.74 652.52 300.70 454.28 197.75 542.67 185.22 556.19 303.66 709.67 378.12 643.59 381.46 627.00 0.000 552.51 232.93 774.87 386.28 438.90 257.63 525.89 124.14 445.99 199.66 743.54 460.60

1.51 1.24 1.17 1.44 0.00 0.00 1.79 1.41 2.35 1.50 0.82 0.98 2.00 1.00 1.24 1.39 2.75 1.59 1.07 1.16 2.00 0.00 1.60 1.54 2.54 1.77 0.80 1.11 2.22 0.83 0.96 1.13 2.54 1.62

1.07 1.01 1.14 1.69 0.00 0.00 1.76 1.64 1.56 1.27 0.34 0.75 0.68 0.87 0.94 1.22 2.08 1.48 1.21 1.18 2.00 0.00 1.21 1.32 1.83 1.58 0.35 0.71 1.33 1.00 0.36 0.78 1.92 1.51

0.80 0.40 0.72 0.45 0.00 0.00 0.82 0.39 0.83 0.39 0.70 0.46 1.00 0.00 0.69 0.47 0.83 0.38 0.59 0.50 1.00 0.00 0.77 0.43 0.92 0.28 0.56 0.50 0.89 0.33 0.71 0.45 0.82 0.39

1.73 0.68 0.47 0.56 0.00 0.00 0.94 0.50 1.04 0.37 0.99 0.97 1.89 0.60 1.44 1.16 1.78 0.80 0.62 0.56 0.00 0.00 0.51 0.62 0.96 0.36 1.03 1.05 2.00 0.87 1.02 1.07 2.23 0.71

Diganta Mukherjee and Saswati Das

157

256

EES

Location & Gender Rural Male

Rural Female

Urban Male

Urban Female

Age Group

Total in NSS

Total individuals in sample

I (L)

I (H, 0.2, 0.3, 0.35)

I (H, 0.6, 0.7, 0.75)

I (H, 0.85, 0.75, 0.8 )

I (EAC)

5-9

26578

240

0.293476

0.417977

0.4699

0.586726

0.903002

10-14

25825

1698

5.002904

4.791946

5.037599

5.472914

6.575024

5-9

23776

305

0.31965

0.513501

0.635473

0.909909

1.282806

10-14

22827

2950

4.306304

5.451614

7.11171

10.71201

12.92329

5-9

13352

92

0.269623

0.323322

0.372753

0.483972

0.689035

10-14

14701

502

2.74131

2.650568

2.763826

2.973811

3.414734

5-9

12178

101

0.238134

0.328133

0.405321

0.578995

0.829364

10-14

13138

814

1.309179

2.162277

3.11539

5.20475

6.195768

Measuring Harm Due to Child Work and Child Labour

Table 7. Measuring rate of harm due to activity (in %)

257

258

Diganta Mukherjee and Saswati Das

We noted earlier that the girl children are more involved in non-wage economic activity and hence they show up less in the labour (L) status but more in the ES or EES status. This fact is now borne out by the relative values of the I (.) measures for the boys and girls in both rural and urban sectors. Whereas the value of I (L) is higher for boys, that of I (H, ., ., ,) and I (EAC) are higher for the girls. This holds true except for the 5 – 9 age group in the rural sector for the measure I (L) and the 10 – 14 age group in the urban sector for I (H, 0.2, 0.3, 0.35). Similar conclusions hold for the other set of parametric configurations (results not shown). In general, as expected, the I ( H, ., .) measure acts as a continuum between the two polar cases represented by I (L) and I (EAC), thus corroborating our idea that a value weighted measure of child activity would be useful in policy analysis depending on the specific (subjective) value judgement of the researcher. By a judicious choice of values for the three parameters α, σ and β, the family of measures I (H, α, σ, β) would be useful for gauging the actual harm due to (excess) activity for the children in a society. We have also illustrated the pattern of harm for a few selected groups with histograms provided at the end of this paper from which the qualitative nature of the distribution of harm among the children may be more closely studied. A closer look at all the histograms (corresponding to the 32 cases as detailed in Table 6) reveal that the learning effect due to light work indeed shows up in the histograms corresponding to the status ES, for 10 – 14 year old girls in both regions and the rural boys. In these categories, there are a significant number of individuals who are measured as having experienced negative harm from activity. These are individuals who have been working, on an average, less than 14 hours per week, which level of activity is considered as skill improving for a child in our set up. Obviously, this is not possible in status L. For the status E also this does not show up in any of the categories but the histograms for this status show quite a lot of variability in some cases. This also shows up in the standard deviation figures in Table 6. So, for status E, we do not get any incidence of negative harm (or skill improving light work) in our sample. But the extent of harm is quite variable over the sample. 4.1. Determinants or correlates of child work We finally come to the second major part of our analysis, that of exploring the possible determinants or correlates of the extent of work by the children (or the consequent harm) in an economy. The candidates for this explanatory role are MPCE, education and occupation related variables and caste and religion

Measuring Harm Due to Child Work and Child Labour

259

dummies as described above. This analysis is done for the eight cases mentioned above. The results are qualitatively similar for all the parametric configurations. Thus we conclude that the regression results are robust with respect to the choice of α, σ and β. So we need not discuss the results for all the parametric configurations separately and may draw our cause and effect inferences from any one of the three regression results presented in Tables 8, 9 and 10. Table 8. Regression results for α = 0.2, sigma=0.3 and β = 0.35 VARIABLE NAME

Estimated Coefficient Rural Male

Rural Female

Urban Male **

Urban Female 3.74E-04 (0.8602)

MPCE

1.56E-05 (0.2788E)

2.12E-04 (0.6045)

-1.12E-03 (-1.792)

GC

-0.15638 (-0.6246)

-0.38603* (-2.398)

-0.30932 (-0.9946)

-0.52067* (-2.376)

SC

0.5675* (2.568)

-0.4309* (-2.861)

0.33395 (0.7726)

-0.42184 (-1.572)

HINDU

2.1718* (8.452)

0.76406* (3.484)

0.19746 (0.2919)

-0.64436 (-1.294)

ISLAM

2.0449* (5.663)

0.12076 (0.4413)

0.653 (0.9323)

-0.41648 (-0.8006)

FGEDU

-0.46912* (-4.483)

-0.27103* (-3.989)

-0.29244* (-2.022)

-8.75E-02 (-0.8761)

0.6086* (3.43)

0.35002* (2.796)

0.31497 (0.9551)

-0.69306* (-3.221)

MGEDU

-0.79842* (-5.559)

-0.1901** (-1.713)

-0.70379* (-4.405)

-0.13805 (-1.097)

SEDU

-0.44193* (-4.774)

-0.27684* (-3.933)

-0.52692* (-3.402)

-0.14729 (-1.471)

2.0396 (5.772)

0.48527 (1.797)

4.9266 (6.164)

1.5828* (2.821)

1938

3256

595

915

0.1532

0.0374

0.1730

0.0326

0.1493

0.0347

0.1603

0.0230

1928

3246

585

905

38.767

14.018

13.601

3.386

FOCU

CONSTANT No. of obs. R

2

R

2

d.f. F t-ratios are in parentheses

* : significant at 5% and ** : significant at 10%

Diganta Mukherjee and Saswati Das

260

Table 9. Regression results for α = 0.6, sigma=0.7 and β = 0.75 VARIABLE NAME MPCE

Estimated Coefficient Rural Male

Rural Female

Urban Male **

Urban Female

-1.86E-05 (-0.3551E)

2.63E-04 (0.8339)

-1.09E-03 (-1.847)

GC

-0.13749 (-0.5851)

-0.34175* (-2.362)

-0.28636 (-0.9812)

-0.39663* (-2.242)

SC

0.55565* (2.678)

-0.38588* (-2.85)

0.3289 (0.8109)

-0.33436 (-1.544)

HINDU

1.9927* (8.262)

0.80288* (4.073)

8.59E-02 (0.1353)

-0.63057 (-1.569)

ISLAM

1.8718* (5.523)

0.22353 (0.9087)

0.52886 (0.8047)

-0.40752 (-0.9705)

FGEDU

-0.42923* (-4.37)

-0.24595* (-4.028)

-0.26791** (-1.974)

-6.39E-02 (-0.7923)

FOCU

0.56285* (3.38)

0.32108* (2.853)

0.30322 (0.9799)

-0.50572* (-2.911)

MGEDU

-0.75466* (-5.598)

-0.21396* (-2.145)

-0.62572* (-4.173)

-0.20631* (-2.031)

SEDU

-0.4281* (-4.926)

-0.25994* (-4.108)

-0.50163* (-3.451)

-0.11642 (-1.44)

CONSTANT

2.4082 (7.261)

1.003 (4.132)

5.1294 (6.84)

2.782* (6.142)

1938

3256

595

915

0.1252

0.0417

0.1666

0.0356

0.1485

0.0390

0.1538

0.0260

1928

3246

594

905

38.539

15.692

12.992

3.707

No. of obs. R

2

R

2

d.f. F

t-ratios are in parentheses * : significant at 5% and ** : significant at 10%

1.13E-04 (0.3222)

Measuring Harm Due to Child Work and Child Labour

261

Table 10. Regression results for α = 0.85, sigma=0.75 and β = 0.8 VARIABL NAME

Estimated Coefficient Rural Male

Rural Female

Urban Male **

Urban Female -1.16E-04 (-0.3845)

MPCE

-1.11E-04 (-0.2353)

4.17E-04 (1.657)

-1.01E-03 (-1.914)

GC

-9.69E-02 (-0.4566)

-0.22515** (-1.952)

-0.29828 (-1.135)

-0.32203* (-2.112)

SC

0.53253* (2.843)

-0.27625* (-2.559)

0.25434 (0.6964)

-0.27096 (-1.451)

HINDU

1.8341* (8.423)

1.0856* (6.908)

-0.28023 (-0.4902)

-0.53882 (-1.556)

ISLAM

1.6888* (5.52)

0.62839* (3.204)

0.21751 (0.3675)

-0.27188 (-0.7512)

FGEDU

-0.3682* (-4.153)

-0.20361* (-4.182)

-0.25291* (-2.069)

-5.22E-02 (-0.751)

FOCU

0.44601* (2.967)

0.24438* (2.724)

0.2682 (0.9625)

-0.36795* (-2.458)

MGEDU

-0.69189* (-5.685)

-0.32459* (-4.081)

-0.47893* (-3.547)

-0.29334* (-3.351)

SEDU

-0.46463* (-5.923)

-0.26372* (-5.228)

-0.47283* (-3.613)

-0.11622 (-1.668)

3.0098 (10.05)

1.9498 (10.08)

5.7031 (8.445)

3.6074* (9.241)

CONSTANT No. of obs.

1938

3256

595

915

2

0.1636

0.0688

0.1555

0.0613

R2

0.1597

0.0662

0.1425

0.0519

d.f.

3256

3256

594

905

41.897

26.644

11.968

6.562

R

F

t-ratios are in parentheses * : significant at 5% and ** : significant at 10%

The first point to note is that the goodness of fit, as measured by R2 is higher for the male children cases compared to the female. Thus it is possible that the measure of harm due to child work for the girls is affected by some other factors not considered in our model in addition to the ones that we include. Regarding the independent variables, MPCE is nowhere significant (except for the urban male, for whom it is significant at 10% level), thus indicating that among the families with working children, income plays almost no role in

262

Diganta Mukherjee and Saswati Das

dictating the extent of activity that a child undertakes. The caste and religious dummies are significant in some cases. For example, coming from the general caste lowers the expected activity (scheduled caste dummy has negative effect for females and positive for male children) and this implies that children from the scheduled tribes are usually more active even to the extent of damaging their educational prospects. Controlling for the caste and religion factors, we find that father and mother’s education has the right sign (negative) in all cases where it is significant. In fact, these variables also turn out to be significant in almost all cases (father’s education not significant for urban female). It is possible that in the rural sector, as average education levels are lower, education has a stronger incremental effect on household decisions and that has shown up in the father’s case, who would have a stronger say in household decisions here. The mother may want to train her girl child for future motherhood and in the process may involve her in excessive household work and this choice could be largely independent of the mother’s level of education. This may have caused the insignificance of mgedu for the girl child in some cases. Child’s education, as expected, has significant negative coefficients in most of the regressions (except female urban). This is simply a confirmation of the fact that working children lose out on education. Father’s occupation has a significant positive coefficient in the regression results for the rural cases. This is surprising, but a possible explanation could be that the contractual characteristics of occupation (which is considered in focu) are not relevant in the rural context and the variable is correlated with some other variable (not in the model) that affects harm positively. It has the right sign (negative) only for the case of urban girl child. We postulated that the unusual sign of focu may also be due to effect of income. It is possible that the relationship between child work and father’s occupation is influenced by income and this may have different qualitative properties for different income groups. To explore this further, we break down the total sample into three roughly equal groups based on the value of the MPCE and estimate the regression model separately for the three groups. These may be called conditional, on MPCE, regression results. But the unusual sign for focu persists in all the three groups for all the four segments of the population and the different parametric configurations studied in this paper. Thus, the puzzle of the sign of focu remains unresolved.

Measuring Harm Due to Child Work and Child Labour

263

5. Conclusion In this paper we have used household level data from National Sample Survey Organization (NSSO) of India for the 55th round (1999 - 2000), to study the pattern of child labour and child work from the perspective of potential harm hence caused to the children. We first comment on the relative magnitude of the usual incidence measures and the harm adjusted measures put forth by us. Secondly, we study the possible determinants or correlates of such harm among education, income and social status related variables. We find that the parents' level of education plays an important role in reducing harm due to activity by the child; thus establishing the linkage between social and human capital outcomes in the family. The child’s own education is also seen as being important in determining this extent. Appendix A. Detailed classification of activity status of children L: Relevant codes for usual principal activity status are 11, 12, 21, 31, 41 and 51. E: Relevant codes are 92, 93 in both usual principal activity status and usual subsidiary activity status. ES: Relevant code is 91 in usual principal activity status and codes 11, 12, 21, 31, 41, 51 are in the usual subsidiary activity status. EES: Relevant code is 91 in usual principal activity status and codes 92, 93 are in the usual subsidiary activity status. Usual activity status: principal and subsidiary: The usual activity status relates to the activity status of a person during the reference period of 365 days preceding the date of survey. The activity status on which a person spent relatively longer time (i.e. major time criterion) during the 365 days preceding the date of survey is considered as the usual principal activity status of the person. A person whose usual principal status was determined on the basis of the major time criterion could have persued some activity for a shorter time throughout the reference year of 365 days preceding the date of survey or for a minor period, which is not less than 30 days, during the reference year. The status in which such activity was pursued was the subsidiary activity status of that person (NSSO 2001). Description of codes: 11: worked in house hold enterprise (self-employed) : own account worker; 12: worked in house hold enterprise (self-employed) : employer; 21: worked as helper in h.h. enterprise (unpaid family worker);

264

Diganta Mukherjee and Saswati Das

31: worked as regular salaried/wage employee; 41: worked as casual wage labour: in public works; 51: in other types of work; 91: attended educational institution; 92: attended domestic duties only; 93: attended domestic duties and was also engaged in free collection of goods (vegetables, roots, fire-wood, cattle feed, etc.), sewing, tailoring, weaving, etc. for household use. References 1. D.P. Chaudhuri, C. Nyland and A. O’Rourke, Child Labour, Poverty and Gender Bias in Education in South Asia: Human Capital or Human Rights?, In G.K. Lieten et. al., Small Hands in South Asia. Child Labour in Perspective, New Delhi: Manohar, 61(2004). 2. S.Das, and D. Mukherjee, Role of Women in Schooling and Child Labour Decision: The case of Urban Boys in India, Social Indicators Research, 82, 463(2007). 3. A Fyfe., Child Labour, Cambridge Polity Press, (1989) 4. International Programme on the Elimination of Child Labour (IPEC) and Statistical Information and Monitoring Programme on Child Labour (SIMPOC), Every Child Counts: New Global Estimates on Child Labour, International Labour Office, Geneva, April 2002. 5. K.P. Kannan,. (eds.) Economics of Child Labour, New Delhi: Deep and Deep, (2001). 6. G.K.Lieten, Child Labour in South Asia: An Account of Numbers, in G.K. Lieten et. al., Small Hands in South Asia. Child Labour in Perspective, New Delhi: Manohar, 37 (2004). 7. NSSO, Level and Pattern of Consumer Expenditure in India, 1999-2000, Ministry of Statistics and Programme Implementation(2001). 8. U.K. Varma,. and A. Satpathy, Declining Trends: Child Labour in India, in G.K. Lieten eds., The Child Labour Problem. Issues and Solutions, Amsterdam: IREWOC, 93.

Agriculture

This page intentionally left blank

DEVELOPMENT STRATEGY: THE STATE AND AGRICULTURE SINCE INDEPENDENCE T.N. SRINIVASAN Samuel C. Park Jr. Professor of Economics, Yale University Stanford center for International Development, Stanford University There is a widespread belief that India is currently in an agrarian crisis, with the spate of suicides by farmers several states since the 1990s seen as a tragic symptom of the crisis. In the large and growing literature on the crisis some common themes emerge: the role of systemic economic reforms since 1991, the opening of the Indian economy to external competition and investment after decades of insulation; the impact on India of implementing the Agreement on Agriculture of the Uruguay Round of Multilateral Trade Negotiations; the alleged neglect of agriculture in the planning process since the mideighties; the decline of public investment in agriculture in response to rising fiscal deficits at the Centre and the States; and above all, the slowing of the growth of agricultural output (particularly food grains) as well as a stagnation in yields per hectare of land since the nineties. This paper argues that the fundamental factor that is at the root of the current state of agriculture is India’s pursuit, until the 1991 reforms, of a state-directed, state-controlled and state-dominated development strategy of import substituting industrialization with emphasis on heavy industry and insulation from the world economy. This strategy completely ignored the lessons of economic history: successful development lies in the transformation of economic structure by shifting a substantial part of the large initial share of labour force in agriculture and other low productivity activities in the informal sector to more productive off-farm activities through rural and urban industrialization with emphasis on labour-intensive manufactures to supply growing domestic and world markets and raising agricultural productivity. Leap-frogging the labour-intensive manufacturing stage of development altogether and focusing on information technology intensive services sector to bring about the transformation is not simply not feasible. This paper elaborates this main point by looking at major policy interventions in agriculture since independence It argues that there was no coherence, and little coordination among the centre, states and other policy making institutions in the decisions on the myriad interventions and their effectiveness in achieving their intended objectives was limited.

267

268

T.N. Srinivasan

1. Introduction India is widely believed to be currently in an agrarian crisis. An official Expert Group claims that “Indian agriculture is currently passing through a period of severe crisis. Although some features of the crisis started manifesting themselves in certain parts of India during the late 1980s, the crisis has assumed a serious dimension since the middle of the 1990s. One of the tragic manifestations of the crisis is the large number of suicides committed by the farmers in some parts of India.” (EGAI6, p13). According to the group, both long-term structural and institutional as well as short term manifestations contributed to the crisis: “The long-term structural features are a sharp decline in the share of agriculture in the Gross Domestic Product (GDP) accompanied by a very low rate of labour force diversification away from agriculture… A large dependence of working population on land has also resulted in a steep decline in per capita land availability. The increasing pressure on land resources is accompanied by severe stress on the availability of water resources in the country and unequal regional distribution of available water. On the credit front, the functioning of the rural cooperative credit institutions has deteriorated in many parts of the country. The emphasis on economic efficiency has led to the neglect of social priorities in lending by the commercial and regional rural banks… The result is growing dependence on non-institutional sources of credit at very high rates of interest. Except for a few crops, the procurement mechanism does not serve the purpose of ensuring minimum prices to agricultural producers in many parts of the country. The crisis has been exacerbated further by rapid environmental degradation and plateauing of the existing agricultural technology. The liberalisation of the economy has failed to give a big push to agricultural exports and to increase income and employment in agriculture. The gradual withdrawal of the state from active participation in development activities has resulted in a steep decline in public investment in agricultural infrastructure in general, and in agricultural science and technology in particular. This has resulted in deterioration of rural infrastructure, stagnation of agricultural research and development, and neglect of extension services.” (ibid, p. 13) Some common themes emerge from the recent literature on agrarian crisis: the role of economic reform since 1991, particularly aspects of opening of the Indian economy to external competition and the implications for India of the Uruguay Round agreement on agriculture; presumed neglect of agriculture in the planning process since the mid-eighties, decline in public investment in agriculture in response to rising fiscal deficits at the centre and states, and above

Development Strategy: The State and Agriculture Since Independence

269

all, the slowing of the growth of agricultural output, particularly food grains, as well as stagnation in yields per hectare since the early nineties. Finance Minister Chidambaram is reported to have expressed concern at the slow growing farm sector and the need for policy attention to issues like stagnant farm yield rates in many major crops, declining per capita availability of food grains and the need for additional public investment, while claiming that “these issues are high on the agenda of the Government and though it has taken many initiatives on these counts, much more needs to be done.” [http://www.hindu.com/2007/11/13/stories200711352621500.htm]. Prime Minister Singh, in his recent address to the Planning Commission, asked for a renewed focus on agriculture and drew attention to the burden of subsidies on food, fertilizers and recently on petroleum. He said “Cabinet colleagues and the Planning Commission [have] to reflect what these mean for our development options and what development options these subsidies are shutting out. Do they mean fewer schools, fewer hospitals, fewer scholarships, slower public investment in agriculture and poorer infrastructure? It is important that we restructure subsidies so that only the really needy and the poor benefit from them and all leakages are plugged.” [http://www.hindu.com/2007/11/12/stories/2007111250781500.htm ] The direct subsidies on food, fertilizers and petroleum in the budgets of central and state governments are substantial. The Centre spent Rs, 53,000 crores or 1.3 percent of GDP in 2006-07 on the three. Food subsidies accounted for little over 40 percent of the total. Estimating the shares of producers and the farmers in fertilizer subsidies is not simple. Still the share of farmers is unlikely to be small. Adding subsidies on food, indirect subsidies on sale of electricity and water from public irrigation to farmers at a price below user cost (or even free of cost in the case of electricity in some states) by the states and other myriad farmer-oriented subsidies, the total farm subsidy burden is large. Central and State governments have intervened in agriculture massively. However, the interventions have not been effective in achieving their objectives, which were not always clear. Also different interventions were not often mutually consistent, let alone reinforcing each other, in achieving a common set of objectives at least social cost. Policy interventions were made in a piecemeal fashion, by different institutions and by central and state governments. Although the five-year and annual plans were in principle the framework through which a coherent set of objectives and a coordinated set of policies could have been formulated, in practice this was not the case.

270

T.N. Srinivasan

The interventions could be classified into four broad categories relating to: (1) agrarian structure (land ownership, land access through share-cropping and tenancy), (2) market structure (regulation of markets, restrictions on futures markets and on the movement of agricultural commodities on private account within India across states and across districts within states, on foreign trade, state trading, and constraints on private trade) (3) prices of inputs and outputs (subsidies on transportation, fertilizer, irrigation water, electricity fuel and credit, procurement and subsidized sale through the public distribution system (PDS) (4) Public investment in irrigation and infrastructure and incentives for private investment in agriculture. The interventions were numerous, extensive, and varied over time in number and their severity in an unanticipated and unpredictable fashion, thus making decision making environment highly uncertain for those with vital interests in agriculture. Also states and centre intervened, often on the same issue. For example, the centre announces its procurement prices at the beginning of each season and many states supplement with their own add-ons to the prices announced by the centre and procure grain on their own account as well. Clearly, the chaotic policy interventions precluded their being mutually consistent in sub-serving a well-defined, intertemporal social objective, though they were rationalized as such.a Twenty-five years ago on the occasion of the Indian Statistical Institute’s Golden Jubilee International Conference on Review of the Indian Planning Process, I presented a paper on “Was Agriculture Neglected in Planning” (Srinivasan27). Although under plausible alternative meanings of the ill-defined phrase, “neglect of agriculture” I could not find such neglect, I concluded that a

The Constitution of India (1950) lays down the distribution of legislative (and policy making) powers of the centre states into three lists: Union list consisting of areas in which the parliament (and central government) have exclusive powers, a state list of areas in which the state legislature (and state governments) have exclusive authority and a concurrent list of areas in which both the parliament (central government) and the state legislatures (and state governments) have power with the important premise that Union law will prevail over state law. However, with the appointment of an extra constitutional body, namely the Planning Commission in March, 1950 through a resolution of the Central Cabinet soon after the adoption of the Constitution in January, 1950 and the institution of five-year plans formulated by the Planning Commission for articulating a national development strategy, the constitutionally assigned of powers of the centre were vastly expanded through administrative actions. The Finance Commission, a constitutionally mandated body, in its report in 1973 remarked “A national plan has necessarily to comprehend the entire range of developmental activities, cutting across the delimitation of powers between Centre and the states. In this process, the Government of India and the Planning Commission have acquired a voice even in matters recognized to be within the Jurisdiction of the States.” (Cited in National Commission on Agriculture15, Part II, p. 95). In the three decades and more since the report, the powers of the central government have expanded even further

Development Strategy: The State and Agriculture Since Independence

271

the lack of neglect did not mean the absence of policy failures. I identified three interrelated failures relating to institutional change, employment generation (or labour “absorption” as it was called then) and in reducing abject poverty. I pointed out the fact that in the thirty years since 1950, the productivity of workers “absorbed” in agriculture had grown at a rate slower than those employed in non-agricultural activities. The expert group (EGAI6, p27) claims that “another important manifestation of the crisis in agriculture is the stagnant if not deteriorating, terms of trade for agriculture after the introduction of economic reforms”. Ashok Mitra13, pp141-142) wrote in the middle seventies that “in recent years, the domestic terms of trade in the country moved continuously in favor of farm products in general and within the farm sector in favor of these specific crops that are marketed by . . . the richer sections of peasantry. This shift in terms of trade can be viewed as mirroring the political arrangement entered into by the urban bourgeoisie with the rural oligarchy . . . the developing shift in terms of trade in favor of the farm sector is a major price paid by the industrial bourgeoisie to cement their political arrangements with the rural oligarchy.” It turned out that the rising trend in agriculture’s TOT on which Mitra based his thesis of a grand alliance between the rural oligarchy and the urban bourgeoisie was confined to the sixties and was not seen either before or thereafter. (Srinivasan27, pp 42-43) The expert group similarly over-interprets the shortterm and reversible trends in terms of trade. The main point of this paper is that, the fundamental or ultimate contributory factors to the current crisis arise from our ignoring the basic fact of economic history: successful economic development lies in transforming economic structure over time by shifting a large proportion of the initial population and work force dependent upon agriculture and other primary activities of low productivity to more productive activities. Historically this transformation was brought about through industrialization, primarily manufacturing. Our development strategy, founded on import-substituting industrialization with emphasis on heavy industry, and rationalized by very long-term growth considerations in an economy insulated from world markets, not only delayed the transformation, but also created a self-fulfilling prophecy that the non-agricultural sectors would not grow at a rapid enough rate to provide more productive employment opportunities for an increasing share of the labour force in agriculture. For example, the National Commission on Agriculture15 stated “An overwhelmingly large proportion of the total labour force, i.e., about 72 percent, is employed as agricultural workers at present.

272

T.N. Srinivasan

Even in the most optimistic view of the creation of additional employment opportunities in the urban, non-agricultural occupations, the transfer of the labour force from agricultural to non-agricultural jobs will be rather slow. But the new labour force will continue to depend on agriculture and allied nonagricultural occupations even in 2000 A.D.” (NCA15, Part III, p. 82). In stark contrast to this pessimistic view, the memorandum of Chairman Jawaharlal Nehru, to the sub-committees of the National Planning Committee of 1938, stated “more important is the planning of different kinds of industries, large, medium and cottage, which alone may effectively mitigate the present pressure on the soil. Within a decade the aim should be to produce a balanced economic structure in which about half the population would depend on agriculture (IIAPR10, p. 55 emphasis added). Alas, after six decades after independence, more than half (60 percent or so) still depend on agriculture. Having failed thus far to expand the share of manufacturing in employment, to think that we can leap-frog the stage of manufacturing in development altogether and rely on the information technology-based service sector to bring about the missed transformation would be a pipe-dream. Also a focus on the agricultural sector, necessary and desirable though it is in the short and medium run, by itself will not address the fundamental failure of our development strategy. In the rest of the paper I will elaborate my main point by briefly looking at the major policy interventions in agriculture.b,c I will argue that these interventions more or less took it for granted that with projected population growth and a stagnant or at best slowly declining share of population dependent on agriculture, the absolute numbers of those employed (as cultivators, tenants and landless workers) will continue to grow. I will contend that even if one viewed the current situation as an agrarian crisis, ad hoc responses to it by modifying the existing set of agricultural policy interventions and adding new ones of the same type would not resolve it. What is needed not only resolve it, b

I will not be covering interventions for which the direct targets were not producers, consumers, traders and other agents involved in agriculture. These include general development programmes, integrated rural intensive agricultural development programmes, agricultural research and extension and public investment. c The literature in the English language alone on Indian agriculture, including reports of committees and commissions appointed by the government at various points of time such as the important Report of the Royal Commission on Agriculture in 1929 and the fifteen volume report of the National Commission on Agriculture of 1976, debates in the two houses of Parliament and state legislatures, as well as reports in the news media, is very vast and diverse. I cannot and will not pretend to cover this vast body of writings. I will draw selectively only on some of the major contributions that are essential to my argumentation.

Development Strategy: The State and Agriculture Since Independence

273

but also eradicate poverty once and for all is an expansion and deepening the reform process to set the economy on a development strategy and path fundamentally different from that of the first from decades of planning. 2. Agrarian Structure: Land Tenure, Ownership, Tenancy and Cultivation 2.1. Flirtation with Co-operative Farming: 1951-66 The National Planning Committee of the Indian National Congress had resolved at the end of 1940 that “the cooperative principle should be applied to the exploitation of land by developing collective and cooperative farms . . . collective or cooperative farms should be begun on ‘cultivatable waste’ land, which should be acquired, where necessary by the state immediately” (IIAPR10, p215). After independence, the committee on agrarian reforms the Congress Party (chaired by J.C. Kumarappa) in its report of 1950, concluded that collective farming to be suitable only for the development of reclaimed waste land. It categorically rejected capitalist farming as its adoption in its view “would deprive the agriculturists of their rights in land [and] turn them into mere wage earners,” and opted for individual peasant farming. However the idea of cooperative farming surfaced in the form “cooperative village management” in the First Five Year plan (1951-56) with the village as the unit of land management with individual families or groups of families cultivating blocks of land allotted by the village management body. Dandekar4 (p. 53) acidly comments that “This was a rather naive concept based on a utopian notion of a village and plain ignorance, or unwillingness to see the truth, about village community functioned.” The Second plan (1956-61), according to Dandekar “offered lip service though with less conviction,” to cooperative village management and the Third Plan (1961-66) made no mention of it and thereafter the concept was quietly dropped. The problem of landless agricultural workers and the need to provide increased employment opportunities (on and off farm) were recognized by planners. Yet as Dandekar4 (pp 84-85) points out ideas on increasing employment opportunities “were not very clear, in any case, they were not elaborated . . . what was said with respect to the landless workers in the First Five Year Plan was plainly evasive.” In particular, it was not understood that the ongoing capital intensive industrialization strategy could not generate rapidly rising employment opportunities for such workers. This basic lack of understanding led to the presumption that the problem of their employment had to be solved within the agricultural sector itself. Dandekar4 (p. 87) adds that “the

274

T.N. Srinivasan

Fifth Plan (1970-75) emphasized that landlessness was a root cause of poverty and that access to land was a major source of employment and income; that such access could be achieved either by a more equitable distribution of land or providing security of tenure to tenants and share croppers who are the actual cultivators.” He stressed that the land reform policy, by keeping a growing rural population on land, simply created a growing population of non-viable farmers (small and marginal farmers). The history of land reforms reviewed in the next section illustrates his finding amply. 3. Land Reform 3.1. Abolition of Intermediaries At the time of independence the prevailing land tenure system was complex going back to the Mughal era and earlier. P.S. Appu1 (pp. xv-xii), characterizes it as follows: “The existence of rent-receiving intermediaries between the actual tillers of the soil at the bottom, and the government at the top, great inequity in the ownership of land, concentration of agricultural lands in the hands of the upper classes, widespread prevalence of insecure tenancies inhibiting the optimum utilization of the tenants’ land, a preponderance of miniscule uneconomic holdings and to the extreme fragmentation and subdivision of holdings.” The post-independence land reform agenda naturally included the abolition of intermediaries, tenancy reform, reducing concentration of land ownership and the consolidation of fragmented holdings. However, not all items in the agenda were effectively implemented. Appu1 concludes that despite the inefficiency and slowness of its implementation, and resistance by intermediaries, the social and economic powers of the intermediaries came to an end with the implementation of legislation. However, he noted (Appu1, p. 79) that the reform had some major weaknesses: it allowed the intermediaries to retain a substantial amount of land for their “personal cultivation,” a term that was so “loosely defined in the legislation that no rights were conferred on tenants-at-will and share croppers,” resulting in millions of tenants and under-tenants being evicted. Also the payment of compensation to the former intermediaries resulted in heavy public expenditure. (Appu1, pp. 72-79) 3.2. Tenancy Reform Tenancy reform in the post-independence period evolved over a period of several decades, with three important guidelines laid down in various five-year

Development Strategy: The State and Agriculture Since Independence

275

plans: (i) there should be an upper bound on rent at one-fifth of gross produce; (ii) tenants should be accorded permanent rights in the land they cultivate subject to a limited right of land owner to resume land under tenancy for “personal cultivation” and (iii) in respect of non-resumable land, landlord-tenant relationships should be ended by conferring ownership rights on tenants. (Appu1, pp. 95-96) Tenancy reform legislations of various states and five year plans had differing definitions of ownership and of tenants, in particular, whether sharecroppers are to be deemed tenants. The First Plan defined small owners as those owning less land than a family holding and middle owners as those holding land in excess of a family holding but less than the limit for resumption for “personal cultivation” (three times the family holding). The second plan defined the “basic” holding as the minimum area needed for profitable cultivation, and a family holding as three times the basic holding. Owners of land less than a basic holding were to be deemed free to resume their entire holding for personal cultivation. Owners holding between one and three basic holdings would be allowed to resume half the area of their holding under tenancy, subject to a lower bound of a basic holding. The Second plan also elaborated the phrase of “personal cultivation” to mean such a cultivator bore the entire risk of cultivation, supervised it himself for a member of his family and supplied a minimum amount of labour himself. Although it recognized that the supply of a minimum amount of labour is difficult to enforce in practice, the Plan suggested that it should be an important criterion for land that is to be resumed for personal cultivation. Appu1 (p. 91) wryly remarks that “all these meticulous exercises in hair splitting in verbal juggling aimed at reconciling the conflicting interests of landowners and tenants ignored the realities of the power equation in the countryside and the character and capability of the administrative machinery . . . the basic fact is that the policy of ‘land to the tiller’ could not have been carried out without hurting the private property rights. But the policymakers were unwilling to wound and afraid to strike.” India’s planners from the fifties till now are also guilty of Appu’s charge of ignoring realities. 3.3. Trends in Land Ownership and Tenancy Before concluding the discussion of tenancy and turning to land ceiling legislation, a brief look at the trends in tenancy and ownership is useful the relevant data are presented in Tables 1-5 and Figure 1. Several conclusions emerge from these data. First, the incidence of tenancy as measured by

T.N. Srinivasan

276

percentage of households leasing in, percentage of area (out of total area owned), leased in or leased out has been steadily decreasing from 1971-92 to 2003 (Table 1). Second, the percentage of tenant holdings as a proportion of operational holdings and the percentage of area leased in every category of the size of operational holding (marginal, small, semi-medium and large) has been declining during 1960-61 to 2002-03 (Tables 2 - 4), with one exception that the percentage area leased in marginal holdings remained unchanged between 199192 and 2002-03. Third, sharecropping has been the dominant farm tenancy accounting for roughly 40% of the leased in area since 1960-61, except for 1970-71 when it was 48%. The proportion of leased-in area under fixed produce rent and also under fixed money rent, after declining between 1960-61 and 1981-82, has been increasing since then (Table 5). Fourth, Figure 1 shows that the absolute number of tenant holdings and the leased-in operated area, after declining until 1981-82, increased thereafter, though not in proportion to the increase in total number of operational holdings. Figure 1: Trends in the extent of tenancy T rends in extent of tenancy 16 14

number/area

12 10

no. of tenant holdings (mill.)

8

leased-in operated area (mill. Ha.)

6 4 2 0 61-62

71-72

81-82 year

Source: NSS (1997)

91-92

Development Strategy: The State and Agriculture Since Independence

277

Table 1: Estimates of reported incidence of leasing out and leasing in of lands by households from various rounds of NSS sl. No.

Characteristic

1971-72 (26th rd)

1982 (37th rd)

1992 (48th rd)

2003 (59th rd)

1

Percentage of household leasing in land

25%

18%

15%

12%

2

Percentage of area leased in to total area owned

12%

7%

9%

7%

3

Percentage of area leased out to total area owned

6%

4%

5%

3%

Source: NSS 2006a Report No. 491 Table 2: Percentage of tenant holdings and area leased in by broad size-class in 2002-03 Size class (ha) or category

% of tenant holdings

% of area leased in

0.002

% share in leased in area

4.7

3.1

0.1

0.002 - 0.5

10.9

9.3

12.3

0.5 - 1.0

12

8.3

17.9

9.8

8.6

30.3

Small

10.7

6.8

22.1

Semi-medium

10.3

6.3

21.8

7.8

4.2

14.6

Large

13.8

6.1

11.2

Over 1.00

10.2

5.8

69.7

9.9

6.45

Marginal

Medium

All sizes

100

Source: NSS 2006b Report No. 492 Table 3: Changes in the percentage of tenant holdings by category of operational holdings, 1960-61 to 2002-03 Percentage of tenant holdings Category

60-61(17th)

70-71(26th)

81-82(37th)

Marginal

24.1

27

14.4

9.3

9.8

Small

25.1

27.8

17.9

14.9

10.7 10.3

91-92(48th)

02-03(59th)

Semi-medium

23.6

24.8

15.9

12.2

Medium

20.5

20

14.5

13.1

7.8

9.5

15.9

11.5

16.7

13.8

23.5

25.7

15.2

11

Large All

Source: NSS 2006b Report No. 492

9.9

T.N. Srinivasan

278

Table 4: Change in percentage of area leased in by category of operational holdings Percentage of area leased in

Category 60-61(17th)

70-71(26th)

81-82(37th)

91-92(48th)

02-03(59th)

Marginal

16.6

18.9

9.7

8.7

8.6

Small

14

14.6

8.5

8.5

6.8

Semi-medium

11.7

11.7

7.3

7.4

6.3

Medium

9.6

8.7

6.6

6.9

4.2

Large

8.3

5.9

5.3

11.4

6.1

10.7

10.6

7.2

8.3

6.5

All sizes

Source: NSS 2006b, Report No. 492 Table 5: Trends in percentage distribution of leased-in area by terms of lease Terms of lease

60-61 (17th)

70-71 (26th)

81-82 (37th)

91-92 (48th) incl n.r. cases

02-03 (59th)

excl n.r. cases

incl n.r. cases

excl n.r. cases

Fixed money

25.6 (23.2)

15.4 (12.7)

10.9 (11.9)

19.0 (23.3)

22.7

29.5 (26.0)

29.8

Fixed produce

12.9 (12.4)

11.6 (10.5)

06.3 (7.6)

14.5 (17.9)

17.4

20.3 (19.2)

20.6

Share of produce

38.2 (42.0)

47.9 (50.7)

41.9 (38.7)

34.4 (42.1)

41.1

40.3 (43.3)

40.8

Other

23.3

25.1

40.9

32.1

18.8

9.9

8.8

All terms

100

100

100

100

100

100

100

Source: NSS 2006b, Report No. 492 Sources of estimates of 17th, 26th and 48th rounds: NSS Report No. 407. Data for 2002-03 relate to the kharif season. Figures in parentheses are the percentages of tenant holdings reporting leased-in area under the terms of lease. n.r. indicates cases for which terms of lease were not recorded.

Development Strategy: The State and Agriculture Since Independence

279

NSS16 (para 3.6.1,p26) claims that “though the measures of land reform undertaken since independence appear to have deterred the growth of exploitative tenancy, there is still a huge proportion of tenanted land in total operated area. What is most remarkable about farming in rural India is the significantly high proportion of total tenanted operated land by a small proportion of holdings.” NSS does not define what is meant by exploitative tenancy, let alone what its growth would have been in the absence of land reform. Nor does it say why the observed concentration in leased-area is unduly high. It is true that marginal farmers operating less than one hectare of land account for 10 percent of operational holdings and 30 percent of leased in land. The 30 percent of holdings operating more than one hectare account for 70 percent of leased land (Table 2). As compared to the share of operated area at 23 percent and 77 percent respectively (NSS20, Table 3.4) for the two groups, leased in area is more concentrated. There is some evidence of eviction of tenants and resumption of land for personal cultivation by landowners in the sharp fall in the percentage of tenant holdings from 25.7 to 15.2 (Table 3) and in area leased from 10.6 to 7.2 between (Table 4) 1970 and 1981-82 in the decade of the Green Revolution as compared to the stability of these percentages prior to 1970-71. Perhaps the most successful of tenancy reforms is “Operation Barga” introduced by the left front government of West Bengal after its assumption of power in 1977. This programme registered the share-croppers (bargadars) and assured their rights. It is widely believed to have contributed to the remarkable acceleration of growth in the output of rice (and of food grains, though less dramatically) from an average of 2.85 percent per year during 195051 to 1975-76 to 4.45 percent per year during 1976-77 to 1998-99. The primary contributor to this remarkable achievement is the acceleration in growth of yield per hectare of land (Sengupta, et al.26; Tables 5, 7, 12 and 14). It would be hasty to conclude that the observed acceleration in output and yield growth t was mostly due to the land reforms (particularly tenancy reforms) of the left front government. Bardhan and Mookherjee3 argue that first, the extent of tenancy in West Bengal was too small to explain the acceleration in growth of total output through acceleration of output on tenanted land. NSS data show leased-in land to total operated had declined from 12.3 percent in 1981-82 to 9.3 percent in 2002-03. Moreover, the effects of tenancy reform could be confounded with the effects of many other changes that took place at the same time, such as changes in the market environment or in farm inputs delivered by the government that may have been correlated with the tenancy reform. Further,

280

T.N. Srinivasan

there is a possibility that reform was implemented in villages with more progressive farmers and the decision whether to register or not by tenants might have been driven in part by productivity improvements arising from reasons unrelated to the registration programme. Their own study was based on disaggregated data from a panel of farms spanning 1981-95 from 89 sample villages. Their main finding is that while there was a statistically significant effect of tenancy Operation Barga on productivity of tenant farmers, there was also a large general equilibrium spillover effect on non-tenant farmers. Thus “the productivity effects of tenancy reforms were overshadowed by supply of farm input services and infrastructure spending by local governments” (Bardhan and Mookherjee3, p. 1). Also “the predicted effect of the programme on average farm yields at the level of village was only 5%, substantially smaller than the effects of farm input supply programmes . . . But the incidence of leasing being very low, the aggregate impact of this was small (ibid, p. 33). Operation Barga of West Bengal illustrates that registering tenants and effectively protecting their legitimate and legal rights alone could improve the productivity of tenant farmers modestly. However, with the incidence of tenancy for the country as a whole being small in 2003 with only 6.5 percent of the operated area being leased in (Table 4), the effect of tenancy reform, though appropriate and desirable, on overall productivity of agriculture will be small. 3.4. Ceilings on Land Ownership The policy on ceiling on land-ownership and agricultural property has been driven solely by consideration of social justice in a context. Appu, writing in 1972 (reproduced in Appu1, p. 268) put the case simply and eloquently. “In this country a simple and effective means for ensuring a measure of social and economic justice will be radical redistribution of land. The country’s industrial sector being small, even if the future programmes of industrialization meet with a great measure of success, for years to come the bulk of India’s population …will have to depend on agriculture for its livelihood. In such a situation social and economic justice calls for a more equitable distribution of the available agricultural land.” (Emphasis added). The origin of the policy of ceilings of landholdings in the postindependence era is the report in 1950 of the Kumarappa Committee on Agrarian Reforms. It influenced all subsequent debates as well as legislation relating to ceilings over several five-year plan periods prior to and after the Green Revolution. The Committee evolved three norms for holdings sizes: Basic, Economic and Optimum. The economic holding was defined as one that

Development Strategy: The State and Agriculture Since Independence

281

would, based on the prevalent agro-economic conditions, afford a reasonable standard of living to the farmer and his family, provide full employment to his family, and a pair of bullocks. Under the assumption that the rehabilitation of the large number of uneconomic holdings would not be feasible, the committee defined a smaller holding than an economic holding, called “basic” and deemed it viable. While viability considerations determined the minimum size of holding as the Basic holding, social justice considerations led the committee defined an upper limit or ceiling as three times the size of Economic holding, which it called the Optimum holding. In effect, the committee expected holdings below the Basic holding to be exempt from any land ceiling laws and only land above the Optimum holding was to be acquired by the state. Khusro11, noting that by the end of the 1950s almost all Indian states had enacted land ceiling legislation, uses the cliché lack of political will for the lack of its effective enforcement. He justified the need for ceiling “for the sake of releasing lands for distribution to the landless and the marginal farmers” (p. xix). He shared the common pessimism about being able to reduce the agricultural population and to alter the land-man ratio in agriculture to any significant extent. He concluded that “The process of a net shift of agricultural population will have to be at work before the impact of the shift can be registered, not to mention the all-important fact that for many years the process may not begin at all. If substantial net shifts . . . are ruled out, it becomes necessary [to impose ceilings] for a more effective absorption of agricultural population within the agricultural sector” (pp. 98-99). Khusro offered his own nine point agenda of an integrated strategy for a new agrarian structure, incredibly assuming away his own assertion about lack of political will for enforcing land reform legislation. The success of the Green Revolution has made the agrarian scene in the nineties to be radically different from what it was at independence. The distinguished agricultural economist, late Professor Dantwala believed that land reform was a lost cause and late Professor Dandekar believed that ceiling laws should be abrogated and that all restrictions on the leasing of agricultural land should be removed. Appu quotes Dandekar as having said “one must admit that [ceiling on land holdings] has totally failed, that it has been circumvented by various means . . . one need not be surprised. The surprising thing is that it was accepted . . .” (Appu1, p. 208). Other distinguished agricultural economists such as V. S. Vyas, Hanumantha and Rao, also recognize the need for a rethinking of land reforms, but surprisingly they believe that there is still a role for ceiling laws.

282

T.N. Srinivasan

3.5. Utter Failure of Land Reforms Even after four decades or more of land reforms, not much has changed in the distribution of land ownership except for the predictable decline in average area owned from 1.78 ha per household in 1961-62 to 0.73 ha in 2003 due to the demographic pressure. However, the proportion of landless households has remained at around 11 percent over four decades (NSS19, statement 2). The gini coefficient of concentration of land ownership has remained virtually constant at around 0.72 (NSS19, statement 3), the share in total number of rural households (of total owned area) owning less than a hectare of land has increased in four decades from 66 percent (8 percent) to 80 percent (23 percent) and that of households owning more than 10 hectares has decreased from 3 percent (28 percent) to 0.5 percent (12 percent) during the same period (NSS, 2003a, statement 3). The distribution of operational holdings, which gives a picture of access to land through tenancy, does not show much of a change either except for the doubling of the number of operational holdings from 51 million in 1960-61 to 101 million in 2002-03 and more than halving of the average area per holding from 2.63 ha to 1.06 ha during the same period again due to demographic pressure. The share of holdings of less than a hectare of land (of total operated area) went up from 39 percent (7 percent) to 70 percent (22 percent) in four decades and the share of holdings of more than 10 ha of area (of total operated area) decreased from 4.5 percent (29 percent) to 0.8 percent (12.5 percent) during the same period (NSS20, Tables 3.2, 3.3 and 3.4). The gini concentration ratio rose from 0.583 and 0.586 respectively in 1960-61 and 1970-71 to 0.629 and 0.641 respectively in 1980-81 and 1991-92 and then fell to 0.624 in 200203 (NSS20, Table 3.5). There was a faster rise in the share of small and marginal holdings as well as their shares in total operated area. Interestingly in West Bengal, the share of marginal holdings (of total area) and the area operated by them rose much more to 88.8 percent (58.3 percent) between 1970-71 and 200203 as compared to corresponding shares of 69.8 percent (22.6 percent) at the All India level. Moreover, the gini concentration ratio fell substantially from 0.494 in 1981-82 (to which it had risen from 0.433 in 1970-71) to 0.430 in 1991-92 and further to 0.313 in 2002-03. This confirms the success of operation Barga of West Bengal. In Kerala there has been a steady fall in concentration ratio from 1970-71. (NSS20, Table 3.6) While the performance of West Bengal and Kerala is expected in this regard, surprisingly Bihar and Jharkand also show a fall in the gini ratio after 1981-82. The two most agriculturally developed states of Punjab and Haryana show the most prominent rise in the concentration ratio

Development Strategy: The State and Agriculture Since Independence

283

since 1970-71. In Punjab the share of land in large operational holdings began to rise after falling between 1970-71 and 1991-92, while in Haryana it fell between 1970-71 and 1981-82 and then rose between 1981-82 and 1991-92 only to fall substantially thereafter. (ibid, Table 3.6) The differences across states in trends in the distribution of operational holdings do not overturn the conclusion that overall, land reform policies had only a very modest impact. It is the pressure of growing number of households cultivating limited land that explains the change in the distribution of ownership and operation. This pressure is the inevitable result of our industrialization failing to develop labour intensive manufacturing to supply domestic and world markets. 3.6. Interventions in Prices and Markets for Agricultural Commodities and Agricultural Inputs Government interventions were extensive, diverse, but without a single overarching objective and coherence. Two observations are in order prior to a discussion of a few of the interventions. First, poverty eradication and ensuring social justice always have been the overarching objectives of policy in India. This does not mean that all policies rationalized by invoking the objectives either were the best compared to alternative policies that could have been used to sub-serve the same objectives or they were effective in reaching the poor for whom they were intended. Second, the objective of attaining self-sufficiency in food grains at the shortest time was rational, given the experience with dependence on food aid from the U.S. under P.L. 480, and of being subjected to the threat of the withholding of aid in order to punish India’s opposition to the Vietnam War at a critical time when India suffered two serious and successive droughts in 1966 and 1967. However it is arguable whether the objective should have been self-reliance in the sense of having the resources to import food on commercial terms when needed rather than doing away with imports altogether as self-sufficiency implies. Also, I would argue that it should have been foreseen (but was not) that the policies (e.g., subsidies on fertilizer, irrigation and electricity, price supports, credit subsidies, etc.) that were introduced in support of the successful adoption and use of the Green Revolution technologies would create vested interests in keeping them in place, even after the need for them had diminished significantly. The interventions relating to PDS and their evolution over time illustrates many of the problems such as the changing objectives, failure in large part to reach the intended beneficiaries, rising budgetary costs, and the implicit costs of distortions created in support of PDS policies. The origins of PDS go back to

284

T.N. Srinivasan

food rationing in metropolitan areas during the Second World War. Over time it has been extended to all of the country, with no means testing and targeting Only recently a means test was introduced, for dividing the households into two groups -- those below the poverty line (BPL households) and above the poverty line (APL households) with the extent of subsidy being higher for the BPL households. The needed supplies for the PDS were acquired at a lower cost than otherwise by cordoning off states (and districts within states) that were deemed food surplus along with a ban on movement of food grains by private traders between states (and between districts within states). The central government set the quantities of food grains to be procured for the PDS and the procurement pricesd. Until years after the Green Revolution, procurement prices were below the post-harvest market prices so that farmers faced an implicit tax if they sold to the system. The PDF therefore acquired its supplies through a system of compulsory levies on wholesale wheat traders and rice millers. A system of minimum support prices (MSP) at which the government stood ready to purchase any amount offered was also instituted so as to protect farmers from a price collapse in bumper harvest years, and also to encourage them to adopt yield raising new technology without fear of a price collapse. MSPs were set below procurement prices. But over time, the distinction between the two disappeared and in effect, the government stood ready to and in fact did purchase whatever was offered at the announced procurement prices. Inevitably, this led to political pressure, which the government rarely resisted, for raising procurement prices, and for lowering quality standards for the grains to be procured. Food Corporation of India FCI, the public sector agency, procured the grains, stored them, and transported them for distribution. The cost of storage and transportation as well as the overheads of FCI were added on to the procurement price to determine the “economic cost” of procurement. The prices at which food grains are issued by the Central Government to states for distribution to BPL and APL families are the “issue prices,” with the Central Government reimbursing the difference between the economic costs and issue prices to the FCI. This is the food subsidy, which amounted to 0.63 percent of GDP in 2006-07. Inefficiency in FCI and increases in procurement prices raise the subsidy cost at unchanged issue prices. d

The Agricultural Price Commission recommends the procurement prices. To save space I will not discuss the add-ons by states to the procurement prices announced by the Centre and the procurement by states.

Development Strategy: The State and Agriculture Since Independence

285

The second objective of the procurement system is to ensure price stability through the maintenance of buffer stockse. With government purchasing any amount offered at minimum support prices, the stocks with the government soared in years of bumper harvest. In July 2001 stocks of rice and wheat rose to 61.7 million tonnes as compared to the operational stock of 24.3 million tonnes for the PDS. With rising stocks, costs of storage mounted and forced the government to allow exports of food grains by private traders with export subsidies as needed. By the same token at times when stocks were low there was pressure on market prices forcing government to allow imports by private traders, as in 2006-07. The purely domestic policies of the PDS, buffer stock for price stabilization and ever increasing minimum support and procurement prices, adversely affected international trade policies of exports and imports of food grains. A superior policy compared to the PDS of cash transfer to the poor to enable them to purchase adequate food and other essential commodities, was not considered seriously because of the belief that non-poor would claim to be poor to take advantage of the transfer. It was not examined whether such leakages would be much more than the documented leakages in the current PDS from the use of bogus ration cards and diversion of PDS supplies. In any case, as has been documented in many studies, for well known reasons significant numbers of the poor do not use the PDS in many parts of the country. Yet, the vested interests in maintaining the current PDS and indeed expanding it are far too strong politically to abolish it in favour of much better alternative safety nets for the poor. Ever since independence there have been many policy interventions aimed at raising agricultural output and yields, particularly of food grains, With their number and intensity of interventions increasing significantly after the onset of the green revolution with its HYVs. Although deriving a large output increase from the cultivation of HYVs required assured supply of water and substantial use of chemical fertilizers, there were no economies of scale per se in their cultivation. Also, millions of small and marginal farmers and tenants could benefit from adopting HYVs, provided they could afford to purchase the necessary inputs, invest in irrigation, able to sell their products without heavy transaction costs and did not face a price collapse in case they produced and marketed large amounts. But because few small and marginal farmers had the ability to store their output beyond the harvest time and to purchase inputs on a e

The stock of food grains was also used for the food-for-work and a few other poverty alleviation programmes.

286

T.N. Srinivasan

cash basis, access to credit was important. Appropriately designed interventions for subsidies, credit, and in rural infrastructure and also irrigation for a limited period of time would have enabled millions of farmers and tenants in areas with a reasonably assured supply of water to adopt HYVs, raised output and improved the distribution of incomes. Unfortunately the interventions were not well designed. First, all farmers regardless of the size of their holdings were entitled to the various subsidies. Thus, the fact that only the small and marginal farmers faced constraints in adopting the green revolution technology did not play any role in the design of interventions. Second, in part because the interventions were not targeted and in part because of their weaker positions in the rural power structure, the small and marginal farmers did not get their due share in the subsidies. Third, because the subsidies were not targeted, their overall budgetary cost was larger even allowing for the lesser participation of small farms. Fourth, the politically stronger large farmers acquired a vested interest, and exercised it effectively, for continuing and increasing subsidies. Fifth, the electricity subsidies, and not charging use cost for surface irrigation, exacerbated the negative externalities from falling water tables of common underground pool, rising water-logging and salinity because of poor drainage and pollution of drinking water from residues of fertilizers and pesticides. Sixth, once farmers adopted the new technology and realized its benefits in their own fields, the subsidies could not be, and were not phased out, because the powerful vested interests in them. The net result was that the budgetary costs of the subsidies rose over time, with the gross subsidy on electricity sale to agriculture alone amounting to a little over 0.5% of GDP in 2007. Seventh, the administrative corruption involved in the dispensation of the subsidies was significant. Eighth and last, the National Rural Employment Guarantee programme, now being extended to the entire rural population, guarantees 100 days of unskilled wage employment supposedly on public works to each rural household opting for it, with a budgetary allocation of roughly 0.25% of GDP in 2006-07. Evidence thus far on the functioning of the programme is mixed: in some districts there have been large leakages of the budgeted amount through corruption. In others, poor have benefited. Even if the scheme had been uniformly successful, it is just a safety net for the poor, and certainly not a means for them to climb out of poverty once and for all. To sum up, the distributions of land ownership and operational holdings have not changed substantially in spite of land reforms. The landless together with owners (or operators of) of small and marginal holdings continue to be the

Development Strategy: The State and Agriculture Since Independence

287

overwhelming majority of rural households. The area under tenancy has remained small and the impact tenancy reform has been extremely modest. The myriad subsidies to agriculture continue although their rationale has eroded over time and their budgetary costs soared. These facts have become common knowledge. Even an outside observer, such as Alan Greenspan8 (p32), has noted that “Growth of agricultural productivity has slowed since the 1980s…. a highly subsidized government-directed agriculture that prevents market forces from adjusting acreage usage is the main culprit. The government in recent years has expended more than 4 percent of GDP on subsidies mainly on food and fertilizers, which state subsidization of power and irrigation has added measurably more.” Realist that he is, Greenspan remarks “Regrettably, the dismantling of large farm subsidies seem no more likely in Delhi than it does in Paris or Washington.” He could have added that while Paris and Washington can afford them, Delhi cannot. The most conspicuous failure of India’s development strategy is the failure to create adequate off farm jobs. Even as late as 2004-05 over 60 percent of usually employed rural males and even higher proportion of rural females are employed in agriculture. To quote Greenspan8 (p320) again, “For India to become a major player in the international arena that it aspires to be, it will need to build factories that entice a very large part of its agricultural workers to urban enclaves, to produce labour-intensive exports the time honored path of the successful Asian Tigers and China.” 4. Agricultural Stagnation, Indebtedness, External Opening and Farmer Suicides 4.1. Deceleration of Growth in Agricultural Output The expert group has claimed that “the most important manifestations of the crisis are deceleration of agricultural growth combined with increasing inefficiency in input use thereby affecting the profitability of agricultural production. …, growth rate of GDP from agriculture decelerated from 3.08 percent during 1980-81 to 1990-91 to 2.61 percent during 1992-92 to 2002-03 at 1999-2000 constant prices. The annual growth rate for all crops taken together decelerated to 1.58 percent during 1990-91 and 2003-04 from a growth rate of 3.19 percent during 1980-81 to 1990-91 (EGAI6, p23). The group, headed by a prominent econometrician, did not elaborate the deceleration is indeed a statistically significant. A comparison of the growth rate in the decade starting in 1980-81 with that in the decade starting in

288

T.N. Srinivasan

1992-93 is arbitrary, except that the reform process started in a systemic way in 1991. A purely econometric analysis of the output that fluctuates from year to year because of weather would have looked for any structural breaks in the series. The group did not do such an analysis and also did not comment on the deterioration in the quality of agricultural statistics relating to area, yield and production, as noted by the National Statistical Commission (2004). Vaidayanathan28 provides a more sophisticated analysis of the data series He subdivides the time span into three: the period 1950-70 covering the pre and the early phases of green revolution, the period 1970-87 covering dynamic phase of the Green Revolution and the liberalization period, 1987-2004. He fits linear quadratic and log-linear quadratic trends to the data series of real gross value added (GVA) in agriculture from National Accounts Statistics (NAS), of gross cropped area (GCA) from Ministry of Agriculture and of yield per hectare of land, defined as the ratio of output to GCA. He finds that the trend growth rate of output per annum was 2.4% during 1950-70, 2.7% during 1970-87 and 2.5% after 1987. In effect, there was no evidence of a significant deceleration in growth rate of output since 1987, although there is some evidence of an increase in volatility of output over time. Growth in crop area decelerated in each sub-period, but with different pace of deceleration A progressive and sharp increase in volatility around the trend was observed. Vaidyanathan comes to the appropriately cautious conclusion that, “Based on these estimates the current concerns about a sharp deceleration in growth of output and yields in the past decades would seem misplaced”. The trends in official indices of gross cropped area, production and yields tell a different story for the period 1987-2004. The annual rate of growth of index production was estimated at around 2.98% in 1950-70 and a lower 2.57% in 1970-87 with no significant trend during 1987-2004. Growth in area showed deceleration within and between the first two periods, again with no significant trend in the last period. Vaidyanathan finds that the correlation between the series of GVA and indices of production, etc., put together by the Ministry of Agriculture was higher than 0.95 during the first two periods, but dropped sharply in the last period. The GVA, by definition, is gross value added, not gross output. There are well known issues relating to the deflation of nominal value added to arrive at real GVA. The Ministry’s indices are derived by weighting gross output of each crop by its share in gross value of output at base year prices, and therefore the series of real GVA and of index of production are not strictly comparable. However, the fact that the correlation between them dropped in the third period

Development Strategy: The State and Agriculture Since Independence

289

(1987-2004) after being high in the first two, suggests that the problems with data gathering and measurement which always existed, have apparently worsened in recent years, as alluded to by the National Statistical Commission. Vaidyanathan28 is again absolutely right in saying that “pending a careful examination of the reasons for the [growing] differences between the series and the possibilities of reconciling them or assessing their relative merits, there is clearly warrant for caution in making inferences about recent trends in agricultural growth.” 4.2. Indebtedness of Farmers The decadal surveys by the NSS of assets and liabilities provide a rich source of data on indebtedness of farmers and their access to institutional credit. The expert group drew extensively on this body of data. I will also very selectively draw on it to make a few points on rural indebtedness, sources of borrowing and interest rates in particular. First, the share of cultivators among rural households has steadily declined from 72.4% in 1971 to 59.7% in 2002 (NSS17, statement 2), with significant interstate variations in the decline. Second, land continues to be the largest component (68% in 2002 and 69% in 1971) of the asset portfolio of cultivators. Interestingly, non-cultivator households have increased the share of land in their portfolio from 32.3% in 1971 to 38.2% in 2002 (ibid, statement 2). Third, although following the nationalization of most commercial banks in 1969, their rural branches increased substantially, and after liberalization many new financial products were offered by banks, still the share of financial assets in the portfolio of cultivator and non-cultivator households did not increase from their low values. However because the share of non-cultivator households in rural households with higher share of financial assets in their portfolio increased, the share of financial assets in the portfolio of all rural households rose from 1.1% in 1971 to 2.2% in 2002. Fourth, the incidence of indebtedness (ignoring some minor issues of comparability over time), having drastically declined from 43% in 1971 to 20% in 1981, has increased slowly since then to 27%. There are substantial interstate differences in the extent of indebtedness, but the pattern of change over time is broadly similar across states. There is no evidence that the incidence of indebtedness rose much more rapidly in the decade of the nineties than it did in the decade of the eighties (ibid, statement 29). However, the debt-asset ratio of rural cultivator households having declined from 4.13% in 1971 to 1.61% in 1991, increased in the nineties to reach 2.49% in 2002 (ibid, statement 35).

290

T.N. Srinivasan

Rural households borrowed more than they repaid in 1971-72, 1981-82, 1991-92 and 2002-03, with differences increasing substantially, from being relatively small in 1971-72 (NSS16, statement 2R and 3R). The proportion of households reporting cash borrowing, after falling from 29.3% (23.4%) for cultivators (non-cultivators) in 1971-72 to 20.6% (16.7%) in 1981-82, began rising steadily to 22.4% (18.4%) in 2002-03. However, there is no jump in the proportion between 1991-92 and 2002-03. The proportion of cultivator households reporting cash repayments fluctuated without trend. However, the proportion of non-cultivator households reporting repayments increased steadily. The expert group (EGAI6, Table 3.2) reports that the share of institutional sources in the debt of cultivator households rose from 7.3% in 1951 to 61.1% in 2002, the most dramatic increase being from 31.7% in 1971 to 63.2% in 1981 after the nationalization of banks in 1969. Since 1981 it has fluctuated without trend. The share of moneylenders, having fallen from 69.7% in 1951 to 16.1% in 1981, began increasing thereafter, particularly in the nineties from 17.5% in 1991 to 27.8% in 2002. Also debt incurred for productive purposes on farm and non-farm businesses, after having risen from 40.1% in 1961 to 71.6% in 1981, has fallen to 67.9% in 2002. Of the 8.7% increase in the proportion of nonproductive debt between 1981 and 2002, 7.7% was accounted for by increase in household expenditure. Clearly, both the rise in the share of moneylenders in debt after 1991, and in the increase in the incurring of debt for household expenditures are undoubtedly disquieting. The incidence of indebtedness also increased after 1981, although there is no evidence of acceleration in the increase in the nineties. However, the debt/asset ratio of rural cultivator households, though it also increased in the nineties, is still very modest at 2.49% in 2002. In sum, the state of farmer indebtedness in the aggregate has indeed deteriorated in the nineties. There are also substantial interstate variations in almost every dimension of indebtedness and its trend over time. Unfortunately, the expert group did not attempt a causal analysis, nor to the best of my knowledge has anyone else. The proportion of interest-free cash debt has declined from 18% in 1981 to 8% in 2002. The dominant share with a simple interest rate has remained stable at around 69%, with only 25 % at a rate of interest of less than 15% per year. Debt at compound rates of interest has doubled from 11% in 1981 to 21% in 2002, with 10% at interest rates between 10% and 15%, and another 10% at interest rates above 15%. Interest rates on rural cash debt from non-institutional sources were basically three: 18% free of interest, 33% at rates below 20% and

Development Strategy: The State and Agriculture Since Independence

291

25%, and 40% at 40% or above. Interest on institutional debt was concentrated at two levels: 57% of debt was at rates between 10% and 15% and another 34% at rates between 15% and 20%. Clearly, the growth of the share of moneylenders in the debt after 1981 and the high rates of interest they charge make the evolving debt scene even more disturbing. 4.3. Agricultural Trade Liberalization The expert group claims that “Agricultural trade has been gradually liberalised beginning with mid-1990s. All-India product lines have been placed under Generalised System of Preferences (GSP). By 2000, all agricultural products were removed from Quantitative Restrictions (QRs) and brought under tariff system. Canalisation of trade in agricultural commodities through state trading agencies was virtually removed and most of the products are brought under Open General Licensing (OGL).” EGAI6 (p. 30) This is a complete misunderstanding of India’s trade policies. First, the term GSP, used in GATT/WTO, denotes generalized system of tariff preferences granted by developed countries on their imports of a subset of commodities from a number of developing countries. It has nothing to do with India’s tariff regime, unless the group has in mind the preferences that India grants to least developed countries. On QRs, what the group must have had in mind was that in the Uruguay Round Agreement on agriculture, all restrictions on agricultural imports were converted into their presumed tariff equivalents. Many WTO members including India took advantage of this process to set tariff bounds on agriculture at very high levels, so that with these high bounds and the reduction in them that they committed to as part of the agreements had no effect on actual applied tariffs on a most favored nation (MFN) basis. For India, the simple average bound tariffs by India on agricultural products was a whopping 114.2%, while the actual average applied level on a MFN basis in 2005 was only 37.6%. Nearly 90% of the bound tariff rates exceeded 50% (WTO31). Since the government has the freedom to raise applied tariffs to their bound levels at its discretion, a large gap between applied and bound levels creates uncertainty about trade policy. In its Trade Policy Review Report on India in 2007 (WR/TPR/S/182), the secretariat of the WTO refers to this uncertainty and notes that during the period of the review India substantially raised tariffs on 27 agricultural products and that India’s use of state trading for food security, marketing and domestic supply seasons is unchanged. Imports of wheat, rice, maize and sorghum continue to be canalized through Food Corporation of India as of April 2006. Of course, the government can and does allow exports and

292

T.N. Srinivasan

imports by private traders, as it thought fit. The expert group simply asserted that canalization was removed without saying when this happened. As recently as in 2001 when mounting stocks led the government to allow exports of food grains, private traders needed permission to export. In 2007, to go by media reports, government has “allowed” wheat to be imported. The variation in trade policy with respect to vegetable oils is another example of short-term considerations influencing them. An appropriate characterization of India’s trade policy with respect to agriculture is not what the expert group claimed. It is primarily driven by shortterm domestic price trends, with a rise in the domestic price of onions or of raw cotton leading to a ban on their exports, or rise in domestic prices (or stocks) of rice or wheat leading to their import (export). There is as yet no long-term liberalized trade in agriculture. The expert group was completely wrong in concluding that agricultural trade liberalization contributed to the crisis. 4.4. Investment in Agriculture Gross Capital Formation in agriculture (GCFA) as a percent of agricultural GDP, both at current prices rose steadily from 10.3% in 2000-01 to 14.1% in 2005-06, having fluctuated around an average of 9.5% between 1990-91 and 2000-01 and 9.5% in the nineties before then.(EGAI6, Table 1.11) Because of the more rapid growth in non-agricultural GDP, GCFA as a proportion of GDP declined from 1.92% in 1990-91 to 1.37% in 1990-00 according to old GDP series and from 2.2% in 1999-00 to 1.9% in 2005-06 according to new GDP series (MOF12, Table 8-19). The decline had begun in the 1980s when rapid growth in non-agricultural GDP began. The share of public sector in GCFA increased from around 18% to 24% between 1999-2000 and 2005-06, having declined from its high value of 43.2% in 1980-81. (EGAI6, Table 1.11) The decline in public sector GCFA is partly due to the decline in total public sector investment as a share of GDP during the period from the late 1980s. GCFA as a proportion of GDP did not decline in recent years, largely due to rises in private investment, which is widely believed to be more production than private investment. This suggests that the argument that decline in GCFA is a contributory factor to the crisis is not consistent with the data. To sum up, among the contributory factors of the agrarian crisis cited by the expert group, only agricultural indebtedness has some firm empirical support. Others, including deceleration in output growth are not firmly established. Per capita net availabilities of cereals and pulses have fluctuated with no pronounced downward trend (MOF12, pp. 5-21) and did not decline.

Development Strategy: The State and Agriculture Since Independence

293

The evidence of a decline in terms of trends for agriculture asserted by the expert group is unconvincing. In fact the ratio of the wholesale price index of manufactured products to the wholesale price index of agricultural products has been declining since 1993-94, suggesting an improvement of agriculture’s terms of trade (MOF12, pp. S-64). The case of trade liberalization and decline in public investment in agricultural having contributed to the crisis is not supported by the data either. 4.5. Farmer Suicides The data on farmer suicides analyzed presented by Nagaraj14 and cited in Sainath22-25 are indeed tragic and alarming. However, Nagaraj14 does not include a deeper-causal analysis. A number of papers on the suicides have been published in the Economic and Political Weekly as well. EGAI6 (Chapter 4) also discusses the available data. Among the many studies, the only study that attempts an analysis of the factors that might be associated with the likelihood of an individual committing suicide is IGIDR9. The study refers to sociological (e.g., precipitating items) and neurological (e.g., predisposition to suicide) aspects of suicide, but its empirical analysis touches on only a few socioeconomic factors that might be involved but not in any depth. No information on neurological aspects was available to the analysts. The methodology of the study was deliberated at a workshop in which economists, sociologists, a psychiatrist, bureaucrats and media persons participated. It was largely based on a primary survey consisting of three components: household interviews, focus group discussion and village level information. The analysis compared a control group of households that had not experienced a suicide but were similar to the households that had experienced a suicide, in terms of ownership of land and other assets. In all, 106 control households were identified in 103 villages. The households that had experienced a suicide cannot obviously be a random sample of the households from the villages studied, nor can the control group of households, since they were in fact chosen to be comparable to the other households. Since most findings of the study are based on a comparison of the two groups of households, because of the non-random aspects of the samples one cannot generalize of the results of the comparison to the population of households of the study villages. It turned out that cotton was the traditional cash crop in the selected districts. The study claims that profitability of cotton cultivation has been declining over the years and attributes the decline to high subsidies by the USA

294

T.N. Srinivasan

leading to price distortions, low import tariffs in India and the failure of the monopoly cotton procurement scheme in Maharashtra. “Withdrawal of the state is evident from declining public investment in agriculture, poor government agricultural extension service, diminishing role of formal institutions in rural financial markets among others.” Unfortunately, the study does not substantiate any of these claims. The age-adjusted suicide mortality rate (SMR) per 100,000 of population has remained in the range of 20-21 for males since 2001 while that for females has been declining since 1999. However, SMR for male farmers trebled from 17 to 53 between 1995 and 2004. Difference in incidence of suicides across caste and size-class of land holding are not statistically significant, though the rates for scheduled caste and scheduled tribe and for marginal and small farmers are slightly higher. Based on police records of cases of suicide for males (females), 31 percent (41 percent) of suicides are attributed to family problems, 24 percent (20 percent) due to illness other than insanity, and 23 percent (20 percent) due to miscellaneous reasons. Economic problems accounted for 11 percent of the suicides of males and 3 percent of females from small farm households. The study is rightly cautious not to over-interpret these data since attributing to a single cause of an event which could have many interrelated causes could be highly subjective and can lead to measurement biases and errors. Of the fourteen risk factors identified in the 111 suicides studied, the most frequent (87%) was indebtedness along with the associated harassment for repayment of loans, with next most frequent being deterioration in social status (74%). Crop failure was associated in 45% of the suicides. Most of the remaining 11 factors related to personal or family problems including illness. At a minimum (maximum), at least 2 (at most 9) of the 14 factors were associated with the 111 suicides, the average being 4.8. In a step-wise logistic regression of log-odds of a household experiencing suicide estimated in the study, outstanding debt and not owning a bullock are the two variables that raise the odds significantly. However, when the regression is restricted to household pairs with similar land holdings, only debt per acre is significant. On the other hand, when the regression restricted to same caste household pairs, not owning a bullock and family size are statistically significant. If we restrict the regression to pairs of households with similar land holdings and the same caste, only not owning a bullock is significant. Unfortunately, not much can be inferred from the study about the agrarian crisis as being a contributory factor to rising suicides. After all any one, farmer or one with a non-farm occupation can accumulate debt for many reasons that

Development Strategy: The State and Agriculture Since Independence

295

he or she cannot service. While not owning a bullock is a serious constraint for a farmer, it is often the case that farmers sell off their bullocks to repay debt or when faced with a serious adverse shock to their income. Contrary to simplistic assertions, including by the expert group, that the burden of debt and trade liberalization are the main causes of suicides, Vaidyanathan30 analysis shows that no more than 20% of households were indebted even in the states where suicides were high or could have been affected by import liberalizations. 5. Summary and Conclusions It is widely believed that India is in the midst of an agrarian crisis. An expert group on Agricultural Indebtedness appointed by the Ministry of Finance and Chaired by the eminent econometrician Professor R.Radhakrishna, former Director of the Indira Gandhi research Institute for Development Research, Mumbai, firmly asserted that “Indian agriculture is currently passing through a period of severe crisis…the crisis has assumed a serious dimension since the middle of the 1990s. One of the tragic manifestations of the crisis is the large number of suicides committed by the farmers in some parts of India.” (EGAI6, p 13) The expert group and contributors to the growing literature on the crisis have attributed the crisis to several factors: the role of systemic economic reforms of 1991; the opening of the Indian economy to external competition and investment after decades of insulation; the impact of India’s implementation of its commitments under the Agreement on Agriculture of the Uruguay Round of Multilateral Trade Negotiations; neglect of Agriculture in the planning process since the mid 1980s; the decline of public investment in agriculture; slowing of the rate of agricultural output; stagnation of yields per hectare of land and growing indebtedness of farmers. The empirical evidence offered in support of most of these factors is weak, if not nonexistent, with one notable exception, namely growing farm indebtedness. First, the process of reforms did not directly impinge on the agricultural and rural sectors. However, the reforms could have affected the two sectors indirectly – for example, fiscal reform and consolidation could have affected public investment in agriculture adversely. In fact, total public investment as well as agricultural investment as a share of aggregate GDP fell in the 1990s. However fiscal consolidation in the form of reduction in fiscal deficits as a per cent of GDP happened only in the first half of the nineties after which the process was reversed. Moreover, faced with the task of reining in

296

T.N. Srinivasan

growth of public expenditures, the authorities chose to do it through cutting capital rather than current expenditures. For example there was no significant reduction in subsidies as a per cent of GDP. This being the case, it is not convincing to argue that fiscal reform forced a reduction in public investment in agriculture. In fact, agricultural investment as a per cent of agricultural GDP rose since 2000-01, so that the reduction as a per cent of aggregate GDP was due to faster growth in non-agricultural GDP relative to agricultural GDP. Also the growth of private investment in agriculture in large part mitigated the effects of the fall in public investment. To the extent that private investment is more productive than public investment, even if it did not substitute rupee for rupee of public investment, it could still have largely, if not more than, offset the output loss from the reduction in public investment. The expert group completely misunderstood India’s commitments under the Agreement of Agriculture of the Uruguay Round. Contrary to its assertion, according to the WTO’s Fourth Review of India’s Trade Policy early in 2007, even as of April 1 2006, canalization and state trading covered foreign trade in rice, wheat and other food grains and some vegetable oils. Although quantitative restrictions on imports were removed in 2002, after an adverse decision against their use by the WTO’s Dispute Settlement Mechanism, many agricultural imports are included among the 300 items whose imports are being monitored since they are considered sensitive. WTO’s Review points out that tariffs continue to be used in support of the overall goals of food self-sufficiency and price stability. Tariffs are raised or lowered from time to time depending on domestic market conditions. India bound its tariffs on agricultural products at a very high level in the Uruguay Round (simple average of 114.2%). With the average applied MFN tariffs in 2005 being 37.6%, there was plenty of room to raise tariffs (up to the bound level) or down (to zero) according to the Government’s discretion. Under the circumstances, it is inappropriate and a wild exaggeration, to claim that increased imports and the resulting competition from the implementation India’s commitments at the WTO have contributed to the agrarian crisis. Although agricultural output growth had slowed in the 1990s as compared to the 1980s, there is no particular reason for this decadal comparison other than the fact that reforms were initiated in 1991. A more convincing periodization of the period 1950-2004 by Vaidyanathan28, 29 leads to different conclusions about growth slow down during 1987-2004, depending on whether data on real gross value added by agriculture (GVA) from National Accounts or the Index of Agricultural production of the Ministry of Agriculture is used in the trend

Development Strategy: The State and Agriculture Since Independence

297

analysis. While GVA data do not support the hypothesis of growth deceleration either in output or yield per hectare of land in the period 1987-2004, the analysis with the Index does. As Vaidyanathan rightly observes, given the conceptual differences between the two data series and the well known deterioration in agricultural data collection and compilation, without an in depth analysis of the differences between the two data series, it will be premature to conclude that the hypotheses of growth deceleration and its contribution to the agrarian crisis, are well founded. Moreover, the data do not show any downward trend in the nineties in the per capita availability of cereals and pulses. The NSS data show that the incidence of indebtedness among farmers has increased slowly from 20% in 1981 to 27% in 2002, although there is no evidence of a faster rate of increase during the decade of the nineties. Moreover, the share of the institutional sources of credit has been fluctuating since 1981 after rising dramatically from 31.7% in 1971 to 63.2% in 1981, in part due to the expansion of bank branches in rural areas after nationalization of banks in 1969. Unfortunately the share of money lenders, having fallen from 69.7% in 1951 to 16.1% in 1981 began rising thereafter reaching 27.8% in 2002. Debt incurred for production purposes also declined after 1981, most of the decline being accounted for by increase in debt-financed household expenditure. Thus the rise in the incidence of farm indebtedness, the share of money lenders as a source of debt finance and in the use of debt for financing household expenses is disquieting. Again without a sound casual analysis of these trends, which no one appears to have attempted thus far, one can neither draw a firm conclusion, as the expert group has done, that they contributed to the agrarian crisis, nor dismiss the hypothesis that they did. The contention of this paper is that the current agricultural situation, whether one describes it as a crisis or not, is the inexorable consequence of India’s development strategy since 1950 until the systematic reforms of 1991. This strategy completely ignored the lesson of economic history that successful development involves transformation of the economic structure through massive shift of work force and population away from low productivity use in agriculture and primary activities and into manufacturing and other tertiary activities. India’s industrialization focused on capital intensive import substitution across the board with emphasis on the development of heavy industries and insulation from world markets. This inevitably led to more than 60% of the rural work force still being engaged in agriculture and allied activities in 2004-05 after five decades of industrialization.

298

T.N. Srinivasan

This contention is confirmed by the history of agrarian reforms since independence. Other than the abolition soon after independence of intermediary tenures, attempts at reforming land ownership and tenancy have not been very effective. Land reforms did not significantly change the distribution of land ownership and tenancy. Concentration in ownership holdings did not change and that in operational holdings has remained unchanged since 1970, after having risen somewhat from its level in 1960-61. The set of interventions introduced in the seventies to support the adoption of high yielding varieties, primarily of rice and wheat, was driven by the need to attain self-sufficiency in food grains. While successful in achieving it, the interventions by their very design created vested interests that successfully lobbied for their perpetuation and increase. The net result was an increased the fiscal burden of subsidies to food, fertilizers and electricity. The interventions associated with the PDS for food grains, motivated primarily by consideration of social justice and providing access to food for the poor, were not cost effective as compared to other policies such as cash transfers for achieving the same objectives. The political economy of PDS also precluded reforms of the system to make it more cost effective. Most importantly, the two sets of interventions associated with the PDS and self-sufficiency in food promoted concentration on cereal agriculture rather than diversification into other crops. The net result is that while consumer demand is shifting away from cereals, the policies on the supply side are still focused on cereals. The focus of future policies has to be to promote labour intensive industrialization in rural and urban areas to supply rising domestic and global demand. Although the growth of the service sector, particularly that of IT enabled services, has been impressive, it would be fool hardy to believe that service sector growth alone would bring about the needed economic transformation in the long run. Of course in short to medium run, effective policies for increasing agricultural productivity and relieving farm distress are essential. The Eleventh Five Year Plan has increased the allocation of resources for agriculture significantly as compared to the tenth plan. An increase in resources, if is not accompanied by policy changes that make the use of resources more effective than in the past would obviously limit what can be achieved with the increase. Vaidyanathan’s29 notes on the agricultural strategy in the eleventh plan emphasizes this point, through he does not phrase the issue in this way. He points out that the plan’s target of achieving a 4% annual growth of agricultural output is very unrealistic as compared to the growth

Development Strategy: The State and Agriculture Since Independence

299

record of 1950-2004 or of recent decade or no. Of course, past growth record is no predictor or a constraint on a more ambitious future growth target, if the latter is associated with the new set of policies different from the past. After all, India was able to accelerate its GDP growth significantly to over 9 percent during 2005-06 to 2007-08, as compared to the so called “Hindu rate of growth” of around 3.75% per year during 1950-80, Vaidyanathan suggests that the policies proposed in the eleventh plan do not differ radically from those of the past, but are merely “more of the same”. He stresses that institutional weaknesses have limited the effectiveness of past policies. Availability of adequate moisture is a major constraint on the use of yield raising (land augmenting) technologies, particularly in rain fed areas accounting for a large share of gross cropped area. Land quality has deteriorated because of imprudent use of water in irrigated areas (in large part due to policy created incentives). For both reasons a shift away from current irrigation practices and from the pattern of investment in creating irrigation capacities is called for. The Eleventh Plan focuses mostly on accelerated expansion of irrigation facilities along the same lines as in the past. Vaidyanathan argues that in the past their implementation has been plagued by serious deficiencies due to lack of accountability and transparent implementation. The plan offers no evidence to believe that these deficiencies have been eliminated. Vaidyanathan points out that “the notion high output prices and input subsidies protect the incomes and livelihood is misplaced. The best way to raise rural in assess and employment is through increasing agricultural productivity” Although policies to increase agricultural productivity are no doubt desirable, realistically, raising productivity and incomes of the rural poor consisting of non-viable farmers and landless cannot be brought about without enabling them to shift from agriculture by creating more productive off farm activities, many of which could be rurally based. With fewer unviable holdings and workers on land, restructuring of cultivation on larger, consolidated and viable holdings would be possible thereby raising productivity. The generation of off-farm employment activities at an adequate pace and amount will not come about without a change in industrialization and foreign trade policies. Also without efficiently functioning land markets and distortion free markets for agricultural inputs and outputs, the small land owners will not be able to divest their land holdings and the needed restructuring of farms into viable holdings will not come about. Had there been an efficient system of land ownership records and a network of thick and efficient markets for land, anyone wishing to acquire land currently in one private wise (for example in farms) for

300

T.N. Srinivasan

another private use (say in SEZs) would have been able to do so at a fair market value without the state having to intervene in a heavy handed and inequitable fashion. Slogans such as no farm land should be allowed to be put to any other use are mere slogans without rationale. A rational policy would insist on land being put to its best social use either for farming or elsewhere. Even after the reforms of 1991 that decisively broke away from insulation from world markets, India is still one of the most protected among developing countries. Our bound tariffs are high and much above their applied levels particularly compared to China. FDI inflow to India continues to be far less as compared to China. China has used FDI in its exporting activities to become the second largest merchandise exporter in the world with a share of 8.8 percent of world exports in 2007, while India was the 26th largest with a share of just 1 percent. Even before China was admitted to the WTO in 2001, it had gained a market share in exports of labour intensive manufacturers, such as garments, fabrics and leather products. India’s market shares remained stagnant or declined during the same period. Dooley et al5 note that China had about 200 million unemployed workers. To bring them into the modern labour force and for political stability they estimate 10-12 million new jobs have to be created each year in urban areas, 30 percent of which will be in the export sector. In their global model, China will be able to do this in a world of plausible rates of capital accumulation and rates of return along a smooth adjustment path. Feenstra and Hong7 in their evaluation of Dooley et al5 find that export growth during 1997-2002 accounted for a third of the employment growth of 7.5-8 million workers per year in China. Although during 2002-05 exports grew faster, and in principle explain the entire employment growth, taking into account the employment generation from the rise in domestic demand, especially for investment, exports still accounted for a third of the total employment. The two authors do not say whether part of the investment growth was itself induced by export growth. Still 2.5 million a year or so in new jobs from exports are still large. India’s total exports in a year in recent years have been less than just the increment in China’s export in year! India’s export growth has obviously not been the driver of growth of GDP or employment. To be able to generate rapid growth in off-farm employment, India has to engage in labour intensive rural and urban industrialization, in part focused on supplying export market. This would require further opening to the world markets and FDI, and removal of infrastructural constraints such as power, rural and urban roads and ports. Shifting the policy focus to one of moving millions

Development Strategy: The State and Agriculture Since Independence

301

from the farms to more productive activities off-farm will not only accelerate growth, but will make it more inclusive as well. Merely chanting the mantra of “inclusive growth” as the eleventh plan does, as if it is a new vision, is like chanting mantras for anything else – to seek divine intervention to bring about a change with having to do anything one self. I will leave it to other to judge whether invocatory chants are productive. Acknowledgments I thank an anonymous referee, A. Vaidyanathan, and participants at my presentations at the International Conference on Comparative Development, Indian Statistical Institute, New Delhi, Platinum Jubilee Lecture Series, Indian Statistical Institute, Kolkata and the 10th Annual Conference on Money and Finance in the Indian Economy, Indira Gandhi Institute for Development Research, Mumbai, for their comments. References 1. P.S. Appu, Land Reforms in India, New Delhi, India: Viking Publishing House (1996). 2. V.B. Athreya , The Journal of Contemporary Management Research, 1(1), 118 (2007). 3. P. Bardhan and D. Mookherjee, Land Reform and Farm Productivity in West Bengal, University of California Berkeley (mimeo) (2007). 4. V.M. Dandekar, The Indian Economy, 1947-1992, Volume 1 Agriculture, Sage Publications, India (1994). 5. Michael Dooley, David Folkerts-Landau and Peter Garber, Direct Investment Rising Real Wages and the Absorption of Excess Labor in the Periphery, Working Paper 10626, Cambridge, MA, National Bureau of Economic Research (2004). 6. EGAI, Report of the Expert Groups on Agricultural Indebtedness, India, Government of India, Ministry of Finance (2007). 7. Robert Feenstra and Chang Hong, China’s Export and Employment, Working Paper 13552, Cambridge, MA, National Bureau of Economic Research (2007). 8. Greenspan, Alan, The Age of Turbulence, New York, Penguin Press (2007). 9. IGIDR, Suicide of Farmers in Maharashtra (author, Srijit Mishra), India, (2006). 10. IIAPR, Report of the National Planning Committee (1998). 11. Ali M. Khusro, The Economics of Land Reform and Farm Size in India, Macmillan India (1973).

302

T.N. Srinivasan

12. MOF, Economic Survey 2006-2007, Government of India (2007). 13. A Mitra, Terms of Trade and Class Relations, London Frank Cass (1977). 14. K. Nagaraj, Pesticide Suicides in India, Madras Institute of Development Studies (2007). 15. NCA, Report of the National Commission on Agriculture, Parts I, II, XIII and XV, Government of India, Ministry of Agriculture (1976). 16. NSS, Household Borrowings and Repayments in India during 1.7.2002 to 30.6.2003 (2006). 17. ––––––––––, Household Assets and Liabilities in India, (as on 30.6.2002), Report 500 (2005a). 18. ––––––––––, Household Indebtedness in India, (as on 30.6.2002), Report 492 (2005b). 19. ––––––––––, Household Ownership Holdings in India, Report 491 (2003a). 20. ––––––––––, Household Operational Holdings, Rural India, 2003, Report 492 (2003b). 21. ––––––––––, Household Operational Holdings, Rural India, 1981-82, Report 407 (1983). 22. P. Sainath, Farm Suicides Rising, Most Intense in 4 States, http://www.thehindu.com/2007/11/12/stories/2007111253911100.htm (2007a) 23. ––––––––––, Farm Suicides After 2001 – A Study, http://www.hindu.com/2007/11/13/stories/2007111352250900.htm (2007b) 24. ––––––––––, Maharashtra: ‘graveyard of farmers’, http://www.hindu.com/2007/11/14/stories/2007111453091100.htm (2007c) 25. ––––––––––, One Farmer’s Suicide Every 30 Minutes, http://www.com/2007/11/15/stories/2007111554771300.htm (2007d) 26. A. Sengupta, J.C. Bhattacharya, and M. Chattopadhyay, South Asian Economic Journal, 15(1), 103-130 (2004). 27. T.N. Srinivasan, Was Agriculture Neglected in Planning? in Review of the Indian Planning Process: Proceedings of the Indian Statistical Institute Golden Jubilee International Conference, D. K. Bose, Ed. Calcutta: Indian Statistical Institute, 1986: 33-46 (1982). 28. A. Vaidyanathan, Essays on Agricultural Development (in preparation) (2007a). 29. ––––––––––, Notes for discussion on Agricultural Strategy for the XIth Plan (personal communication) (2007b). 30. ––––––––––, Agrarian crisis: nature, causes, and remedies, in The Hindu, November 8, http://www.hinduonnet.com/thehindu/thscrip/print.pl?file=2006110804491000 .htm&date=2006/11/08/&prd=th& (2006). 31. WTO, World Tariff Profiles, World Trade Organization (2007).

LAND-USE CHANGES AND AGRICULTURAL GROWTH IN INDIA, PAKISTAN, AND BANGLADESH, 1901-2004* TAKASHI KUROSAKI Institute of Economic Research, Hitotsubashi University, 2-1 Naka, Kunitachi, Tokyo 186-860, Japan. E-mail: [email protected] This paper investigates land-use changes in India, Pakistan, and Bangladesh, associates the changes with long-term agricultural performance, and shows the importance of crop shifts in enhancing aggregate land productivity, which is a source of growth unnoticed in the existing literature. The use of unusually long-term data that correspond to the current borders of India, Pakistan, and Bangladesh for the period 1901-2004 also distinguishes this study from the existing ones. The empirical results show a sharp discontinuity between the pre- and the post- independence periods in all of the three countries: total output growth rates rose from zero or very low figures to significantly positive levels, which were sustained throughout the post-independence period. The improvement in aggregate land productivity explained the most of this output growth. To quantify the effect of crop shifts, a decomposition analysis is applied, which shows that the crop shifts contributed to the productivity growth in all three countries, especially during periods with limited technological breakthroughs. The contribution of the crop shifts was larger in India and Pakistan than in Bangladesh. The decomposition results and changes in crop composition are consistent with farmers’ response to comparative advantage under liberalized market conditions.

1. Introduction To halve, between 1990 and 2015, the proportion of people whose income is less than one dollar a day and to halve, between 1990 and 2015, the proportion of people who suffer from hunger are the first two targets of the Millennium Development Goals. Whether these targets will be achieved critically depends on the performance of the South Asian region where the number of the absolute poor is the largest in the world (e.g., according to World Bank (2001), the number of people living on less than one dollar a day in 1998 was 522 millions in South Asia, out of the global total of 1,199 millions). At the same time, the * The author is grateful for helpful comments to Palapre Balakrishnan, M. Mufakharul Islam, Sunil Kanwar, Yukihiko Kiyokawa, Hiroshi Sato, Shinkichi Taniguchi, Yoshifumi Usami, Haruka Yanagisawa, and participants at the ISI International Conference on Comparative Development, December 2007. 303

304

Takashi Kurosaki

three largest countries in the region, India, Pakistan, and Bangladesh, experienced a rapid agricultural production growth in the second half of the twentieth century. In these countries, the agricultural sector is the largest employer of the poor and the domestic food production is highly important in determining their welfare. Then, how was the agricultural growth achieved and why was there stagnation in the first half of the twentieth century? Why was the growth not sufficient to substantially reduce the number of the poor? How was the agricultural transformation related with market development? These are questions that motivated this article to investigate the source of agricultural growth and changes in land use in these countries during the last century. The importance of agriculture in poverty reduction is emphasized in the latest World Development Report as well (World Bank30). More specifically, this paper describes land-use changes in India, Pakistan, and Bangladesh, associates the changes with long-term agricultural performance, and shows the importance of crop shifts in enhancing aggregate land productivity, which is a source of growth unnoticed in the existing literature.a The use of unusually long-term data that correspond to the current borders of India, Pakistan, and Bangladesh for the period 1901-2004 also distinguishes this study from the existing ones on long-term agricultural development in South Asia.b Some of the previous studies on agricultural production in the colonial period deal with undivided India (e.g., Sivasubramonian24, 25, 26), some deal with British India (Blyn4; Guha8), and others deal with areas of contemporary India (Roy22), but very few investigate the case for areas of contemporary Pakistan and Bangladesh in a way comparable with that for India. If we restrict to Punjab and Bengal, there are several studies with comparative perspectives between Indian Punjab and Pakistan Punjab (e.g., Prabha20; Dasgupta6; Sims23) and between West Bengal and East Bengal (Bangladesh) (e.g., Islam11; Boyce5; Rogaly et al.21 1999; Banerjee et al.2). However, the coverage of these studies is limited --- those a

b

Historical records show that agricultural productivity has increased thanks to the introduction of modern technologies, the commercialization of agriculture, capital deepening, factor shifts from agriculture to nonagricultural sectors, etc. This overall process can be called “agricultural transformation,” and the contribution of each of the factors has been quantified in the existing literature (Timmer28). Datasets are newly compiled by the author (Kurosaki16), using government statistics and revising the author's previous estimates. Using the previous versions of these datasets, Kurosaki12 and Kurosaki13 investigated the performance of agriculture in India and Pakistan for the period c.1900-1995, Kurosaki14 quantified the growth impact of crops shifts in West Punjab, Pakistan for a similar period, and Kurosaki15 extended the analysis for India and Pakistan using data until 2004.

Land-Use Changes and Agricultural Growth in India, Pakistan, and Bangladesh

305

investigating the pre-1947 period did not adjust for the boundary changes, while those comparing the areas corresponding to the current international borders investigated the post-1947 period only. Although it is true that the state of Pakistan did not exist before 1947 and the state of Bangladesh did not exist before 1971, investigating agricultural production trends for “fictitious” Pakistan before 1947 and “fictitious” Bangladesh before 1971 would give us valuable insights, since farming is carried out on land, which is immovable by definition. The article is organized as follows. The next section defines the spatial coverage of the analyses and describes long-term changes in land utilization. Section 3 gives an analytical framework to investigate agricultural growth performance and to structurally associate changes in aggregate land productivity with inter-crop reallocation of land use. Section 4 presents empirical results, contrasting the difference in agricultural growth performance among India, Pakistan, and Bangladesh. Section 5 examines the impact of changes in crop mix, which shows that crop shifts did contribute to agricultural growth in these countries. Section 6 concludes the paper. 2. Changes in Land Utilization in India, Pakistan, and Bangladesh In August 1947, the Indian Empire under British rule was partitioned into India and (United) Pakistan. Before 1947, the Empire was subdivided into provinces of British India and a large number of Princely States. The current international borders are different, not only from provincial/state borders, but also from boundaries of districts (the basic administrative unit within a province). The two important provinces of Bengal and Punjab were divided into India and (United) Pakistan with Muslim majority districts belonging to the latter. In the process, several districts in Bengal and Punjab were also divided. Before 1947, agricultural statistics were collected regularly in all provinces of British India. In contrast, statistical information on the Princely States is limited in coverage and missing for many regions. Because of this reason, the classic and seminal study on agricultural growth in the colonial India by Blyn4 examined the area known as “British India,” which covers all British provinces except for Burma (Burma Province became a separate colony in 1937). British India below corresponds to the area thus defined by Blyn4. Table 1 shows decade-wise statistics on land utilization in British India. In 1901/02,c out of 182 million ha of land for which the information was available, c

“1901/02” refers to the agricultural year beginning on July 1, 1901, and ending on June 30, 1902. In figures with limited space, it is shown as “1902.”

Takashi Kurosaki

306

12.3% was under forest and 50.7% was under cultivation. About one fifth of the total cultivated land lay fallow. In 1941/42, these shares were similar, while the absolute acreage of land under forest or land under cultivation increased, at the annual growth rate of 0.54% and 0.33%, respectively. Besides the land under forest or under cultivation, there was a huge area that was not available for cultivation or was classified as cultivable waste. Most of these lands were barren, with very limited vegetation. Table 1 Decade-wise Land Utilization in India, Pakistan, and Bangladesh, 1901-2002. In million ha. Reported Forest Not Cultiva- Current Net area Total area [1] available ble fallow sown area (total of for waste [4] [5] culti[1]-[5]) cultiva[3] vated tion [2] ([4]+[5]) British India 1901/02

182.01

22.33

33.66

33.73

16.08

76.20

92.28

(100.0%)

(12.3%)

(18.5%)

(18.5%)

(8.8%)

(41.9%)

(50.7%)

1911/12

205.31

25.04

42.20

35.95

20.11

82.01

102.12

1921/22

205.52

26.78

39.58

36.45

18.88

83.84

102.72

1931/32

206.88

26.86

37.87

38.49

18.13

85.54

103.67

1941/42

207.25

27.67

36.86

37.32

19.08

86.32

105.40

(100.0%)

(13.3%)

(17.8%)

(18.0%)

(9.2%)

(41.6%)

(50.9%)

0.32%

0.54%

0.23%

0.25%

0.43%

0.31%

0.33%

Growth rate India 1951/52

287.83

48.89

50.17

40.40

28.96

119.40

148.36

(100.0%)

(17.0%)

(17.4%)

(14.0%)

(10.1%)

(41.5%)

(51.5%)

1961/62

305.35

60.84

50.36

36.58

21.23

136.34

157.57

1971/72

304.02

65.41

41.82

34.66

24.53

137.59

162.12

1981/82

304.11

67.35

39.95

31.85

24.17

140.79

164.97

1991/92

304.84

67.98

40.91

29.40

23.83

142.72

166.55

2001/02

305.01

69.51

41.78

27.36

24.95

141.42

166.36

(100.0%)

(22.8%)

(13.7%)

(9.0%)

(8.2%)

(46.4%)

(54.5%)

0.12%

0.70%

-0.37%

-0.78%

-0.30%

0.34%

0.23%

Growth rate Pakistan 1951/52

46.45

1.39

20.75

9.16

3.54

11.61

15.15

(100.0%)

(3.0%)

(44.7%)

(19.7%)

(7.6%)

(25.0%)

(32.6%)

1961/62

50.99

1.68

18.73

12.46

4.85

13.27

18.12

1971/72

53.55

2.83

20.40

11.11

4.77

14.44

19.21

1981/82

53.92

2.85

19.90

10.86

4.89

15.41

20.30

1991/92

57.61

3.46

24.34

8.85

4.85

16.11

20.96

Land-Use Changes and Agricultural Growth in India, Pakistan, and Bangladesh 2001/02 Growth rate

307

59.28

3.61

24.50

9.13

5.67

16.32

21.99

(100.0%)

(6.1%)

(41.3%)

(15.4%)

(9.6%)

(27.5%)

(37.1%)

0.49%

1.91%

0.33%

-0.01%

0.94%

0.68%

0.75%

Bangladesh 1951/52

14.02

1.27

1.93

1.81

0.63

8.38

9.00

(100.0%)

(9.1%)

(13.8%)

(12.9%)

(4.5%)

(59.7%)

(64.2%)

1961/62

14.28

2.22

2.43

0.75

0.41

8.47

8.88

1971/72

14.28

2.23

2.66

0.30

0.85

8.24

9.09

1981/82

14.29

2.14

2.77

0.25

0.55

8.58

9.13

1991/92

14.84

1.89

3.86

0.48

0.63

7.98

8.61

2001/02

14.84

2.58

3.51

0.32

0.41

8.02

8.43

(100.0%)

(17.4%)

(23.7%)

(2.2%)

(2.7%)

(54.1%)

(56.8%)

0.11%

1.41%

1.06%

-3.45%

-0.87%

-0.09%

-0.13%

Growth rate

Notes: The percentage in parenthesis shows the share in the total reported area. “Growth rate” shows an exponential annual growth rates throughout the per-1947 or the post-1947 period. Data sources: For British India (excluding Burma), Agricultural Statistics of India, Government of India, various issues; For India, Indian Agricultural Statistics, Directorate of Economics and Statistics, Ministry of Agriculture, Government of India, various issues; For Pakistan (corresponding to contemporary Pakistan), Economic Survey, Ministry of Finance, Government of Pakistan, various issues. For Bangladesh (corresponding to contemporary Bangladesh), BBS web pages.

Table 1 also shows decade-wise land utilization in India, Pakistan, and Bangladesh after the Partition in 1947. Each series shows statistics for a geographic area corresponding to the current international borders of the three countries. In 1951/52, just after the partition, 17.0% of the reported land was under forest and 51.5% was under cultivation in India, higher than corresponding figures for Pakistan (3.0% and 32.6%). This shows that Pakistan inherited more barren land than India did. The area under forest in Bangladesh was 9.1%, which is between the share in India and that in Pakistan. The area under cultivation in Bangladesh was 64.2%, much higher than in India and in Pakistan. In India and Pakistan, the area under forests and that under cultivation increased substantially throughout the post-independence period. The annual growth rates were higher in Pakistan (1.91% and 0.75%) than in India (0.70% and 0.23%), well above the figures for British India before independence as well. It is worth mentioning that the area not available for cultivation or cultivable waste decreased in India. In contrast, there was no growth of the area under cultivation in Bangladesh. The annual growth rate of “total area cultivated” was negative (-0.13%), but because of a rapid decline of the area under fallow, the growth rate of “net area sown” was close to zero (-0.09%).

308

Takashi Kurosaki

The post-independence expansion of the cultivated area in India and Pakistan was more impressive and the post-independence stagnation of the cultivated area in Bangladesh was turned into an expansion if we take into account the area sown more than once during the agricultural year. Such changes in cropping patterns with accelerating intensity are the theme of the rest of this paper. Regarding cropping patterns and crop output, there are several sources of information, covering Princely States. Utilizing these sources, Kurosaki16 compiled an updated version of the country-level dataset for India, Pakistan, and Bangladesh, covering a period from 1901/02 to 2003/04, and covering the production of principal crops that are contemporarily important in these countries.d The data compilation procedure for the colonial period is explained in detail by Kurosaki16. Data on the areas that are currently in Pakistan and Bangladesh were subtracted from the database compiled by Sivasubramonian24. Information included in the district-level data in Season and Crop Reports from Punjab, Sind (or Bombay-Sind), the North-West Frontier Province, and Bengal, and the province-level data in Agricultural Statistics of India (before 1947) was utilized in the data compilation. The official data on the area and output of several produces for Bangladesh in the pre-1947 period were revised after consulting the “revision factor” estimated by Islam11. 2.1. Analytical Framework To analyze the growth performance of agriculture in the three countries, the gross output values of these crops are aggregated using 1960 prices,e and denoted by Q. As measures for partial productivity, Q is divided by L (the population estimates of India, Pakistan, and Bangladesh) or by A (the sum of the acreage under the major crops covered in this article). As the first step to d

e

For India, eighteen crops are included: rice, wheat, barley, jowar (sorghum), bajra (pearl millet), maize, ragi (finger millet), gram (chickpea), linseed, sesamum, rapeseed & mustard, groundnut, sugarcane, tea, coffee, tobacco, cotton, and jute & mesta. For Pakistan, twelve major crops are covered: rice, wheat, barley, jowar, bajra, maize, gram, rapeseed & mustard, sesamum, sugarcane, tobacco, and cotton. For Bangladesh, fourteen crops are included: rice, wheat, barley, maize, gram, linseed, sesamum, rape & mustard, groundnut, sugarcane, tea, tobacco, cotton, and jute. These crops currently account for more than two thirds of the total output value from the crop sector in these countries and the contribution of these crops was higher in the colonial period. Ideally, the sum of the value-added evaluated at current prices and then deflated using a price index would be a better measure, but the sum of gross output values at constant prices is used as a proxy due to the absence of reliable data on input prices and quantities before independence. The results reported in this paper are insensitive to the choice of base year (1938/39 and 1980/81).

Land-Use Changes and Agricultural Growth in India, Pakistan, and Bangladesh

309

analyze the changes in agricultural productivity, a time series model for Yt is estimated as lnYt = a + bt + ut ,

(1)

where t is measured in years, a and b are parameters to be estimated, and ut is an error term. Equation (1) is estimated for the logarithm of Q, Q/L, and Q/A, by the ordinary least squares (OLS) method. The larger the coefficient estimate for b, the higher the growth rate of production or productivity. The standard error of regression for equation (1) shows variability, because it indicates how variable the output was around the fitted values in terms of the coefficient of variation. Equation (1) can be extended to investigate the difference-in-difference (DID) of growth rates between the countries. Namely, we estimate lnYtk = (ak0 + ak1Dt) + (bk0 + bk1Dt) t + ukt ,

(1’)

for k = I (India), P (Pakistan), and B (Bangladesh), where Dt is a time dummy variable. For example, the DID estimator bI1 - bP1, when Dt is set to one when t is greater than 1947, captures the difference in growth rate changes observed between India and Pakistan after the Partition. Since both regions are inherently different, the potential level of output (captured by ak0 and ak0 + ak1Dt) and the potential growth rate (captured by bk0) can differ. We are not interested in such a difference. Our interest is on the between-country difference in bk1. If the two regions were exposed to similar exogenous changes in environment, technology, and markets, then the DID estimator bI1 - bP1 can be interpreted as the impact of political regime change, i.e., the Partition. If it is not relevant to assume that the two regions experienced exactly the same changes in environment, technology, and markets, then the DID estimator bI1 - bP1 can be interpreted as the net impact of the regime change and changes in environment, technology, and markets. In this paper, the impact of the Partition using the whole sample period (Dt is set to one when t is greater than 1947) and the impact of Bangladesh’s independence using the subsample after 1947 (Dt is set to one when t is greater than 1971) are investigated. The DID analysis contrasting the pre-1947 and the post-1947 periods for areas delineated by the contemporary international borders is the original contribution of this paper, which becomes feasible thanks to the use of the unusually long time series data. In the next step, to capture long-term changes in the crop mix, the Herfindahl Index of crop acreage was calculated. Let Si be the acreage share of crop i in the sum of the principal crops. The Herfindahl Index is defined as

310

Takashi Kurosaki

H = ΣiSi2

(2)

which can be intuitively understood as the probability of hitting the same crop when two points are randomly chosen from all the land under consideration. Therefore, a higher H implies a greater concentration of acreage into a smaller number of crops. In addition to H, two indices of crop compositions were calculated. The first measure, SRW, is defined as the sum of areas under rice and wheat divided by the sum of areas under cereal and pulses (so-called “foodgrains” in South Asia). This measure shows the tendency to grow the two Green Revolution crops instead of various kinds of coarse grains or pulses. The second measure, SNF, is defined as the sum of Si for non-foodgrain crops, which is a crude measure of the tendency toward growing non-food, pure cash crops. The traditional approach in analyzing agricultural productivity is through growth accounting, estimating the total factor productivity (TFP) as a residual after controlling for factor inputs (Timmer28). As a complement to the TFP approach, Kurosaki14 proposed a methodology to focus on the role of resource reallocation within agriculture --- across crops and across regions. Unlike in manufacturing industries, the spatial allocation of land is critically important in agriculture due to high transaction costs including transportation costs (Takayama and Judge27; Baulch3). Because of this, farmers may optimally choose a crop mix that does not maximize expected profits evaluated at market prices but does maximize expected profits evaluated at farm-gate prices after adjusting for transaction costs (Omamo18, 19). Subjective equilibrium models for farmers provide other reasons for the divergence of decision prices by farmers from market prices. In the absence of labor markets, households need to be selfsufficient in farm labor (de Janvry et al.7), and if insurance markets are incomplete, farmers may consider production and consumption risk or the domestic needs of their families (Kurosaki and Fafchamps17). In these cases, their production choices can be expressed as a subjective equilibrium evaluated at household-level shadow prices. During the initial phase of agricultural transformation, therefore, it is likely that the extent of diversification will be similar at the country level and the more micro levels because, given the lack of well-developed agricultural produce markets, farmers have to grow the crops they want to consume themselves (Timmer29). As rural markets develop, however, the discrepancy between the market price of a commodity and the decision price at the farm level is reduced. In other words, the development of rural markets is a process which allows farmers to adopt production that reflects their comparative advantages more

Land-Use Changes and Agricultural Growth in India, Pakistan, and Bangladesh

311

closely, and thus contributes to productivity improvement at the aggregate level evaluated at common, market prices. Therefore, the effect of crop shifts on productivity is a useful indicator of market development in developing countries. To quantify this effect, changes in aggregate land productivity can be decomposed into crop yield effects, static inter-crop shift effects, and dynamic inter-crop shift effects (Kurosaki14). Let Yt denote per-acre output in year t. Its growth rate from period 0 to period t can be decomposed as (Yt - Y0)/Y0 = [ΣiSi0(Yit - Yi0) + Σi(Sit - Si0)Yi0 + Σi(Sit - Si0)(Yit - Yi0)]/Y0

(3)

where the subscript i denotes each crop so that Yit stands for per-acre output of crop i in year t. The first term of equation (3) captures the contribution from the productivity growth of individual crops. The second term shows “static” crop shift effects, as it becomes more positive when the area under crops whose yields were initially high increases in relative terms. The third term shows “dynamic” crop shift effects, as it becomes more positive when the area under dynamic crops (i.e., crops whose yields are improving) increases relative to the area under non-dynamic crops.f 3. Gross Output and Land Productivity 3.1. Total output and per-capita output The long-term trends of Q (total output value) are plotted in Figure 1. In all of the three countries, the total output value grew very little in the period before independence in 1947 and then grew steadily afterward. However, if we look at the figure in more detail, we observe differences across the three countries and across the decades. During the colonial period, the total output value in Bangladesh declined while that in Pakistan increased. India stood in between. In the post-1947 period, the total output value in Pakistan increased most rapidly, while that in Bangladesh increased slowly. Again, India stood in between. The timing when the growth accelerated further during the post-1947 period also differs across the three countries. f

For each crop, another aspect of land-use changes can be investigated, focusing on the effect of inter-spatial crop shifts on land productivity. Kurosaki14 thus proposed a further decomposition of the crop yield effect for crop i in equation (3) into “District crop yield effects,” “Inter-district crop shift effects (static),” and “Inter-district crop shift effects (dynamic).” Kurosaki14 applied this decomposition to the district-level data of West Punjab from 1901/02 to 1991/92 and found that the inter-district shift effects were important contributor to productivity growth in cotton and rice.

Takashi Kurosaki

312 500

450

400

Index (1959/60=100)

350

300

250

200

150

100

50

India

Pakistan

2002

1998

1994

1990

1986

1982

1978

1974

1970

1966

1962

1958

1954

1950

1946

1942

1938

1934

1930

1926

1922

1918

1914

1910

1906

1902

0

Bangladesh

Figure 1 Agricultural Output (Q) in India, Pakistan, and Bangladesh, 1901/02-2003/04.

To capture the between-country difference quantitatively, Table 2 reports the estimation results of equation (1), first for each decade and then for the preand post- 1947 periods. When we look at the results for each decade, we find that the total output value grew very little up to the Partition in all three countries. Only in Pakistan during the 1900s and 1930s, the growth rate was positive and statistically significant. When the whole pre-1947 period is taken, Q grew at 1.3% per annum in Pakistan and at 0.4% in India, and it declined at 0.2% in Bangladesh, all of which were statistically significant. After the Partition, Q increased in every decade in all three countries with statistical significance. The growth rates were generally higher in Pakistan than in India and Bangladesh. When the whole post-1947 period is taken, Q grew at 3.5% per annum in Pakistan, at 2.7% in India, and at 2.0% in Bangladesh. The column “C.V.” in Table 1 shows how variable was the output around the fitted values in terms of the coefficient of variation. The total output value was the most variable during the 1900s and 1910s but was stabilized since then, possibly due

Land-Use Changes and Agricultural Growth in India, Pakistan, and Bangladesh

313

to the development of irrigation. The stabilization of agricultural output after the Partition is observed in all three countries. During the 1990s, the growth rate in India was 1.7%, a rate lower than the post-independence average of 2.7% but the 1990s were associated with less variability. The similar deceleration in agricultural growth, associated with stabilization, was observed in Pakistan during the 1990s as well. Although these growth rates, except for the negative growth in the pre-1947 period in Bangladesh, seem impressive, they were not high enough to compensate for high population growth rates. This is shown in Figure 2, where the long-term trends of Q/L (agricultural output per capita) is plotted, and in the middle columns of Table 1. In all three countries, including Pakistan, per-capita agricultural output declined in the colonial period. The decline was largest in Bangladesh (-1.2% per annum), followed by India (-0.4% per annum), which were statistically significant. The decline was larger in the 1920s and 1930s than in the 1900s and 1910s. 250

230

210

170

150

130

110

90

70

India

Pakistan

Bangladesh

Figure 2 Agricultural Output Per Capita (Q/L) in India, Pakistan, and Bangladesh, 1901/02-2003/04.

2002

1998

1994

1990

1986

1982

1978

1974

1970

1966

1962

1958

1954

1950

1946

1942

1938

1934

1930

1926

1922

1918

1914

1910

1906

50 1902

Index (1959/60=100)

190

Takashi Kurosaki

314

Since 1947, per-capita output grew at statistically-significant growth rates in India and Pakistan. In both countries, the largest improvement in per-capita output occurred in the 1950s and 1960s. In sharp contrast, in Bangladesh during the post-1947 period, per-capita agricultural output continued to decline, although at a slower rate (-0.26% per annum) but still with statistical significance. In Bangladesh during the 1990s, however, the trend was turned around into a positive one (see also Figure 2). 3.2. Aggregate land productivity The growth of total output (Q) can be decomposed into the contribution from the aggregate land productivity (Q/A) and the growth of cropped areas (A). To investigate how much of the growth (or stagnation) of output was due to the growth of the land productivity, Figure 3 plots the long-term trends of Q/A and the right columns of Table 2 report regression coefficients of the growth equation (1) for Q/A. First, the shape of Figure 3 is very close to that of Figure 1. Figure 3 again indicates the reversal of trends at around 1947 in all 300

Index (1959/60=100)

250

200

150

100

India

Pakistan

2002

1998

1994

1990

1986

1982

1978

1974

1970

1966

1962

1958

1954

1950

1946

1942

1938

1934

1930

1926

1922

1918

1914

1910

1906

1902

50

Bangladesh

Figure 3 Agricultural Output Per Acre (Q/A) in India, Pakistan, and Bangladesh, 1901/02-2003/04.

Land-Use Changes and Agricultural Growth in India, Pakistan, and Bangladesh

315

three countries --- aggregate land productivity stagnated during the pre-1947 period; since the Partition, it continued to grow. A surprising finding is that the reversal of the land productivity occurred before the breakthrough in the cereal production technology known as the “Green Revolution” in the late 1960s. To show this formally, a series of tests are conducted for a structural change of unknown timing for the entire 20th century, following the procedure by Hansen11. First, a time series model of (1’) is estimated by OLS. For all candidate breakdates, Chow statistics for the null hypothesis of no structural change are estimated and their sequence is plotted as a function of candidate breakdates. The year with the highest Chow test statistics is the Quandt statistic, whose statistical significance can be tested by critical values provided by Bai and Perron1. If we can find a statistically-significant breakdate, the sample is then split in two and the test is re-applied to each subsample, following Bai and Perron’s sequential procedure. The breakdate estimates for Q/A in India, Pakistan, and Bangladesh are 1950/51, 1951/52, and 1949/50, respectively. All of the three are statistically significant at the 1% level. The hypothesis of two or more structural breaks is not supported by the data for India and Bangladesh, while the second break at 1934/35 was found with the 5% level significance for Pakistan. The dominant break at around 1950 is thus clearly shown for all three countries, confirming the previous results based on similar methods applied to South Asia (e.g., see Hatekar and Dongre9; Kurosaki14). In the context of India, the Nehruvian era was thus the turning point for agricultural productivity changes. Coefficient estimates for b reported in Table 2 show that, during the pre1947 period, agricultural output per acre stagnated (growth rate was -0.04% per annum and statistically insignificant) in India. Therefore, the growth of total output at 0.43% during the colonial period was totally attributed to the increase in cropped areas in India. During the post-1947 period, Q in India grew at 2.7% per annum while Q/A increased at 2.2%. Therefore, the major contribution to agricultural growth after independence came from the improvement in aggregate land productivity. During the 1980s and 1990s, the growth rates of Q and Q/A were very similar in India, indicating the limited contribution of area expansion to agricultural growth in recent years. The experience in Pakistan was slightly different from that in India. Even during the pre-1947 period, Q/A increased at 0.38% per annum, which was statistically significant. Nevertheless, considering the growth rate of total output at 1.3% in the colonial period, the dominant contribution to agricultural growth in Pakistan came from the increase in cropped areas before independence, as in

Takashi Kurosaki

316

India. During the post-1947 period, Q in Pakistan grew at 3.5% per annum while Q/A increased at 2.3%. Therefore, the major contribution to agricultural growth after independence came from the improvement of the overall land productivity in Pakistan, but the contribution from area expansion was larger in Pakistan than in India. Unlike in India, the contribution from area expansion to agricultural growth continued to be positive in Pakistan during the 1980s and 1990s. In Bangladesh, the growth rate of Q/A was negative before 1947 and its absolute value is close to that of the growth rate of Q. Therefore, the agricultural stagnation in Bangladesh during the British period can be attributed to the stagnation in land productivity, rather than a decrease in cropped areas. During the post-1947 period, Q in Bangladesh grew at 2.0% per annum while Q/A increased at 1.4%. Therefore, the major contribution to agricultural growth came from the improvement in land productivity in Bangladesh, but the contribution from area expansion (land-use intensification) was also substantial, and its contribution continued to be positive during the 1990s. Furthermore, unlike in India or Pakistan, the growth of Q/A did not decelerate during the 1990s in Bangladesh. Comparing the growth rates during the 1990s, the performance of the Bangladeshi agriculture thus surpassed those of India and Pakistan. Table 2 Growth Performance of Agriculture in India, Pakistan, and Bangladesh, 1901-2004. Q (Total output value) Growth rate

C.V.

Q/L (Output per capita) Growth rate

C.V.

Q/A (Output per acre) Growth rate

C.V.

-0.25%

9.8%

India 1901/02 - 10/11

1.31%

11.6%

0.66%

11.6%

1911/12 - 20/21

-0.81%

13.3%

1921/22 - 30/31

0.05%

2.7%

-0.96% **

-0.90%

13.3%

-0.47%

8.1%

2.7%

-0.28%

2.5%

1931/32 - 40/41

0.29%

4.0%

-1.11% **

4.0%

1941/42 - 50/51

-0.52%

4.2%

-1.76% ***

4.2%

0.15% -1.42% **

3.0% 4.0%

1951/52 - 60/61

4.24% ***

5.1%

2.28% ***

5.1%

2.34% ***

1961/62 - 70/71

2.53% **

8.8%

0.32%

8.8%

1.89% **

4.2% 7.2%

1971/72 - 80/81

2.62% **

7.1%

0.41%

7.1%

2.12% ***

5.6%

1981/82 - 90/91

3.21% ***

6.2%

1.07%

6.2%

3.23% ***

3.7%

1991/92-2000/01

1.68% ***

3.5%

-0.27%

3.5%

1.62% ***

2.4%

1901/02 - 46/47

0.43% ***

8.7%

-0.39% ***

10.0%

1947/48-2003/04

2.72% ***

7.5%

0.60% ***

7.6%

-0.04% 2.19% ***

6.2% 6.3%

Land-Use Changes and Agricultural Growth in India, Pakistan, and Bangladesh

317

Pakistan 1901/02 - 10/11

4.32% **

15.1%

2.75%

15.1%

0.99%

10.9%

1911/12 - 20/21

-0.33%

14.6%

-1.18%

14.6%

-0.19%

6.6%

1921/22 - 30/31

-0.64%

10.3%

-1.73%

10.3%

-1.15%

7.4%

1931/32 - 40/41

2.81% ***

5.8%

1941/42 - 50/51

0.05%

6.7%

0.97%

1951/52 - 60/61

3.44% ***

5.2%

1.00%

5.2%

1.66% ***

3.3%

1961/62 - 70/71

5.85% ***

5.2%

2.99% ***

5.2%

3.93% ***

4.5%

1971/72 - 80/81

3.24% ***

3.7%

0.09%

3.7%

1.75% ***

3.3%

1981/82 - 90/91

3.50% ***

5.2%

0.85%

5.2%

2.65% ***

5.3%

1991/92-2000/01

2.30% ***

5.3%

-0.35%

5.3%

1.61% **

5.5%

1901/02 - 46/47

1.30% ***

12.8%

-0.03%

11.9%

0.38% ***

8.6%

1947/48-2003/04

3.48% ***

8.2%

7.7%

2.30% ***

6.4%

-2.92% ***

0.68% ***

5.8% 6.7%

1.86% ** -0.19%

5.1% 3.4%

Bangladesh 1901/02 - 10/11

0.55%

15.6%

-0.55%

15.6%

-0.66%

13.4%

1911/12 - 20/21

-1.63%

12.1%

1921/22 - 30/31

0.52%

7.7%

-0.44%

-2.53% *

12.1%

-1.15%

11.3%

7.7%

0.57%

5.9%

1931/32 - 40/41

-0.98%

7.4%

1941/42 - 50/51

-1.76%

9.2%

-2.13% **

7.4%

-1.19%

8.2%

-1.88% *

9.2%

-0.72%

6.8%

1951/52 - 60/61

1.25% *

6.0%

-1.24% *

6.0%

1.28% **

3.9%

1961/62 - 70/71

3.02% ***

4.1%

0.34%

4.1%

0.96% **

3.1%

1971/72 - 80/81

3.35% ***

3.9%

1.03% **

3.9%

2.29% ***

2.9%

1981/82 - 90/91

2.00% ***

2.9%

2.9%

1.97% ***

2.2%

1.99% ***

4.1%

1991/92-2000/01 1901/02 - 46/47 1947/48-2003/04

2.60% *** -0.24% ** 2.00% ***

5.6%

-0.15% 1.03%

5.6%

10.5%

-1.20% ***

10.6%

6.4%

-0.26% ***

7.1%

-0.21% * 1.39% ***

9.4% 5.8%

Source: Estimated by the author using the dataset described in the text. Note: “Growth rate” provides a parameter estimate for the slope of the log of Q (or Q/L or Q/A) on a time trend, estimated by OLS (see equation (1)). The parameter estimate is statistically significant at 1% ***, 5% **, or 10% * (two sided t-test). “C.V.” shows the coefficient of variation approximated by the standard error of the OLS regression.

3.3. Difference-in-difference From Table 2, it was found that the level of growth performance was highest in Pakistan, followed by India, with Bangladesh at the bottom. However, it is possible that such difference in growth levels reflects the inherent differences among these countries, such as agro-ecological conditions, leading to the

Takashi Kurosaki

318

difference in potential growth rates. To capture the impact of regime shifts, it is better to focus on the difference-in-difference (DID). Therefore, equation (1’) was estimated, whose results are reported in Table 3. Table 3 Difference-in-Difference of Agricultural Growth Rates in India, Pakistan, and Bangladesh, 1901-2004. Q (Total output value)

Q/L (Output per capita)

Q/A (Output per acre)

1. Impact of the Partition, 1947 (a) Difference in growth rates after 1947 (b1 in equation (1’)) India

2.29% ***

1.00% ***

Pakistan

2.18% ***

0.70% ***

2.23% *** 1.91% ***

Bangladesh

2.24% ***

0.94% ***

1.60% ***

(b) Statistical significance of the difference-in-difference (chi2 statistics) India=Pakistan (b1I = b1P)

0.51

3.53 *

6.29 **

Pakistan=Bangladesh (b1P = b1B)

0.15

1.97

5.27 **

Bangladesh=India (b1B = b1I)

0.14

0.14

40.52 ***

India=Pakistan=Bangladesh

0.53

3.67

41.16 ***

2. Impact of Bangladesh’s independence, 1971 (a) Difference in growth rates after 1971 (b1 in equation (1’)) Pakistan

- ** 0.63%

- *** 0.77%

0.00%

Bangladesh

0.19%

0.27%

0.54% ***

(b) Statistical significance of the difference-in-difference (chi2 statistics) Pakistan=Bangladesh (b1P = b1B)

2.71 *

13.58 ***

5.78 **

Source: Estimated by the author using the dataset described in the text. Note: “Difference in growth rates” is estimated by a seemingly unrelated regression (SUR) model treating three countries as a system. The parameter estimate for the difference in growth rates is statistically significant at 1% ***, 5% **, or 10% * (two sided t-test). "Statistical significance of the difference-in-difference" reports chi2 statistics for testing across-equation restrictions on the SUR model. The degrees of freedom for the chi2 statistics are 1 when two countries are compared and 2 when three countries are compared. The null hypothesis that the difference-indifference is zero is rejected at 1% ***, 5% **, or 10% *.

When the pre-1947 and post-1947 performances are compared for Q (total agricultural output), there are no significant difference across the three countries. In all of them, the growth rate of Q increased by 2 percentage points after the Partition. When the DID in Q/L (per-capita output) is compared, the additional growth after the Partition is less in Pakistan than in India or Bangladesh. The difference between India’s and Pakistan’s performances is

Land-Use Changes and Agricultural Growth in India, Pakistan, and Bangladesh

319

marginally significant. This reflects the higher population growth rates in Pakistan after the Partition than in India or Bangladesh. When the pre-1947 and post-1947 performances are compared for Q/A (per-acre output), the top achiever is India, followed by Pakistan, with Bangladesh at the bottom. The pair-wise difference is statistically significant for all three pairs. The null hypothesis that the three countries’ performances are the same is also rejected at the 1% level for Q/A. From these DID results, one is tempted to conclude that the agricultural performance in Pakistan and Bangladesh was adversely affected by the political regime change in 1947, and the adverse impact of the politics was larger in Pakistan. This interpretation assumes that India and United Pakistan experienced exactly the same changes in environment, technology, and markets, which is difficult to accept. It thus makes more sense to interpret these results as that the net effect of various kinds of exogenous macro changes that occurred after 1947 was more negative in Pakistan than in India, with Bangladesh in between. To investigate growth changes that occurred in East Pakistan after it became the independent nation of Bangladesh, the pre-1971 and post-1971 performances are compared between Pakistan and Bangladesh. The subsample after the Partition is used for this exercise. The DID results are reported in the lower half of Table 3. Pakistan’s growth rates declined (Q and Q/L) or remained unchanged (Q/A) after 1971, while Bangladesh’s growth rates remained unchanged (Q) or were increased (Q/L and Q/A). The DID is statistically significant for all three indicators. Therefore, the net effect of exogenous macro changes that occurred after 1971 was more negative in Pakistan than in Bangladesh. The late surge of “Green Revolution” in Bangladesh during the late 1980s and 1990s (Rogaly et al., 1999) could be responsible for these DID results. 3.4. Summary and comparison with previous studies The above findings suggest that, first, the Partition in 1947 reversed the trends of agricultural production in India, Pakistan, and Bangladesh, leading to a sustained growth of total output and land productivity. Factors responsible for this reversal may include the food production campaigns just after the Partition, national programs for agricultural extension and rural development, and institutional reforms including land reforms. Another important factor in increasing crop areas as well as land productivity could be the expansion of irrigation since 1947 in India and Pakistan.

320

Takashi Kurosaki

Second, among the three countries, Pakistan achieved the highest growth throughout the period, and its superior performance was especially significant before 1947. Nevertheless, the performance in India improved after 1947 and that in Bangladesh improved during the latest years. Third, all of the three countries experienced the reversal of the land productivity at around 1950. In all of them, the growth rate of Q/A during the 1950s was positive and statistically significant. It is important to note that the reversal of the land productivity occurred before the breakthrough of the “Green Revolution.” As Kurosaki12 showed, the per-acre yields of rice and wheat were stagnant during the 1950s. The first two points confirm research results found in the existing literature. Considering the fact that we calculate the gross value of output, patterns depicted in Figures 1-3 during the colonial period are reasonably close to those in Sivasubramonian’s24 estimates for the value-added for Undivided India. The overall growth rates during the pre-1947 period reported in Sivasubramonian24 lie within the range of our estimates for the total output value for India, Pakistan, and Bangladesh. A new insight from this study is that the positive growth rate in Undivided India was mostly attributable to the growth that occurred in the areas currently in Pakistan. This paper also confirms Blyn’s4 finding for British India that agricultural production increased until the late 1910s, followed by fluctuations with their average lower than the previous peak. This study decomposes this pattern into contributions from the areas currently in India, Pakistan, Bangladesh separately, to find a contrast that Pakistan areas were most favored before 1947 but Pakistan’s superiority in growth performance was reduced after 1947. The regional contrast after the Partition was demonstrated in earlier studies that compared agricultural performance in West and East Punjab --Prabha20quantified this contrast through investigation on official data and Sims23 explained it through a political-economy approach. This study has added new evidence that the contrast can be extended to the country level between India and Pakistan. Similarly, the stagnation of agricultural production and the decline of per-capita output during the colonial period in areas currently in Bangladesh, which we found in this study, confirms Islam’s11 finding for various regions of (united) Bengal and the recent acceleration of agricultural production in Bangladesh found in this study confirms the dynamic changes reported by Rogaly et al.21. This study has added new evidence that these findings can be extended to the country level between Bangladesh and India (or Pakistan).

Land-Use Changes and Agricultural Growth in India, Pakistan, and Bangladesh

321

The third point was first indicated by Kurosaki12, 13. The point is that even with no changes in land productivity of individual crops and in available land for cultivation, agricultural output can grow by shifting the crop mix toward high value crops. This shift is accelerated when rain-fed land is turned into irrigated land. Although the aggregate output per acre did increase during the 1950s at a statistically significant rate, per-acre productivity of major crops (rice and wheat) did not increase much during the same period. Therefore, one of the most important factors for the reversal at the Partition should have been a change in crop composition toward high value crops, as shown in the next section. 4. Changes in Crop Mix and Their Contribution to Land Productivity 4.1. Trends in crop mix Figure 4 shows the Herfindahl Index (H) of crop acreage over the study period. There are several interesting contrasts among India, Pakistan, and Bangladesh. First, there is a difference in overall levels. In every year, H is the highest in Bangladesh and the lowest in India, with Pakistan in the middle. This seems to reflect the size of the economy and the diversity of agro-ecological conditions. Indian agriculture is the largest and the most diverse among the three, resulting in the lowest crop concentration ratio in India. Second, there is a difference in annual fluctuations: H of India is the most stable and H of Bangladesh is the most variable, with Pakistan in the middle. This again seems to reflect the size of the economy. Third, a distinct pattern emerges after the independence in India and Pakistan --- H fluctuated with no trend before 1947 while it increased continuously since the mid 1950s. In contrast, it is difficult to find such a shift at the Partition for Bangladesh. According to Timmer’s 29 stylization, the one-way concentration of crops since the mid 1950s in India and Pakistan can be interpreted as a stage before a mature market economy with diversified production and consumption at the national level. Fourth, there is a difference in recent trends. In India, the level of concentration accelerated in the 1990s and seems to have reached a plateau in the early 2000s. In contrast, the crop concentration index in Pakistan did not accelerate in the 1990s but it remained at the high level that had already been reached during the late 1980s or early 1990s. This seems to indicate that shifts in acreage toward crops with comparative advantages occurred earlier in Pakistan than in India, possibly reflecting Pakistan’s attempt to liberalize

Takashi Kurosaki

322

India

Pakistan

2002

1998

1994

1990

1986

1982

1978

1974

1970

1966

1962

1958

0.5 1954

0.10 1950

0.55

1946

0.15

1942

0.6

1938

0.20

1934

0.65

1930

0.25

1926

0.7

1922

0.30

1918

0.75

1914

0.35

1910

0.8

1906

0.40

1902

H (Herfindahl index based on area shares)

agricultural marketing during the early 1980s. The recent trend in H for Bangladesh seems to be a negative one. According to Timmer’s29 stylization, when the agriculture of a country enters the next stage with a mature market economy, both production and consumption become diversified at the national level.

Bangladesh (right axis)

Figure 4 Crop Concentration in India, Pakistan, and Bangladesh, 1901/02-2003/04.

Figure 4 may suggest that such transformation occurred first in Bangladesh during the early 1980s, followed by Pakistan in the late 1980s, and finally occurring in India in the early 2000s. The changes in crop composition are shown more concretely in Figures 5 and 6. In all three countries, SRW (the sum of areas under rice and wheat divided by the sum of areas under foodgrains) increased throughout the 20th century and the trend was accelerated during the post-colonial period (Figure 5). Therefore, there is a strong tendency to shift to the two Green Revolution crops instead of various kinds of coarse grains or pulses. However, the trend of SRW in Bangladesh is weak because the rice is highly dominant as the staple crop.

Land-Use Changes and Agricultural Growth in India, Pakistan, and Bangladesh

323

1.00

0.90

SRW

0.80

0.70

0.60

0.50

1994

1998

2002

1998

2002

1990

1986

1982

1978

1974

1970

1966

1962

1958

1954

Pakistan

1994

India

1950

1946

1942

1938

1934

1930

1926

1922

1918

1914

1910

1906

1902

0.40

Bangladesh

Figure 5 Area Share of Rice and Wheat in Total Foodgrains Acreage in India, Pakistan, and Bangladesh, 1901/02-2003/04. 0.30

0.25

0.15

0.10

0.05

India

Pakistan

Bangladesh

Figure 6 Area Share of Non-Foodgrains in Total Acreage in India, Pakistan, and Bangladesh, 1901/02-2003/04.

1990

1986

1982

1978

1974

1970

1966

1962

1958

1954

1950

1946

1942

1938

1934

1930

1926

1922

1918

1914

1910

1906

0.00 1902

SNF

0.20

Takashi Kurosaki

324

The movement of the sum of shares of non-foodgrain crops, SNF, shows again the contrast between Bangladesh on the one hand and India and Pakistan on the other (Figure 6). In Bangladesh, SNF is declining throughout the study period, while in India and Pakistan, it stagnated initially and then it continued to rise in the second half of the 20th century. In India and Pakistan, the rising SNF shows that there is a strong tendency toward growing non-food, pure cash crops. The declining SNF in Bangladesh seems to cast doubt on our previous interpretation that Bangladesh entered the diversified production pattern earlier than India and Pakistan did. However, if we exclude the share of rice, the movement of SNF in Bangladesh became similar to those in India and Pakistan. Here again, Bangladesh is exceptional because the rice is highly dominant. Looking from a different angle, a contrast among the three countries could be attributed to India’s more diversified geography. Using Timmer’s29 stylization, Bangladesh’s agriculture is more like a household economy in a relative sense than Pakistan’s is, and Pakistan’s agriculture is more like a household economy in a relative sense than India’s is. Furthermore, India’s food policy has been more regulatory than Pakistan’s or Bangladesh’s, less exposed to international trade, especially until the Economic Reforms in the early 1990s. These factors might have resulted in a weaker tendency to specialize in a few crops in India than in the other two countries. Whether these conjectures are correct or not should be examined more carefully through investigating production diversification at the household level and food consumption diversification at the national level, which is left for further research. 4.2. Contribution of crop shifts to aggregate land productivity To investigate whether these changes in crop mix were consistent with those indicated by comparative advantage and market development, decomposition (3) was implemented. The results are reported in Table 4. Table 4 Contribution of Crops Shifts to Land Productivity Growth in India, Pakistan, and Bangladesh. Q/A (%)

Contribution share (%)

Pure yield effects

Static shift effects

Dynamic shift effects

1901/02 - 11/12

0.92

-0.05

-0.06

0.81

113.4

-6.3

-7.1

1911/12 - 21/22

-0.24

-0.10

0.07

-0.27

88.5

36.4

-24.9

1921/22 - 31/32

0.03

0.27

-0.08

0.23

14.1

119.1

-33.3

Total

Pure yield effects

Static shift effects

Dynamic shift effects

India

Land-Use Changes and Agricultural Growth in India, Pakistan, and Bangladesh

325

1931/32 - 41/42

-0.36

0.30

-0.06

-0.11

324.5

-277.5

1941/42 - 51/52

-1.49

0.30

-0.02

-1.21

123.0

-24.5

53.0 1.5

1951/52 - 61/62

2.76

0.14

-0.01

2.89

95.3

5.0

-0.3

1961/62 - 71/72

1.55

0.15

0.20

1.90

81.6

8.0

10.3

1971/72 - 81/82

1.83

0.35

0.09

2.28

80.6

15.4

4.0

1981/82 - 91/92

3.10

0.44

0.14

3.68

84.2

12.1

3.7

1991/92-2001/02

0.87

0.59

0.06

1.51

57.5

38.7

3.7

1901/02 - 47/48

-0.15

0.01

0.15

0.00

n.a.

n.a.

n.a.

1947/48-2003/04

2.79

0.21

0.63

3.63

76.7

5.8

17.5

1.84

-0.19

0.09

1.74

105.4

-10.8

5.4

1911/12 - 21/22

0.06

-0.05

0.02

0.02

254.3

-226.2

71.9

1921/22 - 31/32

-0.35

0.03

0.02

-0.31

113.4

-8.5

-4.9

1931/32 - 41/42

1.51

0.16

0.32

1.99

75.8

8.2

16.0

1941/42 - 51/52

-0.80

0.18

0.03

-0.58

136.6

-30.6

-6.0

1951/52 - 61/62

1.03

0.85

0.03

1.92

53.8

44.4

1.8

1961/62 - 71/72

3.37

0.52

0.28

4.16

80.8

12.5

6.7

1971/72 - 81/82

1.72

0.63

0.13

2.49

69.2

25.4

5.4

Pakistan 1901/02 - 11/12

1981/82 - 91/92

2.35

0.07

0.21

2.63

89.3

2.5

8.2

1991/92-2001/02

1.19

0.23

0.02

1.43

82.8

16.1

1.1

1901/02 - 47/48

0.55

-0.03

0.22

0.74

74.5

-4.0

29.6

1947/48-2003/04

2.50

0.48

0.61

3.59

69.7

13.3

17.0

1901/02 - 11/12

1.38

-0.28

-0.11

0.99

139.5

-28.3

-11.1

1911/12 - 21/22

-1.08

0.00

0.00

-1.09

99.5

0.3

0.2

1921/22 - 31/32

0.32

0.09

0.05

0.47

68.8

19.7

11.5

1931/32 - 41/42

-1.99

0.73

-0.06

-1.32

151.2

-55.7

4.5

1941/42 - 51/52

0.71

-0.50

-0.16

0.05

1339.2

-944.5

-294.7

1951/52 - 61/62

1.35

0.20

0.02

1.57

86.0

12.7

1.3

1961/62 - 71/72

0.09

0.20

0.00

0.28

32.4

69.1

-1.5

1971/72 - 81/82

2.04

-0.20

0.14

1.98

103.2

-10.2

7.1

1981/82 - 91/92

2.19

0.22

-0.07

2.35

93.4

9.5

-2.9

1991/92-2001/02

2.46

-0.19

0.05

2.32

106.0

-8.0

2.0

1901/02 - 47/48

-0.18

0.01

0.00

-0.17

107.6

-6.8

-0.8

1947/48-2003/04

2.15

-0.02

0.04

2.18

98.8

-0.8

2.0

Bangladesh

Source: Estimated by the author using the dataset described in the text. Note: Annual growth rates were estimated using the method explained in the text (see eq. (3)). Since both the estimate model and the data treatment for smoothing are different, the total growth rates of land productivity in this table are slightly different from those in Table 2.

326

Takashi Kurosaki

For areas currently in India, first, the contribution of total crop shift effects is substantial, explaining more than 20% of post-independence growth in aggregate land productivity. Second, with more detailed period demarcation, it is shown that the relative importance of crop shift effects has been increasing throughout the post-independence period. During the 1950s, less than 5% of land productivity growth was attributable to crop shift effects; during the 1990s, about 40% was due to crop shifts. Third, the dynamic crop shift effect was an important source of productivity growth only during the 1960s. In other periods, the static crop shift effect was more important than the dynamic effect. Fourth, during the pre-independence period, crop shift effects played a positive role under adverse conditions of declining crop yields. But for the positive contribution from static crop shift effects, the total land productivity growth rates would have been much more negative in the three decades from the 1920s to 1940s. Interestingly, in India during the 1990s, the growth due to improvements in crop yields was reduced compared to the 1980s while the growth due to static crop shifts was higher. As a result, the relative contribution of static shift effects was as high as 39% in the 1990s. This is the highest figure for all the postindependence decades. Therefore, it can be concluded that the changes in crop mix in the 1990s (the decade of economic liberalization in India) were indeed consistent with the comparative advantages of Indian agriculture, leading to an improvement in aggregate land productivity. The middle rows of Table 4 show the decomposition results for Pakistan. The crop yield effect explained about 70% both in pre- and post- independence periods, while the rest was explained mostly by dynamic shift effect before 1947 and by both dynamic and static shift effects after 1947. The importance of the dynamic shift effect before independence could be attributable to the development of the Canal Colony as an agricultural export base in British India. As is discussed in Section 3, the dynamic crop shift effect becomes more positive when the area under dynamic crops increases relative to the area under non-dynamic crops. During the colonial period, rice and cotton were the dynamic crops in West Punjab and the cultivation of these two crops was regionally concentrating into advantageous districts (Kurosaki14). Decade-wise, the importance of crop shift effects in Pakistan was highest during the 1950s and it has been declining since then. This pattern in Pakistan after independence is opposite to India’s. In Pakistan, during the 1950s, more than 45% of land productivity growth was attributable to crop shift effects; during the 1980s and 90s, less than 20% was due to crop shifts. During the

Land-Use Changes and Agricultural Growth in India, Pakistan, and Bangladesh

327

1950s, the contribution of the static shift effect was in a magnitude close to that of yield improvements. These results show that land reallocation toward high value crops was the main engine of agricultural growth during the pre-Green Revolution period after independence in Pakistan. During the 1990s in Pakistan, the growth due to improvements in crop yields declined substantially while the growth due to static crop shifts recovered. As a result, the relative contribution of static shift effects was 16% in the 1990s, a level higher than the postindependence average (13%). Here we find a similarity between India and Pakistan: in both economies, the crop shifts were an important source of land productivity growth in the post-independence period, and especially in the 1990s. In sharp contrast, the contribution of crop shift effects to the improvement in aggregate land productivity was small in Bangladesh (see the lower rows of Table 4). The crop yield effects explained about 100% of changes in aggregate land productivity in Bangladesh, both in pre- and post- independence periods. One possible interpretation of this finding is that Bangladesh is a region where commercialization occurred earlier so that the room for additional crop shifts to increase the land productivity was small already during the first half of the 20th century. In other words, Bangladesh’s agriculture has had a very strong comparative advantage in rice cultivation since the early 20th century (or earlier than that) and this advantage was already exploited when the study period of this paper began. Looking at the decomposition results for each decade in Bangladesh, however, we find that the static shift effects were important sources of aggregate land productivity growth during the 1950s and 1960s. Examining the crop database, we found that these two decades were a period when sugarcane production expanded and sugarcane had higher values per acre than other crops. Therefore, the decomposition results for Bangladesh are also consistent with prevailing market conditions and farmers’ response to comparative advantage. These results thus indicate that the changes in crop mix were an important source of growth in aggregate land productivity in all three countries of India, Pakistan, and Bangladesh, although the contribution of crop shift effects was small in Bangladesh because of the dominance of rice in cultivation. Throughout the post-independence period, there were substantial contributions from both static and dynamic crop shift effects in India and Pakistan.

328

Takashi Kurosaki

5. Conclusion Based on a production dataset from India, Pakistan, and Bangladesh for the period 1901-2004, this article investigated changes in land use, associating the changes with long-term agricultural performance, focusing on the contribution of crop shifts to improvement in agricultural productivity. The empirical results showed a discontinuity between the pre- and the post- independence periods in all of the three countries. Total output growth rates rose from zero or very low figures to significantly positive levels, which were sustained throughout the post-independence period. The improvement in aggregate land productivity explained the most of this output growth. This article also quantified the effects of crop shifts on aggregate land productivity, a previously unnoticed source of productivity growth. It was found that the crop shifts contributed to the productivity growth, especially during periods with limited technological breakthroughs. The contribution of the crop shifts was larger in India and Pakistan than in Bangladesh, where rice is highly dominant as the staple crop. Underlying these changes were the responses of farmers to changes in market conditions and agricultural policies. Agriculture in these countries experienced a period of concentration of crops, when agricultural transformation in terms of output per agricultural worker was proceeding. These trends continued until the early 1980s in Bangladesh, until the early 1990s in Pakistan, and until the early 2000s in India. The performance in the latest periods suggests that agriculture in the region seems to have entered a new phase of diversified production and consumption at the country level (Timmer29). The contrast in the beginning time of the new phase can be attributed to the difference in farmers’ exposure to international prices created by the difference in trade and industrial policies of these countries. In all three countries in the post-1947 period, however, the growth rate of aggregate land productivity was not high enough to cancel the negative growth rate of land availability per capita. The net result was that the growth rate of agricultural output per capita was much smaller than that of output per acre, resulting in a slow pace of poverty reduction in these countries. The crop shift effects identified in this article were not sufficiently strong in this sense. Reducing population growth rates and absorbing more labor force outside agriculture are required to make the growth rate of per-capita agricultural output comparable to that of per-acre agricultural output. Although this paper showed the importance of crop shifts in improving aggregate land productivity, the overall impact is underestimated, because only major crops were covered. Incorporating non-traditional crops into the

Land-Use Changes and Agricultural Growth in India, Pakistan, and Bangladesh

329

framework of this paper would be highly desirable. To quantify the structural determinants of these changes and their net effects on the welfare of rural population, further research is needed, such as analysis of production costs, investigation of livestock activities, etc. These are left for future study. References 1. J. Bai and P. Perron, Econometrica 66, 47 (1998) 2. A.V. Banerjee, P.J. Gertler and M. Ghatak, Journal of Political Economy 110, 239 (2002) 3. B. Baulch, American Journal of Agricultural Economics, 79, 477 (1997). 4. G. Blyn. Agricultural Trends in India, 1891-1947: Output, Availability, and Productivity, Philadelphia: University of Pennsylvania Press. (1966) 5. J.K. Boyce, Agrarian Impasse in Bengal: Institutional Constraints to Technological Change, Oxford: Oxford University Press. (1987) 6. A.K. Dasgupta, Indian Economic and Social History Review 18, 327. (1981) 7. A. de Janvry, M. Fafchamps, and E. Sadoulet , Economic Journal 101, 1400. (1991) 8. S. Guha (ed.), Growth, Stagnation or Decline? Agricultural Productivity in British India, Delhi: Oxford University Press. (1992). 9. N. Hatekar and A. Dongre, Economic and Political Weekly 40, 1432. (2005) 10. B.E. Hansen, Journal of Economic Perspectives 15, 117. (2001) 11. M.M. Islam, Bengal Agriculture 1920-1946: A Quantitative Study, Cambridge: Cambridge University Press. (1978) 12. T. Kurosaki , Economic and Political Weekly 34, .A160. (1999) 13. ----- Economic and Political Weekly 37, 3149. (2002) 14. ----- American Journal of Agricultural Economics 85, 372. (2003) 15. ----- Journal of International Economic Studies 20, 19. (2006) 16. ----- Compilation of Agricultural Production Data for India, Pakistan, and Bangladesh Areas, c.1900-2004, Hitotsubashi University. (2007) 17. T. Kurosaki and M. Fafchamps, Journal of Development Economics 67, 419 (2002) 18. S.W. Omamo, American Journal of Agricultural Economics 80, 116. (1998a) 19. ----- Journal of Development Studies 35, 152. (1998b) 20. C.Prabha, Indian Economic and Social History Review 6, 333. (1969) 21. B. Rogaly, B. Harris-White, and S. Bose, Sonar Bangla? Agricultural Growth and Agrarian Change in West Bengal and Bangladesh, New Delhi: Sage Publications. (1999)

330

Takashi Kurosaki

22. B. Roy, An Analysis of Long Term Growth of National Income and Capital Formation in India (1850-51 to 1950-51), Calcutta: Firma KLM Private Ltd. (1996) 23. H. Sims, Political Regimes, Public Policy and Economic Development: Agricultural Performance and Rural Change in Two Punjabs, New Delhi: Sage Publications. (1988) 24. S. Sivasubramonian, Papers on National Income and Allied Topics I, New York: Asia Publishing House, 231. (1960), 25. ----- Indian Economic and Social History Review 34, 113 (1997) 26. ----- National Income of India in the Twentieth Century. Delhi: Oxford University Press. (2000) 27. A. Takayama and G. Judge, Spatial and Temporal Price and Allocation Models, Amsterdam: North Holland. (1971) 28. C.P. Timmer, Handbook of Development Economics I, Amsterdam: Elsevier Science, 275.(1988) 29. ----- American Journal of Agricultural Economics 79, 621. (1997) 30. World Bank World Development Report 2000/2001: Attacking Poverty, Oxford: Oxford University Press. (2001) 31. ----- World Development Report 2008: Agriculture for Development, Oxford: Oxford University Press. (2008)

Macro and Public Finance

This page intentionally left blank

THE NATURAL INTEREST RATE IN EMERGING MARKETS ASHIMA GOYAL

∗

Indira Gandhi Institute of Development Research Gen. Vaidya Marg, Santosh Nagar, Goregaon (E), Mumbai-400 065 Email: ashima@ igidr.ac.in An optimizing model of a small open emerging market economy (SOEME) with dualistic labour markets and two types of consumers, is used to derive the natural interest rate, terms of trade and potential output. Shocks are classified into generic types that affect the natural interest rates. Since parameters depend on features of the labour market and on consumption inequality, the natural rates and the impact of shocks differ from those in a mature small open economy. Subsistence consumption is found to have the largest effect on the natural rates. It reduces the interest rate, raises natural output and the terms of trade. Technology and infrastructure backwardness reduce natural output. The implications for monetary policy are derived. The effect of managed exchange rates combined with different types of inflation targeting is examined through simulations. Endogenous terms of trade make the supply curve steeper in a SOEME, so partial stickiness of the real exchange rate can be beneficial. In general, domestic inflation targeting, with some weight on the output gap, delivers lower volatility. Output response is higher and volatility lower with fixed terms of trade, demonstrating the flatter supply curve. CPI inflation targeting also does well when terms of trade are credibly fixed.

∗

This paper uses the model developed in Goyal (2007)[10] to address new questions. The earlier paper was presented at the IGIDR-RBI-Northwestern and the ISI conferences in 2007. I thank all those who contributed comments, including an anonymous referee, and T. S. Ananthi, Jayashree and Reshma Aguiar for secretarial assistance. 333

334

Ashima Goyal

1. Introduction Forward-looking aggregate demand and supply curves are used to examine options for monetary policy. They are derived from an open economy dynamic stochastic general equilibrium (DSGE) model with imperfect competitiona and nominal rigiditiesb as well as labour market features that reflect a populous emerging market. Since the behavioral equation coefficients are derived from basic technology, preferences and market structure, they are robust to policy changes, thus meeting the Lucas critique. The calibrated model allows estimation of indicative values for natural rates and the order of magnitude by which they differ from a small open economy (SOE). In DSGE models optimizing labour supply decision drive unemployment — this cannot capture the dimensions of developing economy unemployment. The modeling of two types of labour makes it possible to capture a major aspect. The small open emerging market (SOEME) has a large share of less productive labour in the process of being absorbed into the modern sector. The basic intertemporal optimization model tells us that the steady-state real interest rate must equal the representative consumer’s time discount rate plus the rate of growth of population. The problem is that for a sustained period of transition and rapid catch-up the growth rate of such an economy can be above the growth rate of population. Does that mean that the natural interest rate should equal this growth rate and be considerably higher than benchmark international interest ratesc? Since steady-state values are not of much guidance in transition periods, modern microfoundation-based models of monetary policy offer a good analytical framework in which to examine this question. The framework is also more relevant for the design of actual monetary policy.

a

b

c

Clarida et. al. (1999)[5] surveys this literature, and Clarida et. al. (2001)[6] extends it to an open economy. Woodford (2003)[17] is a rigorous textbook treatment. Obstfeld and Rogoff (1995)[15] is a textbook treatment of a large literature on the new open economy macroeconomics. Prices in their seminal contribution were determined one period in advance. Later treatments use variants of staggered prices, which allows smooth aggregate price adjustment. Such a view seems to guide Indian monetary policy making. For example a deputy governor of the Indian Reserve Bank, writes: “First, real GDP growth has recorded strong growth since 2003-04, averaging 8.6 per cent per annum over the four-year period ending 2006-07. This growth is significantly higher than world economic growth. This would suggest that equilibrium real interest rates for a country like India would be higher than world interest rates. Mohan (2007, pp.5)[14] ”

The Natural Interest Rate in Emerging Markets

335

The natural interest rate is defined as the equilibrium real rate, consistent with a zero or target rate of inflation, when prices are fully flexible. Shocks that change the natural rate open an output gap and affect inflation. Most Central Banks (CBs) have an operating target interest rate. This defines an operating rule, telling the CB how to change its interest rate in response to shocks. Inflation targeting is an example of such a rule. Shocks are of two generic kinds — those affecting demand and those affecting supply. They can be derived from the general equilibrium intertemporal optimization. The special labour market features of the SOEME introduce more shocks affecting the natural rate. Thus the SOEME model helps to identify these shocks, their difference from SOEs, and the implications for policy. Two types of consumers and labour are distinguished in the SOEME, those above subsistence (R), and those at subsistence (P). While the first are able to smooth consumption using international markets, those at subsistence cannot. Their intertemporal elasticity of consumption, productivity and wages are lower and their labor supply elasticity is higher, compared to the first groupd. These features follow from the key difference — high and low productivity. CES aggregation allows the micro diversity to be collapsed to macro aggregates, as is common in the literature. A key difference between the two kinds of economies is that the real exchange rate for the SOEME is depreciated and appreciates over the long-run as development brings it closer to purchasing power parity (PPP). But there are short-run fluctuations. The stronger effects of income on the terms of trade in a SOE imply that fluctuations in the terms of trade make the aggregate supply curve steeper. This is particularly so for an almost closed economy with a large percentage of P. Therefore it may be better to manage the terms of trade, in response to temporary shocks, thus flattening aggregate supply. Variations in the exchange rate can still be used to counter shocks to aggregate supply. A new exogenous variable in the model for the SOEME is consumption of the subsistence group. This has a large impact on the natural rate of interest, potential output, and the equilibrium terms of trade. In general, it reduces the effect of world output on these variables, compared to the SOE. A temporary shock to subsistence consumption imparts a strong negative shock to the natural rate, implying that policy should accommodate such a shock by decreasing the

d

This subsistence-based definition makes the model suitable for analyzing populous emerging markets such as India and China.

336

Ashima Goyal

policy rate. The factors tending to decrease the natural rate are dominant. Subsistence consumption and the gap from world income levels, tend to raise natural output while technology and infrastructure backwardness reduce it. Policy also has to accommodate permanent changes in natural output and equilibrium terms of trade as transition occurs. There are multiple steady states on the way. The welfare consequences of optimizing policy responses to shocks to the natural rate of interest are examined through simulations. Different types of inflation targeting are differentiated by weights in the policy objective function. Flexible targeting of domestic price inflation continues to deliver the best results, as in Goyal (2007) [10] . It involves active use of exchange rate policy to lower inflation, but volatility is lower than under consumer price targeting. The latter does almost as well as domestic inflation targeting when policy credibly fixes the terms of trade. The exchange rate is not itself a target but an instrument affecting both output and inflation. Constant terms of trade are consistent with variation in the nominal exchange rate. As more labour shifts above subsistence with development and the technology and infrastructure gap narrows, permanent changes occur in potential output, and in the equilibrium terms of trade. They eventually converge to world levels. By focusing on outcomes, inflation targeting allows policy to adjust to changes in potential output reflected in inflationary pressures. Changes in equilibrium terms of trade have also to be allowed. Since too many things are going on in the real world, a model can serve as a valuable laboratory. But to have confidence that it captures the crucial aspects of the economy, its response to simple shocks must be similar to that of the real economy — then it can be trusted for more complex shocks (Christiano at. al., 1999) [4]. The behavioral foundation, plus key structural aspects, and the type of shocks modelled increase confidence that the paper does capture crucial aspects. In the Indian economy, however, interest and exchange rate flexibility is relatively recent, there is no formal inflation targeting, and policy optimization is only implicit. The simulation results can be understood as indicating what policy should do in such a structure. The results are intuitive both with respect to theory and structure. Policy has long been targeting a real effective exchange rate or constant terms of trade, it has recently moved to a more flexible nominal exchange rate, interest changes have been smooth and mild, as the simulations recommend. But policy has sometimes failed to accommodate natural rate shocks and kept the level of the policy rate too high. There has, however, been a tendency to accommodate agricultural shocks — which are shocks to the consumption of the poor.

The Natural Interest Rate in Emerging Markets

337

The structure of the paper is as follows. Determinants of the natural interest rate are discussed in Section 2.The basic model (outlined in Appendix A) is adapted to an emerging market in Section 3 and its differences from the SOE model noted. Natural rates are derived in Section 4. Calibrated values of the natural rates, long and short-term optimal policy are obtained through simulations in Section 5. The results on optimal policy are assessed in the light of historical Indian macroeconomic policy in Section 6 before Section 7 concludes. Appendix B derives the employment subsidy that delivers the flexible price equilibrium in a SOEME. 2. The Natural Rate The NKE school (Clarida et. al 1999[5], 2001[6], Woodford, 2003[17]) has developed forward-looking aggregate demand and supply curves from intertemporal optimization by representative consumers and firmse. The rational expectations equilibrium from which aggregate demand is derived comes from the basic consumption Euler, Eq. (1), as a result of household choice of the optimum timing of expenditure. Aggregate real expenditure, Yt , and the price index, Pt, must satisfy the Euler condition (1) at all periods, where rt is the riskless one period nominal interest rate controlled by the Central Bank (CB), Gt is government purchases, and UC is the household’s utility function. The exogenous disturbance, ξ t , captures variations in the household’s impatience to consume and β is its discount factor. 1 + rt = β

−1

⎪⎧ ⎡ U c (Yt +1 − Gt +1 ;ξ t +1 ) Pt ⎤ ⎪⎫ ⎨Et ⎢ ⎥⎬ ⎪⎩ ⎣ U c (Yt − Gt ; ξ t ) Pt +1 ⎦ ⎭⎪

−1

(1)

Monetary policy responds through the interest rate. A natural way to think about monetary policy in this context is through the gap between the policy rate and the natural rate. The natural interest rate was a concept originally defined by Wicksellf, as the equilibrium real rate of return when prices are fully flexible. It is derived from the basic consumption Euler equation when output,

e

f

Investment is not explicitly modeled in this framework but its effect comes in through exogenous variations in productivity — the analysis abstracts from the affect of investment on productive capacity and on marginal utility. The specific form of the interest rate rule Wicksell (1898) advocated for the implementation of price-level targeting was for the CB’s interest rate to rise if prices rose, fall if they fell, and to remain unaltered at whatever level it was at, unless prices changed.

338

Ashima Goyal

Yt , equals its natural level, Yt , and inflation is zero, so that prices are constant (or Pt = Pt+1). In these conditions the policy rate in Eq. (1) equals the natural rate in Eq. (2): rt = rt n .

(

⎧ ⎡ U c Y t +1 ; ξ t +1 ⎪ 1 + rt n ≡ β −1 ⎨ E t ⎢ ⎢ U Y t ;ξ t ⎩⎪ ⎣ c

(

)

) ⎤⎥ ⎫⎪

−1

⎬ ⎥⎪ ⎦⎭

(2)

In the modern approach, the steady-state value of the natural interest rate is defined as the value consistent with a zero or target inflation rate. The natural rate is the real rate of interest that keeps aggregate demand equal to the natural rate of output. This equality follows from the firm’s optimization. Output is at its natural rate when the real marginal cost of supplying each good equals the marginal revenue for any firm that is thinking of changing its price, when all firms charge identical prices. But when this condition holds no firm wants to charge a different price, so there is no inflation. Marginal revenue is the reciprocal of the desired gross mark-up. Since at both Yt and rt n inflation is at zero, Yt is used to define rt n . All these concepts are derived below in the context of our SOEME. The shock or exogenous term rrt that enters the NKE aggregate demand is then the percentage deviation of the natural rate from its steady-state value. The deviation occurs due to real disturbances that change natural output. Loglinearizing Eq. 2, a real shock affecting utility gives rrt , which equals the expectation term on the RHS of Eq. 2: rrt = log (1 + rt n ) + log β

(3)

In case of non-zero target inflation π , nominal interest rate must equal: rt = rrt + π

(4)

A non-zero target inflation can be abstracted from since although it affects average values of output and nominal interest rates, it does not affect the latter’s response to shocks upto a log-linear approximation. Since a log linearization of Eq. 1 implies that rt n = Y t +1 − Y t , to understand the CB’s response to shocks it is necessary to see how these shocks affect the natural output. A temporary shock to natural output changes the natural rate. By substituting Yt = Yt in the firm’s marginal cost and log linearizing, an equation of the form (5) below can be derived. Log linear approximations of

The Natural Interest Rate in Emerging Markets

339

equilibrium conditions are adequate since the policy focus is on small fluctuations around a steady state. The generic disturbances that affect natural output then areg:

(

Yt = f Gˆ t , c t , a t , ϕ ht

)

(5)

Each of these disturbances increases natural output. They can be grouped into those affecting demand, and therefore requiring variation in log output to maintain a constant marginal utility of real income, and those affecting supply and therefore requiring variation in log output to maintain a constant marginal disutility of labour supply. Given these shocks, the change in natural output depends on the intertemporal elasticity of substitution of private expenditure 1/σ, and the elasticity of real marginal cost with respect to a firm’s own output. In the first category are Gˆ t or the normalized deviation of government purchases from their steady-state level, and c or shift in consumer preferences. Technology, at, and labour supply shocks ϕ ht , the latter due to shifts in the disutility of labour function, are in the second, or supply shock category. The effect of each of these shocks on rrt can be obtained by substituting these solutions into Equation (3) for rrt :

(

rrt = g (1 − ρ G ) Gˆ t , (1 − ρ c ) c t , (1 − ρ a ) a t ,ϕ (1 − ρ h ) h t

)

(6)

Each disturbance follows an independent first-order autoregressive process. For stability the respective autocorrelation coefficients (subscripted ρ s) must each be less than unity. The result is that rrt rises for any temporary demand shock and falls for any temporary supply shock. These equations and results are explicitly derived for our model in the section below. Variation in the generic shocks due to openness and underdevelopment is noted. In rational expectations equilibria with stable inflation, interest rates must follow these exogenous variations in the natural rate of interest. Optimal policy requires insulating the output gap from these shocks, so that the CB’s interest rate instrument should move in step with the natural rate. Thus the CB would accommodate supply shocks by lowering interest rates and offset demand shocks by raising interest rates. Since inflation may be stable even with a widening output gap if prices are preset, and output movements are the same under the different types of shocks but the optimal interest response varies, the

g

See Woodford (2003)[17] pp. 249 for a more specific functional form.

340

Ashima Goyal

CB may require information in addition to that contained in output and inflation to fully implement this policy. The required interest rate variation is higher the more temporary the shock. 3. A Small Open Emerging Market Economy Model

Key features of microfoundation based SOE models used to derive optimal monetary policy are intertemporal optimization and labour-leisure tradeoff by consumers; monopolistic competition and product diversity so that producers have pricing power, and output is below the social optimum. The Calvo model of staggered prices generates the sticky prices required for monetary policy to have real effects on output. The optimization results in simple standard aggregate demand (AD) and supply curves (AS) with the difference that they include forward-looking variables. The curves are derived in Appendix A. They can be used to estimate the optimal policy response to shocksh. The basic model has to be adapted to make it relevant to analyze monetary policy in emerging markets with large populations in low productivity employment. The steady-state full employment assumption of equilibrium models is far from adequate in these markets.i We consider a small open emerging market economy (SOEME) with two representative households consuming and supplying labour: above subsistence (R) and at subsistence (P). The product market structure, technology and preferences of R type consumers are the same across all economies. Productivity shocks differ since emerging markets are in transition stages of applying the new technologies becoming available. P type consumers are assumed to be at a fixed subsistence wage, financed in part by transfers from R types. The government intermediates these transfers through taxes on R. It runs a balanced budget so that η TR, t + M t = − (1−η) TP, t where a negative tax is a transfer. M t is government revenue from its monetary operations. The subsidy is calculated to give P a subsistence wage if they work eight hours daily, but they are free to increase their wages by working longer hours. P types are willing to supply more labor hours to the modern sector at a wage epsilon above their opportunity cost or wages in the informal sector. Since each country is of measure zero, it takes world prices as given.

h

i

Appendix A presents a simplified version of the Gali and Monacelli (2005, henceforth GM) small open economy model. This adaptation follows Goyal (2007)[10] . See the latter for detailed derivations, proofs, and systematic comparisons of the SOEME and the SOE.

The Natural Interest Rate in Emerging Markets

341

The intertemporal elasticity of consumption (1/σR), productivity and wages (WR) of R are higher, their labour supply elasticity (1/ϕR) is lower compared to the P, and they are able to fully diversify risk in international capital markets. Ni, t denotes hours of labour supplied by each type. Consumption of each type of good is a weighted average of consumption by the R and the P households, with η as the share of R. Since R and P consume home (H) and foreign (F) goods in the same proportion, Ct is distributed between R and P in the same proportion η, where η is the share of above subsistence households in consumption. The aggregate intertemporal elasticity of substitution, 1/σ, and the inverse of the labour supply elasticity j, ϕ, are also weighted sums with population shares of R and P as weights. Since P lack the ability to smooth consumption, their intertemporal elasticity of consumption approaches zero, so the averaging is done with elasticities, rather than inverse elasticities. The basic consumption Euler and household labor supply are derived for each type. Risk sharing can be derived only for R types. Payoffs D are taken as zero for P types, since they do not hold a portfolio of assets. To solve for St in terms of endogenous Yt and exogenous variables, first substitute CR, t and CP, t for Ct in the aggregate demand equal to supply equation and then substitute out CR, t using risk smoothing. This gives: St = (

Yt )σ D Yt*η C P1−,ηt

(7)

The terms of trade depreciate with a rise in Yt and appreciate with a rise in Yt*; but in a SOEME the former’s effect is magnified. CP, t also affects St, reducing the impact of Yt*. The multiplier factor,

σD =

σR η 1 α ) + ϖα ) − ( (

which affects only the SOEME, is large because the intertemporal elasticity of substitution is small. If σR = 1, then

ω =1 j

This is also the elasticity of price with respect to output in the aggregate supply curve derived. The labour supply elasticity of P can be expected to be high, and their intertemporal elasticity of consumption low. We normalize the latter at zero. Average ϕ is taken as 0.25 in the simulations, implying a labour supply elasticity of 4.

342

Ashima Goyal

and if 1/σP = 0, then σ = σR/η. It also follows that σD < σ. Both rise as η falls or the proportion of P with low intertemporal elasticity of consumption (1/σP = 0) rises. While η affects σ, both η and α affect σD. As α falls σD rises, and as α approaches 0, or the economy becomes closed, σD equals σ, which is its upper bound. In a fully open economy α approaches unity, and σD falls to its lower bound, the value unity in a SOE. The dynamic aggregate supply, which gives domestic inflation as a function of the output gap xt, now becomes:

π H = β E t {π H , t +1} + κ D x t

(8)

The slope for a SOEME is κD = λ(σD + φ). The corresponding value for a closed economy is λ (σ + ϕ) and for a SOE is λ (σα + ϕ), where

σα =

σR (1 − α ) + ϖα

,

σR enters σα since R in the SOEME are identical to the representative SOE consumer. The slope is reduced in an open compared to a closed economy since σ > σD > σα, but the slope can be higher in the SOEME compared to a SOE, even though ϕ is lower for the SOEME, since σD > σα. While σα = 1 if σR = 1, σD always exceeds unity if α < 1. Similar results hold for the more general case of σR ≠ 1. Since the gap between σ and σD is large and varies with η and α, the slope for the SOEME remains larger than in the SOE. The dynamic aggregate demand (AD) equation for the SOEME is derived from the consumption Euler (see Appendix A). Writing it in terms of the output gap gives a term in change in natural output, which defines the shock to the natural rate rrt :

x t = E t { x t +1} −

1

σD

( r − E {π t

t

H , t +1

} − rrt )

(9)

Where rrt = ρ − σ D Γ (1 − ρ a ) a t − σ D (1 − η + Φ ) E t {Δc P , t +1} + σ D ( Θ − Ψ ) E t {Δy t*+1}

and

The Natural Interest Rate in Emerging Markets

Θ = α (ϖ − η ) , d =

343

(1 + ϕ ) , Ψ = η σ − σ d , 1 , Γ= ( D) σ D +ϕ σ D +ϕ

Φ = d ( (1 − η )(σ − σ D ) )

Since σD > σα, the output gap is less responsive to the interest rate in the SOEME compared to the SOE. Thus (8) and (9) are the two AS and AD equations for the SOEME. 4. The Natural Rates of Output and the Terms of Trade

The natural level of output y t is the level where marginal cost is at its desired steady-state level − μ depending on the elasticity of demand. This has been derived for the SOE in the appendix and in our SOEME takes the value: y t = Ω + Γa t − Ψy t* − Φc P ,t + dσ Dκ

(10)

Where Ω=

(1 + ϕ ) , Ψ = η σ − σ d , ν −μ 1 ,d = , Γ= ( D) σ D +ϕ σ D +ϕ σ D +ϕ

Φ = d ( (1 − η )(σ − σ D ) )

(11)

As in the case for marginal cost, the natural output for a SOE has σα instead of σD and no cP, t and ĸ term. Since σ > σD > σα, but σR < σα when σR σα, the latter being the value in the SOE. The calibrations imply that the positive cP, t term dominates. The result is intuitive since the cP, t term captures the lack of maturity of the SOEME; it implies that s t is relatively depreciated compared to the value it will have when the cP, t terms disappear and the SOEME has become a SOE. In the latter case natural terms of trade must be appreciated compared to their value when the consumption gap is positive since of underdevelopment. The cP, t term captures the distance from world consumption levels that has to be overcome in the steady state. 5. Optimal Policy

The model is calibrated for a SOEME and the calibration is used to examine the levels of the natural rates, the types of shocks affecting the natural rates, their relative sizes, how these differ in the SOEME compared to the SOE, and how the former collapses to the latter after development is completed. The calibrated natural rates are conditional on the consumption of the poor, the distance from world incomes, other structural features, and world incomes. There are longerterm implications for policy as the SOEME catches up and the rates change. Short-term policy implications are also drawn for temporary shocks to the natural rate. The relative performance of different types of inflation targeting and of exchange rate intervention is examined through simulations. Because intervention has a rationale in the SOEME, leading to stickiness in the terms of trade, calibration and simulation was also done for this sticky S case. Uncovered interest parity (UIP) holds in some simulations but not when S or E is fixed. Limited capital account convertibility and some intervention that contribute to managing the exchange rates imply that UIP may not hold. The policy reaction function is estimated. The calibration is loosely based on Indian stylized facts. Empirical estimations and the dominance of administered pricing in SOEME’s suggest that past inflation affects current inflation (Fraga et. al., 2004)[8] , so a modification of the AS equation is made to accommodate such behaviour by imposing a share γb of lagged prices:

The Natural Interest Rate in Emerging Markets

ˆ ′t + γ bπ H , t −1 π H , t = γ f β E t {π H , t +1} + λ mc

345

γ f +γ b =1

In most simulations γb is set at 0.2 so γf is 0.8. Because of less than perfectly flexible interest rates, lagged interest rate also enters the AD with a weight of 0.2. The openness coefficient α is set at 0.3; the proportion of R,k η at 0.4; β = 0.99 implies a riskless annual steady-state return of 4 percent; the price response to output, ϕ, is set at 0.25, which implies an average labour supply elasticity of 4. Consumption of the mature economy and of the rich is normalized at unity, five times that of the poor so CP = 0.2. Given η, this gives consistent C values of 0.75, K of 1.1 so that cP = −1.6 and ĸ = 0.1. Initial conditions are normalized at unity so the log value is zero. The natural output y t is derived from the flexible price equilibrium, with an employment subsidy ν = − log (1 − τ ) set so as to correct for market power and for government temptation to change the terms of trade (GM, Section 4). In a SOEME it is also necessary to correct for the deviation from world income levels and poor infrastructure. The Appendix derives this value of the subsidy as ν = μ + log (1 − α ) − κ + log δ . The index of infrastructure δ is taken as 0.5 less than the world level of unityl. An elasticity of substitution between differentiated goods, ε equal to 6, implies a steady-state mark-up, μ, of 1.2. The value of ν- μ derived from the value of α, δ and ĸ is −0.9675. The price setting parameters are such that prices adjust in an average of one year (θ = 0.75), giving λ = 0.24. Since σR = 1 and 1/σP = 0, the implied average intertemporal elasticity of substitution is η(1 − α) + α = 0.58. A negative interest rate effect on consumption requires an intertemporal elasticity large enough so that the substitution effect is higher than the positive income effect of higher interest rates on net savers. Empirical studies have found real interest rates to have weak effects on consumption. Especially in low-income countries subsistence considerations are stronger than intertemporal factors. This is particularly so when the share of food in total expenditure is large. The elasticity Ogaki, Ostry

k

l

GMM regressions of CPI inflation for India (Goyal, 2009)[11] give a coefficient of expected inflation of 0.67. India’s share of imports in GDP was about 20 percent in 2005, and the proportion of population in rural areas 60 percent. In GMM regressions of aggregate demand with monthly data, the one period forward index of industrial production was strongly significant with a coefficient of –0.42. The Hall and Jones (1999)[12] index of social infrastructure gives 63 as the emerging market average compared to 14 as the developed country average, implying a δ of about a quarter. Since there has been extensive building of infrastructure and convergence in infrastructure over the past decade we take a value of 0.5.

Ashima Goyal

346

and Reinhart (1996)[16] estimate in a large cross-country study, varies from 0.05 for Uganda and Ethiopia to a high of 0.6 for Venezuela and Singapore. Our average elasticity compares well with these figures. Burns (2008)[3] estimates that the level of technology employed in developing countries is only one-fourth that in high income countries but technological progress increased 40-60 percent faster in the former than in the latter between the early 1990s and early 2000s. At is normalized at unity for the SOE. Since catch-up has been even faster in India in recent years we take its value to be 0.8 in the SOEME. Table 1: Value of natural rates due to each component Constant term

at = −0.2231 = log (.8)

y* = 0

cp = −1.6

κ = 0.1

Log value of natural rates

Natural rates

Component values

y

−0.4901

−0.1413

0

0.3773

0.0873

−0.1667

Y = 0.85

s

−0.8450

−0.1413

0

1.3373

−0.0127

0.3384

S = 1.4

rr

0.01

0.0024

0

−0.0319

−0.0185

−0.0185

Coefficient values

y

−0.2313

0.6332

−0.1572

−0.2358

0.8734

s

−0.3989

0.6332

−0.5572

−0.8358

−0.1266

rr

0.01

−0.0109

−0.00039

0.0193

These calibrations allow the calculation of the three natural rates and the contribution of each of the exogenous components to the natural rates. Table 1 shows these and also the coefficient values of each exogenous term. For the SOEME natural output is lower than world output while the natural terms of trade are higher. It shows the component values are dominated by cP. The distortions in a SOEME subsumed in the constant terms are also important. The coefficient value of the gap from world income levels is highest for natural output. These results are intuitive since low wages and productivity require a depreciated terms of trade for output to be competitive in world markets. The gap signifies the potential catch-up raising natural output. Subsistence consumption and the gap from world income levels, tend to raise natural output while technology and infrastructure backwardness reduce it. Over the longer-

The Natural Interest Rate in Emerging Markets

347

run policy has to take account of the changes in natural rates. For example, as catch-up occurs, improvements in technology and infrastructure will raise natural output while a rise in subsistence consumption levels or a closing of the potential gap would lower it. Policy makers have to be careful that they are using the correct natural output in their output gap variable so that they accommodate rather than choke changes in natural output. Permanent shocks affect natural output and terms of trade but not natural interest rates, since the change in natural output is permanent. Policy makers may raise interest rates with output if they do not realize that natural output has also risen so that the output gap is unchanged. Although lower rates help absorb labour in productive sectors, higher risk premiums in the SOEME may keep policy rates high. Short-run policy has to respond to temporary shocks. Table 1 shows shocks to subsistence consumption have the largest size among shocks affecting the natural interest rate and tend to reduce it. We take the exogenous force driving the dynamic impulse response as a calibrated 0.1 shock to the period one natural rate. The latter equals ρ = β-1 − 1. The policy response is obtained under discretion with a central bank minimizing different weighted averages of inflation (domestic or consumer), output and interest rate deviations from equilibrium values normalized at unity for the simulations ( L = qY y 2 + q π π 2 + q i i 2 ). The weights attached to the different arguments of the loss function (qs) ensure stability since the weight on inflation exceeds unity. Under strict inflation targeting only inflation has a positive weight of 2. The exchange rate directly affects consumer inflation while it affects domestic inflation through its affect on marginal cost. Monetary policy affects domestic inflation directly by changing the output gap. Domestic inflation is a component of consumer inflation. Table 2 reports some of the simulations. The benchmark set of parameters, for which sensitivity analysis is undertaken, is indicated. The unconditional standard deviations are reported. The square gives a measure of the welfare loss. The effect of monetary policy depends on the lag structure imposed. Since there are many administered prices in the Indian consumption basket we assume:

π t = Pt − Pt −1 ; π H ,t = PH ,t − PH ,t −1; π t = π H ,t −1 + α ( e t − e t −1 − π H ,t )

Ashima Goyal

348

Table 2: Simulations and volatilities

Simulations

Parameters

Standard deviations of (in percentages):

Benchmark:

η = 0.4 α = 0.3 φ = 0.25

Consumer inflation

Output

Domestic inflation

Exchange Rate

Interest rate (initial response)

DIT (cost shock)

qy = 0.7, qπH = 2, qi =1

0.58

0.36

1.08

0.82

0.70 (0.0256)

DIT, 0.01rn

qy = 0.7, qπH = 2, qi = 1

0.46 (−0.02)

0.16 (0.006)

0.31

1.59 (−0.054)

0.39 (0.0133)

DIT, 0.01rn

S Fix

0.31 (−.0036)

0.16

0.31

1.02

0.39 (0.0133)

DIT, 0.01rn

E Fix

0.21 (0.0068)

0.16

0.31

0.00

0.39 (0.0133)

CIT, 0.01rn

qy = 0.7, qπ = 2, qi = 1

0.45 (−0.004)

0.43 (0.0161)

0.59

1.85 (−0.0585)

0.38 (0.0063)

CIT, 0.01rn

S Fix

0.33

0.05 (0.0006)

0.22

1.05

0.54 (0.02)

CIT, 0.01rn

E Fix

0.54

0.30

0.77

0.00

0.84 (0.0296)

DIT, 0.01rn

S sticky model, S fix

0.12

0.32

0.10

0.38

0.11 (0.0039)

CIT, 0.01rn

S sticky model,

0.10

0.32

0.09

0.40

0.08 (0.0001)

CIT, 0.01rn

S sticky model, S fix

0.12

0.33

0.11

0.39

0.12 (0.0041)

Note: The bracketed terms give the value of the variable in the first period of the simulation

That is, consumer price inflation is a weighted average of lagged domestic prices and current depreciation. This structure seems to capture aspects of the relationship between Indian domestic and consumer price inflation, where the latter changes with a lag but is becoming more affected by current import prices as the economy opens out. Domestic inflation tends to lead consumer price inflation although the latter is normally more volatile. A large number of simulations were done, with different variations of the model, to establish robustness. This enabled crosschecking to remove errors. The policy response is on expected lines. There is stability or convergence back

The Natural Interest Rate in Emerging Markets

349

to the initial state by the 12th period. The volatilities and initial simulated variable values (in brackets) in selected simulations are reported in Table 2. The first row reports results of Goyal (2007) [10] in which comparison of different types of discretionary targeting, under a cost shock impacting domestic inflation, showed that domestic inflation targeting (DIT) delivered the lowest volatility and therefore highest welfare. This result continues to hold under a generic shock to the natural interest rate, since DIT delivers lower inflation and exchange rate volatility with only slightly higher interest rate volatility. The initial rise in the policy rate is higher than with CPI inflation targeting (CIT) so that the rise in income is lower, but still volatility is higher with CIT. DIT also performs better than CIT under more kinds of exchange rate management. Therefore the DIT is used as the benchmark again. The reason for higher volatility with CIT is excessive use of the exchange rate channel, combined with lags in the CPI (Table 2 and Figures 1 and 3).

0.04 0.02 0 -0.02 -0.04 -0.06 -0.08

consumer inflation output domestic inflation exchange rate interest rate 1 2 3 4 5 6 7 8 9 10 11 12

Figure 1: CIT: Positive 0.01 shock to natural interest rate

0.035 0.03 0.025 0.02 0.015 0.01 0.005 0 1

2

3

4

5

6

7

8

9

10

11

12

Figure 2: CIT and EFix: positive 0.01 shock to the natural interest rate

Ashima Goyal

350

0.02 0 -0.02 -0.04 -0.06 1

2

3

4

5

6

7

8

9

10

11

12

Figure 3: DIT: positive 0.01 shock to the natural interest rate

0.015

co nsumer inflatio n

0.01

o utput

0.005

do mestic inflatio n

0 -0.005

exchange rate

-0.01

interest rate

-0.015 1

2

3

4

5

6

7

8

9

10

11

12

Figure 4: DIT and Efix, positive .01 shock to natural interest rate The generic response to a rise in the natural interest rate is a rise in the policy rate. But because initially the gap between the policy and the natural interest rate falls, output rises, this raises domestic inflation, but the accompanying currency appreciation reduces consumer price inflation. The rise in the policy rate covers the expected future depreciation and slowly brings output back to steady-state levels (Figure 1). The response to a fall in natural rates to –0.01 (Figure 4 for DIT) is absolutely symmetric, with the signs reversed. Policy rates fall now. CIT has a lower increase in policy rates but greater volatility in every other variable. Considering the cases of exchange rate management, only the exchange and inflation rates change under DIT. With S fixed the only change from the benchmark is a lower appreciation and therefore lower fall in consumer price inflation (Figure 3). With E fixed there would be a slight rise in these prices for a positive shock (Figure 4). They fall with a negative shock. The large response in the policy rate under CIT limits the rise in output, without much gain in further inflation reduction (Figure 1). CIT targeting delivers a much lower rise

The Natural Interest Rate in Emerging Markets

351

in policy rates under perfectly flexible exchange rates but a much higher rise under fixed exchange rates, compared to DIT, so that DIT is more robust. CIT does almost as well only in the case of fixed terms of trade, since the exchange rate channel contributes to stabilizing inflation but excessive volatility is avoided. Under fixed exchange rates the exchange rate channel is not available at all leading to large policy rate response and more volatility. In the simulations above, market expectations are that the terms of trade will adjust to their natural level. If the policy commitment to fixed or managed rates is credible, factoring in fixed terms of trade changes the coefficients of the equations. Simulations were also conducted with this S sticky model. Notable is the much lower rise in the policy rate, much higher output response, and lower inflation and exchange rate volatility. The supply curve is much flatter, supporting the theoretical result of a flatter supply curve with S fixed (Table 2 and Figures 5 and 6). In the S sticky model with S fixed both DIT and CIT have very similar effects. Table 2 gives results for CIT with S fixed and flexible

0.015 0.01 0.005 0 -0.005 -0.01 -0.015 1

2

3

4

5

6

7

8

9

10

11

12

Figure 5: DIT in S sticky model, positive .01 shock to natural interest rate

0.015 0.01 0.005 0 -0.005 -0.01 -0.015 1

2

3

4

5

6

7

8

9

10

11

12

Figure 6: CIT in S sticky model, positive .01 shock to natural rate

Ashima Goyal

352

respectively although the latter is not consistent with a credible belief in fixed terms of trade. CIT does best if the terms of trade are credibly fixed. Although in general the performance of CIT improves with S fixed, its highly variable performance, suggests that DIT is more robust. Table 3 gives the reaction functions in the different cases. Noteworthy is the fall in the weight on the output gap in the S sticky model, again demonstrating the flatter supply curve. There is a weight on the exchange rate under CIT since CPI is a weighted average of the exchange rate and domestic inflation. Under S fix, when the change in exchange rate is moderated, the sign becomes negative while it is positive with flexible exchange rates. Table 3: Reaction Functions

Simulations

Parameters

Benchmark:

η = 0.4, α = 0.3, φ = 0.25

Shock dummy/ Output gap

Domestic inflation

DIT, 0.01rn

qy = 0.7, qπH = 2, qi = 1

0.0133

0.1282

−0.0659

DIT, −0.01rn

E Fix

−0.0133

0.1282

−0.0659

CIT, 0.01rn

qy = 0.7, qπ = 2, qi = 1

0.0063

0.2310

−0.0710

−0.0345

qyv0.7, qπ = 2, qiv1

−0.0063

0.2310

−0.0710

−0.0345

CIT, 0.01rn

S Fix

0.0200

0.0397

0.0728

−0.1036

CIT, 0.01rn

E Fix

0.0296

0.0192

0.0470

−0.0795

DIT, 0.01rn

S sticky model, S fix

0.0039

0.0254

CIT, 0.01rn

S sticky model,

0.0001

0.2837

−0.1266

−0.0137

CIT, 0.01rn

S sticky model, S fix

0.0041

0.0005

0.0086

−0.0207

CIT, −0.01 rn

Coefficients of: Exchange rate

Interest rate

−0.0206

The simulation results should be taken as only indicative since the model is not estimated, and is idealized in many respects. To give inputs for actual policy the lag structures specific to an economy have be built in. Even so, the structural SOEME features together with the microfoundations give useful insights for policy. 6. Indian Episodes

India has largely followed a monetary targeting approach. In the late nineties there was a switch to a multiple indicator approach. There is no formal inflation targeting but the policy statements give both inflation control and facilitating growth as key objectives. A specific value of 5 percent is given as the desirable

The Natural Interest Rate in Emerging Markets

353

rate of inflation, with the aim to bring it even lower in the long-term. Although the exchange rate was said to be market determined after the reforms and twostage devaluation of the early nineties, massive RBI intervention continued in order to absorb foreign inflows. Trend depreciation was allowed all through the nineties in order to cover the inflation differential and maintain the real effective exchange rate set in the early nineties. There was some appreciation due to the weakening of the dollar from 2002 and two-way movement of the nominal exchange rate was allowed from 2004. Foreign exchange reserves had been accumulating steadily since the opening out, but accelerated in this period. Inflation fell in the late nineties and continued low despite high growth and firming international oil prices, but it peaked in March 2007 and 2008. Throughout this period, gradual financial reforms deepened markets; most interest rates stopped being administered, and became an effective policy instrumentm. With the implementation of the liquidity adjustment facility (LAF) in 2001 policy was largely successful in keeping call money rates between the LAF bands determined by the policy repo and reverse repo rates, which began to be changed frequently and smoothly. The economy has suffered from frequent cost shocks, either from a failure of the rains or from international oil price shocks. Since there are administered interventions in the price of food and of oil, their point of impact on prices is known. Administered prices are normally raised after monetary interventions to bring down inflation rates (Bhattacharya and Bhattacharyya, 2001)[2]. Political sensitivity to the consumption of the poor, (in a democracy where they are still about 30 percent of the population, and 50 percent of average consumption basket is spent on food) has normally implied monetary accommodation of government expenditure during a drought, with a tightening immediately afterwards (Dash and Goyal, 2000) [7] . Our optimal monetary policy model does imply that policy should accommodate a temporary shock to the consumption of the poor, but instead of relying solely subsequent monetary tightening to bring down inflation, more nuanced policies that shift down the supply curve could be followed. Because of greater interest and exchange rate flexibility more policy options have now become feasible. We illustrate past policy responses and the variants suggested by our model below.

m

Agarwal (2008) establishes this empirically by examining monetary transmission in different post reform periods.

354

Ashima Goyal

The drought and terms of trade shocks over 1965-67, led to a fiscal tightening, with a cut in budget deficits and public investment. Monetary policy was non-accommodating but not severe. Fiscal and monetary policies were closely linked, as the budget deficit was automatically financed. The oil price plus agricultural supply shock over 1973-75 led to severe monetary and fiscal measures. In both cases there was an unnecessary loss of output. A focus on expanding food supply would have been more effective (Joshi and Little, 1994)[13]. The lesson had been learned by the 1979-80 crisis. There was no cut in public investment, no sudden monetary tightening, no long-term adverse effects on output, and a rapid recovery. But the populist fiscal response to supply shocks was having a cumulative effect in widening the revenue deficit. The response to the early nineties balance of payments crisis included a cut in public investment, an artificial agricultural supply shock as procurement prices for food grains were raised, and a monetary tightening to sterilize capital inflows in 1992-93. Growth revived in 1993-94, and monetary policy was accommodating, but exchange rate volatility in 1995 led to a monetary squeeze that precipitated a slowdown. The monetary stance was relaxed, but reversed again at the first sign of exchange rate volatility. Inflation fell, with the improvements in productivity, and the influence of low global inflation in a more open economy, but industrial growth did not revive until 2003, when Indian interest rates followed falling global interest rates. And even with higher growth, inflation remained low despite an extended period of high global oil prices. Macroeconomic policy was vitiated by the fiscal authority's populism and the monetary authority's tendency to squeeze demand, until it was rescued by falling global interest rates. The fiscal responsibility and budget management act (FRBM) enacted in 2003 was not designed to protect investment while controlling populism and inefficiency, so the fiscal authority continued excessive populism while being conservative with productive expenditure. Off balance sheet items like oil bonds destroyed the spirit of the act while satisfying the letter. The Reserve Bank had more autonomy, since it no longer had to automatically finance deficits; fiscal populism pushed it towards conservatism in order to reduce inflationary expectations. But since populism raised inefficiencies and therefore costs the supply curve shifted up, while monetary tightening reduced demand, resulting in a large negative effect on output for little gain in lower inflation. Monetary policy can use its knowledge of structure to fight inflation. Policy has to tighten only if there is excess demand. But supply shocks have been the

The Natural Interest Rate in Emerging Markets

355

dominant source of inflation. During a catch-up period of rapid productivity growth potential output becomes more uncertain. Excess demand can be removed without output cost if agents are forward looking, but a cost shock creates a short-run tradeoff between inflation and output variability. Since food inflation has high welfare costs, where food is still a large part of the consumption basket, this can be countered in the short-term by exchange rate policy, changes in tax rates, or other fiscal measures. Rise in wages in response to food prices has been important in second round propagation of Indian inflation. Inflation targeting would prevent the second-round inflationary wage-price expectations from setting in that can imply a permanent upward shift in the supply curve from a temporary supply shock. Optimal policy can aim to achieve a stricter inflation target only over the medium-term in order to allow time for temporary supply shocks to peter out. Short-term inflation targeting should be flexible. If two-way movement of the nominal exchange rate is synchronized with temporary supply shocks, and the exchange rate appreciates when there is a negative supply shock, it would lower intermediate goods and food prices. This differs from fixing the exchange rate to bring down high levels of inflation, which often leads to real appreciation and ends in a crisis, as in Latin American exchange-based stabilization episodes. Two-way movement only pre-empts the effect of temporary supply shocks on the domestic price-wage process. The nominal exchange rate reacts to temporary shocks, and the terms of trade to permanent. Productivity improvements would be required to tackle shocks like a permanent rise in global oil prices. In our model, the natural terms of trade must appreciate with a rise in the consumption of the poor. Unless it is productivity driven, such an appreciation would come through inflation. Equilibrium appreciation of natural rates gives policy more freedom to change nominal rates. 7. Conclusion

The optimizing model of a SOEME, with dualistic labour markets and two types of consumers, delivers a tractable model for monetary policy. The basic structure of the forward-looking aggregate demand and supply equations is the same as for the SOE and the closed economy, but the parameters depend on features of the labour market and on consumption inequality. These parameters also affect the natural rates. The SOEME collapses to the SOE model as inequality disappears.

356

Ashima Goyal

Although the slope of the aggregate supply curve is lower in an open compared to a closed economy and more elastic labor supply in the SOEME curve also lowers the slope, the SOEME supply curve is normally steeper than the SOE curve when the terms of trade are endogenous. The narrow section of the population sharing international risk magnifies the changes in the terms of trade due to its determinants, which now also include subsistence consumption. The lower intertemporal elasticity of consumption also magnifies the coefficients of the SOEME natural terms of trade. The positive term due to subsistence consumption levels dominates, implying that the current terms of trade is depreciated and reduction of the consumption gap between the SOEME and the SOE will appreciate the natural terms of trade. The subsistence consumption term is the largest determinant of the natural output as well as the terms of trade. Policy has to be careful to accommodate permanent changes in natural rates taking place as development occurs and the SOEME approaches the SOE. It must make sure it is targeting the output gap and not output as natural output changes. Multiple steady-states occur on the development path. Temporary shocks affect the natural interest rate. Frequent shocks to the consumption of the poor tend to decrease the natural interest rate. Optimal policy requires insulating the output gap from such shocks, so that the CB’s interest rate instrument should move in step with the natural rate. Thus the CB would accommodate supply shocks by lowering interest rates and offset demand shocks by raising interest rates. Comparing the performance of different targeting regimes in response to a temporary shock to the natural rate shows that DIT delivers lower inflation and exchange rate volatility with only slightly higher interest rate volatility. CIT delivers a much lower rise in policy rates under perfectly flexible exchange rates but a much higher rise under fixed exchange rates so that DIT is more robust. Moreover, the volatility of all other variables is higher under CIT. Excessive use of the exchange rate channel under CIT causes higher volatility. CIT does as well only in the case of fixed terms of trade, since the exchange rate channel contributes to stabilizing inflation but excessive volatility is avoided. Under fixed exchange rates the exchange rate channel is not available at all leading to large policy rate response and more volatility. If the policy commitment to fixed or managed rates is credible, factoring in fixed terms of trade changes the coefficients of the equations. Notable in this S sticky model is the much lower rise in the policy rate, much higher output response, lower change in inflation, and in the exchange rate. That is, the supply

The Natural Interest Rate in Emerging Markets

357

curve is much flatter, supporting the theoretical result of a flatter supply curve when S is not endogenous. Goyal (2007)[10] had established that monetary policy does best, in the presence of cost shocks, by flexible targeting of domestic inflation, giving some weight to the output gap and interest rate smoothing. A similar result holds here for generic shocks to the natural rate. Middling exchange rate regimes that imply terms of trade stay close to natural rates do best, and both DIT and CIT work well in such regimes. Limited flexibility of the nominal exchange rate contributes to reducing inflation, but aggressively using the direct exchange channel at short-horizons is not optimal. Neither is fixing the exchange rate. Limiting capital account convertibility can help implement the required exchange rate policy. In particular impediments are required on short-term arbitrage so that the UIP does not hold. The cost of reserve accumulation will imply a higher tax on the rich but there are gains from the reduction in volatility. Lower inflation volatility will especially benefit the poor. The terms of trade can be fixed only in the short-run, but over time have to accommodate changes in the natural terms of trade. The simulations give results in line with theory, but much work remains to be done. First, estimations and analysis for India (Goyal, 2009)[11] suggest that CPI inflation may be forward-looking but domestic inflation is not. This could be expected due to the increasing impact of exchange rates. It will be useful, therefore to simulate the model imposing this restriction. Second, the implications of pricing to market or in importer’s currency can also be explored although this may not be so relevant for commodity imports. Including capital markets may imply a greater role for interest smoothing. Third, SOEMEs have different kinds of nominal and real wage rigidities. It would be particularly useful to model the consequences of real wages rigid in terms of food prices (Goyal, 2009)[11], since this is a feature of populous low-per capita income SOEMEs, and of forward-looking wage setting. Fourth, to explore the consequences of relaxing simplifying assumptions, including on elasticities and on uncorrected steady-state distortions, such as deviations from PPP, and on asset accumulation through the current account. A non-zero current account implies that multiple steady states can exist. Fifth, derive optimal weights for the loss function specifically for a SOEME, given its characteristics. This will allow a more robust welfare analysis.

358

Ashima Goyal

References

1. A. Agarwal, Inflation Targeting in India: An Explorative Analysis’, PhD thesis submitted to IGIDR (2008). 2. K. Bhattacharya and I Bhattacharyya, Economic and Political Weekly XXXVI (51), 4735 (2001). 3. A. Burns, H. Timmer, E. Riordan and W. Shaw, World Bank Report, accessed from http://go.worldbank.org/TC26UFESJ0 on 16th January, 2008 (2008). 4. L.J. Christiano, M. Eichenbaum and C.L. Evans, Handbook of Macroeconomics 1, 65, Elsevier: North Holland (1999). 5. R. Clarida, J. Gali and M. Gertler, Journal of Economic Literature 37 (4), 1661 (1999). 6. R. Clarida, J. Gali and M. Gertler, American Economic Review 91 (2), May, 248 (2001). 7. S. Dash and A. Goyal, Indian Economic Journal 48(1), ( 2000). 8. A. Fraga, I. Goldfajn, and A. Minella, NBER Macroeconomics Annual, Cambridge: MIT Press. (2004) 9. J. Gali and T. Monacelli, Review of Economic Studies 72(3), 707 (2005). 10. A. Goyal, A General Equilibrium Open Economy Model for Emerging Markets, paper presented at ISI International Conference on Comparative Development on 26th May 2008. . An earlier version is available as IGIDR working paper WP-2007-016. (2007) 11. A. Goyal, Journal of Quantitative Economics 22 (1), forthcoming. (2009). 12. R.E. Hall and C.I. Jones, Quarterly Journal of Economics 114(1), 83 (1999) 13. V. Joshi and I.M.D. Little. India: Macroeconomics and Political Economy 1964-1991, Oxford University Press, Delhi. (1994) 14. R. Mohan. ‘Monetary Management in Emerging Market Economies: Concerns and Dilemmas’, Comments by Deputy Governor, Reserve Bank of India at Policy Panel at the NBER Conference on International Dimensions of Monetary Policy at S'Agaro, Catalonia, Spain on June 12.

The Natural Interest Rate in Emerging Markets

359

15. M. Obstfeld, and K. Rogoff, Foundations of International Macroeconomics, Cambridge, Massachusetts: MIT Press ( 1996). 16. Masao Ogaki, Jonathan Ostry, and Carmen M. Reinhart: IMF Staff Papers 43 (March), 38 (1996). 17. M. Woodford. Interest and prices: Foundations of a Theory of Monetary Policy, NJ: Princeton University Press (2003). Appendix A. Deriving aggregate demand and supply in a SOE

The generic form of the objective function the representative consumer maximizes is: ∝

E0 ∑ β tu (Ct , N t )

(A1)

t= 0

Consumption, C, increases and labour, N, decreases the discounted present value of utility with β is the discount factor. Underlying the macro variables is CES aggregation, over i ∈ [0, 1] countries, and j ∈ [0,1] product varieties. Aggregate consumption, C, is derived from CES aggregation of consumption of home and foreign goods ( C H , C F ). If the elasticity of substitution between H and F goods is equal to unity, the CES aggregation collapses to Cobb-Douglas: C t = kC 1H−,αt C Fα ,t

(A2)

Where k = 1/((1 − α)1-α − αα) is a constant and α is an index of openness. The associated consumer price index (CPI) is: Pt = ( PH ,t )

1−α

( PF ,t )

α

(A3)

CH, t is itself an index of consumption of domestic goods derived by CES aggregation with elasticity of substitution ε > 1 over j domestic varieties. CF, t is an index of imported goods, derived by CES aggregation with elasticity of substitution γ = 1 over imported goods j, from i countries of origin, Ci,t. Thus Ci,t is an index over j goods imported from country i and consumed domestically. There is also CES aggregation with elasticity of substitution ε > 1 between j varieties produced within any country i.

360

Ashima Goyal

The other great simplification in a SOE is that foreign variables are independent of home country action, and can be taken as given. Variables with a superscript * indicate foreign countries. The specific form of the utility function is: u (Ci , N i ) =

C i1,−t σ i N i1,+t φi − 1 − σ i 1 + φi

(A4)

Since each country i is assumed to have identical preferences the subscript i can be dropped. The objective function is maximized subject to the period budget constraint: ⎧D ⎫ Pt C t + E t ⎨ t +1 ⎬ ≤ Dt + Wt N t + Tt ⎩ Rt ⎭

(A5)

Where Pt C t = PH ,t C H ,t + PF ,t C F ,t and Rt is the gross nominal yield on a riskless one- period discount bond paying one unit of domestic currency in t + 1 so 1/Rt is its price. Security markets are complete. Dt+1 is the random payoff of the portfolio purchased at t. Differentiating with respect to the two arguments C and N and over time gives the intratemporal optimality condition: C tσ N tφ =

Wt Pt

(A6)

And intertemporal optimality or the consumption Euler: ⎧⎪⎛ C t +1 ⎞ σ ⎛ Pt ⎞ ⎫⎪ ⎟ ⎜ ⎟⎬ = 1 ⎩⎪⎝ C t ⎠ ⎝ Pt +1 ⎠ ⎭⎪

β Rt E t ⎨⎜

(A7)

Log- linearized forms of these FOC’s are: wt − p t = σ c t + ϕ n t c t = E t {c t +1} −

1

σ

( rt − E t {π t +1} − ρ )

(A8) (A9)

Where ρ, the discount rate, equals β-1 − 1 and πt, CPI inflation, is given by πt = pt − pt-1. Small letters normally denote log variables.

The Natural Interest Rate in Emerging Markets

361

Optimal allocation of expenditure between domestic, H, and imported goods, F, gives: C H ,t = (1 − α )

Pt Ct PH ,t

Pt Ct PF ,t

C F ,t = α

(A10) (A11)

Identities and relationships between different types of inflation and real exchange rates are also required. Log- linearization of CPI gives: p t = (1 − α ) p H ,t + α p F ,t

(A12)

The effective terms of trade is: St =

PF ,t PH ,t

(A13)

Or in log terms: p F ,t = s t + p H ,t

Substituting in CPI (Eq. A12) gives: p t = p H ,t + α s t

π t = π H ,t + αΔs t

(A14)

Or That is, CPI inflation is a weighted average of domestic inflation and the terms of trade. The real exchange rate, Q, is related to the terms of trade as follows: Q=

EP * P

q t = et + p t* − p t = s t + p H ,t − p t = (1 − α ) s t Q = S t(

1−α )

The identity p F ,t = e t + p t* is used in the derivation.

(A15)

Ashima Goyal

362

International risk sharing: The consumption Euler for any other country i, with its prices translated into home country prices using the nominal exchange rate, is: ⎛Ci ⎞ β ⎜ t +i1 ⎟ ⎝ Ct ⎠

−σ

⎛ Pt i ⎞⎛ ε ti ⎞ 1 ⎜ P ⎟⎜ ε i ⎟ = R t ⎝ t +1 ⎠⎝ t +1 ⎠

(A16)

Using the equivalent Euler equation for the home country, the definition of Q, and integrating over i ∈ [0, 1] countries to get Ct*, gives: 1

C t = ν C t*Q σ

In logs, c t = c t* + = c t* +

1

σ

qt

(1 − α ) σ

st

(A17)

Symmetric initial conditions and zero net foreign holdings are assumed so that υ = 1. In the symmetric steady state with PPP, C = C* and Q = S = 1 would also hold. Aggregate demand and output equality: For goods market clearing in the SOE, domestic output must equal domestic and foreign consumption ( C H* ) of home goods: Yt = C H ,t + C H* ,t

(A18)

We show below that substituting the allocation FOCs (A10) and (A11) in (A18) and simplifying, with σ = 1, this demand supply equality reduces to: Yt = S tα C t

(A19)

A.1

The allocation of foreign consumption to goods produced in the SOEME is the same as FOC (A11) with P*t C*t instead of Pt Ct. Multiplying and dividing by P*F,t and converting the numerator P*F,t into SOEME prices using the nominal exchange rate gives:

The Natural Interest Rate in Emerging Markets

C H* ,t = α

ε PF*,t Pt* PH ,t PF*,t

363

C t*

Of the two relative prices, the first one compares the price of SOEME goods to all other foreign goods translated into SOEME prices. The second relative price compares the foreign country price index to the price index of all other foreign goods. Thus more SOEME goods are imported as a function of these two relative prices, the weight of foreign goods in the consumption basket, and aggregate foreign consumption. Multiplying and dividing by Pt and substituting Qt: C H* ,t = α Qt

Pt * Ct PH ,t

Substituting the FOC for the SOEME consumer (A10) and that just derived for the foreign consumer, in the aggregate demand = supply Eq. (A18) for the SOEME ( Yt = C H ,t + C H ,t * ), gives: Yt =

(1 − α ) Pt C t + α Q t PH ,t

Pt * Ct PH ,t

Substituting out C t* using risk sharing Eq. (A17): Yt =

(1 − α ) Pt C t + α Q PH ,t

t

1 − Pt CtQ σ PH ,t

Simplifying and assuming σ =1 gives: Yt =

1 1− Pt C t (Q σ ) PH ,t

Ct = S α CtQ

Which simplifies to (A19).

1−

1

σ

Ashima Goyal

364

Determinants of the terms of trade: Substituting risk sharing again (with σ = 1) in aggregate demand equals supply Eq. (A19), we get: Yt = S α C t*Q

(A20)

Substituting Qt = S t1−α Yt = S tα Yt* S t1−α St =

Yt Y*

(A21)

That is, the terms of trade depreciate with a rise in home output relative to world output. Deriving aggregate supply: A simple log-linear production function where output increases with labour input and its productivity, gives marginal cost Eq. (A23) as a function of unit labour costs, from the firms’ optimization, y t = a t + nt

(A22)

mc t = −ν + wt − p H ,t − a t

(A23)

The employment subsidy τ or ν = − log (1 − τ ) , guarantees the optimality of the flexible price outcome, since it induces firms to increase employment to the social optimum. Adding and subtracting pt: mc t = −ν + ( w − p t ) + ( p t − p H ,t ) − a t

Substituting the intratemporal FOC (A8) and CPI (A14): mc t = −ν + σ c t + ϕ n t + α s t − a t

(A24)

Substituting risk sharing (A17), production function (A22), and from A21) y t* = y t − s t : mc t = ν + σ y t* + ϕ y t + s t − (1 + ϕ ) a t = −ν + (σ + ϕ ) y t − (1 + ϕ ) a t

(A25)

The Natural Interest Rate in Emerging Markets

365

The log of gross mark up in steady state, mc, falls as elasticity of demand rises: mc = − log

ε ε −1

≡ −μ

The difference of actual from this optimal marginal cost is: ∧

mc t = mc t − mc

Under Calvo-style staggered pricing, where 1-θ percent of firms change prices in a period, the firm’s optimal price-setting can be shown to give the dynamics of domestic inflation as a function of real marginal cost and discounted expected future inflation (GM Appendix B): ∧

π H ,t = β E t {π H ,t +1} + λ mc t

λ≡

(1 − βθ )(1 − θ ) θ

(A26)

The deviation of marginal cost from its optimum is related to the output gap, x t ≡ y t − y t , or the deviation of y from steady state y t . The latter is derived from mct (A25) by imposing mct = -μ and solving for yt. If σ =1 then: yt =

v−μ + at 1+ϕ

Subtracting yt from y t , substituting for yt from the mct equation (A26) and for y t from above shows how the deviation of mc from its optimal rises with the output gap: ∧

mc t = (σ + ϕ ) x t

Combine with the price setting equation (A26) to get aggregate supply:

π H ,t = β E t {π H ,t +1} + κ x t κ = λ (1 + ϕ)

(A27)

This is the New Keynesian Phillips Curve. It differs from the standard Phillips Curve in including forward-looking variables, which enter since it is derived

Ashima Goyal

366

from microfoundations with optimization over time. Similarly for aggregate demand derived below. Substituting c in the Euler Eq. (A9) with y from the aggregate demand equal to supply Eq. (A19) and log-linearizing gives: y t = E t { y t +1} −

1

σ

( rt − E t {π t +1} − ρ ) − σα E t {Δs t +1}

Converting to domestic prices using π t = π H ,t + αΔs t , y t = E t { y t +1} −

1

σ

( rt − E t {π H ,t +1} − ρ )

Writing in terms of output gaps, x t = E t { x t +1} −

1

σα

( rt − E t {π H ,t +1} − rrt )

(A28)

Aggregate demand is less interest elastic in an open compared to a closed economy, since σ α equals unity if σ =1 and σ α < σ otherwise. rrt = ρ − σ α (1 − ρ a ) a t + χ E t Δy t*+1

(A29)

If σ =1, χ = 0 so y*drops out of the equation. World income then does not affect aggregate demand. All the exogenous shocks affecting AD now come through the rr t term. Appendix B. To derive the unemployment subsidy

The flexible price equilibrium in the SOEME, with variables denoted by an upper bar, must satisfy: 1−

1

ε

= MC t

(B1)

Where ε is the elasticity of demand. Real optimal or steady state marginal cost (1 − τ ) δ W . The subsidy defined above must also equal real unit wage costs: − At P H ,t τ decreases unit labour cost and costs imposed by poor infrastructure increases

The Natural Interest Rate in Emerging Markets

367

it. If world infrastructure quality is unity then δ < 1 measures poor infrastructure in the SOME which adds to the marginal cost facing a firm. The first order condition from the consumer’s choice, the definition of the terms of trade and the aggregate demand equal to supply identity in the SOEME give:

(

)

UN Ct,N t W =− Pt U C ⎡⎣C t , N t ⎤⎦ α

S =

from

Pt Yt = P H ,t C t K α

Y t = CtS K

Substituting the relations above in (B1) we get: 1−

1

ε

=−

δ (1 − τ ) At

α

St

W δ (1 − τ ) Y U N = At Pt KC t UC

Substituting the value of the derivatives of the utility function 1−

1

ε

=

δ (1 − τ ) Yt At

KC t

ϕ

Nt Ct

(B2)

In a SOEME, which takes world output and consumption as given the optimal allocation must satisfy:

−

UN C = (1 − α ) t UC Nt

And from the derivatives of the utility function −

UN = Ct N ϕ UC

Equating the two gives: 1

N = (1 − α ) 1+ϕ

(B3)

Substituting (B3) in (B2) using the definition of the production function gives:

Ashima Goyal

368

1−

1

ε

=

δ (1 − τ )(1 − α ) K

Taking logarithms:

ν − μ = log (1 − α ) − κ + log δ

(B4)

Where μ = log (ε/(ε-1)). The optimal marginal cost or log of the gross mark-up in a flexible price economy is − μ, so setting ν = − log (1 − τ ) such that (B4) holds gives the equivalent of the optimal flexible price equilibrium.

INTERGOVERNMENTAL TRANSFER RULES, STATE FISCAL POLICY AND PERFORMANCE IN INDIA POULOMI ROY Department of Economics, Surendranath College, Kolkata-700009, India, and Department of Economics, Jadavpur University, Kolkata-700032, India Email: [email protected] AJITAVA RAYCHAUDHURI Centre for Advanced Studies, Department of Economics, Jadavpur University, Kolkata-700032, India Email: [email protected] In the federal economy like India intergovernmental transfer policies affect the state revenue and expenditure policies. This paper provides a theoretical model of determining optimal fiscal policy of the state governments in India. State’s optimum fiscal policy depends on the rules applied by transferring agencies in transferring funds to the sub national governments. Three important criteria revenue effort, deficit financing and distance criterion are considered to estimate the weight assigned to these criteria. The comparison of actual state own revenue and expenditure policies with the optimum policy reveals that states are spending more than estimated optimum level and collecting revenues less than the optimum level. The deviation of actual values from the optimum values also give us some idea regarding to which direction the state governments should change its existing revenue and expenditure policies. The analysis is based on fourteen major states in India over the period 1980–2005.

1. Introduction The constitution of India provides independent revenue raising and spending power to both the central and the state governments. It also admits the existence of vertical imbalances in taxing power. The expenditure responsibilities of the state governments on the other hand are higher. The constitution thus directs the central government to transfer resources. Transfers by the central government are meant to bridge the gap between resources required by states to meet their assigned responsibilities and the resources they can raise themselves. Three-tier transfer mechanism exists in India. The central government transfers funds in India via Finance Commission, Planning Commission and 369

370

Poulomi Roy and Ajitava Raychaudhuri

discretionary transfers through various union ministries and agencies. Low taxing power and high expenditure responsibilities make the state governments dependent on the central government for resources. Transfer from the centre covers large part of revenue of the state governments. In this chapter we have studied the impact of intergovernmental transfer on the state fiscal performance. The review of literature (Rao and Singh13 , Rao14, Rao16, Bajpai and Sachs2, Sen and Trebesch17) on state finances and the intergovernmental transfer mechanism in India indicates that most of the studies have examined whether the vertical and horizontal imbalances in the federal transfer mechanism exists and how the design of transfer system can be improved to distribute resources equitably. Ma8 evaluated the intergovernmental transfer mechanism of different countries and suggested methods of determining fiscal capacities of provinces. On the tax side of the state finances, Coondoo et al.4 , Rao and Oommen9 have estimated the tax capacity of the states and the tax effort given by the states in collecting revenue at the state level. Coondoo, Majumder, Mukherjee and Neogi3 examined the relative tax performance of the states in India for the period 1986–87 to 1996–97. Sen18 also calculated the tax effort index of various categories of taxes for 15 major states in India for the period 1991–92 to 1993–94. Rao11 and Bajpai and Sachs2 examined the situation of state finances in India. Rao11 finds that situation of state finances deteriorated after 1990–91. State finances in India are adversely affected by low buoyancy of central transfer. Bajpai and Sachs2 find that reform of the state fiscal system is necessary in order to reduce expenditure and increase revenue. They find that inefficient intergovernmental transfer mechanism in India is responsible for fiscal indiscipline at the state level. Rao12 evaluated the transfer system in India and its possible impact on state fiscal performance. It is argued that the share of incentive linked transfer in India is very small in India and thus its influence on state fiscal performance is very little. Rajaraman and Visstha10 find that an increase in non-matching grants to panchayats affects the local tax effort negatively in districts of Kerala. GR6 argued that the negative relationship between tax effort and grants is arrived by Rajaraman and Vasistha10 because of their assumption that population size represents tax capacity. Assuming the same tax effort over the districts in Kerala GR6 shows that the negative relationship between tax revenues and grants obtained by Rajaraman and Vasistha10 rather represent the negative relationship between grants and taxable capacity.

Intergovernmental Transfer Rules, State Fiscal Policy and Performance in India

371

The paper by Sinelnikov, Kadotchnikov, Trounin and Schkrebela19 relates the rules applied in intergovernmental transfer mechanism for the Russian economy and its impact on the regional optimal tax and expenditure. Another paper by Dahlby5 has derived the optimal tax and expenditure ratios considering borrowing as one of the sources of financing deficit. Sinelnikov, Kadotchnikov, Trounin and Schkrebela19 in their paper considered that expenditure on public goods and services is financed by taxes and transfers but borrowing is not considered. In Dahlby5 intergovernmental transfer mechanism is not considered. Our theoretical model follows from these two papers where we have addressed the impact of rules applied in transferring resources on the optimum fiscal performance of a state considering the fact that deficit is financed by borrowing. Our model is different from these two models in the sense that in our model we have considered the role of transfer along with the case that after devolution of transfers, deficit is financed by market borrowing. The transfer formula used in this model is also relevant for Indian economy. We have tested the theoretical model developed in this chapter using the state level data from the Indian economy. Dahlby5 finds that public debt ratio affects the optimal tax ratio but it does not affect the optimal expenditure ratio. But in our model we find that both the revenue and expenditure ratios depend on public debt ratio. In our study we have assumed a simple formula of transferring resources to the states and derived the optimum revenue and expenditure of a state from the utility maximizing principle. Instead of examining the relationship between total transfer and fiscal performance of a state we have tried to show how weights given by transferring agencies to revenue effort index, deficit financing criterion and distance of per capita income criterion affect the optimum revenue and expenditure of a state. Formula used and weight assigned to various criteria in transferring resources by several Finance Commissions seems to be somewhat arbitrary and subjective. As written in the “Memorandum to The Twelfth Finance Commission of the Government of Gujarat”a in page 21 that “The weights assigned by the various Finance Commissions to the parameters used in the formula of horizontal distribution does not seem to flow from any comprehensive theoretical framework. If there exists a scientific basis for deriving these numbers, none of the Finance Commissions has cared to explain it properly in their reports and hence it gives an impression that they are

a

http://fincomindia.nic.in/pubsugg/memo_guj.pdf

Poulomi Roy and Ajitava Raychaudhuri

372

arbitrary and subjective. In a note appended to the main report of the Fourth Finance Commission, the chairman, Dr. P.V. Rajamannar, had remarked that the selection of a particular set of factors and weights assigned to them for determining the shares has largely remained subjective and continues to be “a gamble on the personal views of the five persons or a majority of them”. In our study we have used a simple formula for transferring resources and statistically estimated the coefficients associated with selected criteria. Thus objective of this study is to find out the weight assigned by transferring agencies to three important criteria in per capita transfer of funds. Having identified these parameters we have estimated the utility maximizing level of optimum revenue and expenditure of fourteen major states in India. Our period of analysis is 1980 to 2005. The chapter is organized in the following way. Section 2 summarizes the criteria used by several agencies in transferring funds in India, Section 3 provides the theoretical model of estimating optimum revenue and expenditure of a state, Section 4 explains the methodology used for empirical analysis, Section 5 represents the data source and variables, section 6 analyses empirical results and Section 7 provides the conclusions derived. 2. Rules Applied in Transferring Resources in India In analyzing intergovernmental transfer mechanism in India it is very important to know the criteria used by different finance and planning commissions. In India, Finance Commission (FC), Planning Commission and different central ministries transfers resources to the states on the basis of a few criteria. In this section we have discussed about the criteria used by different Finance and Planning Commissions. 2.1. Finance Commission’s Transfer Criteria used by different finance commissions in giving grants in aid and sharing income tax and excise tax are summarised below (Table 1): Table 1. Finance Commission Criteria for sharing taxes and grants-in-aid 1st FC

Grants

(i) For seven states to cover their deficits during the period 1951-56; (ii) For eight states to improve their primary education facilities

Share in taxes

Income taxes were shared in the following way: 80 percent on the basis of population and 20 percent on the basis of revenue collection of the state. 40 percent of the net proceeds of excise duties were to be distributed among the states on the basis of population.

Intergovernmental Transfer Rules, State Fiscal Policy and Performance in India

373

2nd FC

Grants

Larger grants in aid were provided to meet development needs of states.

Share in taxes

Income taxes are to be distributed in this way-90 percent pf tax collection was to be distributed on the basis of population and 10 percent on the basis of revenue collection. 25 percent of the net proceeds from excise duties were to be distributed among states.

3rd FC

Grants

(i) Rs. 550 crores to all states except Maharashtra to cover part of their revenue expenditure; (ii) Rs. 45 crores for improvement of communications

Share in taxes

For income taxes, 80 percent was distributed on the basis of population and 20 percent on the basis of revenue collection of the state. In case of excise tax there is an increase in the number of commodities in the divisible pool from 8 to 35 by including all commodities on which duties were collected in 1960-61 but reduced the state’s share from divisible pool from 25 percent to 20 percent.

4th FC

5th FC

6th FC

7th FC

Grants

Rs. 610 crores to cover deficits during the period 1966-71

Share in taxes

80 percent on the basis of population and 20 percent on the basis of revenue collection of the state, income taxes were shared. In case of excise tax the number commodities had been increased to 45. The share of commodities was retained at 20 percent.

Grants

Rs. 638 crores to cover deficits during the period 1969-74

Share in taxes

Population was the major criterion of devolution of income tax. Did not make any change for excise taxes.

Grants

Rs. 2510 crores for fourteen out of twenty one states to cover their nonplan revenue deficit

Share in taxes

Population became the major criterion of devolution of income tax. Did not make any change for excise taxes.

Grants

Rs. 1600 crores to cover deficits of a few poor states during the period 1980-85 and also to upgrade the standard of administration

Share in taxes

Population became the major criterion of devolution of income tax. For excise taxes, 7th FC raised the state’s share to 40 percent of the net proceeds. 25 percent weightage was equally given to population, increase in per capita income of the state, the percentage of poor in each state, a formula for income equalization between states.

Poulomi Roy and Ajitava Raychaudhuri

374

8th FC

9th FC

10th FC

Grants

(i) A small grant of Rs. 1556 crores for the period 1985-90 to cover deficit; (ii) A grant of Rs. 915 crores to certain states to upgrade the standard of administration

Share in taxes

For income taxes, (i) 10 percent on the basis of income tax collection. (ii)Out of remaining 90 percent, 25 percent on the basis of population, 25 percent on the basis of inverse of per capita income of the state multiplied by population, 50 percent on the basis of the distance of per capita income of a state from the highest per capita income state multiplied by population of the state. For excise tax 8th FC raised state share to 45 percent and introduced the same formula as 7th FC for 40 percent of proceeds and retained 5 percent share to distribute that among deficit sates.

Grants

(i) Grant of Rs. 15017 crores to cover deficits of plan and non-plan revenue account during 1990-95. (ii) A special annual grant of Rs. 603 crores towards centre’s contribution to the calamity relief fund-totaling Rs. 3015 crores for five year period, 1990-95. (iii) A grant of Rs. 122 crores to Madhya Pradesh towards expenditure on rehabilitation and relief of victims of Bhopal gas leak.

Share in taxes

For income tax, 9th FC basically followed the above formula with minor modification. Ninth FC added one more criterion, that is, backwardness of sates based on population of scheduled castes and scheduled tribes, number of agricultural labourers in different states as revealed in 1981 census. According to NFC the composite index would correctly reflect poverty and backwardness of a state in large measure. The states having larger share of these components were required to bear substantial expenditure responsibilities. For excise taxes, 9th FC proposed to distribute the entire amount of 45 percent as a consolidated amount. The formula used was: 25 percent on the basis of 1971 census, 12.5 percent on the basis of index of backwardness, 33.5 percent on the basis of per capita income of the state from highest per capita income state, 12.5 percent on the basis of income adjusted total population, 16.5 percent among states with deficits, after taking into account their shares from all sharable taxes.

Grants

(i) Grant of Rs. 15017 crores to cover deficits of plan and non-plan revenue account during 1990-95. (ii) A special annual grant of Rs. 603 crores towards centre’s contribution to the calamity relief fund-totaling Rs. 3015 crores for five year period, 1990-95. (iii) A grant of Rs. 122 crores to Madhya Pradesh towards expenditure on rehabilitation and relief of victims of Bhopal gas leak.

Share in taxes

For income taxes, 10th FC proposed (i) 20 per cent on the basis of population of 1971. (ii) 60 per cent on the basis of distance of per capita income of a state from that of state having highest per capita income.(iii)5 per cent on the basis of area adjusted. (iv) 5 percent on the basis of index of infrastructure. (v) 10 percent on the basis of tax effort. 47.5 percent of net proceeds from excise taxes were distributed among states. Using the same formula as used in sharing of income taxes, 40 percent of excise taxes were distributed among major states. Remaining 7.5 percent taxes were distributed among deficit states.

Intergovernmental Transfer Rules, State Fiscal Policy and Performance in India

11th FC

375

Grants

(i) Rs. 35,359 crores was provided among states facing revenue deficit after devolution of grants. (ii) For upgradation of administration and special problems associated with certain states Rs. 4793 crores was provided. (iii) A total grant of Rs. 10000 crores has been provided to support local bodies at the panchayat level and municipalities at the urban level. To panchayats Rs. 8000 cores and to municipalities Rs. 2000 crores for the five year period (2000-05) were provided.

Share in taxes

(i) 10 percent on the basis of population, (ii) 62.5 percent on the basis of distance of per capita income from that of state having highest per capita income, (iii) 7.5 percent on the basis of are, (iv) 7.5 percent weight is given to index of infrastructure, (v) 7.5 percent weight is given to fiscal discipline.

2.2. Planning Commission’s Transfer The Planning Commission on the other hand transfers resources on the basis of population, per capita income, tax effort, fiscal management, literacy, land reform etc. The transfer of the planning commission is based on a formula where 30 percent of the transfers are in the form of grants and 70 percent as loans. States cannot accept only grants without taking loans. Thus grants and loans are tied together. In the following paragraph we have discussed about the formula used by the planning commission in transferring resources to the states. 60 percent of the planning commission’s transfer is based on population of the state. 15 percent of transfers are based on the following formula. 7.5 percent of resources are transferred on the basis of (a) tax effort (b) fiscal management (include the speed of utilization of committed foreign aid and state’s performance of revenue collection) and (c) progress in respect of national objectives, and another 7.5 percent of transfers are allocated to meet special problems of the states such as, population control, literacy and land reform. The remaining 25 percent of transfers were made on the basis of per capita state domestic product (SDP) based on the following formula. 20 percent was given only to states with less than average per capita SDP on the basis of the inverse formula; and the remaining 5 percent according to the distance formula. The inverse formula is given by ( Pi / Yi ) / ( Σ ( Pi / Yi ) ) which is inversely related to the per capita income of the state. The distance formula is given by: (Yh − Yi ) Pi / ( Σ (Yh − Yi ) Pi ) where Yi and Yh denote per capita SDP of the i-th and the richest state respectively, Pi is the population of the i-th state. The indicator increases as the distance of income of the i-th state from the richest state increases. Keeping these in mind we have used revenue effort, budgetary

Poulomi Roy and Ajitava Raychaudhuri

376

deficit and distance of state per capita income from highest per capita income state as the three important criteria in the devolution of transfer by the central government. It is also observed that population is used as the important criterion of formula based transfer. 3. Theoretical Model Literature on intergovernmental transfer mechanism in India also recommended that formula used in transferring funds should be simple and should not create any fiscal disincentive in a state. In the previous section we have examined first the different criteria used by several Finance and Planning commissions over the period. Instead of transfers by three different bodies separately we have analyzed the central government’s transfers as a whole. Evaluating the formulae used by different Planning and Finance commissions it is observed that population, revenue effort, deficit filling and the distance criteria are the important criteria used by transferring agencies in India. Transfer to a state in India can be divided into formula based transfer (Tritf ) and discretionary transfer (Tritd ) . Thus Trit = Tritf + Tritd

We assume that a part of the transfer is based on a simple formula whereas discretionary transfers depend on the discretion of the central government. Thus we assume that per capita transfer to a state depends on per capita revenue effort index, per capita actual deficit of the state and the distance criterion. Revenue effort index is measured by the difference between the per capita actual revenueb and per capita revenue capacity of the state, per capita actual deficit is calculated as the difference between per capita actual expenditure and per capita actual own revenue and the distance of per capita income from highest per capita income is calculated using the formula stated below (2). Thus transfer to a particular state, i, at any time period t is assumed to follow the following formula: ⎛ Tit − Tˆit ⎛ Tritf ⎞ ⎜ ⎟ = α ⎜⎜ ⎝ Pit ⎠ ⎝ Pit

b

⎞ ⎛ Git − Tit ⎟⎟ + β ⎜⎝ Pit ⎠

⎞ ⎛ Yht − Yit ⎞ ⎟+δ ⎜ P ⎟ t = 1, 2,3,… it ⎠ ⎝ ⎠

(1)

Actual own revenue is defined as the sum of own tax and non-tax revenue on revenue account and non-debt capital receipt on capital account.

Intergovernmental Transfer Rules, State Fiscal Policy and Performance in India

377

Yht − Yit = ( y ht − y it ) Pit

(2)

where,

As total transfer received by a state is assumed to be the sum of formula based transfer and discretionary transfer. Thus, Trit = Tritf + Tritd = α (Tit − Tˆit ) + β (G it − Tit ) + δ DI it + Tritd

(3)

= (α − β )Tit + β G it + Ait

where DI it = ( y ht − y it ) Pit , Ait = δ DI it + Tritd − α Tˆit Here i indicates ith state, Trit is the transfer to the ith state at time t, Pit is the population of the ith state at time t, Ti and Tit are actual own revenue collection and the estimated revenue capacity respectively of the ith state at time t, G it is

⎛ G − Tit ⎞ the total expenditure by the ith state at period t , ⎜ it ⎟ corresponds to per ⎝ Pit ⎠ ⎛ Y − Yit ⎞ capita actual budget deficit of the state i at time t, ⎜ ht ⎟ is the distance of ⎝ Pit ⎠ per capita net state domestic product (NSDP) from highest per capita NSDP in India at time period t, y it is the per capita net state domestic product of ith state at time period t, y ht is the highest per capita NSDP at time period t, Bi (t −1) is the repayment of borrowing of period (t-1) in period t, E it is the government expenditure on goods and services in period t. Tritd is the total discretionary

transfer that does not depend on formula. It is assumed that, α , β , δ > 0 . This implies that as states give more effort to raise its own revenue over and above their revenue capacity then transfer of funds by the central government will increase. We also assume that as deficit of a state increases a part of the deficit will be financed by the central government in India. As the distance of per capita NSDP of the state from that of state having highest per capita NSDP increases the transfer received by the state also increases. For the empirical part we have used the estimated revenue capacity as estimated by the Finance Commission. Here At is a composite term that depends on distance criterion, discretionary transfer and revenue capacity of the state. Now At moves together with the increase in government expenditure. Discretionary transfers in India are

Poulomi Roy and Ajitava Raychaudhuri

378

mainly for centrally sponsored and central plan schemes which comes in certain proportion of state’s expenditure. For example if state spends 75% of total expenditure then centre will give 25%. Thus At can be treated as some function of G. The budget constraintc faced by the state government in a federal country like India is as follows: Git = Tit + Bit + Trit ⇒ E it = Tit + Bit + Trit

(4)

⇒ (1 − θ t )Git = Tit + Bit + Trit

where Git = E it + Bi (t −1)

Here we assume that Bt −1 / G t = θ t . Substituting (3) in (4) we get, Bit = (1 − β − θ t )G it − (1 + α − β )Tit − Ait

(5)

where At = Trt d + δ DI t − α Tˆt Here we assume that the government borrows certain proportion of its B income in each period thus t = bt . We also assume that Yt grows at a constant Yt annual rate of γ . Thus bt = (1 − β − θ t ) g t − (1 + α − β )τ t + a t − γ bt

(6)

where Gt T Tr B Tˆ DI t Tr d ,τ t = t ,τ r t = t , bt = t ,τˆt = t , d t = , ct = t , Yt Yt Yt Yt Yt Yt Yt ∀t a t = δ d t + c t − ατˆt gt =

To determine the optimal own revenue and expenditure in a federal country like India where certain proportion of revenue comes from central transfer we make the following assumptions: 1. Same formula is used in transferring resources to all the states. 2. The decisions concerning both the state budget revenue formation and the procedures of financing the corresponding expenditures are made by the state governments. c

See Appendix A for detailed derivation

Intergovernmental Transfer Rules, State Fiscal Policy and Performance in India

379

3.

Government expenditure and revenue collection in real terms are certain proportion of its real output. It is assumed that G = gY and T = τ Y , where G: real government expenditure, T: real revenue collection (tax, non-tax and non-debt capital receipts), Y: real output, g is the government real expenditure as a proportion to the real output, τ is the real revenue collection as proportion to the real output. Here T is not expressed as a function as is usually used in case of direct tax. 4. Public transfers are not included into regional budget expenditures. 5. Borrowing taken in period (t − 1) is repaid in period t. Objective of the state governments is to maximize utility of the economic agents which depends on its own expenditure and revenue policy. Increase in real expenditure increases utility directly and increase in real revenue collection affects utility through its impact on reduction of real output available in the hands of the economic agents. Thus utility is function of public goods and private goods consumption. G represents the public goods consumption while private goods consumption is proxied by the real output less the output taken away by the government as tax (Y-T). Utility function is defined as follows: ∞

U=

∫ U ( G , (Y t

t

− Tt ) )e − ρ t dt

t =0 ∞

=

∫ Y U ( g , (1 − τ ) )e t

t

t

− ρt

d

t =0 ∞

=

∫ ( w ln g

t

+ (1 − w) ln(1 − τ t ) ) Y0 e −[ ρ −γ ]t d

(7)

t =0

Thus increasing g increases regional utility and increasing τ that takes away certain proportion of real output from the hands of the economic agents. Thus (1 − τ ) is the proportion of real output available in the hands of the economic agents. Higher the value of τ lower is the value of (1 − τ ) and thus lower is the utility derived by economic agents. Here ( w (1 − w) ) measures the provision of public to private goods and services, ρ is the rate of time preference, γ is the rate of growth of real output. We maximize (7) with respect to g ,τ subject to the government intertemporal budget constraint (6). Given the weight assigned to various criteria by the transferring agencies and as government sets a target of rate of growth of b, γ b , we can find out the optimum fiscal policy in different time period for different values of bt , a t and θ t in the following way.

Poulomi Roy and Ajitava Raychaudhuri

380

g t* =

w ( (γ b + γ )bt + a t + (1 + α − β ) )

τ t* = 1 −

(1 − β − wθ t − (1 − w) z ) (1 − w)(1 − β − z ) ( (γ b + γ )bt + a t + (1 + α − β ) ) (1 + α − β )(1 − β − wθ t − (1 − w) z )

(8) (9)

Thus, state authorities’ optimum choice depends upon the rules applied to regional transfer allocation, that is α , β , δ . It can be seen that higher the revenue capacity ratio the lower is the transfer received and thus lower is the optimum expenditure to output ratio. Poorer the state the higher will be the transfer received on the basis of distance criterion and this will affect the optimum expenditure to output ratio positively. For the positive rate of growth of borrowing, more the state can borrow the more it will be able to spend. This is essentially the rule of efficiency in competitive markets where at the margin, rate of substitution between public and private goods for the losing state and gaining states is equalized. Thus transferring one unit of money from one state to another by the way of transfer equalizes gain and loss at the margin. The analysis done here satisfies this criterion. Unfortunately, the finance commissions have mentioned efficiency but they never clearly defined it and incorporated such an efficiency rule in the transfer mechanism. Thus, the transfer rule could not ensure welfare maximization for each participating state. This makes it somewhat susceptible from sustainability point of view. Unless states fulfil the basic criterion of welfare maximization, it is difficult to say how far that will enjoy peoples’ confidence. Now it is important to check how weights assigned to different criteria affects the optimum revenue and expenditure policy of the state. Increase in weight given to α implies a state can receive more funds if it gives more effort in raising own revenue. Again higher the value of β it indicates transferring agencies are giving positive weight to actual expenditure net of own revenue collection of the state. More weight to δ indicates that instead of giving more weight to state’s own revenue and expenditure policy, transferring agency is giving more weight to whether the state is poor or not. Impact of weights assigned to different criteria by the transferring agency on optimum fiscal policy of the state are discussed below. As the transferring agency attaches more weight to revenue effort criterion ( α ) we find that the optimum revenue to output ratio and the optimum expenditure to output ratio will increase. More weight to revenue effort criterion will induce the state government to raise its revenue collection to get the same level of transfer. Given the revenue effort of the government more weight to α

Intergovernmental Transfer Rules, State Fiscal Policy and Performance in India

381

will increase the revenue side of the government budget and thus given the budget constraint optimum expenditure to output ratio will increase. This means that as α increases optimum policy of the government will be to raise revenue more. Now as state’s revenue collection increases this will have negative influence on regional utility. Now to increase utility the state government’s optimum policy will be to increase expenditure to output ratio. (1 − w)(1 − β − z ) [ (γ b + γ )bt ] ∂τ t* = ∂α (1 + α − β ) 2 (1 − β − wθ t − (1 − w) z ) 2 +

(1 − w)(1 − β − z ) [τˆt (1 + α − β )(1 − β − θ t − α − (1 − w) z ) + δ d t + c t ]

(1 + α − β ) 2 (1 − β − θ t − (1 − w) z ) 2 (1 − w)(1 − β − θ t − z )( β + θ t + (1 − w) z ) + (1 + α − β ) 2 (1 − β − θ t − (1 − w) z ) 2 >0 ∂g t* w (1 − τˆt ) > 0 , since τˆt < 1 . = ∂α (1 − β − wθ t − (1 − w) z ) 2

Higher weight assigned to deficit financing criterion increases both optimum revenue-output and expenditure-output ratio. Now we will discuss how weight assigned to distance criterion affects optimum fiscal policy of the state. ⎡( (γ b + γ )bt + a t + (1 + α − β ) ) (α + θ t + z ) ⎤⎦ ∂τ t* = (1 − w) ⎣ >0 2 ∂β (1 + α − β ) 2 ( (1 − β − wθ t − (1 − w) z )

( (γ b + γ )bt + a t + α + (1 − w) z ) ∂g t* =w >0 (1 − β − wθ t − (1 − w) z ) 2 ∂β Increase in weight assigned to the distance criterion indicates that more weight are given to factors other than state own expenditure and revenue policy. Poorer the state is, more funds will be transferred to the state that increases the value of the revenue side of the budget. Given its expenditure policy, optimum policy of the government will be to reduce revenue to output ratio. Given the revenue effort as transfer to the state increases it is optimum to spend more. ∂τ t* (1 − w)d t =− 0 (1 − β − wθ t − (1 − w) z ) ∂δ

382

Poulomi Roy and Ajitava Raychaudhuri

4. Methodology 4.1. Methodology Used by Finance Commission in Calculating Tax Capacity The way the Finance Commission has estimated the taxable capacity of a state is explained below. Taxable capacities of the states for each of the major taxes are calculated first then summing them up the ninth finance commission has estimated the aggregate taxable capacity of the state. Taxes are categorized into six major heads namely: (i) Sales tax (including central sales tax and purchase tax on sugarcane), (ii) state excise duties, (iii) stamp duties and registration fees, (iv) motor vehicles tax and passenger and goods tax, (v) entertainment tax, (vi) tax on agriculture and incomes and a residual category, other taxes. It is difficult to calculate the revenues from agricultural income taxes and other taxes using statistical method. The taxable capacities of this category of taxes are calculated on the basis of projected actual taxes. Taxable capacities of other five categories of taxes have been calculated using pooled time series and cross section data. It is assumed that there is no state specific variation. Thus it is assumed that the intercept and slope parameters are same across states. Time dummies are introduced in the model to capture the inter-temporal shifts. States are divided into three income groups, high income, middle income and lowincome group. For different categories of taxes different variables are considered and taxable capacities are estimated for each of the income groups. The variables that are used to estimate the tax capacities are as follows: State domestic product at factor cost, roads/railway length per 1000 square kilometer, per capita energy sale to ultimate consumer, total registered motor vehicles, proportion of heavy vehicles to total vehicles, consumption of country spirit (PL), seating capacities in cinema hall, proportion of urban population to total population while time dummies are introduced to capture inter-temporal shifts. The ninth Finance Commission has estimated taxable capacities for the period 1989–90 to 1994–95. 4.2. Method of Estimating Revenue Capacity of a State Using the estimated taxable capacities as estimated by Finance Commission for the period 1989–90 to 1994–95 we have estimated the revenue capacities of other years within the period 1980 to 2005 using the average growth of taxable capacity within the estimated period (1989–90 to 1994–95).

Intergovernmental Transfer Rules, State Fiscal Policy and Performance in India

383

We define own revenue capacity as the sum of own tax capacity, non-tax capacity and non-debt revenue receipt capacity. First, we have found the actual ratio of non-tax and tax revenue at constant prices for various years. Then three years moving average method is used. Average of these averages is used to calculate the total revenue (tax + non-tax + non-debt receipt) capacities of the state. 4.3. Method of Estimation of Parameters Ordinary least square method is used on pooled data taken from fourteen major states in India over the period 1980 to 2005. We have estimated the parameters of the transfer formula using the following regression model. ⎛ Tit − Tˆit ⎞ ⎛ Tritf ⎞ ⎛ Git − Tit ⎞ ⎛ Y − Yit ⎞ + δ ⎜ ht ⎟+β ⎜ ⎜ ⎟ = α ⎜⎜ ⎟ ⎟ + c0 ⎟ ⎝ Pit ⎠ ⎝ Pit ⎠ ⎝ Pit ⎠ ⎝ Pit ⎠ +c1 D1 + c 2 D 2 + c 3 D3 + c 4 D 4 + c 5 D5 + c 6 D6 + c 7 D7 +c 8 D8 + c 9 D9 + c11 D11 + c12 D13 + c14 D14 + u it

where D1: Andhra Pradesh, D2: Bihar, D3: Guajrat, D4: Haryana, D5: Karnataka, D6: Kerala, D7: Madhya Pradesh, D8: Maharashtra, D9: Orissa, D10: Punjab, D11: Rajasthan, D12: Tamil Nadu, D13: Uttar Pradesh, D14: West Bengal. Di=1 for i-th state, =0 otherwise, u i t is the random disturbance term. To estimate the coefficients of the above model and eliminate the problem of dummy variable trap we have excluded D11 that is, Punjab and applied ordinary least square method to the above equation. Tritf Pit is the per capita formula based transfer at time period t,

(

)

⎛ Tt − Tˆt ⎜⎜ ⎝ Pit

⎞ ⎟⎟ is defined as the per capita revenue effort by the state over and above ⎠ ⎛ G − Tit ⎞ their per capita revenue capacity at time period t, ⎜ it ⎟ is the difference ⎝ Pit ⎠

between the per capita total public expenditure and per capita own revenue ⎛ Y − Yit ⎞ th collection of the state. ⎜ ht ⎟ is the distance of per capita income of the i ⎝ Pit ⎠ state from highest per capita income of fourteen major states in India. To find out whether the error in prediction (e) is significantly different from zero or not, the paired difference t-test is used. The test statistic that we have

Poulomi Roy and Ajitava Raychaudhuri

384

used is t 0.025, n −1 =

e−0

where n is the number of pairs or number of s n −1 / n differences and s n −1 is the sample standard deviation of e for n-1 observations. Using the estimated coefficients from the above model optimum revenue to output ratio and the optimum expenditure to output ratio are calculated using equations (8) and (9). Actual rates are compared with the optimum rates in order to find out the possible direction of fiscal policy at the state level. Mean of the absolute difference between actual and predicted values relative to actual values is used as a measure of mean relative error in prediction. Thus mean relative error in prediction is calculated using the ⎞ 1⎛ n following formula: ⎜ ∑ x i* − x i x i ⎟ , where x i is the actual value of the n ⎝ i =1 ⎠ variable, x i* is the optimum value of x i , n is number of observations. We assume that a linear relationship exists between A and G and thus At = A0 + zG t ⇒ a t = ( A0 Yt ) + zg t

(10)

We regress the above equation to estimate z for fourteen major states separately. As defined θ t = Bt −1 Gt . We estimated θ t using this relationship only. Target rate of growth of borrowing has been estimated in the following way: First we have estimated ( ( Bt − Bt −1 ) / Bt ) for each year over the period 1980 to 2005. The average of these growth rates has been used as the target rate of growth of borrowing. The rate of growth of SDP ( γ ) has been estimated by regressing ln Yt = ln Y0 + γ t equation for each of the fourteen major states in India. 5. Data Source and Variables 5.1. Data Source This section summarizes the data used in this study. In this chapter we consider the period 1980 to 2005 excluding 2003d, using data on fourteen major states of India. State finance data such as data on grants-in-aid, share in central taxes,

d

In The State Finances issue of RBI bulletin in January 2006 the capital receipts and capital expenditure data of 2003 do not represent public account on net basis but in all other years’ data it represents expenditure and revenue on capital account taking public account on net basis, so to maintain the similarity we have excluded the year 2003.

Intergovernmental Transfer Rules, State Fiscal Policy and Performance in India

385

loans and advances by the central government, borrowing, expenditure on revenue and capital account, own revenue collection of the states are collected from various issues of “State Finances: A study of budgets” published by Reserve Bank of India, India. Revenue capacity of the states is obtained by using the estimated tax capacity data by the finance commission for the period 1989–90 to 1994–95. 5.2. Variables The variablese that have been used in the empirical estimation of the model are defined below: Y: Gross State Domestic Product (GSDP) at constant 1993–94 pricesf Tr: Transfer of the central Government = Share in central taxes, grants and loans; Trd : Discretionary transferg, Bt : Borrowing from all sources other than central government, Tr f : Formula based transfer = Tr − Trd , Bt −1 : Repayment of loans taken in period (t-1) from sources other than loans from the central government; G: State own expenditure net of transfers = Revenue Expenditure + Capital expenditure – Grants and loans other than central ministries grantsh; T: Actual Revenue = Own Tax + Own Non-Tax Revenue + non-debt capital receipt; Tˆ = Own Revenue capacityi, ( (Yht − Yit ) / Pit ) : distance of per capita income of the state from highest per capita income multiplied by population relative to the over all distance of income over the fifteen major states in India, θ t = Bt −1 / G t .

e

All variables are expressed in constant 1993-94 prices. This corresponds to real output as mentioned in the theoretical part. g It is defined as the grants and loans for centrally sponsored and central sector schemes plus all forms assistance by the centre to the states in the form of ‘relief for natural calamities’ and ‘others’ in the Grants part and ‘share of small savings’, ‘relief for natural calamities’, ‘ways and means advance’, ’loans for special schemes’ and ‘others’ in the loan part. h As central ministry’s grants and loans go directly to the block offices they do not appear in state budget whereas other transfers are included in the state expenditure. To find out state’s own expenditure net of transfer we have used this formula. i Methodology used by the finance commission in estimating tax capacity of a state is explained above. f

Poulomi Roy and Ajitava Raychaudhuri

386

6. Empirical Analysis 6.1. Transfer 6.1.1. Formula Based Transfer The regression results are summarized as follows: Appendix C, Table C.1 indicates that over the selected states during the period 1980 to 2005 the estimated weights given to per capita revenue effort index, per capita actual deficit and distance criteria are αˆ = 0.069 , βˆ = 0.128 and δˆ = 0.035 respectively. All the estimated coefficients are found to be positive and significantly different from zero. Our empirical results support our assumptions that α , β and δ are positive and less than one. The value of the condition index is less than 15. This indicates that the multicollinearity problem is not severe. Goodness of fit measured by adjusted R2 (=0.4279) is also good for a pooled data series. Having estimated the parameters of the transfer formula, we have estimated the predicted per capita transfer for each year of the sample period 1980 to 2005 and then estimated the mean transfer over the period using the formula given below. Tritf Pit = 423.55 *** -(227.84) ***d1-(335.59)d -(223.4) ***d3 -(256.56) ***d4 − (268.06) *** d 5 − (252.16) *** d 6 − (302.64) *** d 7 −(260.93) *** d 8 − (228.66) *** d 9 − (272.12) *** d11 − (209.29) *** d12

(

− (317.13) *** d13 − (288.55) *** d14 + (0.069) ** (Tit − Tˆit ) Pit +(0.128)

***

( (Git − Tit )

Pit ) + (.035)

***

( (Yht − Yit )

)

Pit )

where ‘**’ and ‘***’ implies significant at 5 and 10 per cent levels. Table C.2 shows mean per capita transfer received by a state along with their 95 percent levels of confidence. It is predicted that over the period 1980 to 2005 out of fourteen major states in India on an average in per capita terms Maharashtra received lowest transfer whereas Orissa received highest transfer per capita on the basis of the formula used in this model. Last two columns of Table C.2 indicate the 95 percent confidence interval of per capita transfer in fourteen major states. We have tested whether the mean error in prediction is equal to zero or not. The error in prediction (e) is the difference between the actual and the predicted values. The paired difference t-test is used for this purpose. The test statistic

Intergovernmental Transfer Rules, State Fiscal Policy and Performance in India

387

e−0

used is t 0.025, n −1 =

where n is the number of pairs or number of s n −1 / n differences and s n −1 is the sample standard deviation of e. The results obtained in Table 2 suggest that we fail to reject the null hypothesis at 95 percent level of confidence. Table 2. Prediction Error Test of Per Capita Transfer Hypothesis

Mean 0.0000009

mean(e) = 0

df = 349

95%

Confidence

−11.122

Interval 11.122

t = −0.0000(1.00)

The difference between actual and the predicted value are not significantly different from zero at 95 percent level of confidence. Thus the transfer formula used in this chapter predicts quite well the actual per capita transfer to a state in India. The relative error in prediction using the above model is calculated by taking the absolute difference between actual per capita transfer and the predicted transfer relative to the actual per capita transfer over the period 1980 to 2005 in Table 3 below. The average relative error in prediction varies from Table 3. State wise Relative Error in Prediction of Per Capita Transfer States

Mean Relative Error in Prediction of Per Capita Transfer

Andhra Pradesh

0.1233

Bihar

0.1713

Gujarat

0.2178

Haryana

0.3509

Karnataka

0.2273

Kerala

0.1403

Madhya Pradersh

0.0898

Maharashtra

0.1780

Orissa

0.1149

Punjab

0.5208

Rajasthan

0.1020

Tamil Nadu

0.1647

Uttar Pradesh

0.0755

West Bengal

0.2012

Poulomi Roy and Ajitava Raychaudhuri

388

8.98 percent in Madhya Pradesh to 52.08 percent in Punjab over the period 1980 to 2005. 6.1.2. Discretionary Transfer We find a positive relationship between per capita discretionary transfer and per capita expenditure of the states (Table 4). The regression of (10) indicates that as g increases ‘a’ which is a composite function of revenue capacity, distance of per capita NSDP and discretionary transfer, increases. (see Appendix Table C.4) Table 4. Statewise Per Capita Discretionary Transfer (1980–2005) P.C Discretionary Transfer

P.C Expenditure

Andhra Pradesh

135.31 (10)

1449.44 (8)

Bihar

123.14 (13)

810.09 (14)

Gujarat

220.35 (2)

1896.10 (3)

Haryana

208.32 (3)

2104.24 (2)

Karnataka

155.36 (8)

1624.31 (6)

Kerala

146.16 (9)

1515.64 (7)

Madhya Pradersh

131.84 (11)

1151.22 (10)

Maharashtra

174.71 (6)

1874.08 (4)

Orissa

181.16 (5)

1055.33 (11)

Punjab

297.76 (1)

2216.92 (1)

Rajasthan

202.68 (4)

1328.22 (9) 1697.45 (5)

Tamil Nadu

128.60 (12)

Uttar Pradesh

108.44 (14)

922.76 (13)

West Bengal

159.70 (7)

1002.51 (12)

6.2. Comparison of Optimum and Actual Revenue and Expenditure Policy

It is theoretically found that the relationship between the optimum revenueGSDP ratio and the public expenditure-GSDP ratio is as follows j: ⎛ w ⎞ (1 + α − β ) gt = ⎜ (1 − τ t ) . ⎟ ⎝ 1 − w ⎠ (1 − β − z ) j

See Appendix B, equation (B.7)

Intergovernmental Transfer Rules, State Fiscal Policy and Performance in India

389

We regress the following equationk g t = ϖ (1 − τ t )

(11)

w ⎛ 1 + (αˆ − βˆ ) ⎞ ⎜ ⎟ 1 − w ⎜⎝ (1 − βˆ − z ) ⎟⎠

(12)

where

ϖˆ =

for fourteen major states in India separately over the period 1980 to 2005. We observe that fit is very good as the value of adjusted R 2 is above 0.95 and the estimated coefficient is significantly different from zero for all the fourteen major states. Thus the actual data satisfies the relationship from which the optimum revenue-output and expenditure-output ratios are derived. The optimum values estimated using the equations of optimum revenue-output and expenditure-output (Eqs. (8) and (9)) can be well considered as the optimum values. ^ ⎛ ⎞ Using the above relation (12) we have estimated the ⎜ w /(1 − w) ⎟ and thus ⎝ ⎠ wˆ . The optimum revenue-output and expenditure-output ratios are estimated using the estimated coefficients of the transfer formula, estimated value of w for different values of, τˆ, θ and b. The average optimum T/Y ratio and G/Y ratios for the period 1980 to 2005 along with their 95 percent confidence interval are listed below in Table 5 for fourteen selected states in India. Comparing actual and optimum allocation rates it is observed that states are spending more than optimum while raising revenue less than optimum revenue. These optimum rates are dependent on the weights assigned to various criteria by the transferring agencies. The difference in optimum revenue to GSDP ratio on the other hand cannot be explained by only one term. This is a result of all these factors. From (8) and (9) it is clear that such a difference in optimum values lie in difference in ω , a , θ and b . The difference in optimum expenditure to GSDP ratio can be more attributed to the difference in the estimated value of ω . This is the measure of weight assigned to public good consumption. We find a high rank correlation between these two factors.

k

Regression results are presented in Appendix C, Table C.3

Poulomi Roy and Ajitava Raychaudhuri

390

Table 5. Statewise Revenue/GSDP and Expenditure/GSDP Ratios (80-05)

Andhra Pradesh

Bihar$

Gujarat

Haryana

Karnataka

Kerala Madhya Pradesh$$ Maharashtra

Orissa

Punjab

Rajasthan

Tamil Nadu Uttar Pradesh$$$ West Bengal

Actual Mean

Estimated Mean

95% confidence interval

(T/GSDP)

0.1008

0.1140

0.1110

0.1170

(G/GSDP)

0.1656

0.1630

0.1625

0.1636

(T/GSDP)

0.0658

0.0669

0.0614

0.0725

(G/GSDP)

0.1768

0.1763

0.1753

0.1774

(T/GSDP)

0.1041

0.1153

0.1111

0.1195

(G/GSDP)

0.1520

0.1499

0.1492

0.1506

(T/GSDP)

0.1215

0.1364

0.1324

0.1403

(G/GSDP)

0.1609

0.1575

0.1568

0.1582

(T/GSDP)

0.1154

0.1230

0.1180

0.1280

(G/GSDP)

0.1728

0.1711

0.1702

0.1721

(T/GSDP)

0.0975

0.1133

0.1086

0.1179

(G/GSDP)

0.1625

0.1611

0.1603

0.1620

(T/GSDP)

0.0887

0.0986

0.0965

0.1008

(G/GSDP)

0.1575

0.1554

0.1550

0.1558

(T/GSDP)

0.1012

0.1204

0.1158

0.1250

(G/GSDP)

0.1433

0.1402

0.1394

0.1409

(T/GSDP)

0.0724

0.0846

0.0775

0.0916

(G/GSDP)

0.1795

0.1769

0.1755

0.1782

(T/GSDP)

0.1085

0.1302

0.1224

0.1379

(G/GSDP)

0.1534

0.1490

0.1476

0.1503

(T/GSDP)

0.0901

0.0999

0.0964

0.1034

(G/GSDP)

0.1716

0.1696

0.1689

0.1703

(T/GSDP)

0.1094

0.1227

0.1176

0.1278

(G/GSDP)

0.1643

0.1617

0.1607

0.1626

(T/GSDP)

0.0675

0.0783

0.0751

0.0816

(G/GSDP)

0.1525

0.1506

0.1500

0.1511

(T/GSDP)

0.0596

0.0624

0.0577

0.0671

(G/GSDP)

0.1180

0.1177

0.1171

0.1183

Note: $: including Jharkhand, $$: including Chattisgarh, $$$: including Uttarakhand

Intergovernmental Transfer Rules, State Fiscal Policy and Performance in India

391

Mean absolute deviation of actual rates from their optimum rates relative to the actual rates are calculated to find out how far the state’s actual fiscal policy is from their optimum policy. We observe in Table 6 below that such a variation in expenditure to GSDP ratio ranges from 5.02 percent in Karnataka to 20.18 percent in Bihar. Mean absolute deviation of actual revenue to GSDP ratio from optimum revenue-GSDP ratio relative to actual revenue-GSDP ratio varies from 13.12 percent in Karnataka to 26.06 percent in Punjab. Table 6. Relative Variation of Revenue/GSDP and Expenditure/GSDP Ratios (80-05) Mean relative variation Andhra Pradesh Bihar Gujarat Haryana Karnataka Kerala Madhya Pradesh Maharashtra Orissa Punjab Rajasthan Tamil Nadu Uttar Pradesh West Bengal

Min

Max 0.3406

(T/GSDP)

0.1748

0.0277

(G/GSDP)

0.0907

0.0008

0.2420

(T/GSDP)

0.1690

0.0026

0.6753

(G/GSDP)

0.2018

0.0001

0.5876

(T/GSDP)

0.1599

0.0071

0.3968

(G/GSDP)

0.1073

0.0056

0.2987

(T/GSDP)

0.2063

0.0051

0.4656

(G/GSDP)

0.0996

0.0013

0.3566

(T/GSDP)

0.1312

0.0020

0.3727

(G/GSDP)

0.0502

0.0026

0.2037 0.4208

(T/GSDP)

0.1871

0.0276

(G/GSDP)

0.0627

0.0054

0.1728

(T/GSDP)

0.1693

0.0031

0.3377

(G/GSDP)

0.0981

0.0017

0.3618

(T/GSDP)

0.2372

0.0252

0.5436

(G/GSDP)

0.0924

0.0028

0.2547

(T/GSDP)

0.2940

0.0136

0.7399

(G/GSDP)

0.1298

0.0020

0.6113

(T/GSDP)

0.2606

0.0091

0.8692

(G/GSDP)

0.1964

0.0182

0.3706 0.4180

(T/GSDP)

0.1617

0.0062

(G/GSDP)

0.1116

0.0034

0.2635

(T/GSDP)

0.1795

0.0144

0.4048

(G/GSDP)

0.0704

0.0007

0.2124

(T/GSDP)

0.1866

0.0042

0.5904

(G/GSDP)

0.1233

0.0004

0.3359

(T/GSDP)

0.2385

0.0008

0.7883

(G/GSDP)

0.1450

0.0022

0.2884

Poulomi Roy and Ajitava Raychaudhuri

392

Here we find out whether the deviation of actual rates from their optimum rates is significantly different from zero or not. We define the difference between actual and the optimum revenue-GSDP ratio by e (T/Y) and the difference between actual and the optimum public expenditure to GSDP ratio is denoted by e(G/Y). We have tested whether the mean e(T/Y) and the mean e(G/Y) are significantly different from zero or not (Table 7). We reject the null hypotheses mean e(T/Y) = 0 and but fail to reject e(G/Y) = 0 at 95 percent level of confidence. Table 7. Prediction error test of revenue-GSDP ratio & expenditure-GSDP ratio Hypothesis

Mean

Std. Err.

Degrees of freedom = 349

e (T/Y) = 0

−0.0117

0.0010

t = −12.317 (.000)

e(G/Y) = 0

0.0023

0.0014

t = 1.673 (.095)

The fiscal performance at the state level in India reveals that significant change in revenue side of the government budget is required. Above analysis suggests that the state governments can take the optimum values calculated here as the benchmark and can change their actual policies accordingly. From this study we find that there should be increase in revenue to GSDP ratio in all the fourteen major states in India. 7. Conclusions

A model of determination of optimum revenue and expenditure in a federal economy like India has been developed. The model shows how the intergovernmental transfer allocation rule affects the utility maximizing level of revenue to output and expenditure to output ratios of the sub-national governments. The model is developed considering the transfer principle used by different transferring agencies in India. The optimum revenue and expenditure policy of a state government are found to be dependent on the weight assigned to different criteria by the federal government in transferring funds to the state governments. Changing the weights assigned to different criteria federal government can change the utility maximizing revenue to output and expenditure to output rates. Using pooled regression analysis on data taken from fourteen major states in India and for the period 1980 to 2005, we have estimated the weights assigned to various criteria by the transferring agencies. These coefficients are thus statistically estimated not arbitrarily chosen. All the coefficients are found

Intergovernmental Transfer Rules, State Fiscal Policy and Performance in India

393

to be significantly different from zero. All the dummy variables are also found to be significant at 5 percent level of significance. As we have assumed in the theoretical part all the coefficients are found to be positive in sign. We find that on an average Maharashtra received the lowest per capita formula based transfer. On the other hand Orissa received the largest funds per capita from the centre over the period 1980 to 2005 out of fourteen major states in India. We fail to reject the null hypothesis of zero mean error in prediction of per capita transfer on the basis the formula used in this chapter at 95 percent level of confidence. Thus the formula considered in this chapter predicts quite well the actual per capita transfer to a state in India. Optimum revenue and expenditure rates are obtained substituting the estimated coefficients of the transfer formula. The actual revenue to GSDP ratio is found to be lower than the optimum revenue to GSDP ratio in most the selected states. On the other hand, actual expenditure to GSDP ratios in fourteen major states is higher than their optimum values. Given the transfer formula and the estimated parameters of the model, the optimum revenue and expenditure to GSDP ratios calculated here can be considered as the benchmark by the state governments. The deviation of actual values from the optimum values also give us some idea regarding to which direction the state governments should change its existing revenue and expenditure policies. In this chapter we did not consider on what factors discretionary transfer depends but that could be the scope of future research. Optimum fiscal policies are determined with the given weights assigned by the federal government in transferring resources. A model where central government endogenously determines the transfer can be developed as a future research agenda. Appendix A. Derivation of Budget Constraint Gt = Tt + Trt + Bt ⇒ Gt = Tt + Bt + Trt + Bt −1

where Trt = α (Tt − Tˆt ) + β (Gt − Tt ) + δ DI t + Trt d

in the above equation we get, Bt = (1 − θ t )G t − Tt − α (Tt − Tˆt ) − β (Gt − Tt ) − δ DI t − Trtd

or

Poulomi Roy and Ajitava Raychaudhuri

394

Bt = Gt (1 − β − θ t ) − Tt (1 + α − β ) − At

where At = Trtd + δ DI t − α Tˆt or, •

Bt = Gt (1 − β ' − θ t ) − Tt (1 + α ' − β ' ) − At

or, •

Bt = g t (1 − β ' − θ t ) − τ t (1 + α ' − β ' ) − a t Yt

δ ' DI t + Trtd − α 'Tˆt

, α , β , δ are weights assigned by the Yt central government to the various criteria and are determined statistically. These are parameters to the state government. Thus, where a t = At / Yt =

•

b t = g t (1 − β ' − θ t ) − τ t (1 + α ' − β ' ) − a t − γ bt

(A.1)

Appendix B. Derivation of Optimum Revenue and Expenditure Ratio

The problem is to ∞

Max U =

∫ U ( Gt , (Yt − Tt ) ).e − ρt dt =

t =0

∞

∫t =0 YtU ( g t , (1 − τ t ) ).e − ρt dt

∞

=

∫t =0 ( w ln g t + (1 − w) ln(1 − τ t ) ) Y0 e −[ ρ −γ ]t dt

subject to bt = (1 − β ' − θ t ) g t − τ t (1 + α ' − β ' ) − a t − γ bt

where α , β and δ , ( Bt / Yt ) = bt , (Yt Yt ) = γ , a t = At / Yt =

δ ' DI t + Trtd − α 'Tˆt Yt

,

(Gt / Yt ) = g t , (Tt / Yt ) = τ t , ( w (1 − w) ) measures the provision of public to

private goods and services, ρ is the rate of time preference.

Intergovernmental Transfer Rules, State Fiscal Policy and Performance in India

395

The Hamiltonian equation for this problem is as follows H = w ln g t + (1 − w) ln(1 − τ t ) + μ t ( (1 − β − θ t ) g t − τ t (1 + α − β ) − a t − γ bt )

⎛ ⎛ ∂a ⎞ ⎞ ∂H w = + μ t ⎜ (1 − β − θ t ) + θ t − ⎜ t ⎟ ⎟ = 0 ∂g t g t ⎝ ∂g t ⎠ ⎠ ⎝ ⇒

⎛ ∂a w + μ t (1 − β − z ) = 0 where ⎜ t gt ⎝ ∂g t

⇒ μt = − •

⎛ bt −1 ⎞ ⎜∵ θ t = ⎟ (1 γ )gt ⎠ − ⎝

⎞ ⎟ = z (say, constant) ⎠

w g t (1 − β − z )

(B.1)

•

μ g ⇒− = μ g

(B.2)

∂H (1 − w) (1 − w) =− − μ t (1 + α − β ) = 0 ⇒ μ t = − ∂τ t (1 − τ t ) (1 − τ t )(1 + α − β )

(B.3)

•

•

μ (1 − τ t ) ⇒− t = μ t (1 − τ t )

(B.4)

• ∂H = −γμ t = ( ρ − γ ) μ t − μ t ∂bt

(B.5)

•

•

•

μ g (1 − τ t ) ⇒− t = t = = −ρ μ t g t (1 − τ t )

(B.6)

From (B.1) and (B.3) we get ⎛ 1 − w ⎞ (1 − β − z ) ⇒ (1 − τ t ) = ⎜ gt ⎟ ⎝ w ⎠ (1 + α − β )

(B.7)

Suppose the government sets a target on growth rate of b, say γ b . Then the government’s intertemporal budget constraint becomes γ b bt = (1 − β − θ t ) g t − τ t (1 + α − β ) − a t − γ bt or (γ b + γ )bt = (1 − β − θ t ) g t + (1 − τ t )(1 + α − β ) − a t − (1 + α − β )

Poulomi Roy and Ajitava Raychaudhuri

396

By substituting (A2.7) in the above equation we now solve it for τ t and g t in the following way: (1 − β − z ) ⎛1− w ⎞ gt (γ b + γ )bt = (1 − β − θ t ) g t + ⎜ ⎟ (1 + α − β ) (1 + α − β ) ⎝ w ⎠ −a t − (1 + α − β )

or, (γ b + γ )bt = or, g t* =

(1 − β − wθ t − (1 − w) z ) g t − a t − (1 + α − β ) w

w ( (γ b + γ )bt + a t + (1 + α − β ) )

(B.8)

(1 − β − wθ t − (1 − w) z )

Now substituting (B.8) in (B.7) we find

τ t* = 1 −

(1 − w)(1 − β − z ) ( (γ b + γ )bt + a t + (1 + α − β ) ) (1 + α − β )(1 − β − wθ t − (1 − w) z )

(B.9)

The optimum revenue and expenditure policy of the state-government in any time period t depends on the weights assigned to various criteria by the central government, target rate of growth of borrowing as set by the stategovernment, rate of growth of output and borrowing-output ratio of each period. Steady state condition: We can rewrite equation (B.8) and (B.9) in the following way g t = c1bt + c 2 a t + c 0

where c1 =

w(γ b + γ ) w w(1 + α − β ) , c2 = , c0 = (1 − β − wθ t − (1 − w) z ) (1 − β − wθ t − (1 − w) z ) (1 − β − wθ t − (1 − w) z ) •

•

•

•

g b b a a θt wθ t ⇒ t = c1 t t + c 2 t t + c 3 where c 3 = , gt bt g t at g t θt (1 − β − wθ t − (1 − w) z ) •

b a θ ⇒ γ g = − ρ = c1γ b t + c 2γ a t + c 3θˆ where θˆ = t θt gt gt

Intergovernmental Transfer Rules, State Fiscal Policy and Performance in India

397

The steady state condition requires that the above condition is satisfied. At steady state γ g , γ b , γ a and θˆ are constant but they are different in magnitude. At steady state g and (1-τ) grows at the same rate. •

•

•

g t (1 − τ t ) G (Y − Tt ) = = − ρ and t = t =γ −ρ >0 (1 − τ t ) g t (1 − τ t ) Gt

Appendix C. Empirical Results Table C.1: Regression of Transfer Formula Coeff.

t

Sig.

423.55

16.31

0.000***

⎡(T − Tˆ ) P ⎤ ⎣ ⎦

0.069

2.38

[(G − T ) P ]

0.128

[(Yh − Yi ) P ]

Const.

Collinearity Statistics VIF

TOL

Eigen value

CI

0.018**

1.50

0.667

3.819

1.00

5.50

0.000***

1.40

0.715

1.394

1.65

0.035

6.74

0.000***

6.06

0.164

1.037

1.91

Andhra Pradesh

-227.84

-5.74

0.000***

3.12

0.320

1.002

1.95

Bihar

-335.79

-6.17

0.000***

5.86

0.170

1.000

1.95

Gujarat

-223.4

-6.81

0.000***

2.13

0.469

1.000

1.95

Haryana

-256.56

-8.06

0.000***

2.00

0.499

1.000

1.95

Karnataka

-268.06

-7.05

0.000***

2.86

0.349

1.000

1.95

Kerala

-252.16

-6.61

0.000***

2.88

0.346

1.000

1.95

Madhya Pradesh

-302.64

-6.72

0.000***

4.01

0.249

1.000

1.95

Maharashtra

-260.93

-8.30

0.000***

1.96

0.511

1.000

1.95

Orissa

-228.66

-4.63

0.000***

4.82

0.207

1.000

1.95

Rajasthan

-272.12

-6.37

0.000***

3.61

0.277

0.532

2.67

Tamil Nadu

-209.29

-5.89

0.000***

2.50

0.400

0.150

5.05

Uttar Pradesh

-317.13

-6.45

0.000***

4.79

0.208

0.049

8.85

West Bengal

-288.55

-7.05

0.000***

3.32

0.301

0.018

14.7

Durbin Watson=0.894

F=17.72 (.000) Adjusted R-square=0.434

Note: Dependent Variable: (TR/P), const.: represents Punjab, CI: condition index, ‘*’, ‘**’, ‘***’ indicate significant at 10%, 5% and 1% level, respectively

Poulomi Roy and Ajitava Raychaudhuri

398

Table C.2: Per Capita Formula Based Transfer (1980–2005) Mean

Std. Error

95 % Confidence Interval

Andhra Pradesh

442.99 (4)

12.59

416.99

468.98

Bihar$

442.50 (5)

20.37

400.45

484.54

Gujarat

368.07 (12)

11.44

344.47

391.68 310.14

Haryana

289.71 (13)

9.90

269.28

Karnataka

384.95 (10)

7.32

369.83

400.06

Kerala

416.33 (7)

10.37

394.92

437.73

Madhya Pradesh$$

412.75 (9)

17.24

377.18

448.33

Maharashtra

288.52 (14)

7.98

272.04

305.00

Orissa

544.70 (1)

17.90

507.76

581.63

Punjab

517.86 (2)

14.69

487.55

548.18

Rajasthan

446.10 (3)

17.54

409.89

482.31

Tamil Nadu

415.83 (8)

5.89

403.68

427.99

Uttar Pradesh$$$

432.84 (6)

20.38

390.77

474.91

West Bengal

380.53 (11)

10.25

359.38

401.68

Note: $: including Jharkhand, $$: including Chattisgarh, $$$: including Uttarakhand, ‘*’, ‘**’, ‘***’ indicate significant at 10%, 5% and 1% level, respectively Table C.3: Relationship between the Public Expenditure Rate and Revenue Rate State

Coefficient

t(P> t )

Adjusted R-square

Andhra Pradesh

0.1840

39.63(.000)***

0.9843

Bihar$

0.1890

17.56 (.000)***

0.9248

Gujarat

0.1694

33.16(.000)***

0.9777

Haryana

0.1824

28.41 (.000)***

0.9699

Karnataka

0.1952

59.80 (.000)***

0.9931

Kerala

0.1817

65.22 (.000)***

0.994

Madhya Pradesh$$

0.1724

26.31(.000)***

0.9651

Maharashtra

0.1593

40.68 (.000)***

0.9851

Orissa

0.1932

29.50(.000)***

0.9720

Punjab

0.1713

19.63(.000)***

0.9389

Rajasthan

0.1884

33.41(.000)***

0.9781 0.9873

Tamil Nadu

0.1843

44.10 (.000)***

Uttar Pradesh$$$

0.1634

26.95(.000)***

0.9667

West Bengal

0.1256

28.14(.000)***

0.9694

Note: Regression equation: g t = ϖ (1 − τ t ) + u , where ϖ is the estimated coefficient, $: including Jharkhand, $$: including Chattisgarh, $$$: including Uttarakhand. ‘*’, ‘**’, ‘***’ indicate significant at 10%, 5% and 1% level, respectively,

Intergovernmental Transfer Rules, State Fiscal Policy and Performance in India

399

Table C.4: Relationship between Discretionary Transfer Rate and Public Expenditure Rate State

z

a0

F(2,23)

Adjusted R-square

0.294***

−0.0000003***

243.97(.000)***

0.951

Bihar$

0.101

0.0000016***

279.72(.000)***

0.957

Gujarat

0.270***

−0.0000003***

101.69(.000)***

0.890

Haryana

0.176***

−0.0000007***

48.08(.000)***

0.790

Karnataka

0.304***

−0.0000005***

310.61(.000)***

0.961

Kerala

0.339***

−0.0000009***

128.87(.000)***

0.906

Madhya Pradesh$$

0.340***

−0.0000002

324.38(.000)***

0.963

Maharashtra

0.203***

−0.0000002***

50.63(.000)***

0.799

Orissa

0.655***

−0.0000002***

213.11(.000)***

0.944

Punjab

0.037

0.0000002

5.60(.011)**

0.269

Andhra Pradesh

Rajasthan

0.422***

−0.0000006***

297.31(.000)***

0.960

Tamil Nadu

0.251***

−0.0000004***

228.58(.000)***

0.948

Uttar Pradesh$$$

0.519***

−0.0000002

219.39(.000)***

0.946

West Bengal

0.624***

-0.0000006***

67.71(.000)***

0.842

Note: Regression equation: a t = zg t + a 0 y + u , where z is the marginal impact of changes in g on a, $: including Jharkhand, $$: including Chattisgarh, $$$: including Uttarakhand ‘*’, ‘**’, ‘***’ indicate significant at 10%, 5% and 1% level, respectively,

References

1. R.W. Bhal, IMF Staff Papers, 18 (3), 570 (1971) 2. N. Bajpai and J. D. Sachs, The State of State Government Finances in India, Development Discussion Paper No. 719, http://www.ksg.harvard.edu/CID/india/pdfs/719.pdf (1999) 3. D. Coondoo, A. Majumder, R. Mukherjee and C. Neogi, Economic and political weekly, 36(40), 3869(2001) 4. D. Coondoo, A. Majumder and C. Neogi, Taxable Capacity Function: A Note on Specification, Estimation and Application, in Fiscal Federalism in India: Contemporary Issues, ed. D. K. Srivastava (Har-Anand Publications, New Delhi, 2000), 5. B. Dahlby, The Marginal Cost of Funds from Public Sector Borrowing, Department of Economics, University of Alberta, revised September 2004, http://www.uofaweb.ualberta.ca/ipe/pdfs/DahlbyTheMCFfromPublicSector BorrowingDec04.pdf(2004) 6. GR, Economic and Political Weekly, 36(41), 4231, (2001).

400

Poulomi Roy and Ajitava Raychaudhuri

7. Government of Gujarat, Memorandum to the Twelfth Finance Commission, Finance Commission, India, http://fincomindia.nic.in/pubsugg/memo_guj.pdf (2003), 8. J. Ma, Intergovernmental Fiscal Transfer: A Comparison of Nine Countries (Cases of the United States, Canada, the United Kingdom, Australia, Germany, Japan, Korea, India, and Indonesia), prepared for Macroeconomic Management and Policy Division, Economic Development Institute, The World Bank (1997) 9. M. A. Oommen, Economic and Political Weekly, 22(11), 466 (1987) 10. I. Rajaraman and G. Vasishtha., Economic and Political Weekly, 35(33), 2943 (2000) 11. M.G. Rao, Economic and Political Weekly, 37(31), 3261(2002) 12. M.G. Rao, Economic and Political Weekly, 39 (18), 1820 (2004) 13. M.G. Rao and N. Singh., Intergovernmental Transfers: Rationale, Design and Indian Experience, http://econ.ucsc.edu/~boxjenk/cre3.pdf, (1998a) 14. M.G. Rao, An analysis of explicit and implicit intergovernmental transfers in India, http://econ.ucsc.edu/~boxjenk/cre4.pdf (1998b) 15. M.G. Rao, The assignment of taxes and expenditures in India, http://econ.ucsc.edu/~boxjenk/cre1.pdf, (1998c) 16. M.G. Rao, The Political Economy of Center-State Fiscal Transfers in India, http://credpr.stanford.edu/pdf/credpr107.pdf, (2000) 17. T. K. Sen and C. Trebesch., The Use of Socio-Economic Criteria for Intergovernmnetal Transfers: The Case in India, NIPFP Working Paper No. 10, (2004) 18. T. K. Sen, Relative Tax Effort by Indian States, NIPFP Working Papers No. 5, (1997) 19. S. Sinelnikov, P. Kadotchnikov., I. Trounin., and E. Schkrebela, Impact of intergovernmental grants on the fiscal behavior of regional authorities in Russia, mimeo, IET, (2001)

April 20, 2009

17:25

WSPC - Proceedings Trim Size: 9in x 6in

prasad

TAX EVASION AND ADMINISTRATIVE COSTS ROHIT PRASAD Management Development Institute PO No. 60, Mehrauli Rd, Sukhrali, Gurgaon 122001, India E-mail: [email protected] This paper seeks to find if there is any merit in the position that tax evasion is justified in the presence of a leakage of taxpayers’ money. Tax is modeled as a simple redistribution from entrepreneurs to workers. The conclusions are that in an indirect tax regime, an increase in the administrative cost coefficient warrants an increase in the tax evasion coefficient in order to maintain the welfare obtained at the lower administrative cost. However in a direct tax regime, an increase in the administrative cost coefficient must be accompanied by a decrease in the tax evasion coefficient in order to maintain the level of profit or the level of worker utility obtained at the lower administrative cost. With direct taxes, the levels of both profits as well utility cannot be simultaneously preserved with a higher administrative cost coefficient.

1. Introduction A common justification of tax avoidance and tax evasion is that the government puts hard earned taxpayers money to no good use.a This paper seeks to find if there is any merit in this position. The existing literature on tax evasion including papers of Allingham and Sandmo1 and Srinivasan2 analyze the tradeoff between the income saved from evasion and the risk of being caught and penalized. In contrast this paper examines the incentive to evade tax in the presence of a leakage in tax revenue. The revenue lost is referred to as an administrative cost in the tradition of Heller3 and Yitzhaki.4 The analysis is carried out in the context of an Arrow-Debreu5 economy with a continuum of entrepreneurs and workers, and a government which levies a direct tax on the profit of entrepreneurs or an indirect tax on sales.b The proceeds of the tax are redistributed to workers. The inefficiency or corruption of government is modeled as a leakage of tax revenue which a In

India studies have shown that only fifteen percent of the spend on public works reaches the target beneficiary. b This way of modeling the economic agents of the economy owes its origin to the classical school of political economy starting with Ricardo.6 401

April 20, 2009

17:25

402

WSPC - Proceedings Trim Size: 9in x 6in

prasad

Rohit Prasad

neither reaches the intended recipient, nor benefits anyone else. Tax evasion is modeled as an under-reporting of the basis for taxation, without any associated risk of penalty. The paper shows that in a direct tax regime, a higher administrative cost coefficient will increase output and profits but at the expense of the workers welfare. This outcome occurs because workers anticipating a leakage in the transfer payment, work harder to sustain their consumption. The restoration of the workers’ utility and the entrepreneurs’ profit back to the original level require an increase in the tax rate or a reduction in the rate of tax evasion. The level of social welfare goes down with an increase in the administrative cost coefficient. Any economy with a high cost coefficient is Pareto dominated by some suitably defined economy with a low cost coefficient. Given an optimal tax rate at a low administrative cost coefficient, does a shift to a higher cost coefficient economy without a change in the tax rate justify a higher level of tax evasion? This remains an open question, though the paper does identify an example where this is true. The conclusions arrived at for a direct tax economy change significantly for an indirect tax regime. Here an economy without administrative costs is tax neutral. Since the redistribution of taxes eliminates the wedge between the cost borne by the entrepreneur and the consideration received by the worker(both are inclusive of taxes), an indirect tax economy with redistribution of taxes is identical to an Arrow-Debreu economy. A change in the tax rate merely has the effect of scaling prices. When there is an administrative cost however, a wedge is introduced between the price paid by the entrepreneur and the price received by the worker, as the worker does not receive the entire tax paid by the entrepreneur. This makes taxes non-neutral. Any economy with zero administrative costs is identical in terms of equilibria with a positive cost economy where the tax rate is zero(since at this tax rate the administrative cost ceases to play a role). However an economy with a positive cost coefficient and positive taxes cannot be mapped to any zero cost coefficient economy. The introduction of positive administrative cost introduces possibilities that did not exist at zero administrative cost. In the same vein, any economy at an administrative cost greater than zero is identical to an economy with an even higher administrative cost provided the tax rate is set at an appropriately low level with respect to the tax rate in the lower cost economy(thereby dampening the effect of the administrative cost). On the other hand, the tax rate required to equate a

April 20, 2009

17:25

WSPC - Proceedings Trim Size: 9in x 6in

Tax Evasion and Administrative Costs

prasad

403

low cost economy with a high tax, high cost economy may be higher than acceptable limits. It follows that the optimized social welfare in a high cost economy would be at least as high as that in a low cost economy. If the low cost economy happens to achieve an optimal social welfare which is equal to that in the high cost economy, then the socially optimal tax rate is lower in the high cost economy since a lower tax along with the higher cost coefficient is equivalent to a higher tax with a lower cost coefficient. In this case, if taxes are not lowered in response to an increase in the cost coefficient an entrepreneur would be justified in evading taxes for the social good. However if the higher administrative cost economy is greater in social welfare than the lower administrative cost economy, the entrepreneur may not be justified in a higher degree of tax evasion. For a given tax rate, a higher administrative cost coefficient will increase output and profits only if the degree of substitutability between leisure and the consumption good in the utility function of the worker is low. For a given administrative cost coefficient, a reduction in the tax rate will have the same effect. Tax evasion is equivalent to a lowering of tax rates. For a given administrative cost coefficient and a suitably low positive tax rate, a self interested entrepreneur may not choose to evade taxes if the substitutability between leisure and the consumption good is low. The rest of the paper is organized as follows: the first half is devoted to the direct tax model and the second half to the indirect tax model. In Section 2 the baseline model with direct taxes is delineated, existence and uniqueness of equilibrium established and impact of a redistributive tax characterized for a robust sub-class of economies. Next an exogenous administrative cost coefficient is introduced and an equivalence established with the model without administrative cost. Finally the implications of administrative cost for output, profits and worker welfare and consequently for the optimal tax rate are drawn out. In Section 3, a similar analysis is carried out for an indirect tax regime. Section 4 presents concluding remarks.

2. Model with Direct Taxes We consider an economy lasting two time periods t ∈ T = {0, 1}. The economy consists of a continuum of entrepreneurs (E), and workers(W ). There are two goods in the economy - labor l(leisure l) and corn c. Workers are endowed with L units of labor in period 1 and entrepreneurs possess

April 20, 2009

17:25

WSPC - Proceedings Trim Size: 9in x 6in

404

prasad

Rohit Prasad

technology that converts labor supplied by the worker in Period 1 into corn in Period 2. This technology is given by f : R+ → R+ where f satisfies the usual assumptions of differentiability, strict concavity, impossibility of free  production, and possibility of no production. We assume that f (0) = ∞,  i.e. the marginal product at zero input level is infinite and that f (L) > 0. Workers like corn and leisure, while entrepreneurs maximize profits. The 1 ρ worker has a standard CES utility function given by uw (l, c) = (l + cρ )− ρ with −∞ < ρ < 1. 2.1. Government The government levies a tax τ ∈ [0, 1] on the surplus of entrepreneurs and transfers it to the worker in order to maximize a differentiable, perfectly concave, non-decreasing social welfare function defined on the worker welfare and entrepreneur profits. 2.2. Sequence of Activities 2.2.1. Period 1 The following activities take place in the order listed: (1) The labor market meets. The worker exchanges labor for future corn. (2) The labor sold by the worker is employed in the production of corn. 2.2.2. Period 2 The following activities take place in the order listed: (1) Production of corn is realized. (2) The entrepreneur redeems his corn obligation to the worker and pays (corn) tax to the government. (3) The tax is transferred to the worker and consumption takes place. Let qlW ≡ quantity of labor sold by a worker in Period 1, and qcW ≡ quantity of corn demanded by a worker in Period 2. The set of actions available to the worker (qlW , qcW ) is denoted by q W . Let qlE ≡ quantity of labor demanded by an entrepreneur in Period 1, qcE ≡ quantity of corn sold by the entrepreneur in period 2. The set of actions available to an entrepreneur (qlE , qcE ) is denoted by q E . Given the tax rate τ , the actions of the entrepreneur induce a tax obligation of (f (qlE ) − qcE )τ units of corn in Period 2. Let p ≡ the relative price of price of labor to the price of corn, i.e. the real wage.

April 20, 2009

17:25

WSPC - Proceedings Trim Size: 9in x 6in

prasad

Tax Evasion and Administrative Costs

405

2.3. The Budget Set of a Worker The set of feasible allocations AW of leisure and corn are given by the following equations: Labor sold to entrepreneur ≤ Labor endowment: qlW ≤ L

(1)

Leisure consumed = Labor endowment - Labor sold: = L − qlW AW l

(2)

Corn consumed = Corn bought + corn obtained through transfers: W W W W AW c ≡ qk = pql + (f (ql ) − pql )τ

(3)

The set of allocations AW corresponding to actions q W that satisfy these constraints is denoted by ΣW (p) and is called the budget set of the worker. 2.4. The Budget Set of an Entrepreneur We define AE , the final allocation of an entrepreneur as follows: AE c = (f (qlE ) − qcE )(1 − τ ) where S ≡ f (qlE ) − qcE is the gross surplus of the firm E available and π ≡ AE c the profit. The constraints on the set of actions q to an entrepreneur given p are as follows: In Period 1: Cost of labor demanded = Future corn sold: pqlE = qcE

(4)

In Period 2: Corn sold + Corn paid as tax ≤ corn produced: qcE + (f (qlE ) − qcE )τ ≤ f (qlE )

(5)

qcE (1 − τ ) ≤ f (qlE )(1 − τ )

(6)

i.e.

E The set of AE c corresponding to actions q that satisfy these constraints is E denoted by Σ (p) and is called the budget set of the entrepreneur.

2.5. Equilibrium A vector of allocations, and prices (AW , AE ; p) is an equilibrium for a given τ if:

April 20, 2009

17:25

WSPC - Proceedings Trim Size: 9in x 6in

prasad

Rohit Prasad

406

(1) All workers are optimal on their budget sets, i.e. for workers: AW ∈ ΣW (p) and AˆW ∈ ΣW (p) ⇒ uW (AˆW ) ≤ uW (AW ) For entrepreneurs, AE ∈ ΣE (p) AˆE ∈ ΣE (p) ⇒ AˆE ≤ AE

(2) All markets clear, i.e. in the labor market, qlE = qlW In the corn market, qcE = qcW 2.6. Equations of Equilibrium The first order condition for utility maximization for workers is:  M Ul W (A ) ≡ g = p(1 − τ ) + τ f c M Uc

The first order condition for profit maximization for entrepreneurs is: 

f (qlE ) = p These two equations along with the budget constraints of the workers and entrepreneurs and the market clearing condition constitute the equations of equilibrium. c The



boundary equilibrium at which qlW = 0 is ruled out by the assumption f (0) = ∞. The boundary equilibrium at which qlW = L is ruled out by the assumption g(0, c) = ∞.  Therefore the assumption that at equilibrium g = p(1 − τ ) + τ f is without loss of generality.

April 20, 2009

17:25

WSPC - Proceedings Trim Size: 9in x 6in

prasad

Tax Evasion and Administrative Costs

407

2.7. Existence and Uniqueness of Equilibrium Theorem 2.1. For all τ ∈ [0, 1] there exists an equilibrium. Proof. (i) The proof for τ ∈ [0, 1) proceeds in two steps. We first show that the excess demand for labor, EDl = qlE − qlW , is strictly positive for sufficiently small p and strictly negative for sufficiently large p. Given the continuity of the excess demand function and Walras’ Law, this establishes the existence of a price at which all equilibrium conditions are satisfied.d Note that the labor demand function is given by 

f (qlE ) = p 

Given the assumption that f (L) > 0, for sufficiently small p, the quantity demanded will be greater than L the highest possible level of the supply of labor. This shows that for sufficiently small p the excess demand for labor is strictly positive. To show that for sufficiently large p, the excess demand for labor is strictly negative we show that for large enough p, the demand for corn is greater than f (L), the physical limit of production in the economy, i.e. there is an excess demand for corn. From this it follows that there is an excess supply of labor at that price. Note that given the assumptions on the workers’ utility it follows that g(0, c) = ∞, ∀c > 0. We have also assumed that uW is strictly concave. Hence ∃B s.t. g(L, f (L)) < B. For large enough p the worker will be able to to buy f (L), the physical limit of production. For such a p, we define the threshold labor supply l(p) as the quantity of labor that would enable the worker to buy exactly f (L) units of corn. l(p) is given by the equation: pl(p)(1 − τ ) + τ f (l(p)) = f (L) f (L) Note that that l(p) is bounded above by p(1−τ ) and that for any τ ∈  [0, 1), there exists a p large enough that the worker is able to buy f (L) and

d Recall

there are only two goods in our model.

April 20, 2009

17:25

WSPC - Proceedings Trim Size: 9in x 6in

prasad

Rohit Prasad

408

g(L −

p (1 − τ ) > B

(7)

f (L) , f (L)) < B p (1 − τ )

(8)

Condition (8) states that the upper limit on l(p) the threshold labor supply is small enough that the upper bound on g(L, f (L)) continues (L)  to apply at the consumption point (L − pf(1−τ ) ; f (L)). At this p the 

slope of the worker’s budget line m(p , l) = p (1 − τ ) + τ f (l) is greater than the slope of the indifference curve at the consumption point (L − l(p ); f (L)). To see this note: m(p , L − l(p )) > p (1 − τ ) > B

B > g(L −

f (L) , f (L)) > g(L − l(p ); f (L)) − τ)

p (1

The last inequality follows from the observation that g declines as the amount of leisure consumed increases while corn consumed stays the same. Therefore the worker will optimize at a choice of leisure greater than l(p ), and demand an amount of corn greater than f (L), the physical limit. Thus at p , there is excess demand for corn and excess supply of labor. By the continuity of the excess demand function, there exists a price at which equilibrium exists. (ii) For τ = 1, the worker will get all the corn produced in return for the labor supplied. So the objective function is uW (L − qlW ; f (qlW )). By the maximum principle, an optimum exists and the choice of qlW is independent of p. The entrepreneur is indifferent about employing any quantity of labor at any prevailing wage rate, including qlW ∗ the quantity of labor the worker wants to supply. Therefore any positive real wage along with qlW ∗ constitutes an equilibrium when τ = 1. 



Theorem 2.2. Consider an economy with f + lf > 0, ∀l ∈ [0, L].e For all τ ∈ [0, 1] the equilibrium is unique.

e The

function y = A − exp(−l) does not satisfy this condition for l > 1. On the other 1

hand the function y = l n satisfies this condition for all l.

April 20, 2009

17:25

WSPC - Proceedings Trim Size: 9in x 6in

prasad

Tax Evasion and Administrative Costs

409

Proof. From the equations of equilibrium it follows that ql∗ , qk constitute an equilibrium at price p if and only if, p ql∗ (1 − τ ) + τ f (ql∗ ) = qk

(9)



f (ql∗ ) = p g(L −

ql∗ , qk )

(10)



= p (1 − τ ) + τ f



(ql∗ )

(11)

ql∗ ≤ L

(12)

qk

(13)

≤

f (ql∗ )

From this it follows that at equilibrium, 



g(L − ql∗ ; f (ql∗ )ql∗ (1 − τ ) + τ f (ql∗ )) = f (ql∗ )

(14)

From the first two equations of equilibrium it follows that, 

f (ql∗ )ql∗ (1 − τ ) + τ f (ql∗ ) = qk From this it follows that   ∂qk = f + (1 − τ )ql∗ f > 0 ∂ql∗ 

∂q



Given the assumption that f + lf > 0 ∀l ∈ [0, L], ∂qk∗ > 0, i.e. for l a given tax rate if there are two equilibria, the one with the higher output will also have a higher corn consumption on the part of the worker. Now we prove the result by contradiction. Suppose there are two equilibria for a given τ . At the higher output equilibrium E, the worker has less leisure, and more corn, therefore the LHS of (14) is higher in E compared to the lower output equilibrium. But f  declines with an increase in labor, so the RHS of (14) is lower in E compared to the lower output equilibrium. This implies that (14) cannot be satisfied for both the equilibria.

2.8. Impact of Change in Tax Rate 



Theorem 2.3. Consider an economy with f + lf > 0 ∀l ∈ [0, L]. In this economy an increase in tax rate will lead to a decrease in output.

April 20, 2009

17:25

WSPC - Proceedings Trim Size: 9in x 6in

prasad

Rohit Prasad

410

Proof. From Theorems 2.1 and 2.2, we know that there is a unique equilibrium therefore it is possible to examine the outcome of a tax increase. As shown earlier in (14), 



g(L − ql∗ ; f (ql∗ )ql∗ (1 − τ ) + τ f (ql∗ )) = f (ql∗ ) We first note that given the concavity of f ,  ∂qk = (f − f ql∗ ) > 0 ∂τ

(15)

i.e. qk is a positive function of τ . As shown in the proof of Theorem 2.2,    ∂qk = f + (1 − τ )(f + ql∗ f ) > 0 ∂ql∗

(16)

i.e. qk is a positive function of ql∗ . Suppose an increase in tax rate results does not lead to a decrease in the quantity of labor supplied. Then from (15) and (16) the worker must be consuming more corn and no more leisure. This implies that g the marginal rate of substitution of corn for leisure will go up. On the other hand the  marginal product f will stay the same or decrease as the quantity of labor has not gone up. Thus the equilibrium equation will not be satisfied. Corollary 2.1. The real wage and worker’s welfare increase with the tax rate. Proof. From Theorem 2.3 it follows that as the tax rate increases, output falls. Therefore marginal product must increase, and by the equation of equilibrium, so must the real wage p. Notice, ∂qkW = qlW (1 − τ ) ≥ 0 ∀qlW ∈ [0, L] ∂p This means that for any quantity of leisure consumed the worker has at least as much corn at a higher real wage, and strictly more for all qlW ∈ (0, L]. Therefore the budget set at a higher real wage is a superset of the budget set at a lower real wage, and the worker must be better off. Corollary 2.2. Profits fall with an increase in the tax rate.

April 20, 2009

17:25

WSPC - Proceedings Trim Size: 9in x 6in

prasad

Tax Evasion and Administrative Costs

411

Proof. From Theorem 2.3, output falls with an increase in the tax rate. The surplus of the entrepreneur is given by π = (f (ql∗ ) − p ql∗ )(1 − τ ) Substituting the entrepreneur’s profit maximizing condition we get π ∗ = (f (ql∗ ) − f  (ql∗ )ql∗ )(1 − τ ) From the concavity of f, (f (l) − f  (l)l), decreases with a decrease in output. (1 − τ ) also decreases with an increase in τ . The result follows.

2.8.1. Example 1 Consider an economy characterized by the following parameters: L = 1

2

√

10; f (l) = 30l − 12 l2 ; uW = l 3 c 3 . In this economy l = 110− 4900+1200τ . 6−τ It can be checked that as tax increases, output falls, real wage increases, the worker is better off, and profits falls.

2.9. The Direct Tax Model with Tax Evasion Suppose entrepreneurs can evade tax with impunity. Instead of paying tax on the entire surplus they pay tax on a proportion (1 − η) ∈ [0, 1]. It is   obvious that a tax rate τ and an evasion coefficient η , is equivalent to a   tax rate τ = (1 − η )τ , and an evasion coefficient η = 0.

2.10. The Direct Tax Model with Administrative Cost We now consider a variation of the model in which the tax collected from the entrepreneurs is not entirely transferred to the intended recipients, the workers. A proportion κ ∈ [0, 1] of tax revenue is lost on account of administrative costs. This variation implies two changes in the equations of equilibrium. Firstly, the amount of corn consumed by the worker is now given by the equation qkW = pqlW + (f (qlW ) − pqlW )τ (1 − κ) For this reason the workers’ utility maximization equation must factor the administrative cost, i.e. now

April 20, 2009

17:25

412

WSPC - Proceedings Trim Size: 9in x 6in

prasad

Rohit Prasad  M Ul W (A ) ≡ g = p(1 − τ (1 − κ)) + τ (1 − κ)f M Uc

The profit maximization condition of the entrepreneur remains the same. As in Theorem 2.1, it can be shown that there exists an equilibrium for all τ ∈ [0, 1], η ∈ [0, 1] and κ ∈ [0, 1]. Further, as in Theorem 2.2, if   f + lf > 0 ∀l ∈ [0, L], then the equilibrium is unique. 2.11. Comparing the Direct Tax Models With and Without Administrative Cost Theorem 2.4. An equilibrium at tax rate τ ∈ [0, 1] with cost coefficient κ ∈ [0, 1] is identical in terms of output, real wage, worker welfare, and gross surplus S of entrepreneurs to an equilibrium at tax rate τ (1 − κ) without administrative cost. Proof. Follows easily from the observation that the equations of equilibrium in the the model with tax rate τ and cost coefficient κ are identical to the equations of equilibrium in the the model with tax rate τ (1 − κ) and no administrative cost. The equivalence classes of τ and κ that yield the same equilibrium output, real wage, worker welfare, and gross surplus of entrepreneurs are given dτ τ = 1−κ ≥ 0. by the equation τ (1 − κ) = C and indicated by Fig. 1. Note dκ From Theorem 2.3, output declines and real wage and worker welfare increase with an increase in the tax rate. Therefore output and gross surplus decline and real wage and worker welfare increase in the NW direction of the equivalence map. The net profit of entrepreneurs does not share the same equivalence class as the worker’s welfare, or the surplus. This is because π = (1 − τ )S. Therefore the net profit of entrepreneurs is lower at the equilibrium with tax rate τ ∈ (0, 1] with cost coefficient κ ∈ (0, 1] compared to the equilibrium at tax rate τ (1 − κ) with κ = 0. In both equilibria, the gross surplus is the same, as the output and real wage are identical, but in the equilibrium with tax rate τ ∈ (0, 1] with cost coefficient κ ∈ (0, 1] entrepreneurs are worse off as they incur tax at the rate τ as opposed to τ (1 − κ). In fact, as we move along the workers’ indifference curve given by τ (1 − κ) = C, the gross surplus remains the same but profit decreases.

April 20, 2009

17:25

WSPC - Proceedings Trim Size: 9in x 6in

Tax Evasion and Administrative Costs

prasad

413

Note also that for given τ as κ increases, the economy moves to a lower effective tax rate τ (1−κ), and a higher output and surplus for entrepreneurs. The net profit will therefore increase with κ for a given τ . The following corollary summarizes the conclusions. Corollary 2.3. The slope of the indifference curve of profits is positive and lower than the slope of the indifference curve of the worker at every point in the τ − κ space. From Corollary 2.3, the equivalence classes of τ and κ that yield the same equilibrium profit will have a lower slope than the equivalence classes of τ and κ that yield the same welfare and surplus. They are also shown in the same figure.f

Fig. 1.

τ − κ Equivalence curves with direct taxes

f I have conducted the analysis with a constant κ. However the result on existence would go through with κ being any continuous function of the tax revenue. The result on uniqueness and decline in output with an increase in tax rate would go through with κ being any non-increasing function of tax revenue. The more complicated model has been omitted for expositional simplicity.

April 20, 2009

17:25

414

WSPC - Proceedings Trim Size: 9in x 6in

prasad

Rohit Prasad

Corollary 2.4. Assume the government is maximizing a differentiable, perfectly concave, non-decreasing social welfare function defined on the worker welfare and entrepreneur profits by choosing τ ∈ [0, 1] for an exogenously fixed κ. The social welfare cannot increase as κ increases. If social welfare at the lower administrative cost coefficient is maximized at a strictly positive tax rate, then it must strictly decrease with an increase in administrative cost coefficient. Proof. Consider any two cost coefficients κ1 and κ2 with κ2 > κ1 . Consider any τ > 0. For any economy defined by (τ, κ2 ), there exists an economy    (τ , κ1 ) such that τ (1 − κ1 ) = τ (1 − κ2 ). At (τ , κ1 ), the worker welfare is the same and profits higher than at (τ, κ2 ). Therefore the equilibrium at  (τ , κ1 ) pareto dominates the equilibrium at (τ, κ2 ). If τ = 0 then the social welfare is the same in both cases. The result follows. 2.12. Tax Evasion in a Direct Tax Regime Corollary 2.5. An equilibrium at tax rate τ ∈ [0, 1] with cost coefficient κ ∈ [0, 1] and tax evasion coefficient η ∈ [0, 1] is identical in terms of output, real wage, worker welfare, and profit of entrepreneurs to an equilibrium at tax rate (1 − η)τ (1 − κ) without administrative cost and without tax evasion. Proof. Follows in a straightforward manner from Theorem 2.4 and the remark in Sec. 2.9. It is clear from the above analysis that given τ ∈ (0, 1] and κ ∈ [0, 1), the entrepreneur always has an incentive to fix η = 1 to maximize profits. However, any increase in η makes the worker worse off. Consider a tax rate τ > 0 with an evasion coefficient η > 0 and a cost coefficient κ. If the administrative cost coefficient increases from κ  to κ without a change in the tax rate, then a lower level of tax evasion  η < η is required to restore the profits or the worker utility to the original levels(note both cannot be simultaneously restored to their original levels).  Equivalently a higher level of tax τ > τ is required to restore either the profits or the worker utility to the original levels. Suppose τ  > 0 maximizes social welfare with a cost coefficient κ and an evasion coefficient η > 0. Would the optimal tax rate increase with an increase in κ, given a constant η? Alternatively, would η need to decrease with an increase in κ, given a constant τ  to achieve the social optimum?

April 20, 2009

17:25

WSPC - Proceedings Trim Size: 9in x 6in

Tax Evasion and Administrative Costs

prasad

415

Given that restoration of either the level of utility or the level of profits enjoyed by the worker and the entrepreneur respectively following an increase in the administrative cost coefficient requires an increase in the tax rate (for a given evasion coefficient), one would surmise that the optimal tax rate would increase with an increase in the cost coefficient, given the tax evasion parameter. Alternatively, the level of tax evasion would need to come down given the tax rate. Example 2 shows that this need not always be true.

2.12.1. Example 2 Consider the economy of Example 1 characterized by the following param1

2

eters: L = 10; f (l) = 30l − 12 l2 ; uW = l 3 c 3 . Suppose the government is choosing τ to maximize a social welfare function defined over worker utility and profits given by ω = 100000u.95π .05 . We numerically compute the optimal choice of τ at two administrative cost coefficients κ1 = 0 and κ2 = 0.3. As can be seen from Table A1 and Table A2 in the Appendix, the optimal tax rate at κ1 = 0 is 0.4 with a corresponding social welfare of 422135 utils. The optimal tax rate at κ2 = 0.3 is 0.21 with a corresponding social welfare of 418678 utils. The optimal tax rate and the level of social welfare have fallen with an increase in the administrative cost coefficient.

3. Model with Indirect Taxes Consider a variation of the previous model in which the government levies an indirect tax on corn sales instead of a direct tax on corn surplus. For every unit of corn sold, τ ∈ R+ units are collected by the government and transferred to the worker.

3.1. The Budget Set of a Worker W We define AW , the final allocation of a worker as follows: AW l = L − ql , W W W and Ac ≡ qk = pql (1 + τ ). The constraints on the set of actions q W available to a worker given p, the relative price of labor and corn (gross of tax), are as follows: In Period 1: Labor sold to entrepreneur ≤ Labor endowment:

qlW ≤ L

(17)

April 20, 2009

17:25

WSPC - Proceedings Trim Size: 9in x 6in

prasad

Rohit Prasad

416

Corn purchased = Wage Income: qcW = pqlW

(18)

Total corn consumed by worker = Corn purchased + Corn transferred by government: qkW = pqlW + τ pqlW = pqlW (1 + τ )

(19)

The set of allocations AW corresponding to actions q W that satisfy these constraints is denoted by ΣW (p) and is called the budget set of the worker. 3.2. The Budget Set of an Entrepreneur We define AE , the final allocation of an entrepreneur as follows: AE c = f (qlE ) − qcE . The constraints on the set of actions q E available to an entrepreneur given p are as follows: In Period 1: Cost of labor demanded = Future corn sold: pqlE = qcE

(20)

In Period 2: Corn sold + Corn paid as tax ≤ corn produced: qcE (1 + τ ) ≤ f (qlE )

(21)

E The set of AE c corresponding to actions q that satisfy these constraints E is denoted by Σ (p) and is called the budget set of the entrepreneur.

3.3. Equilibrium A vector of allocations, prices and policy (AW , AE ; p) for τ ∈ R+ is an equilibrium if: (1) All workers are optimal on their budget sets, i.e. for workers, AW ∈ ΣW (p) and AˆW ∈ ΣW (p) ⇒ uW (AˆW ) ≤ uW (AW ) For entrepreneurs, AE ∈ ΣE (p) AˆE ∈ ΣE (p) ⇒ AˆE ≤ AE

April 20, 2009

17:25

WSPC - Proceedings Trim Size: 9in x 6in

prasad

Tax Evasion and Administrative Costs

417

(2) All markets clear, i.e. in the labor market, qlE = qlW In the corn market, qcE = qcW 3.4. Equations of equilibrium The first order condition for utility maximization for workers is M Ul W (A ) ≡ g = p(1 + τ ) M Uc The first order condition for profit maximization for entrepreneurs is 

f (qlE ) = p(1 + τ ) These two equations along with the budget constraints of the workers and entrepreneurs and the market clearing condition constitute the equations of equilibrium. The tax rate is the exogenous variable chosen by the government. The proof for existence of equilibrium ∀τ ≥ 0 proceeds along similar lines as the proof of Theorem 2.1. As in the proof of Theorem 2.2, for   economies with f + lf > 0,g the equilibrium is unique. 3.5. Impact of Change in Tax Rate Theorem 3.1. change in the tax rate has no impact on the level of output, worker welfare, or profit. Proof. Consider two economies, one with tax rate τ and the other with tax rate τ  . We prove the result by showing that there exists an equilibrium in the economy with tax rate τ if and only if there exists a corresponding equilibrium in the economy with tax rate τ  with the same level of output, worker welfare, and profit. Notice that the equilibrium equations at the tax rate τ with price p are identical to the equilibrium equations at the tax rate τ  at a price equal to p(1+τ ) . Therefore the two equilibria are identical except for the scaling of 1+τ  prices. g Please

see footnote i

April 20, 2009

17:25

WSPC - Proceedings Trim Size: 9in x 6in

prasad

Rohit Prasad

418

3.6. The Indirect Tax Model with Tax Evasion Suppose entrepreneurs can evade tax with impunity. Instead of paying tax on the entire sale of corn they pay tax on a proportion (1 − η) ∈ [0, 1]. It   is obvious that a tax rate τ and an evasion coefficient η , is equivalent to   a tax rate τ = (1 − η )τ , and an evasion coefficient η = 0. 3.7. The Indirect Tax Model with Administrative Cost Coefficient Equal to One We now consider a variation of the model in which the tax collected from the entrepreneurs is not transferred to the intended recipients, the workers. A proportion κ ∈ [0, 1] of tax revenue is lost due to administrative costs. This variation implies two changes in the equations of equilibrium. Firstly, the amount of corn consumed by the worker is now given by the equation qkW = pqlW (1 + (1 − κ)τ ) For this reason the workers’ utility maximization equation must factor the administrative cost, i.e. now M Ul W (A ) ≡ g = p(1 + (1 − κ)τ ) M Uc The profit maximization condition of the entrepreneur remains the same, i.e. 

f (qlE ) = p(1 + τ ) The proof for existence and uniqueness of equilibrium for all τ ∈ [0, 1], η ∈ [0, 1] proceeds along similar lines as the proof of Theorem 2.1 and Theorem 2.2. In this model the invariance of the economy with respect to taxes is no longer obtained. We first present the results for the case where κ = 1, i.e. all the corn collected as tax is spent on administration. Theorem 3.2. Consider an economy where workers’ utility is character1 ized by the form uW = (lρ + cρ )− ρ with 0 < ρ < 1, and where κ = 1. In this economy, output falls with an increase in tax rates. Proof. From the equations of equilibrium,

April 20, 2009

17:25

WSPC - Proceedings Trim Size: 9in x 6in

prasad

Tax Evasion and Administrative Costs

419



(1 + τ )g(L − ql∗ ;

 f (ql∗ )ql∗ ) = f (ql∗ ) 1+τ

Computing the MRS for the CES utility function we get  (L − ql∗ )(1 + τ ) ρ−1 ) = f (ql∗ ) f  (ql∗ )ql∗  L − ql∗ ρ−1 ⇔ (1 + τ )ρ ( ) = (f (ql∗ ))ρ ∗ ql

(1 + τ )(

Denoting get

L−ql∗ ql∗

(22) (23)

by ω and differentiating both sides with respect to τ we

  dql∗ (1 + τ )ρ (ρ − 1)Lω ρ−2 [ + ρ(f (ql∗ ))ρ−1 f (ql∗ )] = ω ρ−1 ρ(1 + τ )ρ−1 dτ ql∗ 2

The term in brackets on the LHS is negative as 0∗ < ρ < 1 and f is dq concave. The term on the RHS is positive. Therefore dτl must be negative.

Theorem 3.3. Consider an economy where workers’ utility is character1 ized by the form uW = (lρ + cρ )− ρ with 0 < ρ < 1, and where κ = 1. In this economy, profits decrease with an increase in the tax rate. Proof. The profit of the entrepreneur is given by: π = f (qlE ) − qcE Substituting the entrepreneur’s budget set we get π = f (qlE ) − pqlE (1 + τ ) Substituting the entrepreneur’s profit maximizing condition we get π ∗ = f (qlE ) − f  (qlE )qlE From the concavity of f , the profit increases with output. From Theorem 3.1, output decreases with an increase in the tax rate. The conclusion follows. No conclusion can be arrived at about the welfare of the worker from an increase in taxes. We present an example where an increase in taxes results in a decrease in output and a Pareto decline in the economy.

April 20, 2009

17:25

WSPC - Proceedings Trim Size: 9in x 6in

prasad

Rohit Prasad

420

3.7.1. Example 3 Consider an economy characterized by the following parameters: L = 0.5 + c0.5 )−2 . In this economy l = 10; f (l) = 30l − 12 l2 ; uW = (l √ τ +41− 481+τ 2 +82τ . It can be checked that as tax increases, output falls, 2 real wage decreases, the entrepreneur and worker are worse off, and tax revenue decreases. However in the indirect tax model output need not necessarily go down with an increase in taxes. 3.7.2. Increase in Output with the Tax Rate In the following economy output increases with an increase in the tax rate. 3.7.3. Example 4 Consider an economy described by the following parameters. L = 1; uW = 2 min{l, c}; f (l) = 10l − l2 .h In this economy the equilibrium labor supply is given by √ 11 + τ − τ 2 + 18τ + 117 ∗ ql = 2 The derivative with respect to τ dql∗ 1 2τ + 18 = − √ dτ 2 4 τ 2 + 18τ + 117 is greater than zero. The equilibrium labor supplied is less than the endowment at tax rates up to one million! In this economy output and profits increase with the tax rate. From the worker’s utility function the optimization implies that at equilibrium L − ql∗ = qk Therefore the indirect utility v W = min{L − ql∗ , qk } is a declining function of τ . h Notice

that Leontieff utilities do not fall within the space of utility functions chosen in the specification of the model. The analysis of the economy with Leontieff utilities can be used to understand the behaviour of ‘near-Leontieff economies’, which do fall within the ambit of the specification, and whose outcomes would be close to the outcomes of the Leontieff economy, by continuity of all our functional forms.

April 20, 2009

17:25

WSPC - Proceedings Trim Size: 9in x 6in

prasad

Tax Evasion and Administrative Costs

421

3.8. The General Indirect Tax Model with Administrative Costs We now map any economy characterized by [τ, κ] with τ ∈ R+ and κ ∈ [0, 1]  to an economy with tax rate τ  and κ = 1. Theorem 3.4. The indirect tax economy with tax rate τ ∈ R+ and κ ∈ [0, 1] is equivalent in terms of output, worker welfare, and entrepreneur   κτ and κ = 1. profit to the indirect tax economy with tax rate τ = 1+(1−κ)τ Proof. Solving the equations of equilibrium for the economy with tax τ and κ ∈ [0, 1] we get 

g(L − qlW ; f (qlW )qlW

 1+τ (1 + (1 − κ)τ ) ) = f (qlW ) 1+τ 1 + (1 − κ)τ

Recall the equilibrium condition for the indirect tax economy with tax τ  and κ = 1 is 

g(L − qlW ;

  f (qlW )qlW )(1 + τ ) = f (qlW )  1+τ



kτ Setting τ = 1+(1−κ)τ , it is easy to see that the two economies will have the same output and consumption by the worker, and therefore worker welfare. The price in the economy with tax τ and κ ∈ [0, 1] is given by the equation: 

f (qlW ) = p(1 + τ ) 

In the economy with tax rate τ = ing equation is: f  (qlW ) = p

kτ 1+(1−κ)τ

(24) 

and κ = 1 the correspond-

1+τ 1 + (1 − κ)τ

(25)

The LHS of 24 and 25 are equal as shown through the equilibrium condition of the two economies. It follows that, p=

p 1 + (1 − κ)τ

April 20, 2009

17:25

WSPC - Proceedings Trim Size: 9in x 6in

prasad

Rohit Prasad

422

The entrepreneur profit in the economy with tax τ and κ ∈ [0, 1] in terms of p is given by f (ql∗ ) − pql∗ (1 + τ ) = f (ql∗ ) − p ql∗

1+τ 1 + (1 − κ) tau 

This is also the entrepreneur profit in the economy with tax rate τ = kτ  1+(1−κ)τ and κ = 1. Corollary 3.1. Consider an economy where workers’ utility is character1 ized by the form uW = (lρ + cρ )− ρ with 0 < ρ < 1. At a given τ > 0, as we increase the cost coefficient κ, the output and profits decrease. Proof. Notice that for any τ, κ the corresponding effective rate τ  in the  economy with κ = 1 is an increasing function of κ. 

τ + τ2 ∂τ = >0 ∂κ 1 + (1 − κ)τ Therefore as we increase κ the resulting equilibrium is the same as the   equilibrium in the economy with κ = 1 and a tax rate τ > τ . The result follows from Theorem 3.1 and Theorem 3.2. Corollary 3.2. Consider an economy where workers’ utility is character1 ized by the form uW = (lρ + cρ )− ρ with 0 < ρ < 1. At a given κ > 0, as we decrease the tax rate τ , the output and profits increase. Proof. Notice that for any τ, κ the corresponding effective rate τ  in the  economy with κ = 1 is an increasing function of τ . 

κτ ∂τ = >0 ∂τ 1 + (1 − κ)τ As we decrease τ the resulting equilibrium is the same as the equilibrium   in the economy with κ = 1 and a tax rate τ < τ . The result follows from Theorem 3.1 and Theorem 3.2. Corollary 3.3. Consider the economy of Example 4. L = 1; uW = 2 min{l, c}; f (l) = 10l − l2 . At a given τ ∈ [0, 1000000],i as we increase the cost coefficient κ, the output and profits increase. i The

bound on τ is set to ensure nonnegative values of the variables at equilibrium.

April 20, 2009

17:25

WSPC - Proceedings Trim Size: 9in x 6in

prasad

Tax Evasion and Administrative Costs

423

Proof. As shown in Corollary 3.1, when we increase κ the resulting equilibrium is the same as the equilibrium in the economy with κ = 1 and a  tax rate τ > τ . The result follows from the conclusion of Example 4. Corollary 3.4. Consider the economy of Example 4. L = 1; uW = 2 min{l, c}; f (l) = 10l − l2 . At a given κ ∈ (0, 1), as we increase the tax rate in the range [0, 1000000], the output and profits increase. Proof. As shown in Corollary 3.1, as we increase κ the resulting equilibrium is the same as the equilibrium in the economy with κ = 1 and a tax  rate τ > τ . The result follows from the conclusion of Example 4. The equivalence classes of τ and κ that yield the same equilibrium outcomes are indicated by Figs. 2 and 3.

Fig. 2.

τ − κ Equivalence curves with indirect taxes for strong substitutes

April 20, 2009

17:25

WSPC - Proceedings Trim Size: 9in x 6in

prasad

Rohit Prasad

424

Fig. 3.

τ − κ Equivalence curves with indirect taxes for strong complements

From the equivalence map it is clear that any economy characterized by a rate of tax τ and cost coefficient κ can be mapped on to an equivalent economy with 

τ =

 kτ ;κ = 1 1 + (1 − κ)τ 

Notice that the indifference curve for a given τ  with κ = 1, is a hyperτ bola with an asymptote at 1+τ . Corollary 3.5. An equilibrium at tax rate τ ∈ [0, 1] with administrative cost coefficient κ ∈ [0, 1] and tax evasion coefficient η ∈ [0, 1] is identical in terms of output, real wage, worker welfare, and profit of entrepreneurs to    κ(1−η)τ and κ = 1, η = 0. an equilibrium at tax rate τ = 1+(1−κ)(1−η)τ Proof. Follows in a straightforward manner from Theorem 3.4 and the remark in Section 3.6.

April 20, 2009

17:25

WSPC - Proceedings Trim Size: 9in x 6in

prasad

Tax Evasion and Administrative Costs

425

3.9. Increase of Social Welfare with an Increase in the Administrative Cost Coefficient 2 −→ R+ be a continuous social welfare function defined on Let W : R+ worker utility and entrepreneur profit. Let us assume there is an upper bound on the tax rate given by τ . Given the differentiability of all functions, W is a continuous function of τ ∈ [0, τ ] at any given cost coefficient κ ∈ [0, 1]. Therefore by the maximum principle, there exists a maximum social welfare at any κ.  Suppose we increase the administrative cost coefficient to κ > κ. From Theorem 3.4 and the equivalence map it is clear that all combinations  of utility and profit possible at κ are also possible at κ . However there  are some combinations of utility and profit possible at κ that are not possible at κ as the tax rate that would establish equivalence lies outside the domain. Therefore social welfare cannot decrease with an increase in the administrative cost coefficient. From example 4, it is clear that an increase in the tax rate at any given cost coefficient could result in an increase in profit and an increase in welfare for a suitably defined social welfare function. Therefore it is possible that social welfare increases with an increase in the administrative cost coefficient. If social welfare remains constant with an increase in κ, then there exists an optimal tax lower than the optimal tax chosen at the lower cost coefficient which will achieve the maximum.

3.10. Tax Evasion in an Indirect Tax Regime It is clear from the above analysis that given τ ∈ (0, 1] and κ ∈ [0, 1), the entrepreneur may not have an incentive to fix η = 1 to maximize profits. For instance, in the economy of Example 4, the entrepreneur wants an effective rate of tax greater than zero. On the other hand in the same example, the workers’ welfare is maximized at an effective tax rate equal to zero, therefore the worker is best off with η = 1! Now consider a socially optimal tax rate τ  with an evasion coefficient  η. If the administrative cost coefficient increases from κ to κ , then τ  may  no longer be optimal. If the optimal social welfare with κ is equal to the optimal welfare with κ, then the optimal effective rate of tax is lower at  κ . If the government does not change the tax rate, the socially optimal strategy of the entrepreneur would be to increase the evasion coefficient.

April 20, 2009

17:25

426

WSPC - Proceedings Trim Size: 9in x 6in

prasad

Rohit Prasad 

If on the other hand, the optimal social welfare with κ is greater than the optimal welfare with κ, then the optimal effective rate of tax may not be  lower at κ . One cannot in general draw any conclusion about the optimal evasion response of the entrepreneur. 4. Concluding Remarks Tax evasion is often justified on grounds of a happy coincidence of self interest and societal interest in the context of high tax administration costs. In fact, both tax evasion and administrative costs represent leakages from tax revenue and one might expect that an increase in one kind of leakage would require a decrease in the other kind to restore the economy to its original condition. But this holds true only in a direct tax regime, not in an indirect tax regime. This qualitative difference makes the two tax regimes quite different in terms of the appropriate evasion response to a change in administrative costs. This paper finds that tax evasion by entrepreneurs increases output and is warranted by self interest in a direct tax regime. However in an indirect tax regime it may reduce output and profits if labor and the consumption good are strong complements in the utility function. With direct taxes, restoration of the economy to its original state is not possible after an increase in the administrative cost coefficient. Higher costs always lead to a decline in social welfare. One can restore either the worker’s utility level of the entrepreneur’s profit level. Both require an increase of the tax rate, or alternatively a decrease in the evasion rate. Given this, it is surprising to find an example showing that the optimal tax rate need not increase with an increase in the cost coefficient. Or in other words the appropriate evasion response may not involve a reduction of the evasion rate. Finding examples and general conditions under which the appropriate evasion coefficient would decrease with an increase in the administrative cost coefficient is an open problem. With indirect taxes higher administrative costs always justify an increase in tax evasion for the greater good in cases where the increased administrative costs do not create higher social welfare possibilities as they may sometimes do. However if an increase in the administrative cost does lead to an increase in social welfare, then greater tax evasion may not be justified either in self interest or for societal welfare. Future directions of research include making the tax evasion coefficient and the administrative cost coefficient endogenous, introducing public

April 20, 2009

17:25

WSPC - Proceedings Trim Size: 9in x 6in

prasad

Tax Evasion and Administrative Costs

427

goods, and modeling the game between the government and the taxpayer in the context of a general equilibrium model. Appendix A. Table A1.

Outcomes with No Administrative Cost

Tax Rate

Labor

Corn Consumption

Utility

Profit

Welfare

0.00 0.10 0.20 0.30 0.36 0.37 0.38 0.39 0.40 0.41 0.42 0.43 0.44 0.45 0.50 0.60 0.70 0.80 0.90 1.00

6.67 6.64 6.6 6.57 6.56 6.55 6.55 6.55 6.54 6.54 6.54 6.54 6.53 6.53 6.52 6.49 6.46 6.43 6.41 6.38

155.56 157.23 158.88 160.49 161.45 161.61 161.76 161.92 162.08 162.24 162.39 162.55 162.71 162.86 163.64 165.17 166.67 168.15 169.60 171.04

43.21 43.65 44.09 44.52 44.77 44.81 44.86 44.90 44.94 44.98 45.02 45.07 45.11 45.15 45.35 45.76 46.16 46.55 46.93 47.31

22.22 19.81 17.45 15.13 13.76 13.53 13.30 13.08 12.85 12.63 12.40 12.18 11.95 11.73 10.61 8.42 6.26 4.14 2.05 0.00

417938 419622 420933 421804 422076 422100 422118 422130 422135 422133 422124 422108 422084 422053 421771 420453 417698 412439 401351 0

April 20, 2009

17:25

WSPC - Proceedings Trim Size: 9in x 6in

prasad

Rohit Prasad

428 Table A2.

Outcomes with Thirty Percent Administrative Cost Coefficient

Tax Rate

Labor

Corn Consumption

Utility

Profit

Welfare

0.00 0.10 0.16 0.17 0.18 0.19 0.20 0.21 0.22 0.23 0.24 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00

6.67 6.64 6.63 6.63 6.63 6.63 6.62 6.62 6.62 6.62 6.61 6.60 6.58 6.56 6.54 6.52 6.50 6.48 6.46

155.56 156.73 157.43 157.55 157.66 157.78 157.89 158.01 158.12 158.24 158.35 159.04 160.17 161.29 162.39 163.48 164.56 165.62 166.67

43.21 43.52 43.71 43.74 43.77 43.80 43.83 43.86 43.89 43.92 43.95 44.13 44.43 44.73 45.02 45.31 45.60 45.88 46.16

22.22 19.87 18.47 18.24 18.01 17.78 17.55 17.31 17.08 16.85 16.62 15.25 12.99 10.76 8.55 6.37 4.22 2.10 0.00

417938 418466 418635 418651 418663 418672 418677 418678 418676 418669 418658 418502 417843 416545 414354 410800 404840 393225 0

References 1. M. C. Allingham and A. Sandmo, Journal of Public Economics 1, 323(November 1972). 2. T. N. Srinivasan, Journal of Public Economics 2, 339 (1973). 3. W. P. Heller and K. Shell, The American Economic Review 64, 338(May 1974), Papers and Proceedings of the Eighty-sixth Annual Meeting of the American Economic Association. 4. S. Yitzhaki, The American Economic Review 69, 475(June 1979). 5. K. J. Arrow and G. Debreu, Econometrica 22, 265(July 1954). 6. D. Ricardo, On the Principles of Political Economy and Taxation (John Murray, London, 1817).

INEQUALITY, PUBLIC INVESTMENT AND DEFICITS IN INDIA ERROL D’SOUZA Economics Area, Indian Institute of Management, Ahmedabad, and Visiting Senior Fellow, Institute of South Asian Studies, National University of Singapore. Email: [email protected] With economic growth as a priority goal of the state it is a puzzle as to why public investment in India declined since the mid 1990s despite no significant reduction in fiscal deficits. This paper advances the proposition that public investment affects the returns to the distribution of factor endowments differentially. The rise in inequality then turns the attention of the state towards redistribution. When public expenditures are financed by borrowing, increased inequality that creates pressures for redistributive transfers crowds out public investment. Future income generation gets adversely affected by a reversal of public investment which makes creditors impose borrowing constraints on the state. This can take the form of the enactment of fiscal responsibility legislation.

1. Introduction India’s fiscal deficit has deteriorated since the mid 1990s and now ranks amongst the worst in the world (Kochhar, 2006). Amongst emerging markets, only Turkey and Argentina have larger fiscal deficit to GDP ratios. The large fiscal deficit has been a persistent feature of the macro economy. Even though the balance of payments crisis of 1991 did result in the initiation of some fiscal restraint this was reversed in the mid 1990s. The deficit reduction reversed in part due to the low buoyancy of tax revenues as the tax system is narrowly based on indirect taxes and manufacturing and a few services, and customs revenues declined as trade has been liberalized (Rao, 2005). The deterioration in revenues was also accompanied by expenditure pressure after 1996-97 due to the substantial increase in the government pay and pension bill associated with the recommendation of the Fifth Pay Commission (Acharya, 2002). However, even as early as 1994, the Indian government decided not to accept further IMF loans as it sought to increase current social expenditures (e.g. cheap power to 429

430

Errol D’Souza

farmers and households) to politically consequential groups (Kohli, 2006, p.1363). Finally, expenditures surged on account of a rise in interest payments as financial repression was reduced and government borrowings took place at market rates of interesti (Acharya, 2002). The rise in fiscal deficits has given rise to concern about its macroeconomic impact and its sustainability (Lahiri and Kannan, 2004). Fiscal consolidation has become a salient policy objective and is sought to be achieved in India via the Fiscal Responsibility and Budget Management Act which became effective from July 2004. This Act specifies annual targets for fiscal correction and seeks to reduce the fiscal deficit to 3 per cent of GDP by March, 2008. A Task Force was also set up for drawing up the medium term framework for fiscal policies so as to achieve the targets as specified in the Act. With an adjustment path spelt out there is concern about whether the burden of adjustment will fall on public investment and other important items of expenditure such as operations and maintenance expenditures. Given the pressure on current expenditures deficits have been reduced mainly by cutting public investment and especially social and physical infrastructure spending. As a decline in public investment constrains growth there have been concerns raised about the need to step up this component of expenditure (Ahluwalia, 2002; Kochhar, 2006; Lahiri and Kannan, 2004). In fact, Kohli (2006) argues that the decline of public investment (and the buoyancy of private investment) is a “key element of India’s economic growth ‘story’ in the 1990s”. A significant feature of the pre-crisis 1980s is the growth in public investment that fueled the economic growth of that period. The 1990s and beyond by contrast has been associated with declining public investment (see Figure 1). The decline in public investment at a time when more expenditures are required on power, water, and rural infrastructure, is growth constraining to the extent that public investment is known to crowd in private investment in India (Serven, 1996; Murty and Soumya, 2006). Given that economic growth is a priority goal of the state (Kohli, 2006) this is a puzzle. The standard explanation which is an event driven one has been that the high levels of debt incurred in the 1980s and the subsequent balance of payments crisis of 1991 shifted the focus of fiscal policy towards the low level of government savings and resulted in the initiation of a fiscal restructuring and compression of public expenditures. As the cash flow stream associated with public investment in infrastructure is such that high costs are incurred in the present and the returns though high, accrue over the long run, postponing lumpy and costly public

Inequality, Public Investment and Deficits in India

431

investment spending is far easier for a government than cutting current expenditures. Expenditure compression is therefore linked to investment expenditure cuts. Figure 1: Public Investment and Fiscal Deficits in India 12

10

Per cent of GDP

8

6

4

2

1 19 72 19 73 19 74 19 75 19 76 19 77 19 78 19 79 19 80 19 81 19 82 19 83 19 84 19 85 19 86 19 87 19 88 19 89 19 90 19 91 19 92 19 93 19 94 19 95 19 96 19 97 19 98 19 99 20 00 20 01 20 02 20 03 20 04

19 7

19 7

0

0

Year Source: RBI, Handbook of Statistics on the Indian Economy

Public Investment

Fiscal Deficit of Centre & States

It is often argued that policy changes, stops, and reversals tend to be episodic. They are often triggered by discrete changes or shocks such as banking and balance of payments crises, changes in government, changes in global interest rates, and even leverage exercised by international financial institutions (Krueger, 1993). The event driven explanation has merit in ex post identifying the exact timing of a policy change. However, it does not interpret a policy change as part of a process – an event may be part of a larger process and identifying the structure of a process can offer an altogether different understanding of policy changes. The event or shock may then be a factor that hastens or hinders a policy change that would nevertheless have eventually occurred if the underlying economic processes had unfolded undisturbed. Our focus in this paper is on identifying the process behind public investment reversals in India. We consider a two period representative agent economy with utility over consumption in the two periods. In the first period the agent receives an endowment and a lump-sum redistributive transfer from government. The agent decides how much to save/invest and this amount invested along with complementary public investment contributes to the output of the second period.

432

Errol D’Souza

The government borrows and allocates this amount between redistributive transfer expendituresii and public capital expenditures. The government is unable to credibly commit to the composition of public expenditures before it borrows. Accordingly one of our objectives is to inquire as to whether this lack of commitment has implications for the extent of borrowing by the government as well as its public expenditure program. We inquire into the implications of increased inequality on government borrowing and the composition of public expenditures. The evidence on rising inequality during the liberalization period is unequivocal. Mahendra Dev and Ravi (2007) find that inequality in consumption as measured by Gini coefficients has increased significantly for both rural and urban areas from 1983 to 2004-05, with the rate of increase being higher for urban as compared to rural areas. Deaton and Dreze (2002) sum up their findings as follows – “We find strong indications of a pervasive increase in economic inequality in the nineties. This is a new development in the Indian economy: until 1993-94, the all-India Gini Coefficients of per capita consumer expenditure in rural and urban areas were fairly stable. Further, it is worth noting that the rate of increase of economic inequality in the nineties is far from negligible”. The rise in inequality is associated with an increase in the skewness of the distribution of income. The increased concentration of income at the top makes redistribution more attractive for the median voter rather than public capital expenditure. A rise in inequality (interpreted standardly as a decline in median income relative to mean income) increases the preference of the median income voter towards transfers and redistributive expenditures and away from public capital expenditure. As inequality increases we demonstrate that the government becomes more attentive to increasingly distressed median voter preferences that are decisive in electoral outcomes. Such governments reallocate public expenditure towards transfers and away from public investment. However, the rise in fiscal deficits and transfer expenditures by the government were contemporaneous. After years of discussion the Fiscal Responsibility and Budget Management (FRBM) Act was passed in 2004 which specified annual targets for the reduction of the deficit along with a reduction in debt liabilities. This limits the expenditure that government can incur and raises the question as to whether the result that increased inequality results in a reduction in capital expenditures by a government sensitive to median voter preferences holds when financing of those expenditures is on the basis of borrowing rather than taxes as demonstrated in D’Souza(2007). A related question we address is whether a rise in inequality results in voter preferences

Inequality, Public Investment and Deficits in India

433

for limiting the borrowing capacity of government as occurs when a fiscal responsibility legislation is introduced. We turn to examining these questions in the next section. The rise in transfers in India were accompanied by high fiscal deficits and borrowings were deployed towards transfer expenditures. This reduced emphasis of public expenditure policy on public investment is tantamount to signaling a reduction in potential growth. Lenders who witness the constraints on economic growth associated with the increased emphasis of public expenditures on transfers would then have an incentive to impose a ceiling on borrowing by the government as a way of securing their returns. This paper accordingly argues that inequality and the composition of public expenditures were the important underpinnings to the passage of a fiscal responsibility act which constrains the growth of government expenditure. The approach we adopt presumes the existence of a large number of heterogeneous individuals who are beneficiaries of broadly targeted categories of public expenditure programmes – public investment and transfer programmes. As these components of public expenditure are of a broad nature and of universalistic design, they cannot be targeted easily to the specific demands of well-defined groups of voters. As a result the divergence of opinions in evaluating such programmes tend to be unidimensional and they run from left to right. Those with higher incomes favour public investment expenditures whereas those with relatively lower incomes favour transfers. This is an instance of general-interest politics (Persson and Tabellini, 2002). The motivation of politicians is assumed to be opportunistic – they are officeseeking politicians. Candidates and parties that covet power will announce intentions to carry out those policies and programmes that carry the greatest public appeal. As a result the tinge to their policy platforms will be in the direction of the median voter. Of course, as pointed out by Alesina (1988), such an explanatory framework is valid only if prospective rulers have the ability to commit to their actions. If not, prospective voters recognize the incentive of parties and candidates to resort to their own preferred policies on the resolution of the competition for political power, and do not give credence to their announced policy platforms. The model we adopt to explain government borrowing and the composition of public expenditures has no role for government credibly committing as to how it will allocate borrowed resources. In the same vein we do not address issues such as the desire to maintain reputations with voters over the long run as a political determinant of policy

434

Errol D’Souza

choices. We restrict ourselves to the situation where the median alternative is the programme/policy decision. 2. Borrowing and Public Expenditures As borrowing implies repayments, at the very least, we require a two period model. Here we consider any agent with utility over consumption ( C i ) in the two periods ( i = 1, 2 ). U = u ( C1 ) + θ u ( C 2 )

= log C1 + θ log C 2

(1)

where, θ is the impatience for period 2 consumption. On date 1 the individual receives an endowment Y1 , but no capital is inherited from the past. The individual also receives a redistributive transfer that is lump-sum, R . The agent decides how much to consume ( C1 ) and how much to save and invest ( K 1 ). The first period budget constraint then isiii C1 = Y1 + R − K 1

(2)

The amount invested in the first period along with public investment G produces an output in the second period given byiv Y2 = F ( K 1 , G ) = α K 1G

(3)

Consumption in the second period, then, is that output left over after repayments ( ℜ ) are made for the borrowings made by the government to finance its expenditures. C 2 = Y2 − ℜ

(4)

The government borrows an amount D (for debt) and allocates this between redistributive transfers and public capital expenditures D = R+G

(5)

The government may not repay the sum of interest and principal (1 + r ) D in full. In case it defaults, then, creditors get only a fraction (η ) of second period output due to enforcement and collection costs as the institutions of investment are weak (Gertler and Rogoff, 1990).

Inequality, Public Investment and Deficits in India

435

ℜ = min {(1 + r ) D,ηY2 }

(6)

The repayment on debt by the government is financed by taxing individuals at a proportionate rate t in order to cover the repayments. Thus, with tY2 = ℜ , second period consumption may be written as C 2 = (1 − t ) Y2

(7)

Having borrowed the government is free to choose what it wants to do with the money borrowed. It could just redistribute the money or spend it on public capital formation that raises date 2 output. Creditors are thus prompted to ask themselves the question that if they lend in period 1, will the government choose to invest enough to make ηY2 ≥ (1 + r ) D . If the government does not invest enough creditors won’t be repaid in full. Thus, creditors would be interested in fathoming how much they should safely lend. Public capital expenditures are significant in that they raise second period output and make it more likely that borrowings will be repaid. However, the government is unable to credibly commit to the composition of public expenditures before it borrows. We thus first consider the case of an individual who takes as parametric the composition of public expenditure and the tax rate and chooses consumption ( C1 , C 2 ) and private capital formation ( K 1 ) . The second period constraint is C 2 = (1 − t )Y2 = (1 − t )α K 1G

or,

C2 = K1 (1 − t ) α G

Add the first period constraint C1 = Y1 + R − K 1

to the second period constraint to obtain the intertemporal budget constraint C1 +

C2 = Y1 + R (1 − t ) α G

(8)

Errol D’Souza

436

Maximizing (1) subject to (8) gives C1 =

1 (Y1 + R ) 1+θ

C 2 = (1 − t )α G K1 =

θ 1+θ

θ 1+θ

(9)

(Y1 + R)

(Y1 + R )

(10)

(11)

Substituting (9) and (10) into (1) gives the maximized lifetime utility of an individual as U = log

1 θ (Y1 + R) + θ log(1 − t )α G (Y1 + R ) 1+θ 1+θ

or, U = θ log(1 − t )α G + log

θθ + (1 + θ ) log(Y1 + R ) θ +1 (1 + θ )

(12)

The preferences of an individual for the composition of public expenditures will depend on whether the government repays the borrowing or defaults. Let us first take the case of default. Then, the tax rate in the second period equals the fraction of second period output that creditors will be able to recover, i.e., t = η . Then, lifetime utility under the condition of non repayment (N) will be U N = θ log (1 − η ) α ( D − R ) + log

θθ + (1 + θ ) log (Y1 + R ) 1+θ (1 + θ )

(13)

where we have used the government budget constraint to replace G in the first term. We can find the preference for transfers by the individual as ∂U Ν θ 1+θ =− + =0 D − R Y1 + R ∂R

(14)

The level of redistributive transfers that an individual would prefer under default then is

Inequality, Public Investment and Deficits in India

RN =

(1 + θ ) D − θ Y1 1 + 2θ

437

(15)

An individual would therefore prefer a level of transfers that is increasing with the debt and decreasing with income if non-repayment is to occur. We now turn to the case of repayment ( P ) of the borrowing. Then, from the second period optimal consumption plan (10) it must be the case that tα G

θ 1+θ

(Y1 + R ) = (1 + r ) D

(1 + r ) D

or, t =

αG

θ

1+θ

(16)

(Y1 + R )

Maximized lifetime utility in the case of repayment will then be U P = θ log(1 − t )α ( D − R ) + log

θθ + (1 + θ ) log(Y1 + R ) (1 + θ ) 1+θ

(17)

where, t is given by (16). Hence, the redistributive transfers preferred by an individual when there is repayment would be ∂U P θ θ 1+θ tR + =− − =0 D − R 1− t Y1 + R ∂R

(18)

Comparing (14) and (18), ( ∂U P ∂R ) < ( ∂U N ∂R ) provided t R > 0 . Thus, the level of redistributive transfers preferred by an individual in the instance where the government does not repay (case N) is greater than when there is repayment (case P). This is what our hunch would be as well. If there is going to be repayment in the second period then given a fixed outgo of (1 + r ) D it makes sense to ensure that second period income is as far above this fixed outgo as possible. That would occur if complementary public capital spending which increases second period income was increased and redistributive transfers lowered. On the other hand, if there is going to be non-repayment in the second period, then, as a fraction η of second period income would be lost to creditors, it is advantageous to reduce this outgo by reducing second-period income. A reduction in public capital expenditure achieves this purpose and diverts

438

Errol D’Souza

borrowings to redistributive transfers that increase income in the first period. Hence, an agent would prefer higher redistributive transfer expenditures when non repayment by government is to occur than when repayment will transpire in the second period. Now, if the individual concerned is the person with median income who is decisive in swaying policy makers as we are presuming, then, an increase in inequality as represented by a decline in median income relative to the mean will result in policy makers making larger redistributive transfers and investing less of the resources borrowed. This is obvious when we rewrite U P from (17) as U P = θ log(1 − t )α ( D − R ) + log

θθ

(1 + θ )

1+θ

+ (1 + θ ) log(Y + G )

+(1 + θ ) ⎡⎣ log(Y m + G ) − log (Y + G ) ⎤⎦

where, Y is the mean income and Y m the median income. The first part is the maximized lifetime utility of the individual with mean income which we write in short as U P (Y ) . The second part with the term log (Y m + G ) (Y + G ) in square brackets is the log difference of median from mean income. Then, P m ∂U P ∂U (Y ) (1 + θ ) (Y − Y ) = + ∂R ∂R (Y + G ) (Y m + G )

(19)

Thus, an increase in inequality as measured by an increase (Y − Y m ) will result in larger redistributive transfers in the situation where the government repays. As a result if we set Y1 = Y m in (15), transfers will also rise in the case of non-repayment. That transfers rise with inequality implies that second period income will be lower than otherwise. This raises an issue in creditors minds as to what are the safe limits to lending. It is safe to lend as long as the amount of public capital expenditure results in second period output that covers the cost of repayment in case of default, i.e., as long as ηY2 ≥ (1 + r ) D . Thus, it is safe to extend credit as long as U N − U P = 0 , where, to recall, U N = θ log (1 − η ) α G N + log

θθ

(1 + θ )

1+θ

+ (1 + θ ) log (Y1 + R N )

Inequality, Public Investment and Deficits in India

439

Similarly,

UP

⎡ ⎤ ⎢ ⎥ 1 r D + ( ) θθ = θ log α G P + θ log ⎢1 − + log ⎥ (1 + θ ) 1+θ ⎢ α G P θ (Y1 + R P ) ⎥ 1+θ ⎣ ⎦

+(1 + θ ) log(Y1 + R P ) Then, U N − U P = 0 is the expression given by –

0 = θ log

(1 − η ) G N GP

⎡ ⎤ ⎢ ⎥ 1+ r ) D ( Y + RN ⎥ − (1 + θ ) log 1 − θ log ⎢1 − Y1 + R P ⎢ αG P ⎛ θ ⎞ Y + R P ⎥ ) ⎜ ⎟( 1 ⎢ ⎥ ⎝1+θ ⎠ ⎣ ⎦

Exponentiating this equality, ⎡ ⎤ ⎥ ⎡ (1 − η ) G ⎤ ⎢ 1+ r ) D ( ⎥ 1= ⎢ ⎥ ⎢1 − P ⎣ G ⎦ ⎢ α G P ⎛ θ ⎞ (Y1 + R P ) ⎥ ⎜ ⎟ ⎢ ⎥ ⎝ 1+θ ⎠ ⎣ ⎦ N

θ

−θ

⎡ Y1 + R N ⎤ ⎢ P ⎥ ⎣ Y1 + R ⎦

− (1+θ )

Solving for D we obtain the limit beyond which creditors would be reluctant to extend credit as 1+θ ⎧ N − θ ⎫ ⎞ ⎪ ⎛ θ ⎞ P ⎪ P N ⎛ Y1 + R α⎜ ⎬ ⎟ (Y1 + R ) ⎨G − (1 − η ) G ⎜ P ⎟ 1 θ Y R + + ⎝ ⎠ ⎝ 1 ⎠ ⎪⎩ ⎪⎭ D= >0 1+ r

(20)

We thus see that creditors will not extend credit beyond D because that may require them to incur a penalty where they are not repaid the amount borrowed. Amongst the terms in (20) that increase the limit of borrowing include – (1) A greater fraction of second period output that creditors receive when default occurs – higher η (2) The greater is the productivity of capital in the economy – higher α (3) The lower is the interest rate – lower r (4) The higher is the weight on second period consumption – higher θ

440

Errol D’Souza

That there exists a debt ceiling implies that U Ν is steeper than U P so that when D < D the government repays and when D > D it does not repay the debt. What happens to the debt ceiling if inequality increases? We answer this question by inquiring as to what happens to U Ν and U P when median income increases without any alteration in mean income. In the equation for U Ν let the income be that associated with the person of median income. Then, an increase in median income affects the maximized lifetime utility of the individual in the case where a choice has been made to not repay the borrowing as follows ∂U N 1+θ = m m ∂Y1 Y1 + R N

(21)

Similarly, the increase in median income affects the maximized lifetime utility of the individual in the case where a choice has been made to repay the amount borrowed in the following way

θ (1 + r ) D ∂U P 1+θ = + m m ∂Y1 Y1 + R P ⎡ P⎛ θ ⎞ m ⎤ m P P ⎢α R ⎜ 1 + θ ⎟ (Y1 + R ) − (1 + r ) D ⎥ (Y1 + R ) ⎝ ⎠ ⎣ ⎦

(22)

As R N > R P , it is clearly the case that ( ∂U N ∂Y1m ) < ( ∂U P ∂Y1m ) . An increase in inequality by increasing the utility of non- repayment relative to repayment results in the decrease of the amount that the government may borrow. Thus as inequality increases (which we interpret as a decline in Y1m for a given mean income), the decrease in maximized lifetime utility is larger in the case of non-repayment than in the case of repayment. As a result the borrowing limit in (20) will be lower. This arises because as the distribution of income in the economy becomes more unequal there is pressure from the median citizen on policymakers to make more redistributive transfers. As a result the government spends less on public capital expenditures and this lowers second period output. The fraction of second period output that can be acquired by creditors in the event of default then decreases and this results in the response by creditors to reduce the amount which government may borrow. This is interpreted by us as an imposition of borrowing limits as for instance in a fiscal responsibility legislation. As the government is unable to commit to an increase in public capital expenditure prior to receiving funds from creditors, the creditors require the government budget constraint to be tightened so that the damage to future growth is contained.

Inequality, Public Investment and Deficits in India

441

3. Conclusion

The rise in fiscal deficit and the decline in public investment are major constituents of the Indian economic growth story since 1990. As economic growth has been a priority objective for the state in India and as public investment crowds in private investment this is a puzzle. The standard event driven explanation for this is that the increased fiscal deficits of the 1980s proved to be unsustainable and this required fiscal restructuring and a compression of public expenditures. As a result there was a pressure to introduce legislation to contain the deficit through the enactment of fiscal responsibility legislation. The long gestation period that accompanies public investment projects which generate returns over the longer run also makes it more appealing to contain expenditures by reducing public investment. Our approach, however, provides an endogenous explanation for this puzzle. We argue that public investment affects individuals differentially – those with higher capital factor endowments benefit more in terms of income returns from an increase in public investment than those with lower factor endowments. This creates incentives for those with above median incomes to influence the composition of public expenditures towards capital expenditures. Incomes above the median grow faster as government responds to these influence effects and inequality rises. A rise in inequality makes redistribution more attractive to the median voter and a government attentive to such preferences now reallocates expenditures towards transfers and away from public investment. Contemporaneously with the rise in inequality there was the global integration of the Indian economy which required that tax rates be moderated and the tax system modernized. This resulted in a reduction in the tax-GDP ratios in the economy and a rise in borrowings in order to finance public expenditures. Borrowings can be used for redistribution or to finance public capital formation. Borrowings may also be repaid or not re-paid. Borrowings that finance public capital expenditure increase future income and increases the repayment capacity of the government. Redistributive transfers, however, raise current incomes whilst leaving unaffected future income. As inequality increases and government resorts to redistributive transfers, creditors bothered about the repayment capacity of the government will end up limiting the amount that the government may borrow. A rise in inequality is thus associated endogenously with the imposition of borrowing limits as occurs in the enactment of a fiscal responsibility legislation.

Errol D’Souza

442

References

1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22.

S.Acharya, Economic & Political Weekly. 1515 (2002). M.S. Ahluwalia, Journal of Economic Perspective., 16(3), 67 (2002) A. Alesina, American Economic Review.78, 796 (1988) E. Baldacci., A.L. Hillman, and N.C. Kojo, European Journal of Political Economy. 20, 517 (2004) A. Banerjee, and T. Piketty, MIT, Department of Economics, Working Paper 03,32, (2003). A. Deaton, and J. Dreze, Economic & Political Weekly. 3729 (2002) E. D’Souza, Indian Journal of Labour Economics, 45(4) (2001) E. D’Souza, Working Paper, Institute of South Asian Studies, NUS (2007) E. D’Souza, Macroeconomics, Pearson Education, forthcoming (2008) M.Gertler and K. Rogoff, Journal of Monetary Economics. 26, 245 (1990) K.Kochhar, A Sustainable Fiscal Policy for India - An International Perspective, Oxford University Press, 44 (2006) A.Kohli, Economic & Political Weekly. 1251 (2006) A.Kohli, Economic & Political Weekly. 1361 (2006) A.O.Krueger, Political economy of policy reform in developing countries. Cambridge MA, MIT Press. (1993) A.Lahiri, and R. Kannan, Fiscal Policies and Sustainable Growth in India, Oxford University Press. 23 (2004) S. Mahendra Dev and C. Ravi, Economic & Political Weekly. 509 (2007) K.N. Murty and A. Soumya, IGIDR, Working Paper 2006-003, Mumbai (2006). M.Obstfeld and K. Rogoff, Foundations of International Macroeconomics, MIT Press. (1997) T.Persson and G. Tabellini, Political Economics Explaining Economic Policy, MIT Press. (2002) I. Rajaraman, A Sustainable Fiscal Policy for India An International Perspective, Oxford University Press. 8 (2006) M. Govinda Rao, Journal of Asian Economics. 16, 993 (2005) L. Serven, Policy Research Working Paper 1613, The World Bank. (1996)

Inequality, Public Investment and Deficits in India

443

Endnotes: i Prior to 1991 government borrowing was accommodated through hikes in the statutory liquidity ratio (SLR) that was imposed on commercial banks. This enabled the placement of government securities at sub-market rates. By 1990 SLR was 39 per cent of bank assets. ii Transfer expenditure in this paper is a generic expression that includes not just those expenditures classified as transfers in an economic classification of government budgets but also implicit transfersii. Implicit transfers include for example the rent component of public sector wages. Public spending on wages of unproductive and surplus employees is akin to a transfer rather than government consumption expenditure. Similarly, not charging user fees on many public utilities such as electricity, water, etc., is a substantial form of transfer that does not necessarily get reported as an explicit subsidy or transfer payment in the accounts of the government. Directed credit programmes stipulating how much of a banks’ portfolio goes into lending say to agriculture, also constitute a form of implicit transfers. iii Note that output can only be produced if there is complementary public capital expenditure as F ( K 1 ,0 ) = 0 in (14). Hence, we presume that creditor behaviour is separable from tax behaviour and a part of the received endowment in period 1 is lent to the government to enable complementary public expenditure. Otherwise with borrowings repaid by tax financing and the government budget inter-temporally balanced we would obtain Ricardian consequences. iv We follow Obstfeld and Rogoff (1997, p.382) who assume that the marginal product of capital is approximately constant over the small scale on which the country can invest.

This page intentionally left blank

Environment

This page intentionally left blank

CLIMATE CHANGE AND THE KYOTO PROTOCOL PARKASH CHANDER Department of Economics, National University of Singapore, Singapore E-mail: [email protected]

1. Introduction This paper interprets the Kyoto Protocol in terms of game theory. Calling upon both positive and normative economics, it analyzes the issues at stake in the current international negotiations on climate change. The negotiations on climate changea, that have been taking place since the late 1980's within the United Nations institutions, are obviously a worldwide process, judging by the length of the list of the participating countries.b But these negotiations, prior to the Kyoto meeting, had led only to a "framework convention", signed in 1992 in Rio de Janeiro, that was little more than a declaration of intent. The real issue then was: are the continuing negotiations eventually going to lead to a sustainable agreement bearing on effective actions that is also worldwide? Or will they lead to a breaking up of the countries into separate blocks, each acting to the best of its own interests? The Kyoto Protocol, signed in December 1997, has been a major development in the post-Rio evolution of these negotiations. Its importance lies mainly in the fact that it requires some countries to take effective actions that would become binding on them once they ratify it. The Protocol does not require all countries to take specific actions. As our summary presentation reports more in details below, commitments to quantified emission reduction or limitation are mentioned only for the so-called "Annex-1" partiesc. The role of the other countries in the agreement, while not ignored, is much less precisely specified.

a

b c

For a thorough account of the scientific evidence on the state of the problem, the reader is referred to the work of the Intergovernmental Panel on Climate Change (IPCC), and in particular to the contribution of its Working Group III (see under IPCC 1995 in our references below). 178 in Rio, 159 in Kyoto and 161 in Buenos Aires. "Annex-1" (to the Rio Convention text) countries are the OECD countries, the former Soviet Union countries, and the Eastern European economies in transition. 447

Parkash Chander

448

One natural question that arises is whether the Kyoto Protocol is to be considered as just an "Annex-I" Protocol; or is it to be seen, after further thought and beyond the appearances, as a worldwide Protocol? Below, we defend the second thesis in terms of game theory. 2.

Main features of the Protocol

We briefly note the main features of the Protocold that are important from the point of view of our analysis: 1. 2. 3.

4.

The Protocol proposes dated emission quotas, expressed in percentages of 1990 emissions, for Annex-I countries, to be met around 2012. It proposes the principles of (a) emission trading by countries (or by their entities) and of (b) joint implementation by Annex-I countries. It proposes a clean development mechanism (CDM) as a way to involve non-Annex-1 countries (especially developing ones) in some particular form of joint implementation and emission trading. It allows trade in emissions only among those countries which ratify the Protocol. It is also expected that trade in emissions will not be allowed with countries that may not fulfil their obligations under the Protocol.

We may also note some of the features that the Protocol does not have: 1.

2.

3.

d

The Protocol does not set targets in terms of the accumulated stock of greenhouse gases. Its object is not a trajectory of stock of greenhouse gases, but it is emission flows per year from some point of time onwards. No explicit emissions ceilings have been proposed for non-Annex-1 countries and such ceilings, if at all, have to be negotiated in future rounds. While the parties to the Protocol are expected to enforce the commitments made by them within their own countries, no sanctions are specified if a ratifying country does not fulfill its obligations under the Protocol, except for the above provision on being excluded from

In Kyoto, the text of the protocol was adopted unanimously by the delegates of the 159 countries that participated in the negotiations. Signing of the text by governments and ratification by parliaments are the following stages of the process. The US, under the Bush administration, subsequently refused to ratify the protocol.

Climate Change and the Kyoto Protocol

449

emission trading. A compliance regime, including possible sanctions for non-compliance, is yet to be specified in the course of future negotiations. 3. Economics of the issues at stake Consider the n countries of the world (indexed by i = 1,…, n) each of which enjoys an aggregate consumption level xi , equal to the aggregate value of its production activities yi minus damages Di which consist of lost production due to global pollution.e The production activities of country i are described most simply by an increasing and strictly concave production function yi = g i (ei ), where ei is the fossil fuel energy input. Assume that the units have been so defined that a unit of fossil fuel use generates a unit of emissions as a by product. The emissions of country i are thus equal to ei . Accordingly, g i′ (ei )(= dg i (ei ) / dei ) is the marginal product of fossil fuel energy or the marginal cost of abatement, depending on the context. Damages in each country depend on the total emissions of all countries, i.e., on ∑in=1 ei . They are represented by an increasing damage cost function Di = d i (∑nj=1 e j ), which for simplicity is taken to be linear.f Each country's net output is thus given by the expression n

xi = g i (ei ) − d i ∑ e j , j =1

(1)

where d i > 0 is the damage per unit of emissions or, equivalently, the benefit per unit of abatement of country i. 3.1. The world optimality

Ignoring distributional issues, the optimal world consumption is equal to the maximum of ∑in=1 xi with respect to the n variables e1 ,…, en . Let (e1* , …, en* ) be the vector of emissions of the n countries that achieve such a world optimum. These are obtained as a solution to the first order conditions for a maximum, i.e.,

e

f

Several studies give estimates of these damage costs (see e.g. Fankhauser9and Nordhaus and Yang12). According to some estimates, damages as a percentage of GNP from a hypothetical doubling of CO2 concentration for developing countries are substantially larger than for developed countries. The main reasons for the high estimates for developing countries are health impacts and the high proportion of global wetlands found in these countries. Numerical estimates of damages in some regions of the world are given in Table 1 below.

450

Parkash Chander n

g i′ (ei* ) = ∑ d j , i = 1,…, n. j =1

(2)

Thus, at the world optimum, the marginal abatement cost of each country must be equal to the sum of marginal damages of all countries. Notice that the world efficient emissions are independent of the actual or current emissions of the countries. They depend only on the total marginal damage of the world.g Negotiations on climate change must aim, at least in principle, to achieve the world efficient emissions. This would of course require transfers among the countries so as to balance the costs and benefits of attaining the world efficient emissions. We argue below that the Kyoto Protocol can be seen as a step in this direction and that a sequence of such steps can indeed lead ultimately to the world efficient emissions and optimal consumption. 3.2. Reference emissions

How does a country decide how much to emit? Low emissions imply a low production according to the function g i , whereas high emissions entail high damages according to the function Di. Classical economics reasoning suggests that each country can achieve its domestic optimum by maximizing its consumption level xi with respect to ei as defined in (1), taking as given all variables e j with j ≠ i. If all countries adopt such behaviour, a Nash equilibrium between countries would prevail. This is given by the vector of emissions (e1 ,…en ) such thath g i′ (ei ) = d i , i = 1,…, n.

(3)

We note two characteristics of this Nash equilibrium: (i) the equilibrium emissions (e1 ,…, en ) are clearly not equal to the world efficient emissions, as can be seen by comparing (2) and (3), and (ii) ei > ei* for each i, since g i is concave and ∑nj=1 d j > d i for each i. Thus, the world efficient emissions are lower than those prevailing at the non-cooperative Nash equilibrium. However, there is little empirical evidence to support that the countries do indeed decide their emission levels in this rational manner. Fulfilment of conditions (3) that characterize the Nash equilibrium requires domestic policies

g

h

However, the production functions, g i , may change over time. Consequently, the world efficient emission levels may also change even if damages remain unchanged. Uniqueness of this vector is ensured under our assumptions of concavity of the functions g i and linearity of the functions Di .

Climate Change and the Kyoto Protocol

451

that involve either an energy tax or appropriately priced pollution permits such that the energy price including the tax or the permit price is equal to the domestic marginal damage cost d i . Such domestic policies are often called “no regrets policies”. If the firms in a country have strong lobbying power, they may be able to influence their government to keep the energy prices low. Since profit maximization by firms implies equality between the marginal product and the price of energy, this will lead to emissions ei which are higher than ei and such that g i′ (ei ) < d i , thus preventing the nationally rational policy from being adopted. If the firms and the government in each country behave in this manner, a different equilibrium - also non-cooperative in nature -would result, called the "market solution" by Nordhaus and Yang 12 or "business-as-usual" by others. Another reason why a nationally rational policy may not come about is that firms in a country may simply not be profit maximizers, as is the case with large public sector enterprises in some non-market economies. In such cases, the domestic equilibria are neither of the "business-as-usual" nor of the "nationally rational" type, and energy prices do not induce any well defined emission policy — except for a generally low concern for efficient use of energy. In sum, at least three types of country behaviour are possible. But whatever be a country’s behaviour if its firms maximize profits and markets are competitive, its marginal abatement cost must be equal to the (average) domestic fossil fuel price in real terms. Given the concavity of the production function g i , it follows that the higher the domestic fossil fuel price, the higher the marginal cost of abatement. As seen from Table 1 below such a relationship indeed holds (except in case of China, where, as is known, firms do not necessarily maximize profits).i In particular, the energy prices in the US are systematically lower and so is the marginal abatement cost. Moreover, for the three market economies of the US, the EU, and Japan, the higher the energy prices, the higher the marginal abatement costs.j For the other countries, we cannot say much, not only because of lack of data but also because they are either non-market or less developed economies, or both. The marginal abatement cost of the US is low compared to that of the EU or Japan, it is next only to that of China, and significantly below that of India. Since the marginal damage cost of the US, which is the largest economy in the world, cannot be lower than, say that of the EU, this suggests that the US i

j

The marginal cost of abatement may seem exceptionally high in case of Japan, but this is because of its large dependence on natural gas, pric of which is relatively high, and less on coal and oil. Coal in Japan is a noticeable exception; but its use there is considerably lower.

Parkash Chander

452

emissions are determined by the "business-as-usual" policy rather than by optimization at the national level.k On the other hand, domestic oil prices are kept high in India by imposing import tariffs not out of concern for the environment but to avoid an adverse balance of payment. The last column of Table 1 presents an educated guess about the type of domestic equilibrium that is likely to be prevailing in each country/region. Table 1 – Retail prices (in US$ per unit) of industrial fossil fuels, marginal abatement cost and damage cost in selected countries or regions Annual damage cost as % of GDP***

Type of domestic equilibrium Conjectured

Heavy fuel oil for industry* (per ton)

Steam coal for industry* (per ton)

Natural gas for industry* (per 10kcalGV)

Marginal abatement cost/ton, for first 100 M ton reduction**

US

138.00

35.27

136.62

$ 12

1.3

g i '(ei ) = p i < d i

EU

187.4

76.0

182.0

$ 40

1.4

g i '(ei ) = p i ≥ d i

Japan

172.86

49.90

423.12

$ 350

1.4

g i '(ei ) = p i ≥ d i

India

191.15

19.36

Na

$ 22

Na

?

FSU

Na

Na

Na

$ 22

0.7

?

China

150.60

30.12

Na

$ 3.5

4.7

?

*Source: Energy Prices and Taxes 1996 **Source: Ellerman and Decaux7 ***Source: Fankhauser 9

4. A world treaty in the making

Let (e1 ,…, en ) be some vector of reference emissions. They may be the Nash equilibrium or the business-as-usual emissions. Or worse, they may be the outcome of a generally low concern for the efficient use of energy. In either case, the reference emissions are higher than the world efficient emissions. Reducing the emissions from the reference levels to the world efficient levels requires each country i to reduce its emissions by ei − ei* , imposing costs and benefits that are unlikely to be equal across the countries: some may have high abatement cost, i.e., g i (ei ) − g i (ei* ) , and little benefit, i.e., d i ∑nj=1 (e j − e*j ), k

This is clearly a case of government, and not market, failure.

Climate Change and the Kyoto Protocol

453

while others may have low abatement costs and high benefits. Since the emission reductions must be agreed upon voluntarily by all countries, we need a scheme of transfers so as to balance the costs and benefits of reducing emissions. Chander and Tulkens (1995, 1997) indeed propose such a scheme. In the present context, it is defined as Ti ={g i (ei ) − g i (ei* )} −

n di ⎧ n * ⎫ ⎨ ∑ g j (e j ) − ∑ g j (e j )⎬ , i = 1,…, n, j j = = 1 1 ⎭ ∑ dj ⎩ n

(4)

j =1

where Ti > 0 means a receipt by country i , while Ti < 0 means a payment by i. The first expression within the braces on the right is equal to country i ’s total abatement cost, and the second expression within the braces is equal to world’s total abatement cost. The scheme thus requires country i not to bear its own abatement cost g i ( ei ) − g i (ei* ), but to bear instead a damage-weighted proportion, d i / ∑nj=1 d j , of world’s total abatement cost. Clearly, ∑in=1Ti = 0, which ensures a balanced budget if an international agency were established to implement the scheme. Notice the role played by the reference emissions (e1 ,…, en ) in the calculation of the transfers (T1 ,…, Tn ). Chander and Tulkens3, 4 assume the reference emissions to be equal to the Nash equilibrium emissions and show that the scheme enjoys several game theoretic properties. In particular, besides leading to the world efficient emissions, it implies coalitional stability in the sense that not only each country is individually better off, but also each coalition of countries is better off compared to what they would get by adopting any alternative arrangement among themselves in terms of emissions and transfers. But what if the reference emissions are not equal to the Nash equilibrium emissions? In particular, if these are equal to the business-as-usual emissions of the type discussed earlier. It turns out that the game theoretic properties of the scheme are robust with regard to the reference emissions. If (e1 ,…, en ) are equal to the business-as-usual emissions, then the corresponding transfers (T1 ,…, Tn ) have the same game theoretic properties as when they are equal to the Nash equilibrium emissions. This is seen intuitively as follows: (a) the business-asusual emissions are generally higher than the Nash equilibrium emissions, and (b) given (a) the payoff that a coalition can achieve for itself is lower, since the emissions of members not in the coalition are higher. The first row of Table 2 provides an example of a vector of reference emissions. These have been estimated by Ellerman and Decaux 7 on the basis of MIT’s EPPA multi-regional and multi-sector computable general equilibrium

454

Parkash Chander

model of economic activity, energy use and carbon emissions. We use these estimated emission levels in our arguments below and for obvious reasons refer to them as the business-as-usual emissions. 4.1. Competitive emission trading

Unlike the scheme above, the Kyoto Protocol does not propose any transfers among the countries. It only proposes ceilings or caps on the emissions of some countries, and these caps are obviously not equal to the world efficient emissions. Yet, as argued below, the Kyoto Protocol can be interpreted as a scheme of transfers and a step towards reaching the world efficient emissions. To see this, let us redefine the above scheme of transfers in terms of emission quotas and trade. This requires us to first introduce the concept of a “competitive emission trading equilibrium”. A competitive emissions trading equilibrium with respect to emission quotas ( e10 ,…, en0 ) is a vector of emissions ( eˆ1 ,…, eˆn ) and a price γˆ > 0 (expressed in units of the consumption good per unit of emissions) such that for each country i = 1,…, n, eˆi = arg max ( g i (ei ) + γˆ (ei0 − ei ) ) n

n

i =1

i =1

∑ eˆi = ∑ ei0 .

(5) (6)

The first order conditions for maximization imply g i′ (eˆi ) = γˆ, i = 1,…, n. This means that competitive trade in emissions enables the countries to relocate the production and emission activities so as to maximize their total output while keeping their total emissions restricted to ∑in=1 ei0 , since by definition 0 ∑in=1 eˆi = ∑in=1 ei and g ′(eˆi ) = g ′(eˆ j ) for all, i, j = 1,…, n. In a competitive emission trading equilibrium, the countries trade in their “pollution rights” which are equal to their emission quotas ( e10 ,..., en0 ), at a given market price γˆ , and at that price, demand and supply of pollution rights are equal. The amount γˆ (ei0 − eˆi ) represents the value of payment, in units of the consumption good, for the purchase of pollution rights at the world market price γˆ if (ei0 − eˆi ) is negative or receipt from the sale of pollution rights if (ei0 − eˆi ) is positive.

Climate Change and the Kyoto Protocol 455

Parkash Chander

456

Now, define emission quotas (e10 ,…, en0 ) from the world efficient emissions e1* ,..., en* and the reference emissions (e1 ,..., en ) such that for each country i, n

(ei0 − ei* )∑ d j = { g i (ei ) − g i (ei* )} − j =1

n di ⎧ n * ⎫ ⎨ ∑ g j ( e j ) − ∑ g j (e j ) ⎬ . j j = 1 = 1 ⎭ ∑dj ⎩ n

(7)

j =1

The left hand side of this expression is what country i pays (or receives) if it buys (sells) pollution rights in amount (ei0 − ei* ) at price γ * ≡ ∑nj=1 d j . In view of (2), γ * = g i′ (ei* ) = g ′j (e*j ), i, j = 1,…, n. Which means that (e1* ,…, en* ) and γ * are nothing but the competitive emission trading equilibrium relative to the pollution rights (e10 ,..., en0 ) . And the right hand side is equal to the transfer Ti advocated above to achieve world efficiency and a stable agreement. Note that while the world efficient emissions (e1* ,…, en* ), as defined in (2), are independent of the reference emissions (e1 ,…, en ), the pollution rights (en0 ,…, en0 ), as defined in (7), are not. In fact, since the world efficient emissions are independent of the reference emissions and thus fixed, there is a one-to-one correspondence between (e1* ,…, en* ) and (e1 ,…, en ) . This means that if the countries are agreeable to the reference emissions (e1 ,…, en ), then they should also be agreeable to the assignment of pollution rights (e10 ,…, en0 ) and competitive trade in emissions, since by definition these would not only lead to the world efficient emissions (e1* ,…, en* ), but also to transfers that make each country or coalition of countries better-off relative to the reference emissions and consumptions. This shifts the argument from an agreement on pollution rights to an agreement on reference emissions (e1 ,…, en ). However, reaching an agreement on reference emissions might not be easy. This has reference to the following two problems: First, the current Nash or business-as-usual reference emissions (e 1 ,…, en ) that determine the transfers (T1 ,…, Tn ) or equivalently the pollution rights (e10 ,…, en0 ) may be considered unfair, especially by those countries which are in the early stages of their economic development. They currently have comparatively low emissions, while the emissions of developed countries are high. In the future, when they would have developed, the currently developing countries will have higher emissions and they might argue that those should be used as reference emissions instead of the current ones. Thus, the scheme of transfers, while Paretian (everyone is better off) with respect to the current Nash or business-as-usual reference emissions might be considered unsatisfactory by the developing countries. For instance, as seen from the first row of Table 2, India’s estimated reference emissions are nearly one-fourth of those of the US

Climate Change and the Kyoto Protocol

457

and substantially less than one-third of China. Obviously, India is unlikely to accept such low reference emissions compared to those of China and the US. Second, if the reductions in the emissions, i.e. ei − ei* , are very large (as proposed by some countries), they are politically not feasible, at least in the short run. 4.2. The Kyoto Protocol

The Kyoto Protocol can be seen to address both these issues. Since the emissions of developing countries in general and of India and China in particular have not been subjected to ceilings, their emissions will rise as a result of their ongoing economic development and those of the Annex-1 countries will fall as a result of abatements and remain fixed at the levels agreed upon at Kyoto until at least further negotiations take place. With time the emissions of developing countries will become comparable to those of Annex-1 countries – likely to be sooner in case of China than India – and these might be then subjected to ceilings. Furthermore, the Kyoto Protocol only requires relatively small reductions for the immediate future, leaving further reductions for later periods. In other words, the Kyoto Protocol is not inconsistent with the ultimate goal of reaching an agreement on appropriate reference emissions (e1 ,..., en ) in some future round of negotiations. In fact, it can be viewed as a step towards it. For reaching an agreement on reference emissions the countries may have to first agree on adopting some equity principle. The currently considered baselines of business-as-usual or historically grandfathered emissions are clearly problematic. Similarly, the uniform per capita emissions, being advocated by India and China, are also unacceptable: if emissions cannot be grandfathered then by the same logic population size cannot be grandfathered either. A scheme of differential standards of emissions per unit of GDP is more likely to be acceptable, but it does not resolve the problem completely. As all the economies grow and their emissions rise, the standards may have to be revised from time to time and made more stringent. Whatever be the equity principle for determining the pollution rights, it seems unlikely from the figures in the first and second rows of Table 2 that the minimal emission reductions or non-reductions implied by the Kyoto Protocol would be inconsistent with it. This seems to be especially true in case of India, which unlike China has rather low emissions. What this means in policy terms is that the developing countries should not oppose the Kyoto Protocol and leave the issue of pollution rights, on which they have repeatedly insisted, to future

Parkash Chander

458

negotiations. Implementation of the Kyoto Protocol will not only reduce the emissions of Annex-1 countries and thus improve the global environment, but will also strengthen the position of the developing countries in future rounds of negotiations as their emissions will continue to rise as their economies grow and become comparable to those of Annex-1 countries. 5. Alternative forms of emission trading

If each Annex -I country were to meet its Kyoto commitment ei0 on its own, the world output will be equal to ∑in=1 g i (ei0 ), which by definition is less than ∑in=1 g i (eˆi ) , where eˆi ’s are the competitive trading equilibrium emissions, as defined in (5) and (6). In fact, as can be easily seen, competitive emission trading allows the countries of the world to restrict the total world emissions to their aggregate Kyoto commitment e 0 = ∑in=1 ei0 at least cost. Competitive trade in emissions thus enables the countries to reduce the world emissions efficiently. As seen above each country or coalition of countries gains from competitive trade in emissions. However, this does not imply that each country or coalition of countries would be willing to participate in competitive emission trading. For that to be true we must show further that no country or coalition of countries can gain even more by forming a separate bloc and trading emissions only among themselves. An argument based on the theory of market games indeed shows that no coalition of countries can be better off compared to the competitive emission trading equilibrium by forming a separate bloc. Let S ⊂ N be a bloc of countries whose members decide, given their aggregate emission quota ∑i∈S ei0 , to adopt some joint policy of their own such as trading only among themselves or engaging in some other bilateral/ multilateral agreements. The maximum payoff of such a bloc of countries is then given by w( S ) = max ∑ g i (ei ) subject to ∑i∈S ei = ∑i∈S ei0 . i∈S

(8)

This is the maximum total gross output that the countries in bloc S can jointly achieve, given their aggregate emission quota.l Consider again (eˆ1 ,…, eˆn ), the competitive trading equilibrium emissions relative to (ei0 ,…, en0 ). We show that the payoff of members of S under the competitive equilibrium is not lower than their payoff when they form a separate l

We ignore the damages because they remain the same, since the aggregate emission quota is fixed.

Climate Change and the Kyoto Protocol

459

bloc (as defined in (8)). This would establish that no country or coalition of countries will have incentives to form a separate bloc and not participate in competitive emission trading. This is in fact straightforward. Indeed, we only need to show that ∑i∈S g i (eˆi ) ≥ w( S ). Using (5), this is equivalent to 0 ∑i∈S ( g i (eˆi ) + γˆ (ei − eˆi )) ≥ ∑i∈S g i (e~i ) where (e~i )i∈S is the solution to (8). Since 0 ∑i∈S e~i = ∑i∈S ei , we must show that ∑i∈S g i (eˆi ) + γˆ (∑i∈S e~i − ∑i∈S eˆi ) ~ ≥ ∑i∈S g i (ei ). This inequality is true since each g i is concave and γˆ = g i′ (eˆi ) in competitive emission trading equilibrium. Therefore, g i (eˆi ) + γˆ (e~i − eˆi ) ≥ g i (e~i ), ei − eˆi ) is positive or negative. irrespective of whether (~ This leads to the conclusion that no country or coalition of countries will have an incentive to form a separate bloc and not participate in competitive emission trading. Thus, the outcome of competitive trade in emissions among the countries cannot be improved upon by the formation of coalitions of countries, such as separate trading blocs. We are thereby rediscovering — in fact, just applying — a general property of competitive equilibria known as their "core" property, which says that competitive equilibria belong to the core of an appropriately defined cooperative gamem. 5.1. Free trade in emissions

While the Kyoto Protocol allows trade in emissions among the Annex-1 countries, it leaves open the questions of extent and nature of such trading.n Economic and game theoretic considerations can be further called upon to resolve these issues. As to the extent of trading, that is, the number of participants in the trade, the market equilibrium theory generally favours trade among the largest number of economic agents. This is also implied by the previous argument against the formation of separate trading blocs or any other form of “coalitions” that restrict trade. Indeed, it is not to the benefit of any country or group of countries to form a coalition and act independently of the other countries.

m

n

The present game is a pure market game where externalities play no role, since, once the emission quotas are fixed, the public good aspect of the problem disappears. One is left with only the private goods-type problem of allocating the emissions between the countries. Note, however, that this is a game for an economy with production, and not that of the usual pure exchange type. An attempt was made to address this at the Conference of Parties in Buenos Aires in November 1998.

460

Parkash Chander

Thus, it is in world's overall economic interest that non-Annex- 1 countries, whose emissions are not subject to quotas, be nevertheless allowed to participate in the trading process. The clean development mechanism (CDM) contains provisions to that effect. A policy implication of our claim is that this mechanism be designed so as to make it as open as possible to the largest number of countries. The fact that no quotas were assigned to these countries is irrelevant if the full benefits of trade in emissions are to be realized. Similarly, it is irrelevant whether or not a country ratifies the Protocol or has not met its commitment under the Protocol. Excluding a country from trade in emissions on any pretext hurts all. As to the nature of trading, the same body of theory advocates that the institutions governing the trades be designed so as to ensure that they be as competitive as possible — competitiveness meaning here that all participants behave as price takers. It is indeed only for markets with that property that efficiency, coalitional stability and worldwide maximal benefits are established. Regulatory provisions that restrict competitiveness in the emissions trading process are thus to be avoided. Such as, for instance, provisions allowing for market power to be exerted by some traders so as to influence price formation to their advantage, as well as regulatory controls that would impede sufficient price flexibility; or still, as proposed by some, limiting the quantities that can be traded. As is well known, the larger the number of participants, the more competitive the market is likely to be: our argument favouring a large extent of the market is thus also one that favors competitiono. Large numbers are admittedly neither the only way nor a sufficient condition to ensure the competitive character of a market, but they are a powerful factor. Table 2 gives a numerical illustration of the outcome of competitive trade in emissions.p The competitive equilibrium price of emissions γˆ is estimated to be equal to $24.75 per ton in 1985 dollars. Country i is an exporter of emission reductions if ei0 > eˆi and an importer if ei0 < eˆi . . Country i' s gain from emissions trade is equal to γˆ (ei0 − eˆi ) − ( g i (ei0 ) − g i (eˆi )) if it is an exporter and g i (eˆi ) − g i (ei0 ) − γˆ (eˆi − ei0 ) if it is an importer – both are positive, since the price γˆ is equal to the marginal cost of abatement and g i is concave. Exporting country i will not gain from trade if it is paid only its actual cost of o

p

Our argument on the role of markets to achieve coalitional stability is also reinforced by a central result in economic theory (Debreu and Scarf (1963); Edgeworth (1881)) according to which only competitive equilibria are coalitionally stable, if the number of traders is large. Additional details can be found in Ellerman and Decaux 7 who also consider other trading regimes.

Climate Change and the Kyoto Protocol

461

abatement i.e. g i (eˆi ) − g i (eˆi′ ) and all the gains from trade in that case will go to the importing countries. Competitive emission trading thus distributes the gains from trade among the exporters and importers in exactly the same way as it does in case of competitive trade in commodities. Among the developing countries, China turns out to be the single largest exporter of emissions followed by India.q Among the Annex-1 countries, the US turns out to be the single largest importer followed by the EU. All countries gain from emission trading. The gains are substantial for both sides indicating the need for cooperation among the developed and developing countries for institutionalizing such trade. Though, as the numerical example illustrates, all countries gain from trade in emissions, yet for several reasons there might be opposition to such trade from both developed and developing countries alike. 5.2. The clean development mechanism

Since restricting trade in emissions among the Annex-1 countries alone may affect both Annex-1 and non Annex-1 countries, this raises the question how to involve the non-Annex-1 countries in emission trade without having them committed to any emission quotas? r This is difficult, but not impossible.s For example, one can calculate the impact of a tax increase on fossil fuel energy in a developing country and offer to transfer to the developing country an amount which is equal to the market value of the consequent reduction in its emissions. However, the developing countries might fear that participation in any form of trade in emissions will amount to some sort of acceptance of emission quotas on their part. Developing countries like India and China have often expressed the view that the problem of climate change has been created by the industrialized countries and therefore it is these countries which should first reduce their emissions, no matter how, before the developing countries can consider accepting any quotas. The developing countries may not participate in emission q

r

s

There is however a practical difference between Annex-I trading and the modelling of global trading which tries to mimic a perfect CDM which may implicitly impose nominal quotas on nonAnnex-I countries. One colleague has expressed this problem as follows: “… should we allow Mexico to “sell” permits to the US if it is not guaranteed that Mexico will really reduce emissions accordingly?” The recently proposed nuclear agreement between India and the US is a case in point, as it promises cleaner technologies to help India meet its energy needs. What would be the impact of this agreement on India’s emissions and therefore how much emission reductions can the US claim to have imported?

462

Parkash Chander

trade also because the clean development mechanism is often interpreted as a form of trade that distributes the gains from trade entirely to the importing (read Annex-1) country and none to the exporting (read non Annex-1) country.t More specifically, it has been often proposed that rather than paying the exporting developing country i the market value at the competitive price, i.e. γˆ (ei0 − eˆi ), the importing countries may pay only the actual cost of abatement, i.e. g i (ei0 ) − g i (eˆi ), which (given the strict concavity of the function g i (ei ) ) is strictly less than γˆ (ei0 − eˆi ). This form of trade in emissions can be easily given effect by the importing countries by “offering” to cover the cost, and cost alone, of abatement activities in developing countries on a project- by- project basis.u Both developed and developing countries might thus oppose the establishment of trade in emissions, though for entirely different reasons. In fact, the above mentioned positions or perceptions concerning trade in emissions seem to have been behind the deadlock at the negotiations held in Buenos Aires in 1999. Given the above stated problems concerning trade in emissions between the developing and developed countries, the ultimate solution might be to first reach an agreement on the reference emissions that, as shown, can lead to welldefined pollution rights or entitlements for each country.v The Kyoto Protocol is not inconsistent with such a solution and in fact, as noted earlier, it is a step towards it. Regardless of whether or not competitive trade in emissions is established, the developing countries stand to benefit from the implementation of the Kyoto Protocol. If the Annex-1 countries meet their Kyoto commitments, the international prices of fossil fuels will fall which would accelerate economic growth in developing countries.w The energy exporting non Annex -1 countries,

t

It is ironic that the same countries which generally extol the virtues of competitive markets should look for other forms of trading when it suits them. u Though in theory this is not the only possible outcome and institutions can be set up to promote more competitive trading, there is a widespread concern among the developing countries that this will not be the case and a project-by-project approach is more likely to be adopted. v Besides facilitating competitive emissions trade among Annex-1 and non Annex-1 countries which would reduce the burden of Annex-1 countries of meeting their Kyoto commitments, assignment of such pollution rights or entitlements would create stronger incentives for the development and adoption of cleaner technologies even by the non-Annex-1 countries. w In fact, the non-Annex-1 countries would benefit even more if, as some Annex-1 countries have suggested, no trade in emissions is to be established among Annex-1 countries and each country is to meet its Kyoto commitment on its own. This is so because then Annex-1 countries will not have access to the Russian “hot air” and the actual total reductions in emissions of Annex –1 countries will be much larger.

Climate Change and the Kyoto Protocol

463

however, might suffer economic losses because of (a) less revenue from energy exports and (b) higher prices of energy-intensive exports from Annex-1 regions. As shown by Babiker, Reilly and Jacoby (2000), other non Annex-1 countries such as India and China with a different mix of imports and exports might be better off. In sum, future negotiations on climate change should aim at reaching an agreement on reference emissions. Such an agreement on reference emissions, as shown, will lead to an agreement on pollution rights and facilitate competitive trade in emissions. The Kyoto Protocol is not inconsistent with such an objective and, in fact, it can be viewed as a step towards it. References

1. M. Babiker, John. M. Reilly. and Henry. D. Jacoby, Energy Policy 28, 525 (2000). 2. P. Chander, and H. Tulkens, European Economic Review, 36(2/3), 288 (April, 1992). 3. P. Chander and H. Tulkens, International Tax and Public Finance, 2(2), 279 (1995). 4. P. Chander and H. Tulkens, International Journal of Game Theory, 26, 379 (1997). 5. P. Chander, H Tulkens, J-P Ypersele Van, and S, Willems, Theory for the Environment, Festschrift in honor of Karl-Göran Mäler, (eds.) P. Dasgupta, B. Kriström and K. -G. Löfgran, Edward Elgar, 98(2002). 6. P. Chander, International Journal of Game Theory, 35, 539 (2007). 7. A.D. Ellerman, and A. Decaux, “Analysis of post-Kyoto CO 2 emissions trading using marginal abatement curves”, MIT Joint Program on the Science and Policy of Global Change, Report No. 40, Massachusetts Institute of Technology (October, 1998). 8. Energy Prices and Taxes, (1998). 9. S, Fankhauser, Valuing Climate Change: The Economics of Greenhouse, Earthscan Publications, London (1995). 10. M. Germain, P.H Toint,, H Tulkens and A. de Zeeuw, , Transfers to sustain cooperation in international stock pollutant control (1998). 11. Intergovernmental Panel on Climate Change (IPCC), Climate Change 1995: The Economic and Social Dimensions of Climate Change, Contribution of Working Group III to the Second Assessment Report of the Intergovernmental Panel on Climate Change, Cambridge University Press (1995). 12. W.D. Nordhaus, and Z.Yang, American Economic Review, 86(4), 741 (1996).

This page intentionally left blank

Finance

This page intentionally left blank

RECENT TRENDS IN MICROFINANCE INSTITUTIONS: SOME THEORETICAL IMPLICATIONS SUMAN GHOSH Department of Economics, Florida Atlantic University Email: [email protected] ERIC VAN TASSEL Department of Economics, Florida Atlantic University 777, Glades Road, Boca Raton FL 33431 In this paper, we offer a theoretical formalization of the phenomenon known as mission drift. In recent years, there have been claims that the entry of large donors with deep pockets have led to a mission drift phenomenon whereby microfinance institutions (MFI) who were previously catering to the poorest agents have drifted towards catering to the ‘better off’ poor. Our aim in this chapter would be two-fold: one is to trace very briefly some of the current ongoing debates and the concomitant empirical work that sheds light on those debates. Second, we want to highlight a few of the theoretical ramifications that arise as a result of the most recent studies and debates.

1. Introduction Credit constraints have been widely acknowledged as a primary impediment to the development process and are now a part of conventional thinking of development economists. A great deal of effort has been directed towards closing perceived funding gaps. Such efforts are manifested in a broad range of activities, from subsidizing small business loans to reforming the legal and judicial structures governing credit transactions. One approach in particular that has received a growing amount of attention in academic and policy circles alike is the role of microfinance.* This movement aims to extend small amounts of capital to poor borrowers throughout the world, typically to facilitate income generating self-employment activities. Inspired by the remarkable achievements of institutions such as the Grameen Bank in Bangladesh and the Bank for Agriculture (BAAC) in

*

See Aghion and Morduch (2005) and Morduch (1999b) for a survey of microfinance institutions. 467

468

Suman Ghosh and Eric Van Tassel

Thailand, the microfinance model has been studied and replicated throughout the developing world.† Between 1997 and 2005 the number of microfinance institutions increased from 618 to 3,133 and the number of people who received credit from these institutions rose from 13.5 million to 113.3 million (DaleyHarris, 2006). Given the overwhelming spread of this phenomenon the UN declared 2005 as the International Year of Microcredit. In this context, it is not surprising that academic research on this topic mushroomed during this period. As a result of the pioneering work by Stiglitz (1990), Besley and Coate (1995), Ghatak (1999), Ghatak and Guinnane (1999) and others, we now have a much better understanding of issues pertaining to issues such as group loans and joint liability. What is rather striking about all this is that as the enthusiasm about microfinance grows and the model continues to spread throughout the world, a casual survey of some of the most recent work from top scholars in this field has titles such as: “Can Micro-credit Bring Development?, Is Microfinance too Rigid?, Does Microfinance Really Help the Poor?”.‡ It is in this context that we find motivation for our current write-up. The recent critiques of microcredit suggest that perhaps the microfinance revolution is falling short of initial expectations. Our aim in this paper would be twofold: one is to trace very briefly some of the current ongoing debates and the concomitant empirical work that sheds light on those debates. Second, we want to highlight a few of the theoretical ramifications that arise as a result of the most recent studies and debates. Our objective is modest in that we explore a few leads on how some rather simple theoretical models can be extended in order to encompass the issues at hand. 2. Mission Drift in Microfinance Institutions? A current debate amongst microfinance practitioners centers around the question whether to pursue sustainability or to maintain an allegiance to the poverty minimization mission with which the microfinance revolution started. The debate is summarized in Robinson (2001), as the financial systems approach and the poverty lending approach. Basically the point is whether there is a necessary trade-off between being commercially viable and maximizing poverty reduction. ACCION International, a network based in Boston, that provides support for

† ‡

For discussion on the Grameen Bank model, see Yunus (2003). “Can Micro-credit Bring Development?” By Ahlin and Jiang, “Is Microfinance too Rigid?” By Karlan and Mullainathan and “Does Microfinance Really Help the Poor?” by Morduch.

Recent Trends in Microfinance Institutions: Some Theoretical Implications

469

microlenders in the United States and Latin America has been highly influential in shaping donors’ minds over the past decade, advocating subsidization for start-up costs only and pushing hard for commercial orientation. The view here is that the best hope to reach the greatest number of poor and near-poor households is to get access to commercial capital in amounts that are only possible if institutions transform themselves into fully chartered banks. There is mixed academic evidence regarding the success of such an agenda. Studies such as Hulme and Mosley (1996) and Copestake et.al (2005) indicate that it is the better off poor rather than the starkly poor who stand to benefit the most from the wave of commercialization. Other studies such as Khandker (2005) seem to point in the other direction. The study by Cull et.al (2007) is the only convincing study which explores the issue as to whether more profitability is associated with a lower depth of outreach to the poor and whether there is a deliberate shift away from serving poor clients to wealthier clients in order to achieve financial sustainability. This phenomenon has also been termed as “mission-drift” and is exemplified in a pioneering debate between Grameen Bank founder Mohammad Yunus and the billionaire turned microfinance expert, Pierre Omidyar.§ In their study, Cull et.al disaggregate their data set by lending type, namely, joint liability contracts, village banking and individual based lending. They find that institutions that make smaller loan sizes are not necessarily less profitable. But they do find evidence that larger loan sizes are associated with lower average costs for both individual-based lenders and group lenders. Since larger loan size is often taken to imply less outreach to the poor, their results do indeed have negative implications. For individual-based lenders, the pattern of results they find are consistent with disincentives for depth of outreach- i.e., the personnel expenses devoted to identifying borrowers worthy of larger loans could deter institutions from serving the poorest segments of society. Regarding mission drift they find that larger individual based lenders and group-based lenders tend to extend larger loans and lend less frequently to women. Older individual-based lenders also do worse on outreach measures than younger ones. While this is not evidence of mission drift in the strict sense (i.e., that pursuit of improved financial performance reduces focus on the poor), the results for larger and older microbanks are consistent with the idea that as institutions mature and grow, they focus increasingly on clients that can absorb larger loans.

§

See October 30, 2006 article “Millions for Millions” in The New Yorker.

470

Suman Ghosh and Eric Van Tassel

We believe that the theoretical literature on microfinance has failed to adequately address such issues and hence to either explain or predict the analytics that might govern such phenomenon. The literature has talked about the ‘better off poor’ and the ‘starkly poor’ though the objective function adopted to model the MFIs is generally the standard profit maximization or client maximization such as in the recent study by McIntosh and Wydick (2005). One weakness of these objective functions is that they fail to distinguish between poor agents in terms of depth of poverty. A poor person who is just below the poverty line is treated equivalently as one who is much further below the poverty line. This is important because a number of recent policy debates and empirical studies emphasize that depth of poverty is a critical factor in determining the impact of microfinance. With this in mind, a natural first step to analyzing the impact of MFI lending would be to adopt a different sort of metric, such a weighted poverty gap. While it may be arguable exactly what a particular MFI’s mission might be, as analysts we can use the weighted poverty gap to scrutinize the impact the MFI’s lending policies have on poverty. In doing so we might find that in certain contexts, the MFI is behaving as if it is trying to minimize a poverty gap with a specific weight. If one was to interpret mission drift as a reduction in the weight on the poverty gap, then one way to pin down whether an MFI is drifting is to establish a link between optimal lending policy and the specific weight on the poverty gap. In the next section of the paper we explore some of these ideas by developing a simple model of subsidized MFI lending. We assign the MFI a poverty gap as an objective function and look at the impact of different lending policies. In Section 2.1.1 we assume agents hold the same initial wealth levels and in Section 2.1.2 we introduce heterogeneity in wealth. 2.1. The Model Consider a one period setting with a population of N agents. Each agent has access to a risky production project that requires a $1 investment of capital. If the project is successful it generates revenue R > 1 and if unsuccessful the revenue is 0. Assume that agent j has a probability of project success of p j . Agent j is endowed with an initial wealth w j and we assume w j < 1 for all j . To run the production project agent j requires a loan of exactly L j = 1 − w j . Loans are provided by lenders. Loan contracts are limited in liability and characterized by loan size L j and interest rate r j . The agents have access to either a monopolist moneylender or a MFI for the loan. In particular, there are private banks and a single microfinance

Recent Trends in Microfinance Institutions: Some Theoretical Implications

471

institution (MFI). The bank’s objective is to maximize profit. The general objective of the MFI is to reduce poverty through the distribution of production loans. A lending policy of the MFI describes a subset M of the population and for each j ∈ M , a specific loan contract ( rt j , Ltj ) . The objective of the MFI in period t is to select a lending policy to minimize poverty. To define poverty in the economy, we use an exogenous poverty line y > 0 . Given a profile of strategies for all lenders and agents, we can calculate an equilibrium expected income y tj for agent j in period t , for all j ∈ N .** The (expected) poverty level in equilibrium is then defined as

∑ ( y − y tj )

α

, where α ≥ 1 .††

j

yt < y

We assume that the objective of the MFI is to select a lending policy to minimize

∑ ( y − y tj )

α

.

y tj < y

As is well known, the weight α on the poverty gap determines the degree to which the measure depends on relative inequality below the poverty line. As α rises, the poorest of the poor have the dominant influence on the level of poverty in the economy. As α → ∞ , it is fairly clear the MFI’s lending policy will be to devote all of its resources to getting loans to the absolute poorest of the poor. On the other hand, when α =1 the lending policy may look very different. To start, we consider weight of α = 1 . In the model we assume that lenders have access to an unlimited supply of capital at the interest rate i ≥ 0 . However, unlike the banks, the MFI is endowed with a subsidy Z ≥ 0 that it can use to help finance its lending operations. In distributing loans, the MFI is restricted to offering loans contracts that earn the MFI a non-negative expected profit taking into account the subsidy Z . 2.1.1. Identical Wealth Endowments In this section we consider a case where each agent holds an identical wealth endowment. To simplify the analysis we assume this endowment is zero. We also assume that each agent has a probability of success of either p h or p l ,

** ††

Expected income does not include disutility of effort. This weighted poverty gap belongs to the class of measures from Foster, Greer and Thorbecke (1984).

472

Suman Ghosh and Eric Van Tassel

where p h > p l . In the population fraction λ of the agents are type p h . Lenders cannot observe agents’ individual types but do observe the distribution of types, λ . For a given set of M agents, to cover costs, the commercial bank can afford to offer a “competitive” interest rate as low as r c where (λ p h + (1 − λ ) p l )(1 + r c ) − (1 + i ) = 0

(1)

Note that as long as 1 + r c ≤ R , bank lending is profitable, otherwise it is not. Given a poverty line y , we can then calculate an interest rate r such that R − (1 + r ) = y . At this interest rate, if the agent has a successful project he will be classified as non-poor. If it happens to be the case that r c ≤ r , the competitive interest rate offered by banks leaves successful agents with a payoff such they are classified as non-poor. In this event, given the objective function we have assigned to the MFI, the MFI has no role to play in the credit market. If on the other hand, r c > r then the MFI faces an opportunity to reduce poverty through subsidized lending. The question facing the MFI is exactly how to do this. One possibility is for the MFI to charge r on every loan it issues. Obviously each agent then prefers a loan from the MFI over a bank. On each loan the MFI issues, the MFI must allocate a subsidy of z = −(λ p h + (1 − λ ) p l )(1 + r ) + (1 + i )

(2)

Note that as long as r c > r , the subsidy z > 0 . Since the total available subsidy is Z , the MFI can afford exactly X loans where zX = Z . Then it must be the case that either X ≥ M or X < M . When the prior holds, subsidy is not scarce and the optimal lending policy for the MFI is to simply issue loans to the entire population at the interest rate r . When X < M , the MFI faces a choice. One option is to charge the interest rate r on each loan to a subset X of the population M . On each of these loans the MFI offers a rate such that if successful, the agent will no longer be poor. We refer to this lending strategy as maximizing impact of outreach.‡‡ The alternative option facing the MFI is to charge a higher interest rate on each loan, thus decreasing the subsidy per loan, but increasing the number of loans issued. We say that when the MFI uses its available subsidy to maximize the number of loans issued, the MFI is maximizing outreach.

‡‡

Note that given the objective function, there is no reason for the MFI to charge a rate below r .

Recent Trends in Microfinance Institutions: Some Theoretical Implications

473

To identify the optimal lending strategy for the MFI we can explicitly measure the poverty level associated with each strategy. For α = 1 we get two “unweighted” poverty gaps that we can then compare. Result 1: If bank lending is unprofitable, the MFI should maximize impact of outreach in order to minimize the unweighted poverty gap, otherwise the MFI can optimally pursue either lending strategy. Proof. See appendix. In a market where banks cannot afford to offer pooling contracts, we find that the optimal lending policy for the MFI is maximize the impact of outreach by offering the interest rate r on each loan. The intuition for this finding is as follows. On every new loan the MFI issues, the MFI expects a loss that must be covered by a subsidy. This is true even when the MFI charges an interest rate that extracts the entire revenue from the agent. On each loan issued there is a positive amount of subsidy that is used and these subsidy “dollars” have absolutely no impact on poverty reduction. Hence, if the MFI is trying to minimize the sum of unweighted poverty gaps across the population, the MFI should minimize the number of clients by charging r on each loan. For the sake of illustration, suppose the MFI is offering only one loan. If the MFI obtains a very small amount of extra subsidy the MFI can use it to reduce the interest rate on the existing loan or offer a new loan to a second agent. Since a new loan will not reduce poverty, the MFI should optimally use the extra subsidy to reduce the rate on the existing loan. What this suggests is that when credit markets are unprofitable, there is an efficiency loss associated with maximizing outreach. Every time a new loan is issued there is a specific amount of subsidy that is spent which has absolutely no impact on poverty. If these dollars spent on creating new loans were instead used to lower the rates on existing loans, poverty would be lower. Clearly this discussion depends critically on the assumption that α = 1 . When α is higher the MFI may find it is worthwhile to utilize subsidy to increase the number of loans. These findings might be useful in interpreting MFI lending policies in the following manner. For example, an MFI pursuing a policy of maximizing outreach is effectively trying to minimize a heavily weighted poverty gap. On the other hand, if an MFI appears to have a narrow outreach with heavily subsided loans, the MFI is trying to minimize an unweighted poverty gap. Also keep in mind that when α > 1 , the MFI should

474

Suman Ghosh and Eric Van Tassel

not necessarily pursue the lending strategy of maximizing the number of loans issued. It is important to acknowledge that tradeoff between maximizing the number of clients and the size of the efficiency loss on each loan issued. 2.1.2. Heterogeneous Wealth Endowments In the previous analysis we assumed the agents had identical endowment and hence, we ignored issues regarding depth of poverty alleviation. With that purpose, now we consider a population in which agents have different initial wealth levels. We assume fraction θ of the population N has wealth level w A and fraction 1 − θ has wealth level w B , where 1 > w A > w B ≥ 0 . One implication of this is that agents now require different loan sizes to invest in their projects. Since lenders observe the loan size requested by the agent lenders can observe agents’ wealth ex ante. With this information agents can be partitioned into two groups according to their wealth endowments. To simplify we assume each group has exactly 0.5N agents. Consider the group with wealth w j for j ∈ { A, B} . Assume that in this group fraction λ j are type p h and the remainder are type p l . Since agents hold initial wealth, they are willing to borrow only if the expected profit from borrowing exceeds their wealth endowment. For a loan of size 1 − w j at interest rate r , a type i agent will prefer to borrow rather than not when p i ( R − (1 + r )(1 − w j )) ≥ w j .

(8)

These participation constraints for the agents are sensitive to the interest rate being offered on the loan. For a type i agent the maximum interest rate the agent is willing to borrow at is 1 + rˆi , j =

pi R − w j p i (1 − w j )

(9)

Note that this maximum rate is higher for the type p h agent than for the type p l agent. If a bank lends to the group of agents with wealth w j , the interest rate the bank can afford to offer depends on what types of agent want to borrow. If both types of agents borrow then the bank can afford to offer an interest rate as low as r jc where 1 + r jc =

1+ i λ j p h + (1 − λ j ) p l

(10)

Recent Trends in Microfinance Institutions: Some Theoretical Implications

475

As long as r jc < rˆi , j both types of agents choose to borrow at this pooling rate offered by banks. As we did in the previous section we can compare the interest rate offered by the banks with the rate that eliminates poverty. For agents with wealth level w j , define the interest rate r j as that where R − (1 + r j )(1 − w j ) = y . When r jc > r j the MFI can reduce poverty by using its subsidy to offer a pooling rate lower than the competitive rate being offered by banks. Say the MFI offers a loan size (1 − w j ) at the interest rate r j . Then to cover the expected loss on this loan the MFI requires a per loan subsidy of z j = {−(λ j p h + (1 − λ j ) p l )(1 + r j ) + (1 + i )} (1 − w j )

(11)

Note that the size of this subsidy is dependent on the initial wealth level of the agent as well as the fraction of high quality agents in the group. For a given group of agents with identical wealth the MFI faces lending policy options similar to those described in the previous section of the paper. That is, for a wealth level w j , the MFI can focus on maximizing depth of outreach or simply outreach itself. Once the MFI identifies optimal lending policy for a given wealth level, the next question is which wealth group to focus on. That is, if the MFI has a scarce subsidy Z , should the MFI use the subsidy to finance loans to the group of agents with wl or w h ? We focus our attention on a case where our two wealth groups are being offered pooling rates by banks, namely rAc and rBc . Furthermore, to keep things interesting suppose that each of these rates exceed the poverty elimination rate r j associated with the agent’s wealth level. The first question is what kind of subsidy is required to make a loan to each group. We find the following. Result 2.1: If λ A ≥ λ B , the higher wealth agents require a smaller per loan subsidy than the lower wealth agents. As long as the average quality is higher in the pool of agents with higher wealth,§§ it easily follows that the necessary subsidy to the higher wealth agents is less. Furthermore, when higher wealth agents require a smaller subsidy, we find the following.

§§

This is not an uncommon assumption, for example, see McIntosh and Wydick (2005).

476

Suman Ghosh and Eric Van Tassel

Result 2.2: If λ A ≥ λ B , the MFI will have a greater impact on the poverty gap if the MFI targets its subsidized loans to the agents with higher wealth. Proof. See Appendix. We find that with scarce subsidy, the MFI can minimize an unweighted poverty gap by focusing on the less poor. An important assumption driving this result is that the group of agents with higher wealth has a higher fraction of high quality agents. When this assumption holds, the less poor require a smaller subsidy and the implication is that the MFI can achieve more poverty reduction by concentrating its scarce subsidy on the less poor. The above analysis helps us to think about the issue of mission drift in microfinance institutions. What our results indicate is that given a current level of subsidy, if suppose there is a push for self-sustenance such that there is a gradual removal of subsidies then it is indeed the very poor who will be affected by this trend. The obvious follow up question then remains as to whether subsidies are at all necessary for the successful functioning of MFIs. This issue has been raised in the current literature in recent years which is the subject of our enquiry in the next section. 3. Whither Subsidies? A Theoretical Enquiry Much of the enthusiasm on the microfinance revolution rests on an enticing “win-win” proposition which is basically that eventually the microfinance institutions can eschew subsidies and achieve financial sustainability and hence be able to grow without the constraints imposed by donor budgets. A key tenet is that poor households demand access to credit, not “cheap” credit. This vision has been translated into a series of “best practices” circulated widely by the CGAP,*** the UNDP and other donors. Until recently there has been almost no formal analysis of the validity of these claims and arguments. Morduch (1999a) is the first formal analysis that takes up the question as to whether the continuing subsidies are necessary for the functioning of the microfinance institutions. He shows that in order to become subsidy independent, the Grameen Bank would have needed to increase the lending rates by 75% between 1985 and 1996. Thus he argues that much of the success of microfinance has been crucially dependent on the role of continuing subsidies,

***

The Consultative Group to Assist the Poor (CGAP), is a consortium of 33 public and private development agencies working together to expand access to financial services for the poor.

Recent Trends in Microfinance Institutions: Some Theoretical Implications

477

which tends to contradict the ‘win-win’ proposition made by the proponents of CGAP. In a recent cross country study by Cull et. al (2007), where the financial institutions are united by claiming strong commitments to achieving selfsufficiency, the authors find that the average share of funding made up of subsidy exceeds 20%. The above two papers convincingly show that subsidies are still essential for the functioning of microfinance institutions. Thus what is needed is to enquire into the costs and benefits of subsidies rather than question the validity of subsidies in microfinance programs. Unfortunately there are only few studies that attempt to do this. Townsend and Yaron (2001) for the BAAC in Thailand and Khandker (2005) for the Grameen Bank are some of those. Both of these studies suggest that the social benefits of these microfinance institutions exceed the costs. Now, to effectively study the impact of costs and benefits of subsidies one has to take into account the existing lenders which do not receive subsidies. The tension between the MFIs and the existing private moneylenders has been the subject of controversy amongst microfinance practitioners in recent times.††† In contrast, the theoretical literature on microfinance has not addressed the analytics that govern the interaction between the subsidized MFIs and the private moneylenders. As Morduch (2000) argues “practitioners need to know much more about problems that arise when multiple programs- some subsidized, some not - coexist” (page 626). We believe that such a study would definitely give us a better understanding on the costs and benefits of subsidizing MFIs.‡‡‡ With that objective we sketch a brief model based on Ghosh & Van Tassel (2007) which addresses those issues. 3.1. A Model of Moneylender and Subsidized MFI Interaction Consider a two period model with a large population of N identical risk neutral agents. Every agent owns a risky one period production project in both periods. The project requires an investment of $1 in capital and generates a revenue of R > 1 if successful and 0 if unsuccessful. The probability of project success is p t ( e t ) = e t , where e t ∈ [0,1] denotes the effort level chosen by the agent in

†††

See Aug 19th 2006 article “Microsharks” in The Economist. Jain (1999) study interactions between formal and informal institutions but his question was different from this paper’s. A few papers have pointed out potential perverse effects from introducing subsidized credit. For example Hoff and Stiglitz (1998) show that loss of scale economies in the lending process can lead to an increase in interest rates, and Bose (1998) finds that if borrowers are heterogeneous in risk and lenders are asymmetrically informed, cheap credit may wind up increasing interest rates. To our knowledge, none of these papers look at subsidized credit in a dynamic setting, which is the main focus of our work.

‡‡‡

Suman Ghosh and Eric Van Tassel

478

period t . The agent’s disutility of effort is given as g ( et ) , where as is standard, g ' > 0 and g '' > 0 . An agent’s wealth at the start of period t is denoted wt ≥ 0 . We assume that w1 = 0 for all agents.§§§ Thus, to invest in the first period the agent requires a $1 loan. As before, the agents have access to either a monopolist moneylender or a MFI for the loan. Loan contracts are characterized by limited liability. In particular, a loan contract specifies a loan size Lt and an interest rate rt , where the agent is obligated to repay min {(1 + rt ) Lt , R} in the event of project success and 0 otherwise. Both the moneylender and the MFI has access to an unlimited supply of capital at a fixed interest rate, which we normalize to zero. However, unlike the moneylender, the MFI is endowed with a subsidy Z t at the start of each period. The MFI uses this subsidy by allocating a non-negative portion of the subsidy to each loan contract it issues. In particular we assume that the MFI is constrained to selecting a lending policy such that e tj (1 + rt j ) Ltj − Ltj + Z t j ≥ 0 for all j ∈ M , where Z t j ≥ 0 is the subsidy allocated to loan contract (rt j , Ltj ) and

Z tj ≤ Z t . ∑ j M

At the start of t = 2 , an agent’s wealth w2 is equal to the net

∈

revenue from the previous period. Hence, w2 = R − (1 + r1 )( L1 = 1) if the agent was successful with his first period project and w 2 = 0 otherwise. As stated before, the general objective of the MFI is to reduce poverty through the distribution of production loans. In each period the MFI selects a lending policy. A lending policy describes a subset M of the population and for each j ∈ M , a specific loan contract ( rt j , Ltj ) . The objective of the MFI in period t is to select a lending policy to minimize poverty in the current period. In this analysis, we adopt a measure where α is sufficiently high, though short of a “Rawlsian” measure where α → ∞ . Under this weighting, the MFI’s priority is to raise the expected income of the poorest agents, but can still benefit if the less poor also experience an increase in expected income. The timing of the game is as follows. In the first period lenders offer loan contracts to the agents. Since w1 = 0 all agents have zero self-finance, so the next relevant move is for agents to accept loans and exert effort. After project outcomes are realized, each agent is assigned a second period wealth level w2 . §§§

As argued by Morduch (2000) one of the main obstacles facing the poor in trying to access credit is lack of collateral. Assuming all agents begin with zero wealth captures this and simultaneously allows us to concentrate on the moral hazard problem. We do observe heterogeneous wealth levels in the second period as an endogenous outcome of the model. Introducing heterogeneous wealth endowments in the first period is certainly an interesting direction for future research.

Recent Trends in Microfinance Institutions: Some Theoretical Implications

479

At the start of the second period, both lenders offer loan contracts. Agents select a pair (c 2 , s 2 ) , i.e., a consumption-savings pair for period 2 and then choose a loan contract. After t = 2 project outcomes are realized and the game concludes. We briefly state the two main results that we derive from the above setup.**** Our first result is regarding the period 2 lending behavior of the lending agencies. Next we give the results for both periods combined. Before we present our first result we define the variable θ which can be interpreted as the measure of outreach for a given lending policy, where θ ∈ (0,1] . Thus θ = 1 would be a situation where every borrower gets a loan. Result 3.1: The successful agents from period 1 who can hold sufficient equity in their period 2 projects take a loan from the moneylender. The interest rate r ** that the moneylender charges is a decreasing function of the subsidy level of the MFI. For Z 2 ≤ Z 2 , the MFI offers loans of size L 2 = 1 to all agents with a first period credit history. For Z 2 > Z 2 , the MFI offers loans of size L 2 = 1 to all agents with a first period credit history until θ = θ , after which it uses additional subsidy to lend to agents with no credit history.

As the MFI’s outreach θ increases, this puts downward pressure on the interest rate the moneylender must charge to attract borrowers. This pressure originates from the choice of the MFI to make his second period loans available to basically anyone who does not get a loan from the moneylender. The availability of the MFI loans gives the successful agents an option to consume their first period net-earnings and take out an MFI loan. This option raises the bargaining power of the successful agents when they negotiate with the moneylender. To persuade agent to re-invest their net-earnings, the moneylender is forced to lower the interest rate he charges. As long as this rate is not too low, the moneylender can still make a profit. In this scenario agents with positive wealth are catered to by the moneylender and agents with no wealth are issued subsidized loans by the MFI. Note that the lending policy not only raises the expected incomes of agents with zero wealth but also those agents borrowing from the moneylender, which is relevant to the MFI as long as the expected incomes of these agents are below the poverty line. In the above result, the MFI’s lending policy does not assume that the MFI observes the exact credit history of the borrowers, i.e., whether success or

****

For detailed proofs of the results, see Ghosh-Van Tassel (2007).

480

Suman Ghosh and Eric Van Tassel

failure. It is entirely based on the fact whether borrowers have a credit history or not. The reason the MFI targets new clients with subsidy dollars in excess of Z 2 is to keep its lending policy compatible with the profitable operations of the moneylender. If the MFI deviates by using subsidy to achieve an outreach exceeding θ , the MFI winds up effectively undercutting the moneylender. In doing so, the MFI lending policy encourages at least some agents with positive wealth to stop self-financing their projects and instead, to consume their wealth and apply for subsidized loans. This generates a situation where the MFI is using subsidy dollars to ration loans to a set of agents, of which now some are agents that otherwise could be earning income by accessing loans from the private market. Such an outcome is inconsistent with the priority of the MFI to focus on the poorest of the poor. Result 3.2: Under certain standard conditions there exists a Z 2 such that for all Z 2 < Z 2 , the moneylender’s two period expected payoff is higher when an MFI is present than when no MFI is present.

This result offers some interesting insight on the question of how a monopolistic moneylender might respond to the entry of a subsidized MFI. In a setting where poor agents lack sufficient equity to make lending profitable, a moneylender faces a negative return on his initial distribution of loans. If an MFI enters and offers first period loans to these agents this clearly represents an increase in the moneylender’s first period profit, in that it minimize first period losses. In the process of lending to the poor in the first period, the MFI creates a subset of profitable agents that the moneylender can then cater to in the second period. Under this scenario the moneylender takes in a positive profit in the second period without having to incur the concomitant loss in the first period. Thus we can have a range of second period subsidy levels for which the moneylender earns a higher expected profit from having an active MFI in the market. In a sense, one might argue that it is the donor’s subsidy level which determines the harmonious relationship between the monopolistic moneylender and the MFI. This is relevant given some of the recent controversies over the tension between the interactions of the MFI and private lenders.†††† We also calculate a critical upper bound on the subsidy level, above which excessive subsidies will trigger a decrease in the incentives for agents to work hard and

††††

See Aug 19th 2006 article “Microsharks” in The Economist.

Recent Trends in Microfinance Institutions: Some Theoretical Implications

481

save income. This issue is relevant to Morduch’s (1999b) argument, in which he claims that excessive subsidies have led to the failure of many previous microfinance programs. Our paper provides one explanation for why such an assertion could be true. By limiting the subsidy levels available to the MFI, the behavior of moneylender and the MFI are complementary, in that the loans provided by the money lender work in fashion with the MFI’s mission of poverty alleviation. 4. Conclusion

It appears that the success of microfinance institutions over the last decade or so has come with new challenges. In this paper we highlight some of the ongoing debates about microfinance and look at several implications for theoretical study. This paper addresses two key issues. One is from the theoretical standpoint of how to model the concept of “mission-drift”? We believe that the existing theoretical models have yet to make much progress on this important policy issue. An aim of our paper is to examine how mission drift might be studied using a formal framework. In our examination of this topic we stress the importance of adopting a metric such as the weighted poverty gap rather than an ad hoc MFI objective function. Secondly, though there is a rapidly growing empirical literature on the questions regarding costs and benefits of subsidizing microfinance in the credit market, there is still much less theoretical work. In the third section of our paper we offer some discussion on a study of the interaction between subsidized microfinance and a profit maximizing money lender. Acknowledgments

We would like to thank Parikshit Ghosh and Alexander Karaivanov for comments and suggestions. We would also like to thank conference participants at the 2007 Econometric Society Summer Meetings at Duke University, the 4th Annual Conference on Economic Growth and Development at the Indian Statistical Institute Delhi and the 2007 Missouri Economics Conference. Appendix Proof of Result 1. Consider the case where the MFI maximizes impact of outreach. The MFI issues loans to X agents, where X < M and charges interest rate r on each loan. The rest of the group, M − X either borrows from a bank if r c ≤ R − 1 , otherwise these agents earn 0. Under this lending policy

482

Suman Ghosh and Eric Van Tassel

the poverty gap at the end of the period is X (λ p h + (1 − λ ) p l )( y − [ R − (1 + r )]) + X (λ (1 − p h ) + (1 − λ )(1 − p l )) y + ( M − X )( y − q )

(3)

where q = (λ p h + (1 − λ ) p l ) ⎡⎣ R − (1 + r c ) ⎤⎦ if R − (1 + r c ) ≥ 0 and q = 0 otherwise. The alternative for the MFI to decrease the subsidy per loan. In doing so, the MFI increases the interest rate and raises the number of borrowers. Say the MFI increases the rate on each loan so that he can raise the number of borrowers to X + 1 . To obtain X + 1 clients, the MFI allocates a per loan subsidy of z = Z ( X + 1) −1 and charges r where (λ p h + (1 − λ ) p l )(1 + r ) − (1 + i ) + z = 0

(4)

Under this lending policy the poverty gap at the end of the period is ( X + 1)(λ p h + (1 − λ ) p l )( y − [ R − (1 + r )]) + ( X + 1)(λ (1 − p h ) + (1 − λ )(1 − p l )) y + ( M − X − 1)( y − q )

(5)

We can then subtract (3) from (5) to get the marginal impact on the poverty gap from increasing outreach to a higher number of borrowers. This gives us X (λ p h + (1 − λ ) p l ) [ r − r ] − (λ p h + (1 − λ ) p l ) [ R − (1 + r ) ] + q

(6)

In general, the sign of this marginal impact depends on whether banks are providing loans in addition to the MFI. If R − (1 + r c ) ≥ 0 , so that bank lending is profitable, the marginal impact of raising MFI outreach is zero. That is, if banks can afford to offer pooling contracts to the population then it is not important whether the MFI maximize outreach or impact of outreach. Either lending strategy leads to the same level of poverty in the economy. For the case where R − (1 + r c ) < 0 , so banks do not lend, the marginal impact of increasing outreach is −(λ p h + (1 − λ ) p l ) R + (1 + i )

(7)

Since this expression is positive, as the MFI increases outreach the poverty gap rises. Hence, when banks cannot afford to offer a pooling contract, the MFI can minimize poverty in the economy by maximizing impact of outreach rather than outreach itself. QED

Recent Trends in Microfinance Institutions: Some Theoretical Implications

483

Proof of Result 2.2. Assume that λ A ≥ λ B . For agents with wealth level w j , if the MFI offers exactly one loan at the rate r j this generates a poverty gap

PG j = { y − (λ j p h + (1 − λ j ) p l ) ⎡⎣R − (1 + r j )(1 − w j ) ⎤⎦} +(

N − 1) y − (λ j p h + (1 − λ j ) p l ) ⎡⎣ R − (1 + r jc )(1 − w j ) ⎤⎦ 2

{

}

(12)

The poverty gap for group A is less than the poverty gap for group B if PG A < PG B or y [(λ A p h + (1 − λ A ) p l ) − (λ B p h + (1 − λ B ) p l )]{1 + ( +(

N − 1) R} 2

N − 1)(1 + i )( w A − w B ) > 0 2

(13)

Since we have assumed that λ A ≥ λ B , this inequality holds. QED References

1. A. B. Aghion and J. Morduch, The Economics of Microfinance. MIT Press, Cambridge MA(2005). 2. C. Ahlin and N. Jiang, Journal of Development Economics 80, 1 (2008) 3. T. Besley and S. Coate, Journal of Development Economics 46, 1 (1995). 4. P. Bose Journal of Development Economics 56, 265 (1998). 5. J. Conning, Journal of Development Economics 60, 51 (1999). 6. J. Copestake, P. Dawson, J.P. Fanning, A. McKay and K. WrightRevolledo, Development Policy Review, 60(1), 703 (2005). 7. R. Cull, A. Demirguc-Kunt and J. Morduch, Economic Journal 117, F107 (February, 2007). 8. S. Daley-Harris, State of the Micro-credit Summit Campaign Report, Washington D.C: Microcredit Summit Campaign, (2006), 9. J. Foster, J. Greer, and E. Thorbecke, Econometrica 52,761 (1984). 10. M. Ghatak, Journal of Development Economics 60(1), 27 (1999). 11. M. Ghatak, and T. Guinnane, Journal of Development Economics 60(1), 27 (1999). 12. S. Ghosh, and E. Van Tassel, Working Paper, Florida Atlantic University, (2007), 13. K. Hoff, and J. Stiglitz, Journal of Development Economics 55, 485 (1998). 14. D. Hulme and P. Mosley Finance Against Poverty, London: Routledge (1996). 15. S. Jain, Journal of Development Economics, 59(2), 419 (1999),

484

Suman Ghosh and Eric Van Tassel

16. D. Karlan, and S. Mullaianthan Working Paper, MIT (2006). 17. S. R. Khandker, The World Bank Economic Review, 19(2), 263 (2005), 18. C. McIntosh and B. Wydick., Journal of Development Economics, 78, 271 (2005). 19. J. Morduch, Journal of Development Economics, 60, 229 (1999a). 20. J. Morduch, Journal of Economic Literature, 37 (4), 1569 (1999b). 21. J. Morduch, World Development, 28 (4), 617 (2000). 22. J. Morduch, Working Paper, New York University (2006). 23. M. Robinson, The Microfinance Revolution: Sustainable Banking for the Poor, Washington D.C: The World Bank (2001). 24. J. Stiglitz, World Bank Economic Review 4(3), 351 (1990), 25. R. Townsend and J. Yaron, The Credit Risk-Contingency System of an Asian Development Bank, Federal Reserve Bank of Chicago Economic Perspectives, Third Quarter, 31 (2001). 26. M. Yunus, Moneylenderer to the Poor: Microlending and the Battle Against World Poverty, Public Affairs, Perseus Book Group, New York, NY (2003).