Advances on International Economics [1 ed.] 9781443881333, 9781443878289

Citation preview

Advances on International Economics

Advances on International Economics Edited by

Carmen Díaz-Roldán and Javier Perote

Advances on International Economics Edited by Carmen Díaz-Roldán and Javier Perote This book first published 2015 Cambridge Scholars Publishing Lady Stephenson Library, Newcastle upon Tyne, NE6 2PA, UK British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library Copyright © 2015 by Carmen Díaz-Roldán, Javier Perote and contributors All rights for this book reserved. No part of this book may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without the prior permission of the copyright owner. ISBN (10): 1-4438-7828-6 ISBN (13): 978-1-4438-7828-9

CONTENTS

List of Tables .............................................................................................. xi List of Figures........................................................................................... xiii Part I: Introduction Chapter One ................................................................................................. 3 Recent Developments on International Economics Carmen Díaz-Roldán and Javier Perote 1.1. Introduction 1.2. Modelling International Economics 1.3. International Trade 1.4. Macroeconomic Aspects of International Trade and Finance 1.5. International Factor Movements and International Business Part II: Modelling International Economics Chapter Two .............................................................................................. 21 Recent Episodes on a Globalized Economy: An Historical Approach Nieves Carmona-González 2.1. Introduction 2.2. Globalization of the Economy in an Historical Perspective 2.3. Globalization: The Result of Integration and Internationalization Processes 2.4. Economic Integration as the Engine of Globalization: Some Examples 2.5. Economic Crises: Lessons from the Past 2.6. Conclusions Chapter Three ............................................................................................ 39 Monetary Unions under Financial Shocks: Do Fiscal Rules Matter? Carmen Díaz-Roldán and Alberto Montero 3.1. Introduction 3.2. The Fiscal Rules in Monetary Unions 3.3. The Macroeconomic Model

Contents

vi

3.4. The Costs and Benefits of Following a Fiscal Rule 3.5. Summary and Conclusions Chapter Four .............................................................................................. 63 A Generalized NEG Wage-Type Equation Fernando Bruna 4.1. Introduction 4.2. The Demand Side of the NEG Basic Model 4.3. Supply Side and Generalized Wage-Type Equation 4.4. Inclusion of Human and Physical Capital 4.5. Conclusions Chapter Five .............................................................................................. 83 The Stationarity of the Inflation in Latin-American Countries Reviewed when Additive Outliers are Detected Dionisio Ramírez-Carrera and Gabriel Rodríguez 5.1. Introduction 5.2. The Issue of Outlier Detection and Testing for Unit Roots with Additive Outliers 5.3. Monte Carlo Results 5.4. Empirical Application 5.5. Conclusions Part III: International Trade Chapter Six .............................................................................................. 117 The Effectiveness of Development Aid in Promoting Economic Growth Laura Gómez-Wingley 6.1. Introduction 6.2. Theoretical Background and Current Situation 6.3. Methodology 6.4. Results 6.5. Conclusions Chapter Seven.......................................................................................... 139 Offshoring and Productivity from a Time-Series Perspective Pablo Agnese 7.1. Introduction 7.2. A Simple Model 7.3. Empirical Analysis 7.4. Effects on Labor Productivity 7.5. Final Remarks

Advances on International Economics

vii

Chapter Eight ........................................................................................... 159 Sectorial Differences in Export Performance: The Role of the DecisionMaker’s Gender Helena Marques 8.1. Introduction 8.2. Explaining Differences in Export Propensity and Intensity 8.3. What Can be Learned from Worldwide Data 8.4. Gender Effects across Sectors and Firm Sizes 8.5. Conclusions Chapter Nine............................................................................................ 183 Tariff Reform and Reduction of Regulatory Barriers: An Analysis for Female and Male Workers María C. Latorre 9.1. Introduction 9.2. Data on Gender and Tanzanian Economic Structure 9.3. The Model 9.4. Results 9.5. Conclusions Part IV: Macroeconomic aspects of International Trade and Finance Chapter Ten ............................................................................................. 209 Detecting De Facto Exchange Rate Regimes María del Carmen Ramos-Herrera and Simón Sosvilla 10.1. Introduction 10.2. De Iure and De Facto Exchange Rate Regimes Classifications 10.3. Implicit Fluctuation Bands 10.4. Credibility Tests of Fluctuation Bands 10.5. Concluding Remarks Chapter Eleven ........................................................................................ 225 How Do International Economic Conditions Affect Household Economic Expectations? An Analysis for the Peripheral Countries Cristina Vilaplana Prieto 11.1. Introduction 11.2. Econometric Model 11.3. Data and Descriptive Statistics 11.4. Results 11.5. Conclusions

viii

Contents

Chapter Twelve ....................................................................................... 247 Assessing the Evolution of Economic Integration in Latin America: Is Mercosur passé? Alejandro Jacobo and Ileana Jalile 12.1. Introduction 12.2. On MERCOSUR Trade Dynamism 12.3. Disentangling the Determinants of Tariff AD Barriers in MERCOSUR 12.4. Concluding Comments Chapter Thirteen ...................................................................................... 271 Normative vs. Positive Fiscal Policies: An International Analysis on Their Contribution to Economic Growth Fernando Callejas, Isabel Martínez and Miguel A. Tarancón 13.1. Positive Economics versus Normative Economics 13.2. Importance of Fiscal Policy to the Economy of a Country and Its Impact on Economic Growth 13.3. Procyclical Fiscal Policy 13.4. Analysis of the Implementation of Fiscal Policies in the Economies of Different Geographical Areas Part V: International Factor Movements and International Business Chapter Fourteen ..................................................................................... 299 Financial Market Risk Before and After the Subprime Crisis Isabel A. Palomero, Javier Perote and Andrés Mora-Valencia 14.1. Introduction 14.2. Empirical Characteristics of Financial Time Series 14.3. Non-Normality and the Frequency of Log-Returns 14.4. Estimating Value-at-Risk of Financial Series 14.5. Value-at-Risk in Highly Volatile Scenarios 14.6. Conclusions Chapter Fifteen ........................................................................................ 321 Financialization of the Economy and Changes in Industrial Relations Santos Ruesga and Manuel Trujillo 15.1. Introduction: The Beginning of Financial Globalization 15.2. The Concept of Financialization 15.3. The Effect of Financialization on the Labour Market 15.4. Empirical Evidences 15.5. Summary and Conclusions

Advances on International Economics

ix

Chapter Sixteen ....................................................................................... 343 The China Effect on Foreign Reinvestment in Asia Salvador Gil-Pareja, Rafael Llorca and Jordi Paniagua 16.1. Introduction 16.2. Background 16.3. Empirical Methodology and Data 16.4. Results and Discussion 16.5. Conclusions

LIST OF TABLES

Table 3-1: Government Deficit (-)/Surplus (+) and Debt in the EU-27 (% of GDP) Table 3-2: Losses after a Common Contractive Demand Shock Table 4-1: The Generalized Wage Equation in Several NEG Models Table 5-1: Size of the ADF Test: ARFIMA (0, d, 0) Errors Table 5-2: Size of the ADF Test: ARFIMA (0, d, 1) Table 5-3: Size of the ADF Test: ARFIMA (1, d, 0) Table 5-4: Size of the ADF Test: ARFIMA (0, d, 0) Table 5-5: Size of the ADF Test: ARFIMA (0, d, 1) Table 5-6: Size of the ADF Test: ARFIMA (1, d, 0) Table 5-7: Standard and New Unit root Tests Table 5-8: ADF Test corrected for Additive Outliers using ߬ௗ Table 6-1: Description of the Variables and Data Sources Table 6-2: Models for Testing the Impact of ODA on Economic Growth Table 7-1: Labor Productivity Growth Rate, Broad Sector Averages (%) Table 7-2: Summary Statistics, 1976-2008 Table 7-3: Two Equation Models for Different Industries Table 7-4: Offshoring Contribution to Productivity Table 8-1: Survey Countries and Sectors Table 8-2: Description of Variables Table 8-3: Mean Differences by Top Manager and Sole Owner Gender Table 8-4: Gender, Exporting and Firm Size by Sector (Sample %) Table 8-5: Export Propensity by Sectors Table 8-6: Export Intensity by Sectors Table 8-7: Summary of Gender Effects Table 9-1: Factors' Remuneration with Respect to Total Value Added in the New and Previous Datasets (in percentage) Table 9-2: Tariff Levels, Market Shares and Regulatory Barriers (in Percentage) Table 9-3: Aggregate Results (% Change from Initial Equilibrium) Table 9-4: Production (Percentage Change from Initial Equilibrium) Table 9-5: Factor Remunerations (Percentage Change from Initial Equilibrium) Table 9-6: Sectoral Composition of Factors’ Remunerations Across Main Sectoral Classifications (in Percentage)

xii

List of Tables

Table 9-7: Factor Intensities Across Sectors (in Percentage) Table 9-8: Sensitivity Analysis Table 11-1: Description of the Special Eurobarometer Surveys Used Table 11-2. Descriptive Statistics for Macroeconomic Variables Table 11-3: Estimation of the Bivariate Ordered Model with and without Mixed Effects Table 11-4: Predicted Probabilities from the Bivariate Ordered Model with Mixed Effects Table 11-5: Simulation of the Effect over Expectations as a Consequence of a Change in Macroeconomic Conditions Table 12-1: Applied Bilateral Tariffs Baseline Model Table 12-2: Bilateral Tariff Baseline Model Before/After 2009 Table 12-3: Bilateral Tariff Baseline Model with Product Fixed Effect Table 12-4: Antidumping Initiations Incidence Before/After 2009 Table 12-5: Antidumping Initiations Incidence Model: Fixed Effects Table 13-1: Summary of Contributions Concerning the Impact of Fiscal Policy on Economic Growth Table 13-2: Germany. Positive Economic – Normative Economic Adjust Table 13-3: France. Positive Economic – Normative Economic Adjust Table 13-4: Italy. Positive Economic – Normative Economic Adjust Table 13-5: Spain. Positive Economic – Normative Economic Adjust Table 13-6: European Union. Positive Economic – Normative Economic Adjust Table 13-7: United Kingdom. Positive Economic – Normative Economic Adjust Table 13-8: Japan. Positive Economic – Normative Economic Adjust Table 13-9: Brasil. Positive Economic – Normative Economic Adjust Table 13-10: Relationship 2002-2012 Table 14-1: Log-Likelihood of ARMA-GARCH Models Table 14-2: Fitted Density Parameters for the IBEX35 Table 14-3: Descriptive Statistics for Log-Returns of Different Indices Table 14-4: AR(1)-GARCH(1,1) Estimates in the Pre-Crisis Period Table 14-5: AR(1)-GARCH(1,1) Estimates in the Crisis Period Table 15-1: Empirical Outcomes Table 16-1: World's Largest Companies Table 16-1: Main contributions on the China Effect Table 16-2: Variable Description Table 16-3: List of Countries Table 16-4: Results

LIST OF FIGURES

Figure 5-1: Quarterly Latin-American Inflation Series Figure 7-1: Labor Productivity Growth Rate, Broad Sector Averages (%) Figure 7-2. Offshoring intensity and IT-Manufacturing (%) Figure 7-3: Semiconductors: Offshoring Contribution to Productivity Figure 7-4: Electronic Parts: Offshoring Contribution to Productivity Figure 7-5: Industry Machinery: Offshoring Contribution to Productivity Figure 7-6: Electronic Equipment: Offshoring Contribution to Productivity Figure 11-1: Household Expectations (EH) and General Economic Expectations (EE) Differences and Real GDP Growth Rate Figure 11-2: Difference between Optimistic and Pessimistic Household Expectations (EH) and General Economic Expectations (EE) and GDP Growth Rate Figure 11-3: Effect of General Economic Expectations over Household Expectations Figure 12-1: Trade Openness: MERCOSUR Countries 1991-2011 Figure 12-2: Growth Rates of TO and GDP: Argentina 1991-2011 Figure 12-3: Growth Rates of TO and GDP: Brazil 1991-2011 Figure 12-4: Growth Rates of TO and GDP: Uruguay 1991-2011 Figure 12-5: Growth Rates of TO and GDP: Uruguay 1991-2011 Figure 12-6: Intraregional Trade Index for MERCOSUR 1994-2009 Figure 12-7: Herfindahl-Hirschmann Index: MERCOSUR 1991-2012 Figure 13-1: Positive Economics versus Normative Economics Figure 13-2. Fiscal Policy Behavior: Procyclical and Countercyclical Figure 13-3: Germany. Evolution GDP. PE and Tx Figure 13-4: France. Evolution GDP. PE and Tx Figure 13-5: Italy. Evolution GDP. PE and Tx Figure 13-6: Spain. Evolution GDP. PE and Tx Figure 13-7: European Union. Evolution GDP. PE and Tx Figure 13-8: United Kingdom. Evolution GDP. PE and Tx Figure 13-9: Japan. Evolution GDP. PE and Tx Figure 13-10: Brazil. Evolution GDP. PE and Tx Figure 14-1: Daily Log-Returns of IBEX35 and Dow Jones Indices Figure 14-2: Level and Log-Return of the IBEX35 Figure 14-3: Histogram and Descriptive Statistics of the Log-Return of IBEX35 and Dow Jones Indices

xiv

List of Figures

Figure 14-4: QQ-Plots for the Daily Log-Return of IBEX35 under Normal and Student’s t Distributions Figure 14-5: QQ-Plot for the Monthly Log-Return of IBEX35 under the Normal Distribution Figure 14-6: Histogram and Descriptive Statistics of the Monthly Log Returns of the IBEX35 Index Figure 14-7: Log-Returns and 95% VaR for the IBEX35 under the Normal and Student’s t Figure 14-8: Log-Returns and 99% VaR for the IBEX35 under the Normal and Student’s t Figure 14-9: Histogram and Normal and Student’s t Fitted Densities of the Standardized IBEX35 (Left Tails) Figure 15-1: Relationship Between the Variation of Financialization Rate* and the Variation of the Unemployment Rate for 22 OECD Countries for the Period 1995 – 2007 Figure 15-2: Relationship Between the Variation of Financialization Rate* and the Variation of the Employment Compensation in Percentage of Value Added for 20 OECD Countries for the Period 1995 – 2007 Figure 16-1: GDP Distribution in 1950, 1973, 2001 and 2010 Figure 16-2: World's FDI Share Figure 16-3: FDI in Asia Figure 16-4: FDI in South America Figure 16-5: FDI in Southern and Eastern Europe

PART I: INTRODUCTION

CHAPTER ONE RECENT DEVELOPMENTS ON INTERNATIONAL ECONOMICS CARMEN DÍAZ-ROLDÁN* AND JAVIER PEROTE

The goal of this chapter is to offer a broad overview of the recent developments on International Economics, both in the theoretical and empirical areas. The recent financial and economic crisis raised questions concerning the usefulness of several paradigms accepted by both academia and the advising government institutions. In such a framework, new theories and practical lessons stemming from empirical analyses are revealed as useful tools for international policy recommendations.

1.1. Introduction In a globalized world characterized by huge international capital mobility, interest in International Economics has been renewed in academic circles as well as in economic policy forums and supranational institutions. On one hand, capital markets have reached a remarkable development, due to increasing financial innovations and the widespread liberalization of capital movement. As a consequence, the evolution of the exchange rate has come to depend more and more on the movement of capital instead of on the international trade of goods. This greater reliance on capital movements is reflected in the high volatility of the values reached by the exchange rates. On the other hand, the integration process evolution is quite complex and, indeed, different worldwide. Examples provided by Latin-American *

Corresponding author: Carmen Díaz-Roldán. Facultad de Derecho y Ciencias Sociales. Universidad de Castilla-La Mancha. 13071 Ciudad Real (Spain). Tel: +34-926-295300 Ext. 6657. E-mail: [email protected].

4

Chapter One

countries contrast with the European case of the monetary union based on a new currency, namely the euro. Integration processes are intended to increase the flow of international trade and investment and, through a more efficient use of resources, a favourable result in the evolution of productivity and economic growth. As was the case for the credibility of anti-inflationary policy, the elimination of transaction costs and exchange rate uncertainty would be achieved much more clearly in a monetary union than in a system of fixed exchange rates. But the recent financial and economic crisis has cast doubt on the success of monetary unions, given the lack of independent monetary and exchange rate policies. With those considerations in mind, in this book we provide some contributions with the intention of explaining the most recent developments in International Economics.

1.2. Modelling International Economics Developed economies are interrelated in international markets through the interchange of goods and services, and the mobility flows of capital and labour. It is well known that the increasing vulnerability of open economies depends on the degree of openness of the country, as well as on its size. The literature on International Trade and Economic Integration states that an open and large economy generates externalities on the outside world. Those characteristics are precisely the ones which define globalization, and the transmission and the effects of externalities depend crucially on the particular economic framework of the involved countries. Questions such as the exchange rate and the monetary regimes, the development of the financial system, and the role played by institutions, are not trivial. After the Second World War, the international economic order envisioned by the Bretton Woods Treaty gave way to the neo-liberal economic order that remains prevalent today, promoting advancement towards a close global economy. This phase was consolidated at the end of the 20th century thanks to the gradually growing influence of the General Agreement on Tariffs and Trade (GATT), the growing presence, at the global level, of international companies that work as integrated production systems, the expansion of high capital mobility and a noticeable tendency towards the homogeneity of the models of development. The Keynesianism that had led the economic policies of the 20th century focused on the prominent role of the public sector. The oil crisis of the 70’s, however, questioned not only Keynesian models but also the

Recent Developments on International Economics

5

usefulness of economic policies in general. Nowadays, in the 21st century, the recent problems generated by the economic and financial crisis have triggered a debate on the role of economic policies. The question is to what extent a specific monetary policy regime would impose a restriction on policymakers; in particular, the cost of losing independence in the use of exchange rates and monetary policy, and the restrictions derived from the fiscal discipline required for supporting monetary agreements. As an example, we can think about the expected success of the European Economic and Monetary Union (EMU), which relates to the benefits of the single currency, the higher degree of integration of financial markets, and also to the sound public finances guaranteed by the set of fiscal rules provided by the EMU. When signing the Stability and Growth Pact (SGP), Member States committed themselves to reaching a mediumterm budgetary position close to balance. The Maastricht Treaty mainly stresses that every Member State of the EMU should avoid excessive deficits, and the reference values for deficit-to-GDP and debt-to-GDP ratios have in fact worked as an explicit fiscal rule. But, in practice, the policy orientation of the SGP has not been fully satisfied. This has opened a debate about the utility and effectiveness of fiscal rules within the EMU, and on their complementarities with discretionary fiscal policy measures and automatic stabilisers to deal with short-run fluctuations. For these reasons, it is useful to investigate how to deal with monetary (financial) shocks in a monetary union following fiscal rules. The particular interest here is in discovering interactions among those member countries showing a relatively high level of public debt, and those that seem to follow a more strict fiscal discipline. In this book, Chapters 2 and 3 are devoted to the analysis of the extent to which the recent episodes affecting the global economy can be explained by International Economics theories and models, paying special attention to the particular macroeconomic framework of monetary unions. As addressed by Krugman (1998), there is a new approach for studying International Economics and International Trade. This new approach, the so called New Economic Geography (NEG), emerges from the study of geography itself and brings together disciplines such as urban, environmental, ethnic and gender economics, as well as location, distribution and the spatial organization of economic activities (see Krugman (1991), Krugman and Venables (1995) as seminal contributions). The theoretical bases rely on the same exploited by the “New Trade” and “New Growth” theories, i.e., general equilibrium models that derive aggregate behaviour from individual maximization. Stemming from economic growth, some NEG models try to explain the long-term growth of labour productivity

6

Chapter One

endogenously. These models incorporate the role of externalities associated with the accumulation of capital and technological innovation in the context of imperfect competition. Some of these models show identication problems due to two types of observational equivalence; the between-equivalence, which appears when comparing the implications of NEG models with those of alternative models, mainly based on technological externalities; and the withinequivalence, produced using the implications of NEG models based on vertical linkages among rms, along with those of NEG models based on factor mobility, are compared. In Chapter 4, an analysis of the observational equivalence of NEG theory is offered. In spite of the apparent empirical success of the NEG wage equation, some authors have asserted its observational equivalence, i.e., the consistency of those results with alternative frameworks, such as Urban Economics theories. However, few efforts have been made to empirically test this equivalence, with one reason possibly stemming from the lack of a commonly accepted approach for doing so. New empirical approaches propose a procedure for showing that many empirical wage equations are actually proxying an underlying production function, augmented with locational information about the economic scale of the nearest neighbour(s). The method begins by presenting a NEG setting with capital stock, which encompasses several wage equations found in the literature. A baseline wage-type equation is then estimated by redefining the key variables of the model in several ways which are different from those that are commonly considered by the NEG empirical literature. The findings are similar to those of a standard NEG wage equation, which is the essence of the observational equivalence of NEG theory. Finally, in Chapter 5, a key question for macroeconomic modelling is addressed. The proper inference of statistical parameters proves to be crucial for the forecast of the evolution of macro variables, and according to the Lucas critique, changes to the values of the parameters have serious implications for the design and implementation of economic policies. In an environment of growing economic interdependence and financial turmoil, the behaviour of the series of inflation is not a trivial question. Therefore, a new branch of econometric techniques is devoted to the analysis of the quarterly behaviour of macro variable series.

Recent Developments on International Economics

7

1.3. International Trade Traditionally, classical theories of International Trade are designed to answer two basic and closely related questions: What are the causes of trade? Why do countries trade? In addition, questions also include: How would countries import and export products in different industries? And what are the effects of international trade on domestic production and consumption? But, during the second half of the 20th century, the lines of research have been widened, focusing on questions such as the role played by the external aid that is provided by benefactor countries to developing ones, relocation phenomena, or the sectorial differences produced by the results of expenditure on research and development, and expenditure on human capital. The reason behind the amplification of these fields of study relies on the new ways of examining the performance of international relationships, and the factors contributing to economic growth favoured by the development of new technologies. Among the observed stylized facts of the last decades, the Official Development Assistance (ODA) flows have increased considerably. Consequently, a new branch of the literature has been devoted to the study of the impact of ODA on economic growth and the characteristics of recipient countries and donors’ management practices, which influence its effects. Chapter 6 analyzes new models, studying the impact of international aid on economic growth and welfare relative to the characteristics of recipient countries and donors’ management practices. These new growth theories and studies take into account the endogeneity of aid and other variables as well as time and individual fixed effects. Results confirm the effectiveness of foreign aid in promoting economic growth, though it presents diminishing marginal returns. Furthermore, its effectiveness decreases if the country has recently been involved in armed conflicts, which probably reflects the lower quality of the institutions in such situations. On the other hand, aid fragmentation is found to have a positive effect on growth. In conclusion, aid has proved to be effective in encouraging economic growth, though there needs to be further consideration of aid heterogeneity and its impact on other aspects of development. As a complement to foreign aid, and with regard to relocation phenomena, the process of globalization stimulates the availability of resources; foreign direct investment, and the outsourcing and offshoring methods of relocation; greater economic integration; and even the development of stronger social cohesion. In most of the cases, the increase in trade appears to be the result of offshoring, with manufacturing

8

Chapter One

becoming fragmented across borders because firms benefit from comparative cost advantages, the underlying economic logic being the reduction in cost. Through the relocation of economic activities, there is a transfer of jobs to other countries, if services are contracted with foreign companies or establish a base in sites outside the country of origin. Traditionally, relocation from one country to another was based on the production process (manufacturing or even accounting). But nowadays, it is also related to the supply of technical and administrative services. In particular, the increase of offshoring – or, in other words, the relocation of some processes – of the production of a firm has not yet received the attention it deserves. Of equal importance to the first goal of reducing costs, there are several consequences of offshoring that are changing the traditional patterns of production. One of the most important is related to the reduction of the cost of labour units and its implications for productivity and employment. Within the study of the subject of offshoring and productivity, Chapter 7 proposes a simple framework for estimating the contribution of offshoring strategies to the growth rate of labour productivity from a time series perspective. This framework is useful to assess the impact of offshoring on skill upgrading and the labour share. Both empirical questions are quite relevant and need to be answered, especially considering the recent years of slow growth. The results would suggest that offshoring can improve labour productivity in industry. Moreover, offshoring is found to be the source of important changes among industries with different skills (skill upgrading) and an important factor behind the fall of the labour share. As stated before, the classical theory of International Trade explains inter-industrial trade; that is, how countries import and export products in different industries. However, the classical theory does not take into account intra-industry trade, which occurs when a country exports and imports products which belong to the same sector or industry, and it also does not consider the weight of multinationals. On the contrary, the new theories on International Trade incorporate intra-industry trade, trade between similar countries, economies of scale and product differentiation, and also the role played by firm heterogeneity in the overall performance of International Trade. The literature on heterogeneous firms states that when there are fixed costs of exporting, only the most productive firms export (extensive margin) and the export intensity (intensive margin) changes with the variable costs of exporting for those firms. On the other hand, the literature on gender, entrepreneurship and firm performance has found that women are less likely to become entrepreneurs, and, after having entered the

Recent Developments on International Economics

9

market, firms where one of the owners is a woman show a weaker performance (measured, among other variables and also by productivity). Related to that, the new theories of endogenous growth have emphasized the role of externalities derived from the accumulation of capital. Endogenous growth models have assumed that knowledge spillovers come from expenditure on research and development, and from expenditure on education and the training of workers, what is known as human capital. Summing up, it seems that women used to be less skilled workers than men, and putting together the contributions from both strands of literature, the role of the firm’s decision-maker’s gender in determining sectorial differences in export performance (propensity and intensity) turns out to be a new branch of research. Chapter 8 studies the role of the decision-maker’s gender in export performance when there are sectorial differences. The gender dummy is defined for firms where women are either the firm’s top manager or the firm’s sole owner. In general, it is found that there are significant differences at several levels between firms managed and owned by either gender, but these differences depend on firm size (correlated with productivity) and the sector in which the firm operates. Moreover, the gender differences across sectors are not always negative. After having controlled for firm size and sector of activity, women present a greater gap in the amount of sales they allocate to foreign markets (export intensity) rather than in deciding to start exporting (propensity to export), so the variable costs of exporting may be a greater constraint relative to the fixed costs. From another point of view, Chapter 9 is also devoted to analysing the impact of gender. Nowadays, a recent approach when studying International Trade analyses the impact of gender on tariffs, Foreign Direct Investment (FDI) and regulatory barriers’ liberalization using computable general equilibrium models. FDI and facilitation of the new entry of domestic firms proves to be more important for workers than trade. This process is less beneficial for women because the less skilled workers and the ones that are more involved in the contracting agriculture sector used to have a lower salary.

1.4. Macroeconomic Aspects of International Trade and Finance In the last years, capital markets have reached a remarkable level of development, due to increasing financial innovation and a widespread liberalization of the movement of capital. Consequently, nowadays the

10

Chapter One

exchange rate is more linked to the movement of capital than it is to the international trade of goods and services. This evidence has resulted in a sharp increase of the volatility of exchange rates. One of the main issues, especially controversial and unresolved within the international context, is the choice of the optimal exchange rate regime. A system of flexible exchange rates ensures the existence of external balance. This, in turn, would allow the government to focus on achieving their internal goals. In addition, it would tend to insulate the economy from the effects of external shocks – particularly those affecting trade balance – and also guarantees greater autonomy with regard to monetary policy. On the contrary, a system of fixed exchange rates imposes greater discipline for the authorities: the fixed exchange rate would act as an “anchor” to gain credibility and would lead to an anti-inflationary policy. However, in recent years, fixed exchange rates have proved to be extremely fragile and difficult to maintain. The final reason is the extraordinary development experienced by the international capital markets, which makes the system of fixed exchange rates highly vulnerable to speculative attacks on a massive scale. The classical Mundell-Fleming model follows the so-called impossible trinity. This concept refers to the impossibility of simultaneously maintaining the following options: fixed exchange rates, monetary autonomy, and open capital markets, leading to the bipolar view (Obstfeld and Rogoff, 1995). This trilemma concludes that given a high mobility of capital and being more exposed to financial markets, there are only feasible corner solutions, i.e. hard peg regimes (dollarization, currency unions and currency boards) and free floating regimes. To shed light on those questions, Chapter 10 surveys the theoretical and empirical literature on exchange-rate regimes and also summarizes the main consensus available regarding the categorization of de iure and de facto exchange rate regimes. One of the theoretical explanations of exchange rate movements is based on the expectations about the behaviour of the government. According to this theory, the exchange rate moves from one equilibrium to another depending on how agents formulate their expectations. But the role played by expectations is not only a useful tool for analysing exchange rate movements; on the contrary, it is one of the most studied issues, both from the theoretical and empirical viewpoint, when trying to explain a wide range of economic phenomena. Nowadays, this issue has acquired a renewed interest in policy circles, given that agents’ expectations are one of the explanatory variables of the current economic

Recent Developments on International Economics

11

and financial crisis. The aim of Chapter 11 is to shed further light on the determinants of individuals’ expectations with respect to household finances and the general economic situation. The empirical study is performed relating to the case of Spain. This peripheral European country is of particular interest, since, in spite of the lower level of development in the Spanish economy, it has experienced a remarkable process of growth since its integration with the EU. But the trade deficits involved because of the increase in external openness has represented an unassailable obstacle to recovery after the current financial and economic crisis. In the model proposed in this chapter, individual prospects regarding personal situations are simultaneously determined with people’s opinions for the future economic situation, and also expectations for the general economy are an endogenous variable in the equation of expectations for personal financial situations. According to the results, it seems that families tend to believe that their personal financial situation is going to be better than forthcoming general economic conditions, and that general economic expectations are more sensitive to changes in economic growth rather than to contractions of the labour market. In addition to the evolution of the exchange rate and the role played by expectations, International Trade developments are also highly dependent on the development of international agreements, usually leading to strengthened economic linkages and, consequently, promoting a deeper integration between countries. Keeping in mind the European integration process, there have been several attempts to promote economic integration among the Latin American countries. Although the proposals have been mainly based on the creation of a common market, aimed to favour competitiveness and economic convergence, they have also served to renew interest in analysing the convenience of a monetary union. Nevertheless, there has been no explicit proposal for forming a monetary union similar to the EMU. In practice, the discussions on the creation of a monetary union have been more academic than political, because it does not seem that there are enough arguments to support the creation of a single common currency, or even to adopt another country’s currency. Theoretically, according to the analysis initiated in Mundell (1961), countries interested in sharing a common currency should form an Optimum Currency Area (OCA), i.e., a region characterized not only by trade integration, but also by price and wage flexibility, similar inflation rates and factor mobility among countries; and also by the mirroring of the external shocks the area would have to deal with. However, the Latin American countries would be far

12

Chapter One

from satisfying most of the criteria for establishing a currency area  see Edwards (2006) for a review. Given that a common market requires some degree of macroeconomic harmonization among the involved economies, this harmonization process usually leads to a deeper economic integration that also contributes to achieving a greater economic convergence. Moreover, when the countries which belong to an OCA are exposed to similar external shocks, there would be no serious reason to use the exchange rate policy to increase competitiveness. This argument has been exposed as one of the potential benefits of forming a monetary union, since unexpected exchange rate fluctuations would hurt the integration process. In this sense, as pointed out by Eichengreen (1998, p. 7), “…Whether or not exchange rate movements threaten regional integration depends on two things: the depth of that integration and the source of the disturbances in response to which exchange rate moves”. Chapter 12 briefly describes the trade dynamism of the four founding members of the Southern Common Market (MERCOSUR) during 19912012 in order to present a comprehensive picture of the current state of regional integration. The trade openness ratio, the intraregional trade index and the Herfindahl-Hirschmann index show that the countries reached higher levels of openness during 2002-2012, intraregional trade was more intense in the first decade than in the second, and the diversification of exports by country of destination has increased since 2001. The assessment is completed with the exploration of the determinants of tariff and non-tariff barriers. The use of pre- and post-2008 crisis trade and protection data reveals that discriminatory measures implemented by Argentina and Brazil do not provide a better treatment for their MERCOSUR partners, but quite the opposite. To preview the chapter’s conclusion: MERCOSUR is not passé, but it has somewhat shifted from its optimum path to success. Among the implications of any integration process, we found changes in the macroeconomic framework of reference. Consequently, changes in the performance and implementation of economic policies should be taken into account. As has been addressed in the Introduction, a debate on the role of economic policies has been opened following the financial and economic crisis. It is well known that the success of particular policies depends not only on the improvement of macroeconomic indicators, but also on macroeconomic conditions such as monetary and fiscal policy regimes, and also the exchange rate adjustment adopted. The economic framework characteristics are particularly relevant in monetary unions, where fiscal policy is the only demand policy aimed at achieving the

Recent Developments on International Economics

13

stabilization goal, and monetary autonomy is obtained at the cost of losing direct control over the exchange rates. In the EMU, the fiscal policy is oriented to achieve output stabilization in the short-run, through the use of the public deficit and automatic insurance mechanisms. In the long-run, the fiscal policy should guarantee the sustainability of public finances, and it should also contribute to economic growth through the structure of revenues and expenditures, and public investment in physical and human capital (European Central Bank, 2004). Keeping these points in mind, recent research contributions consider the impact of fiscal policies that have been implemented in recent years (see for instance, Barrios et al. (2010), Riguzzi (2011), and Karras (2012), among others). Some of these studies analyse times of growth and times of recession to determine, through comparative analyses between the two periods, the degree of effectiveness of the instruments of fiscal policy. In the case of fiscal policy, many questions arise relating to its behaviour and effectiveness, especially in times of recession. For example, what are the effects of a tax cut on the economy? How much does it matter whether they are financed by corresponding spending cuts or by corresponding increases in public debt, compared to the scenario of not implementing a tax cut? (Mountford and Uhlig, 2009) How persistent should spending cuts or tax increases be? What degree of fiscal stimulus should be considered? (Parker, 2011) Chapter 13 attempts to answer all of these questions in the scenario of the current crisis, where both fiscal consolidations and the increase of economic growth are needed.

1.5. International Factor Movements and International Business The 2007 financial crisis spread into a global economic shock and it was transmitted to the EMU. In fact, in 2009, there was negative growth in the Eurozone. In an attempt to restore trust in the financial system and recover growth rates, the macroeconomic policies in most economies focused mainly on short-term actions such as the bailout of banks and insurance companies, expanding money supplies and implementing large fiscal stimulus packages. Both the United States Federal Reserve and the European Central Bank have undertaken the largest monetary policy action in world history. Regarding the reforms and long-term responses, no significant measure has been implemented. In particular, the lack of fundamental changes in the banking and financial markets is one of the main concerns of some contributions to the International Monetary Fund’s

14

Chapter One

publications (see Blanchard and Milesi-Ferreti (2009) and Merrouche and Nier (2010), among others). One of the main concerns regarding financial markets is that the volatility in these markets has experienced a dramatic increase during the subprime and sovereign debt crises. Nevertheless, despite the well-known fact of the non-normality of asset returns in these scenarios, most financial institutions still report value at risk (VaR) measures based on the Gaussian distribution and thus underestimate their risk exposition. Chapter 14 revises the main empirical features of high frequency financial data and studies the relative VaR performance of the methodologies based on either Normal or Student’s t distributions. The analysis shows a clear outperformance of the latter that is found, in particular, in the context of high volatility scenarios such as the recent financial crises. The empirical analyses in this chapter focus on stock indices, but the results may be extended to other variables such as interest rates or exchange rates. Since the 1970s, capitalism has undergone profound changes following a new wave of technological innovations and changes in social organization, with the result being the configuration of a new economic model. One of the peculiarities of this new economic model concerns the growing prominence of financial activities, which increase their importance in the direct production of goods and services, and set the standard level of return on capital for all economic activities, including the non-financial sector. The new model has been accompanied by a rapid removal of trade barriers and the free movement of capital around the world, which has contributed to the expansion of financial activities. This process has accelerated the speed of investment flows, changing the strategic decisions of companies that are particularly oriented to meeting the expectations of their investors, which are focused on immediate short-term interests – trying to ensure the profitability of their investments, compromising the investment and production decisions of the firms in the long-term. This evidence has direct effects on the development of the labour market. In this new scenario, labour suffers increasing pressure to improve its productivity, which is achieved by greater flexibility conditions, in order to adapt the labour to increasingly competitive market conditions. This process tends to worsen working conditions, in order to keep labour costs adequate and increase production efficiency, i.e. the efficiency measure that ensures the profitability of the shareholders’ investment. In chapter 15, an empirical analysis is performed to assess some of the impacts of economic financialization on labour relations for a sample of developed countries. The empirical results seem to establish a negative

Recent Developments on International Economics

15

and significant relationship between the advance of economic financialization, and some key aspects concerning the equilibrium in the labour market – unemployment and labour compensation – and income distribution. Finally, we cannot forget to mention the emerging markets phenomenon. In recent years, several large economic areas have experienced a rapid industrialization process and sustained growth. The most remarkable characteristic is that those economies have reached a considerable positive growth differential over the OECD average, due to the lower cost of their production processes and, in some cases, the rapid accumulation of technologies or even the catching-up phenomenon. For that reason they have been called the emerging markets: the ASEAN (Association of South East Asian Nations), the BRIC countries (Brazil, Russia, India and China), as well as Mexico, Indonesia and Turkey (see Hanson (2012) for an analysis). The shift of international capital from the west to the east has configured a new set of actors in the international economy. The consequences of the rebalanced economic power implied by the shifting wealth process are wide and intense (OECD, 2011). The effects in terms of trade, financial flows and the catch-up of new technologies are impressive. But also, there are gains in growth, economic convergence and social cohesion. The countries involved are mainly large emerging economies, particularly China, which has been a strong worldwide competitor for years. Moreover, new multinational enterprises (MNEs), rooted in Asia and Latin America, are the new corporate giants at the expense of western and Japanese MNEs. China’s companies have shifted to deliver complete high-end products. To illustrate the phenomena, Chapter 16 examines foreign direct investment (FDI) diversion, in particular the impact of firm-level foreign investments in China on FDI inflows into Asia. FDI in high performing recipients has mixed effects on its neighbours. FDI agglomeration around certain areas may divert new FDI from similar surrounding countries. However, MNEs seeking to expand their international production might spill-over FDI to border regions with favourable endowments. By means of the gravity equation, this research estimates the diversion of FDI extensive margin and aggregate flows in Asia due to greenfield investments and re-investments in China during 2003-2009. Results show that while new foreign projects in China deter FDI, re-investments expand FDI in Asia.

16

Chapter One

References Bajo-Rubio, O. and Díaz-Roldán, C. (2003). “Insurance mechanisms against asymmetric shocks in a monetary union: a proposal with an application to EMU”, Recherches Economiques de Louvain, 69, 73-96. Bajo-Rubio, O. and Díaz-Roldán, C. (2007). “Vulnerability to shocks in EMU: 1991-2004”, Czech Journal of Economics and Finance, vol. 57, nº 5-6, 225-234. Bajo-Rubio, O. and Díaz-Roldán, C. (2011). Macroeconomic analysis of monetary union: A general framework based on the Mundell-Fleming model, Springer, Berlin. Bajo-Rubio, O., Díaz-Roldán, C. and Esteve, V. (2006). “Is the budget deficit sustainable when fiscal policy is nonlinear? The case of Spain, 1964-2001”, Journal of Macroeconomics, Vol. 28, nº 3, 596-608. Bajo-Rubio, O., Díaz-Roldán, C. and Esteve, V. (2009). “Deficit sustainability and inflation in EMU: an analysis from the fiscal theory of the price level”, European Journal of Political Economy, vol 25, 525-539. Barrios, S., Langedijk, S. and Pench, L. (2010). “EU fiscal consolidation after the financial crisis. Lessons from past experiences”, Economic Paper 418, Directorate-General for Economic and Financial Affairs, European Commission. Blanchard, O. and Milesi-Ferreti, G.M. (2009). “Global imbalances: In Midstream?”, International Monetary Fund Staff Position Note 09/29. Díaz-Roldán, C. (2004a). “International monetary policy coordination under asymmetric shocks”, International Advances in Economic Research, 10, 72-82. —. (2004b). “On the desirability of coordinated supply-side intervention: does a monetary union matter?”, Czech Journal of Economics and Finance, 54, 243-251. Edwards, S. (2006). “Monetary unions, external shocks and economic performance: A Latin American perspective”, International Economics and Economic Policy 3, 225-247. Eichengreen, B. (1998). “Does Mercosur need a single currency?”, Working Paper 6821, National Bureau of Economic Research. European Central Bank (2004). “Fiscal policy influences on macroeconomic stability and prices”, Monthly Bulletin, April, 45-57. Hanson, G.H. (2012). “The Rise of Middle Kingdoms: Emerging Economies in Global Trade”, Journal of Economic Perspectives, American Economic Association, vol. 26(2), pages 41-64.

Recent Developments on International Economics

17

Karras, G. (2012). “Trade openness and the effectiveness of fiscal policy: some empirical evidence”, International Review of Economics 59 (3), 303-313. Krugman, P. (1991).“Increasing Returns and Economic Geography”, Journal of Political Economy, 99 (3), 483-499. —. (1998). “What’s new about the new economic geography?”, Oxford Review of Economic Policy 14 (2), 7-17. Krugman, P. and Venables, A.J. (1995). “The Seamless World: A Spatial Model of International Specialization”, CEPR Discussion Paper 1230. Merrouche, O. and Nier, E. (2010). “What caused the global financial crisis? Evidence on the drivers of financial imbalances 1999-2007”, International Monetary Fund Working Paper 10/265. Mountford, A. and Uhlig, H. (2009). “What Are the Effects of Fiscal Policy Shocks?”, Journal of Applied Econometrics, 24(6), 960-992. Mundell, R. A. (1961). “A theory of optimum currency areas”, American Economic Review 51, 657-665. Obstfeld M. and Rogoff K. (1995). “The mirage of fixed exchange rates”, Journal of Economic Perspectives 9, 73-96. OECD (2011). "Shifting Wealth: A Window of Opportunity", in OECD, Perspectives on Global Development 2012: Social Cohesion in a Shifting World, OECD Publishing. Parker, J. A. (2011). “On Measuring the Effects of Fiscal Policy in Recessions”, Journal of Economic Literature 49(3), 703–718. Riguzzi, Marco, (2011). “Openness and the effect of Fiscal Policy”, Department of Economics, University of Bern, Switzerland.

PART II: MODELLING INTERNATIONAL ECONOMICS

CHAPTER TWO RECENT EPISODES ON A GLOBALIZED ECONOMY: AN HISTORICAL APPROACH NIEVES CARMONA-GONZÁLEZ*

The internationalization of economies has taken place since ancient times. Over the gradual stages of economic history, technological challenges have contributed to the development of international linkages leading to a growing economic interdependence. Moreover, it could be said that economic integration has been the engine of globalization. This chapter is devoted to the analysis of the extent to which the recent episodes affecting the global economy stem from the past and how International Economics theories and models have tried to explain them.

2.1. Introduction Among economists, there is not always an interest in analysing economic phenomena from an historical approach. Apart from “Economic History” or the “History of Economic Thought” as areas of research, there are not many more areas devoted to learning from the past. The Journal of Literature Classification System offers a broad description of the fields of studies in Economics, but we can easily compare section B, Schools of Economic Thought and Methodology, with the rest of sections and find a simple conclusion: History seems not to be in fashion. This observation has been placed in the arena of debate about the methods of research over time (see Blaugh (2001) for an interesting review). In Economics, the papers or books that base their introductions or motivations on the historical roots of the issues they address are scarce. But, undoubtedly, the episodes of current times are always a product of their own history. *

Universidad Francisco de Vitoria, Madrid (Spain). E-mail: [email protected].

22

Chapter Two

International Economics phenomena are no different. To when can the international dimension of Economics be dated? Why are economies more interrelated today than ever before? Why should economists analyse the stylized facts in an historical perspective? Nowadays, we talk about competitiveness and globalization as characteristics of the 21st century, but there is nothing new about that. The exchange of goods and services, as the engine of International Trade, has been favoured over the course of history by the evolution of transport and communications, but the basics work in a similar way: economies specialize in goods in which they present an advantage, and benefit from the difference in relative prices. Nowadays, distances are measured by the time taken to realise a transaction; so, the border effect is now more dependent on technologies of communication and information rather than purely on physical or geographical conditions. The globalization of trade and markets has different characteristics in the 21st century than those of the era of discoveries in the 16th century, but the roots of International Trade are the same. The current economic and financial crisis has been transmitted worldwide in a different manner than the preceding ones, but is not the first collapse of a financial bubble. Tulip mania and railway mania are good examples of this. In this chapter, we will review some historic episodes that have led to the current international economic situation.

2.2. Globalization of the Economy in an Historical Perspective The term globalization has been used since the end of the 20th century to highlight the technological advances that have favoured the extension of market forces beyond national borders. International transactions, both trade and financial flows, have been in operation for centuries. But the end of 20th century was when technological progress reached an unprecedented high level, allowing for the widest expansion of economic activity in history. The increases in cross-border trade, investment, and migration have been so large that Friedman (2005) has outlined that current globalization is “farther, faster, cheaper, and deeper” than earlier examples. In the last years, the engine of globalization has been the root of the development of new information technologies, making possible an unprecedented expansion of the markets; and the growing interdependence derived from economic integration among economies produces positive externalities derived from international trade agreements. Free trade zones are organized around major areas, with geographic advantages for trade. In those regions, countries agree to reduce or eliminate trade barriers, and

Recent Episodes on a Globalized Economy: An Historical Approach

23

over the decades, the governance of the international community has been responding to the globalization processes (or increasing internationalization) by strengthening the international monetary and financial systems. As defined by the World Bank, “Globalization is an inevitable phenomenon in human history that’s been bringing the world closer through the exchange of goods and products, information, knowledge and culture. But over the last few decades, the pace of this global integration has become much faster and more dramatic because of unprecedented advancements in technology, communications, science, transport and industry” (www.worldbank.org). Economic globalization is a continuous process, involving the increasing integration of economies worldwide, through trade and financial flows, and also the result of innovation and technological progress. The movements of factors across international borders also imply a broader cultural and political expansion. According to the International Monetary Fund (IMF, 2000), there are four basic aspects that characterize a globalization process: (i) increase in international trade and transactions, (ii) increase of capital and investment international movements, (iii) migration of people, and (iv) the dissemination of knowledge. Those four pillars are classic issues that have been studied by International Economics for decades: International Trade, Movements of factors: capital and labour, and Research and Investment spill-overs. Over the course of history, the economic integration of widely separated countries has promoted trade and economic development. After World War II, this re-integration was based on the intra industry trade aimed to exploit economies of scale. In the second half of the 20th century and the beginning of the 21st century, globalization has been driven by technological advances and by a geopolitical evolution that has allowed the integration of West–East and North–South economies. In such a globalized environment, the comparative advantages offered by international trade have become essential to countries remaining competitive. And as Krugman (1996) stressed, the new trade theories (based on imperfect competition and market failures) offer considerable scope even for government intervention. Following Krugman, we can briefly summarize the main points of view for explaining the sense of International Trade over time. The first viewpoint is that which is offered by Mercantilists. According to them, trade is viewed as a global struggle or war; countries behave as corporations, were ‘high-value’ sectors can be distinguished; and trade policy is mainly concerning about creating jobs. On the contrary, the Classical model of trade is essentially stated by David Ricardo and formalized by John Stuart Mill, being the main subject of

24

Chapter Two

International Economics and related subjects taught in universities. Under the Classical position, the purpose of trade is imports, not exports. Exports must be produced when import suppliers are crass enough to demand payment. In other words, an export is an indirect way to produce an import, when it is more efficient than producing certain goods domestically. The third viewpoint is that of the Strategists. In the basic classical model, competition is perfect. But in practice, the technological progress of industries seems to be generated by the mutual spill-overs of knowledge from national producers, and exports, therefore, may create comparative advantages. The failures of the classical model were the dominant subjects of the theoretical and empirical research into international trade during the 1980s. When industries generate strong spillovers, strategists support government intervention in domestic firms to take advantages of competition or monopoly profits. Finally, in the Realists’ point of view, the classical analysis of international trade refutes the mercantilist views, and the qualifications with regard to classical trade theory create arguments for intervention. Summing up these viewpoints, in Krugman’s words (Krugman, 1990), “The new trade theory suggests a more complex view. The potential gains from trade are even larger in a world of increasing returns, and thus, in a way, the case for free trade is all the stronger”.

2.3. Globalization: The Result of Integration and Internationalization Processes Nowadays, as stated in the Introduction, we are observing an unprecedented phase of economic development, qualified as a process of globalization. In fact, in the last decades, increases in cross-border trade, investment, and migration seem to be larger than ever. Broadly recognized, globalization means that world trade and financial markets are becoming more integrated, but globalization is not a recent phenomenon. If we understand globalization as a greater interconnection among regions and economies, we can interpret globalization from a historical perspective and recognize that it is not something new. In the words of Bernanke (2006): “History may provide some guidance … the process of global economic integration has been going on for thousands of years, and the sources and consequences of this integration have often borne at least a qualitative resemblance to those associated with the current episode”. The first globalization in history can be dated to when the Romans were developing an Empire. They built roads, water channels, imposed a

Recent Episodes on a Globalized Economy: An Historical Approach

25

legal system and forced other countries to use their coins. Jacques Le Goff (1980) said that “the Romans had the sense and the project [to] spread their territories all over the habited world”. Then, there was a real attitude of globalization. As Temin (2001) addressed, “… the economy of the early Roman Empire was primarily a market economy. The parts of this economy located far from each other were not tied together as tightly as markets often are today, but they still functioned as part of comprehensive Mediterranean market”. Later, after the fall of the Roman Empire, Feudalism turned this development backwards. The second globalization took place in the era of Great Discoveries and the beginning of the colonial empires. According to the World Systems Theory, underlying the works of Wallerstein (see Wallerstein 1974, 1980, 1989, 2011), the origin of the modern world system can be dated back to the 16th century. But the success of international trade was interrupted several times by wars and fights among the European monarchies. The precedents of the modern international economic and political order can be found in the writings of Francisco de Vitoria and the scholars of the School of Salamanca. This group of theologians and jurists conciliated the ideas of Thomas Aquinas with the practical problems of morality, economics, and jurisprudence. Their doctrines represented the end of the medieval concepts and opened the way to economic Liberalism and the development of International Law. Man became the center of the laws, and for the first time, the rights to life, private property, freedom of thought and human dignity were considered to be natural rights. Around 1771, with the Industrial Revolution, the third globalization began and reached its consolidation in the 19th century. This step of the globalization process occurred when economic Liberalism was substituted in place of Mercantilism, and governments began to incorporate the democratic system into parliaments. Economic liberalism became a reality at the end of the 18th century. In 1776, An inquiry into the nature and causes of the Wealth of Nations by Adam Smith was published. In the Wealth of Nations, Adam Smith defends some viewpoints which constitute the root of some key concepts in International Economics. Regarding foreign trade, it is well known his contribution are of an absolute advantage, but perhaps this is not extended to his opinions on the expansion of economic activity over the world. According to Smith, world wealth is not stabile and free international trade would be the way to increase the welfare level of countries. To achieve this goal, it would be necessary to specialize in a wide form of goods and activities, and increase production, which would make international cooperation obligatory. Through those ideas, Smith introduced both international agreements and

26

Chapter Two

coordination issues into the International Economics literature. Another of Smith’s contributions is related to the microeconomic foundations of international trade theory. In his work, Smith developed the concepts of the labour theory of value, the market balance price, and the perfect competition and the conditions of this kind of market. Those questions, using microeconomic information, allow the explanation of the foreign trade movements of international economies, the effect of open economies on welfare compared to closed economies, and also the composition of foreign trade. Finally, another key concept explained by Smith was that of arbitrage, or the law of one price in terms of the Wealth of Nations writings. The law of one price means that as a result of arbitrage the price difference of a high-grade security between two markets disappears and the price becomes one price. Nowadays, the arbitrage concept is used to refer to price equalization for all markets: factor markets, goods and services, and also exchange rate markets (in Kucukaksoy (2011), a detailed description of Smith’s contributions to International Economics literature can be found). In parallel to the publication of the Wealth of Nations, the Declaration of Independence of the United States of America was also signed in 1776. Later, in 1846, the Corn Law in England was repealed, liberalizing foreign trade, and in 1860, the Free Trade Agreement between France and England was signed. The European colonization of Africa and Asia, led to new methods of trade and capital flows, and there were also high levels of migration to colonize the new continents. Economic liberalism was a reality. But the third globalization was interrupted during the First World War and the Great Depression. Later, in the 20th century, after the Great Depression of the 30s, the Keynesians economic policies of the 20th century focused on the prominent role of the public sector and public intervention as a way of recovering. We can date the fourth globalization, after the Second World War, to the Bretton Woods Treaty, where a big effort was made to develop international financial and commercial institutions, such as the International Monetary Fund, the World Bank and the International Trade Organization. The Bretton Woods Treaty was signed with the purpose of stabilizing world currencies and establishing credit for international trade. The international economic order envisioned by the conference gave way to the neo-liberal economic order that is prevalent today, promoting advancement towards a close global economy. After the Second World War, the Bretton Woods Treaty was when the modern international economic order was promoted.

Recent Episodes on a Globalized Economy: An Historical Approach

27

Although the basic Keynesian model describes a closed economy, in 1944, John Maynard Keynes led the British delegation at the Bretton Woods Conference, where financial and commercial institutions such the International Monetary Fund, the World Bank and the International Trade Organization were founded. The presence in the world scenario of international firms was consolidated thanks to the gradual generalization of the General Agreement on Tariffs and Trade (GATT). So, the development of a new international order had begun. Following the emergence of Keynesian ideas, economic policies were used to stabilize the economy, expanding and contracting it to smooth out booms and slumps, in a context of fixed exchange rates under the Bretton Woods system. Moreover, the adoption of market economic systems contributed to increase production worldwide, through an intense growth of international trade and investment. Governments responded by establishing international agreements to promote trade in goods, services, and investment. As a result, international policies opened economies to global markets. But the oil crisis of the 70s brought into question not only Keynesians models but also the usefulness of economic policies. For the New Classical Macroeconomics, Keynesian models had meant a failure, both theoretically and empirically. So, alternative theoretical equilibrium models were proposed. According to them, economic agents form rational expectations and cyclical fluctuations are due to unanticipated changes. As a consequence, economic policy could only be effective if it was not perfectly anticipated but merely provisional. The fifth globalization began after the fall of the Berlin Wall on the 9th November, 1989, and the dissolution of the Soviet Union in 1991. Both dates mark not only the collapse of the soviet bloc but also the end of the bipolar system of the Cold War that divided the world into two sides, one socialist and another capitalist. Later, the former Soviet Union and the allies of the Warsaw Pact began an accelerated process of incorporation into the world market, the capitalist system and globalization tendencies. From the 1990s onwards, there was a fight for the markets that was translated into a wave of negotiations, signatures and the performing of free trade agreements that widened the interchanges between national borders. As a result, the World Trade Organization was created in 1995.

28

Chapter Two

2.4. Economic Integration as the Engine of Globalization: Some Examples As we have pointed out, globalization is mainly an economic process that is defined by the integration of several national economies. Economic globalization can be understood as the process of increasing economic integration between countries, leading to the emergence of a global marketplace or a single world market. In a globalized environment, the comparative advantages offered by international trade have become essential to countries remaining competitive. As we have seen, over the course of history, the economic integration of widely separated countries has promoted trade and economic development. After World War II, this re-integration was based on the intra industry trade, aiming to exploit economies of scale. In the second half of 20th century and the beginning of the 21st century, globalization has been driven by technological advances and by a geopolitical evolution that has allowed for the integration of West–East and North–South economies. In academic circles, the evolution of the world economy has been studied not only by doctrines but also through theoretical models, particularly in the last years. From the seminal contribution of ChaseDunn (1991), theoretical economic models and data have been incorporated to study the evolution of countries that configure the current structure of the world economy. Using a formal model of size, trade and growth, Alesina, Spolaore and Warcziag (2005) explore the idea that the country’s size is related to the trade regime. They also review the theory and evidence of economic performance and the history of international economic integration, starting from the notion of the city-state of Renaissance Europe, as an example of a territorial unit that was able to take advantage of world markets. But the international linkages leading to globalization do not always involve integration at every country area level. As economic integration advances at the supranational level, regional separatism emerges in some countries. For small territories the supranational community is seen as too big, while ethnic and cultural minorities feel themselves economically viable in the context of a common market. As per the study of Alesina et al. (2005), an economic union lowers the costs of independence for small countries by providing them with a free trade area. As a consequence, sometimes the globalizing dynamics revert and lead to a denationalizing process (Sassen, 2006). It is widely recognized that, in any case, economic integration among economies produces positive externalities. For that reason, countries have

Recent Episodes on a Globalized Economy: An Historical Approach

29

always tried to take advantages from international trade agreements. Free trade zones are organized around major areas with geographic advantages for trade. In those regions, countries agree to reduce or eliminate trade barriers. Countries involved in an integration process with other more advanced countries expect to enjoy significant GDP growth rates, as well as to grow faster than those countries, so they can converge with them. Since a common market requires some degree of economic harmonization, the harmonization process usually leads to a deeper economic integration that also contributes to achieving a greater economic convergence. In Europe, the Single Market Program was aimed at the removal of all kinds of non-tariff barriers that in fact prevented the existence of a single internal market within the former European Community. To this end, a legislative programme was developed, aimed at setting the essential requirements for completion of the internal market, including the removal of physical barriers (i.e., administrative formalities and border controls), technical barriers (i.e., security norms and regulations in production), and fiscal barriers (i.e., harmonization of indirect taxes), as well as the opening of public procurement to foreign producers. Finally, the Treaty on the European Union (known as the Maastricht Treaty), which came into effect in 1993, has meant a new step in the process of European integration. We can find examples of economic convergence in the case of transition economies, in particular the Central and Eastern European countries which are already members of the European Union (EU), the attempts of integration among Latin American countries, the Association of South East Asian Nations (ASEAN), and the Arab Maghreb Union. We should also mention a group of emerging countries, Brazil, Russia, India and China, named BRIC by O’Neill (2011). Given their large size and their importance in the global economy, the evolution of these countries generates big externalities. According to the latest estimates, they will turn into the engines of global growth in the next five years (Orgaz et al. 2011). Other countries comparable to the BRICs are Mexico, Indonesia, South Korea and Turkey. But the BRIC group is one of the most reliable emerging economies, since they have grown at a much faster rate than developed economies. The dynamism of the BRIC economies, according to the latest estimates, will turn them into the engines of global growth in the next five years (Orgaz et al. 2011). They have reached a high and stable growth rate that, joint with a slowdown in population growth, has favored the increase of per capita income and the convergence towards advanced economies. As we have just seen, in recent years, several large economic areas are experiencing a rapid industrialization process and sustained growth. They

30

Chapter Two

are called the emerging markets: the ASEAN, the BRIC countries, as well as Mexico, Indonesia and Turkey. The most remarkable characteristic is that those large emerging economies have reached a considerable positive growth differential over the OECD average due to the lower cost of their production processes and, in some cases, the rapid accumulation of technologies or even the catching-up phenomenon. However, according to Wallerstein (op. cit.), the modern world system is characterized by the economic control of the world order, divided into core, semi-periphery and periphery world areas. In addition, the economic interdependence is not symmetric: the core owns the highest level of development, while the periphery acts as supplier to the core area. One of the main contributions of Wallerstein is the criticism of global capitalism from a macroeconomic point of view. Along the same lines, Arrighi and Moore (2001) study the cycles of capital accumulation starting from the Spanish and Portuguese empire, and the city-state of the Renaissance as the origin of the capitalism. They point out that from the time of the ancients, over the course of economic history each centre of economic power has had a hegemony cycle. First, a commercial development creates a network of trade, followed by a worldwide expansion. Finally, a financial crisis leads to a change to the centre of economic power. This final step, the shift in the centre of economic gravity, has been named the shifting wealth phenomenon (OECD, 2011). But this sort of redistribution of the world economy is not new. The phenomenon of shifting wealth stimulates the availability of resources, foreign direct investment, and the methods of relocation, a greater economic integration, and even the development of stronger social cohesion. Through economic activities relocation, there is a transfer of jobs to other countries, if services are contracted with foreign companies or establish a base in sites outside the country. Traditionally, relocation from one country to another was based on the production process (manufacturing or even accounting), but nowadays, it is also related to the supply of technical and administrative services. The consequences of the rebalanced economic power implied by the shifting wealth process are wide and intense. The effects in terms of trade, financial flows and the catch-up of new technologies are the most impressive. But there are also gains in growth, economic convergence and social cohesion. The countries involved are mainly large, emerging economies, particularly China, which has been a strong worldwide competitor for years. The growth in the volume of international trade is related to reduced transport costs and trade agreement reforms. The emerging countries have

Recent Episodes on a Globalized Economy: An Historical Approach

31

also contributed to the increase of international specialization. The explanation is as simple as the classical theory of comparative advantage explaining international trade: the less efficient country (low-income) specializes in the production and export of goods in which its absolute disadvantage is less (labour intensive sectors). This is the good in which the country has a comparative advantage. On the other hand, the country should import (from middle-income countries) goods in which its absolute disadvantage is higher (capital-intensive sectors); that is, the goods in which it has a comparative disadvantage. In the late eighteenth century, Adam Smith argued that under free trade, each country should specialize in producing those goods in which they have an absolute advantage (or could produce in a more efficient way than other countries), and import those other in which countries have an absolute disadvantage (or produce less efficiently). This international specialization (or international division of labour) would lead to an increase in world production, which would be shared by the countries participating in trade. In the early nineteenth century, David Ricardo established that even if a country has an absolute disadvantage in the production of both goods compared to other states, if the relative costs (cost measured in terms of the other good) are different, the interchange of both goods is possible and mutually beneficial. The less efficient nation should specialize in producing and exporting the good in which its absolute disadvantage is lower. This is the good in which the country has comparative advantage. On the other hand, the country should import the good in which its absolute disadvantage is greater; that is, the good in which it has a comparative disadvantage. As can be seen, the roots of the contemporary shifting wealth process rely on the traditional theories which explain International Trade.

2.5. Economic Crisis: Lessons from the Past The first author who wrote about financial crisis was Mackay (1841). He described three historical moments of speculation between 1635 and 1720: tulip mania, the project of Mississippi, and the bubble of the South Sea Company. Thus, the first great financial crisis was the speculative bubble known as tulip mania on the Amsterdam Stock Exchange, and its end was the first stock market crash (see Garber (2000), Kindleberger (1991) and Thompson (2007), among others). This crisis, originating in the commercial prosperity of Holland, led to a sharp increase in money supply

32

Chapter Two

that eventually cheapened credit and encouraged the speculative fever around tulips. The next financial crisis took place in France. In 1716, John Law established the Banque Générale in France, which became the first central bank of the nation. Law contributed to “The Mississippi Bubble” that began grouping the trading companies of Louisiana into a single monopoly, and finished with the collapse of the Banque Générale and the devaluation of the Mississippi Company’s shares. The Mississippi bubble was contemporaneous with England’s South Sea Company bubble that started in London in 1720. This time, the speculation was due, among other things, to the new trade privileges of Great Britain in the Spanish colonies in America. The South Sea Company carried out the first securitization of debt in history. In the nineteenth century, there were other speculative crises linked to the expansion of transport. Inspired by the ideas stemming from Adam Smith and David Ricardo’s economic thoughts about free trade and the triumph of liberalism programs, the European and the United States blocs of economic power undertook a lengthy process to convert the transport lines. By the early 1930s, railway mania – an English term coined to describe the global passion for railways – had begun. In 1829, the first line opened in the United States of America. In just six years, a total of 1,500 kilometers of railway lines were laid. Some years later, in 1835, Britain had already built 720 kilometers. Other countries such as Germany, France, Belgium, the Netherlands and Italy were beginning to take their first steps. Classic economist ideas were widespread in a time when the industry was undergoing unprecedented development. Their main concern was economic growth and several related issues such as distribution, the theory of value and International Trade, among others. And one of their main objectives was to counteract the restrictive mercantilist ideas of competition that were still dominant at the time. With so-called Neoclassical economics, the focus moves to microeconomics, and the main objective of economic analysis becomes the functioning of the market system and its role in allocating resources. The Neoclassical macroeconomic analysis, represented by Irving Fisher, Knut Wicksell or Pigou, consisted in extending the Neoclassical microeconomic principles to the overall economy. According to these principles, market economic fluctuations would be eliminated if markets are left to act freely and, therefore, economic policies would not be necessary to influence the determination of activity levels.

Recent Episodes on a Globalized Economy: An Historical Approach

33

However, market mechanisms were not able to solve the problems that arose in the 1930s, when activity levels and employment fell around the world, to an extent that this period was called the Great Depression. The global collapse of capitalism during this time was difficult to explain from the dominant Neoclassical theory. It was inexplicable that the economic system did not show any tendency to return to the position of equilibrium and full employment through the market’s mechanisms themselves. The Great Depression of 1929 was a disaster and represented an intellectual failure to economists. As the economic depression was prolonged, the discrediting of Neoclassical theory increased. With the crisis of 1929, Classical and Neoclassical theories fell into disrepute. Monetary policy had failed. The Classical theory believed that after the fall of the quantity of money, wages and prices would also fall, and automatically, the economy would return to full employment. But employers refused to cut prices and workers refused to earn less. In this environment, the economic thought of Keynes became more important. In fact, Roosevelt did not hesitate to follow the policies recommended by Keynes when he was elected president of the United States of America, in 1933, to revive the American economy. Its aim was to increase employment through increasing public spending. As we have seen, economic theories have tried to explain economic crises. In general, they are a product of several steps that begin with a currency crisis that ends in a financial crisis and finally collapses in a more generalized economic crisis. A currency crisis can be defined as a serious difficulty in defending a particular exchange rate. The origin used to be a speculative attack carried out by operators in the foreign exchange market. The current economic and financial crisis is considered the worst since the Great Depression. But since the late twentieth century, there have been precedents such as the speculative attacks on the currencies of the European Monetary System in 1992 and 1993, the debt crisis in Mexico in 1994, devaluations and banking crises in Asia in 1997 and 1998, and the 2001 crisis in Argentina. These episodes showed the increasing difficulty of a country of acquiring the reputation needed to sustain a system of fixed exchange rates. Among other reasons, the source of such crises is the spectacular growth of the capital markets, followed by the continuous liberalization and deregulation of capital movements that have taken place in recent years. In some economies, the real estate sector is also affected, contributing to the collapse of the housing market and the decline in consumption, with a consequent slowdown in economic activity. This has contributed to the fact that the current economic crisis has become global.

34

Chapter Two

Theoretically, there are two types of explanatory theories of currency crises. The first one offers as an explicative argument, an unfavourable evolution of the “basics” or economic fundamentals: the so-called “first generation” models. Usually, crises are due to unsustainable monetary or fiscal policies, and/or the deterioration in competitiveness (sometimes as a consequence of abandoning the fixed exchange rate system). The second theory relies on the presence of expectations about the behaviour of the government, which are known as the self-fulfilling expectations: the so-called “second generation” models. The explanation is quite simple, and consequently the speculative attacks are self-fulfilling because if the attack is expected to occur, it will happen. Therefore, the exchange rate moves from the one equilibrium to another depending on how agents formulate their expectations. But the interaction between currency crises and weak financial systems feature in the latest episodes and the current financial and economic crisis. Those episodes may be described by the “third generation” models. In such events, exchange rates played a relatively minor role and the currency crises arose as part of a wider financial crisis. The root of the problem is the existence of underdeveloped financial systems which are poorly supervised and regulated. Along these lines, Paul Krugman (2002) emphasizes the role of the high external debt of firms in emerging countries. In an attempt to explain the economic crisis, among the postKeynesians, Minsky (1994) stated that financial fragility is a common feature of capitalist economies, linked to the economic cycle itself. So, high financial fragility leads to a high risk of financial crisis. As the economy grows and profits are expected, firms rely on their ability to borrow. More loans lead to more investment, the economy grows further and lenders trust in being able to recover the money without trouble. If firms have trouble refinancing debts, they begin to sell their most profitable assets to obtain liquidity. But when any firm fails, lenders cease to provide credit. This is the so-called Ponzi scheme debt financing: if there is not a new injection of liquidity, allowing refinancing, a real economic crisis will occur. From another point of view, Krugman (1999) noticed that sometimes the source of the problem is moral hazard. Emerging countries undertake excessive investment in risky projects, accompanied by a high external debt. But when the confidence of foreign investors decreases, they will withdraw their investments, causing a collapse of the exchange rate that translates into a financial crisis. The loss of confidence also reduces asset prices and the drop in investment is reinforced. The result is a recession

Recent Episodes on a Globalized Economy: An Historical Approach

35

that favours the collapse of asset prices, and endorses the initial loss of confidence (Krugman, 2002). Chang and Velasco (2000) indicated that the source of the problem is a lack of liquidity. If foreign currency debt is bigger than the availability of foreign credit, it turns in a loss of confidence, leading to a run on bank deposits. Then investment decreases, provoking the collapse of the exchange rate and, finally, a financial crisis. The trigger for the recent crisis was a lack of liquidity in the banking system in the United States of America in 2008, linked to the collapse of the USA housing bubble which had peaked in 2007. The insolvency of banks and the lack of availability of credit damaged the confidence of investors, and had a negative impact on capital markets. As credit facilities decreased and international transactions were reduced, economic growth slowed worldwide. As addressed in a report by the Brookings Institution: “The U.S. economy has been spending too much and borrowing too much for years and the rest of the world has depended on U.S. consumption as a source of global demand” (Bailly and Elliot, 2009). After the collapse of financial institutions, economic activity has decreased and unemployment has increased, contributing to a global economic recession. In Reinhart and Rogoff (2009), a full summary of the financial crisis can be seen.

2.6. Conclusions In this chapter, we have offered an analysis of how the recent episodes affecting the global economy stem from the past, and how they can be explained by the International Economics theories and models. Phenomena such as the globalization process and the economic crisis are not new. In fact, they are the product of the evolution of economic linkages and interactions between countries. Over time, technological advances have favoured the extension of market forces beyond national borders, and international transactions, both trade and financial flows, have been in operation for centuries. The growing interdependence derived from economic integration among economies has also brought benefits due to positive externalities derived from international trade agreements. In addition, the governance of the international community has been responding to the increasing internationalization processes by strengthening the international monetary and financial systems. In that environment, globalization can be understood as the process of increasing economic integration between countries, leading to the

36

Chapter Two

emergence of a global marketplace or a single world market. If globalization means a greater interconnection among regions and economies, we can interpret globalization from a historical perspective and recognize that it is not something new. In a similar way, it can be recognized that the current economic and financial crisis has been transmitted worldwide in a different manner than the preceding ones, but is not the first collapse of a financial bubble. Economic theories have always tried to explain economic crises in order to learn from the experiences. In general, economic crises have been the products of several steps that begin with a currency crisis, lead to a financial crisis and, finally, turn into a more generalized crisis, featuring a decline in consumption with a consequent slowing down of economic activity levels. Nowadays, the development of international linkages, leading to a growing economic interdependence, has contributed to the fact that the current economic crisis has become global.

References Alesina, A; Spolaore, E. and Warcziag, R. (2005). “Trade, growth and the size of countries”, chapter 3 in Handbook of Economic Growth, Vol. 1 B (edited by Aghion, P. and Durlauf, S.N.), Elsevier. Baily, M. N. and Elliott, D. J. (2009). "The U.S. Financial and Economic Crisis: Where Does It Stand and Where Do We Go From Here?", Brooking Institution Report. http://www.brookings.edu/papers/2009/0615_economic_crisis_baily_e lliott.aspx Chang, R. and Velasco, A. (2000). “Financial fragility and the exchange rate regime”, Journal of Economic Theory 82, 1-34. Chase-Dunn, Ch.K. (1991). Global Formation: Structures of the World Economy, Basil Blackwell, London. Bernanke, B. S. (2006). “Global Economic Integration: What's New and What's Not?”, Speech at the Federal Reserve Bank of Kansas City's Thirtieth Annual Economic Symposium, Jackson Hole, Wyoming. http://www.federalreserve.gov/newsevents/speech/bernanke20060825a .htm Blaug, M. (2001). “No History of ideas, please, we’re economists”, Journal of Economic Perspectives 15 (1), 145-164. Friedman, T. (2005). The World is Flat. A Brief History of the TwentyFirst Century, Farrar, Straus and Giroux (eds.), New York. Garber, P. M. (2000). Famous First Bubbles, The Massachusets Institute of Tecnology Press, Cambridge (MA).

Recent Episodes on a Globalized Economy: An Historical Approach

37

International Monetary Fund (2000). Globalization: Threats or Opportunity Kindleberger, C. P. (1991). Manias, Panics, and Crashs: Una historia de las crisis financieras, Ariel, Barcelona. Krugman, P. R. (1990). Rethinking International Trade, Cambridge, MA, MIT Press. Krugman, P. (1999). “¿Hay alguna salida?”, in La globalización de la economía y las crisis financieras, Lecciones de Economía Pedro Barrié de la Maza, Instituto de Estudios Económicos de Galicia, A Coruña, 31-39. Krugman P. R. (1996). “Making sense of the competitiveness debate”, Oxford Review of Economic Policy vol. 12, nº 3, 17-25. Krugman, P. (2002). “Crises: the next generation” in Assaf Razin, Elhanan Helpman, and Efraim Sadka, eds., Economic policy in the international economy: essays in honor of Assaf Razin, Cambridge. Kucukaksoy, I. (2011). “Adam Smith’s conceptual contributions to international economics: Based on the Wealth of Nations”, Business and Economic Horizons vol. 4, nº 1, 108-119. Le Goff, Jacques (1980). Time, work and culture in the Middle Ages, University of Chicago Press, Chicago and London. Mackay, C. (1841). Memoirs of Extraordinary Popular Delusions and the Madness of Crowds, London, Office of the National Illustrated Library http://www.econlib.org/library/Mackay/macExCover.html Minsky, H. (1994). “The financial instability hypothesis”, en P. Arestis y M. Sawyer (eds.), Handbook of Radical Political Economy, Edward Edgar Publishing, Cheltenham, U.K. Orgaz, L., Molina, L. and Carrasco, C. (2011). “El creciente peso de las economías emergentes en la economía y gobernanza mundiales, los países BRIC”, Documentos ocasionales, Nº 1101, Banco de España, Madrid. O´Neill, J. (2001). Building Better Global Economic BRIC, Goldman Sachs, Global Economic Paper 66. Reinhart C.M. and Rogoff, K.S. (2009). This time is different: Eight centuries of financial folly, Princeton University Press, Princeton. Sassen, S. (2006). Territory, Authority, Rights: From Medieval to Global Assemblages, Princeton University Press, Princeton. Temin, P. (2001). “A market economy in the early roman empire”, The Journal of Roman Studies, vol. 91, 169-181. Thompson, E. A. (2007). “The tulipmania: Fact or artifact?” Public Choice 130(1–2): 99-114.

38

Chapter Two

Wallerstein, I. (1974). The Modern World-System, vol. I: Capitalist Agriculture and the Origins of the European World-Economy in the Sixteenth Century, Academic Press, New York. —. (1980). The Modern World-System, vol. II: Mercantilism and the Consolidation of the European World-Economy, 1600-1750, Academic Press, New York. —. (1989). The Modern World-System, vol. III: The Second Great Expansion of the Capitalist World-Economy, 1730-1840's, Academic Press, San Diego. —. (2011). The Modern World-System, vol. IV: Centrist Liberalism Triumphant, 1789–1914. University of California Press, Berkeley.

CHAPTER THREE MONETARY UNIONS UNDER FINANCIAL SHOCKS: DO FISCAL RULES MATTER? CARMEN DÍAZ-ROLDÁN* AND ALBERTO MONTERO

The recent problems generated by the economic and financial crisis have led to some debate on the role of economic policies. The question is to what extent a specific monetary policy regime would impose a restriction on policymakers; in particular, the cost of losing independence in the use of the exchange rate and monetary policy, and the restrictions derived from the fiscal discipline required for supporting monetary agreements. As an example, we can think on the expected success of the European Economic and Monetary Union (EMU), related to the benefits of the single currency, the higher degree of integration of financial markets, and also the sound public finances guaranteed by the set of fiscal rules provided by EMU. When signing the Stability and Growth Pact (SGP), Member States committed themselves to reaching a medium-term budgetary position close to balance. The Maastricht Treaty stresses as basic that the Member States of the EMU should avoid excessive deficits, and the reference values for deficit-to-GDP and debt-to-GDP ratios have in fact worked as an explicit fiscal rule. But, in practice, the policy orientation of the SGP has not been fully satisfied. This has opened a debate about the utility and effectiveness of fiscal rules in EMU, and on their complementarities with discretionary fiscal policy measures and *

Corresponding author: Carmen Díaz-Roldán. Facultad de Derecho y Ciencias Sociales. Universidad de Castilla-La Mancha. 13071 Ciudad Real (Spain). Tel: +34-926-295300 Ext. 6657. E-mail: [email protected].

40

Chapter Three

automatic stabilisers to deal with short-run fluctuations. The aim of this chapter is to investigate how to deal with monetary (financial) shocks in a monetary union following fiscal rules. In particular, we will analyse the interaction among those members showing a relatively high level of public debt and those that seem to follow a more strict fiscal discipline.

3.1. Introduction The recent financial crisis is considered to be the worst crisis since the Great Depression of the 1930s. After the collapse of financial institutions there, has been a decline in economic activity and an increase of unemployment that have contributed to a global economic recession. There are several explanations for such a big crisis (see Reinhart and Rogoff (2009) for a survey of financial crises), but there is no consensus about how it could be avoided. The 2007 financial crisis spread into a global economic shock and it was transmitted to the European Economic and Monetary Union (EMU). In 2009, Eurozone growth became negative. The macroeconomic policy responses have focused mainly on short-term actions such as expanding money supplies and implementing large fiscal stimulus packages. Both the U.S. Federal Reserve and the European Central Bank have undertaken the largest monetary policy action in world history. Regarding long-term responses, no significant measure has been implemented. In particular, the lack of fundamental changes in the banking and financial markets is one of the main concerns of some contributions to the International Monetary Fund’s publications (see Blanchard and MilesiFerreti (2009) and Merrouche and Nier (2010), among others). The macroeconomic problems generated by the economic and financial crisis have led to some debate on the role of economic policies. But most of the macroeconomic models do not seem to capture specifically the role of financial markets. As far as we know, a financial crisis is generally modelled as a monetary negative shock, and theoretical findings reveal that the effects of the monetary (financial) shocks depend on the international linkages and channels of transmission of monetary policy, which are related to the particular exchange rate regime (see Díaz-Roldán (2004), for example). Regarding exchange rate regimes, some experiences (such as the speculative attacks on currencies in the European Monetary System in 1992-1993, the default on Mexican debt in 1994, the devaluations and the banking crises across Asia in 1997-1998, the Argentine crises in 2001, and the financial crisis of 2007 … followed by a global recession) have shown the increasing difficulty for a country to build the reputation needed to

Monetary Unions under Financial Shocks: Do Fiscal Rules Matter?

41

sustain a fixed exchange rate system. The ultimate reason is the spectacular growth of world capital markets, following the continuous liberalization and deregulation of capital movements that have occurred in the last years. So, if a government’s compromise of maintaining a certain exchange rate is not believed as credible by financial markets, huge speculative attacks at such a massive scale can occur. All this has led to some authors (e.g., Obstfeld and Rogoff, 1995) suggesting that, in the near future, the choice faced by a country will be either to maintain a flexible exchange rate or adopt a common currency, rather than a fixed exchange rate with other related countries. Moreover, from a macroeconomic point of view, it is clear that a system of fixed exchange rates (and full capital mobility) implies that there is only one system-wide monetary policy. National currencies would become perfect substitutes through the irrevocable fixing of exchange rates if they became equally appropriate for the three classical functions of money, namely: unit of account, store of value and medium of exchange. For those reasons, a monetary union has been suggested as an alternative to a system of fixed exchange rates. As an example, the expected success of the EMU was related to the benefits of the single currency, the higher degree of integration of financial markets, and also to the sound public finances guaranteed by the set of fiscal rules provided by the EMU. When signing the Stability and Growth Pact (SGP), Member States committed themselves to reaching a mediumterm budgetary position close to balance. The Maastricht Treaty stresses as basic that the Member States of the EMU should avoid excessive deficits, and the reference values for deficit-to-GDP and debt-to-GDP ratios have worked, in practice, as an explicit fiscal rule. But, in practice, the policy orientation of the SGP has not been fully satisfied. This has opened a debate about the utility and effectiveness of fiscal rules in the EMU, and on their complementarities with discretionary fiscal policy measures and automatic stabilisers to deal with short-run fluctuations.

3.2. The Fiscal Rules in Monetary Unions In the EMU, fiscal policy is the only available stabilization tool and the implementation of fiscal policies is constrained by fiscal discipline. In such an environment, fiscal rules have been considered a useful way of combining the achievement of policy aims with discipline and control over economic variables. In academic circles, the studies that have explicitly considered the needed of fiscal rules are scarce. As far as we know, Ballabriga and Martínez-Mongay (2003) estimated monetary and fiscal rules for the eurozone, concluding that monetary policy rules should be

42

Chapter Three

accompanied by an explicit public deficit objective. Debrun et al. (2008) study the relationship between fiscal discipline and fiscal rules in the EU25, and they have found that fiscal rules lead to more stable budget policies and less pro-cyclical fiscal policies. Brzozowski and SiwiĔskaGorzelak, (2010) analyse the impact of fiscal rules on fiscal policy volatility. From their results, they conclude that rules based on deficit control are more destabilizing than those based on imposing a limit to public debt. From another point of view, Díaz-Roldán and Montero-Soler (2009, 2011) also explore the usefulness of fiscal rules. In particular, DíazRoldán and Montero-Soler (2011) analyse the convenience of using fiscal rules for the New Member States (NMS) of the EMU, and they found that the success of fiscal policy decisions depend on the symmetric or asymmetric nature of the shocks to deal with. More recently, Díaz-Roldán (2013) has analysed the convenience of using fiscal rules for different subsets of countries of the EMU, concluding that the success of fiscal policy decisions depends on the symmetric or asymmetric nature of the shocks to deal with, and also on the particular economic conditions of each country. But, in practice, after the financial and economic crisis, fiscal rules have recovered a new interest due to the potential usefulness that fiscal rules could provide in helping economies to recover. Schaechter et al. (2012) present details about the rules’ key design elements – particularly in support of their enforcement – and conclude that the “next generation” fiscal rules are increasingly complex as they should combine the objectives of sustainability with the need for flexibility in response to shocks, thereby creating new challenges for implementation, communication, and monitoring. From another point of view, and with the EMU debt crisis in mind, Wyplosz (2013) has asserted that solving the fiscal discipline problem requires adequate institutions or rules, or both. On the one hand, the large risk posed by fiscal imbalances to any monetary area’s stability justifies close rules-based coordination in budgetary policies. But, on the other hand, the fiscal discipline imposed by the monetary agreements could limit the scope of fiscal policies for stabilization and their implications for economic growth. Summing up, in a monetary union, fiscal policy is the only demand policy aimed at achieving the stabilization goal, and monetary autonomy is obtained at the cost of losing direct control over the exchange rates. Therefore, member states of a monetary union face special difficulties when dealing with external shocks. In the EMU, fiscal policy is oriented to achieve output stabilization in the short-run, through the use of the public deficit and automatic insurance mechanisms. In the long-run, the fiscal policy should guarantee the sustainability of public finances, and it should

Monetary Unions under Financial Shocks: Do Fiscal Rules Matter?

43

also contribute to economic growth through the structure of revenues and expenditures, and the public investment in physical and human capital (European Central Bank, 2004). However, as mentioned before, in the EMU, the management of fiscal policy is constrained by the limits imposed on the deficit and the lack of a federal budget. With these considerations in mind, we are interested in analysing the possibilities of fiscal policy when dealing with demand (financial) shocks. In addition, one of our contributions will also be to explore such issues in the novel economic framework provided by a monetary union scenario, where we will consider the possibility of following an explicit fiscal rule to guarantee a medium-term budgetary position close to balance. To that aim, this chapter is devoted to investigating how to deal with monetary (financial) shocks in a monetary union. In particular, we will analyse the interaction among those members showing a relatively high level of public debt and those that seem to follow a more strict fiscal discipline. We will examine the consequences of monetary (financial) shocks when there is a single monetary policy and the domestic authorities are constrained by the fiscal discipline imposed by the monetary agreements of a monetary union following an explicit fiscal policy rule.

3.3. The Macroeconomic Model Our main purpose is to analyse and investigate how to deal with monetary (financial) shocks in a monetary union, when the central bank and the fiscal authorities follow policy rules. To that end, we will develop a simple model for a monetary union to analyse different combinations of fiscal policies. We will follow Díaz-Roldán and Montero-Soler (2009). Our starting point is a “small” monetary union formed by two symmetric countries, where a nominal exchange rate disappears among countries. Variables are defined as logarithmic deviations from their equilibrium levels (see Appendix for details). The aggregate demand and the aggregate supply functions for each country are as follows:

y1 y2

 a'p1 r b'p 2 r cy 2  hg 1  v1  a'p 2 r b'p1 r cy1  hg 2  v 2 y1 t'p1  s1 y 2 t'p 2  s 2

(1) (2) (3) (4)

44

Chapter Three

Equations (1) and (2) represent the aggregate demand function for each member country of the monetary union, where y1 and y2 are outputs; 'p1 and 'p2 are inflation rates; g1 and g2 are the budget deficits, i.e., the fiscal policy instrument; and v1 and v2 capture any kind of expansionary demand shock. Equations (3) and (4) represent the aggregate supply function for each member country of the monetary union, where s1 and s2 capture any kind of contractive supply shock. We find that a contractive supply shock (s1, s2 > 0), always leads to a fall in output in both countries, while positive demand shocks (v1, v2 > 0) lead to positive effects on the output and prices of the country of origin of the shock. But when the demand shock is transmitted across the two countries, the sign of the coefficients depends on which channel of transmission prevails. The channels of transmission of such disturbances are those related with international trade: the aggregate demand, the interest rate, and the real exchange rate (i.e., relative prices in monetary unions). When a country’s aggregate demand increases, imports of foreign goods also increase, and the result is the called “locomotive” effect, i.e., the effects on the output and prices of the country of origin of the shock are transmitted to the other country with the same sign. In our case, we would find an aggregate demand expansion with an output expansion and a rise in prices in the two countries. Regarding the interest rate, the development and integration of financial markets implies that interest rates are determined by world markets. High interest rates attract cross-border investments until equilibrium is reached, but under imperfect capital mobility, domestic interest rates will diverge from world interest rates, inducing capital inflows or outflows, depending on whether they are higher or lower than the world rates. When changes in the real exchange rate prevail, the result is the “beggar-thy-neighbour” effect, i.e., the effects on the output and prices of one country are transmitted abroad with the opposite sign. The reason is that a real exchange rate depreciation (appreciation) in an economy means an appreciation (depreciation) in the other, which leads to an aggregate demand expansion (recession) in that economy, and to a recession (expansion) in the other. Looking at the coefficients of the equations of the model (see Appendix) the “beggar-thy-neighbour” effect prevails when countries are particularly concerned by inflation targeting and output stabilization. This would be the case for a monetary union following a monetary policy rule. Solving (1) to (4), we obtain the reduced forms:

y1 = A hg1 + A v1 ± B hg2 ± B v2  C s1  D s2

(5)

Monetary Unions under Financial Shocks: Do Fiscal Rules Matter?

y2 = A hg2 + Av2 ± B hg1 ± B v1  C s2  D s1 'p1 = A’hg1 + A’v1 + B’hg2 + B’v2 + C’s1 + D’s2 'p2 = A’hg2 + A’v2 + B’hg1 + B’v1 + C’ s2 + D’s1

45

(6) (7) (8)

To take into account the role of fiscal rules, we will follow Ballabriga and Martínez-Mongay (2003). So, we will consider a fiscal rule which relates an explicit public deficit target (in terms of the GDP), go, with public debt deviations (in terms of the GDP) with respect to its optimal level (d-1 – do), and the output level y:

g io

 [G ( d i , 1  d io )  Ty i ] i = 1, 2

(9)

The public deficit adjusts according to the following path, where

0 d U d 1: gi

(1  U ) g io  Ug i , 1

(10)

Adding together the variables that are given in period 1, we obtain the simplified fiscal rules for each member country of the union:

g1 g2 ( d i , 1  d io )

k1  Oy1 k 2  Oy 2

(11) (12)

> 0, then ki < 0, indicating a country with a Notice that if relatively high level of debt. And the opposite holds for ki > 0, indicating a country with a relatively low level of debt. We will assume that fiscal authorities will try to minimize their loss function constrained by the economic framework (given by the reduced form of the macroeconomic model), and the explicit fiscal rule. Their goals are to minimize output changes, with stabilization purposes, and to minimize public deficit changes in order to achieve fiscal discipline. Regarding inflation, since our model describes a monetary union, we assume full delegation of prices control to the monetary authority; therefore, public deficit will be the only policy instrument available. In this framework, the set of policy makers’ decisions are the following: (i) Independent decision and no fiscal rule in any country, (ii) Coordinated decision and no fiscal rule in any country, (iii) Independent decision and fiscal rule in both countries, (iv) Coordinated decision and

Chapter Three

46

fiscal rule in both countries, and (v) Coordinated decision and fiscal rule only in one country. Notice we are assuming that coordination (the cooperative symmetric solution) means to implement an identical policy response. The most interesting is case (v), because (i) to (iv) give trivial results. In the case of no country adopting fiscal rules (or the adoption of fiscal rules in both countries), cooperation would not be the best solution when the shocks have asymmetric effects on the output. The reason is that when facing shocks which can lead to different effects, the best policy response would be to use different fiscal policies. When both countries adopt a fiscal rule, the results differ from the case with no fiscal rules only in the size of the coefficients: graphically, it is just a change of scale. Next, we show the optimization problems. Solving them, we will obtain the set of optimal (fiscal) policies, i.e., the optimal level of public deficit (see Appendix for details). (i) Independent decision and no fiscal rule in any country

y i2  Vg i2

min Li gi

s.t. yi = yi (…) i = 1,2. where Li

yi2  Vg i2 i 1, 2 is the loss function of the fiscal authority.

And in order to describe the concern regarding deficit control, we assume V > 1. N

Solution (see Appendix): g1

G1N v1 r G2N v 2  G3N s1 r G 4N s 2

(ii) Coordinated decision and no fiscal rule in any country

1 º ª1 «¬ 2 L1  2 L2 »¼

min " g1 , g 2

s.t. y1 = y1 (…) y2 = y2 (…) C

Solution (see Appendix): g1

G1C v1 r G 2C v 2 r G3C s1 r G 4C s 2

Monetary Unions under Financial Shocks: Do Fiscal Rules Matter?

47

(iii) Independent decision and fiscal rule in both countries

min Li gi

yi2  Vg i2

s.t. yi = yi (…) gi = gi (…) i = 1,2. Solution (see Appendix):

g1N , R

G1N , R v1 r G2N , R v2  G3N , R s1 r G4N , R s 2

(iv) Coordinated decision and fiscal rule in both countries

min "

g1 , g 2

1 º ª1 « 2 L1  2 L2 » ¬ ¼

s.t. y1 = y1 (…) y2 = y2 (…) g1 = g1 (…) g2 = g2 (…) Solution (see Appendix):

g1C , R

G1C , R v1 r G 2C , R v 2 r G3C , R s1 r G4C , R s 2

(v) Coordinated decision and fiscal rule only in one country, we assume that country 1 follows the fiscal rule:

min " g1 , g 2

1 º ª1 « 2 L1  2 L2 » ¬ ¼ s.t. y1 = y1 (...) y2 = y2 (...)

g1

k1  Oy1

Solution (see Appendix); notice we found no symmetric solution.

g1C , R g 2C

G1C , R v1  G2C , R v 2 r G3C , R s1 r G4C , R s 2 G1C v 2 r G2C v1 r G3C s 2 r G 4C s1

Given that fiscal authorities aim to minimize public deficit changes, in order to achieve fiscal discipline, the best response would be the one showing minimal deviation from the equilibrium level. Since it would be tedious to compare the coefficients for each solution, in the next section we will perform an empirical application. We will assign numeric values to the coefficients, in order to evaluate the loss of adopting the policy responses given by the analytical solutions.

48

Chapter Three

3.4. The Costs and Benefits of Following a Fiscal Rule As we mentioned in the introduction, there is a debate about the utility and effectiveness of fiscal rules, and on their complementarities with discretionary fiscal policy measures and automatic stabilisers to deal with short-run fluctuations. Particularly, in EMU, the Maastricht Treaty stressed as basic that the Member States of EMU should avoid excessive deficits, and the reference values for deficit-to-GDP and debt-to-GDP ratios have worked in practice as an explicit fiscal rule. But the success of any kind of policy remains an empirical question. In Tables 3-1, the government deficit ()/surplus (+) and government debt in percentage of GDP is shown for the EU-27. In 2013, the government deficit and government debt of the EU-27 was 3.3 and 87.4 respectively (both in percentage of the GDP)1. These figures are above the 3 and 60 limits required by the Maastricht Treaty. Moreover, the recent financial crisis is not a good environment, and contributes to create difficulties when deciding how to finance the public deficit. In such a context, the scope of fiscal policies in a monetary union seems to be reduced. If we look at the figures, in 2023, 14 of the EU-27 countries show debt figures above 60% (Belgium, Germany, Ireland, Greece, Spain, France, Italy, Cyprus, Hungary, Malta, the Netherlands, Austria, Portugal and the United Kingdom). Regarding the deficit ratio, nine countries (Ireland, Greece, Spain, France, Cyprus, Poland, Portugal, Slovenia and the United Kingdom) show figures above 3%. In order to make an empirical application of our theoretical findings, we will make the following assumptions. The shocks suffered by the countries are identical in size (normalized to 1); in other words, they are perfectly symmetric in size. The shocks may differ in the sign: expansive (+) or contractive (); so they are perfectly asymmetric in their effects. Next, we will give numerical values to the parameters of the reduced form according to the following criteria. In the fiscal rule, the response of the public deficit to changes in output will be neutral (O = 0.5) to underline the relevance of the debt level: higher than the target (k = 0.9) or lower (k = 0.9). For comparability reasons, we assign the value 1 to the aggregate supply slope (t = 1), and in the loss function, we assume that fiscal authorities are more concerned about fiscal discipline, than about stability (V = 1.3)2.

1 2

Source: Eurostat. For the rest of the values see Appendix.

Monetary Unions under Financial Shocks: Do Fiscal Rules Matter?

49

Table 3-1: Government Deficit (-)/Surplus (+) and Debt in the EU-27 (% of GDP) 2006

2007

2008

2009

2010

2011

2012

2013

-1,5 61,5

-0,9 58,9

-2,4 62,2

-6,9 74,5

-6,5 80,2

-4,4 82,7

-3,9 85,5

-3,3 87,4

0,4 87,9

-0,1 84

-1,0 89,2

-5,6 96,6

-3,8 96,6

-3,8 99,2

-4,1 101,1

-2,6 101,5

1,9 21,6

1,2 17,2

1,7 13,7

-4,3 14,6

-3,1 16,2

-2,0 16,3

-0,8 18,4

-1,5 18,9

-2,4 28,3

-0,7 27,9

-2,2 28,7

-5,8 34,6

-4,7 38,4

-3,2 41,4

-4,2 46,2

-1,5 46

5,2 32,1

4,8 27,1

3,2 33,4

-2,7 40,7

-2,5 42,8

-1,9 46,4

-3,8 45,4

-0,8 44,5

-1,6 68

0,2 65,2

-0,1 66,8

-3,1 74,6

-4,2 82,5

-0,8 80

0,1 81

0,0 78,4

2,5 4,4

2,4 3,7

-3,0 4,5

-2,0 7,1

0,2 6,7

1,1 6,1

-0,2 9,8

-0,2 10

2,9 24,6

0,2 24,9

-7,4 44,2

-13,7 64,4

-30,6 91,2

-13,1 104,1

-8,2 117,4

-7,2 123,7

-5,7 106,1

-6,5 107,4

-9,8 112,9

-15,7 129,7

-10,9 148,3

-9,6 170,3

-8,9 157,2

-12,7 175,1

2,4 39,7

2,0 36,3

-4,5 40,2

-11,1 54

-9,6 61,7

-9,6 70,5

-10,6 86

-7,1 93,9

-2,3 63,7

-2,7 64,2

-3,3 68,2

-7,5 79,2

-7,0 82,7

-5,2 86,2

-4,9 90,6

-4,3 93,5

-3,4 106,3

-1,6 103,3

-2,7 106,1

-5,5 116,4

-4,5 119,3

-3,7 120,7

-3,0 127

-3,0 132,6

-1,2 64,7

3,5 58,8

0,9 48,9

-6,1 58,5

-5,3 61,3

-6,3 71,5

-6,4 86,6

-5,4 111,7

-0,6 10,7

-0,7 9

-4,4 19,8

-9,2 36,9

-8,2 44,5

-3,5 42

-1,3 40,8

-1,0 38,1

EU-27

Deficit/surplus Debt BELGIUM

Deficit/surplus Debt BULGARIA

Deficit/surplus Debt CZECH REPUBLIC

Deficit/surplus Debt DENMARK

Deficit/surplus Debt GERMANY

Deficit/surplus Debt ESTONIA

Deficit/surplus Debt IRELAND

Deficit/surplus Debt GREECE

Deficit/surplus Debt SPAIN

Deficit/surplus Debt FRANCE

Deficit/surplus Debt ITALY

Deficit/surplus Debt CYPRUS

Deficit/surplus Debt LATVIA

Deficit/surplus Debt Source: Eurostat

Chapter Three

50

Table 3-1 (cont.): Government Deficit (-)/Surplus (+) and Debt in the EU-27 (% of GDP) 2006

2007

2008

2009

2010

2011

2012

2013

-0,4 17,9

-1,0 16,8

-3,3 15,5

-9,4 29,3

-7,2 37,8

-5,5 38,3

-3,2 40,5

-2,2 39,4

1,4 6,7

3,7 6,7

3,2 14,4

-0,7 15,5

-0,8 19,5

0,2 18,7

0,0 21,7

0,1 23,1

-9,4 65,9

-5,1 67

-3,7 73

-4,6 79,8

-4,3 82,2

4,3 82,1

-2,1 79,8

-2,2 79,2

-2,7 62,5

-2,3 60,7

-4,6 60,9

-3,7 66,5

-3,5 66

-2,7 68,8

-3,3 70,8

-2,8 73

0,5 47,4

0,2 45,3

0,5 58,5

-5,6 60,8

-5,1 63,4

-4,3 65,7

-4,1 71,3

-2,5 73,5

-1,5 62,3

-0,9 60,2

-0,9 63,8

-4,1 69,2

-4,5 72,5

-2,5 73,1

-2,6 74,4

-1,5 74,5

-3,6 47,7

-1,9 45

-3,7 47,1

-7,5 50,9

-7,8 54,9

-5,1 56,2

-3,9 55,6

-4,3 57

-4,6 69,4

-3,1 68,4

-3,6 71,7

-10,2 83,7

-9,8 94

-4,3 108,2

-6,4 124,1

-4,9 129

-2,2 12,4

-2,9 12,8

-5,7 13,4

-9,0 23,6

-6,8 30,5

-5,5 34,7

-3,0 38

-2,3 38,4

-1,4 26,4

0,0 23,1

-1,9 22

-6,3 35,2

-5,9 38,7

-6,4 47,1

-4,0 54,4

-14,7 71,7

-3,2 30,5

-1,8 29,6

-2,1 27,9

-8,0 35,6

-7,5 41

-4,8 43,6

-4,5 52,7

-2,8 55,4

4,2 39,6

5,3 35,2

4,4 33,9

-2,5 43,5

-2,5 48,8

-0,7 49,3

-1,8 53,6

-2,1 57

2,3 45,2

3,6 40,2

2,2 38,8

-0,7 42,6

0,3 39,4

0,2 38,6

-0,6 38,3

-1,1 40,6

-2,8 42,7

-2,8 43,7

-5,0 51,9

-11,4 67,1

-10,0 78,4

-7,6 84,3

-6,1 89,1

-5,8 90,6

LITHUANIA

Deficit/surplus Debt LUXEMBOURG

Deficit/surplus Debt HUNGARY

Deficit/surplus Debt MALTA

Deficit/surplus Debt NETHERLAND

Deficit/surplus Debt AUSTRIA

Deficit/surplus Debt POLAND

Deficit/surplus Debt PORTUGAL

Deficit/surplus Debt ROMANIA

Deficit/surplus Debt SLOVENIA

Deficit/surplus Debt SLOVAKIA

Deficit/surplus Debt FINLAND

Deficit/surplus Debt SWEDEN

Deficit/surplus Debt UNITED KINGDOM

Deficit/surplus Debt Source: Eurostat

Monetary Unions under Financial Shocks: Do Fiscal Rules Matter?

51

In terms of a macroeconomic model, the current economic crisis, due to the problems of the financial system and the difficulty in borrowing, has led to increases in the real interest rate (a common contractive monetary shock), decreasing the level of income of the economy. The attempts to deal with the crisis by implementing expansionary demand policies would result in an increase of public deficit and debt, which are already very high in most advanced countries. Particularly in the EU countries, contractive fiscal policies are being implemented to reduce the size of government deficits and to recover the confidence of financial markets, in an attempt to avoid the risk of debt default. But it is also true that a contractive fiscal policy will tend to cause a drop in activity levels, exacerbating the recession and making further deficit reduction more difficult. According to the loss functions of the optimization problems described in section 3.2, in Table 3-2 we have computed losses when the countries of the monetary union are hit by a common contractive demand shock (v1 < 0 + v2 < 0), leading to contractive effects on output (y1=  1.4103, y2=  0.2565) and on inflation ('p1=  1.7317, 'p2=  1.0372), but different in size. Table 3-2: Losses after a Common Contractive Demand Shock (v1 < 0 + v2 < 0) NO FISCAL RULES IN ANY COUNTRY

FISCAL RULES IN BOTH COUNTRIES

BEGGAR-THY-

L1 = 2.7155

L1 = 2.9522

L1 = 2.0594

L1 = 3.0522

NEIGHBOUR

L2 = 2.0448

L2 = 2.1147

L2 = 0.0643

L2 = 2.1233

"

= 2.0076

" = 2.0359

"

"

LOCOMOTIVE

L1 = 2.8646

L1 = 2.9575

L1 = 3.9831

L1 = 3.8749

EFFECT

L2 = 0.0564

L2 = 0.0483

L2 = 3.4963

L2 = 3.2447

" = 3.2466

"

"

"

EFFECT

= 3.1298

FISCAL RULE ONLY IN ONE COUNTRY (Country 1) High debt Low debt k1 < 0 k1 > 0

= 2.6688

= 3.2749

= 2.0433

= 3.1316

Note: L1 and L2 are losses when countries act individually (Nash solution). While " is the loss when countries act in a coordinated way (Cooperative solution). In the light coloured cells, the best response is not to implement an identical fiscal policy.

From the figures in Table 3-2, we can conclude that the best fiscal policy response would be to:

52

Chapter Three

Ɣ Not cooperate when only those countries with a high debt adopt a fiscal rule (but not the rest of the countries), the whole union has suffered a common financial shock (losses are larger for the cooperative solution), and the beggar-thy-neighbour effect prevails. In these cases, the cooperative solution ( " ) shows a higher loss than the Nash solution (L1 and L2). Ɣ Not cooperate if no country (or all the countries) have adopted a fiscal rule, and the whole union has suffered a common financial shock (losses are larger for the cooperative solution), when the locomotive effect prevails. Again, the cooperative solution ( " ) shows a higher loss than the Nash solution (L1 and L2).

3.5. Summary and Conclusions The macroeconomic problems generated by the economic and financial crisis have led to some debate on the role of economic policies. After the collapse of financial institutions, there has been a decline in economic activity and an increase of unemployment that have contributed to a global economic recession. The macroeconomic policy responses have focused mainly on short-term actions such as expanding money supplies and implementing large fiscal stimulus packages. But no significant measure has been implemented as a long-term response, representing the lack of fundamental changes in the banking and financial markets, which is one of the main concerns of international institutions. These problems have a special dimension in monetary unions, given that fiscal policy is the only demand policy aimed at achieving the stabilization goal, and monetary autonomy is obtained at the cost of losing direct control over the exchange rates. Therefore, member states of a monetary union would face special difficulties when dealing with external shocks. On the one hand, the large risk posed by fiscal imbalances to any monetary area’s stability justifies close rules-based coordination in budgetary policies. But, on the other hand, the fiscal discipline imposed by the monetary agreements could limit the scope of fiscal policies for stabilization, and their implications for economic growth. In such an environment, fiscal rules have recovered a new interest due to the potential usefulness that fiscal rules could provide in helping economies to recover.

Monetary Unions under Financial Shocks: Do Fiscal Rules Matter?

53

Appendix The Macroeconomic Model y1

Dr  E ( p 2  p1 )  Jy 2  g 1  f1

(1.A)

y2

Dr  E ( p1  p 2 )  Jy1  g 2  f 2

(2.A)

ª1 'p1  'p 2  'p o º»  H y1  y 2 ¬2 ¼ 2

(3.A)

r

P«

From (1.A) to (3.A), we obtain the aggregate demand for each country.

'w1

'p1Ec  Mu1  I'prod 1  z1w

(4.A)

'p1 'w1  I'prod1  z1p n1 y1  prod 1 p1Ec p1c , 1

(5.A) (6.A) (7.A)

p1c Kp1  (1  K ) p 2 u1 { l1  n1

(8.A) (9.A)

From (4.A) to (9.A), we obtain the aggregate supply for each country. The “beggar-thy-neighbour” effect prevails when countries are particularly concerned by inflation targeting and output stabilization (P and H, in the monetary rule – equation (3.A) – are high enough).

Aggregate Demand Coefficients

a

h vi

DP  2 E 2div

b

DP  2 E 2div

c

DH  2J 2div

, where

div

1 div f D DP O E E  'p  rA  p 2 , 1  p1, 1  i div div div div div

2  DH 2

,

Chapter Three

54

Aggregate Supply Coefficients

t

1

M

si

1 1  'pc , 1  l  prod  ( z p  z w )

M

M

Reduced Form y1 = 

(a  t )

t (hg1 + v1) ±

(b  ct )

den

t (hg2 + v2) 

( a  bc ) t  a

2

b

2 s1 

den

den

( b  ac )

t s2

den

(10.A) y2 = 

(a  t )

t (hg2 + v2) ±

(b  ct )

den

t (hg1 + v1) 

( a  bc ) t  a

den

2

b

2 s2 

den

( b  ac )

t s1

den

(11.A) 'p1 = 

(a  t )

(hg1 + v1) +

(b  ct )

den

(hg2 + v2) +

( b  ct ) c  ( a  t )

den

s1 +

den

( b  ac )

s2

den

(12.A) 'p2 = 

(a  t )

(hg2 + v2) +

(b  ct )

den

(hg1 + v1) +

( b  ct ) c  ( a  t )

den

den

den = (ct +b)2 – (a + t)2 < 0 (a  t )

t > 0, B =

den

A’= 

> 0, B’=

(b  ac ) den

2

b

2

den

>0

den and D’=

( a  bc ) t  a

t

den

(a  t )

t > 0, C =

den

( b  ac ) and D =

(b  ct )

(b  ct ) den

>0

> 0, C’=

(b  ct ) c  ( a  t ) den

( b  ac ) den

(13.A)

A=

s2 +

>0

>0

s1

Monetary Unions under Financial Shocks: Do Fiscal Rules Matter?

55

Independent Decision and No Fiscal Rule in Any Country Reaction functions: N q1

ABh

2

2 2 [A h  V ]

ABh

N , q2

2 2 [A h  V ]

2 A h

N , q3

2 2 [A h  V ]

N , q4

ACh 2 2 [A h  V ]

N , q5

ADh 2 2 [A h  V ]

Nash: N G1

N G2



2 2 2 2 2 3 A h( A h  V )  A B h 2 2 2 2 2 ( A h  V )  ( ABh )

ABhV 2 2 2 2 2 ( A h  V )  ( ABh )

N G3

2 3 2 2 A BDh  ACh ( A h  V ) 2 2 2 2 2 ( A h  V )  ( ABh )

N G4

2 3 2 2 A BCh  ADh ( A h  V ) 2 2 2 2 2 ( A h  V )  ( ABh )

Coordinated Decision and No Fiscal Rule in Any Country C g1

C C C C C q1 g 2  q 2 v 2  q 3 v1  q 4 s1  q 5 s 2 C C C C C q1 , q 2 , q 3 , q 4 ! 0, q 5 g1 ( g 2 )

2 ABh

C q1 [( A

2

2

2 2  B )h  V ]

( AC  BD ) h

C q4 [( A

2

2

 B )h

2

 V]

2 ABh

C , q2 [( A

2

2 2  B )h  V ]

( AD  BC )

C , q5 [( A

2

(A

C , q3

2 2  B )h  V ]

[( A

2

2

2  B )h

2 2  B )h  V ]

Chapter Three

56

Cooperative solution: C G1



(A

2

2 2 2 2 2  B ) h[( A  B ) h  V ]  ( 2 ABh ) h 2 2 2 2 2 2 [( A  B ) h  V ]  ( 2 ABh )

2 2 2 2 2 2 2 2 ABh[( A  B ) h  V ]  h ( A  B ) 2 2 2 2 2 2 [( A  B ) h  V ]  ( 2 ABh )

C G2

C G3

3 2 2 2 2 ABh ( BC  AD )  [( A  B ) h  V ]h ( BD  AC ) 2 2 2 2 2 2 [( A  B ) h  V ]  ( 2 ABh )

C G4

3 2 2 2 2 ABh ( BD  AC )  [( A  B ) h  V ]h ( BC  AD ) 2 2 2 2 2 2 [( A  B ) h  V ]  ( 2 ABh )

Independent Decision and Fiscal Rule in Both Countries Reaction functions: N ,R q1

B

N ,R , q2

B

N ,R , q3

Ah

A

1

N ,R , q4

C

N ,R , q5

Ah

h

D

N ,R , q6

Ah

k1OV

2 Ah (1  O V )

Nash: 1

N ,R G1



N ,R G2

0

h



k1VO 2 h (1  VO )( A  B )

2 ( B  A)(1  VO )  k1VO 2 h (1  VO )( A  B )

k1VO 2 h (1  VO )( A  B )

k1VO  2 h ( A  B )( A  B ) h (1  VO )( A  B )

2 ( BD  AC )(1  VO )  k1VO ( A  B ) 2 h (1  VO )( A  B )( A  B )

k1VO 2 h (1  VO )( A  B )

2 ( BC  AD )(1  VO )  k1VO ( A  B ) 2 h (1  VO )( A  B )( A  B )

( BD  AC )

N ,R G3

( BC  AD ) N ,R G4

 h ( A  B )( A  B )

Monetary Unions under Financial Shocks: Do Fiscal Rules Matter?

57

Coordinated Decision and Fiscal Rule in Both Countries C,R g1

C,R C,R C,R C,R C,R C,R q1 g 2  q 2 v 2  q3 v1  q 4 s1  q5 s 2  q 6

g1 ( g 2 )

C ,R C ,R C ,R C ,R C ,R q1 ! 0, q 5 , q4 , q2 , q3

C ,R q6  0

when ( Ak 1 2 AB

C,R q1

2

2

C ,R q6 ! 0

2 AB

C,R , q2

2

2

C,R , q3

( A  B )h

( AD  BC ) 2 2 ( A  B )h

when

( Ak 1  Bk 2 )  0

 Bk 2 ) ! 0

(A  B ) C,R q5

and

C,R , q6

1 h

C,R q4

,

( AC  BD ) 2 2 ( A  B )h

( Ak1  Bk 2 ) OV 2 2 2 ( A  B ) h (1  O V )

Cooperative solution: 1

( Ak1  Bk 2 )VO 2 2 h (1  VO )( A  B )

2 2 (1  VO )( A  B )  ( Ak1  Bk 2 )VO 2 2 h (1  VO )( A  B )

C ,R G1



C ,R G2

0

C ,R G3

2 2 ( Ak1  Bk 2 )VO 2 AB ( BC  AD )  ( A  B )( BD  AC )  2 2 2 2 h( A  B) ( A  B) h (1  VO )( A  B )

h





( Ak1  Bk 2 )VO 2 2 h (1  VO )( A  B )

2 2 2 2 [ 2 AB ( BC  AD )  ( A  B )( BD  AC )](1  VO )  ( A  B ) ( Ak1  Bk 2 )VO 2 2 2 h (1  VO )( A  B ) ( A  B ) C ,R G4

2 2 ( Ak1  Bk 2 )VO 2 AB ( BD  AC )  ( A  B )( BC  AD )  2 2 2 2 h( A  B) ( A  B) h (1  VO )( A  B ) 2 2 2 2 [ 2 AB ( BD  AC )  ( A  B )( BC  AD )](1  VO )  ( A  B ) ( Ak1  Bk 2 )VO 2 2 2 h (1  VO )( A  B ) ( A  B )

Coordinated Decision and Fiscal Rule Only in One Country, We Assume that Country 1 Follows the Fiscal Rule C ,R g1

g1 ( g 2 )

C ,R C ,R C ,R C ,R C ,R C ,R q1 g 2  q 2 v 2  q 3 v1  q 4 s1  q 5 s 2  q 6

,

Chapter Three

58

C ,R C ,R C ,R C ,R C ,R C ,R ! 0 when k1 > 0 q1 , q 2 , q3 , q 4 , q5 ! 0 and q 6 C,R q1

C,R q4

C g2

2 AB ( 2  VO )

2 AB ( 2  VO )

C,R , q2 2 2 2 ( A (1  VO )  B ) 2 ( AC (1  VO )  BD ) 2 2 2 ( A (1  VO )  B ) h

,

2

2

C,R , q3

2

( A (1  VO )  B ) h 2 ( AD (1  VO )  BC )

C,R q5

2 2 2 ( A (1  VO )  B ) h

h VOk1 A

C,R q6

,

1

2 2 2 ( A (1  VO )  B ) h

C C C C C C q1 g 2  q 2 v 2  q 3 v1  q 4 s1  q 5 s 2  q 6 C C C C C C q1 , q 2 , q 3 , q 4 , q 5 ! 0 and q 6 ! 0 when k1 > 0 C q1

C q3

g 2 ( g1 )

2 2 ABh ( 2  VO ) 2 2 2 2 2 h [ A  B (1  VO )]  V

C q2

,

2 ABh ( 2  VO )

C q4

2 2 2 2 2 h [ A  B (1  VO )]  V

,

[A

2

2 2 2 2 2 h [ A  B (1  VO )]  V 2 ( AC  BD (1  VO )) 2 2 2 2 2 h [ A  B (1  VO )]  V

2

C q5

( AD  BC (1  VO )) 2

h [A

2

2

2

 B (1  VO )]  V

2

,

2 2  B (1  VO )]h ,

,

VOk1B

C q6

2

h [A

2

2 2 2  B (1  VO )]  V

Cooperative solution: Country 1: 2

2

2

2

2

2

VOk1 A[ B ( 2  VO )  h ( A  B (1  VO ))  V ]

reg1 = C,R G1

C,R G2

C,R G3

2 2 2 2 2 2 2 2 2 2 h[( A (1  VO )  B )( h ( A  B (1  VO ))  V )  ( ABh ( 2  VO )) ]



1 h



reg1 2 2 AB ( 2  VO )V



2 2 2 2 2 2 2 2 2 2 ( A (1  VO )  B )( h ( A  B (1  VO ))  V )  ( ABh ( 2  VO ))

 reg1

2 2 2 2 2 2 2 2 ABh ( 2  VO )[ AD  BC (1  VO )]  [ AC (1  VO )  BD ]( h ( A  B (1  VO ))  V ) 2 2 2 2 2 2 2 2 2 2 ( A (1  VO )  B )( h ( A  B (1  VO ))  V )  ( ABh ( 2  VO ))

 reg1

Monetary Unions under Financial Shocks: Do Fiscal Rules Matter?

C,R G4

59

2 2 2 2 2 2 2 2 ABh ( 2  VO )[ AC  BD (1  VO )]  [ AD (1  VO )  BC ]( h ( A  B (1  VO ))  V ) 2 2 2 2 2 2 2 2 2 2 ( A (1  VO )  B )( h ( A  B (1  VO ))  V )  ( ABh ( 2  VO ))

 reg1

Country 2: 2

reg2 =

C G1

C G2 C G3

C G4

2

2

2

2

VOk 1 B[ A ( 2  VO ) h  A (1  VO )  B ] 2 2 2 2 2 2 2 2 2 2 ( A (1  VO )  B )( h ( A  B (1  VO ))  V )  ( ABh ( 2  VO ))

2 2 2 2 2 2 2 2 ( ABh ( 2  VO ))  ( A  B (1  VO ))( A (1  VO )  B ) h 2 2 2 2 2 2 2 2 2 2 ( A (1  VO )  B )( h ( A  B (1  VO ))  V )  ( ABh ( 2  VO ))

 reg 2

 reg 2 2 2 2 2 2 2 2 2 ABh ( 2  VO )[ AD (1  VO )  BC ]  [ AC  BD (1  VO )]( h ( A  B (1  VO ))  V ) 2 2 2 2 2 2 2 2 2 2 ( A (1  VO )  B )( h ( A  B (1  VO ))  V )  ( ABh ( 2  VO ))

 reg 2

2 2 2 2 2 2 2 2 ABh ( 2  VO )[ AC (1  VO )  BD ]  [ AD  BC (1  VO )]( h ( A  B (1  VO ))  V ) 2 2 2 2 2 2 2 2 2 2 ( A (1  VO )  B )( h ( A  B (1  VO ))  V )  ( ABh ( 2  VO ))

Values for the empirical application Beggar-thy-neighbour H = 0.6 P = 0.8 D = 0.9 E = 0.1 J = 0.1 M=1 a = 0.3622 b = 0.2047 c = 0.1338 h = 0.7874 A = 0.7824 B = 0.1944 C = 0.2436 D = 0.0897 V = 1.3 O = 0.5 k1 =  0.9 or k1 = 0.9

Locomotive H = 0.6 P = 0.8 D = 0.9 E = 0.1 J = 0.1 M=1 a = 0.9126 b =  0.8349 c =  0.0679 h = 0.9708 A = 0.6727 B = 0.3175 C = 0.3488 D = 0.2718 V = 1.3 O = 0.5 k1 =  0.9 or k1 = 0.9

 reg 2

60

Chapter Three

References Ballabriga F. and Martínez-Mongay C. (2003). “Has EMU shifted monetary and fiscal policies?”, in Buti, M. (ed.): Monetary and fiscal policies in EMU. Interactions and coordination, Cambridge University Press. Blanchard, O. and Milesi-Ferreti, G.M. (2009). “Global imbalances: In Midstream?”, International Monetary Fund Staff Position Note 09/29. Brzozowski M. and SiwiĔska-Gorzelak J. (2010). “The impact of fiscal rules on fiscal policy volatility”, Journal of Applied Economics, Vol. 3, No. 2. Debrun X., Moulin L., Turrini A., Ayuso-i-Casals J. and Kumar M.S. (2008). “Tied to the mast? National fiscal rules in the European Union”, Economic Policy, Vol. 23, No. 4. Díaz-Roldán, C. (2004). “International monetary policy coordination under asymmetric shocks”, International Advances in Economic Research, Vol.10, No. 1, 72-82. Díaz-Roldán, C. and Montero-Soler, A. (2009). “Las políticas de demanda en uniones monetarias: ¿Son necesarias las reglas de política fiscal?”, paper presented at the XVI Encuentro de Economía Pública/ XVI Meeting on Public Economics, Granada. Díaz-Roldán, C. and Montero-Soler, A. (2011). “Fiscal policy in the European Monetary Union: How can fiscal discipline be achieved?”, Argumenta Oeconomica Vol 27 nº 2, pp. 5-14 Díaz-Roldán, C. (2013). “Déficit, deuda y reglas fiscales en una unión monetaria”, Cuadernos de Información Económica nº 233, 11-23. European Central Bank (2004). “Fiscal policy influences on macroeconomic stability and prices”, Monthly Bulletin, abril, 45-57. Merrouche, O. and Nier, E. (2010). “What caused the global financial crisis? Evidence on the drivers of financial imbalances 1999-2007”, International Monetary Fund Working Paper 10/265. Obstfeld, M. and Rogoff, K. (1995). “The miracle of fixed exchange rates”, Journal of Economic Perspectives 9, 73-96. Reinhart C.M. and Rogoff, K.S. (2009). This time is different: Eight centuries of financial folly, Princenton University Press, Princenton. Schaechter, A., Kinda, T., Budina, N., and Weber, A. (2012). “Fiscal Rules in Response to the Crisis.Toward the “Next-Generation” Rules”. IMF Working Paper, Fiscal Affairs Department, WP/12/187 Wyplosz, C. (2013). “Fiscal rules: Theoretical issues and historical experiences” Chapter 12 in Alesina, A. and Giavazzi, F. (ed.) Fiscal

Monetary Unions under Financial Shocks: Do Fiscal Rules Matter?

61

policy after the financial crisis, 495-525. National Bureau of Economic Research, University of Chicago Press.

CHAPTER FOUR A GENERALIZED NEG WAGE-TYPE EQUATION FERNANDO BRUNA*

This chapter discusses several controversial issues of the New Economic Geography (NEG) theory, focusing on some problems of interpretation regarding the estimation results of a wage equation. In order to do that, several wage equations found in the literature are encompassed under the derivation of a new generalized wage-type equation, with marginal cost as a dependent variable. A testable equation controlling for human and physical capital stocks is also derived.

4.1. Introduction1 The basic form of the so-called wage equation of the New Economic Geography (NEG) predicts that “nominal manufacturing wages” depend on the accessibility to markets, as captured by an index of Market Access or Market Potential. This relationship has been studied in a large empirical literature. However, the NEG “is not easy to test” (Head and Mayer, 2006). For reasons of tractability, the theory uses strong simplifying assumptions. More generally, the Marshallian or observational equivalence of the NEG refers to the difficulty in discerning between alternative theories of location. It seems that this situation has created several common misunderstandings about the NEG. This chapter discusses various reasons that hinder the connection between the estimation results of a wage equation and the specific explanation offered by the NEG. In particular, it is emphasized that the *

Universidad de A Coruña. E-mail: [email protected] I am grateful to the seminar participants at the University of A Coruña, the Autonomous University of Madrid and the XV Conference on International Economics. I especially thank Eduardo Giménez, Carlos Llano, José Luis Zofío and Jesus Lopez-Rodriguez for their helpful comments on previous drafts.

1

64

Chapter Four

“wage” equation is the result of imposing a market clearing condition on a profit equation. The dependent variable is actually marginal costs. Redding and Venables (2004) concluded that it is more accurately an equation for the price of the composite immobile factor of production. They interpreted that factor as labour. However, going from marginal costs to wages requires additional assumptions (Head and Mayer, 2004b; Combes et al., 2008, chap. 12) that might have not been sufficiently highlighted. The “wage equation” has been viewed as an explanation of the spatial distribution of different phenomena: “manufacturing wages”, under a literal interpretation of the basic model, or “economic activity” (Redding, 2011), in a broader sense. These alternative views imply different options when measuring the variable on the left-hand side of the testable equation. Redding and Venables (2004), as well as other authors, have chosen income per person. From a literal interpretation of the wage equation, income per capita would be a measure of welfare rather of factor prices. However, the robustness analysis of Redding and Venables (2004) using manufacturing wages per worker shows an extremely similar pattern of estimation results. Tests carried out by the author of this chapter (not shown) using regional European data also prove that the estimation of a wage equation is robust with regard to alternative dependent variables related to income or wages for the aggregate economy, manufacturing or services. The correlations between these variables are very high because crosssectional analysis is extremely sensitive to the relative levels of “development”. The higher the sample heterogeneity, the higher the correlations are. In other words, all these measures are affected by the level of total factor productivity. Of course, the NEG avoids explaining agglomeration through exogenous technological differences (Krugman, 2011). However, “technology”, and its spatial distribution, always underlies any measure of factor prices, income or Market Potential.2 These issues are discussed here in the context of a new generalized wage-type equation encompassing other wage equations found in the literature. Additionally, the derivation of a wage-type equation including human and physical capital is used to illustrate the difficulties when interpreting the estimation results. The model sketched here3 is mainly 2 See Duranton and Puga (2004), Head and Mayer (2004b) or Brakman et al. (2009, chap. 5). 3 Estimations carried out by Head and Mayer (2011) or the author of this chapter reveal that the “wage equation” is robust with regard to fixed effects panel data. However, discussing the interpretation of panel data estimations and the role of

A Generalized NEG Wage-Type Equation

65

based on Redding and Venables’s (2004) model and subsequent ones by Head and Mayer (2004b), Breinlich (2006), and Head and Mayer (2006). NEG models have been described many times, but the contribution of this paper is to accommodate different derivations under a common notation. The rest of the chapter is organized as follows. Section 4.2 presents the demand side of the NEG basic model. Section 4.3 introduces the generalized form of the wage equation. Using this framework, the inclusion of human and physical capital is derived in section 4.4. Section 4.5 concludes.

4.2. The Demand Side of the NEG Basic Model The NEG distinguishes two sectors. A perfect competitive sector produces a single homogeneous good under constant returns to scale (CRS). The second sector produces a large variety of differentiated goods and is composed of firms exhibiting internal increasing returns to scale (IRS) and operating under a market structure of monopolistic competition. Here, these two sectors are noted as ‫ ܥ‬and ‫ܯ‬, respectively (Fingleton, 2006; Fingleton and Fischer, 2010). In the basic NEG model, the two sectors are termed “agriculture” and “manufacturing”. As Fujita et al. (1999, p. 58) point out, “agriculture” can be interpreted as the “‘residual’, perfectly competitive sector that is the counterpart to the action taking place in the increasing-returns, imperfectly competitive manufacturing sector”. For Redding and Venables (2004) the ‫ ܯ‬sector can be interpreted as a composite of manufacturing and service activities, while Fingleton and Fischer (2010) define it as services. Alternatively, Baldwin et al. (2003, p. 13) argue that the key distinction in the basic core-periphery model is that the ‫ ܥ‬sector intensively uses the internationally immobile factor. Indeed, in some models, ‫ ܥ‬is identified with a nontraded good (housing sector). However, the possible presence of nontraded goods affects the selection of the appropriate proxy for the dependent variable of the “wage” equation – see equations (21) and (22) below. Some researchers seem to interpret the dependent variable of the “wage” equation as the level of “nominal manufacturing” wages. However, on the one hand, an assumption of sectoral factor mobility guarantees factor equalization across sectors, justifying a proxy of the dependent variable by measures for the aggregate economy. On the other hand, the “nominal wage” of the ‫ ܯ‬workers refers to their wage in terms different rates of technological progress is beyond the scope of this paper, which only refers to the abundant cross-sectional literature.

Chapter Four

66

of the ‫ ܥ‬numeraire sector (Baldwin et al., 2003, chap. 2). Therefore, the ‫ܥ‬ and ‫ ܯ‬notation for the two sectors, according to their market structure, is a reminder that we are not sure about what the two NEG sectors are. The model presented here focuses on the ‫ ܯ‬sector, but is introduced with the demand side for both sectors. The basic model assumes that every consumer shares the same Cobb-Douglas tastes for the two types of goods. Alternatively, though it is not essential for later arguments, it is useful to assume a different Ͳ ൏  ߤ௝ ൏ ͳ parameter of preferences for each type of good in each region ݆. The upper-level step of the problem of the representative consumer in region ݆ is to divide her total income ܻ௝ between the consumption of the two aggregated goods: ఓ

ଵିఓೕ

ƒš ܷ௝ ൌ ‫ܯ‬௝ ೕ ‫ܥ‬௝

ெೕ ǡ஼ೕ

(1) s.t. ‫ܩ‬௝ெ ‫ܯ‬௝ ൅ ܲ௝஼ ‫ܥ‬௝ ൌ ܻ௝ where ܲ௝஼ is the price of the ‫ ܥ‬goods and ‫ܩ‬௝ெ is a “price index” of ‫ܯ‬ goods. Therefore, the amount of consumption of region ݆ in ‫ ܯ‬goods is: ‫ܯ‬௝ ൌ ߤ௝ ܻ௝ Τ ‫ܩ‬௝ெ ൌ ‫ܧ‬௝ெ ൗ‫ܩ‬௝ெ (2) ெ ‫ܧ‬௝ is the expenditure of region ݆ in all the varieties of the ‫ ܯ‬good. ߤ௝ ൌ ‫ܧ‬௝ெ ൗܻ௝ ൌ ‫ܧ‬௝ெ ൗ‫ܧ‬௝ . is the share of ‫ ܯ‬consumption in income. Here, total income, ܻ௝ , is the same as total expenditure, ‫ܧ‬௝ , because the model does not include intermediate goods. Considering heterogeneous preferences (ߤ௝ ), as Combes et al. (2008b, chap. 12), allows for what in a model with intermediate goods (Fujita et al., 1999, chap. 14) would be different sectoral shares of costs in intermediate goods, or different sectoral composition. The intermediate inputs are included in Table 4-1 below. After deciding the optimal consumption of the composite index of ‫ܯ‬ goods, the representative consumer of region ݆ decides the quantity of consumption for each ‫ ܯ‬variety. The demand of ‫ ܯ‬goods in any region ݆ is derived from the maximization of a Dixit-Stiglitz CES subutility function for the consumption ‫ݔ‬ሺ‫ݒ‬ሻ௝ of each ‫ ܯ‬variety ‫ ݒ‬ൌ ͳǡ ǥ ǡ ܸ. Given that the utility function ‫ܯ‬௝ embodies a preference for diversity and there are IRS in the ‫ ܯ‬sector, each firm produces a distinct variety. If the “world” is composed of ܴ regions (݅ ൌ ͳǡ ǥ ǡ ܴ), the number of varieties potentially available (ܸ) in region ݆ is the number of firms and varieties (݊௜ ) produced in all the regions: ܸ ൌ σோ௜ୀଵ ݊௜ . In equilibrium, all goods produced in each region ݅ are demanded by ݆ in the same quantity. Therefore, the representative consumer in ݆ solves the following problem:

A Generalized NEG Wage-Type Equation ఙ ఙିଵ ఙିଵ ቎෍ ෍ ‫ݔ‬ሺ‫ݒ‬ሻ௜௝ఙ ቏ ௜ୀଵ ௩ୀଵ ோ ோ

ƒš ‫ܯ‬௝ ൌ ௫೔ೕ

௡೔

ఙ ఙିଵ ఙିଵ ൥෍ ݊௜ ‫ݔ‬௜௝ఙ ൩ ௜ୀଵ

67

ோ

ൌ

(3)

•Ǥ –Ǥ ෍ ݊௜ ‫݌‬௜௝ ‫ݔ‬௜௝ ൌ ‫ܧ‬௝ெ ௜ୀଵ

where ߪ ൐ ͳ is the elasticity of substitution between any pair of varieties, ‫ݔ‬௜௝ is the amount of consumption in ݆ of the variety produced in ݅, and ‫݌‬௜௝ is the delivery price of that variety. The first-order conditions of this problem for a representative variety from region ݅ and a variety ݃ produced anywhere give equality of marginal rates of substitution to price ratios: ‫ݔ‬௜௝ ଵΤఙ ‫݌‬௚௝ (4) ൌ ‫ݔ‬௚௝ ଵΤఙ ‫݌‬௜௝ ௖ ) of the good produced by a The ݆-market optimum consumption level (‫ݔ‬௚௝ ݃-firm is obtained by plugging the value of ‫ݔ‬௜௝ from equation (4) into the expenditure constraint: ‫ܧ‬௝ெ ௖ ିఙ ൌ ‫݌‬௚௝ (5) ‫ݔ‬௚௝ ଵିఙ σோ௜ୀଵ ݊௜ ‫݌‬௜௝ This relation is also true for the representative variety from region ݅. Therefore, keeping the denominator as a sum across ݅ varieties, ݆consumption of a variety produced by an ݅-region firm is: ‫ܧ‬௝ெ ‫ܧ‬ெ ௖ ିఙ ିఙ ௝ ൌ ‫݌‬௜௝ ൌ ‫݌‬ (6) ‫ݔ‬௜௝ ௜௝ ଵିఙ σோ௜ୀଵ ݊௜ ‫݌‬௜௝ ܵ௝ெ Redding and Venables (2004) call the term ‫ܧ‬௝ெ ൗܵ௝ெ “market capacity”. It gives the position of the demand curve facing each firm in market ݆. ௖ , is Equation (6) says that the consumption of a variety ݅ in market ݆, ‫ݔ‬௜௝ inversely related to the delivery price of that variety, ‫݌‬௜௝ , and to the index, ଵିఙ ܵ௝ெ ൌ σோ௜ୀଵ ݊௜ ‫݌‬௜௝ . This latter index is called “supplier access” by Redding and Venables (2004). Here, it is termed “supply” (Head and Mayer, 2006) or the “competition” index. ܵ௝ெ measures the level of competition between ‫ ܯ‬varieties in market ݆, given the characteristic tastes of consumers. The assumption ߪ௜ ൌ ߪ ൐ ͳ implies negative exponents in the ‫݌‬௜ଵିఙ terms. Therefore, through equation (6), it will be difficult to obtain a high market share in location ݆ served by a large number of lowprice sources.

Chapter Four

68

Plugging equation (6) into ݆’s subutility function, the optimal utility భ

level is ‫ܯ‬௝ ൌ ‫ܧ‬௝ெ ܵ௝ெ ഑షభ , which can be re-written as ‫ܯ‬௝ ൌ ‫ܧ‬௝ெ ൗ‫ܩ‬௝ெ after ଵΤሺଵିఙሻ

defining ‫ܩ‬௝ெ ‫ܵ ؠ‬௝ெ . This allows the interpretation of ‫ܩ‬௝ெ as an aggregate “price index” of ‫ܯ‬. Fujita et al. (1999, chap. 4) to obtain this index through the dual problem of the restricted maximization in equation (3). The minimum cost of attaining ‫ܯ‬௝ results in ‫ܩ‬௝ெ ‫ܯ‬௝ , where: ோ

௡೔

ଵ ଵିఙ

‫ܩ‬௝ெ ൌ ቎෍ ෍ ‫݌‬ሺ‫ݒ‬ሻଵିఙ ௜௝ ቏ ௜ୀଵ ௩ୀଵ

ோ

ଵ ଵିఙ

ଵିఙ ൌ ൥෍ ݊௜ ‫݌‬௜௝ ൩

(7)

௜ୀଵ

Head and Mayer’s (2006) ܵ௝ெ notation is preferred here to the traditional ‫ܩ‬௝ெ because it makes the negative exponents of ‫݌‬௜ more transparent and avoids defining ‫ܩ‬௝ெ as a nonobservable price. ‫ܩ‬௝ெ is a price because it is the unit cost of utility derived from the consumption of ‫ ܯ‬goods. “[…] Just as ‫ ܯ‬can be thought of as a utility function, ‫ ܩ‬can be thought of as an expenditure function” (Fujita et al., 1999, p. 47). Brakman et al. (2009, chap. 3) call ‫ܩ‬௝ெ a consumption-based or exact price index. Baldwin et al. (2003, chap. 2) label it “perfect” price index. The original adjectives used by Krugman (1992) were “true” or “ideal” price index. Firms of the same region are assumed to have the same free-on-board price. Trade costs are assumed to be borne by consumers, so firms follow a mill pricing policy. If ‫݌‬௜ is the mill price of a good produced in region ݅, the delivered price in market ݆ is assumed to be ‫݌‬௜௝ ൌ ܶ௜௝ ‫݌‬௜ , where ܶ௜௝ ൒ ͳ are “iceberg” transport or trade costs: for every unit shipped, only ͳΤܶ௜௝ units arrive at the destination, while the rest melts during transport. Therefore, for every unit consumed in ݆ at a price ‫݌‬௜௝ , ܶ௜௝ units must be shipped. From equation (6), ݆ effective demand to ݅ becomes: ‫ܧ‬௝ெ ௗ (8) ൌ ܶ௜௝ ଵିఙ ‫݌‬௜ିఙ ெ ‫ݔ‬௜௝ ൌ ܶ௜௝ ‫ݔ‬௜௝ ܵ௝ When ݅ ് ݆, these sales are the exports from region ݅ to region ݆. ߶௜௝ ‫ؠ‬ ܶ௜௝ ଵିఙ is what Baldwin et al. (2003, chap. 2) call “phi-ness” of trade. It ranges from ߶௜௝ ൌ Ͳ, where ܶ௜௝ and ߪ are high enough to eliminate all trade, to ߶௜௝ ൌ ͳ, for full economic integration. In order to get the value of total exports from region ݅ to ݆, Redding and Venables (2004) express in values the volume of export in equation (8) and aggregate it across all the varieties produced in region ݅. The resulting “trade equation” reflects bilateral trade flows in Anderson and van Wincoop’s gravity-type equation:

A Generalized NEG Wage-Type Equation

݊௜ ‫݌‬௜ ‫ݔ‬௜௝ ൌ ߶௜௝ ݊௜ ‫݌‬௜ଵିఙ

69

‫ܧ‬௝ெ

(9) ܵ௝ெ where the term ݊௜ ‫݌‬௜ଵିఙ measures the “supply capacity” of the exporting region. Redding and Venables (2004) proxy market capacities, ‫ܧ‬௝ெ ൗܵ௝ெ , with the estimates for importing region dummies in the gravity equation (9). Therefore the crucial role of ܵ௜ெ in NEG theory is measured by an unobservable and constant individual effect for the importing region in an equation of bilateral trade. With those estimates, the authors construct the key variable of the model, Market Potential. Given the effective demand from ݆-market in equation (8), total demand to a representative ‫ ܯ‬firm in region ݅ will be the sum of what it sells to the world markets4: ோ ோ ‫ܧ‬௝ெ ‫ݔ‬௜ ‫ ؠ‬෍ ‫ݔ‬௜௝ ൌ ‫݌‬௜ିఙ ෍ ܶ௜௝ ଵିఙ ெ ൌ ‫݌‬௜ିఙ ܴ‫ܲܯ‬௜ (10) ܵ௝ ௝

௝

where ܴ‫ܲܯ‬௜ stands for Real Market Potential (ܴ‫ܲܯ‬௜ ). In summary, using alternative notations, the Market Potential of a firm/region ݅ is: ோ ோ ‫ܧ‬௝ெ ఙିଵ ܴ‫ܲܯ‬௜ ൌ ෍ ߶௜௝ ெ ൌ ෍ ߤ௝ ܶ௜௝ ଵିఙ ‫ܧ‬௝ ‫ܩ‬௝ெ (11) ܵ௝ ௝

௝ ெ ܵ௝

ଵିఙ

ൌ ‫ܩ‬௝ெ ൌ σோ௜ୀଵ ߶௜௝ ݊௜ ‫݌‬௜ଵିఙ . It is where the competition index is ଵିఙ ൌ ͳ for the domestic sales (݅ ൌ ݆), although assumed that ߶௜௝ ൌ ܶ௜௝ research with areal data needs to consider a proxy for internal trade costs. “Krugman Market Potential” (Head and Mayer, 2004a) in equation (11) is a phi-ness of trade weighted by sum of market capacities. This term is relabelled as “Real Market Potential” (ܴ‫ܲܯ‬௜ ) by Head and Mayer (2006). These authors reserve the adjective “nominal” for a Harris’s (1954) index such as σோ௝ ߶௜௝ ‫ܧ‬௝ , because is a pure measure of the size of the available market, equivalent to assuming ܵ௝ெ ൌ ܵ ெ ൌ ͳ. The adjective “real” underlines the importance of discounting expenditures by the supply index ܵ௝ெ . However, Head and Mayer’s (2006) adjective “real” may be a misleading analogy with the deflation of nominal monetary values, because the ‫݌‬௜ଵିఙ terms in ܵ௝ெ have a negative exponent. “Real” becomes more confusing when expenditure is measured in deflated monetary units, 4

Many variants of the NEG models are isomorphic irrespective of the agglomeration mechanism they assume (Robert-Nicoud, 2005). Additionally, Head and Mayer (2011) show that the wage equation prediction arises under diverse conditions. However, the interpretation of the empirical results may be different.

70

Chapter Four

as it is common in empirical research. Therefore, the competition effects of ܵ௝ெ seem to be better described by the expression “Market Access” (Redding and Venables, 2004). Despite this, the name “Real Market Potential” has two virtues. On one side, it stresses the continuity from the old-style Regional Science to the NEG, as commented on by Fujita et al. (1999, chap. 3). On the other hand, it avoids the confusion with the WTO’s definition of “market access” (Head and Mayer, 2011). Empirical works using Harris’s (1954) measure of Market Potential are frequently criticised because that measure does not allow interpretation of the estimating results of a wage equation in terms of the structural parameters. When proxying trade costs by distances, Harris’s index implies an ad hoc assumption of -1 for the distance exponent, instead of estimating it through equation (9). However, a trade elasticity to distance of -1 is an extremely robust empirical finding (Head and Mayer, 2014). Moreover, the different measures of Market Potential share the same crucial features (Bruna, Lopez-Rodriguez, et al., 2014) and are highly correlated (Breinlich, 2006; Head and Mayer, 2006). Additionally, with distance exponents close to -1, any of them overweighs the nearest neighbours. The meaning of geographical distances as trade cost also remain unclear and measuring internal distances is a serious challenge (Bruna, Faíña, et al., 2014). In any case, for any proxy variable of Market Potential, the main argument of the present paper is that the empirical results of a wage equation do not enable unambiguous interpretation in terms of the structural parameters of the model.

4.3. Supply Side and Generalized Wage-Type Equation Krugman’s (1980) classic assumptions for the ‫ ܯ‬sector are the following: labour is the only production factor; there are no economies of scope; and there are CRS during production, which involves a marginal input requirement. Production also involves a fixed input, inducing IRS. Because of that, consumers’ preference for variety and the unlimited number of potential varieties, no firm will choose to produce the same variety supplied by another firm. This means that each variety is produced in only one location by a single specialized firm, so the number of firms in operation is the same as the number of available varieties (Fujita et al., 1999). Keeping the ‫ݔ‬௜ notation of the demand equation (10), the production function considered here for the ‫ ܯ‬firm in region ݅ is: ݂ (12) ‫ݔ‬௜ ൌ െ݂ ൅ ‫ܣ‬௜ ‫ܫ‬௜ ൌ ‫ܣ‬௜ ൬െ ൅ ‫ܫ‬௜ ൰ ‫ܣ‬௜

A Generalized NEG Wage-Type Equation

71

where ‫ܫ‬௜ is a compound input. ‫ܣ‬௜ is a Ricardian technology, which means that the marginal input requirement is ܿ௜ ൌ ͳΤ‫ܣ‬௜ . ݂ is a fixed cost defined in units of output. Therefore, ‫ܫ‬௜ ൌ ܿ௜ ሺ݂ ൅ ‫ݔ‬௜ ሻ and the fixed input requirement, ܿ௜ ݂, is allowed to vary across regions. If, for now, ‫ݍ‬௜ is the price index of ‫ܫ‬௜ , the cost of producing ‫ݔ‬௜ is ‫ݍ‬௜ ܿ௜ ሺ݂ ൅ ‫ݔ‬௜ ሻ. Marginal cost, the price of the compound input in efficiency units, is ݉௜ ൌ  ‫ݍ‬௜ ܿ௜ . A firm’s total output is given by the sum of what it sells to the world markets, ‫ݔ‬௜ ൌ σோ௝ ‫ݔ‬௜௝ , and its total income is σோ௝ୀଵ ‫݌‬௜ ‫ݔ‬௜௝ ൌ ‫݌‬௜ ‫ݔ‬௜ . Therefore, firms of the ‫ ܯ‬sector, facing given factor prices in ݉௜ , maximize the following profit function with respect to their mill prices ‫݌‬௜ : (13) ߨ௜ ൌ ‫݌‬௜ ‫ݔ‬௜ െ ݉௜ ሺ݂ ൅ ‫ݔ‬௜ ሻ where the effective demand (‫ݔ‬௜ ) is taken from equation (10). If each firm takes the competition index ܵ௝ெ in ܴ‫ܲܯ‬௜ as a given4, profit maximization implies that firms choose price as a mark-up over marginal costs: ߪ ‫݌‬௜ ൌ ݉ (14) ߪെͳ ௜ At these optimum mill prices, profits are: ͳ (15) ߨ௜ ൌ ݉௜ ൬ ‫ ݔ‬െ ݂൰ ߪെͳ ௜ The demand function in equation (10) at optimum prices is: ‫ݔ‬௜ ൌ ఙ

ିఙ

ቀ ቁ ݉௜ିఙ ܴ‫ܲܯ‬௜ . Therefore, the profits of ݅ become a function of its ఙିଵ Real Market Potential: ሺߪ െ ͳሻఙିଵ ଵିఙ (16) ݉௜ ܴ‫ܲܯ‬௜ െ ݂݉௜ ߨ௜ ൌ ߪఙ Equation (16) is similar to the “profit equation” derived by Combes et al. (2008, chap. 12). It is at the origin of the “wage equations” derived below. When a market equilibrium condition is imposed in equation (16), the wage-type equation gives a relationship between marginal costs and the spatial distribution of expenditure, as captured by ܴ‫ܲܯ‬௜ .5 Free entry and exit in response to profits or losses ensures that the long-run profits are zero. Therefore, Redding and Venables (2004) use the production level at which firms break even to calculate the maximum remuneration that firms can afford to pay to factors. In order to get rid of ݂, equation (10) allows the calculation of the production level at which profits are zero: ‫ݔ‬௜ ൌ ሺߪ െ ͳሻ݂ ൌ ‫ݔ‬ҧ . Therefore, from the effective demand, active firms at location ݅ attain this level of output and break even if, and only if, the mill price they charge satisfies: 5

See the discussion about a possible variable elasticity of demand in Baldwin et al. (2003, chap. 5).

Chapter Four

72

ͳ (17) ܴ‫ܲܯ‬௜ ‫ݔ‬ҧ Through the mark-up pricing rule (14), equation (17) can be expressed with marginal cost as the dependent variable: ‫݌‬௜ఙ ൌ

݉௜ ൌ

ଵ ఙ

ோ

‫ܧ‬௝ ߪെͳ ͳ ߪെͳ ͳ ൬ ܴ‫ܲܯ‬௜ ൰ ൌ ቌ ෍ ߤ௝ ܶ௜௝ ଵିఙ ெ ቍ ߪ ‫ݔ‬ҧ ߪ ‫ݔ‬ҧ ܵ௝

ଵ ఙ

(18)

௝

Equation (18) is called here the “generalized wage-type equation”. It reveals a relation between Real Market Potential and the maximum marginal cost that a firm can afford to pay. The word “generalized” is used to emphasize that the dependent variable is marginal cost. The “wagetype” equation makes reference to the name “wage equation”6 used by Fujita et al. (1999, chap. 4) in a model in which labour is the only production factor. Baldwin et al. (2003, p. 19) prefer the expression “market-clearing condition” in order to emphasize the assumption of zero profits. The presence of Real Market Potential, ܴ‫ܲܯ‬௜ ൌ σோ௝ ܶ௜௝ ଵିఙ ‫ܧ‬௝ெ ൗܵ௝ெ , in the profit equation (16) captures the two typical NEG effects. As summarized by Redding (2011), IRS imply that firms want to concentrate production while transport costs imply that they want to be close to large markets. This is called “home market effect” and provides a “backward linkage”. Therefore, firms close to large markets can pay higher marginal costs. On the other hand, the counteracting force promoting ‫ ܯ‬sector dispersion is the “market crowding” or “competition” effect derived from discounting expenditures by ܵ௝ெ in ܴ‫ܲܯ‬௜ . The supply index ܵ௝ெ is a trade cost weighted sum of supply capacities and measures the degree of competition in market ݆. As more firms choose one region, the market there becomes more crowded, lowering the Real Market Potential, until another region is more profitable (Head and Mayer, 2004a). Head and Mayer (2006) discuss two paths to equilibrium. Spatial equilibrium requires that markets clear and no mobile agent has a unilateral incentive to relocate. In a spatial equilibrium, firms have the same profits in all regions, so if ܿ௜ is considered as given, any shock in the demand to a region will be followed by an adjustment in their use of factors and/or by an adjustment in its factor prices. Therefore, the relative magnitudes of price or quantity adjustment to cross-regional variation in 6

This nonstrategic behaviour is innocuous. (Fujita et al., 1999, chap. 4; Combes et al., 2008, chap. 9).

A Generalized NEG Wage-Type Equation

73

demand depend chiefly on the mobility of factors. One strand of the literature makes the polar assumption of factor price equalization. Redding and Venables (2004) pioneered what Head and Mayer (2006) call the second polar path towards spatial equilibrium, that loads all the response to demand differences into factor prices. When labour is not the only production factor, as in Redding and Venables’s (2004) model, the assumptions about factor mobility determine which factor prices are not equalized across regions. Therefore, the lefthand side variable of equation (18) becomes a function of the price of immobile factors. Redding and Venables interpret these factors as labour. The full general equilibrium explored in Fujita et al. (1999) involves specifying factor endowments. Alternatively, Redding and Venables “take expenditure and output in each country as exogenous, and ask, “What wages can manufacturing firms in each location afford to pay?” Accommodating the notation to distinguish two inputs, now ‫ݍ‬௜ designates the price of an internationally immobile factor with input share ߠ. ‫ݖ‬௜ is the price of a mobile factor with input share ߰. If there are CRS during production, ߰ ൌ ͳ െ ߠ. Alternatively, ߠ and ߰ can be viewed as parameters describing the degree of mobility of the underlying production factors. Therefore, similarly to Breinlich’s (2006) specification, marginal costs are: ట (19) ݉௜ ൌ ‫ݍ‬௜ఏ ‫ݖ‬௜ ܿ௜ And the generalized wage-type equation takes the form: ଵΤఏ

ߪെͳ ͳ ͳ ͳ ଵ ‫ݍ‬௜ ൌ ൥ ܴ‫ܲܯ‬௜ ൗఙ ట ൩ ߪ ‫ݔ‬ҧ ଵΤఙ ‫ܿ ݖ‬௜

(20)

௜

Given that ‫ݖ‬௜ is the price of a mobile factor, Redding and Venables (2004) assume that it is equalized across regions, so ‫ݖ‬௜ ൌ ‫ݖ‬. These authors seem to be thinking of the Footloose Capital model (Baldwin et al., 2003, chap. 3). In this model, each ‫ ܯ‬firm requires just one unit of mobile capital. Capital owners spend their income locally, so a long-run spatial equilibrium implies the international nominal equalization of the return on capital. The following section provides an alternative assumption. Under Redding and Venables’s (2004) interpretation, the price of the immobile factor is the wage level: ‫ݍ‬௜ ൌ ‫ݓ‬௜ . Simplifying ܿ௜ ൌ ‫ିܣ‬ଵ and ௜ taking logarithms in equation (20), their testable cross-sectional wage equation including an intercept (‫ )ܥ‬becomes: ͳ ͳ (21) Ž ܴ‫ܲܯ‬௜ ൅ Ž ‫ܣ‬௜ Ž ‫ݓ‬௜ ൌ ‫ ܥ‬൅ ߠߪ ߠ Assuming as well that ‫ݍ‬௜ ൌ ‫ݓ‬௜ , Head and Mayer (2004b) provide an alternative version of equation (20). In equation (19), labour is

74

Chapter Four

distinguished from other primary factors without reference to their geographical mobility. Now, the wage-type equation takes the following form: ͳ (22) ߠŽ ‫ݓ‬௜ ൅  ߰Ž ‫ݖ‬௜ ൌ ‫ ܥ‬൅ Ž ܴ‫ܲܯ‬௜ ൅ Ž ‫ܣ‬௜ ߪ where ‫ ݖ‬might be the price of nontrade factors, for instance. The left-hand side of equation (22) is a cost-share weighted sum of logged primary factor prices. Head and Mayer (2004b) interpret that a natural proxy for this dependent variable is the log of GDP per capita. Similarly, despite the appearance of equation (21), it seems that Redding and Venables (2004) considered that the assumption ‫ݍ‬௜ ൌ ‫ݓ‬௜ was restrictive and their empirical analysis recovered the meaning of ‫ݍ‬௜ in equation (20). They proxied ‫ݍ‬௜ by GDP per capita, because GDP includes the income of all immobile factors. In other words, under both approaches, the proxy variable to measure the left-hand side of these wage-type equations is not the wage level. This assessment of the “wage” equation has an important consequence when trying to interpret the empirical estimations in terms of structural parameters. The estimate of Ž ܴ‫ܲܯ‬௜ in equation (21) cannot be interpreted as an estimate of ͳΤߪ anymore. It would be necessary to measure ‫ܣ‬௜ in order to deduce a value for ߪ from the estimation. On the contrary, equation (22) allows the direct estimation of ߪ. However, as mentioned in section 0, the high correlation between alternative dependent variables makes it observationally equivalent to equation (21). Table 4-1 summarizes how some NEG models can be interpreted under the specification of marginal costs in equation (19). Starting from the 1999 book by Fujita, Krugman and Venables, the table presents some models with different emphases on theory and on econometrics. The distinction ట about what is noted as ‫ݍ‬௜ఏ and ‫ݖ‬௜ in Table 4-1 is a matter of convenience. ‫ݓ‬௜ is chosen as the dependent variable of the generalized wage-type equation in order to encompass the models under a common framework. The derivation of each specification from the generalized wage-type equation is available from the author upon request. The last column in Table 4-1 translates the marginal input requirement, ܿ௜ , into the productivity parameters of the ‫ ܯ‬sector. A few models in the table consider an empirical wage equation with control variables proxying for total factor productivity (‫ܣ‬௜ ). Several of those models consider human capital (݄௜ ) and one of them distinguishes the wages of unskilled labour (‫ݓ‬௜௨ ). ‫ܩ‬௜ெ , defined in equation (7), appears in the models with “forward linkages”, where the ‫ ܯ‬consumption varieties are also intermediate inputs and their true price index is included in equation (19). The following section presents the wage-type equation in the last row of Table 4-1.

A Generalized NEG Wage-Type Equation

75

Table 4-1: The Generalized Wage Equation in Several NEG Models ࣒

࢓࢏ ൌ ࢗࣂ࢏ ࢠ࢏ ࢉ࢏  ֜If ‫ݍ‬௜ ൌ ‫ݓ‬௜ : ࢝࢏ ൌ ൤ Authors, model

࣌ି૚ ૚ ࣌

ࣂ

૚ ഥ ൗ࣌ ࢞

ࡾࡹࡼ࢏

૚ൗ ࣂ ૚ൗ ૚ ૚ ࣌ ࣒ ൨ ࢠ࢏ ࢉ ࢏

૚ൗ ࢉ

࣒

ࢠ࢏ ͳ ͳ

Fujita et al. (1999, c. 4) Fujita et al. (1999, c. 15) Fujita et al. (1999, c. 14); Puga (1999)

ͳ ͳ ߠ

‫ܩ‬௜ெ

Redding and Venables (2004) Head and Mayer (2004a) Breinlich (2006) Head and Mayer (2006) Fingleton (2006); Fingleton and Fischer (2010) Redding and Schott (2003), if ࢛࢝࢏ ൌ ࢝࢏ Τࢎ࢏

ߠ ߠ ߠ ͳ

‫ ݖ‬ఊ ‫ܩ‬௜ெ ‫ݖ‬௜ଵିఏ ‫ݖ‬ట ͳ

Bruna (2015)

࢏

‫ܣ‬ ‫ܣ‬௧ ‫ܣ‬

ଵିఏ ଵିఏିఊ

‫ܣ‬௜ ‫ܣ‬௜ ‫ܣ‬௜ ‫ܣ‬eš’ሺߚ݄௜ ሻ ఉ

ͳ

ͳ

‫ܣ‬௜ ; ݄௜

1

1

݄௜

ߠ

‫ݖ‬ట

ቀ‫݇ܣ‬௜ఈ ݄௜ ቁ

ఉ ఉ

ఏ

4.4. Inclusion of Human and Physical Capital In order to consider the role of capital stock as an immobile factor, it is convenient to disaggregate ‫ݍ‬௜ in equation (20) as ‫ݍ‬௜ ൌ ‫ͳݓ ݅ߙݎ‬െߙ , with ‫ݎ‬௜ ݅ being the user cost of capital in region ݅. If this last price is not equalized across countries, and if it is difficult to obtain data about it, the dependent variable of the generalized wage-type equation comes close to the one in equation (22). Under this latter interpretation, the testable equation would not be affected by the inclusion of physical capital stock. However, as might be expected, the estimates of Market Potential change dramatically when a wage equation is controlled for physical and human capital (Breinlich, 2006; Bruna, Faíña, et al., 2014). Alternatively, it is natural to wonder about the interpretation of Redding and Venables’s (2004) wage equation when capital adopts the form of capital stock, as in the constructed capital model (Baldwin et al., 2003, chap. 6) or, particularly, in Li’s (2012) model with immobile specific capital. The international trade literature considers investment as a produced commodity subject to trade. Once installed as an addition to the capital stock, investment becomes immobile. If capital is considered as mobile ex-ante, in some degree or another, but immobile ex-post, then past decisions about the location of capital goods are going to condition firms’ productivity during long time horizons. However, the time horizon in the NEG model is empirically ambiguous. Cross-sectional regressions with

76

Chapter Four

variables in levels are usually considered to represent “long-term” relationships. This may mean that we are taking current data to study remote forces responsible for originating a particular spatial distribution of economic activity. Indeed, Krugman (2011) recognized that NEG models were talking about the past, and not so much about the current forces of agglomeration. This might not be what many researchers try to test, as evidenced by their discussion about the estimates of ߪ in terms of current industries. The time horizon of the model is related to the problem of endogeneity. It might be said that the NEG is concerned with fundamental determinants of income (Redding and Venables, 2004) while developing accounting searches for proximate ones (Caselli, 2005). Under a “longterm” perspective, human and physical capital are endogenous and dependent on forces such as institutions and government policies, which are also endogenous. All these relationships are shaped by geography through historical processes and it is not easy to disentangle the causal channels. For instance, the spatial distribution of human capital (Redding and Schott, 2003), as well as other variables, can be correlated with Market Potential. However, a distance exponent close to -1 makes any proxy variable for Market Potential to capture neighbouring effects or similarities. Therefore, the whole empirical debate is about the compromise between endogeneity and the omission of relevant variables, the selection of the proper exogenous instruments and the unambiguous statistical information captured by each empirical variable. These issues are part of the problem of interpreting the estimation results of a wage equation. Similarly to Head and Mayer’s (2006) approach to human capital, the alternative proposed here is to include human and physical capital stock in the productivity parameter as immobile and exogenous factors. Per capita capital stock could be viewed as embodied in the firms paying immobile workers. Including exogenous human and physical capital in the model does not solve the problems identified above. However, the resulting testable wage-type equation presents several advantages. It permits evaluating a possible upward bias in the estimate of Market Potential due to the omission of relevant variables (Fingleton, 2006) and ascertaining the “direct effects” of Market Potential (Breinlich, 2006). Additionally, the specification can help to capture the effects of exogenous education, infrastructure and transport policies (Bruna, Faíña, et al., 2014). Finally, the similarity of the resulting equation to an expanded production function is an useful reminder of the problem of measuring our ignorance in development accounting, particularly in the context of the discussion in

A Generalized NEG Wage-Type Equation

77

section 4.1 about the empirical dependent variable of the wage-type equation. The starting point is to replace equation (12) by a Jones’s (1997) production function, but omitting any reference to Minceranian regressions: ‫ݔ‬௜ ൌ െ݂ ൅ ‫ܭ‬௜ఈ ሺ‫ܤ‬௜ ݄௜ ‫ܮ‬௜ ሻଵିఈ (23) where ‫ܤ‬௜ is a labour-augmenting technological index, while ‫ܭ‬௜ , ‫ܮ‬௜ and ݄௜ are physical capital, raw labor and the average level of human capital, respectively. Then, if physical capital stock per worker is noted as ݇௜ , the firm’s production function becomes: ‫ܮ‬௜ ൌ െ݂ ൅ ‫ܣ‬௜ ݇௜ఈ ݄ଵିఈ ‫ܮ‬௜ ‫ݔ‬௜ ൌ െ݂ ൅ ‫ܤ‬௜ଵିఈ ݇௜ఈ ݄ଵିఈ (24) ௜ ௜ ଵିఈ ൌ ‫ ܣ‬is a common marginal labour requirement in where ‫ܣ‬௜ ൌ ‫ܤ‬௜ “effective” labour units (‫ܮ‬௜ Τ݇௜ఈ ݄ଵିఈ ). Therefore, marginal costs take the ௜ ట form ݉௜ ൌ ‫ݓ‬௜ఏ ‫ݖ‬௜ ܿ௜ ൌ ‫ݓ‬௜ ܿ௜ ൌ ‫ݓ‬௜ Τ‫݇ܣ‬௜ఈ ݄ଵିఈ , with ‫ݖ‬௜ ൌ ߠ ൌ ߰ ൌ ͳ. ௜ However, even if ݇௜ and ݄௜ are considered as determinants of the productivity of labour, the possible presence of other production factors makes relevant the ߠ parameter. The following three examples show why this issue is pertinent to the interpretation of empirical results. The first case is a generalization of equation (24), such as: ఏ ట (25) ‫ݔ‬௜ ൌ െ݂ ൅ ൣ‫݇ܣ‬௜ఈ ݄௜ ଵିఈ ‫ܮ‬௜ ൧ ܼ௜ where ܼ௜ is a factor with price ‫ݖ‬௜ , and ‫ݖ‬௜ ൌ ‫ݖ‬. From this production function, the last row of Table 4-1 shows the proposed version of the generalized wage-type equation (20): ߪ െ ͳ ିଵΤఙ ିట ଵΤఏ Τ ఉ (26) ‫ݓ‬௜ ൌ ‫ ܣ‬൬ ‫ݔ‬ҧ ‫ ݖ‬൰ ܴ‫ܲܯ‬௜ଵ ఏఙ ݇௜ఈ ݄௜ ߪ which it is supposed to verify ߙ ൅ ߚ ൌ ͳ. The exponent of ܴ‫ܲܯ‬௜ is ͳΤߠߪ. A second example with a different interpretation of ݄௜ is based on Mankiw-Romer-Weil’s (1992) production function, in which human capital is considered to be ݄௜ times more productive than raw labour: ఉ (27) ‫ݔ‬௜ ൌ െ݂ ൅ ‫ܭܣ‬௜ఈ ሺ݄௜ ‫ܮ‬௜ ሻఉ ‫ܮ‬௜ ଵିఈିఉ ൌ െ݂ ൅ ‫݇ܣ‬௜ఈ ݄௜ ‫ܮ‬௜ Unlike the previous example, the wage equation derived from here verifies ߙ ൅ ߚ ൏ ͳ and the exponent of ܴ‫ܲܯ‬௜ is ͳΤߪ: ߪ െ ͳ ିଵΤఙ Τ ఉ (28) ‫ݓ‬௜ ൌ ‫ܣ‬ ‫ݔ‬ҧ ܴ‫ܲܯ‬௜ଵ ఙ ݇௜ఈ ݄௜ ߪ The third example starts with a CRS production function in per capita . In the spirit of models with spillovers units, such as ‫ݕ‬௜ ൌ ‫ܣ‬௜ ݇௜ఈ ݄ଵିఈ ௜ (Baldwin et al., 2003, chap. 7), now the productivity parameter depends on neighbours productivity or “scale”. From the empirical point of view, this ఞ is equivalent to assuming ‫ܣ‬௜ ൌ ‫ܲܯܴܣ‬௜ for ߯ ൐ Ͳ and the estimation results are identical to those of equation (26). Variants of this approach

78

Chapter Four

have been used in different strands of the literature by Clemente et al. (2009), Fischer (2011) and Holl (2012), among others. This example illustrates once more the difficulty in interpreting the estimation results of a wage-type equation in terms of specific NEG interactions.

4.5. Conclusions This chapter discusses some controversial issues of the NEG basic model and derives a new generalized wage-type equation with marginal costs as the dependent variable. This equation is used to encompass many previous wage equations found in the literature and to derive a testable equation including human and physical capital as explanatory variables. The NEG model is difficult to test and presents problems of observational equivalence. The estimation results of a wage-type equation can be interpreted in several ways. These are probably the reasons why some misunderstandings are frequent. In terms of a testable equation and real data, doubts remain about issues such as: the definition of the NEG sectors; the selection of the proper dependent variable; the measurement of the competition index; the production function of the agglomeration sector; the degree of geographical mobility of the production factors; and the time horizon studied by each researcher. Examples of other problematic issues include: the influence of the distance decay parameter when measuring Market Potential; the measurement of internal markets; the selection of control variables and exogenous instruments; and the role of technological differences and remote history in cross-sectional regressions with variables in levels. This discussion is a reminder of the lessons learned by Head and Mayer (2004b) from past work. On the one hand, methods should be connected closely to the theory but should not depend on features of models that were included for tractability. On the other hand we do not want to confirm the validity of NEG based on results that are also consistent with alternative theories.

A Generalized NEG Wage-Type Equation

79

References Baldwin R.E., Forslid R., Martin P., Ottaviano G. and Robert-Nicoud F. (2003). “Economic Geography and Public Policy”, Princeton University Press. Brakman S., Garretsen H. and van Marrewijk C. (2009). “The New Introduction to Geographical Economics”, Cambridge University Press. Breinlich, H. (2006). “The spatial income structure in the European Union—what role for Economic Geography?”, Journal of Economic Geography 6(5), 593–617. Bruna F., Faíña A. and Lopez-Rodriguez J. (2014). “Market Potential and the curse of distance in European regions”, MPRA Paper 56747, University Library of Munich. Bruna F., Lopez-Rodriguez J. and Faíña A. (2014). “Market Potential and Spatial Dependencies in the European Regions”, Regional Studies (resubmitted). Caselli, F. (2005). “Accounting for Cross-Country Income Differences”, in Aghion P. and Durlauf S. (eds.): Handbook of Economic Growth, pp. 679-741, Elsevier. Clemente J., Pueyo F. and Sanz F. (2009). “Market potential, European Union and growth”, Journal of Policy Modeling 31(5), 719–730. Combes P.-P., Mayer T. and Thisse J.-F. (2008). “Economic geography: the integration of regions and nations”, Princeton University Press. Duranton G. and Puga D. (2004). “Micro-foundations of urban agglomeration economies”, in Henderson J.V. and Thisse J.-F. (eds.): Handbook of Regional and Urban Economics 4, pp. 2063–2117, North Holland. Fingleton, B. (2006). “The new economic geography versus urban economics: an evaluation using local wage rates in Great Britain”, Oxford Economic Papers 58(3), 501–530. Fingleton B. and Fischer M.M. (2010). “Neoclassical theory versus new economic geography: competing explanations of cross-regional variation in economic development”, The Annals of Regional Science 44(3), 467–491. Fischer, M.M. (2011). “A spatial Mankiw–Romer–Weil model: theory and evidence”, The Annals of Regional Science 47(2), 419–436. Fujita M., Krugman P. and Venables A.J. (1999): “The spatial economy: cities, regions and international trade”, The MIT Press.

80

Chapter Four

Harris, C.D. (1954). “The Market as a Factor in the Localization of Industry in the United States”, Annals of the Association of American Geographers 44(4), 315–348. Head K. and Mayer T. (2004a). “Market potential and the location of Japanese investment in the European Union”, Review of Economics and Statistics 86(4), 959–972. Head K. and Mayer T. (2004b). “The empirics of agglomeration and trade”, in Henderson J.V. and Thisse J.-F. (eds.): Handbook of regional and urban economics 4, pp. 2609–2669, North Holland. Head K. and Mayer T. (2006). “Regional wage and employment responses to market potential in the EU”, Regional Science and Urban Economics 36(5), 573–594. Head K. and Mayer T. (2011). “Gravity, market potential and economic development”, Journal of Economic Geography 11(2), 281–294. Head K. and Mayer T. (2014). “Gravity Equations: Workhorse, Toolkit, and Cookbook”, in Helpman E., Rogoff K. and Gopinath G. (eds.): Handbook of International Economics 4, pp. 131–195, Elsevier. Holl, A. (2012). “Market potential and firm-level productivity in Spain”, Journal of Economic Geography 12(6), 1191–1215. Jones, C.I. (1997): “Convergence revisited”, Journal of economic Growth 2(2), 131–153. Krugman, P. (1980). "Scale economies, product differentiation, and the pattern of trade", American Economic Review 70(5), 950–959. —. (1992). “A Dynamic Spatial Model”, NBER Working Paper 4219, National Bureau of Economic Research. —. (2011). “The New Economic Geography, Now Middle-aged”, Regional Studies 45(1), 1–7. Li, Y. (2012). “Capital mobility, diminishing returns and wage inequality”, Journal of International Trade and Economic Development 21(3), 321–345. Mankiw N.G., Romer D. and Weil D.N. (1992). “A Contribution to the Empirics of Economic Growth”, The Quarterly Journal of Economics 107(2), 407–437. Puga, D. (1999). “The rise and fall of regional inequalities”, European Economic Review 43(2), 303–334. Redding, S.J. (2011). “Economic Geography: A Review of the Theoretical and Empirical Literature”, in Bernhofen D., Falvey R., Greenaway D. and Kreickemeier U. (eds.). Palgrave Handbook of International Trade, pp. 497–531, Palgrave Macmillan.

A Generalized NEG Wage-Type Equation

81

Redding S.J. and Schott P.K. (2003). “Distance, skill deepening and development: will peripheral countries ever get rich?”, Journal of Development Economics 72(2), 515–541. Redding S.J. and Venables A.J. (2004). “Economic geography and international inequality”, Journal of International Economics 62(1), 53–82. Robert-Nicoud, F. (2005). “The structure of simple ‘New Economic Geography’ models (or, On identical twins)”, Journal of Economic Geography 5(2), 201–234.

CHAPTER FIVE THE STATIONARITY OF THE INFLATION IN LATIN-AMERICAN COUNTRIES REVIEWED WHEN ADDITIVE OUTLIERS ARE DETECTED DIONISIO RAMÍREZ-CARRERA AND GABRIEL RODRÍGUEZ*

The proper inference of statistical parameters proves to be crucial for the forecast of the evolution of macroeconomic variables. According to the Lucas critique, changes on the parameters’ values have serious implications for the design and implementation of economic policies. In an environment of growing economic interdependence and financial turmoil, the behaviour of inflation series is not a trivial question. This chapter analyses the behaviour of eight quarterly Latin-American inflation series when additive outliers are detected, determining if they follow a stationarity or a unit root process. Additive outliers have undesirable effects on the inference of statistical parameters. In the context of integrated data, it is well recognized that outliers also affect the properties of unit root tests. The application of standard unit root tests and the more robust M-tests suggest rejection of the null hypothesis of a unit root when additive outliers are not controlled. However, using Perron and Rodríguez’s (2003) statistic IJd to test for the presence and location of the additive outliers, the Latin-American inflation series show a strong presence of large sized additive outliers. When the ADF statistic is corrected for the presence of additive outliers, the results are quite different, which is consistent with other works.

*

Corresponding author: Gabriel Rodríguez, Departamento de Economía, Pontificia Universidad Católica del Perú (PUCP), Av. Universitaria 1801, Lima 32, Perú, Email Address: [email protected]

84

Chapter Five

5.1. Introduction Nowadays, the main target of central banks is to control inflation at low levels, so that inflation persistence and price-level stickiness have become some of the most important issues of the study of inflation, since this variable plays a central role in the design of monetary policy. In this sense, knowing the degree of the persistence of inflation and its underlying causes gives the central bank essential information about how to reach the target through its policy instrument. It may also have a great impact on the expectations and behaviours of private agents. Prices, wages and any other nominal variables should be adjusted attending to inflation, which may result in an inflationary spiral because of the influence of expectations over their own persistence. If policy makers’ decisions are based on the wrong knowledge of the statistical properties of inflation, the economic policy measures will be ineffective in the sense of Lucas (1976). As a consequence, it is very relevant to be able to estimate the true behaviour of inflation rates to avoid problems arising from Lucas’ critique about the stability of the parameters value. Moreover, inflation persistence has important implications for economic and financial theories like the Taylor rule, the Philips curve, the Non-Accelerating Inflation Rate of Unemployment (NAIRU) and the Fisher effect1. However, in practice, it is difficult to measure the existence of persistence in inflation rates, in spite of the importance of this statistical property in macroeconomic theories. So, there is not a consensus about the degree and stability of inflation persistence. Although recent works, focusing on development countries, like O’Reilly and Whelan (2005), Batini (2006), and Pivetta and Reis (2007), have found evidence of persistence in post-war inflation series, according to Levin and Piger (2004), there is a lot of controversy about the statistical techniques employed. A normal way to evaluate inflation persistence is to determine whether it is stationary. Previous studies have analysed the persistence in inflation rates through unit root tests to determine whether the null hypothesis of unit root —Ho: I(1)— can be rejected in favour of the alternative hypothesis of stationarity —Ha: I(0)—. If null hypothesis cannot be rejected, it means that price shocks permanently affect the level of inflation and tend to push the inflation rate away from its steady state

1 For a deeper review of these implications, see Fuhrer (2009), Romero-Ávila and Usabiaga (2009) or Caporale and Paxton (2013).

Stationarity of the Inflation in Latin-American Countries Reviewed

85

forever. However, if the inflation rate follows a stationary process, all shocks will have a short-lasting effect and will fade over time. The traditional unit root test is the Augmented Dickey-Fuller (ADF) test. This test consists of regressing the first difference of the variable on a constant, its lagged level and k lagged differences: ο‫ݕ‬௧ ൌ ߤ ൅ ߙ‫ݕ‬௧ିଵ ൅ σ௞௜ୀଵ ݀௜ ο‫ݕ‬௧ି௜ ൅ ߝ௧

(1)

where yt is the inflation rate. The null hypothesis is rejected if the coefficient Į is significantly less than zero. The lagged differenced terms remove serial correlation in the errors and their numbers depend on the criteria used2. Since the seminal work of Nelson and Plosser (1982), a great number of works have studied inflation persistence using unit root tests. However, there is not a clear conclusion in the empirical literature around whether inflation is a stationary or a unit root process3. Gadea and Mayoral (2006) point out that most of the economic literature only consider I(1) or I(0) processes. Nevertheless, it is possible to find situations where this consideration may be too simplistic, since many variables react to shocks in a different way. Thus, some shocks are not permanent but its effects take a long time to disappear. This can be the reason why the order of integration as the measure of inflation persistence, using unit root tests, shows divergent results which are not consistent with either an I(1) or an I(0) process, especially in the case of the United States4. Finally, the majority of the research on inflation stationarity has focused on developed countries (G7 or OECD), but there are few findings on this issue with regard to the Latin American nations and most of them have been done focusing only on a single country5. So, in this work, we shed light on the inflation properties of eight Latin American countries (Argentina, Bolivia, Chile, Colombia, Ecuador, Peru, Uruguay and Venezuela) using appropriate statistical techniques to deal with series that have suffered high inflation episodes with an average inflation rate, in some cases, of more than 500%. Such episodes have occurred frequently 2

For further details about the criteria of lag length selection see Ng and Perron (1995), and Ng and Perron (2001). 3 In Henry and Shields (2004) and Arize (2011) can be found a comprehensive overview of works with results in favor and against the existence of a unit root in inflation rates. 4 See Kumar and Okimoto (2007). 5 See Cati et al. (1999), Thornton (2008), Chiquiar et al. (2010) or Oliveira and Petrassi (2014).

86

Chapter Five

during the 80s and 90s, but have disappeared after the introduction of inflation targeting6. Therefore, according to Chan (1998), those extreme values are inconsistent with the rest of the series and can be considered as outliers. Additive outliers affect inference of parameters in different circumstances. For example, they affect the inference of the autoregressive and moving average estimates of ‫ܣܯܴܣ‬ሺ‫݌‬ǡ ‫ݍ‬ሻ models (see Cheng and Liu (1993), and Chan (1992, 1995)). They also affect other topics like causality tests (see Baldé and Rodríguez (2005)) and fractional estimates (see Fajardo et al. (2009), and Chareka et al. (2006)). In the context of a unit root, additive outliers have also been analysed since the contribution of Franses and Haldrup (1994). These authors show that additive outliers contaminate the limiting distribution of the unit root statistics (also see Vogelsang (1999), and Perron and Rodríguez (2003)). Vogelsang (1999) suggests using Mtests based on GLS de-trended data because they are robust with regard to the presence of negative moving average autocorrelation, which is induced by the presence of additive outliers. Another alternative procedure is to estimate an ADF statistic corrected for dummy variables related to the identified additive outliers in a preliminary step. Rodríguez (2004) used four Latin-American inflation series and shows that even the M-tests indicate a rejection of the null hypothesis of a unit root. When applying an ADF corrected for dummy variables, some countries show a non-rejection of the null hypothesis of a unit root, indicating non-stationarity of the inflation series which is an opposite result obtained from the standard unit root tests. The procedure mentioned above needs to know the location of the additive outliers. Perron and Rodríguez (2003) have suggested a powerful test, denoted by Ϙd, which works with first-differenced data7. This procedure is more powerful than others based on levels of the data, for example (see Perron and Rodríguez (2003) for a detailed discussion). These authors claim that Ϙd is powerful even for departures from the unit root case8. The purpose of this note is to show that this claim is correct.

6

See Roache (2013). Of course, there are many other procedures to identify outliers, for example, those proposed in Tsay (1986), Chang et al. (1988), Shin et al. (1996), Chen and Liu (1993) and Gómez and Maravall (1992). Another interesting approach is proposed by Lucas (1995a, 1995b), and Hoek et al. (1995). See Rodríguez (2004) for a comparison with other approaches. 8 However, this procedure is not robust to departures from the assumption of normality in the errors. It is mentioned and discussed by Perron and Rodríguez 7

Stationarity of the Inflation in Latin-American Countries Reviewed

87

We do it to analyse the empirical size of the ADF statistic (using Ϙd to locate additive outliers) when the DGP contains ARFIMA (p,d,q) errors. The experiment deals with different values of the fractional parameter (d) to observe different departures from the unit root hypothesis. Also, the different structure of autocorrelation is analysed (moving average and autoregressive). The Monte-Carlo simulations show that the ADF statistic corrected for dummy variables associated with the additive outliers suffers from size distortions in only a few cases. For example, when the moving average parameter is close to -1, the empirical size is greater than the nominal size. Negative autoregressive autocorrelation has impacts on the size of the ADF statistic too. However, most of these issues are fixed when the sample size increases from ܶ ൌ ͳͲͲ to ܶ ൌ ʹͲͲ in simulations. In general, the ADF test appears to be slightly undersized. When the fractional parameter is higher, distortions appear but at the same time as when correlation is higher. Therefore, the fractional parameter itself does not cause problems with or distortions in the size of the ADF test. After simulations, we present an empirical application using the quarterly inflation series of eight countries, from 1970:1 until 2010:4. We use a sample of eight countries and the spirit of this exercise is very similar to Rodríguez (2004), where only four countries were used. In this study, we add more countries and more observations. In particular, the Phillips and Perron (1988) statistic shows a strong rejection of the null hypothesis of a unit root, which is not rare given the sensitivity of this statistic to the presence of strong negative moving average correlations, which is the case here because additive outliers are clearly present and the literature has shown that they are related to this type of correlation. Similar results are obtained with the ADF statistic and even with the M-tests and MPT tests using GLS de-trended data as suggested by Elliott et al. (1996) and Ng and Perron (2001), respectively. Only Uruguay and Venezuela show non-rejection of the null. However, when applying the ADF test augmented by dummy variables related to the location of the additive outliers identified by the procedure Ϙd, none of the countries reject the null hypothesis of a unit root.

(2003). See also Burridge and Taylor (2006) for more evidence about this drawback and the correction they propose based on the extreme value theory.

88

Chapter Five

This chapter is organized as follows. Section 5.2 presents the model, discusses the issue of outlier detection and briefly revises the method proposed by Perron and Rodríguez (2003). In section 5.3, we present the results from the simulations. Section 5.4 shows the empirical application and Section 5.5 concludes.

5.2. The Issue of Outlier Detection and Testing for Unit Roots with Additive Outliers The issue of outlier detection in the unit root framework is the approach taken by Perron and Rodríguez (2003), which is based on Vogelsang (1999)9. The data-generating process entertained is of the following general form: ‫ݕ‬௧ ൌ ݀௧ ൅ σ௠ ௝ୀଵ ߜ௝ ‫ܦ‬൫ܶ௔௢ǡ௝ ൯௧ ൅ ‫ݑ‬௧

(2)

where ‫ܦ‬ሺܶ௔௢ǡ௝ ሻ௧ ൌ ͳ if ‫ ݐ‬ൌ ܶ௔௢ǡ௝ and 0 otherwise. This permits the presence of m additive outliers occurring at dates ܶ௔௢ǡ௝ ሺ݆ ൌ ͳǡ ǥ ǡ ݉ሻ. The term dt specifies the deterministic components. In most cases, ݀௧ ൌ ߤ if the series is non-trending, or ݀௧ ൌ ߤ ൅ ߚ௧ if the series is trending. The noise function is integrated in order one, i.e., ‫ݑ‬௧ ൌ ‫ݑ‬௧ିଵ ൅ ‫ݒ‬௧ ; where ‫ݒ‬௧ is a stationary process. While Perron and Rodríguez (2003) use an ‫ܣܯܴܣ‬ሺ‫݌‬ǡ ‫ݍ‬ሻ for the process ‫ݒ‬௧ , in this paper, we assume that ‫ݒ‬௧ is an ‫ܣܯܫܨܴܣ‬ሺ‫݌‬ǡ ݀ǡ ‫ݍ‬ሻ process. As shown in Perron and Rodríguez (2003), the original procedure of Vogelsang (1999) has severe size distortions when applied in an iterative fashion to search for additive outliers. The reason for this is that the limiting distribution of the statistic is only valid in the first step of the

9

Let ‫ݐ‬ఋ෡ ሺܶ௔௢ ሻ denote the t-statistic for testing į = 0 in (1). Following Chen and Liu (1993), the presence of an additive outlier can be tested using ߬ௗ ൌ ‫ೌ்݌ݑݏ‬೚ ห‫ݐ‬ఋ෡ ሺܶ௔௢ ሻห. Assuming that ߣ ൌ ܶ௔௢ Ȁܶ remains fixed as ܶ grows, Vogelsang (1999) showed that as ܶ ՜ λ, the limiting distribution of ‫ݐ‬ఋ෡ ሺܶ௔௢ ሻ is ଵ non-standard. More precisely, ‫ݐ‬ఋ෡ ሺܶ௔௢ ሻ ֜ ‫ܪ‬ሺߣሻ ൌ ܹ ‫ כ‬ሺߣሻȀሺ‫׬‬଴ ܹ ‫ כ‬ሺ‫ݎ‬ሻଶ ݀‫ݎ‬ሻଵȀଶ , where ܹ ‫ כ‬ሺߣሻ denotes a demeaned standard Wiener process. If (2) also includes a time trend, ܹ ‫ כ‬ሺߣሻ will denote a detrended Wiener process. Furthermore, from the continuous mapping theorem it follows that, ߬ ֜ ‫݌ݑݏ‬ఒఢሺ଴ǡଵሻ ȁ‫ܪ‬ሺߣሻȁ ‫ כܪ ؠ‬. This distribution is invariant with respect to any nuisance parameters, including the correlation structure of the noise function.

Stationarity of the Inflation in Latin-American Countries Reviewed

89

iterations as specified in Theorem 1 of Perron and Rodríguez (2003). In subsequent steps, the asymptotic critical values used need to be modified. Perron and Rodríguez (2003) have proposed a more powerful iterative strategy using a test based on first-differences of the data. Consider data generated by (2) with ݀௧ ൌ ߤ, and a single outlier occurring at date ܶ௔௢ with magnitude į. Then, (3) ο‫ݕ‬௧ ൌ ߜሾ‫ܦ‬ሺܶ௔௢ ሻ௧ െ ‫ܦ‬ሺܶ௔௢ ሻ௧ିଵ ሿ ൅ ‫ݒ‬௧ where ‫ܦ‬ሺܶ௔௢ ሻ௧ ൌ ͳ if ‫ ݐ‬ൌ ܶ௔௢ (0, otherwise) and ‫ܦ‬ሺܶ௔௢ ሻ௧ିଵ ൌ ͳ if ‫ ݐ‬ൌ ܶ௔௢ ൅ ͳ (0, otherwise). If the data are trending, a constant should be included. In this case, we are interested in ߬ௗ ൌ ‫ೌ்݌ݑݏ‬೚ ห‫ݐ‬ఋ෡ ሺܶ௔௢ ሻห, where ෡ ఋ భȀమ ෠ ሺ଴ሻିோ ೡ ೡ ሺଵሻሻȀଶሿ

‫ݐ‬ఋ෡ ሺܶ௔௢ ሻ ൌ ሾሺோ෠

and ܴ௩ ሺ݆ሻ is the autocovariance function of

10

‫ݒ‬௧ at delay j . To detect for multiple outliers, we can follow a strategy similar to that suggested by Vogelsang (1999), by dropping the observation labelled as an outlier before proceeding to the next step. The important feature is that, unlike the case of the test based on levels, the limit distribution of the test ߬ௗ is the same as each step of the iterations when dealing with multiple outliers. The disadvantage of this procedure, compared to that based on the level of the data, is that the limiting distribution depends on the specific distribution of the errors ‫ݒ‬௧ , though not on the presence of serial correlation and heteroskedasticity11. This problem is exactly the same as that for finding outliers in stationary time series. In this chapter, we analyse the empirical size of the ADF test corrected for detected additive outliers when errors ‫ݒ‬௧ are ARFIMA(p,d,q) processes. It is equivalent to using the t-statistic for testing that ߙ ൌ ͳ in the following regression: ௣ାଵ ௞ ‫ݕ‬௧ ൌ ߤ ൅ ߙ‫ݕ‬௧ିଵ ൅ σ௜ୀ଴ σ௠ ௝ୀଵ ߜ௜௝ ‫ܦ‬ሺܶ௔௢௝ ሻ௧ି௜ ൅ σ௜ୀ଴ ݀௜ ο‫ݕ‬௧ି௜ ൅ ‫ݒ‬௧ (4) where ‫ܦ‬ሺܶ௔௢ǡ௝ ሻ௧ି௜ ൌ ͳ if ‫ ݐ‬ൌ ܶ௔௢ǡ௝ ൅ ݅ and 0 otherwise, with ܶ௔௢ǡ௝ ሺ݆ ൌ ͳǡ ʹǡ ǥ ǡ ݉ሻ being the dates of the outliers identified using the statistic ߬ௗ . Notice that ݇ ൅ ʹ one-time dummy variables have to be included in (4) to remove all possible influences of the additive outliers.

்ି௝ ܴ෠௩ ሺ݆ሻ ൌ ܶ ିଵ σ௧ୀଵ ‫ݒ‬ො௧ ‫ݒ‬ො௧ା௝ with ‫ݒ‬௧ the least-squares residuals obtained from regression (3). Then, ܴ෠௩ ሺ݆ሻ is a consistent estimate of ܴ௩ ሺ݆ሻ. 11 The dependence of the distribution or departures of the normality of ‫ݒ‬௧ has been mentioned by Perron and Rodríguez (2003). However, Burridge and Taylor (2006) deals with this issue using extreme value theory. 10

90

Chapter Five

5.3. Monte Carlo Results In order to analyse the empirical size of the ‫ ܨܦܣ‬statistic, we consider the following experiment. Let ‫ݕ‬௧ follow (1) where ‫ݑ‬௧ ൌ ‫ݑ‬௧ିଵ ൅ ‫ݒ‬௧ (a unit root process) and ‫ݒ‬௧ is an ‫ܣܯܫܨܴܣ‬ሺ‫݌‬ǡ ݀ǡ ‫ݍ‬ሻ process, that is ߩሺ‫ܮ‬ሻሺͳ െ ‫ܮ‬ሻௗ ‫ݒ‬௧ ൌ ߠሺ‫ܮ‬ሻߝ௧ , where ߝ௧ is an ݅Ǥ ݅Ǥ ݀Ǥ ܰሺͲǡͳሻ. More precisely, in one case we consider ‫ ݌‬ൌ ͳሺߩሺ‫ܮ‬ሻ ൌ ͳ െ ߩ‫ܮ‬ሻ and ‫ ݍ‬ൌ Ͳ, that is ߩሺ‫ܮ‬ሻሺͳ െ ‫ܮ‬ሻௗ ‫ݒ‬௧ ൌ ߝ௧ , while in the other ‫ ݌‬ൌ Ͳ case and ‫ ݍ‬ൌ ͳሺߠሺ‫ܮ‬ሻ ൌ ͳ ൅ ߠ‫ܮ‬ሻ, that is ሺͳ െ ‫ܮ‬ሻௗ ‫ݒ‬௧ ൌ ߠሺ‫ܮ‬ሻߝ௧ . The fractional parameter is ݀ ‫ א‬ሾെͲǤͶͺ‫Ͳ݋ݐ‬ǤͶͺሿ with a step of ͲǤͳʹ. Each Table and each value of ߠ or ߩpresent three rows named “without”, “with”, and “total”. The row named “without” indicates the size of the ‫ ܨܦܣ‬statistic when no additive outliers have been found. The word “with” indicates the size of the ‫ ܨܦܣ‬statistic when additive outliers have been identified. Therefore, the row entitled “total” simply means the sum of the two previous rows. If the size is correct, we expect that this row should be close to the nominal size of ͷǤͲΨ. In order to save space, we present only selected Tables. Each experiment is performed using 10,000 replications, nominal size at ͷǤͲΨ, and we use tabulated critical values (Table 1 of Perron and Rodríguez (2003)) for ܶ ൌ ͳͲͲ and ܶ ൌ ʹͲͲ. Other extensive Tables are available upon request. In all Tables, the total iterative procedure is applied; that is, we search for all outliers and finish the procedure when no outliers are found. Two sets of Tables are presented. In one case, the lag length of (4) is fixed at ݇ ൌ ͳ, while in the other case, we use the procedure t-sig proposed by Campbell and Perron (1991) for ݇ ‫ א‬ሾͲǡͷሿ. In each Table, three cases are presented. In the first case, no outliers are in the process; that is, ߜ௜ ൌ Ͳ for ݅ ൌ ͳǡ ʹǡ ͵ǡ Ͷ. In the second case, we consider medium sized additive outliers: ߜ௜ ൌ ͷǡ ͵ǡ ʹǡ ʹ. The final case is for large sized additive outliers; that is, ߜ௜ ൌ ͳͲǡ ͷǡ ͷǡ ͷ. In summary, the design of the experiment closely follows that of Perron and Rodríguez (2003). When there are outliers, a maximum of four additive outliers are considered and they are located at positions ͲǤʹͲܶ, ͲǤͶͲܶ, ͲǤ͸Ͳܶ and ͲǤͺͲܶ, respectively. Table 5.1 shows the results for the case where errors are ‫ܣܯܫܨܴܣ‬ሺͲǡ ݀ǡ Ͳሻ. The first set of columns show the case where no outliers are present in the data. The other columns show medium and large sized additive outliers, respectively. The results show that the size of the ADF is oversized for every ݀ ൏ Ͳ. More negative values of d imply more oversized ‫ ܨܦܣ‬tests. This is true for the case where no outliers are found and when they are present in the data. For other values of ݀, the ‫ ܨܦܣ‬is

Stationarity of the Inflation in Latin-American Countries Reviewed

91

slighthly undersized but close to the nominal size of ͷΨ. Given these results, in what follows, we do not consider cases where ݀ ൏ Ͳ. Tables 5.2a–5.2c show the size of the ‫ ܨܦܣ‬test for ‫ܣܯܫܨܴܣ‬ሺͲǡ ݀ǡ ͳሻ errors; that is, when a moving average correlation exists. In order to save space, we only show the results for ݀ ൌ ͲǤͲͲ, ͲǤʹͶ, and ͲǤͶͺ. Table 5.2a indicates that the ‫ ܨܦܣ‬test is oversized for ߠ ൌ െͲǤͺ and for ߠ ൌ െͲǤͶ. A small distortion is also found for ߠ ൌ ͲǤͺ. In all other cases of ߠ and for cases where there are or are not additive outliers, the exact size is close to ͷΨ. Table 5.2b shows the case for ݀ ൌ ͲǤʹͶ. Again, the ‫ ܨܦܣ‬test is oversized for ߠ ൌ െͲǤͺ but distortions are smaller than before. In all other cases, the size is better although slightly undersized. When ݀ ൌ ͲǤͶͺ (Table 5.2c) – that is, when the memory of the errors is large – the size of the ADF test is very close to the nominal size of ͷΨ. It is true for when there are or are not additive outliers and for both sample sizes. In summary, there are some difficulties when ߠ goes to െͳ but the performance is better when ݀ goes to ͲǤͷ.

Chapter Five

92

Table 5-1: Size of the ADF Test: ARFIMA (0, d, 0) Errors* ߜͳ ൌ Ͳǡ ߜʹ ൌ Ͳǡ ߜ͵ ൌ Ͳǡ ߜͶ ൌ Ͳ ܶ ൌ ͳͲͲ ݀ ൌ  െͲǤͶͺ

݀ ൌ  െͲǤ͵͸

݀ ൌ  െͲǤʹͶ

݀ ൌ  െͲǤͳʹ

݀ ൌ ͲǤͲͲ

݀ ൌ ͲǤͳʹ

݀ ൌ ͲǤʹͶ

݀ ൌ ͲǤ͵͸

݀ ൌ ͲǤͶͺ

ܶ ൌ ʹͲͲ

ߜͳ ൌ ͷǡ ߜʹ ൌ ͵ǡ ߜ͵ ൌ ʹǡ ߜͶ ൌ ʹ ܶ ൌ ͳͲͲ

ܶ ൌ ʹͲͲ

ߜͳ ൌ ͳͲǡ ߜʹ ൌ ͷǡ ߜ͵ ൌ ͷǡ ߜͶ ൌ ͷ ܶ ൌ ͳͲͲ

ܶ ൌ ʹͲͲ

Without

0.792

0.940

0.075

0.051

0.000

0.000

With

0.038

0.048

0.779

0.938

0.737

0.979

Total

0.830

0.988

0.854

0.989

0.737

0.979

Without

0.487

0.767

0.032

0.021

0.000

0.000

With

0.023

0.037

0.514

0.792

0.428

0.761

Total

0.510

0.805

0.546

0.813

0.428

0.761

Without

0.229

0.385

0.009

0.005

0.000

0.000

With

0.011

0.018

0.242

0.403

0.188

0.361

Total

0.240

0.403

0.251

0.408

0.188

0.361

Without

0.088

0.122

0.002

0.001

0.000

0.000

With

0.004

0.007

0.092

0.129

0.076

0.116

Total

0.092

0.129

0.094

0.130

0.076

0.116

Without

0.037

0.036

0.000

0.000

0.000

0.000

With

0.001

0.002

0.038

0.038

0.037

0.038

Total

0.038

0.038

0.038

0.038

0.037

0.038

Without

0.022

0.022

0.000

0.000

0.000

0.000

With

0.001

0.001

0.023

0.023

0.023

0.022

Total

0.023

0.023

0.023

0.023

0.023

0.022

Without

0.021

0.026

0.000

0.000

0.000

0.000

With

0.001

0.002

0.022

0.028

0.022

0.026

Total

0.022

0.028

0.022

0.028

0.022

0.026

Without

0.023

0.032

0.000

0.000

0.000

0.000

With

0.001

0.002

0.024

0.033

0.022

0.030

Total

0.024

0.034

0.024

0.033

0.022

0.030

Without

0.022

0.030

0.000

0.000

0.000

0.000

With

0.001

0.001

0.026

0.033

0.025

0.035

Total

0.023

0.031

0.026

0.033

0.025

0.035

*Lag length fixed at one

Stationarity of the Inflation in Latin-American Countries Reviewed

93

Table 5-2a: Size of the ADF Test: ARFIMA (0, d, 1) Errors with d=0.00* ߜͳ ൌ Ͳǡ ߜʹ ൌ Ͳǡ

ߜͳ ൌ ͷǡ ߜʹ ൌ ͵ǡ

ߜͳ ൌ ͳͲǡ ߜʹ ൌ ͷǡ

ߜ͵ ൌ Ͳǡ ߜͶ ൌ Ͳ

ߜ͵ ൌ ʹǡ ߜͶ ൌ ʹ

ߜ͵ ൌ ͷǡ ߜͶ ൌ ͷ

ܶ ൌ ͳͲͲ ߠ ൌ  െͲǤͺͲ

ߠ ൌ  െͲǤͶͲ

ߠ ൌ ͲǤͲͲ

ߠ ൌ ͲǤͶͲ

ߠ ൌ ͲǤͺͲ

ܶ ൌ ʹͲͲ

ܶ ൌ ͳͲͲ

ܶ ൌ ʹͲͲ

ܶ ൌ ͳͲͲ

ܶ ൌ ʹͲͲ

Without

0.819

0.846

0.250

0.221

0.000

0.000

With

0.043

0.045

0.651

0.692

0.845

0.891

Total

0.862

0.891

0.901

0.913

0.845

0.891

Without

0.094

0.103

0.012

0.006

0.000

0.000

With

0.005

0.005

0.110

0.114

0.086

0.101

Total

0.099

0.108

0.122

0.120

0.086

0.101

Without

0.037

0.036

0.000

0.000

0.000

0.000

With

0.001

0.002

0.038

0.038

0.037

0.038

Total

0.038

0.038

0.038

0.038

0.037

0.038

Without

0.059

0.060

0.000

0.000

0.000

0.000

With

0.002

0.002

0.045

0.053

0.054

0.056

Total

0.061

0.062

0.045

0.053

0.054

0.056

Without

0.094

0.098

0.000

0.000

0.000

0.000

With

0.002

0.003

0.064

0.079

0.082

0.090

Total

0.096

0.101

0.064

0.079

0.082

0.090

*Lag length fixed at one

Chapter Five

94

Table 5-2b: Size of the ADF Test: ARFIMA (0, d, 1) Errors with d=0.24*

ߠ ൌ  െͲǤͺͲ

ߠ ൌ  െͲǤͶͲ

ߠ ൌ ͲǤͲͲ

ߠ ൌ ͲǤͶͲ

ߠ ൌ ͲǤͺͲ

ߜͳ ൌ Ͳǡ ߜʹ ൌ Ͳǡ

ߜͳ ൌ ͷǡ ߜʹ ൌ ͵ǡ

ߜͳ ൌ ͳͲǡ ߜʹ ൌ ͷǡ

ߜ͵ ൌ Ͳǡ ߜͶ ൌ Ͳ

ߜ͵ ൌ ʹǡ ߜͶ ൌ ʹ

ߜ͵ ൌ ͷǡ ߜͶ ൌ ͷ

ܶ ൌ ͳͲͲ

ܶ ൌ ʹͲͲ

ܶ ൌ ͳͲͲ

ܶ ൌ ʹͲͲ

ܶ ൌ ͳͲͲ

Without

0.264

0.223

0.048

0.027

0.000

ܶ ൌ ʹͲͲ 0.000

With

0.013

0.001

0.266

0.223

0.246

0.220

Total

0.277

0.234

0.314

0.250

0.246

0.220

Without

0.023

0.025

0.001

0.000

0.000

0.000

With

0.001

0.002

0.023

0.025

0.021

0.024

Total

0.024

0.027

0.024

0.025

0.021

0.024

Without

0.021

0.026

0.000

0.000

0.000

0.000

With

0.001

0.002

0.022

0.028

0.022

0.026

Total

0.022

0.028

0.022

0.028

0.022

0.026

Without

0.023

0.024

0.000

0.000

0.000

0.000

With

0.001

0.001

0.024

0.024

0.025

0.026

Total

0.024

0.025

0.024

0.024

0.025

0.026

Without

0.033

0.028

0.000

0.000

0.000

0.000

With

0.001

0.001

0.029

0.029

0.042

0.035

Total

0.034

0.029

0.029

0.029

0.042

0.035

*Lag length fixed at one

Stationarity of the Inflation in Latin-American Countries Reviewed

95

Table 5-2c: Size of the ADF Test: ARFIMA (0, d, 1) Errors with d=0.48* ߜͳ ൌ Ͳǡ ߜʹ ൌ Ͳǡ

ߜͳ ൌ ͷǡ ߜʹ ൌ ͵ǡ

ߜͳ ൌ ͳͲǡ ߜʹ ൌ ͷǡ

ߜ͵ ൌ Ͳǡ ߜͶ ൌ Ͳ

ߜ͵ ൌ ʹǡ ߜͶ ൌ ʹ

ߜ͵ ൌ ͷǡ ߜͶ ൌ ͷ

ܶ ൌ ͳͲͲ ߠ ൌ  െͲǤͺͲ

ߠ ൌ  െͲǤͶͲ

ߠ ൌ ͲǤͲͲ

ߠ ൌ ͲǤͶͲ

ߠ ൌ ͲǤͺͲ

ܶ ൌ ʹͲͲ

ܶ ൌ ͳͲͲ

ܶ ൌ ʹͲͲ

ܶ ൌ ͳͲͲ

ܶ ൌ ʹͲͲ

Without

0.059

0.041

0.004

0.001

0.000

0.000

With

0.003

0.002

0.058

0.040

0.053

0.040

Total

0.062

0.043

0.062

0.041

0.053

0.040

Without

0.033

0.052

0.000

0.000

0.000

0.000

With

0.002

0.003

0.030

0.053

0.030

0.049

Total

0.035

0.055

0.030

0.053

0.030

0.049

Without

0.022

0.030

0.000

0.000

0.000

0.000

With

0.001

0.001

0.026

0.033

0.025

0.035

Total

0.023

0.031

0.026

0.033

0.025

0.035

Without

0.018

0.020

0.000

0.000

0.000

0.000

With

0.001

0.001

0.032

0.041

0.048

0.055

Total

0.019

0.021

0.032

0.041

0.048

0.055

Without

0.024

0.019

0.000

0.000

0.000

0.000

With

0.001

0.001

0.044

0.048

0.078

0.079

Total

0.025

0.020

0.044

0.048

0.078

0.079

*Lag length fixed at one

Tables 5.3a–5.3c show size of the ‫ ܨܦܣ‬test for ‫ܣܯܫܨܴܣ‬ሺͳǡ ݀ǡ Ͳሻ errors; that is, when autoregressive autocorrelation exists. Again, in order to save space, we only show results for ݀ ൌ ͲǤͲͲ,ͲǤʹͶ, and ͲǤͶͺ. Table 5.3a indicates that the ‫ ܨܦܣ‬test has a good, exact size except for the case where ߩ ൌ െͲǤͺ and when the process is contaminated for medium and large sized additive outliers. It is worth mentioning that the distortions are smaller compared to the previous Tables and we observe that size is better when the sample size is larger. Table 5.3b shows the case for ݀ ൌ ͲǤʹͶ. In this case, the ‫ ܨܦܣ‬test has an exact size close to the ͷΨ, although we observe small oversized results when ߩ ൌ ͲǤͺ. This results is more evident when ݀ ൌ ͲǤͶͺ (Table 5.3c), even when there are no outliers in the process. This problem is not fixed when the sample size is larger. It is more evident for the extreme values of ߩሺെͲǤͺܽ݊݀ͲǤͺሻ.

Chapter Five

96

Table 5-3a: Size of the ADF Test: ARFIMA (1, d, 0) Errors with d=0.00*

ߠ ൌ  െͲǤͺͲ

ߠ ൌ  െͲǤͶͲ

ߠ ൌ ͲǤͲͲ

ߠ ൌ ͲǤͶͲ

ߠ ൌ ͲǤͺͲ

ߜͳ ൌ Ͳǡ ߜʹ ൌ Ͳǡ

ߜͳ ൌ ͷǡ ߜʹ ൌ ͵ǡ

ߜͳ ൌ ͳͲǡ ߜʹ ൌ ͷǡ

ߜ͵ ൌ Ͳǡ ߜͶ ൌ Ͳ

ߜ͵ ൌ ʹǡ ߜͶ ൌ ʹ

ߜ͵ ൌ ͷǡ ߜͶ ൌ ͷ

ܶ ൌ ͳͲͲ

ܶ ൌ ʹͲͲ

ܶ ൌ ͳͲͲ

ܶ ൌ ʹͲͲ

ܶ ൌ ͳͲͲ

ܶ ൌ ʹͲͲ

Without

0.037

0.037

0.080

0.053

0.018

0.004

With

0.001

0.001

0.026

0.018

0.139

0.095

Total

0.038

0.038

0.106

0.071

0.157

0.099

Without

0.035

0.037

0.009

0.004

0.000

0.000

With

0.002

0.002

0.046

0.044

0.041

0.039

Total

0.037

0.039

0.055

0.048

0.041

0.039

Without

0.037

0.036

0.000

0.000

0.000

0.000

With

0.001

0.002

0.038

0.038

0.037

0.038

Total

0.038

0.038

0.038

0.038

0.037

0.038

Without

0.035

0.036

0.000

0.000

0.000

0.000

With

0.002

0.002

0.033

0.035

0.037

0.038

Total

0.037

0.038

0.033

0.035

0.037

0.038

Without

0.040

0.040

0.000

0.000

0.000

0.000

With

0.001

0.001

0.035

0.040

0.049

0.046

Total

0.041

0.041

0.035

0.040

0.049

0.046

*Lag length fixed at one

Stationarity of the Inflation in Latin-American Countries Reviewed

97

Table 5-3b: Size of the ADF Test: ARFIMA (1, d, 0) Errors with d=0.24*

ߠ ൌ  െͲǤͺͲ

ߠ ൌ  െͲǤͶͲ

ߠ ൌ ͲǤͲͲ

ߠ ൌ ͲǤͶͲ

ߠ ൌ ͲǤͺͲ

ߜͳ ൌ Ͳǡ ߜʹ ൌ Ͳǡ

ߜͳ ൌ ͷǡ ߜʹ ൌ ͵ǡ

ߜͳ ൌ ͳͲǡ ߜʹ ൌ ͷǡ

ߜ͵ ൌ Ͳǡ ߜͶ ൌ Ͳ

ߜ͵ ൌ ʹǡ ߜͶ ൌ ʹ

ߜ͵ ൌ ͷǡ ߜͶ ൌ ͷ

ܶ ൌ ͳͲͲ

ܶ ൌ ʹͲͲ

ܶ ൌ ͳͲͲ

ܶ ൌ ʹͲͲ

ܶ ൌ ͳͲͲ

ܶ ൌ ʹͲͲ

Without

0.031

0.044

0.010

0.016

0.001

0.000

With

0.001

0.001

0.012

0.019

0.022

0.031

Total

0.032

0.045

0.022

0.035

0.023

0.031

Without

0.025

0.034

0.001

0.000

0.000

0.000

With

0.001

0.002

0.022

0.033

0.025

0.033

Total

0.026

0.036

0.023

0.033

0.025

0.033

Without

0.021

0.026

0.000

0.000

0.000

0.000

With

0.001

0.002

0.022

0.028

0.022

0.026

Total

0.022

0.028

0.022

0.028

0.022

0.026

Without

0.022

0.024

0.000

0.000

0.000

0.000

With

0.001

0.002

0.025

0.025

0.027

0.026

Total

0.023

0.026

0.025

0.025

0.027

0.026

Without

0.072

0.056

0.000

0.000

0.000

0.000

With

0.001

0.002

0.064

0.073

0.096

0.092

Total

0.073

0.058

0.064

0.073

0.096

0.092

*Lag length fixed at one

Chapter Five

98

Table 5-3c: Size of the ADF Test: ARFIMA (1, d, 0) Errors with d=0.48* ߜͳ ൌ Ͳǡ ߜʹ ൌ Ͳǡ

ߜͳ ൌ ͷǡ ߜʹ ൌ ͵ǡ

ߜͳ ൌ ͳͲǡ ߜʹ ൌ ͷǡ

ߜ͵ ൌ Ͳǡ ߜͶ ൌ Ͳ

ߜ͵ ൌ ʹǡ ߜͶ ൌ ʹ

ߜ͵ ൌ ͷǡ ߜͶ ൌ ͷ

ܶ ൌ ͳͲͲ ߠ ൌ  െͲǤͺͲ

ߠ ൌ  െͲǤͶͲ

ߠ ൌ  െͲǤͲͲ

ߠ ൌ ͲǤͶͲ

ߠ ൌ ͲǤͺͲ

ܶ ൌ ʹͲͲ

ܶ ൌ ͳͲͲ

ܶ ൌ ʹͲͲ

ܶ ൌ ͳͲͲ

ܶ ൌ ʹͲͲ

Without

0.031

0.044

0.010

0.016

0.001

0.000

With

0.001

0.001

0.012

0.019

0.022

0.031

Total

0.032

0.045

0.022

0.035

0.023

0.031

Without

0.025

0.034

0.001

0.000

0.000

0.000

With

0.001

0.002

0.022

0.033

0.025

0.033

Total

0.026

0.036

0.023

0.033

0.025

0.033

Without

0.021

0.026

0.000

0.000

0.000

0.000

With

0.001

0.002

0.022

0.028

0.022

0.026

Without

0.022

0.024

0.000

0.000

0.000

0.000

With

0.001

0.002

0.025

0.025

0.027

0.026

Total

0.023

0.026

0.025

0.025

0.027

0.026

Without

0.072

0.056

0.000

0.000

0.000

0.000

With

0.001

0.002

0.064

0.073

0.096

0.092

Total

0.073

0.058

0.064

0.073

0.096

0.092

*Lag length fixed at one

The previous results (undersized or oversized results) may be due to the selection of the lag length, which has been fixed to unity. In order to observe if this issue is important, we present similar simulations to the previous Tables, but now the lag length is selected using the procedure tsig as suggested by Campbell and Perron (1991), considering a ݇ ‫ א‬ሾͲǡͷሿ. Table 5.4 presents results for ‫ܣܯܫܨܴܣ‬ሺͲǡ ݀ǡ Ͳሻ errors and for ݀ ൒ Ͳ. The message is that the ‫ ܨܦܣ‬test has an exact size close to the ͷΨ. In some cases, it presents a slightly smaller exact size.

Stationarity of the Inflation in Latin-American Countries Reviewed

99

Table 5-4: Size of the ADF Test: ARFIMA (0, d, 0) * ߜͳ ൌ Ͳǡ ߜʹ ൌ Ͳǡ

ߜͳ ൌ ͷǡ ߜʹ ൌ ͵ǡ

ߜͳ ൌ ͳͲǡ ߜʹ ൌ ͷǡ

ߜ͵ ൌ Ͳǡ ߜͶ ൌ Ͳ

ߜ͵ ൌ ʹǡ ߜͶ ൌ ʹ

ߜ͵ ൌ ͷǡ ߜͶ ൌ ͷ

ܶ ൌ ͳͲͲ ߠ ൌ ͲǤͲͲ

ߠ ൌ ͲǤͳʹ

ߠ ൌ ͲǤʹͶ

ߠ ൌ ͲǤ͵͸

ߠ ൌ ͲǤͶͺ

ܶ ൌ ʹͲͲ

ܶ ൌ ͳͲͲ

ܶ ൌ ʹͲͲ

ܶ ൌ ͳͲͲ

ܶ ൌ ʹͲͲ

Without

0.043

0.037

0.000

0.000

0.000

0.000

With

0.003

0.002

0.047

0.040

0.040

0.036

Total

0.046

0.039

0.047

0.040

0.040

0.036

Without

0.029

0.024

0.000

0.000

0.000

0.000

With

0.002

0.001

0.028

0.021

0.028

0.023

Total

0.031

0.025

0.028

0.021

0.028

0.023

Without

0.028

0.023

0.000

0.000

0.000

0.000

With

0.001

0.001

0.026

0.022

0.024

0.021

Total

0.029

0.024

0.026

0.022

0.024

0.021

Without

0.027

0.023

0.000

0.000

0.000

0.000

With

0.001

0.001

0.025

0.022

0.022

0.020

Total

0.028

0.024

0.025

0.022

0.022

0.020

Without

0.023

0.021

0.000

0.000

0.000

0.000

With

0.001

0.001

0.021

0.021

0.019

0.021

Total

0.024

0.022

0.021

0.021

0.019

0.021

*Lag length selected using the sequential t – sig method

Tables 5.5a–5.5c are similar to Tables 5.2a–5.2c, but select the lag length with the procedure t-sig. In all Tables, the ‫ ܨܦܣ‬test is oversized but clearly different or smaller compared to the case where ݇ ൌ ͳ. We found an oversized ‫ ܨܦܣ‬test only when ݀ ൌ ͲǤͲ. In other cases, when the memory is larger, the results are better, in particular for ܶ ൌ ʹͲͲ. .

Chapter Five

100

Table 5-5a: Size of the ADF Test: ARFIMA (0, d, 1) Errors with d=0.00*

ߠ ൌ  െͲǤͺͲ

ߠ ൌ  െͲǤͶͲ

ߠ ൌ ͲǤͲͲ

ߠ ൌ ͲǤͶͲ

ߠ ൌ ͲǤͺͲ

ߜͳ ൌ Ͳǡ ߜʹ ൌ Ͳǡ

ߜͳ ൌ ͷǡ ߜʹ ൌ ͵ǡ

ߜͳ ൌ ͳͲǡ ߜʹ ൌ ͷǡ

ߜ͵ ൌ Ͳǡ ߜͶ ൌ Ͳ

ߜ͵ ൌ ʹǡ ߜͶ ൌ ʹ

ߜ͵ ൌ ͷǡ ߜͶ ൌ ͷ

ܶ ൌ ͳͲͲ

ܶ ൌ ʹͲͲ

ܶ ൌ ͳͲͲ

ܶ ൌ ʹͲͲ

ܶ ൌ ͳͲͲ

ܶ ൌ ʹͲͲ

Without

0.356

0.282

0.130

0.087

0.000

0.000

With

0.018

0.012

0.302

0.233

0.396

0.285

Total

0.374

0.294

0.432

0.320

0.396

0.285

Without

0.072

0.057

0.005

0.003

0.000

0.000

With

0.005

0.003

0.078

0.056

0.072

0.053

Total

0.077

0.060

0.083

0.059

0.072

0.053

Without

0.043

0.037

0.000

0.000

0.000

0.000

With

0.003

0.002

0.047

0.040

0.040

0.036

Total

0.046

0.039

0.047

0.040

0.040

0.036

Without

0.051

0.039

0.000

0.000

0.000

0.000

With

0.002

0.001

0.044

0.039

0.045

0.039

Total

0.053

0.040

0.044

0.039

0.045

0.039

Without

0.048

0.041

0.000

0.000

0.000

0.000

With

0.001

0.001

0.050

0.040

0.038

0.039

Total

0.049

0.042

0.050

0.040

0.038

0.039

*Lag length selected using the sequential t – sig method

Stationarity of the Inflation in Latin-American Countries Reviewed

101

Table 5-5b: Size of the ADF Test: ARFIMA (0, d, 1) Errors with d=0.24*

ߠ ൌ  െͲǤͺͲ

ߠ ൌ  െͲǤͶͲ

ߠ ൌ ͲǤͲͲ

ߠ ൌ ͲǤͶͲ

ߠ ൌ ͲǤͺͲ

ߜͳ ൌ Ͳǡ ߜʹ ൌ Ͳǡ

ߜͳ ൌ ͷǡ ߜʹ ൌ ͵ǡ

ߜͳ ൌ ͳͲǡ ߜʹ ൌ ͷǡ

ߜ͵ ൌ Ͳǡ ߜͶ ൌ Ͳ

ߜ͵ ൌ ʹǡ ߜͶ ൌ ʹ

ߜ͵ ൌ ͷǡ ߜͶ ൌ ͷ

ܶ ൌ ͳͲͲ

ܶ ൌ ʹͲͲ

ܶ ൌ ͳͲͲ

ܶ ൌ ʹͲͲ

ܶ ൌ ͳͲͲ

ܶ ൌ ʹͲͲ

Without

0.115

0.057

0.019

0.008

0.000

0.000

With

0.007

0.003

0.112

0.054

0.114

0.054

Total

0.122

0.060

0.131

0.062

0.114

0.054

Without

0.029

0.023

0.000

0.000

0.000

0.000

With

0.001

0.001

0.029

0.022

0.026

0.022

Total

0.030

0.024

0.029

0.022

0.026

0.022

Without

0.028

0.023

0.000

0.000

0.000

0.000

With

0.001

0.001

0.026

0.022

0.024

0.021

Total

0.029

0.024

0.026

0.022

0.024

0.021

Without

0.030

0.023

0.000

0.000

0.000

0.000

With

0.001

0.000

0.028

0.020

0.026

0.018

Total

0.031

0.023

0.028

0.020

0.026

0.018

Without

0.033

0.024

0.000

0.000

0.000

0.000

With

0.001

0.001

0.027

0.022

0.029

0.021

Total

0.034

0.025

0.027

0.022

0.029

0.021

*Lag length selected using the sequential t – sig method

Chapter Five

102

Table 5-5c: Size of the ADF Test: ARFIMA (0, d, 1) Errors with d=0.48* ߜͳ ൌ Ͳǡ ߜʹ ൌ Ͳǡ

ߜͳ ൌ ͷǡ ߜʹ ൌ ͵ǡ

ߜͳ ൌ ͳͲǡ ߜʹ ൌ ͷǡ

ߜ͵ ൌ Ͳǡ ߜͶ ൌ Ͳ

ߜ͵ ൌ ʹǡ ߜͶ ൌ ʹ

ߜ͵ ൌ ͷǡ ߜͶ ൌ ͷ

ܶ ൌ ͳͲͲ ߠ ൌ  െͲǤͺͲ

ߠ ൌ  െͲǤͶͲ

ߠ ൌ ͲǤͲͲ

ߠ ൌ ͲǤͶͲ

ߠ ൌ ͲǤͺͲ

ܶ ൌ ʹͲͲ

ܶ ൌ ͳͲͲ

ܶ ൌ ʹͲͲ

ܶ ൌ ͳͲͲ

ܶ ൌ ʹͲͲ

Without

0.059

0.045

0.002

0.001

0.000

0.000

With

0.004

0.002

0.055

0.044

0.054

0.043

Total

0.063

0.047

0.057

0.045

0.054

0.043

Without

0.027

0.024

0.000

0.000

0.000

0.000

With

0.001

0.001

0.024

0.022

0.024

0.022

Total

0.028

0.025

0.024

0.022

0.024

0.022

Without

0.023

0.021

0.000

0.000

0.000

0.000

With

0.001

0.001

0.021

0.021

0.019

0.021

Total

0.024

0.022

0.021

0.021

0.019

0.021

Without

0.023

0.021

0.000

0.000

0.000

0.000

With

0.000

0.001

0.020

0.021

0.022

0.019

Total

0.023

0.022

0.020

0.021

0.022

0.019

Without

0.029

0.021

0.000

0.000

0.000

0.000

With

0.000

0.000

0.021

0.020

0.029

0.022

Total

0.029

0.021

0.021

0.020

0.029

0.022

*Lag length selected using the sequential t – sig method

Finally, Tables 5.6a–5.6c are similar to Tables 5.3a–5.3c, but use the procedure t-sig. The results indicate that the ADF test has a good size for almost every case. More clearly, the exact size is under the nominal size for ߩ ൏ Ͳ but is closer to ͷΨ when ߩ ൒ Ͳ. The conclusion suggests that using a data dependent rule to select the lag length fixes the problems detected before12. 12

It is worth to mention that we simulated data using another DGP. Let ሼ‫ݕ‬௧ ሽǡ ‫ א ݐ‬Ժ be a weakly stationary process. Let ሼ‫ݖ‬௧ ሽǡ ‫ א ݐ‬Ժ be a process contaminated by additive outliers, which is described by ‫ݖ‬௧ ൌ ‫ݕ‬௧ ൅ σ௠ (5) ௝ୀଵ ‫ݓ‬௝ ܺ௝ǡ௧ where ݉ is the maximum number of outliers and the unknown parameter ‫ݓ‬௝ indicates the magnitude of the ݆௧௛ outlier. The ܺ௝ǡ௧ is a random variable with probability distribution ܲ‫ݎ‬ሺܺ௝ ൌ െͳሻ ൌ ܲ‫ݎ‬ሺܺ௝ ൌ ͳሻ ൌ ‫݌‬௝ Ȁʹ and ܲ‫ݎ‬ሺܺ௝ ൌ Ͳሻ ൌ

Stationarity of the Inflation in Latin-American Countries Reviewed

103

Table 5-6a: Size of the ADF Test: ARFIMA (1, d, 0) Errors with d=0.00*

ߠ ൌ  െͲǤͺͲ

ߠ ൌ  െͲǤͶͲ

ߠ ൌ ͲǤͲͲ

ߠ ൌ ͲǤͶͲ

ߠ ൌ ͲǤͺͲ

ߜͳ ൌ Ͳǡ ߜʹ ൌ Ͳǡ

ߜͳ ൌ ͷǡ ߜʹ ൌ ͵ǡ

ߜͳ ൌ ͳͲǡ ߜʹ ൌ ͷǡ

ߜ͵ ൌ Ͳǡ ߜͶ ൌ Ͳ

ߜ͵ ൌ ʹǡ ߜͶ ൌ ʹ

ߜ͵ ൌ ͷǡ ߜͶ ൌ ͷ

ܶ ൌ ͳͲͲ

ܶ ൌ ʹͲͲ

ܶ ൌ ͳͲͲ

ܶ ൌ ʹͲͲ

ܶ ൌ ͳͲͲ

ܶ ൌ ʹͲͲ

Without

0.045

0.040

0.061

0.043

0.008

0.002

With

0.001

0.001

0.021

0.011

0.089

0.055

Total

0.045

0.041

0.082

0.054

0.087

0.057

Without

0.044

0.037

0.006

0.005

0.000

0.000

With

0.002

0.002

0.045

0.037

0.045

0.040

Total

0.047

0.039

0.051

0.042

0.045

0.040

Without

0.043

0.037

0.000

0.000

0.000

0.000

With

0.003

0.002

0.047

0.040

0.040

0.040

Total

0.046

0.039

0.047

0.040

0.040

0.040

Without

0.046

0.037

0.000

0.000

0.000

0.000

With

0.002

0.001

0.039

0.034

0.040

0.041

Total

0.048

0.038

0.039

0.034

0.040

0.041

Without

0.053

0.040

0.000

0.000

0.000

0.000

With

0.001

0.001

0.048

0.036

0.045

0.044

Total

0.054

0.041

0.048

0.036

0.045

0.044

*Lag length selected using the sequential t – sig method

ͳ െ ‫݌‬௝ . Therefore, ܺ௝ is the product of a Bernouilli (‫݌‬௝ ) and a Rademacher random variables; the latter equals ͳ or െͳ, both with probability ͳȀʹ. Furthermore, ‫ݕ‬௧ and ܺ௝ are independent random variables. The model (5) is based on the parametric models proposed by Fox (1972). This DGP is also used by Franses and Haldrup (1994), Fajardo et al. (2009), among others. In order to save space, results from this model are not included but they indicate very similar conclusions as in the previous DGP. Such Tables are available upon request.

Chapter Five

104

Table 5-6b: Size of the ADF Test: ARFIMA (1, d, 0) Errors with d=0.24* ߜͳ ൌ Ͳǡ ߜʹ ൌ Ͳǡ ߜ͵ ൌ Ͳǡ ߜͶ ൌ Ͳ

ߠ ൌ  െͲǤͺͲ

ߠ ൌ  െͲǤͶͲ

ߠ ൌ ͲǤͲͲ

ߠ ൌ ͲǤͶͲ

ߠ ൌ ͲǤͺͲ

ߜͳ ൌ ͷǡ ߜʹ ൌ ͵ǡ ߜ͵ ൌ ʹǡ ߜͶ ൌ ʹ

ߜͳ ൌ ͳͲǡ ߜʹ ൌ ͷǡ ߜ͵ ൌ ͷǡ ߜͶ ൌ ͷ

ܶ ൌ ͳͲͲ

ܶ ൌ ʹͲͲ

ܶ ൌ ͳͲͲ

ܶ ൌ ʹͲͲ

ܶ ൌ ͳͲͲ

ܶ ൌ ʹͲͲ

Without

0.026

0.023

0.011

0.013

0.000

0.000

With

0.001

0.000

0.012

0.010

0.021

0.024

Total

0.027

0.023

0.022

0.022

0.021

0.024

Without

0.026

0.022

0.000

0.000

0.000

0.000

With

0.001

0.002

0.026

0.022

0.025

0.024

Total

0.027

0.023

0.026

0.022

0.025

0.024

Without

0.028

0.023

0.000

0.000

0.000

0.000

With

0.001

0.001

0.026

0.022

0.024

0.022

Total

0.029

0.024

0.026

0.022

0.024

0.022

Without

0.030

0.022

0.000

0.000

0.000

0.000

With

0.001

0.001

0.029

0.020

0.028

0.024

Total

0.031

0.023

0.029

0.020

0.028

0.024

Without

0.057

0.037

0.000

0.000

0.000

0.000

With

0.001

0.001

0.048

0.040

0.054

0.044

Total

0.058

0.038

0.048

0.040

0.054

0.044

*Lag length selected using the sequential t – sig method

Stationarity of the Inflation in Latin-American Countries Reviewed

105

Table 5-6c: Size of the ADF Test: ARFIMA (1, d, 0) Errors with d=0.48* ߜͳ ൌ Ͳǡ ߜʹ ൌ Ͳǡ ߜ͵ ൌ Ͳǡ ߜͶ ൌ Ͳ

ߠ ൌ  െͲǤͺͲ

ߠ ൌ  െͲǤͶͲ

ߠ ൌ ͲǤͲͲ

ߠ ൌ ͲǤͶͲ

ߠ ൌ ͲǤͺͲ

ߜͳ ൌ ͷǡ ߜʹ ൌ ͵ǡ ߜ͵ ൌ ʹǡ ߜͶ ൌ ʹ

ߜͳ ൌ ͳͲǡ ߜʹ ൌ ͷǡ ߜ͵ ൌ ͷǡ ߜͶ ൌ ͷ

ܶ ൌ ͳͲͲ

ܶ ൌ ʹͲͲ

ܶ ൌ ͳͲͲ

ܶ ൌ ʹͲͲ

ܶ ൌ ͳͲͲ

ܶ ൌ ʹͲͲ

Without

0.024

0.022

0.006

0.007

0.000

0.000

With

0.001

0.000

0.015

0.014

0.021

0.021

Total

0.025

0.022

0.021

0.021

0.021

0.021

Without

0.023

0.021

0.000

0.000

0.000

0.000

With

0.001

0.001

0.023

0.020

0.020

0.019

Total

0.024

0.022

0.023

0.020

0.020

0.019

Without

0.023

0.021

0.000

0.000

0.000

0.000

With

0.001

0.001

0.021

0.021

0.022

0.020

Total

0.024

0.022

0.021

0.021

0.022

0.020

Without

0.025

0.021

0.000

0.000

0.000

0.000

With

0.000

0.001

0.023

0.021

0.026

0.021

Total

0.026

0.022

0.023

0.021

0.026

0.021

Without

0.061

0.032

0.000

0.000

0.000

0.000

With

0.001

0.000

0.055

0.041

0.065

0.055

Total

0.062

0.032

0.055

0.041

0.065

0.055

*Lag length selected using the sequential t – sig method

5.4. Empirical Application The Latin-American inflation series offer a good example of the strong presence of large sized additive outliers in a possible non-stationary time series. Figure 5.1 shows quarterly inflation series for eight Latin-American countries: Argentina, Bolivia, Chile, Colombia, Ecuador, Peru, Uruguay and Venezuela. The frequency is quarterly and the sample spans from 1970:1 until 2010:4. Many or all of these countries have experimented with different stabilization programs to stop high inflation episodes. Intervention of this kind, in most of these cases, has introduced additive outliers in the evolution of their inflation series.

Chapter Five

106

Figure 5-1: Quarterly Latin-American Inflation Series Argentina

Bolivia

800

500 400

600

300 400 200 200 100 0

0

-200

-100 1975

1980

1985

1990

1995

2000

2005

2010

1975

1980

1985

1990

1995

2000

2005

2010

1995

2000

2005

2010

1995

2000

2005

2010

Chile Colom bia

160 16 120 12 80 8 40 4 0 0 -40 1975

1980

1985

1990

1995

2000

2005

2010

-4 1975

1980

1985

1990

Peru

Ecuador 800

30 25

600 20 400

15 10

200

5 0 0 -200

-5 1975

1980

1985

1990

1995

2000

2005

1975

2010

1980

1985

1990

Venezuela

Uruguay 40

28 24

30 20 16

20

12 8

10

4 0

0 1975

1980

1985

1990

1995

2000

2005

2010

1975

1980

1985

1990

1995

2000

2005

2010

Stationarity of the Inflation in Latin-American Countries Reviewed

107

The periods of high inflation in Argentina and Peru happened between 1985 and 1990, when the most important stabilization programs were applied. In the case of Argentina, the most well-known governmental plans were the Austral Program (June 1985), the program of February 1987, the Austral II Program (October 1987), the Spring Program (August 1988), the BB Program (1989), the Bonex Program (January 1990) and the Cavallo’s Program (March 1991), where the dates in parentheses correspond to the start date of the programs. In the Peruvian case, we can mention two principal stabilization programs. These are the Salinas’ Program (September 1988) and the Fujimori’s Program (JulyAugust 1990). In the Bolivian case, the episode of high inflation was in the middle of the 1980s. Many small stabilization programs were applied during the period between 1982 and 1984, but it was the program applied in August 1985 which stopped the high inflation. High inflation in Chile began around 1975. Diverse programs were applied between 1975 and 1977 until the shock plan applied at the end of 1977-1979. A piece of research related to this note is Rodríguez (2004), where four LatinAmerican countries were analyzed. In this study, we add more countries and more observations. For more details related to the inflationary process in some of these countries, see Rodríguez (2004). One important issue from Figure 5.1 is the following. Observing the vertical axis, four countries (Argentina, Bolivia, Chile and Peru) show huge additive outliers. The other countries also show the presence of additive outliers, but their magnitudes are very different (smaller) in comparison with the four countries mentioned above. It implies that procedure ߬ௗ will identify more additive outliers in the former countries and fewer additive outliers in all the other countries. Table 5.7 shows results from the application of standard and new unit root tests. We apply two standard unit root statistics: the Phillips and Perron, and the Augmented Dickey and Fuller statistics (see Phillips and Perron (1988), and Said and Dickey (1984), respectively). Other tests are the M-tests based on GLS detrending data as suggested by Ng and Perron (2001). In all cases, lag length has been selected using the dependent recursive rule named t-sig as proposed by Campbell and Perron (1991)13. Almost in all cases, all statistics suggest a rejection of the null hypothesis of a unit root. This result is particularly clear for the Phillips and Perron (1988) test where the rejection is strong. It is not surprising if we 13

We also use the MAIC approach as suggested by Ng and Perron (2001). Results are very similar and conclusions are not modified. We present the results using the t-sig method to be coherent (in terms of comparison) with the method used in simulations.

Chapter Five

108

remember that this statistic is very oversized when there are strong negative moving average correlations. Given the evidence that this type of correlation implies the presence of additive outliers (see Franses and Haldrup (1994), and Vogelsang (1999)), the results of the PP test are not rare. The results are clearer in countries like Peru where the size of the additive outliers are huge. Even the robust M-tests proposed by Stock (1999) indicate a rejection of the null hypothesis of a unit root, except for the cases of Uruguay and Venezuela. Table 5-7: Standard and New Unit root Tests Phillips-Perron Value Argentina Bolivia Chile Colombia Ecuador

a

-9.475

a

-6.594

k 1 3

Value a

-4.976

a

-4.615

-2.812

b

12

-2.641

-3.125

b

8

-0.683

-2.702

c

‫ܵܮܩߙܼܯ‬

ADF

c

k 1 3

Value a

-29.142

a

-54.306

b

‫ܵܮܩݐܼܯ‬

‫ܵܮܩܤܵܯ‬

ܲܶ‫ܵܮܩ‬

Value

Value

Value

k

0.131

a

a

1

a

3

b

12

a

-3.817

a

0.096

a

b

-2.378

0.206

a

-5.211

0.841 0.451

12

-11.509

8

-1.720

-0.822

0.477

12.672

8

8

c

b

c

c

8

8

-2.243

Peru

a

-9.380

5

-2.769

c

-7.339

5

-10.101

Uruguay

-2.404

8

-1.450

8

Venezuela

-3.186b

5

-2.199

5

b

-1.911

0.260

2.211

3.354

b

-2.247

0.222

b

b

2.426

5

-3.452

-1.292

0.374

7.092

8

-4.089

-1.414

0.345

6.011

5

Lag length selected using the recursive method t-sig; a,b,c indicate statistically significancy at 1.0%, 5.0%, and 10.0% respectively

Table 5-8 shows the results of the ADF statistic corrected for the presence of the additive outliers. The results indicate a non-rejection of the null hypothesis of a unit root for all countries, implying that LatinAmerican inflation series are non-stationary. Rodríguez (2004) found a similar result but only for Argentina and Peru using a smaller sample size. Charemza et al. (2005) also find results in favour of non-stationarity using a large set of inflation series, when the innovations are treated as draws from a symmetric stable Paretian distribution with infinite variance. However, more recently, Caporale and Paxton (2013), using a different unit root strategy, have arrived at opposite results for a pool of five LatinAmerican countries and a shorter period sample. In this case, when structural breaks are considered, inflation series are found to be

Stationarity of the Inflation in Latin-American Countries Reviewed

109

stationary14. This suggests that an appropriate treatment of extreme values is important in this context to avoid divergent results. Table 5-8: ADF Test corrected for Additive Outliers using ࣎ࢊ Country

Value

Coefficient

k

Outliers

Argentina

-1.723

0.880

7

3

Bolivia

-0.131

0.977

13

16

Chile

-2.353

0.865

12

14

Colombia

-0.329

0.986

8

3

Ecuador

-0.899

0.949

12

2

Peru

1.423

1.098

19

19

Uruguay

-1.378

0.958

10

3

Venezuela

-1.469

0.904

7

5 a,b,c

Lag length selected using the recursive method t-sig; indicate statistically significancy at 1.0%, 5.0%, and 10.0% respectively

5.5. Conclusions This chapter analyses the empirical size of the ‫ ܨܦܣ‬statistic when there are additive outliers and ‫ܣܯܫܨܴܣ‬ሺ‫݌‬ǡ ݀ǡ ‫ݍ‬ሻ errors. Results indicate that a few cases imply oversized ‫ ܨܦܣ‬tests. In most cases, the statistic is slightly undersized or very close to the nominal size of ͷΨ. There are some difficulties when ߠ goes to െͳ or when ߩ goes to ȁͳȁ. An empirical application for eight Latin-American countries indicates the difficulties that standard and new unit root tests have in verifying if there is or is no a unit root in the inflation time series. When hyperinflationary episodes are ignored, the analysis is incomplete and this leads to the wrong conclusions. An application of an ADF test with dummies associated with the location of the identified additive outliers confirms that all inflation time series are non-stationary. It is a similar 14 A similar approach is applied by Ozcan (2013) to examine the stationarity of inflation rates in 11 Mediterranean countries.

110

Chapter Five

result to that obtained by Rodríguez (2004) but using a larger sample size and more countries. Results also are consistent with those found by Charemza et al. (2005), where the results are in favour of the nonstationarity of a large set of inflation series when the innovations are treated as draws from a symmetric stable Paretian distribution with infinite variance. It is relevant to say that an appropriate treatment of extreme values is important. Our results suggest important policy implications. Shocks to inflation in Latin-American countries have long-lasting effects on the inflation rates. Therefore, this makes it more difficult for monetary authorities to forecast future movements in the inflation rates based on its past history, which may be interpreted as further evidence in favour of the Lucas critique about the instability of parameters.

References Arize, A. C. (2011). Are Inflation Rates Really Non-stationary? New Evidence from Non-linear STAR Framework and African Data. International Journal of Economics and Finance 3(3), 97-108. Baldé, T., and G. Rodríguez (2005). Finite Sample Effects of Additive Outliers on the Granger-Causality Test with an Application to Money Growth and Inflation in Peru. Applied Economics Letters 15, 841-844. Batini, N. (2006). Euro Area Inflation Persistence. Empirical Economics 31(4), 977-1002. Burridge, P., and A. M. R. Taylor (2006). Additive Outlier Detection via Extreme-Value Theory. Journal of Time Series Analysis 27(5), 685701. Campbell, J. Y., and P. Perron (1991). Pitfalls and Opportunities: What Macroeconomists Should Know About Unit Roots. In O. J. Blanchard and S. Fisher (eds.), NBER Macroeconomic Annual, Vol. 6, 141-201. Caporale, T., and J. Paxton (2013). Inflation Stationarity During Latin American Inflation: Insights from Unit Root and Structural Break Analysis. Applied Economics 45, 2001-2010. Cati, R., M. Garcia, and P. Perron (1999). Unit Roots in the Presence of Abrupt Governmental Interventions with an Application to Brazilian Data. Journal of Applied Econometrics 14, 27-56. Chan, W. (1992). A Note on Time Series Model Specification in the Presence of Outliers. Journal of Applied Statistics 19, 117-124. —. (1995). Outliers and Financial Time Series Modelling: A Cautionary Note. Mathematics and Computers in Simulation 39, 425-430.

Stationarity of the Inflation in Latin-American Countries Reviewed

111

—. (1998). Outlier Analysis of Annual Retail Price Inflation: A CrossCountry Study. Journal of Actuarial Practice 6, 149-172. Chareka, P., F. Matarise, and R. Turner (2006). A Test for Additive Outlier Applicable to Long-Memory Time Series. Journal of Economic Dynamics & Control 30, 595-621. Charemza, W. W., D. Hristova, and P. Burridge (2005). Is Inflation Stationary? Applied Economics 37(8), 901-903. Chang, I., G. C, Tiao, and C. Chen (1988). Estimation of Time Series Parameters in the Presence of Outliers. Technometrics 30, 193-204. Chen, C., and L. Liu (1993). Joint Estimation of Model Parameters and Outlier Effects in Time Series. Journal of the American Statistical Association 74, 427-431. Chiquiar, D., A. E. Noriega, and A. Ramos-Francia (2010). A Time Series Approach to Test a Change in Inflation Persistence: The Mexican Experience. Applied Economics 42(24), 3067-3075. Elliott, G., T. Rothenberg, and J. H. Stock (1996). Efficient Tests for an Autoregressive Unit Root. Econometrica 64, 813-836. Fajardo Molinares, F., V. A. Reisen, and F. Cribari-Neto (2009). Robust estimation in long-memory processes under additive outliers. Journal of Statistical Planning and Inference 139, 2511-2525. Fox, A. J. (1972). Outliers in Time Series. Journal of the Royal Statistical Association 74, 140-146. Franses, P. H., and N. Haldrup (1994). The Effects of Additive Outliers on Tests for Unit Roots and Cointegration. Journal of Business & Economic Statistics 12, 471-478. Fuhrer, J. C. (2009). Inflation Persistence. Federal Reserve Bank of Boston, WP 09-14. Gadea, M. D., and L. Mayoral (2006). The Persistence of Inflation in OECD Countries: A Fractionally Integrated Approach. MPRA Paper No. 815, posted 14. November 2006 Gómez, V., and A. Maravall (1992). Time Series Regression with ARIMA Noise and Missing Observations. Program TRAMO. European University Institute, Working Paper ECO 92/81. Henry, O. T., and K. Shields (2004). Is There a Unit Root in Inflation? Journal of Macroeconomics 26, 481-500 Hoeck, H., A. Lucas, and H. K. vand Dijk (1995). Classical and Bayesian Aspects of Robust Unit Root Inference. Journal of Econometrics 69, 25-79. Kumar, M. S., and T. Okimoto (2007). Dynamics of Persistence in International Inflation Rates. Jounal of Money, Credit and Banking 39(6), 1457-1479.

112

Chapter Five

Levin, A. T., and Piger (2004). Is Inflation Persistence Intrinsic in Industrial Economies? European Central Bank, Working Paper Series 334. Lucas, A. (1995a). An Outlier Robust Unit Root Test with an Application to the Extended Nelson-Plosser Data. Journal of Econometrics 66, 153-173. —. (1995b). Unit Root Tests Based on M Estimators. Econometric Theory 11, 331-346. Lucas, R. E. (1976). Econometric Policy Evaluation: A Critique. Carnegie-Rochester Conference Series on Public Policy 1, 19-46. Nelson, C. R., and C. I. Plosser (1982). Trends and Random Walks in Macroeconomic Time Series. Journal of Monetary Economics 10, 139162. Ng, S., and P. Perron (1995). Unit Root Tests in ARMA Models with Data-Dependent Methods for the Selection of the Truncation Lag. Journal of the American Statistical Association 90, 268-281 Ng, S., and P. Perron (2001). Lag Length Selection and the Construction of Unit Root Tests with Good Size and Power. Econometrica 69, 15191554. Ozcan, B. (2013). Are Inflation Rates Stationary in 11 Mediterranean Countries? Evidence from Univariate and Panel Unit Root Tests. Eurasian Journal of Business and Economics 6 (12), 79-96. Oliveira, F. N., and M. Petrassi (2014). Is Inflation Persistance Over? Revista Brasileira de Economia 68(3), 393-422. O'Reilly, G., and K. Whelan (2005). Has Euro-Area Inflation Pesistence Changed Over Time? The Review of Economics and Statistics 87(4), 709-720. Perron, P., and S. Ng (1996). Useful Modifications to Unit Root Tests with Dependent Errors and their Local Asymptotic Properties Review of Economic Studies 63, 435-463. Perron, P., and G. Rodríguez (2003). Searching for Additive Outliers in Nonstationary Time Series. Journal of Time Series Analysis 24(2), 193-220. Phillips, P. C. B., and P. Perron (1988). Testing for a Unit Root in Time Series Regression. Biometrika 75, 335-346. Piveta, F., and R. Reis (2007). The Persistence of Inflation in the United States. Journal of Economic Dynamics & Control 31, 1326–1358. Roache, S. K. (2013). Inflation Persistence in Brazil – A Cross Country Comparison. IMF Working Paper, WP14/55.

Stationarity of the Inflation in Latin-American Countries Reviewed

113

Rodríguez, G. (2004). An Empirical Note about Additive Outliers and Nonstationarity in Latin-American Inflation Series. Empirical Economics 29, 361-372. Romero-Ávila, D., and C. Usabiaga (2009). The Hypothesis of a Unit Root in OECD Inflation Revisited. Journal of Economics and Business 61, 153-161. Said, S.E., and D.A. Dickey (1984). Testing for Unit Roots in Autoregressive-Moving Average Models of Unknown Order. Biometrika 71, 599-608. Shin, D. W., S. Sharkar, and J. H. Lee (1996). Unit Root Tests for Time Series with Outliers. Statistics and Probability Letters 30, 189-197. Stock, J. H. (1999). A Class of Test for Integration and Cointegration. In Engle, R. F., and H. White (eds.), Cointegration, Causality and Forecasting. A Festschrift in Honour of Clive W. J. Granger, Oxford University Press, 137-141. Thornton, J. (2008). Inflation and Inflation Uncertainty in Argentina: 1810-2005. Economics Letters 98, 247-252. Tsay, R. S. (1986). Time Series Model Specification in the Presence of Outliers. Journal of the American Statistical Association 81, 132-141. Vogelsang, T. J. (1999). Two Simple Procedures for Testing for a Unit Root when there are Additive Outliers. Journal of Time Series Analysis 20, 237-252.

PART III: INTERNATIONAL TRADE

CHAPTER SIX THE EFFECTIVENESS OF DEVELOPMENT AID IN PROMOTING ECONOMIC GROWTH LAURA GÓMEZ WIGLEY*

Official Development Assistance (ODA) flows have considerably increased in the last decades. This article studies the impact of ODA on economic growth and the characteristics of recipient countries and donors’ management practices which influence its effects. In order to achieve this, a model of the impact of aid is developed based on the new growth theories and studies, taking into account the endogeneity of aid and other variables and time and individual fixed effects. It is estimated with the system Generalized Method of Moments with data from the period 19952010 corresponding to 27 African countries among the least developed ODA recipients. Results confirm the effectiveness of foreign aid in promoting economic growth, though it presents diminishing marginal returns. Furthermore, its effectiveness decreases if the country has been recently involved in armed conflicts, which probably reflects the lower quality of institutions in such situations. On the other hand, aid fragmentation is found to have a positive effect on growth. In conclusion, aid is proved to be effective in encouraging economic growth, though there has to be further consideration of aid heterogeneity and its impact on other aspects of development.

*

London School of Economics and Political Science

118

Chapter Six

6.1. Introduction The main objective in administering Official Development Assistance (ODA)1 is to promote the economic development and welfare of developing countries. Although economic growth and development are different, albeit related, concepts, in this article we will focus on economic growth as an essential facet of economic development, ignoring other aspects. Therefore, the existence of ODA is based on the premise that it will have a positive effect on the growth of the receiving country. Generally it is thought that this positive effect can take place through two channels: supplementing the domestic capital in the receiving country or reducing its fiscal burden. A review of existing literature shows that not only there is no consensus on these channels, there is not even unanimity on the positive impact of aid. Juselius et al. (2001) distinguish between three main groups of opinion on this subject. Some consider aid to be ineffective; others that it is partially effective (but not perfect) or that its efficacy is conditioned by specific circumstances in the receiving country and, finally, there are those who defend aid as effective or as a moral obligation of rich countries, who should contribute to eradicate poverty. Despite this lack of consensus, aid flows have acquired considerable dimensions ever since the end of the Second World War and especially since the 60s. Solely in the period we analyse, from 1995 to 2010, the African countries considered have received annually on average more than 13% of their gross domestic product in purchasing power parity (GDP PPP). Given the magnitude of these flows, it seems relevant to pose the question of the effectiveness of ODA in promoting economic growth. The aim of this article is to contrast whether Official Development Assistance has a positive effect on growth and if so, what the exact nature of that relationship is (in particular, if it has diminishing marginal returns). In addition, we want to analyse which characteristics of receiving countries and donors’ management practices influence this relationship. The rest of the variables we will use are control variables. A growth model derived from Barro’s model (1991) and used, among others, by Tezanos et al. (2009) is developed. The estimation is carried out with the system Generalized Method of Moments (GMM) with data from 27 African countries among those with the lowest income per capita corresponding to 1

ODA includes all flows to countries and territories on the Development Assistance Committee (DAC) List of ODA Recipients which are concessional in character with a grant element of at least 25 per cent and provided by official agencies, such as multilateral institutions or state and local governments.

The Effectiveness of Development Aid in Promoting Economic Growth 119

the period 1995-2010. The results indicate that aid is effective in the countries and period considered, that it presents diminishing marginal returns, that its effectiveness decreases after armed conflicts and that a greater fragmentation of these flows favours growth. Therefore, the idea of the effectiveness of ODA is reinforced, though there still is much ground to cover concerning the heterogeneity of aid and its impact on other aspects of development, as well as taking into account the limitations faced when collecting reliable and extensive data. The rest of this chapter is structured as follows: Section 6.2 presents the theoretical basis, Section 6.3 covers the methodology used, Section 6.4 analyses the results of the study and, finally, the last section summarizes the main conclusions.

6.2. Theoretical Background and Current Situation Since the sixties, the effectiveness of aid in promoting economic growth has generated great interest among academics and policy-makers. According to Hansen and Tarp (2000), aid efficacy has been analysed at a micro and macro level and studies have been carried out focusing both on one country as well as comparing among countries. Some argue that a so called micro-macro paradox exists, stated by Mosley (1987), for they consider that aid effectiveness has been proven at a micro level, but not so at a macro level. Nevertheless, a review of literature proves that the majority of empirical studies find a positive relationship between aid and growth at a macroeconomic level, rather than a negative one. In any case, results differ greatly between studies which use the same or similar databases depending on the methodology applied. In the last decades, we can find three distinct generations of comparative studies among countries concerning aid effectiveness, as stated by Hansen and Tarp (2000). The first ones were based on Harrod-Domar growth models and the two breach model of Chenery-Strout-Bruno. Aid was perceived as an exogenous increase in the stock of capital of the receiving country and the causal relationship was thought to be aid-savingsinvestment-growth. The results in general showed a positive effect of aid on growth through an increase in savings. In the second generation, focus shifted to the nature of the relationship between aid and growth. The underlying analytical framework was consistent with the Harrod-Domar growth model or with a simple version of the neoclassical Solow-Swan model and some authors chose to use reduced-form equations. Investment was considered the main determinant of growth and, in general, the results of these studies were congruent with

120

Chapter Six

a positive link between aid and investment. Regarding papers which used reduced-form equations, regressions suggested a positive relationship between aid and growth when there was a positive link between savings and growth, therefore being congruent with the aforementioned causality relationship: from aid to savings to investment to growth. Finally, in the third generation of studies there has been a shift in focus as studies have started working with panel data and a large number of countries due to a greater availability of statistical information. Additionally, the analytical basis has changed to include the new growth theory which has been developed. Other variables are taken into account, such as institutions and human capital and endogenous growth models have been developed to treat the possible endogeneity of aid and other variables. Furthermore, studies are considering the possibility that the relationship between aid and growth is non-linear, for reasons such as the Dutch syndrome analysed by Rajan and Subramanian (2005) and the possible existence of limits to the absorptive capacity of aid in recipient countries, as explained by Hansen and Tarp (2001), who found diminishing returns of aid. Some of these changes were introduced in previous papers, but it has been in this generation when they have acquired a greater relevance. The results of these studies also find in general a positive link between aid and growth, though there is still much debate since it is obvious that aid is not equally effective in all cases. Among the studies which conform this third generation, we must highlight those which consider that there are certain characteristics of donors and of recipient economies which condition aid effectiveness. The study of Burnside and Dollar (2004) is especially remarkable; they reached conclusions which, although criticised and contested, raised the possibility that there are essential conditions or prerequisites for aid to be effective, specifically, the existence of sound economic policies in the recipient country. In our model we introduce variables which condition the impact of aid, such as the existence of an effective rule of law and control of corruption, based on the aforementioned study of Burnside and Dollar (2004), as well as on the results of Chauvet and Guillaumont (2004). This last work also leads us to include the existence of commercial shocks as a proxy for the possible presence of unfavourable effects on growth, when aid will supposedly be more effective. Also, based on this one and other studies, like Collier and Hoeffler (2004), we introduce the interaction between aid and a variable which reflects whether the country has been recently involved in armed conflicts, because in those situations the efficacy of aid can change. Finally, a variable is included to indicate whether it is a landlocked country as a proxy for structural disadvantages

The Effectiveness of Development Aid in Promoting Economic Growth 121

the country may suffer. Generally authors use the proportion of territory between the tropics to consider geographical disadvantages, but in this context that variable would not make sense. On the other hand, we consider other variables regarding aid flows and donors’ management practices which may affect growth. It is the case of aid fragmentation and volatility, as Bulir and Hamann (2008) and Djankov et al. (2009) state respectively. If aid is pro-cyclical and volatile, economic cycles will be amplified and the uncertainty about the amount of aid a country will receive will distort the fiscal behaviour of recipient governments and the investment decisions in the country. In addition, it generates fluctuations in the exchange rate, generally appreciating the local currency, therefore decreasing its competitiveness in international markets. Regarding fragmentation, it is though that if aid comes from a large number of donors and there is a lack of coordination among them, aid effectiveness will decrease, given that the time and money the recipient has to dedicate to deal with donors will increase and there will be less control over the use of the aid. Finally, Quiñones and Tezanos (2012) talk about an emerging fourth generation which takes into account aid heterogeneity when considering its effects. For example, Quattara and Strobl (2008) distinguish between technical aid, food aid, etc. when analysing the effectiveness of aid. After this brief review, one can see that there is an extended literature on aid effectiveness at a macroeconomic level. The majority of articles find a positive relationship between aid and growth (with or without conditioning factors) or they do not obtain conclusive results, though we must not forget that lack of proof of an effect is not the same as proof of a lack of effect. Few studies conclude that aid is ineffective and even these can be called into question because of their methodology. The main issue lies in the underlying growth model used when studying aid effectiveness. Without consensus regarding this theoretical model, it is difficult to reach conclusive results. In any case, in the following section we build a model to study the impact of aid basing ourselves on growth theory and some of the aforementioned papers.

6.3. Methodology 6.3.1. Model: The Impact of ODA on the Growth Rate We develop a model to study the impact of aid on the growth rate and its possible interaction with other variables. Therefore, the goal is not to formulate a growth model, but we must include all the variables we

122

Chapter Six

consider relevant according to growth theory and important papers on this subject in order for the model to have a sufficient explanatory capability and avoid biased results. Just like Tezanos et al. (2009), we choose a Barro type model with certain modifications. The theory indicates which variables should be included in the model, although the availability of statistical data conditions which ones can be effectively introduced. Now we will formulate the model: ‫ܩ‬௜௧ ൌ ߙ௜ ൅ ߬௧ ൅ ߚ Ž൫‫ݕ‬௜௧బ ൯ ൅ ߜଵ ‫ܣ‬௜௧ ൅ ߜଶ ‫ܣ‬ଶ௜௧ ൅ ࢽĄࡾ࢏࢚ ‫ܣ‬௜௧ ൅ ࣂĄࡰ࢏࢚ ൅ ࣐Ą࡯࢏࢚ ൅ ‫ݑ‬௜௧ (1) Where ߙ௜ and ߬௧ represent the fixed effects associated with each individual (country) and each time period respectively, ࢽĄ=(ߛଵ ,ߛଶ ,ߛଷ ,ߛସ ), ࣂĄ=(ߠଵ ,ߠଶ ) and ࣐Ą=(߮ଵ ,߮ଶ ,߮ଷ ,߮ସ ,߮ହ ,߮଺ ,߮଻ ,଼߮ ,߮ଽ ,߮ଵ଴ ,߮ଵଵ ) are vectors of parameters and ‫ݑ‬௜௧ is the disturbance term of the model, which we will assume behaves like a white noise. The model variables are described below, though you can find more information and the sources of the data in Table 4.1. For each country i and time period t: - ‫ܩ‬௜௧ is the annual rate of variation of GDP per capita PPP (GDPpc PPP). - ‫ݕ‬௜௧బ is the initial level of GDPpc PPP. The parameter ߚ reflects the possibility of conditional convergence among countries, as long as it is negative. - ‫ܣ‬௜௧ is the annual average ODA country i receives in period t (as a percentage of its GDP PPP). The parameter ߜଵ shows the relationship aid-growth and ‫ܣ‬ଶ௜௧ is included to account for the possibility of that relationship not being linear. Therefore, the parameter ߜଶ reflects the possible existence of diminishing marginal returns (if it is negative). - ࡾ࢏࢚ Ą=(ܴ‫ܮ݈݁ݑ‬௜௧ ǡ ‫݀݊ܽܮ‬௜ ǡ ‫݄ܵܿܧ‬௜௧ ǡ ‫݂݊݋ܥ‬௜௧ ) is a vector of variables which represents characteristics of aid recipients that can condition the effects of aid. Thus, ࢽ reflects the possible mechanisms of impact that ODA can have over economic growth. The variables which make up the vector are: rule of law and control of corruption (ܴ‫)ܮ݈݁ݑ‬, economic shocks (‫)݄ܵܿܧ‬, landlocked (‫ )݀݊ܽܮ‬and post-conflict scenarios (‫)݂݊݋ܥ‬. - ࡰ࢏࢚ Ą=(ܸ‫݈݋‬௜௧ ǡ ‫݃ܽݎܨ‬௜௧ ) is a vector of variables which define characteristics of aid flows and donors’ management practices which also influence the impact of aid. It is expressed as a function of aid fragmentation (‫ )݃ܽݎܨ‬and volatility (ܸ‫݈݋‬ሻ. The vector of parameters ࣂ captures the effects of these variables on economic growth. - ࡯࢏࢚ Ą=(‫ܿݑ݀ܧ‬௜௧బ ǡ Žሺ‫݌ݔܧ‬௜௧బ ሻǡ ‫݀݊ܽܮ‬௜ ǡ ܴܰ௜௧ ǡ ‫݄ܵܿܧ‬௜௧ ǡ ܵ‫ܮ݁ݐܽݐ‬௜௧ ǡ ‫ܿܽݎܨ‬௜ ǡ ‫݈݂݊ܫ‬௜௧ ǡ ‫ݒ݋ܩ‬௜௧ ǡ ܱ‫݊݁݌‬௜௧ ǡ ܲ‫ܾ݋‬௜௧ ሻ (2)

The Effectiveness of Development Aid in Promoting Economic Growth 123

is a vector which includes other variables relevant to explaining growth and, therefore, ࣐ measures the effects of these variables on growth. There are many differences in growth theory and economic policies regarding these variables. For this model, we have chosen among those which have been consistently used in previous papers, such as Barro (1991, 1996), Sachs and Warner (1997), Sala-i-Martin (1997), Collier and Gunning (1999) y Rajan and Subramanian (2005). It is not an exhaustive list, but unless one is willing to do four million regressions, the number of regressors has to be limited. Human capital is approximated by the initial level of education (‫ܿݑ݀ܧ‬௧బ ) and the initial health status (‫݌ݔܧ‬௧బ ), for human capital influences the productive and innovative capacities of the country and, therefore, its growth. The country’s geography is approximated by being landlocked or not (‫ )݀݊ܽܮ‬and the endowment of natural resources (ܴܰ). Countries which are landlocked experience more difficulties to participate in international trade since they have much higher transport costs due to their lack of direct access to sea routes. Regarding the impact of having a large endowment of natural resources, there is controversy as to whether it is negative or positive because different papers reach different results depending on the control variables, the sample of countries and the methodology used. The variable of economic shocks (‫ )݄ܵܿܧ‬is generally thought to be important for these countries since their exports are mainly based on a few primary products and they are very dependent on them. Shocks are measured through changes in the terms of trade. The demographic variable (ܲ‫ )ܾ݋‬measures the difference between the growth of the economically active population and that of total population, in order to control the effect of changes in the composition of the population on the growth rate. If it is positive, then the working-age population who provides labour (whether they are employed or unemployed) is growing and in theory this will cause an increase in the GDP per capita. The variable of rule of law and control of corruption (ܴ‫ )ܮ݈݁ݑ‬is used to introduce the effect of the quality of some of the country’s institutions, which can influence its potential to grow and the distribution of resources among agents and generations. We also include another variable, the index of ethnic fractionalization (‫ )ܿܽݎܨ‬as a proxy for social stability in the country. This index remains more or less constant across time. The economic environment and macroeconomic policies are introduced in the model through the inflation rate to take into account macroeconomic instability and the government’s ability and capacity to

124

Chapter Six

control the economy, as stated by Fischer (1993). We also consider government final consumption expenditure as a proxy for the size of the public sector and its intervention in the economy, as well as the openness to trade of the country (measured as the share of the sum of exports and imports to GDP) In any case, there is no consensus among academics about what variables should be included and they differ greatly from one paper to another one, so discussion remains open over which variables we should include.

6.3.2. Econometric Estimation Procedure In order to determine the estimation method, we must take into account the model, variables and database we have to work with. In this case, panel data allows us to take advantage of both transversal and temporal variability, so the model must reflect this. Aside from this issue, the main obstacle to analyse this data lies in the lack of strict exogeneity of some of the variables; that is to say, they are correlated with past and/or present values of the error term (they are predetermined or endogenous variables). If the method of ordinary least squares were used, we would obtain inconsistent and biased estimators. The variables are classified as stated below: Endogenous. ‫ܣ‬, interaction variables, ‫݈݂݊ܫ‬, ‫݄ܵܿܧ‬, ܱ‫݊݁݌‬, ‫ݒ݋ܩ‬. If aid is given guided by altruistic motives, aid allocation is negatively correlated to the growth rates of recipient countries. In the case of inflation, a decrease in the growth rate can lead to the government applying expansionary policies, which can cause inflation. The government’s final consumption expenditure can also be affected by changes in the growth rate during the period considered, because public expenditure is generally anti-cyclical. On the other hand, exogenous changes in this period’s growth rate can lead to a variation in the amount of imports and exports and, therefore, affect the openness of the country to international trade. Additionally, we must consider the relationship between commercial shocks and aid, given that if a country suffers an adverse shock to its economy, the aid flow towards this country tends to increase. Predetermined. ‫ݕ‬௧బ ǡ ‫ܿݑ݀ܧ‬௧బ , ‫݌ݔܧ‬௧బ , ܴ‫ܮ݈݁ݑ‬, ܲ‫ܾ݋‬. Variables of initial conditions are treated as predetermined, for they are correlated to past values of the error term. In addition, the double direction of causality between growth and institutions leads us to treat it as a predetermined variable. The demographic variable is also considered predetermined

The Effectiveness of Development Aid in Promoting Economic Growth 125

because changes in growth will affect fertility decisions in subsequent periods. Exogenous. ‫ܿܽݎܨ‬, ‫݀݊ܽܮ‬, ܴܰ. These variables are not considered to be correlated to past or present values of the error term. Taking into account all these issues, we choose regression models with instrumental variables to estimate our model. Thus, we use as instruments of non-exogenous variables others with which they are not correlated and which are orthogonal to the error term. There are several estimators to analyse dynamic panel data, mainly those of Arellano and Bond (1991) and those of Arellano and Bover (1995) and Blundell and Bond (1998). In both cases, it is a particular case of instrumental variables: the generalized method of moments (GMM). These methods are appropriate, according to Roodman (2009), in situations where there are: 1) Panel data with “small T and large N”, that is to say, many individuals or transversal observations and few time periods. 2) Independent variables which are not strictly exogenous, for they are correlated with past and possibly present values of the error term. 3) Fixed individual effects, which in our case are fixed country effects. 4) Heteroskedasticity and autocorrelation within individuals, but not across observations of different individuals. The Arellano-Bond method is known as difference GMM. In this paper, we work with the Arellano-Bover/Blundell-Bond method, as it allows us to introduce more instruments and therefore its estimation is more efficient. It is known as system GMM because it builds a system of equations, the original in levels and the transformed one (in differences), thus eliminating the correlation with fixed effects. In the model we have developed to analyse the impact of aid there are many variables and it is not possible to estimate a model with all of them if we want to avoid an excessive number of instruments, which could result in inefficient estimations. Following the general rule stated by Roodman (2009), the number of instruments is limited so that it does not surpass the number of groups (countries) analysed. Therefore, we estimate reduced equations subsequently substituting the least statistically significant variables, though always maintaining the aid variables (A y A2). Consequently, the estimation of the model is carried out with the software Stata applying the system GMM method and controlling for fixed effects (country specific), time specific effects and the endogeneity of certain independent variables. In order to control for time effects we

Chapter Six

126

include time dummies to make it easier for the assumption that there is no correlation across individuals in the idiosyncratic shocks to hold and thus increase the estimators’ robustness. In addition, we apply the small-sample correction in the estimation of the covariance matrix (therefore obtaining t statistics instead of z statistics and test F instead of Wald ߯ ଶ . The estimation is done in two steps applying the Windmeijer correction (2005) to avoid the downwards bias of standard errors and it is done over orthogonal deviations instead of differences, which is recommended when the panel is unbalanced. Finally, the instruments matrix is restricted, creating an instrument for each variable and lag distance, thus reducing the bias present in small samples when the number of instruments is close to the number of observations.

6.3.3. Variables, Sample and Time Structure 6.3.3.1. Dependent Variable The annual cumulative growth rate of GDPpc PPP is used as the dependent variable in this model. It is calculated using the following formula: ‫ܩ‬௜ǡሺ௧బǡ்ሻ ൌ ͳͲͲ ‫ כ‬ቆ

೅ష೟బ ௬೔ǡ೅

ට௬೔ǡ೟ െ ͳቇfor country i, with T being the last బ

year and ‫ݐ‬଴ the initial year of the period considered. In most of the articles on aid effectiveness this variable is used as the regressand of the model, though it only reflects the economic growth of the country and not its development, which is the objective of the majority of types of aid. However, development is a difficult concept to measure and, therefore, we use the growth rate as the dependent variable, for it is one of the easiest components of development to quantify and there is ample information available about it. Additionally, though it is possible to achieve economic growth without development, it is neither easy nor plausible to achieve the opposite situation. 6.3.3.2. Independent Variables Aid, whose impact on economic growth is the focus of our study in this paper, is measured by the Official Development Assistance. It does not include all aid flows, but in those countries with the lowest income per capita, ODA accounts for the majority of those flows, and therefore it is an appropriate variable to approximate the aid received by the country.

The Effectiveness of Development Aid in Promoting Economic Growth 127

The endowment of natural resources, which in theory remains constant in the short term, is measured as the average of the annual proportion of exports of agricultural products, minerals and fuels to GDP. The time period is divided into two, calculating an average for the two first fouryear periods and another one for the other two (to maximize available data). Time dummies have also been included to improve the estimation of the model. The variables present in the model and their data source are described in Table 4.1. The averages have been calculated as the arithmetic average of the annual values. 6.3.3.3. Sample of Countries The great majority of countries in the DAC’s List of ODA Recipients in 2012 and 2013 are African, and consequently the analysis is centred on them; in particular, on those African countries which are considered least developed and of low income (according to the classification by gross national product per capita in 2010). There are 36 countries with these characteristics, nine of which have been excluded due to a lack of statistical information, even though excluding them could possibly lead to biased results. It is the case of Comoros, Djibouti, Equatorial Guinea and Sao Tomé and Príncipe, which are countries with less than one million inhabitants and do not offer much information about their economic situation, as well as South Sudan, country which was set up in July of 2011. The other excluded countries are Somalia – a country which through the last decades has suffered a lot of division and political instability as well as armed conflicts-, Chad, the Democratic Republic of Congo and Liberia since this last one has gone through two civil wars and there was a lot of political instability before the elections of 2005. In the regressions where the variable ‫ ݄ܵܿܧ‬or its interaction with aid is included, we exclude Sierra Leona from our analysis due to lack of information. Table 6-1: Description of the Variables and Data Sources Variable

‫ܩ‬ሺ‫ݕ‬ሻ

Description

Data source

Annual cumulative average of the variation of GDPpc PPP during the period analysed. Constant prices, 2011 international dollars.

Prepared by the author based on data from the World Bank (2014)

Chapter Six

128

Ž൫‫ݕ‬௧బ ൯

Natural logarithm of GDPpc PPP of the initial year of the period being considered. Constant prices, 2011 international dollars.

World Bank (2014)

‫ܣ‬

Annual average of the ratio of ODA to GDP.

DAC (2014)

‫݃ܽݎܨ‬

Annual average of a Hirschman-Herfindal index constructed as the sum of the squares of the relative share of net aid provided by each ODA donor over the total sum of ODA in current US dollars (Values from 0- maximum fragmentation- to 10.000-null fragmentation).

Prepared by the author based on data from the DAC (2014)

ܸ‫݈݋‬

Annual average of the ratio of variances of the ratio of ODA to GDP and of the ratio of government revenues (taxes and donations) to GDP. Variable used by Bulir and Hamann (2008).

Prepared by the author based on data from the World Bank (2014) and the African Development Bank Group (2014).

Dummy variable (value 1 if the country is landlocked and value 0 in any other case).

Author’s own work

Annual average of two indicators of governance: rule of law and control of corruption.

Kaufmann, Kraay and Mastruzzi (2013)

‫݂݊݋ܥ‬

Dummy variable which is 1 if the country has been involved in a conflicto in the period considered or in the four previous years and 0 in any other case.

Centre for the Study of Civil War (2009)

‫݄ܵܿܧ‬

Annual average of the difference between the percentage of the index of export prices over the index of import prices (base year 2000=100) and 100.

Prepared by the author based on data from the African Development Bank Group (2014).

‫݀݊ܽܮ‬

ܵ‫ܮ݁ݐܽݐ‬

The Effectiveness of Development Aid in Promoting Economic Growth 129

Annual average of the gross enrolment ratio in primary education in each country during the period preceding the one considered.

World Bank (2014)

Natural logarithm of the life expectancy in the initial year of the period analysed (years).

World Bank (2014)

‫ݒ݋ܩ‬

Annual average of the ratio of government final consumption expenditure to GDP.

World Bank (2014)

‫݈݂݊ܫ‬

Annual average of the inflation rate (measured by the anual variation of the GDP deflator).

World Bank (2014) and African Development Bank Group (2014)

ܱ‫݊݁݌‬

Annual average of the sum of exports and imports to GDP.

World Bank (2014)

‫ܿܽݎܨ‬

Ethnic fractionalization index. Values from 0 (a unique ethnic group) to 1 (complete fractionalization)

‫ܿݑ݀ܧ‬௧బ

ln(‫݌ݔܧ‬௧బ )

ܴܰ

Annual average of the share of exports of agricultural products, fuels and minerals to GDP.

ܲ‫ܾ݋‬

Difference between the growth of the economically active population and that of total population.

Database on Fractionalization of Alesina et al. (2003) World Trade Organization (2014) World Bank (2014)

6.3.3.4. Time Period and Structure The time period considered is 16 years, from 1995 to 2010, and we divide it into 4 consecutive periods of four years. In the majority of studies about the effectiveness of aid, authors use periods of four, five or more years, for example ten. Choosing periods of four years allows us to maximise the available data and to smooth short-term fluctuations in the growth rate.

6.4. Results The results of the different models are presented in the columns of Table 5.1. The coefficients of the variables which appear in each model can be found in each table and in brackets, the p-values. The temporal dummy variables do not appear, but they have been included in the

130

Chapter Six

estimation and they are not significant in any of the estimated models. It is clear that the variables which have had a statistically significant impact in most of the estimations are aid and the interaction between aid and postconflict situations. In the case of aid, if its share in GDP increases by a percentage point, the positive effect on the growth rate will be between 0.47 and 0.64 percentage points, ceteris paribus. This figure could seem small but considering the magnitude of the aid flows to these countries, it has a considerable effect on their growth. This result is coherent with the majority of studies done on the effectiveness of aid, which, as stated in previous sections, find a positive link between aid and growth. Additionally, this relationship presents diminishing marginal returns, for the quadratic term of aid is negative and statistically significant in two of the regressions; that is to say, the bigger the aid flow is in terms of GDP, less effect will an additional percentage point of aid have in promoting growth. These results are in agreement with those of Hadjimichael et al. (1995) and Hansen and Tarp (2001). It can be caused by the existence of limits to the absorption capacity of aid in these countries, due to their structural and institutional characteristics, or because increases in aid flows can lead to the Dutch syndrome, that is to say, they can cause appreciations of the domestic currency, decreasing their competitiveness in international markets. The interaction variable ‫ ݂݊݋ܥݔܣ‬is negative and statistically significant in all the regressions it appears in. In other words, ceteris paribus, aid will be less effective when the country has recently been involved in a conflict. These results do not coincide with those of Collier and Hoeffler (2004), who concluded that aid had a greater positive effect on economic growth after conflicts. However, they are in agreement with the results of Chauvet and Guillaumont (2004), who state that aid is less effective in situations of political instability, which is common after armed conflicts. In any case, these results can be explained by the fact that after conflicts there are more possibilities for opportunism and groups in power can more easily “expropriate” aid. Therefore, it could be that this variable is a better proxy for institutions than ܵ‫ܮ݁ݐܽݐ‬. Furthermore, in these situations, there are normally less incentives or opportunities for long term investments and this could lead to a less productive use of aid flows. In addition, we must consider that the heterogeneity of aid has not been taken into account and certain types of aid may be more effective than others in these situations. Consequently, the coefficient can simply be reflecting the ineffectiveness of the kind of aid received by the country in question, but we cannot state that all types of aid would be ineffective in these cases.

The Effectiveness of Development Aid in Promoting Economic Growth 131

Aid fragmentation is statistically significant in three of the regressions and it indicates a positive link between fragmentation and growth. That is to say, a decrease of a percentage point in aid fragmentation will translate into a decrease of approximately 0.002 percentage points of growth. Generally, it has been argued that a greater fragmentation could negatively influence growth, because there would be more transaction and opportunity costs with more donors. The results of this paper indicate that for our sample of countries and time period the opposite happens. The explanation can lie in the fact that fragmentation does not necessarily have to mean lack of coordination and, furthermore, this situation allows a recipient country to “diversify” its sources of income. In other words, it will depend less on the aid flows coming from a single country and it will have more bargaining power if the donor wants to tie its aid. On the other hand, donor countries could give out aid according to their comparative advantages (they could provide different types of aid depending on their relative capabilities or endowments) and having a broad base of donors, recipient countries can benefit from the comparative advantages of more countries. The other interaction variables are not statistically significant and neither is aid volatility. Regarding the rest of variables, the only ones with a significant impact on growth are demography and the endowment of natural resources of the country. With the first variable we want to include the effect of change in the composition of the population on the growth rate and it is significant in three of the regressions where it is included. The model estimates that if the economically active population grows a percentage point more than the total population, the growth rate will increase more or less a percentage point. In African countries demographic phenomena are especially important since they have not yet experimented the demographic transition. In this continent you can find the highest child and youth dependency ratios, as while fertility rates remain high, infant mortality rate is decreasing. This population pyramid puts a strain on economic growth since it weighs down the productive capacity per capita and it generally comes with lower savings and investment rates. The endowment of natural resources has a positive impact on economic growth. An increase of a percentage point in ܴܰ translates into an increase of 0.14 percentage points in the growth rate. The issue of the effect of an abundance of natural resources on economic growth has been largely debated, with many authors arguing that it is detrimental while others state that institutions are what determine whether a country will take advantage of its potential to grow or not. Gylfason (2000) defended that the endowment of natural resources has effects on the incentives to

Chapter Six

132

accumulate human capital and, therefore, the negative effects of abundant natural resources which can be found in existing literature can, up to a point, be reflecting that effect. He argues that a high endowment of natural resources can cause low-skill intensive natural-resource based industries to quickly develop and therefore incentives to invest in education will decrease. In this model education is already taken into account with the variable ‫ܿݑ݀ܧ‬, so the coefficient of natural resources would not include that effect and it would only reflect the benefits of having additional income in the country originating from the exploitation of natural resources. Furthermore, in general countries with abundant natural resources have smaller government deficits, higher relative entry flows of foreign direct investment and higher investment rates, all of which can promote growth. Lastly, the results of the Hansen test and the second order autocorrelation test in the model in differences, which you can find in the last rows of Table 5.1, indicate that the instruments and the procedure GMM of Arellano-Bover/Blundell-Bond are valid. The Hansen test checks the exogeneity of the instruments as a group and the second-order autocorrelation test checks for the existence of autocorrelation in the disturbance term since its existence would make certain lags not valid as instruments. Table 6-2: Models for Testing the Impact of ODA on Economic Growth (1)

(2)

(3)

(4)

(5)

(6)

(7)

‫ܖܔ‬൫࢚࢟૙ ൯

-3.2629 (0.381)

-3.7531 (0.277)

-5.6565 (0.272)

-2.1832 (0.708)

-0.3877 (0.937)

-2.7827 (0.589)

-0.0164 (0.998)

࡭

0.6389 (0.052)

0.4667 (0.072)

0.5335 (0.024)

0.5136 (0.138)

0.5983 (0.003)

0.5396 (0.183)

0.6189 (0.034)

࡭૛

-0.0187 (0.075)

-0.0149 (0.041)

-0.0034 (0.589)

-0.0015 (0.885)

-0.0049 (0.380)

-0.0019 (0.869)

-0.0087 (0.269)

ࡲ࢘ࢇࢍ

-0.0009 (0.372)

-0.0010 (0.224)

-0.0017 (0.103)

-0.0019 (0.045)

-0.0015 (0.005)

-0.0021 (0.077)

-0.0014 (0.244)

ࢂ࢕࢒

0.0284 (0.346)

࡭࢞ࡸࢇ࢔ࢊ

0.0360 (0.835)

࡭࢞ࡾ࢛࢒ࢋࡸ

-0.0644 (0.810)

The Effectiveness of Development Aid in Promoting Economic Growth 133

-0.3436 (0.005)

࡭࢞࡯࢕࢔ࢌ

-0.3728 (0.014)

-0.3700 (0.001)

-0.3733 (0.006)

-0.3563 (0.000)

0.0000 (0.982)

࡭࢞ࡱࢉࡿࢎ ࡱࢊ࢛ࢉ࢚૙

-0.0316 (0.284)

0.01463 (0.772)

-0.0252 (0.731)

ln(ࡱ࢞࢖࢚૙ )

4.7431 (0.681)

5.4920 (0.582)

1.9641 (807)

ࡵ࢔ࢌ࢒

-0.0012 (0.723)

-0.0355 (0.594)

-0.0552 (0.136)

-0.0248 (0.756)

-0.0026 (0.975)

1.0023 (0.211)

0.9998 (0.082)

0.9895 (0.180)

0.3016 (0.830)

0.8942 (0.797)

ࡾ࢛࢒ࢋࡸ

1.0941 (0.081)

ࡼ࢕࢈

0.0010 (0.968)

ࡱࢉࡿࢎ

-0.0099 (0.781)

ࡻ࢖ࢋ࢔ ࡲ࢘ࢇࢉ

1.4198 (0.747)

ࡳ࢕࢜

0.1202 (0.377)

ࡸࢇ࢔ࢊ

1.2605 (0.652)

ࡺࡾ

0.1468 (0.061)

0.1420 (0.090)

0.1341 (0.481)

0.0951 (0.401)

0.0702 (0.451)

0.1111 (0.311)

0.0218 (0.652)

Prob > F Instruments Groups Hansen test AR(2) test

0.000 26 27 0.805 0.221

0.000 26 27 0.585 0.309

0.000 26 27 0.271 0.641

0.000 25 26 0.485 0.844

0.000 25 27 0.474 0.880

0.000 25 26 0.437 0.824

0.000 27 27 0.252 .473

Dependent variable: annual cumulative growth rate of GDP per capita PPP Regression with panel data. System GMM estimation in two steps Sample: Number of observations = [95,97] Number of periods: 4 (1995-1998, 1999-2002, 2003-2006, 2007-2010) Instruments: in each regression we use the instruments of the variables included in that variable Predetermined variables: ‫ݕ‬௧బ ǡ ‫ܿݑ݀ܧ‬௧బ , ‫݌ݔܧ‬௧బ , ܴ‫ܮ݈݁ݑ‬, ܲ‫ܾ݋‬, (1 and 2 lags) Endogenous variables: ‫ ܣ‬y ‫ܣ‬ଶ (2 and 3 lags). Interaction variables: ‫݈݂݊ܫ‬, ‫݄ܵܿܧ‬, ܱ‫݊݁݌‬, ‫( ݒ݋ܩ‬2 lags)

134

Chapter Six

6.5. Conclusions Aid flows towards developing countries have acquired great relevance in the last decades and, consequently, the issue of their efficacy has become a focus of attention. The estimation of the model developed in this paper allows us to conclude that aid has had a significant beneficial impact on growth, a result which is consistent with the majority of studies done regarding aid effectiveness in the last decades. Thus, we find empirical confirmation of the efficacy of aid in promoting economic growth, which supports the exhortation to provide more development assistance contained in the United Nations Millennium Declaration of 2000. Nevertheless, aid is not equally effective in all cases and, furthermore, it presents diminishing marginal returns, which is congruent with recent studies on this subject. Regarding aid fragmentation, it is generally considered to have negative effects on economic growth, but in this paper a greater fragmentation has a positive effect. As explained before, this could be due to the existence of a greater coordination in spite of the fragmentation, to the fact that it could allow a recipient to diversify its sources of income and provide it with more negotiating power regarding tied aid, in addition to allowing the country to take better advantage of comparative advantages among donor countries (only if these provide different kinds of aid according to these advantages). Also, the results of this model indicate that aid is less effective after conflicts, which could be due to the fact that in these situations there generally is more political instability, institutions do not function as well and there are less incentives to invest with a medium or long term perspective. Additionally, it could be that the type of aid provided in these cases is not the most appropriate or effective. In any case, these results should not lead to a decrease in the amount of aid given in these situations, as this is when a country can need aid the most, rather it should encourage more in-depth study of the specific impact of aid in post-conflict situations to analyse the particular reasons of its ineffectiveness and to find out whether different kinds of aid have different effects. Regarding other variables which have an impact on growth, the demographic effects and the endowment of natural resources are statistically significant. The abundance of natural resources has a positive influence on the growth rate, which could be explained because it is not the fact of having an abundance of these resources which lead to less growth, rather it depends on having governments which know how to avoid the risks that these resources can involve, such as the change in incentives to accumulate human capital. Demographic changes are one of

The Effectiveness of Development Aid in Promoting Economic Growth 135

the specific challenges African countries face compared to other developing or developed countries, for the demographic transition has not yet taken place and there is a high child dependency ratio which weighs down the economic growth of the country. There are many courses of action which could be taken to expand and improve this study. Several problems have arisen when constructing the database for this empirical work, mainly the lack of statistical data. Therefore, when more data is available, it would be interesting to include other variables in the model to capture more determinants of growth or to improve how certain characteristics of the economies analysed are measured. For example, human capital could be better estimated with literacy rates or net enrolment rates in different education levels. On the other hand, using the dependent variable used by Tezanos et al. (2009), the growth rate of the GDP per capita excluding the tenth income decile, could throw more light on the real impact of aid. However, for now it is not possible since we do not have enough available information to introduce these variables. In this study we have used the variables mentioned in previous sections to be able to have a wide sample with data for almost all countries and to avoid as much as possible the bias which comes with omitting countries with not much data available (which are generally those with the lowest income). Even so, our panel is unbalanced since in several countries data availability is limited for some years, generally due to political instability. This empirical work could also be expanded testing the robustness of the results, using another sample and time period. Additionally, it would be interesting to introduce aid heterogeneity in the study, taking into account the different kinds of aid so as to analyse the differences in effectiveness among them. As we have previously mentioned, one of the main problems when studying the relationship between aid and growth in these countries is the quality, availability and reliability of the data collected, which in many cases makes us use less desirable variables about which we can find more information. In addition, the existence of economic and political interests can determine the allocation of aid and whether it is tied or not, thus distorting its effectiveness. These problems should not deter us from studying this issue, but lead us to consider the results as limited by the existing information and methodology and interpret them accordingly. In conclusion, we find a significant positive link between aid and growth using a dynamic model which takes into account the endogeneity of aid. This result helps support the defence of the usefulness of ODA, though we are only taking into account its effectiveness at a

136

Chapter Six

macroeconomic level and not in other facets of development. Therefore, it is advisable to continue studying the effects of aid considering its heterogeneity and its possible effects on other aspects of development.

References African Development Bank Group (2014). Open Data for Africa. Socio Economic Database Enero 2014, available at http://opendataforafrica.org/dqcelid/afdb-socio-economic-database-jan2014 Alesina, A., Devleeschauwer, A., Easterly, W., Kurlat, S. & Wacziarg, R. (2003). Fractionalization. Journal of Economic Growth. 8(2): 155-94. Ethnic Fractionalization Database available at http://www.nsd.uib.no/macrodataguide/set.html?id=16&sub=1 Arellano, M. & Bond, S. (1991). Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment Equations. Review of Economic Studies, 58: 277-97. Arellano, M. & Bover, O. (1995). Another Look at the Instrumental Variable Estimation of Error-Components Modelos. Journal of Econometrics, 68: 29-51. Barro, R. (1991). Economic Growth in a Cross Section of Countries. The Quarterly Journal of Economics, 106(2): 407-43. —. (1996). Determinants of Economic Growth: A Cross-Country Empirical Study. NBER Working Papers 5698, National Bureau of Economic Research, Inc. Blundell, R. & Bond, S. (1998). Initial Conditions and Moment Restrictions in Dynamic Panel Data Models. Journal of Econometrics. 87: 115-43 Bulir, H. & Hamann & A.J. (2008). Volatility of Development Aid: From the Frying Pan into the Fire. World Development, 36(10): 2048-66. Burnside, C. & Dollar, D. (2004): Aid, policies and growth: reply. American Economic Review, 94: 781-84. Centre for the Study of Civil War (2014): Data on Armed Conflict (version 4- 2014), available at http://www.pcr.uu.se/research/ucdp/datasets/ucdp_prio_armed_conflict _dataset/ Chauvet, L. & Guillaumont, P. (2004): Aid and growth revisited: Policy, economic vulnerability and political instability. En B. Tingodden, N. Y. Stern and I. Kolstad (Eds): Towards pro-poor policies – Aid, Institutions and Globalization (pp. 95-109), Washington DC: World Bank - Oxford University Press.

The Effectiveness of Development Aid in Promoting Economic Growth 137

Collier, P. & Gunning, J.W. (1999). Why Has Africa Grown Slowly? Journal of Economic Perspectives, 13( 3): 3-22 Collier, P. & Hoeffler, A. (2004). Aid, policy and growth in post-conflict societies. European Economic Review, 48: 1125-45. DAC (2014). Creditor Reporting System 2014, available at http://stats.oecd.org/Index.aspx?datasetcode=CRS1 Djankov, S., Montalvo, J.G., Reynol-Querol, M. (2009): Aid with multiple personalities. Journal of Comparative Economics, 37: 217-29. Fischer, S. (1993). The Role of Macroeconomics Factors in Growth. NBER Working Paper 4565 Hadjimichael, M.T., Ghura, D., Mühleisen, M., Nord, R., Uçer, E.M. (1995). Sub-Saharan Africa: Growth, savings and investment, 1986-93. Occasional Paper 118, Washington D.C. International Monetary Fund. Hansen, H., Tarp, F. (2000). Aid effectiveness disputed, Journal of International Development, 12: 375-98 Hansen, H, Tarp, F. (2001). Aid and growth regressions. Journal of Development Economics, 64: 547-70 Juselius, K., Moller, N.F., Tarp, F., (2011). The Long-Run Impact of Foreign Aid in 36 African Countries. WIDER Working Paper, 51. Kaufmann, D., Kraay, A. & Mastruzzi, M. (2013). The Worldwide Governance Indicators (WGI) project, available at http://info.worldbank.org/governance/wgi/index.aspx#home Mosley, P. (1987). Overseas Aid: Its Defense and Reform. Wheatsheaf Books. Quiñones, A. & Tezanos, S. (2012). Innovación, ayuda y crecimiento: ¿un “trío” de conveniencia?, Documentos de trabajo sobre cooperación y desarrollo 2012/13. Cátedra de Cooperación Internacional y con Iberoamérica – Universidad de Cantabria. Quattara, B. & Strobl, E. (2008). Aid, Policy and Growth: Does Aid Modality Matter? Review of World Economics, Springer. 144(2): 34765. Rajan, R.G. & Subramanian, A. (2005). What Undermines Aid’s Impact on Growth? IMF Working Paper, 126. Roodman, D. (2009). How to Do xtabond2: An Introduction to “Difference” and “System” GMM in Stata. Stata Journal, 9(1): 86136. Sachs, J.D. & Warner, A.M. (1997): Sources of Slow Growth in African Economies. Journal of African Economies, 6(3): 335-76. Sala-i-Martin, X. (1997): I just run four million regressions. American Economic Review. 87(2): 178-83

138

Chapter Six

Tezanos, S., Madrueño, R. & Guijarro, M. (2009). Impacto de la ayuda sobre el crecimiento económico. El caso de América Latina y el Caribe. Cuadernos Económicos de ICE, 78: 187-220. Windmeijer F. (2005). A finite simple correction for the variance of linear efficient two-step GMM estimators. Journal of Econometrics, 126(1): 25-51. World Bank (2014). World Development Indicators 2014, available at http://data.worldbank.org/data-catalog/world-development-indicators World Trade Organization (2014). Merchandise trade and commercial services database, available at http://stat.wto.org/Home/WSDBHome.aspx?Language=E

CHAPTER SEVEN OFFSHORING AND PRODUCTIVITY FROM A TIME-SERIES PERSPECTIVE PABLO AGNESE*

This chapter proposes a simple framework for estimating the contribution of offshoring strategies to the growth rate of labor productivity from a time-series perspective. For this empirical question, I consider a group of Japanese industries during the recent years of slow growth. The results should be interpreted with caution, yet clearly suggest that offshoring can improve labor productivity in the Semiconductors industry.

7.1. Introduction The present chapter takes up the study of offshoring and productivity using a time-series perspective for the industry, something on which the literature has not yet produced a clear consensus. To achieve this, we will take a look at a major event in the Japanese economy: the ‘lost decade’  that period of economic contraction which spans from somewhere in the 1990s to the present day, and which is characterized by a slowdown of the growth rate of productivity. With this landscape in mind, our main objective will be to assess the impact of offshoring strategies on the performance of industries. We will propose a simple model that allows the net effect of offshoring on the growth rate of labor productivity to be derived for different *

FH Duesseldorf (University of Applied Sciences), Department of Business Studies, Universitaetsstr., Gebäude 23.32, 40225 Düsseldorf, Germany; and IZA Institute for the Study of Labor, Schaumburg-Lippe-Str. 5-9, 53113 Bonn, Germany. Email: [email protected]. Older versions of this chapter are available as IZA DP-7323 and Duesseldorf WP in Applied Management and Economics 23.

Chapter Seven

140

industries. Our empirical exercise will simulate the growth rate of labor productivity after the East Asian crisis (1997), assuming that the offshoring strategies remained unchanged at that initial level. Considering that over the last few years, the offshoring of Japanese activities has really leaped forward (see Agnese, 2012), it is of interest to see whether these activities may have prevented a more pronounced downturn of economic activity.1 The Japan Industrial Productivity (JIP) Database (RIETI, 2011) will provide the data on a vast number of economic activities classified as 108 industries, out of which we will be using four for reasons that will become clear below. Our results, which are in line with the recent literature, point to the existence of positive effects of offshoring on the growth rate of labor productivity for some of the industries (see Amiti and Wei, 2009, for the US; Girma and Görg, 2004, for the UK; and Hijzen et al., 2010, for Japan). The industries are classified by the JIP Database as information technology (IT) manufacturing industries. Table 7-1 condenses some comparative information on the broad sectors within the database for which data were available (1976-2008), and shows that IT-Manufacturing industries are traditionally more productive. Even when all sectors have been hit by the slump, the IT-Manufacturing sector has still fared reasonably well. Table 7-1: Labor Productivity Growth Rate, Broad Sector Averages (%) 19762008

19761980

19811990

ITManufacturing

5.81

8.14

IT

4.32

5.99

Manufacturing

3.53 2.8

All

19912000

20012008

9.51

3.4

2.73

6.68

3.21

1.73

6.24

6.86

2.19

-0.63

3.68

4.69

2.78

-0.09

Source (all tables and figures): JIP Database (RIETI, 2011), own calculations.

Figure 7-1 complements the previous information and shows in more detail the downward trend of labor productivity for the same broad sectors. 1

The recent Japanese experience is very well documented in several places and from different angles (see Caballero et al., 2008, Fukao and Kwon, 2006, Hayashi and Prescott, 2002, or Krugman, 1998, among many others).

Offshoring and Productivity from a Time-Series Perspective

141

This general decline in the performance of the economy was greatly felt especially after the 1997 crisis, when the growth rate of labor productivity stood at negative levels for many of the years that followed. Figure 7-1: Labor Productivity Growth Rate, Broad Sector Averages (%) a. Waning growth, 1976-2008

b. The aftermath of the Asian crisis, 1997-2008

24

12

20

Asian crisis (1997)

16

8

12

4

8 0 4 0

-4

-4

Avg Avg Avg Avg

-8 -12 1980

1985

IT-Manufacturing IT Manufacturing all

1990

1995

Avg Avg Avg Avg

-8 -12 2000

1998

2005

2000

IT-Manufacturing IT Manufacturing all

2002

2004

2006

2008

Since the IT-Manufacturing industries were the least affected by the stifling economic conditions during the aftermath of the Asian crisis, we might want to know whether the industries there do share some specific traits that could explain their not-so-bad performance. IT industries are at the forefront of innovation and technological developments; offshoring, in particular, can be thought of as a managerial innovation whereby industries, as with any other technological improvement, can become more efficient. We need now to measure offshoring before going any further. Our offshoring definition for the industry is similar to Feenstra and Hanson (1996), and is given by the following expression: os

I 3 Q D

where the first term is defined as the share of intraindustry purchases of inputs (I) in total (nonenergy) inputs (Q), and the second term is an economy-wide import share defined as the imports of the good from the same industry abroad (3) in the domestic demand (D). The idea of this is to proxy offshoring by the import content of inputs. Given that the data for both terms are taken from the same source (this is not generally so), it is expected that the measurement errors underlying such an index will be significantly reduced. Moreover, this is a narrow definition of offshoring as it only accounts for the intermediate inputs that a local industry imports from the same industry abroad. I think this narrow

142

Chapter Seven

measure makes for a better choice when using aggregate industry-level data, as it will clearly diminish the aggregation bias (on this issue, see Fukao and Arai, 2013). Unfortunately, the structure of our database does not allow for a distinction between destination countries (e.g. low and high-income countries). Therefore, I am assuming that, first, offshoring is homogeneous across firms of the same industry and, second, the index above is mainly grasping the intraindustry trade that usually takes place with lower income countries, hence, offshoring. To drive the point home, we need to highlight the trend of the offshoring intensity of IT-Manufacturing industries among the most representative ones within the database.2 Figure 7-2a shows an important difference regarding these strategies. Furthermore, Figure 7-2b zooms in on a small set of IT-Manufacturing industries that show different trends. Two of these industries, Semiconductors and Electronic Parts, are distinctly above the sector’s average, whereas the other two, Industry Machinery and Electronic Equipment, are clearly below. Seemingly, highly productive IT-Manufacturing industries can have very different offshoring strategies. With the exception of Industry Machinery, the industries show a high growth rate of labor productivity during the post-crisis years (1997– present). It is left to wonder if offshoring might have had anything to do with these experiences. In addition, we might want to know if offshoring played any role in the major changes that took place during those years (e.g. a significant drop in the labor income share). This chapter is organized as follows. Section 7.2 goes over the model. Section 7.3 discusses the data and methodology, and shows the estimation of the model for the four industries discussed earlier. Section 7.4 uses the models from the previous section to produce a dynamic accounting exercise on the effects of offshoring on productivity. Finally, section 7.5 concludes.

2 Among these were some non-IT Manufacturing and some IT non-Manufacturing. Note that highly developed industries, while being exposed to new technologies more rapidly, can also engage in what has come to be known as `services' offshoring. The use of a narrow measure of offshoring prevents us from further breaking down the data into `materials' and `services offshoring', as it is sometimes done (see, for instance, Amiti and Wei, 2009, for the US, or Ito and Tanaka, 2010, for Japan).

Offshoring and Productivity from a Time-Series Perspective

143

Figure 7-2. Offshoring intensity and IT-Manufacturing (%)

a. In perspective, 1976-2008 4.0

b. Selected industries, 1976-2008 12

IT-Manufacturing (avg.) IT-Manufacturing + other (avg.)

3.5

Semiconductors Electronic parts Industry machinery Electronic equipment IT-Manufacturing (avg.)

10

3.0

8

2.5 6 2.0 4

1.5

2

1.0

0

0.5 1980

1985

1990

1995

2000

2005

1980

1985

1990

1995

2000

2005

7.2. A Simple Model Here, I propose a model that I will later use to track down the possible effects of offshoring on productivity growth. For this purpose, we assume a Cobb-Douglas production function as below:

Y

Aos, t K D N 1D

(1) where Y is the output supplied by the firm, K its stock of capital, N its employment level, Į and 1Į the input elasticities, and A the technology shifter  which is dependent on the offshoring index as defined above and satisfies that Ac(os) ! 0 , and a time trend t denoting other time-varying factors. The marginal productivities in this case are given by: wY DAos, t K D 1N 1D MPK wK MPN

wY wN

1  D Aos, t K D N D

(2) (3)

Costs are defined as usual, as the total expenditures on inputs: C

RK  WN

where R and W are the input prices, real interest rate and real average wages. Cost minimization, given a certain level of output, determines that the ratio of marginal productivities be equal to the ratio of factor prices,

Chapter Seven

144 MP N

W



MP N

W

R MP K R that is MP K . Expressing the latter as implies that the slope of the isoquant (e.g. the marginal rate of substitution) and isocost are C W both equal. From the cost definition above, we know that K  R   R N

wK

W



MP N



>

1D



K

@

MP K D N R , and from (2) and (3) we know that . and wN Finally, the cost minimization equilibrium condition is, as always: ª 1  D K º §W · « » ¨ R ¸ N D ¬ ¼ © ¹ From here, it is possible to derive the conditional factor demand for both capital and labor. Clearing up N in the equilibrium condition above, N 1DD WR K , which we plug then into (1) to obtain the conditional we get  1 D 1 D W 1 D factor demand for capital K Aos , t 1D  R Y . To obtain the

>

@

conditional factor demand for labor, we substitute the last expression into (1), which yields the following symmetric expression: N

D D ª º 1 § 1  D · § R · ¸ ¨ ¸ »Y « Aos, t ¨ W D © ¹ © ¹ ¼» ¬«

(4)

where employment is dependent, among other things, on the offshoring intensity index. Aggregating (1) and (4) to the industry level, taking logs, and adding lags, errors, and the time subscript, we get an estimable system which is useful for tracking down the effects of offshoring on the industry’s labor productivity: y t D 1  I1 y t 1  E1 nt  Gk t  O1ost  W 1t  H 1 (5)

nt

D 2  E 2 nt 1  Jwt  I 2 y t  O2 ost  W 2 t  H 2

(6)

Tt

yt  nt

(7)

where the small letters are the variables in logs with their corresponding coefficients, to wit: I1 and E 2 the output and employment inertia coefficients,3

3

E1

and

G

the employment and capital elasticities to output,

Introducing dynamics here allows us to conceive the existence of some frictions within the labor market. For the labor demand in particular, these can be interpreted as the adjustment costs employers face when significant training and firing costs are present---and these frictions are consistent with the presence of

Offshoring and Productivity from a Time-Series Perspective

J

O2 the offshoring semielasticities, W 1 and W 2 the trend coefficients, and H 1 and H 2 the normally the wage elasticity to employment,

O1

145

and

distributed errors with constant variance. Finally, let T be the productivity of labor, which is defined as the ratio of output to the labor input. Since both output and labor are expressed in logarithms, we can approximate labor productivity by (7)  and then track down the changes in T 'T as we set out to do originally. We also expect the following signs for the coefficients: 0  I1  1 and 4 0  E 2  1 , as to guarantee dynamic stability, E1 ! 0 and G ! 0 , J  0 (a negatively sloped labor demand), I 2 ! 0 and O1 ! 0 and O 2  0 . On the latter two coefficients, a few remarks are in order. Amiti and Wei (2009) identify four possible channels through which offshoring can affect A and, as per our model above, also T. First, when industries decide to relocate activities to overseas locations, they relocate the less efficient parts and average productivity increases due to a compositional effect (static efficiency gain). Second, the remaining workers may become more efficient if industries can restructure in a way that pushes out the technology frontier (restructuring). Third, industries can learn to improve the way activities are performed by importing services (learning externalities). And fourth, productivity could increase due to the use of new material or service input varieties (variety effects). Our model cannot distinguish the exact channel of the productivity gain arising from offshoring, yet we can assume that these are embedded into O1 and O2 . We can further suspect that T in particular will be more directly affected by the compositional and restructuring effects. Our focusing on T (and not A) stems from the following reasons. First, the literature has yet to produce more evidence on this link as labor is the main factor involved in offshoring strategies. Second, by allowing for a two-equation specification as in (5)-(6), we can reduce the endogeneity of inputs present in simpler single equation specifications. Lastly, and related involuntary unemployment which, in turn, is a likely outcome of offshoring practices for some workers. E1

G

The constant returns to scale hypothesis would require that 1 E  1 E 1. 2 2 Notice that down below we do not constrain the equations as to fulfill this hypothesis (the estimation results are not that different from one another anyhow). Non-constant returns would imply that (1) and (4) should be slightly changed but this is of no real importance for the empirical analysis below. 4

146

Chapter Seven

to the previous reason, our simulations still offer a simple and straightforward exercise of the contribution of offshoring to labor productivity while accounting for cross-equation dynamics.

7.3. Empirical Analysis 7.3.1. Data and Methodology The data I use for this study come from the JIP Database (RIETI, 2011), ranging from 1970 to 2008 and including 108 industries from different branches of economic activity, with services (42) and manufacturing (52) being the main categories. The four industries under study are categorized as both manufacturing and IT-intensive industries (19 in total). Table 7-2 summarizes the main information on the time-series that I use in the estimation below. Notice that, as in all tables and figures, the industries are ordered by their ‘offshoring intensity’, namely: 1) semiconductor devices and integrated circuits, 2) electronic parts, 3) general industry machinery, and 4) electronic equipment and electric measuring instruments. We must now briefly discuss the selection of these four particular industries. In general, the industries considered stand out in terms of output and employment, are at the forefront of innovation and technological development (less in the case of Industry Machinery), and display a clear pattern in the offshoring intensity index in recent years (two above average and two below).

Offshoring and Productivity from a Time-Series Perspective

147

Table 7-2: Summary Statistics, 1976-2008 Mean

N: employment (workers) W: average real wages (million yen) Y: real output (million yen) K: real net capital stock (million yen) os: offshoring index (%) ǻș: labor productivity growth rate (%)

Max. Semiconductors

Min.

Std. dv.

145,192

236,251

7,252

80,217

4.12

16.46

1.06

3.38

1,245,543

4,274,040

14,16

1,145,483

4,715,221

12,665,640

402,851

3,764,188

2.28

11.57

0.04

3.45

9.58

77.24

-33.93

18.85

Electronic Parts

N: employment (workers) W: average real wages (million yen) Y: real output (million yen) K: real net capital stock (million yen) os: offshoring index (%) ǻș: labor productivity growth rate (%)

463,964

570,711

213,386

118,487

3.05

7.54

0.84

2.01

2,412,535

6,686,689

240,717

1,769,861

2,928,973

6,185,476

526,448

1,994,487

2.03

6.25

0.31

1.5

8.04

36.89

-30.39

14.33

Industry Machinery

N: employment (employees) W: average real wages (million yen) Y: real output (million yen) K: real net capital stock (million yen) os: offshoring index (%) ǻș: labor productivity growth rate (%)

436,49

503,17

351,355

32,846

4.86

6.64

3.33

0.92

2,819,787

3,841,516

1,550,403

642,478

5,594,873

9,077,593

2,217,262

2,241,890

1.45

3.48

0.6

0.72

1.92

19.57

-13.45

6.83

148

Chapter Seven Electronic Equipment

N: employment 113,889 146,309 72,26 20,44 (workers) W: average real 5.28 11.59 0.78 3.09 wages (million yen) Y: real output 882,535 1,770,692 153,761 475,369 (million yen) K: real net capital 1,006,303 1,842,645 218,022 500,279 stock (million yen) os: offshoring index 1.37 2.89 0.57 0.59 (%) ǻș: labor productivity growth 14.77 323.61 -58.43 60.57 rate (%) Note: 33 observations (1976-2008); Y is gross value added (factor prices); Y, K, and W were deflated using the GDP deflator (JIP Database 2011). Variables not in logs.

It is true that a somewhat arbitrary decision on what industries to use must be taken, for the idea is to proxy offshoring as best as possible and then see how this affects productivity. Due to the relatively high level of aggregation of the data, we must aim at those industries that have the technological edge (those like Semiconductors and Electronics Parts, and Electronics Equipment) or else the tradition (Industry Machinery) for undertaking offshoring practices. Note that the industry labeled as Industry Machinery, even when presumably lacking the technological edge, is typically characterized by large scale mass-production firms where offshoring presents itself as a possible, and sometimes the only, avenue for further productivity increases.5 The estimation strategy involves the Autoregressive Distributed Lagged (ARDL) approach by Pesaran (1997), Pesaran and Shin (1999), and Pesaran et al. (2001). The ARDL yields consistent estimates for the short and long-run that can be used when regressors are either I(1) or I(0), as it is in our case. For each industry, I estimate a two-equation system that allows me to track down the changes in labor productivity and assess the contribution of offshoring. Equations (5) and (6) are first estimated separately and evaluated against a set of diagnostic tests. Both equations are then estimated jointly 5

I tried working with some other industries within IT-Manufacturing too, but either the results were not significant or the industries were not relevant their weight in terms of output or employment was small, or the industry itself was not bound to engage in offshoring, e.g. pottery, smelting of non-ferrous metals, etc.

Offshoring and Productivity from a Time-Series Perspective

149

with the three-stage least squares method (3SLS), which accounts for potential endogeneity and cross-equation correlation. The potential endogeneity of some of the variables is something we should take into consideration. In particular, for the offshoring index, endogeneity can be further magnified by the presence of measurement errors. To solve for this, we instrument wages, capital, and the offshoring index, with the past values of wages and capital.6 As a final step, we should check on the validity of the long-run relationships among the growing variables in the models. For this, I reparametrize the estimated equations as error correction models (ECMs) and obtain the cointegrating vectors (CVs) among the I(1) variables. Even when the ECM on its own gives proof of cointegration of the time-series involved, I also use Johansen’s cointegration procedure (Johansen, 1988) to check whether the long-run relationships conform with those obtained through the estimation of the two-equation model.7

7.3.2. Estimation Tables 7-3a to 7-3d present the two-equation models for all four industries. Note that in all cases, the coefficients are properly signed (e.g. as hypothesized above), and in most cases turn out to be significant at conventional levels. The offshoring coefficients, however, turn out with a lower significance and, in some cases, are not significant at all (this is denoted with an asterisk in the tables).8 For the Semiconductors industry, they are significant at 5% in both equations (Table 7-3a); for the 6

We also tried other exogenous instruments for the offshoring index in particular, namely: the investment in information technology used to produce software and hardware, but the results were not changed significantly. Moreover, the endogeneity of offshoring does not pose so serious a problem for industry level data as it does for firm level data. Regardless, the validity of the instruments and of the overidentifying restrictions must still be checked---this I do below by means of a conventional Sargan test. Finally, the time trend is omitted because it is generally non-significant and would drastically change the results in some of the specifications. On this, it must also be pointed out that additional information about other technology shifters such as the number of patents or R&D expenditures was not available, so the results should be interpreted with care. 7 The cointegration analysis described in this last paragraph is left out for reasons of space but available on request. 8 The dynamic structure of all four models is rather unpromising too. However, it should be stressed that the introduction of dynamics is due to the fact that they are a source of frictions which can bring about involuntary unemployment and this is consistent with the offshoring story.

150

Chapter Seven

Electronic Parts industry, it is significant only in the production function at the 5% level (Table 7-3b); for Industry Machinery, it is only (yet highly) significant in the labor demand equation, at a 1% level (Table 7-3c); and for Electronic Equipment, it is neither significant in the labor demand nor in the production function. Because the number of instruments exceeds the number of regressors in the proposed models, we must test for the validity of the overidentifying restrictions. Under the null hypothesis that these are valid, the Sargan

F2

 k  p with k denoting the number of statistic is distributed as instruments and p the number of estimated coefficients. Not rejecting the test at conventional levels (e.g. above 5%) is indicative of the exogeneity of the instruments used. In spite of these results which are not totally convincing, it should be noted that all offshoring coefficients are properly signed, namely, a negative sign in the employment equations and a positive one in the output equations. This is in agreement with the economic intuition as pointed out before (see Amiti and Wei, 2009).

7.4. Effects on Labor Productivity We can now use the estimated models to obtain the contributions of the offshoring index to the growth in labor productivity. These contributions are computed through a dynamic simulation of the estimated models as follows: we first fix the offshoring index in each industry at the level of the certain arbitrary year, then we solve the model, and finally we retrieve the new path of the endogenous variable. For us, that initial year corresponds with the beginning of the Asian crisis in 1997. The endogenous variable is the labor productivity growth rate ( 'T )  as has been shown to perform badly from 1997 onwards. To illustrate this, I plot the results of these simulations as Figures 7-3a, 7-4a, 7-5a, and 7-6a, along with the simulated trajectories of the offshoring index as Figures 7-3b, 7-4b, 7-5b, and 7-6b. The first set of figures (“a”) show both the actual trajectory of the growth rate of labor productivity and the simulated trajectory, had the offshoring index remained at the 1997 value. Notice that the average growth rates for both trajectories are made explicit in the figures. The second set of figures (“b”) shows both the actual trajectory of the offshoring index and the trajectory fixed at the 1997 value. Therefore, what we get from Figures 7-3 to 7-6 below is the individual contribution of offshoring to the changes in labor productivity or, in other

Offshoring and Productivity from a Time-Series Perspective

151

words, what the growth rate of productivity would have been if offshoring had remained at its 1997 level. As can be seen from the figures, for all four industries, offshoring went up during 1997-2008, for some more than others. Table 7-3a: Semiconductors

Table 7-3b: Electronic Parts

2-Eq. system (1976-2008; total sys. obs. 66), 3SLS Dependent Dependent variable: yt variable: nt-1 cn cn -5.80 t. 1.01 [0.000] t. [0.002] yt0.08 nt0.04 [0.485] [0.562] 1 1 0.79 wt -0.63 [0.000] nt [0.000] 0.58 yt 0.82 [0.000] kt [0.021] os 6.04 os -1.88 [0.049] [0.022] t t

2-Eq. system (1976-2008; total sys. obs. 66), 3SLS Dependent Dependent variable: yt variable: nt-1 c cn 2.64 nt -5.80 t. [0.000] . [0.018] 0.19 yt0.41 nt[0.129] [0.028] 1 1 -0.50 0.86 wt [0.000] nt [0.005] 0.59 0.20 yt [0.000] kt [0.101] -1.44 os 10.88 os [0.333] [0.016] t* t

r²

0.995

0.985

S

[0.052]

[0.134]

r² S

r²

0.943

0.938

S

[0.348]

[0.685]

0.975

[0.750]

[0.888]

Table 7-3d: Electronic Equipment

Table 7-3c: Industry Machinery 2-Eq. system (1976-2008; total sys. obs. 66), 3SLS Dependent Dependent variable: yt variable: nt-1 c nt cn -7.75 . 5.45 [0.000] t. [0.004] yt0.17 nt 0.17 [0.067] [0.380] -1 1 1.20 w -0.26 [0.000] nt [0.000] t 0.28 yt 0.39 [0.000] kt [0.011] os 2.60 o -3.67 [0.000] [0.382] st t*

0.988

2-Eq. system (1976-2008; total sys. obs. 66), 3SLS Dependent Dependent variable: yt variable: nt-1 c c nt 2.73 nt -5.84 . [0.000] . [0.099] 0.47 yt0.04 nt[0.009] [0.794] 1 1 -0.29 0.58 wt [0.000] nt [0.215] 0.29 0.86 yt [0.000] kt [0.002] -3.96 os 17.54 os [0.196] [0.539] t* t* r²

0.966

0.825

S [0.817] [0.145] Note: p-values in brackets; r² the adjusted r-squared; S the p-value for the Sargan test; * offshoring coefficient not significant.

Chapter Seven

152

Figure 7-3: Semiconductors: Offshoring Contribution to Productivity a. Productivity growth rate (%)

b. Offshoring intensity (%)

60

12

40

10

20

8

0

6

11.57

-20

4 Actual (avg. 9.28) Simulated (avg. 5.47)

-40

Actual Fixed at 1997

2 1.54

0

-60 1998

2000

2002

2004

2006

2008

1998

2000

2002

2004

2006

2008

The Semiconductors industry is, according to the data, highly involved in offshoring practices. Being perhaps among the most technologically oriented industry within and outside of Japan, the Japanese semiconductor sector has achieved significantly high levels of offshoring intensity in the past few years (Figure 7-3b), especially through the hands of the big players like Toshiba and Renesas.1 Our analysis suggests that had offshoring remained unchanged at its 1997 level, then the productivity growth rate would have been lower (5.47 on average, instead of 9.28, as seen in Figure 7-3a). Most multinational companies can be said to have interests in several industries. Such is the case, for instance, of Panasonic, Fujistsu, Sony, Toshiba and the Hitachi Group, just to name a few, with interests in several and varied industries. However, their contribution over the last years to the growth of the electronics industry cannot go unnoticed.2 As for the offshoring trend, it is positive but not as important as in the Semiconductors industry (see Figures 7-4b and 7-6b),3 yet it is more than 1

These two alone account for around 8% share of the international market (iSuppli Corporation supplied rankings, 2011). See Wakasugi (1988) for the study on the evolution of the Semiconductors industry in Japan and how it acquired its international competitive capability through fierce competition. 2 Japanese electronics firms are highly respected worldwide, as is documented by the OECD Information Technology Outlook (2010). There, 44 economies were reported as bases for the top 250 ICT-firms in 2009: 75 (30%) were based in the United States, 52 were based in Japan, and 18 in Chinese Taipei. 3 Firms within the Electronics parts and Electronic equipment industries cannot be easily distinguished from one another for their many daily activities frequently overlap both classifications.

Offshoring and Productivity from a Time-Series Perspective

153

enough to have a larger impact on productivity growth (see Figures 7-4a and 7-6a; and see also Yamada, 1990). The traditional reluctance of Japanese firms in general, and of those within the electronics sector (broadly defined) in particular, is starting to show signs of breaking down, not only because of the more aggressive Asian competitors but also because of the increased risk profile of many firms in Japan's postearthquake and post-tsunami economy (see WSJ, 2011). Figure 7-4: Electronic Parts: Offshoring Contribution to Productivity a. Productivity growth rate (%)

b. Offshoring intensity (%)

40

7

20

6

0

5

5.40

-20

4 Actual (avg. 3.97) Simulated (avg. -4.90)

Actual Held at 1997

-40

3 2.30

-60

2 1998

2000

2002

2004

2006

2008

1998

2000

2002

2004

2006

2008

On the other hand, Industry Machinery (Figures 7-5a and 7-5b) is not in the least bit affected by the positive trend of offshoring in recent years. According to our analysis, labor productivity is unaffected in spite of the considerable increase in offshoring. Moreover, and as suggested before, offshoring might well not be employment friendly for this particular industry. Indeed, for reasons of costs and proximity, many Japanese firms are starting to relocate their low-end activities to China (where large numbers of Japanese speakers can be found), as well as to some other Southeast Asian countries with an abundant and cheap labor force.4

4

See the Browne and McCarthy (2007) for an analysis on the offshoring opportunities of Japan in China and India; and Ito and Tanaka (2010) for evidence on Japan which is consistent with this section.

Chapter Seven

154

Figure 7-5: Industry Machinery: Offshoring Contribution to Productivity a. Productivity growth rate (%)

b. Offshoring intensity (%)

6

3.6

4

3.2

3.48

2

2.8

0 2.4 -2 2.0

-4

Actual (avg. 5.82) Simulated (avg. 5.83)

-6

1.6

-8

1.2 1998

2000

2002

2004

2006

Actual Held at 1997 1.59

2008

1998

2000

2002

2004

2006

2008

Table 7-4 sums up the results of the dynamic accounting exercise. The first two columns show the values of the offshoring index for 1997 and 2008 respectively, which correspond to the two ends of the simulation period. The next columns exhibit, respectively, the difference for that period, the contribution in terms of productivity growth rate, and the contribution per percentage point (p.p.). Figure 7-6: Electronic Equipment: Offshoring Contribution to Productivity a. Productivity growth rate (%)

b. Offshoring intensity (%)

40

3.0

30

2.8

20

2.6

10

2.4

0

2.2

-10

2.59

2.0 1.98

-20 -30

1.8

Actual (avg. 4.81) Simulated (avg. 3.92)

Actual Held at 1997

1.6 1.4

-40 1998

2000

2002

2004

2006

2008

1998

2000

2002

2004

2006

2008

As noted before, the Semiconductors industry is the only one showing significant estimated coefficients of the offshoring variable in both equations. In terms of contribution to productivity, however, both electronics industries produce larger numbers – yet the contribution of the

Offshoring and Productivity from a Time-Series Perspective

155

Electronic Equipment industry should be interpreted carefully due to the lack of significance of both coefficients. Lastly, we found no effect on productivity for the less IT-intensive industry, labeled as Industry Machinery. Table 7-4: Offshoring Contribution to Productivity

Semiconductors

osΌΔΔΒ

os΍΋΋Γ

ǻos*

Cont. to ǻș**

Cont. per 1 p.p.

1.54

11.57

10.03

3.81

0.38 2.86

Electronic parts†

2.30

5.40

3.10

8.87

Industry machinery†

1.59

3.48

1.89

؄0

؄0

Electronic equipment‡

1.98

2.59

0.61

0.89

1.46

* In percentage points; ** actual minus simulated average (Figs. 4a, 5a, 6a, 7a). † Offshoring coefficients partially significant or ‡ not significant in estimation.

The results in this section are in line with the firm-level literature. Hijzen et al. (2010), for instance, find positive effects in Japan on the productivity levels of the offshoring firm as long as it concerns intrafirm offshoring (e.g. sourcing of intermediate inputs to foreign affiliates within a particular multinational firm), but not arm’s-length offshoring (e.g. sourcing to unaffiliated foreign firms). Notice that even when our results refer to the industry level, our preference for a narrow measure of offshoring is very much in agreement with their definition of intrafirm offshoring. In particular for them, a 1 percentage point increase in the intrafirm offshoring intensity raises TFP by 0.12%, suggesting that the median offshoring firm has a TFP growth rate which is 0.6 percentage point higher than when it does not engage in offshoring  this is consistent with our results about the effects on labor productivity, as shown in the last column of Table 7-4. There, a 1 percentage point increase in offshoring, narrowly defined, contributes to an increase in labor productivity of 0.38% and 2.86% in the Semiconductors and Electronic Parts industries respectively. The other two, Industry Machinery and Electronic Equipment, show no effect or are unambiguously nonsignificant.

7.5. Final Remarks The subject of offshoring and productivity is still in its early days when compared to studies dealing with the more direct and not so friendly employment effects. As we have seen, productivity improvements can be achieved within some industries riding on the offshoring wave. Here, we

156

Chapter Seven

have focused on a small group of highly productive IT-Manufacturing industries in Japan during the long-lived (and still ongoing) slump. We have uncovered significant positive effects on the growth rate of labor productivity in the Semiconductors industry that are of the order of 3.81 average p.p. during 1997-2008 (or 0.38 per 1 p.p. of increase in the offshoring index). In addition, we have obtained positive and large, yet marginally significant effects, for the Electronic Parts industry (8.87 average p.p. and 2.86 per 1 p.p. of increase in offshoring). Finally, the effects in the Electronic Equipment industry were statistically not significant and those in Industry Machinery were non-existent. Despite the lack of uniformity in the results, our analysis points to the importance of offshoring strategies for some industries, precisely at a time where they most need it, not only because of the slump, but also because of the increased competition of neighboring countries.

References Agnese, P. (2012). Employment Effects of Offshoring across Sectors and Occupations in Japan. Asian Economic Journal 26 (4), 289-311. Amiti, M., Wei, S-J. (2009). Service Offshoring and Productivity: Evidence from the US. World Economy 32, 203-220. Browne, J., McCarthy, J. (2007). Japan's Offshore Evolution: Baby steps toward China and India, Forrester Report. Caballero, R., Hoshi, T., Kashyap, A. (2008). Zombie Lending and Depressed Restructuring in Japan. American Economic Review 98 (5), 1943-77. Feenstra, R., Hanson, G. (1996). Globalization, Outsourcing, and Wage Inequality. American Economic Review 86 (2), 240-245. Fukao, K., Kwon, H. (2006). Why did Japan's TFP Growth Slow Down in the Lost Decade? An Empirical Analysis Based on Firm-Level data of Manufacturing Firms. Japanese Economic Review 57 (2), 195-228. Fukao, K., Arai, S. (2013). Offshoring Bias in Japan's Manufacturing Sector. RIETI Discussion Paper 13-E-002. Girma, S., Görg, H. (2004). Outsourcing, Foreign Ownership, and Productivity: Evidence from UK Establishment-Level Data. Review of International Economics 12 (5), 817-832. Hayashi, F., Prescott, E. (2002). The 1990s in Japan: A Lost Decade. Review of Economic Dynamics 5, 206-235. Hijzen, A., Tomohiko, I., Todo, Y. (2010). Does Offshoring Pay? FirmLevel Evidence from Japan. Economic Inquiry 48 (4), 880-895.

Offshoring and Productivity from a Time-Series Perspective

157

Ito, K. Tanaka, K. (2010). Does Material and Service Offshoring Improve Domestic Productivity? Evidence from Japanese Manufacturing Industries. RIETI Discussion Paper 10-E-010. Johansen, S. (1988). Statistical analysis of cointegrating vectors. Journal of Economic Dynamics and Control 12, 231-254. Krugman, P., 1998, It's back: Japan's slump and the return of the liquidity trap, Brooking Papers on Economic Activity 2, 137--187. OECD Information Technology Outlook 2010, OECD Publishing. Pesaran, M., (1997). The Role of Economic Theory in Modelling the Long-Run. Economic Journal 107 (440), 178-191. Pesaran, M., Shin, Y. (1999). An Autoregressive Distributed-Lag Modelling Approach to Cointegration Analysis, 371-413, in Strom, S. (ed.), Econometrics and Economic Theory in the Twentieth Century: The Ragnar Frisch Centennial Symposium, Cambridge University Press, Cambridge. Pesaran, M., Shin, Y. Smith, R.J. (2001). Bounds Testing Approaches to the Analysis of Level Relationships. Journal of Applied Econometrics 16 (3), 289-326. Wakasugi, R. (1988). Research and Development and Innovations in High Technology Industry: The Case of the semiconductor industry. The Japanese Economy 17 (1), 3-35. Yamada, B, (1990). Internationalization Strategies of Japanese Electronics Companies: Implications for Asian Newly Industrializing Economies (NIEs). Working Paper 28 OECD Development Centre.

CHAPTER EIGHT SECTORIAL DIFFERENCES IN EXPORT PERFORMANCE: THE ROLE OF THE DECISION-MAKER’S GENDER* HELENA MARQUES** The heterogeneous firms literature has proposed that when there are fixed costs for exporting, only the most productive firms export (extensive margin). For those firms, export intensity changes with the variable costs of exporting. On the other hand, the literature on gender, entrepreneurship and firm performance has been finding that women are less likely to become entrepreneurs and, after having entered the market, firms where one of the owners is a woman show a weaker performance. Putting together both strands of the research, the role of the firm’s decision-maker’s gender in determining sectoral differences in export performance is investigated using a sample of exporting and non-exporting firms provided by the World Bank’s Enterprise Surveys database. The gender dummy is defined for firms where women are either the firm’s top manager or the firm’s sole owner. It is found that there are significant differences between firms managed and owned by either gender, but these differences depend on firm size and the sector in which the firm operates. Moreover, the gender differences across sectors are not always negative. After controlling for firm size and sector of activity, women present a greater gap in the amount of sales they allocate to foreign markets (export intensity) rather than in deciding to start exporting, so the variable costs of exporting may be a greater constraint relative to the fixed costs. *

This research was carried out using data from the Enterprise Surveys (http://www.enterprisesurveys.org) collected by The World Bank. Thanks are due to The World Bank for having made the data freely available for research. The usual disclaimer applies. ** Department of Applied Economics. University of the Balearic Islands, Palma de Mallorca, Spain. E-mail: [email protected]

160

Chapter Eight

8.1. Introduction The empirical observation is that not all firms export has led to important applied work, such as the work of Bernard and Jensen (2004) for the United States; Roberts and Tybout (1997) for Colombia; Aitken et al. (1997) for Mexico; and Clerides et al. (1998) for Colombia, Mexico and Morocco, among many others. This applied work actually preceded the theoretical work of, among others, Melitz (2003), Yeaple (2005), and Bernard et al. (2007), who proposed that, when there are fixed costs of exporting, only the most productive firms start to export (extensive margin). For those firms, export intensity (intensive margin) changes with the variable costs of exporting (Manova, 2013).1 In a completely unrelated way, the literature on gender, entrepreneurship and firm performance has been finding a gender effect that is unfavourable to women: at market entry, women are less likely to become entrepreneurs (e.g., Koellinger et al., 2013); after having entered the market, firms where one of the owners is a woman show a weaker performance (e.g., Bardasi et al., 2011). This literature suggests that most firms owned by women are young and small, and have more difficulty in obtaining credit (e.g., Amin, 2010). In addition, women may self-select into routine sectors with lower mean productivity (e.g., Bardasi et al., 2011). The firm’s small size (which hinders productivity and the ability to obtain credit) and the sector’s high routine level (which is linked to lower mean productivity) are two channels through which export propensity and intensity could be negatively affected by gender. Although both effects are suggested by previous research, neither of them has been explicitly studied before. Additionally, the gender and firm performance literature commonly defines the gender dummy for firms where one of the owners is a woman. Although it has been recognized that partial ownership (and participation in management teams) does not signify decision-making power, little has been done to correct this caveat. Part of the problem has been the lack of availability of information on the gender of the main decision-maker in the firm: the top manager, in larger firms with multiple ownership; or the owner, in smaller firms where there may be only one owner who is also the manager (see discussion in Aterido and Hallward-Driemeier, 2011). Taking into account the contributions from both strands of research, this chapter investigates to what extent there are gender differences in the exporting activities (propensity and intensity) of firms of different sizes 1

For comprehensive surveys on the exporting and investment decisions of heterogeneous firms see Helpman (2006), Greenaway and Kneller (2007), Bernard et al (2012) and Melitz and Redding (2014).

Sectorial Differences in Export Performance

161

operating in different sectors in a large number of countries worldwide. This is done by taking a decision-making perspective on the gender issue in management and ownership, and defining the gender dummy for firms where women are either the firm’s top manager or the firm’s sole owner. This novel definition is allowed because of the availability of information on the gender of the manager in the World Bank’s Enterprise Surveys database since 2008.2 The exporting and non-exporting firms sampled in the database have already made their market entry, so those women who are the sole owners of a firm have already decided to become entrepreneurs. The first stage decision faced by the firms studied is whether to enter foreign markets by exporting (propensity to export). If they decide to export, in a second stage, they face a decision regarding how much of their sales are destined to foreign markets (export intensity). Two equations are estimated, one for the propensity to export, and the other for export intensity. It is found that there are significant differences at several levels between firms managed and owned by either gender. Size effects are found in all cases. There are gender effects at the sector level, but these do not follow a clear sign pattern. In what follows, Section 8.2 examines the role of firm size, sector characteristics and gender in determining export propensity and intensity as suggested by previous research. Section 8.3 describes the main features of the World Bank’s Enterprise Surveys data used in the empirical analysis of the following section. Section 8.4 investigates the existence of gender effects in exporting across sectors and firm sizes. Section 8.5 concludes.

8.2. Explaining Differences in Export Propensity and Intensity The heterogeneous firms literature states that, when there are fixed costs of exporting, the decision to export is primarily dependent on the firm’s productivity and in turn the firm’s productivity level is correlated to its size (Manova, 2013). However, once a firm has become an exporter, the share of sales destined to foreign markets is primarily dependent upon the variable costs it may face in a given market. As a consequence, the firm’s size is not expected to play a role in determining export intensity, but it is expected to be a determinant of the propensity to export. 2

A related issue is whether women are part of the top management team and actually have decision-making power. Unfortunately, the World Bank’s Enterprise Surveys database does not contain information on who holds decision-making power. All that can be done is to infer such power through the figures of the top manager or of the sole owner.

162

Chapter Eight

On the other hand, the gender and firm performance literature states that the firm’s productivity level is influenced by gender and that firms owned by women are smaller on average (see, for example, the discussion in Bardasi et al., 2011). Hence productivity can be approached by using the information on its correlation with firm size. Although this is conceptually less satisfactory than directly obtaining the firm’s productivity level, when data is scarce (particularly capital stock data), the use of firm size (highly correlated with productivity) preserves sample size and estimates can be obtained for a large number of developed and developing countries. Moreover, smaller firms have more difficulty in obtaining credit (Amin, 2010) and productivity is constrained by the availability of internal finance (Chen and Guariglia, 2013). As a consequence, smaller firms may be less productive on average and also less likely to export (e.g., Bernard and Jensen, 2004; Roberts and Tybout, 1997; Aitken et al., 1997; Clerides et al., 1998). Most firms owned by women are small firms (Bardasi et al., 2011; Aterido and Hallward-Driemeier, 2011). Thus gender differences may arise through a size effect. Besides, export propensity and export intensity may both be affected by the variable costs of exporting, but fixed costs should affect only the decision to start exporting, not how much is exported. Hence size effects should determine the propensity to export, not export intensity. But if access to credit is an issue, size may operate indirectly by facilitating or hindering it. In this case, difficulty in financing fixed costs would lower the propensity to export, but difficulty in financing variable costs would lower both export propensity and intensity. Furthermore, productivity and firm size differ according to the sector’s degree of routineness (Costinot et al., 2011). High-routine sectors present lower mean productivity, as well as a larger share of small firms, so operating in that group might hinder exporting. If women are overrepresented in high-routine sectors, the export activity of firms they own may also be hindered.

8.3. What Can Be Learned from Worldwide Data The data used in testing the explanations provided in the previous section come exclusively from the Enterprise Surveys, an initiative of the World Bank that contains country surveys on firm-level data stratified for various manufacturing and services sectors (and on a second level also for firm size). Thus the data are representative of the economic structure of each country. As the surveys are harmonized across countries, there is international comparability and the data for various countries can be

Sectorial Differences in Export Performance

163

pooled together. Here, two main waves are used, that of 2006-07 and that of 2009-10. Those years represent respectively 35.23% and 47.69% of the observations available, although for a few countries, surveys may have been conducted one year in advance or with a one year delay. Table 8-1 details the countries and sectors available in the sample. The number of firms available for analysis varies across countries and sectors, but it is larger than 300 for each of the 12 manufacturing and services sectors. Overall, there is a worldwide representation of over 30 countries per sector, totaling at least 100 countries when all sectors are pooled together. Table 8-2 details the construction of the variables used in the analysis. The first explanatory variable is the gender of the firm’s main decisionmaker (the top manager or the sole owner). Information on the gender of the top manager (femCEO) is available only from 2008. However, use of data prior to 2008 is made by constructing a second gender variable, female sole owner (femSO), which takes value 1 when two conditions are met: (i) a firm has one sole owner (the answer to the survey question “How many owners does the firm have?” is equal to “1”); (ii) one of the firm’s owners is female (the answer to the survey question “Is any of the firm’s owners female?” is equal to “Yes”). By crossing the information contained in these two questions it is possible to identify those firms that have one sole owner who is a woman. In this way, the two gender variables constructed here represent only those management and ownership cases where a woman is the main decision-maker. Most of the previous work has failed to implement this point, although that failure has been acknowledged as a caveat, mostly due to lack of data. The other explanatory variable is the firm’s size (size) measured by the number of permanent, full-time employees of the firm in the survey year: small (between 5 and 19 workers), medium (between 20 and 99 workers), and large (more than 100 workers). Finally, the analysis is carried out separately for each sector classified by the degree of sector routineness (routine) as indicated by Costinot et al. (2011): low routine (Electronics, Chemicals & Pharmaceuticals), medium routine (Metals & Machinery, Non-Metallic & Plastic Materials, Other Manufacturing, Other Services), and high routine (Textiles, Garments, Food, Retail & Wholesale Trade, Hotels & Restaurants, Construction & Transportation).

164

Chapter Eight

Table 8-1: Survey Countries and Sectors COUNTRIES Afghanistan, Albania, Algeria, American Samoa, Andorra, Angola, Antigua and Barbuda, Argentina, Armenia, Aruba, Australia, Austria, Azerbaijan , Bahamas, Bahrain, Bangladesh, Barbados, Belarus , Belgium, Belize, Benin, Bermuda, Bhutan, Bolivia, Bosnia and Herzegovina , Botswana, Brazil , Brunei Darussalam, Bulgaria , Burkina Faso, Burundi, Cambodia, Cameroon, Canada, Cape Verde, Cayman Islands, Central African Republic, Chad, Channel Islands, Chile, China, Colombia, Comoros, Congo , Costa Rica, Croatia, Cuba, Curacao, Cyprus, Czech Republic , Democratic Republic of Congo, Democratic People's Republic of Korea, Denmark, Djibouti, Dominica, Dominican Republic, Ecuador, Egypt, El Salvador, Equatorial Guinea, Eritrea, Estonia , Ethiopia, Faeroe Islands, Falkland Islands (Malvinas), Fiji, Finland, France, French Guiana, French Polynesia, FYR Macedonia, Gabon , Gambia, Georgia, Germany, Ghana, Gibraltar, Greece, Greenland, Grenada, Guadeloupe, Guam, Guatemala, Guinea, Guinea Bissau, Guyana, Haiti, Honduras, Hong Kong, Hungary , Iceland, India, Indonesia, Iran, Iraq, Ireland, Isle of Man, Israel, Italy, Ivory Coast, Jamaica, Japan, Jordan, Kazakhstan , Kenya, Kiribati, Kosovo , Kuwait, Kyrgyz Republic , Lao PDR, Latvia , Lebanon, Lesotho, Liberia, Libya, Liechtenstein, Lithuania , Luxembourg, Macao, Madagascar, Malawi, Malaysia, Maldives, Mali , Malta, Marshall Islands, Mauritania, Mauritius, Mexico, Micronesia, Moldova , Monaco, Mongolia , Montenegro , Morocco, Mozambique, Myanmar, Namibia, Nepal, Netherlands, Netherlands Antilles, New Caledonia, New Zealand, Nicaragua, Niger, Nigeria, Northern Mariana Islands, Norway, Oman, Pakistan, Palau, Panama, Papua New Guinea, Paraguay, Peru, Philippines, Poland , Portugal, Puerto Rico, Qatar, Republic of Korea, Reunion, Romania , Russia , Rwanda, Saint Helena, Saint Kitts and Nevis, Saint Pierre and Miquelon, Samoa, San Marino, Sao Tome and Principe, Saudi Arabia, Senegal, Serbia , Seychelles, Sierra Leone , Singapore, Sint Maarten (Dutch part), Slovak Republic, Slovenia, Solomon Islands, Somalia, South Sudan, South Africa, Spain, Sri Lanka , St. Martin (French part), St Kitts and Nevis, St Lucia, St Vincent and Grenadines, Sudan, Suriname, Swaziland, Sweden, Switzerland, Syrian Arab Republic, Taiwan, Tajikistan , Tanzania, Thailand, Timor Leste, Togo, Tonga, Trinidad and Tobago, Tunisia, Turkey , Turkmenistan, Turks and Caicos Islands, Tuvalu, Uganda, Ukraine , United Arab Emirates, United Kingdom, United Republic of Tanzania, United States, Uruguay, Uzbekistan, Vanuatu, Venezuela, Vietnam, Western Sahara, Yemen, Yugoslavia, Zambia, Zimbabwe

SECTORS Textiles, Garments, Food, Metals & machinery, Electronics, Chemicals & pharmaceuticals, Non-metallic & plastic materials, Other manufacturing, Retail & wholesale trade, Hotels & restaurants, Other services, Construction & transportation

The dependent variables are export intensity (expint), measured by the percentage of the establishment’s annual sales that were direct exports in the survey year, and an indicator of being an exporter (exporter) that takes a value 1 if the value of export intensity is positive. Table 8-3 shows the means differences tests by gender of the decisionmaker for the variables used. In the sample, a woman is the top manager of 16% of firms and the sole owner of 10% of them. All the variables present significant gender differences both by management and by ownership. Women owners and managers are overrepresented in the group of small

Sectorial Differences in Export Performance

165

firms (61.91% and 55.06% of small firms are respectively owned or managed by a woman) and underrepresented in the group of large firms (11.08% and 14.19% of small firms, respectively). Moreover, a lower proportion of female owned and managed firms export: 12.07% and 13.17% respectively, compared to over 17% for male owners and managers. Export intensity is lower as well: 4.46% and 5.61% respectively, compared to more than 7% for male owners and managers. Table 8-2: Description of Variables IDENTIFICATOR femCEO femSO

size

expint exporter

routine

DESCRIPTION Dummy taking value 1 if the firm’s top manager is female (included in the surveys from 2008) Dummy taking value 1 if any of the firm’s owners is female and the firm has one sole owner (percentage of the firm held by the largest owner is 100) Dummy variable(s) for the number of permanent, full-time employees of the firm in the survey year: small (between 5 and 19 workers), medium (between 20 and 99 workers), large (more than 100 workers) Export intensity (percentage of the establishment’s annual sales that were direct exports in the survey year) Dummy variable taking value 1 if expint is positive Dummy variable(s) for the degree of sector routineness as indicated by Costinot et al (2011): low routine (electronics, chemicals and pharmaceuticals), medium routine (metals and machinery, non-metallic and plastic materials, other manufacturing, other services), high routine (textiles, leather, garments, food, retail and wholesale trade, hotels and restaurants, construction and transportation)

Table 8-4 shows the sample distribution of gender, exporting, and firm size across routine-differentiated sectors. Small firms represent 40-60% of the firms in high and medium routine sectors, but in low routine sectors medium-sized firms predominate and about a third of the firms are large. The share of exporters is 50% higher in low routine, compared to high and medium routine sectors, although mean export intensities are very similar across routine types, evidencing the distinct behaviour of the extensive and intensive margins. With respect to gender, high routine sectors have the highest shares of firms managed (20-30%) or owned (12-18%) by women. In conclusion, women owners and managers are overrepresented in high routine sectors, which are also those with the largest share of small firms, but are not necessarily less export-oriented. This is especially true for the Garments sector, where 25% of the firms export an average of 15% of their sales (this is only surpassed by Electronics, which has an export

166

Chapter Eight

intensity of 18%). Therefore, as highlighted also by previous research (e.g., Costinot et al., 2011; Aterido and Hallward-Driemeier, 2011; Koellinger et al., 2013), the choice of sector may determine the firm’s behaviour.

8.4. Gender Effects across Sectors and Firm Sizes The descriptive analysis of the sample data revealed that women are overrepresented in high and medium routine sectors, where 40-60% of the firms are small firms. In this section, those differences are explored in greater detail, carrying out routine-differentiated sector-level regressions for the propensity to export, measured by the probability of firm ݅in country ݆ and sector ‫ ݏ‬becoming an exporter ൫ܲ‫ݎ‬ሺܺ௜ ൌ ͳሻ൯, and for the intensity of exports, measured by the percentage of sales exported (‫ܫ‬௜ ሻ. The equations to estimate for each sector are written under the generic form: ‫ݎܸܽ݌݁ܦ‬௜ ൌ ߙ଴ ൅ ߙଵ ‫ݎ݁݀݊݁ܩ‬௜ ൅ σ௞ୀଶǡଷ ߚ௞ ܵ݅‫݁ݖ‬௞௜ ൅ σ௞ୀଶǡଷ ߛ௞ ‫ݎ݁݀݊݁ܩ‬௞௜ ܵ݅‫݁ݖ‬௞௜ ൅ ߜ௝ ൅ ‫ݑ‬௜ where firm size comprises three categories3 and ߜ௝ represents country dummies. The propensity to export equation is estimated by a Probit model where ‫ݎܸܽ݌݁ܦ‬௜ ൌ ߩ௜ ൌ Ȱିଵ ൫ܲ‫ݎ‬ሺܺ௜ ൌ ͳሻ൯, and ܺ௜ is a dummy taking a value of 1 if firm ݅ exports and 0 otherwise. The estimation results are presented in Table 8-5A for management and Table 8-5B for ownership. The export intensity equation is estimated by a Tobit model where ‫ݎܸܽ݌݁ܦ‬௜ ൌ ‫ܫ‬௜ is numerical and continuous but censored between 0 and 1. The estimation results are presented in Table 8-6A for management and Table 8-6B for ownership. On average, firm size and the decision-maker’s gender explain about 20% of the variation in the propensity to export and the export intensity, going up to about 60% in the Food sector. In all cases, the size effect is negative for small firms, and positive for medium firms (firms with more than 20 employees) and especially for large firms (firms with more than 100 employees). So, as highlighted by the heterogeneous firms’ literature, medium and especially large firms are more likely to export. They also 3 The size dummy comprises three categories as given in Enterprise Surveys data: small if having 5-19 workers (base group), medium (20-99 workers), large (over 100 workers). In the regressions the omitted reference category is the group of small firms, so that the constant represents the size effect in small firms and the gender dummy at the intercept represents the gender effect in small firms.

Sectorial Differences in Export Performance

167

export a higher percentage of their total sales, which can reflect the indirect effect of the difficulty of small firms in financing the variable costs of exporting. This result is independent of the decision-maker’s gender. Within each size category, gender effects are sector-specific, although there is no pattern related to the degree of routineness. Thus, export activity in the Food sector and, to a lesser extent, in the Metals & Machinery sector, shows a negative gender effect in small firms but this effect becomes positive in medium and large firms. The opposite happens in the Hotels & Restaurants sector and, to a lesser extent, in the Electronics sector. These results are found both in management (Tables 85A and 8-6A) and in ownership (Tables 8-5B and 8-6B). Other sectors present gender effects that are more specific to each firm size, decision-making and activity type. Thus, the Chemicals & Pharmaceuticals sector shows a negative (positive) gender effect of small (medium) firms’ management on export intensity (Table 8-6A) and a positive (negative) gender effect of small (medium and large) firms’ ownership on both types of export activity (Tables 8-5B and 8-6B). The Construction & Transport sector shows a negative gender effect of small firms’ ownership on the propensity to export (Table 8-5B), but also a positive (negative) gender effect of small and medium (large) firms’ management and a negative (positive) gender effect of small and medium (large) firms’ ownership on export intensity (Tables 8-6A and 8-6B). In the Retail & Wholesale sector, there is a negative gender effect of large firms’ ownership on their propensity to export (Table 8-5B) and of small firms’ management on their export intensity (Table 8-6A). For other sectors, gender effects are present only in export intensity and exhibit non-linearities with respect to firm size (Tables 8-6A and 86B).4 This is the case of the Garments sector, which presents a negative gender effect in both management and ownership, and of the Non-Metallic & Plastic Materials sector, with a positive gender effect in management but a negative gender effect in ownership.

4

The existence of non-linearities in sectorial behavior was studied in the context of the New Economic Geography literature by, among others, Forslid et al (2002) and Marques (2005). Empirical surveys of this literature are provided by Head and Mayer (2004) and Overman et al (2003).

Chapter Eight

N

Factor

% %

7.04

N N

7,075

% %

5.61

Chi2

5.32***

Mean t-test

N N

57,489

%

7.14

% N

6,345

N

Is the sole owner female? No Yes % %

4.46

Chi2

9.67***

Meant-test

exporter 36,434 18.43 7,075 13.17 112.93*** 57,489 17.83 6,345 12.07 132.63*** size 36,583 7,105 431.46*** 57,871 6,362 531.34*** small 42.71 55.06 47.42 61.91 medium 34.58 30.75 32.99 27.00 large 22.72 14.19 19.59 11.08 routine 36,571 7,100 650.57*** 57,846 6,360 415.29*** Notes: The null hypothesis of the means t-test for numerical variables equals to zero the difference between the mean for the “No” (male) subsample and the mean for the “Yes” (female) subsample. The Pearson Chi2 test constructed from the tabulations of factor variables is based on sums of squares.

36,434

N

Is the top manager female? No Yes

expint

Numerical

Variables

Table 8-3: Mean Differences by Top Manager and Sole Owner Gender

168

Female top manager

Gender Female sole owner

Exporting Export Exporters intensity (mean) Small firms

Medium firms

Firm size Large firms

169

High routine 19.98 11.93 13.88 6.20 51.09 31.19 17.72 Textiles 18.26 6.64 30.06 14.34 39.50 33.38 27.11 Garments 31.66 18.47 25.43 15.81 48.03 30.09 21.88 Food 15.27 10.48 20.92 8.10 40.65 36.29 23.06 Retail & wholesale 21.47 12.63 6.37 1.78 60.64 27.44 11.92 Hotels & restaurants 26.08 17.22 2.92 1.31 57.19 31.38 11.43 Construction & transport 8.22 8.08 6.64 2.48 43.08 35.52 21.40 Medium routine 10.48 7.24 19.19 7.33 48.97 32.45 18.58 Metals & machinery 7.62 4.55 28.77 9.94 40.35 36.92 22.74 Non-metallic & plastic 11.12 5.83 21.91 7.40 40.01 36.38 23.61 materials Other manufacturing 12.48 7.91 19.74 8.31 54.99 30.64 14.37 Other services 9.80 9.88 11.14 4.28 50.87 29.95 19.18 Low routine 15.27 6.17 30.50 9.37 33.26 39.29 27.45 Electronics 12.06 4.07 32.19 18.07 27.96 39.42 32.62 Chemicals & pharmaceuticals 16.19 6.65 30.11 7.38 34.48 39.26 26.26 TOTAL 16.26 9.90 16.78 6.79 49.28 32.12 18.60 Notes: Export intensity is defined as the share of sales that are directly exported. Small, medium and large firms are those having 5-19 workers, 2099 workers and more than 100 workers, respectively. The degree of sector routineness follows the classification in Costinot et al (2011). A more detailed definition of variables can be found on Table8- 2.

Sectors

Table 8-4: Gender, Exporting and Firm Size by Sector (Sample %)

Sectorial Differences in Export Performance

Chapter Eight

High routine Retail & Hotels & Construction & Textiles Garments Food wholesale restaurants transport -1.908*** -1.401*** -2.655*** -2.124*** -3.784*** -2.368*** Constant (0.654) (0.253) (0.585) (0.216) (0.505) (0.434) -0.0759 -0.0983 -0.650** -0.231 1.382*** 0.114 GENDER (0.195) (0.221) (0.277) (0.152) (0.418) (0.440) 1.555*** 1.545*** 0.680** 0.553*** 1.748*** 0.331 Medium firm (0.210) (0.276) (0.291) (0.121) (0.365) (0.302) 0.438 -0.400 0.732* 0.260 -2.045*** 0.113 X GENDER (0.489) (0.345) (0.427) (0.239) (0.713) (0.576) 2.465*** 2.353*** 1.789*** 0.784*** 1.654*** 0.0630 Large firm (0.281) (0.384) (0.361) (0.164) (0.402) (0.323) -0.576 -0.471 0.747* -0.219 -2.544*** -0.206 X GENDER (0.515) (0.566) (0.433) (0.306) (0.737) (0.553) 62 48 69 70 30 35 Countries 246.45*** 231.73*** 367.92*** 408.33*** 61.66*** 80.28*** Joint Chi2 1,816 2,143 4,029 9,318 644 1,031 Observations 688.3*** 531.6*** 749.5*** 451.1*** 101.6*** 95.75*** Model Chi2 -16301 -22386 -25608 -81064 -4730 -14490 Log pseudo ll 0.471 0.387 0.639 0.174 0.238 0.118 Pseudo R2 Notes: Probit regression with country dummies. Dependent variable: exporter dummy (1 if exporter is defined as having positive export intensity). Regressors: (1) female top manager dummy (1 if the firm’s manager is female), indicated by GENDER; (2) size dummy (small if having 5-19 workers (base group), medium (20-99 workers), large (over 100 workers)). Robust standard errors indicated in parentheses. Weighted regression using survey sampling weights. The same results are obtained by a similarly weighted probit regression with standard errors clustered by survey sampling stratification sectors. *** p