Sustainable Agriculture and Agribusiness in Iran [1st ed.] 978-981-13-6282-8, 978-981-13-6283-5

Table of contents :
Front Matter ....Pages i-xvii
Introduction to Sustainable Agriculture and Agribusiness in Iran (Masoomeh Rashidghalam)....Pages 1-7
Front Matter ....Pages 9-9
Agricultural Risk Management Through Weather-Based Insurance in Iran (Esmaeil Pishbahar, Sahar Abedi, Ghader Dashti, Ali KianiRad)....Pages 11-28
Determining Conditional Value at Risk (CVaR) for Food Industry Companies’ Stocks Portfolios in the Tehran Stock Market (Sahar Abedi, Esmaeil Pishbahar)....Pages 29-45
Assessing Climate Change Impacts on Land Use in Iran: The Spatial Fractional Multinomial Logit Modeling Approach (Khadijeh Alefi, Mohammad Ghahremanzadeh)....Pages 47-59
Front Matter ....Pages 61-61
Estimating the Non-use Values and Related Compensative Surplus of Arasbaran Forests in Iran: An Application of the Choice Experiment Method (Maryam Haghjou, Babollah Hayati, Esmaeil Pishbahar, Morteza Molaei)....Pages 63-77
Optimized Reservoir Management for Meeting Conflicting Stakeholder Preferences: Methodological Innovations with Evidence from Iran (Omid Zamani, Hemen Nader, Masoomeh Rashidghalam, Frank A. Ward)....Pages 79-100
Willingness to Pay for IPM: An Application of the Heckman-Copula Approach (Esmaeil Pishbahar, Javad Hosseinzad, Sahar Abedi, Pariya Bageri)....Pages 101-114
Front Matter ....Pages 115-115
Testing the Law of One Price Under Nonlinearity: An Application to Iranian Broiler and Egg Markets (Mohammad Ghahremanzadeh, Fatemeh Faryadi, Ghader Dashti)....Pages 117-133
Assessing the Relationship Between Marketing Mix and Customer Satisfaction: Evidence from Iranian Dairy Companies (Esmaeil Pishbahar, Roya Ferdowsi, Babollah Hayati)....Pages 135-149
Testing for Neutrality and Super-neutrality of Money: Evidence from Iran’s Agricultural Sector (Esmaeil Pishbahar, Zahra Rasouli)....Pages 151-161
Effects of Oil Prices and Exchange Rates on Imported Inputs’ Prices for the Livestock and Poultry Industry in Iran (Esmaeil Pishbahar, Parisa Pakrooh, Mohammad Ghahremanzadeh)....Pages 163-182
Back Matter ....Pages 183-187

Citation preview

Perspectives on Development in the Middle East and North Africa (MENA) Region

Masoomeh Rashidghalam   Editor

Sustainable Agriculture and Agribusiness in Iran

Perspectives on Development in the Middle East and North Africa (MENA) Region Series editor Almas Heshmati, Sogang University, Seoul, Korea (Republic of)

This book series publishes monographs and edited volumes devoted to studies on the political, economic and social developments of the Middle East and North Africa (MENA). Volumes cover in-depth analyses of individual countries, regions, cases and comparative studies, and they include both a specific and a general focus on the latest advances of the various aspects of development. It provides a platform for researchers globally to carry out rigorous economic, social and political analyses, to promote, share, and discuss current quantitative and analytical work on issues, findings and perspectives in various areas of economics and development of the MENA region. Perspectives on Development in the Middle East and North Africa (MENA) Region allows for a deeper appreciation of the various past, present, and future issues around MENA’s development with high quality, peer reviewed contributions. The topics may include, but not limited to: economics and business, natural resources, governance, politics, security and international relations, gender, culture, religion and society, economics and social development, reconstruction, and Jewish, Islamic, Arab, Iranian, Israeli, Kurdish and Turkish studies. Volumes published in the series will be important reading offering an original approach along theoretical lines supported empirically for researchers and students, as well as consultants and policy makers, interested in the development of the MENA region.

More information about this series at http://www.springer.com/series/13870

Masoomeh Rashidghalam Editor

Sustainable Agriculture and Agribusiness in Iran

123

Editor Masoomeh Rashidghalam Department of Agricultural Economics University of Tabriz Tabriz, Iran

ISSN 2520-1239 ISSN 2520-1247 (electronic) Perspectives on Development in the Middle East and North Africa (MENA) Region ISBN 978-981-13-6282-8 ISBN 978-981-13-6283-5 (eBook) https://doi.org/10.1007/978-981-13-6283-5 Library of Congress Control Number: 2019930572 © Springer Nature Singapore Pte Ltd. 2019 This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, express or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd. The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721, Singapore

To my parents, Mohammad and Zari

Contents

1

Introduction to Sustainable Agriculture and Agribusiness in Iran . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Masoomeh Rashidghalam

Part I 2

3

4

5

6

Risk Management and Climate Change

Agricultural Risk Management Through Weather-Based Insurance in Iran . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Esmaeil Pishbahar, Sahar Abedi, Ghader Dashti and Ali KianiRad Determining Conditional Value at Risk (CVaR) for Food Industry Companies’ Stocks Portfolios in the Tehran Stock Market . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Sahar Abedi and Esmaeil Pishbahar Assessing Climate Change Impacts on Land Use in Iran: The Spatial Fractional Multinomial Logit Modeling Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Khadijeh Alefi and Mohammad Ghahremanzadeh

Part II

1

11

29

47

Natural Resources and Environmental Economics

Estimating the Non-use Values and Related Compensative Surplus of Arasbaran Forests in Iran: An Application of the Choice Experiment Method . . . . . . . . . . . . . . . . . . . . . . . . . Maryam Haghjou, Babollah Hayati, Esmaeil Pishbahar and Morteza Molaei

63

Optimized Reservoir Management for Meeting Conflicting Stakeholder Preferences: Methodological Innovations with Evidence from Iran . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Omid Zamani, Hemen Nader, Masoomeh Rashidghalam and Frank A. Ward

79

vii

viii

7

Contents

Willingness to Pay for IPM: An Application of the HeckmanCopula Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 Esmaeil Pishbahar, Javad Hosseinzad, Sahar Abedi and Pariya Bageri

Part III

Agricultural Prices and Commodity Market Analysis

8

Testing the Law of One Price Under Nonlinearity: An Application to Iranian Broiler and Egg Markets . . . . . . . . . . . 117 Mohammad Ghahremanzadeh, Fatemeh Faryadi and Ghader Dashti

9

Assessing the Relationship Between Marketing Mix and Customer Satisfaction: Evidence from Iranian Dairy Companies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135 Esmaeil Pishbahar, Roya Ferdowsi and Babollah Hayati

10 Testing for Neutrality and Super-neutrality of Money: Evidence from Iran’s Agricultural Sector . . . . . . . . . . . . . . . . . . . . 151 Esmaeil Pishbahar and Zahra Rasouli 11 Effects of Oil Prices and Exchange Rates on Imported Inputs’ Prices for the Livestock and Poultry Industry in Iran . . . . . . . . . . 163 Esmaeil Pishbahar, Parisa Pakrooh and Mohammad Ghahremanzadeh Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183

Editor and Contributors

About the Editor Masoomeh Rashidghalam is Researcher at University of Tabriz. She did her B.Sc. and Ph.D. in Department of Agricultural Economics at University of Tabriz and holds a M.Sc. from Tarbiat Modares University. She was a visiting scholar in Sweden and South Korea (in Department of Economics at Jönköping International Business School (JIBS) and Department of Economics at Sogang University). She was an assistant in organization of “ADB Workshop on Urbanization in Asia,” Seoul. Dr. Rashidghalam’s areas of expertise are agricultural production economics, farm management, poverty and inequality, child well-being, labor economics, urbanization, and technical efficiency measurement. She has a wide range of teaching experience in econometrics, agricultural production economics, and microeconomics. She has written a book: Measurement and Analysis of Performance of Industrial Crop Production: The Case of Iran’s Cotton and Sugar Beet Production, 2018, published by Springer. She has contributed different chapters to six books.

Contributors Sahar Abedi Department of Agricultural Economics, University of Tabriz, Tabriz, Iran Khadijeh Alefi Department of Agricultural Economics, University of Tabriz, Tabriz, Iran Pariya Bageri Department of Agricultural Economics, University of Tabriz, Tabriz, Iran Ghader Dashti Department of Agricultural Economics, University of Tabriz, Tabriz, Iran ix

x

Editor and Contributors

Fatemeh Faryadi Department of Agricultural Economics, University of Tabriz, Tabriz, Iran Roya Ferdowsi Department of Agricultural Economics, University of Tabriz, Tabriz, Iran Mohammad Ghahremanzadeh Department University of Tabriz, Tabriz, Iran

of

Agricultural

Economics,

Maryam Haghjou Department of Agricultural Economics, University of Tabriz, Tabriz, Iran Babollah Hayati Department of Agricultural Economics, University of Tabriz, Tabriz, Iran Javad Hosseinzad Department of Agricultural Economics, University of Tabriz, Tabriz, Iran Ali KianiRad Agricultural Economics-Research, Deputy-APERDRI-Ministry of Agriculture, Tehran, Iran Morteza Molaei Department of Agricultural Economics, Urmia University, Urmia, Iran Hemen Nader Department of Agricultural Economics, University of Zabol, Zabol, Iran Parisa Pakrooh Department of Agricultural Economics, University of Tabriz, Tabriz, Iran Esmaeil Pishbahar Department of Agricultural Economics, University of Tabriz, Tabriz, Iran Masoomeh Rashidghalam Department of Agricultural Economics, University of Tabriz, Tabriz, Iran Zahra Rasouli Department of Agricultural Economics, University of Tabriz, Tabriz, Iran Frank A. Ward Department of Agricultural Economics and Agricultural Business, New Mexico State University, Las Cruces, NM, USA Omid Zamani Department of Agricultural Market Analysis, Institute of Agricultural Economics, Christian-Albrechts-University Kiel, Kiel, Germany; Leibniz-Institut für Agrartechnik und Bioökonomie e.V. (ATB), Potsdam, Germany

Abbreviations

AD ADF AHP AIC AMH ANP ARCH ARIMA ASC BDS BIC BM CCC CORE CS CVaR C-vine CVM DCC DEMATEL df DF-GLS D-vine EBA ESTAR EVT FAO FGM FIML

Associated diploma Augmented Dickey–Fuller Analytic hierarchy process Akaike’s information criterion Ali–Mikhail–Haq Analytic network process Autoregressive conditional heteroskedasticity Autoregressive integrated moving average Alternative specific constant Brock, Dechert, and Scheinkman Bayesian information criterion Block maxima Constant conditional correlation Center for Operations Research and Econometrics Compensative surplus Conditional value at risk Canonical vine Contingent valuation methods Dynamic conditional correlation Decision-making trial and evaluation laboratory Degrees of freedom Dickey–Fuller generalized least squares Drawable vine Economics and business administration Exponential smooth transition autoregressive Extreme value theory Food and Agriculture Organization Farlie–Gumbel–Morgenstern Full information maximum likelihood

xi

xii

FS GARCH GEV GIS GMM GP GPD GTL HEGY IPCC IPM JML KPSS LGP LOP LR LRD LRN LRSN LSW MCA MCDM M-GARCH MLE MRE OLS OPEC p.d.f. PCC POT SA SACF SBC SM STAR TOPSIS VAR VaR VCC WBCI WRT WTP

Abbreviations

Fisher–Seater Generalized autoregressive conditional heterogeneity Generalized extreme value Geographic information system Generalized method of moments Goal programming Generalized Pareto distribution Generalized Tukey’s lambda Hylleberg–Engle–Granger–Yoo Intergovernmental Panel on Climate Change Integrated pest management Joint maximum likelihood Kwiatkowski–Phillips–Schmidt–Shin Lexicographical goal programming Law of one price Likelihood ratio Long-run derivative Long-run neutrality Long-run superneutrality of money Lucas, Sargent, and Wallace Multi-criteria analysis Multi-criteria decision making Multivariate-GARCH Maximum likelihood estimation Macroeconomic rational expectations hypothesis Ordinary least squares Organization of the Petroleum Exporting Countries Probability density function Pair copula constructions Peak over threshold Stakeholder analysis Sample autocorrelation function Schwarz–Bayes criterion Sequential method Smooth transition autoregressive Technique for order of preference by similarity to ideal solution Vector autoregression Value at risk Variable conditional correlation Weather-based crop insurance With respect to Willingness to pay

List of Figures

Fig. Fig. Fig. Fig.

2.1 2.2 2.3 3.1

Fig. 3.2 Fig. Fig. Fig. Fig.

3.3 3.4 4.1 5.1

Fig. Fig. Fig. Fig.

6.1 6.2 6.3 6.4

Fig. 6.5

Fig. 7.1 Fig. 8.1

Fig. 9.1

Fig. 10.1

C-vine tree structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . D-vine tree structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Indemnity charts for wheat in phenology stages . . . . . . . . . . . Dynamic conditional correlations of the return series in the dairy portfolio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Dynamic conditional correlations of the return series in the sugar portfolio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C-Vine tree structure for dairy portfolio . . . . . . . . . . . . . . . . . C-Vine tree structure for sugar portfolio . . . . . . . . . . . . . . . . . Political divisions of the Iran’s counties . . . . . . . . . . . . . . . . . An example of choice card used in calculation of non-use values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Geographical position of Mahabad . . . . . . . . . . . . . . . . . . . . . General framework of the study . . . . . . . . . . . . . . . . . . . . . . . Power/interest matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Hierarchical structure in order to prioritize the use of dam’s water . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Power/interest matrix for the stakeholder groups Note DF, AO, MM, and WA stand for downstream farmers, agricultural organization, Mahabad municipalities and Water Resource Authority, respectively . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Integrated pest management components. Source Baraki (2010) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Procedure to test for the weak and the strong version of the LOP with a nonlinear price differential. Source Emmanouilides and Fousekis (2012) . . . . . . . . . . . . . . . . . . . General structure of the decision tree for prioritization of dairy companies based on customer satisfaction (Source Set by the author based on previous studies) . . . . . . . Confidence interval for bk (without including structural break in unit root test) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

.. .. ..

21 22 25

..

39

. . . .

. . . .

40 42 43 53

. . . .

. . . .

67 80 82 83

..

86

..

88

. . 102

. . 124

. . 138 . . 158 xiii

xiv

Fig. 10.2 Fig. 11.1

List of Figures

Confidence interval for bk (with including structural break in unit root test) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159 Relationship between the oil market, exchange rate market, and agriculture market . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165

List of Tables

Table 2.1 Table 2.2 Table 2.3 Table 2.4 Table 2.5 Table 2.6 Table 2.7 Table 2.8 Table 2.9 Table 3.1 Table 3.2 Table 3.3 Table 3.4 Table 3.5 Table 3.6 Table 3.7 Table 4.1 Table 4.2 Table 4.3

Denotation and properties of bivariate elliptical and Archimedean copula families . . . . . . . . . . . . . . . . . . . . . . . . Copula selection and parameter estimation result for C-vine model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Copula selection and parameter estimation result for D-vine model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Vuong and Clarke tests to compare C-vine and D-vine models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Goodness of fit for wheat yield . . . . . . . . . . . . . . . . . . . . . . Computing premium amount for rainfed wheat in Miyaneh (Rial) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Estimation of the parameter for Clayton, Gumbel, and Gaussian copulas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Amount of trigger value and stop-loss . . . . . . . . . . . . . . . . . Liability insurance (Rial) . . . . . . . . . . . . . . . . . . . . . . . . . . . Data descriptive statistics for portfolio include ‘stocks dairy companies’ (dairy portfolio) . . . . . . . . . . . . . . . . . . . . . . . . . Data descriptive statistics for portfolio include ‘stocks sugar companies’ (sugar portfolio) . . . . . . . . . . . . . . . . . . . . . . . . . DCC-GARCH estimation results . . . . . . . . . . . . . . . . . . . . . Threshold values and ML GPD parameters estimation . . . . . Copula selection and parameters’ estimation for the dairy portfolio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Copula selection and parameters’ estimation for the sugar portfolio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VaR and CVaR estimation in each portfolio . . . . . . . . . . . . Summary of estimated models and their information criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Estimates of land use transitions. . . . . . . . . . . . . . . . . . . . . . Marginal effects on land use transitions in spatial multinomial logit model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

..

15

..

20

..

22

.. ..

23 23

..

23

.. .. ..

24 24 24

..

37

.. .. ..

37 38 41

..

41

.. ..

42 43

.. ..

54 55

..

57 xv

xvi

List of Tables

Table 5.1 Table 5.2 Table 5.3 Table 5.4 Table 5.5 Table 5.6 Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table

6.1 6.2 6.3 6.4 6.5 6.6 6.7 6.8 6.9 6.10 6.11 7.1 7.2 7.3 7.4 7.5 7.6 8.1 8.2 8.3 8.4 8.5 8.6 8.7 8.8 8.9 9.1 9.2 9.3 9.4 10.1 10.2

Table 10.3

Attributes and attribute levels . . . . . . . . . . . . . . . . . . . . . . . . Descriptive statistics of variables . . . . . . . . . . . . . . . . . . . . . Frequency distribution of alternative specific constant (ASC) for non-use features among responses . . . . . . . . . . . . Estimation of mixed logit model for non-use values . . . . . . Willingness to pay for extracting and ranking of non-use values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Estimating monthly compensative surplus of non-use values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . TFN values. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Survey of most relevant actors for the project . . . . . . . . . . . Importance of the stakeholders . . . . . . . . . . . . . . . . . . . . . . . Local weights of criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . Local weights of sub-criteria . . . . . . . . . . . . . . . . . . . . . . . . Local weight of each alternative relevant to sub-criteria . . . . Results of combination GP and FAHP model . . . . . . . . . . . Drinking water allocated in different months . . . . . . . . . . . . Volume of reservoir . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Electricity production in different priorities (unit kW) . . . . . Volume of empty space for flood control . . . . . . . . . . . . . . . Properties of bivariate copula families . . . . . . . . . . . . . . . . . WTP averages for each environmental class . . . . . . . . . . . . . Copula selection results for Probit-t model . . . . . . . . . . . . . . Copula selection results for Logit-t model . . . . . . . . . . . . . . Estimation results for Probit-t model . . . . . . . . . . . . . . . . . . Estimation results for Logit-t model . . . . . . . . . . . . . . . . . . . DF-GLS unit root test for broiler price differentials . . . . . . . DF-GLS unit root test for egg price differentials . . . . . . . . . Luukkonen et al. (1988) linearity test for broiler . . . . . . . . . Luukkonen et al. (1988) linearity test for egg . . . . . . . . . . . Results of final model for broiler and egg . . . . . . . . . . . . . . BDS test results for broiler price differentials . . . . . . . . . . . . BDS test results for egg price differentials . . . . . . . . . . . . . . LOP test results for broiler price differentials . . . . . . . . . . . . LOP test results on egg price differentials . . . . . . . . . . . . . . Total relationship matrix for ranking of dairy companies . . . Mutual interaction matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . Weighted normalized super matrix . . . . . . . . . . . . . . . . . . . . Prioritization of companies based on TOPSIS method . . . . . LRN and LRSN restrictions . . . . . . . . . . . . . . . . . . . . . . . . . Process of selecting the restrictions and deciding about the LRN and LRSN . . . . . . . . . . . . . . . . . . . . . . . . . . Results of unit root tests . . . . . . . . . . . . . . . . . . . . . . . . . . . .

.. ..

67 70

.. ..

71 71

..

72

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

73 84 84 88 88 89 89 91 92 92 93 93 107 109 110 110 111 111 125 125 126 126 127 128 129 130 130 143 144 145 146 154

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . 156 . . 157

List of Tables

Table 10.4 Table 10.5 Table Table Table Table

11.1 11.2 11.3 11.4

Table 11.5

xvii

Results of LRN and LRSN—without considering structural breaks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Results of LRN and LRSN—with considering structural breaks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Summary of literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Result of stationary tests (1995–2014) . . . . . . . . . . . . . . . . . Result of seasonal unit root test (1995–2014) . . . . . . . . . . . . Result of ARIMA-MGARCH model for variables (1995–2014) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Result of correlation with R-Vine model (1995–2014) . . . . .

. . 158 . . . .

. . . .

158 166 175 176

. . 177 . . 178

Chapter 1

Introduction to Sustainable Agriculture and Agribusiness in Iran Masoomeh Rashidghalam

1.1 Background Seventy-eight million people in Iran live in an area of 1648,195 km2 . About 29.2% of the population lives in urban areas with an annual growth rate of 0.9% while the remaining 70.8% lives in rural areas with an annual growth rate of 1.9%. Iran’s oil and natural gas reserves are among the world’s largest, and its economy depends significantly on the extraction of these resources. Iran is primarily an agricultural country with the agriculture sector playing an important role in its economy. According to statistics, the agriculture sector accounts for about 33% of Iran’s employment and over 25% of its GDP. Iran’s total cultivated area is about 15 million hectares spread in its western, north-eastern and north-western provinces. The average rainfall is about 250 mm which is less than one-third of the mean rainfall in the world. About 84.8% of Iran is located in an arid climate zone. Currently, about 4.5% of the country is covered by forests. A time trend review of the forest cover shows that in the twentieth century about 18 million hectares were covered by forests, which went down to 12 million ha and now only 7.3 million ha is covered. Considering these statistics Iran’s agriculture suffers from various problems. There is increasing population growth, soil erosion, water crises, overuse of pesticides and chemical fertilizers, low moisture, declining groundwater resources, and destruction of resources, all of which threaten sustainable agriculture and agribusiness. Iran is a transitional society in which the agricultural sector has an important role to play in meeting society’s basic needs and also for national development, creating rural and rural employment, and food security. The country needs a dynamic and sustainable move for optimizing the available resources and increasing the quality and quantity of agricultural products with an emphasis on maintaining or enhancing the quality of these resources. M. Rashidghalam (B) Department of Agricultural Economics, University of Tabriz, Tabriz, Iran e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2019 M. Rashidghalam (ed.), Sustainable Agriculture and Agribusiness in Iran, Perspectives on Development in the Middle East and North Africa (MENA) Region, https://doi.org/10.1007/978-981-13-6283-5_1

1

2

M. Rashidghalam

Therefore, considering the importance and necessity of sustainable agriculture in Iran, this book examines different aspects of this issue in this country. This volume is a collection of ten selected empirical studies on agricultural economics and agribusiness of Iran. Nineteen authors have contributed their research to this volume. The core purpose of editing this book is to identify situation and problems of sustainability in Iran’s agricultural sector. In addition to it, the studies also talk about main challenges of Iran’s sustainable agriculture and agribusiness and suggest policy recommendations to decision makers and government agencies. In this volume, readers are also introduced to key microeconomic principles which are applied to agriculture from a suitability perspective. To the best of our knowledge this is the first work that collects research papers in different fields of Iran’s sustainable agriculture and covers wide range of topics. Therefore, a main contribution of this volume lies in applying and examining important and extensive microeconomics theory in Iran’s agriculture sector. The sustainability of agriculture and agribusiness in Iran will interest scholar and students of agricultural and environmental economics.

1.2 Summary of Individual Studies This volume is a collection of ten empirical studies on Iran’s sustainable agriculture and agribusiness. Chapters of this volume are grouped into three domains: risk management and climate change; natural resources and environmental economics; and agricultural prices and commodity market analysis. This Chapter gives a brief introduction of the book. The studies jointly provide an up-to-date situation and problems of Iran’s sustainable agriculture. In addition, they give policy recommendation to decision makers and agricultural producers. Full description of the chapters is presented following: Part A. Risk Management and Climate Change Part A covers three chapters which assess agricultural risk management and analyzes the climate change effects on agricultural production and land use. First study (Chap. 2), Agricultural Risk Management Through Weather-based Insurance in Iran, by Esmaeil Pishbahar, Sahar Abedi, Ghader Dashti, and Ali KianiRad, examines weather-based crop insurance (WBCI) for rain-fed wheat using weather variables (for example, temperature and precipitation) and yields of rain-fed wheat (“Sardari variety”) during the period 1987–2013. It measures the dependence structure between weather indices and wheat yields using the Vine copula functions; it also calculates the indemnity function and premium amounts. The results of this study show that the D-Vine model is better than the C-Vine model for describing joint distribution, and hence it uses this model to compute the premium for rainfed wheat. The premium was calculated in four levels of coverage (50, 80, 90, and 100%). The results show that its amount at the 80% coverage level was 588,320 rial. The computing premium in WBCI is less than current insurance premium which is reasonable.

1 Introduction to Sustainable Agriculture and Agribusiness in Iran

3

Second study (Chap. 3), Determining Conditional Value at Risk (CVaR) for Food Industry Companies’ Stocks Portfolios in the Tehran Stock Market, by Sahar Abedi and Esmaeil Pishbahar, determines the value at risk (VaR) and conditional value at risk (CVaR) of food industry companies’ stock portfolios on the Tehran stock market. This study applied the DCC-GHARCH-EVT model to capture the tail distribution of portfolio risks and time-varying behavior. In addition, it used the Vine copula to describe the dependence structure between dairy and sugar portfolios’ returns. The results show that the returns are dependent and that the two markets experienced extreme events. The sugar portfolio’s returns were more volatile and extreme losses were more likely than extreme rewards in this portfolio. In addition, the VaR and CVaR results also indicate that the dairy portfolio is less risky as compared to the sugar portfolio. Third study (Chap. 4), Assessing Climate Change impacts on Land-Use in Iran: The Spatial Fractional Multinomial Logit Modeling Approach, by Khadijeh Alefi and Mohammad Ghahremanzadeh, assesses the effect of climate change’s variables on the land share of annual crop groups, including cereals, legumes, vegetables, cucurbits, forages, and industrial crops in Iran. It analyzed four climatic parameters—temperature, precipitation, wind speed, and humidity. The study also estimates two spatial fractional multinomial logit models using agronomic and climate data from 336 counties for two periods, 2006–2007 and 2012–2013. The results show that the crop groups responded to climate change and this response has increased over time. Part B. Natural Resources and Environmental Economics Part B discusses natural resources and environmental economics in Iran and covers three chapters. First study (Chap. 5) Estimating the Non-use Values and Related Compensative Surplus of Arasbaran Forests in Iran: An Application of the Choice Experiment Method by Maryam Haghjou, Babollah Hayati, Esmaeil Pishbahar, and Morteza Molaei, estimates the non-use value of Arasbaran forests in Iran using the choice experiment technique. The results show that the total non-use value of Arasbaran forests is about 1704.199 billion rial (about 40.57 million USD) in which 76% is the “option” value and 17 and 7% are existence and bequest values, respectively. Moreover, the results of the estimation of compensating for the surplus non-use value shows that this would be worth about 16038182 rial (about 381.86 USD) per month. Findings of this study also show that factors like respondents’ educational levels, income, number of annual visits, and favorable attitudes toward Arasbaran forests, have positive and significant effects on their willingness to pay (WTP) for the non-use value of the forests. These results can serve as a guideline for decision makers and for planning forest protection programs which can help attract public participation in the conservation and sustainable use of this valuable natural resource in Iran. Second study (Chap. 6) Optimized Reservoir Management for meeting Conflicting Stakeholder Preferences: Methodological Innovations with Evidence from Iran by Omid Zamani, Hemen Nader, Masoomeh Rashidghalam, and Frank A. Ward, applies a hybrid approach using a multi-criteria analysis (MCA), stake-

4

M. Rashidghalam

holder analysis (SA), and goal programming (GP) to allocate water reservoirs with the final aim of addressing conflicting demands between potential stakeholders and users to get water from the Mahabad reservoir in Iran’s West Azerbaijan province. The results of this study showed that the economic value of water for irrigated agriculture has major influence on the most economically efficient method to allocate reservoir water. Nevertheless, food security goals have an important influence on reservoir allocation. The optimized cropping pattern showed that the price of wheat cultivation was an important determinant of how reservoir water should be allocated for improved economic efficiency. Third study (Chap. 7) Willingness to Pay for IPM: An Application of the HeckmanCopula Approach by Esmaeil Pishbahar, Javad Hosseinzad, Sahar Abedi, and Pariya Bageri, uses the Heckman-copula approach to study Iranian farmers’ willingness to pay for IPM. The results of this study show that the logistic distribution of the selection equation and student’s t-distribution for the outcome equation are suitable. Moreover, the number of used IPM operations and knowledge of chemical pesticides’ risks have a positive and significant effect on the probability of WTP for IPM to avoid the negative effects of pesticides. Part C. Agricultural Prices and Commodity Market Analysis Part C contains four chapters examining agricultural prices and analyzing commodity markets in Iran. First study (Chap. 8) Testing the Law of One Price Under Nonlinearity: An application to Iranian Broiler and Egg Markets by Mohammad Ghahremanzadeh, Fatemeh Faryadi, and Ghader Dashti, tests the validity of the law of one price (LOP) in Iranian broiler and egg markets. It uses daily retail prices in West Azerbaijan, East Azerbaijan, Ardebil, Tehran, and Zanjan provinces for the period 2006–2014. The results of this study confirm the existence of nonlinear behavior in the different series. Emmanouilides and Fousekis’s (2012) nonlinear unit root test, which is an auxiliary regression for the ESTAR model, shows that both broiler and egg markets are well integrated and LOP holds in all market pairs. For the broiler market, a strong version of LOP holds for Tehran-West Azerbaijan, Tehran-Ardebil, Tehran-Zanjan, West Azerbaijan-Ardebil, and the West Azerbaijan-Zanjan pairs and a weak version of LOP holds for the Tehran-East Azerbaijan and West Azerbaijan-East Azerbaijan pairs. In the egg market as well, a strong version of LOP holds for all market pairs except for Tehran-Ardebil. Second study (Chap. 9) Assessing the Relationship between Marketing Mix and Customer Satisfaction: Evidence from Iranian Dairy Companies by Esmaeil Pishbahar, Roya Ferdowsi, and Babollah Hayati, tries to achieve three aims: identifying the effective factors of customer satisfaction with dairy products based on marketing mix elements, using the DEMATEL-ANP integrated techniques to determine the relative importance of each of these factors, and using the TOPSIS technique to rank seven Iranian dairy companies based on the level of their customer satisfaction. The results show that the product and promotion criteria were “effective” and between them, the “product” criterion had a greatest impact on the other criteria while the price and the location criteria were the “affected” criteria. Also, the sub-criterion “product

1 Introduction to Sustainable Agriculture and Agribusiness in Iran

5

diversity” in the product criteria, the “appropriate price” sub-criterion in the price criteria, the sub-criterion “coordination of the distribution channel and timely delivery” in the criteria place, and the sub-criterion “advertising and promotional and incentive measures” in the criteria promotion were more important. The results following the TOPSIS method show that PEGAH Company got the first place based on customer satisfaction while MIHAN and KALE were in the second and third place. RAMAK, KOHSARAN, SUTICH, and PAK were ranked next in terms of customer satisfaction. Third study (Chap. 10) Testing for Neutrality and Super-Neutrality of Money: Evidence from Iran’s Agricultural Sector by Esmaeil Pishbahar and Zahra Rasouli, uses the Fisher–Seater (FS) approach to test long-run neutrality (LRN) and longrun super-neutrality of money (LRSN) in Iran. To formulate LRN and LRSN and to extract the restrictions required by them, the study first introduces the long-run derivative. FS assumes a log-linear bivariate system of the stationary and invertible ARIMA models. To use the FS approach, the integration order of variables should be known. Hence, the study uses different unit root tests (DF-GLS, ADF, KPSS, and HEGY). In addition, it also does the Zivot–Andrews test, which includes a structural break in the data. The results show that M2 is neutral with respect to real GDP and real agricultural output. However, for nominal agricultural output neutrality of M2 is rejected. The neutrality results for the nominal GDP vary depending on the unit root test’s results. The findings indicate that the super-neutrality of M2 is confirmed with respect to real GDP. Forth study (Chap. 11) Effects of Oil Prices and Exchange Rates on Imported Inputs’ prices for the Livestock and Poultry Industry in Iran by Esmaeil Pishbahar, Parisa Pakrooh, and Mohammad Ghahremanzadeh, examines the correlation between oil prices and exchange rates with input prices for the livestock and poultry industry for two periods 1995–2004 (pre-crisis) and 2005–2014 (post-crisis), using the Vine copula ARIMA–MGARCH approach. The results show that due to a change in the intensity and type of correlation between oil prices, exchange rate and inputs in the second period as compared to the first period, inputs had a positive and high correlation with oil prices and a negative correlation with the exchange rate. Therefore, imported inputs in the poultry and livestock industry were impacted by global developments, the starting of the shocks in oil prices due to the 2005 Iraq-US war, the on-going global financial crisis and the increase in global prices for agricultural inputs.

1.3 Final Words The primary market for this book includes researchers, universities, administrative of agricultural institutions, producers, policy makers, and NGOs. This volume can also serve as complementary reading to text books on applied economics, agricultural, and natural resource economics. Students and professionals in agricultural economics, resource economics, risk management, and food policy as well as economists will

6

M. Rashidghalam

benefit from the valuable lessons presented in this book which demonstrate real world examples of the principles under discussion. Agricultural planners and decision makers in government agencies will also benefit from an expanded range of case studies in different regions of the country which could be applied in agricultural policy-making process. These include agricultural insurance, price, marketing, and natural resource economics. There are many books on agricultural economics that have been published recently. The most important feature of this volume which distinguishes it from other volumes is that its scope includes a wide range of topics related to agricultural economics. This book brings together economic theory, its application, and policy recommendation of agriculture sector in a developing country. I am grateful to all of those with whom I have had the pleasure to work during this work. I feel fortunate to express my deep sense of reverence and gratitude to Professor Almas Heshmati for his support, immaculate guidance, constructive criticism, and constant encouragement. I acknowledge the help and support of all faculty members of department of agricultural economics at University of Tabriz especially Dr. Pishbahar and Dr. Ghahremanzadeh for their assistance and valuable contributions. They all encouraged my work and encouraged me to do better work at all times. My sincere gratitude goes to the chapter’s authors who contributed their time and expertise to this book. Nobody has been more important to me in the pursuit of this book than the members of my family. I would like to thank my parents for encouraging me to pursue my dreams. My twin sister Haniyeh deserves my special thanks for her endless emotional support and love. Lastly, I am thankful to my lovely and lively nieces Fatemeh, Zahra, Asra, and Amanda whose innocence refreshed me during my difficult times. I hope that one day they can read this book and understand why I spent so much time in front of my computer and my hope is fulfilled now.

Reference Emmanouilides CJ, Fousekis P (2012) Testing for the LOP under nonlinearity: an application to four major EU pork markets. Agric Econ 43(6):715–723

Masoomeh Rashidghalam is Researcher at University of Tabriz. She did her B.Sc and Ph.D. in Department of Agricultural Economics at University of Tabriz and holds an M.Sc from Tarbiat Modares University. She was Visiting Scholar in Sweden and South Korea (in Department of Economics at Jönköping International Business School (JIBS) and Department of Economics at Sogang University). She was Assistant in organization of “ADB Workshop on Urbanization in

1 Introduction to Sustainable Agriculture and Agribusiness in Iran

7

Asia,” Seoul. Dr. Rashidghalam’s areas of expertise are: agricultural production economics, farm management, poverty and inequality, child well-being, labor economics, urbanization, and technical efficiency measurement. She has a wide range of teaching experience in econometrics, agricultural production economics, and microeconomics. She has written a book: Measurement and Analysis of Performance of Industrial Crop Production: The Case of Iran’s Cotton and Sugar Beet Production, 2018, published by Springer. She has contributed different chapters to six books.

Part I

Risk Management and Climate Change

Chapter 2

Agricultural Risk Management Through Weather-Based Insurance in Iran Esmaeil Pishbahar, Sahar Abedi, Ghader Dashti and Ali KianiRad

2.1 Introduction The agriculture sector is one of the most important sectors which structurally plays an important role in sustainable development. However, this sector is dependent on natural factors making it an essentially risky business. Nowadays, climate change and poor weather conditions are a major threat to sustainable development as they are damaging agricultural activities and food security. Risks are inevitable but manageable elements. One of the main policies in agricultural risk management is insurance. Insurance is an appropriate mechanism for reducing risks, stabilizing producers’ incomes, and providing security to the economy. It also provides an opportunity for attracting sustainable capital and technology in the agricultural sector. Agricultural insurance reduces poverty and also the double burden on the environment through sustaining production and providing farmers with income stability. In other words, by sustaining farmers’ incomes, the utilization of water and soil resources will also be sustainable which will enable achieving sustainable agricultural development. However, traditional insurance schemes are difficult to execute due to asymmetric information. Such challenges increase premium rates, indemnities, and assessing the cost of the damage accurately. Therefore, an insurer is forced to pay additional E. Pishbahar (B) · S. Abedi · G. Dashti Department of Agricultural Economics, University of Tabriz, Tabriz, Iran e-mail: [email protected] S. Abedi e-mail: [email protected] G. Dashti e-mail: [email protected] A. KianiRad Agricultural Economics-Research, Deputy-APERDRI-Ministry of Agriculture, Tehran, Iran e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2019 M. Rashidghalam (ed.), Sustainable Agriculture and Agribusiness in Iran, Perspectives on Development in the Middle East and North Africa (MENA) Region, https://doi.org/10.1007/978-981-13-6283-5_2

11

12

E. Pishbahar et al.

fees for assessing damages (Ofoghi et al. 2011). The costs of traditional insurance schemes should be supported by governments, but unfortunately in developing countries the governments often do not have sufficient funds for such schemes (Aziznasiri 2011). In Iran, the agricultural insurance law was approved in 1983 to insure agricultural crops against losses due to natural disasters. In 1984, agricultural production insurance operations started with two crops, cotton and sugar beet; these insurance operations now cover more than 153 crops. Currently, there are more than 2,000,000 insured people and the amount of received premium is about 2,357 milliard Rial. Paid indemnity is about 11,000 milliard Rial (the Iran Agricultural Insurance Fund 2013). This shows that the current insurance programs have not been reasonably balanced between the received premium and the paid indemnity. This problem makes the Iran Agricultural Insurance Fund dependent on subsidies that are about 65.6% of the premium. The high costs associated with traditional agricultural insurance and asymmetric information make this kind of insurance an inefficient tool in risk management. In addition, the excessive subsidies by the government mean that a new approach to insurance is needed (Aziznasiri 2011). On the other hand, climate change can critically affect agricultural productivity. Consequently, serious technological adjustments will be necessary to reduce the high ‘transaction costs’ and the negative impact of climate change on agricultural productivity (Bokusheva 2010). The experiences of developed and developing countries show that employing index-based insurance schemes such as weather-based crop insurance (WBCI) can resolve a number of basic problems. Unlike traditional insurance, in index-based insurance schemes, the premium and indemnity are determined based on indices and their impact on product losses. Since the indices are clear and the insured people cannot impact the indices, this insurance system can solve the problem of moral hazard as indemnification is not dependent on yield losses. Besides, unlike traditional insurance, there is no need for an insurance company to visit farmers’ fields to determine premiums or to assess damages. If the index amount is less or more than the trigger value, the insurer pays for the loss (Ofoghi et al. 2011). According to the benefits of index-based insurance and its successful results in many countries, it is expected that agricultural insurance in Iran will also use this insurance for sustainable development. Wheat as a strategic crop has a special position in both production and consumption in Iran. According to the Ministry of Agriculture of Iran (2012), there was 6.4 million hectares of land under wheat cultivation, and the share of ‘rainfed land’ was about 61.3%. Agricultural insurance covered 5.8 million hectares of agronomic land in which the highest share of coverage was for rainfed wheat. The East Azerbaijan Province is the second province in Iran with 7% of its total area under wheat cultivation, and Miyaneh County is the largest producer of wheat in this province with 20% area and 20% products. Numerous studies have been done on WBCI in Iran and also in other countries. For instance, Ofoghi et al. (2011) studied WBCI for rainfed wheat in the city of Maragheh. They used the Archimedean copula to investigate the dependence structure between weather indices and yields. The authors computed the insurance premium, and their

2 Agricultural Risk Management Through Weather-Based Insurance …

13

results showed that the premium was more than the current program’s premium leading to the potential possibility of greater indemnification. In separate studies, Ardestani (2012) and Ghahremanzadeh et al. (2014) investigated the feasibility of WBCI in rainfed wheat risk management. They determined the factors that influenced WBCI. Zhu et al. (2008), Turvey and Belltawn (2009), and Bokusheva (2010) did research in developing countries and showed that WBCI did not have the problems associated with traditional insurance and that WBCI had been successful. This insurance system simulates yields with attention to weather indices for computing the premium. Therefore, the dependence structures between yields and weather indices have to be determined. Traditionally, the literature has used a simple regression and linear correlation coefficient. Nevertheless, these methods have serious problems (e.g., a one-sided relationship, using one weather index and using normal distribution). Therefore, an investigation of the dependence structure with attention to the variable ‘joint distribution’ can lead to reliable results. Karuaihe et al. (2006) showed that one index may not adequately explain yield variations. Thus, there is a need to use more than one index. Hence, searching for flexible multivariate distributions has made ‘copula modeling’ increasingly popular in many fields of its application (Brechmann and Schepsmeier 2013). The term ‘copula’ comes from the Latin term referring to ‘link, join, or connect’ (Chen et al. 2013). A group of marginal distributions can be connected with a copula to create a joint distribution (Nelsen 2006). Though the simple copula functions are better than other dependence structure measurement methods, they have limitations in the arbitrary dimension. Multivariate data often has complex dependence patterns such as non-symmetry and dependence in the extremes. Multivariate copulas lack the flexibility of modeling dependence among a large number of variables. Though their generalization leads to some improvements, they become complex in their structure. They also have other limitations such as parameter restrictions (Brechmann and Schepsmeier 2013). A high-dimensional data analysis requires flexible multivariate stochastic models which can consider inherent dependency patterns. Considerable attempts have been made to increase the flexibility of multivariate copula models. Vine copulas are one of the best-received attempts (Czado et al. 2013). Vine copulas were first proposed by Joe (1996) and were developed in more detail by Bedford and Cooke (2001, 2002) and Kurowicka and Cooke (2006). In their research on risk management on ‘the Euro Stoxx 50,’ Brechmann and Czado (2011) used the high-dimensional vine copula. In their study of financial returns, Dibmann et al. (2013) used the ‘regular vine copula’ and Goodwin (2012) showed that the vine copula models were more flexible and efficient in explaining the dependence structure in modeling for crop insurance and reinsurance contracts. According to other studies, using vine copula methods for describing the dependence structure between wheat yields and weather indices can present reliable results. Consequently, in this study we use weather-based crop insurance for the ‘Sardari rainfed wheat variety’ in Miyaneh County as an efficient instrument for risk management. We describe the dependence structure between yields and weather indices using the vine copula functions.

14

E. Pishbahar et al.

The rest of the paper is organized as follows. Section 2.2 provides methodology, while empirical results are presented in Sect. 2.3. Section 2.4 gives the conclusion and recommendations.

2.2 Methodology Sklar (1959) introduced copulas in ‘Sklar’s theorem.’ The role of copulas in the relationship between univariate margins to gain multivariate distribution functions was explained by Sklar’s theorem (Nelsen 2006). Let F be a joint distribution function, with margins F 1 , …, F d . Then, there exists a copula function in the unit hypercube [0, 1]d with uniform U(0, 1) marginal distributions that can be obtained as:   C(u 1 , . . . , u d )  F F1−1 (u 1 ), . . . , Fd−1 (u d )

(2.1)

The density function of the copula, c, can be derived using Eq. 2.1 and marginal density functions, f i (Goodwin et al. 2011): c(u 1 , . . . , u d ) 

f (F1−1 (u 1 ), . . . , Fd−1 (u d )) ∂ d C(u 1 , . . . , u d )  d −1 ∂u 1 . . . ∂u d i1 f i (Fi (u i ))

(2.2)

There are many families of copulas in the literature. Two of the most frequently used copula families are Archimedean and elliptical copulas. Archimedean copulas including Clayton, Gumbel, Frank, and Joe are obtained by generator functions (ϕ), while elliptical copulas are gained directly by inverting ‘Sklar’s theorem.’ The most famous examples of elliptical copulas are Gaussian and Student-t copulas (Brechmann and Schepsmeier 2013) (see Table 2.1 for details). Vine copulas: Vines are a flexible graphical model for specifying the so-called pair-copula constructions (PCCs) that depict a multivariate distribution. Vine or a set of linked trees factorize the multivariate probability density function into pair-copulas (Czado et al. 2013). Each pair-copula can be selected from arbitrary families. Hence, vines combine the advantages of multivariate copula modeling giving more flexibility to dependency modeling. Specifically, tail dependency and asymmetry can be considered (Brechmann and Schepsmeier 2013). According to Aas et al. (2009), a threedimensional joint density function can be factorized as: f (x1 , x2 , x3 )  f 1 (x1 ) · f (x2 |x1 ) · f (x3 |x1 , x2 ) By Sklar’s theorem, it can be written as: f (x2 |x1 ) 

c12 (F1 (x1 ), F2 (x2 )) · f 1 (x1 ) · f 2 (x2 ) f (x1 , x2 )  f 1 (x1 ) f 1 (x1 )

(2.3)

  θ δ θ δ 1/δ 1/θ 1 − 1 − 10−[(− log[1−(1−u) ]) +(− log[1−(1−v) ]) ]

  −1/δ 1/θ 1 − 1 − (1 − u¯ θ )−δ + (1 − v¯ θ )−δ − 1

(− log(t))θ

t)−1 − log( exp(−θ exp(−θ )−1 )

− log(1 − (1 − t)θ )

(t −θ − 1)δ

(− log[1 − (1 − t)θ ])δ

(1 − (1 − t)θ )−δ − 1

Gumbel

Frank

Joe

BB1

BB6

BB7

⎡

δ

⎤1/θ      θ θ − − log 1−(1−δu)θ −log 1−(1−δv)θ   1−(1−δ) 1−(1−δ) · 1−(1−δ)θ ⎦

1/δ −1/θ 

δ

δ 1+ u +v

1−⎣1−10

 −1/θ max u −θ + v−θ − 1; 0 }



Source Brechmann and Schepsmeier (2013) and Fischer (2002)

− log( 1−(1−δt) ) 1−(1−δ)θ

θ

1/θ  1 − (1 − u)θ + (1 − v)θ ) − (1 − u)θ (1 − v)θ

(t −θ − 1)/θ

Clayton

BB8

  θ 1 exp − (− ln u)θ + (− ln v) θ

 −θu −θv −1) − θ1 ln 1 + (e −1)(e e−θ −1

–

Student-t

   P −1 (u), −1 (v)   tv, p tv−1 (u), tv−1 (v)

–

Gaussian

Functional form

Generator function

Name

δ ∈ (0, 1]

θ ≥ 1,

δ>0

θ ≥ 1,

δ≥1

θ ≥ 1,

δ≥1

θ > 0,

θ ≥1

θ ∈ R\{0}

θ ≥1

4 δθ

0

1

θ

dx

t log(t)(1 − t)2(1−θ )/θ dt

c/θ 0 exp(x)−1

θ

0

1

 −[1 − (1 − t)θ ]δ+1

    − log 1 − (1 − t)θ × (1 − t) 1 − (1 − t)θ dt

1+

0

  1  (1 − δt)θ − 1 − log θ (1 − δ) − 1  × (1 − tδ)(1 − (1 − tδ)−θ ) dt

4 δθ

 (1 − (1 − t)θ )−δ − 1 dt × (1 − t)θ −1

1+

0

1

+4

2 δ(θ +2)

4 θ2

4 θ

1+ δθ4

1−

1−

1−

1− 1 θ

ρ ∈ (−1, 1), v > 2 θ θ +2

arcsin(ρ)

2 π

θ >0

arcsin(ρ)

2 π

Kendall’s τ

ρ ∈ (−1, 1)

Parameter range

Table 2.1 Denotation and properties of bivariate elliptical and Archimedean copula families

1

(0,0)

1

1

(2− δ , 2 − 2 θ )

1

(0, 2 − 2 δθ )

1

(2− δθ , 2 − 2 δ )

(0, 2 − 21/θ )

(0, 0)

(0, 2 − 21/θ )

(2−1/θ , 0)

  √  2tv+1 − v + 1 1−ρ 1+ρ

0

Tail dependence (lower, upper)

2 Agricultural Risk Management Through Weather-Based Insurance … 15

16

E. Pishbahar et al.

 c12 (F1 (x1 ), F2 (x2 )) · f 2 (x2 )

(2.4)

and c23|1 (F(x2 |x1 ), F(x3 |x1 )) · f (x2 |x1 ) · f (x3 |x1 ) f (x2 , x3 |x1 )  f (x2 |x1 ) f (x2 |x1 )  c23|1 (F(x2 |x1 ), F(x3 |x1 )) · f (x3 |x1 )  c23|1 (F(x2 |x1 ), F(x3 |x1 ))c13 (F1 (x1 ), F3 (x3 )) · f 3 (x3 ) (2.5)

f (x3 |x1 , x2 ) 

Equation 2.3 can be rewritten as: Marginal

Unconditional pairs

      f (x1 , x2 , x3 )  f 1 (x1 ) · f 2 (x2 ) · f 3 (x3 ) × c12 (F1 (x1 ), F2 (x2 ) · c13 (F1 (x1 ), F3 (x3 )) Conditional pairs

   × c23|1 (F(x2 |x1 ), F(x3 |x1 ))

(2.6)

Therefore, the joint density in Eq. 2.3 can be explained with bivariate copulas C 12 , C 13 , and C 23|1 with densities c12 , c13 , and c23|1 , or the ‘pair-copulas.’ As the decomposition in Eq. 2.3 is not unique, and there are many PCCs, Bedford and Cooke (2001, 2002) introduced a graphical model called the ‘Regular vine copula’ (R-vine) to classify them. The R-vine copula with n-elements is a set of linked trees,1 v  {T1 , . . . , Tn−1 }. Tree j has n + 1 − j nodes and n − j edges. The edges in tree j become nodes in tree j + 1, and two nodes in tree j + 1 can be joined by an edge if the corresponding edges in tree j share a node (proximity condition). The R-vine specification embraces many possible pair-copula decompositions. Therefore, Aas et al. (2009), Brechmann et al. (2010), and Dibmann et al. (2013) suggest the strongest dependency in the ‘first tree’ to be considered. In general, they can be classified into three main groups, that is, R-vines, C-vines2 , and D-vines.3 C-vines and D-vines have been extensively used in the literature. C-vine trees have a star structure (Brechmann and Schepsmeier 2013). C-vines are characterized by a root node in each tree, or in other words the root node is connected to all the other nodes of the tree (Czado et al. 2013). D-vine trees have a path structure (Brechmann and Schepsmeier 2013); this means that each node does not have a degree more than two (Czado et al. 2013). The C-vine density with root nodes 1, …, d, and a d-dimensional D-vine density can be determined, respectively, as: f (x) 

d  k1

1 Tree is an

f k (xk )

d−1 d−i 

ci,i+ j|1:(i−1) (F(xi |x1 , . . . , xi−1 ),

i1 j1

acyclic connected graph with nodes and edges, and a graph is a set of nodes N and a set of edges E, which connect these nodes. 2 Canonical vine. 3 Drawable vine.

2 Agricultural Risk Management Through Weather-Based Insurance …

F(x j+1 |x1 , . . . , xi−1 )|θi,i+ j|1:(i−1) ) f (x) 

d  k1

f k (xk )

d−1 d−i 

17

(2.7)

c j, j+i|( j+1):( j+i−1) (F(x j |x j+1 , . . . , x j+i−1 ),

i1 j1

F(x j+i |x j+1 , . . . , x j+i−1 )|θ j, j+i|( j+1):( j+i−1) )

(2.8)

Now, the main problem is getting the conditional distribution functions. For a paircopula in tree j + 1, this can be established by using the previous trees’ pair-copulas by sequentially applying:   ∂C xυ j |υ− j F(x|υ− j ), F(υ j |υ− j )|θ (2.9) h(x|υ, θ )  F(x|υ)  ∂ F(υ j |υ− j ) where υ j is an arbitrary element of υ and υ− j denotes the (d − 1)-dimensional vector υ excluding υ j (Brechmann and Schepsmeier 2013). It is assumed that datasets have uniform margins in [0, 1] or the so-called copula data. Thus, data has to be transformed with empirical marginal distribution functions. (v, B, θ ) is a R-vine copula specification where (v) is a d-dimensional regular vine and B  {Be |i  1, . . . , n − 1; e ∈ E i } is a set of copulas with Be or bivariate copula or pair-copula and the pair-copula parameters θ  θ (B(v)). To begin with, we have to determine the vine tree structure. Then, for a given vine tree structure, we choose the pair-copula families. Finally, we estimate the parameters of the pair-copulas. We adopted C-vine and D-vine and the selection mechanism suggested by Dibmann et al. (2013) to choose the optimal ordering of data to define the vine. In this mechanism: 1. The empirical Kendall’s τ for all possible variable pairs should be calculated. 2. The spanning tree has to be selected so that it maximizes the sum of absolute empirical Kendall’s τ . 3. For each edge in the selected spanning tree, a copula is chosen and its parameters are estimated by the sequential method (SM) and the joint maximum likelihood (JML) approach. 4. Applying Eq. 2.9, we created pseudo-observations. 5. For {T 2 , … T d }, iterate 1–4. The optimal copula functions are chosen with utilization of the minimized values of AIC and BIC. Simulation from vines is discussed in Aas et al. (2009). First, u i is sampled: i  1, …, d independent uniform on [0, 1]. Then, set: ⎧ x1 ⎪ ⎪ ⎪ ⎪ x ⎪ ⎨ 2 x3 ⎪ . ⎪ ⎪ ⎪ .. ⎪ ⎩ xn

 u1  F −1 ( u 2 |x1 )  F −1 ( u 3 |x1 , x2 )  F −1 ( u d |x1 , . . . , xd−1 )

(2.10)

18

E. Pishbahar et al.

# To determine F( x j #x1 , . . . , x j−1 ) for each j, the h-function in Eq. 2.9 is used, and at the end, 10,000 random observations for the yield are simulated, but they are in [0, 1]. Therefore, we applied the inverse cumulative distribution function for yield to turn the observations into real form. We used three goodness-of-fit tests—Anderson–Darling, Kolmogorov–Smirnov, and chi-square—to determine an appropriate marginal distribution function. Determining the premium: In general, the pricing for contracts uses expected losses. To this end, the joint distribution function of weather indices and crop yields is estimated and then simulated data for yields is obtained. Forecasting the yield amount for next year is obtained using the ARIMA model. Then, critical values of crop yield, yC , in different levels of coverage (e.g., 100, 90, 80, and 50%) are: yC  yfcast × COV

(2.11)

where yfcast is the forecasted yield and COV is coverage level. Next, we compared yC and 10,000 simulated observations of yield. In case each of the observations is less than yC , the insurer must pay the indemnity. The expected loss or fair premium is computed as: Fair premium  Ave[Max(yc − y, 0)] ∗ P

(2.12)

where P is the guaranteed price for wheat (Aziznasiri 2011). According to Skees et al. (1997), the loading factor should add to expected losses to cover transaction costs. Thus, the actual premium is calculated by: Actual premium 

Fair premium 0.9

(2.13)

The indemnity function: To design indemnity, a weather variable is selected in every phenology stage which is strongly related to wheat yields. For this, we used the Clayton, Gumbel, and Gaussian copula parameters. As θ → 0 or ρ → 0, it represents independence, and as θ → ∞ or |ρ|→ 1, it shows a high dependence. Then, for every selected variable, we determined the trigger value and stop-loss and designed the indemnity function with attention to the phenology stage. Given i as an amount of weather variable, g(i) is an indemnification function. There are two methods for indemnification that depend on the phenology stage—index deficiency and index increasing. In index deficiency, the indemnification function is: ⎧ ⎪ (if i < μ) ⎨ L,∗ −i , (if μ < i < i ∗ ) g(i)  L ii∗ −μ (2.14) ⎪ ⎩ 0, ∗ (if i > i ) where μ and i* are stop-loss and trigger values, respectively. μ and i* are the minimum and maximum amount of weather variables in the last year. This method is

2 Agricultural Risk Management Through Weather-Based Insurance …

19

suitable for crops in that the increasing of weather variables’ values has a positive effect on crop growth. In contrast, in index increasing indemnification the function is written as: ⎧ ⎪ ⎨ L , ∗ (if i > μ) i−i ∗ < i < μ) g(i)  L μ−i (2.15) ∗ , (if i ⎪ ⎩ 0, (if i < i ∗ ) In Eq. 2.15, μ and i* are the maximum and minimum amount of weather variables in the last year, respectively. This method is suitable for crops in that the decreasing of weather variables’ values has a positive effect on crop growth (Ofoghi et al. 2011). In our study, average temperature, cumulative rainfall index, relative humidity, and faster wind speeds are important variables for Sardari rainfed wheat variety in Miyaneh County. According to Nourmohammadi et al. (2005), the phenology stages can be classified into six stages to investigate the effects of these variables on yield—germination, tillering and hibernation, stem extension, heading, ripening, and harvest. In Miyaneh, these stages are from October 7 to December 21; December 22 to March 19; March 20 to May 5; May 6 to June 21; June 22 to July 22; and July 23 to August 22, respectively. In the five stages, the average temperature, cumulative rainfall index, and relative humidity affect crop yields. While in the harvest stage, faster wind speeds influence yields. These indices are modified for every year as: T  wGer TGer + wTil TTil + wSteam TSteam + wHead THead + wRipe TRipe CRI  wGer CRIGer + wTil CRITil + wSteam CRISteam + wHead CRIHead + wRipe CRIRipe RH  wGer RHGer + wTil RHTil + wSteam RHSteam + wHead RHHead + wRipe RHRipe U  UHarvest (2.16) where T is the average temperature (°C), CRI is the cumulative rainfall (mm), RH is the relative humidity (percent), U is the faster wind speed (knot), and w represents a sub-period’s weight which is obtained by the standard regressions of the weather variables on wheat yields or Y (kg per hectare). The information on Sardari rainfed wheat variety’s yields and weather variables was collected for the period 1987–2013. Data was obtained from the Ministry of Agriculture of Iran and the Iran Meteorological Organization.

2.3 Results and Discussion First, to simplify the calculation, a number was attributed to each copula data as: YECDF  1, TECDF  2, CRIECDF  3, HECDF  4, UECDF  5

20

E. Pishbahar et al.

As discussed, the order of the variables has to be selected. For C-vine, the root nodes for each tree have to be determined, and for D-vine, the order of the variables in the first tree needs to be chosen. Therefore, we calculated the empirical Kendall’s τ for all pairs and compared 60 different kinds of C-vines and D-vines. Then, based on the maximizing rank correlation in each vine tree, we selected the order of the variables. The root node order in the dataset for C-vine was determined as CRI, T , U, Y , and finally RH. The estimation of corresponding parameters’ results and selection of suitable copula for the C-vine model are presented in Table 2.2. We considered a large variety of copula functions for each pair, and based on the minimized value of AIC/BIC, we opted for the best one. JML has a larger log-likelihood value; thus, we used JML’s results for defining the tree structure. The C-vine tree structure implied by the estimates is given in Fig. 2.1. This structure is ‘R 3.1.1’ software’s output. In fact, the tree structure represents the joint density function for wheat yield and weather indices. We have three elements on each edge: (i) the selected copula family, (ii) Kendall’s τ , and (iii) the estimated parameter. Similarly, the order of variables for the D-vine model is Y , CRI, RH, T , and U, in the first tree. The results of the parameters’ estimation for D-vine are reported in Table 2.3. The last row in Table 2.3 shows that JML is better than SM. The D-vine tree structure is presented in Fig. 2.2. To compare the C-vine and D-vine models, we did the Vuong and Clarke tests. In the Vuong test, the null hypothesis showed that the C-vine model was better, but the Clarke test showed that the two models were statistically equal. The results of these two tests are presented in Table 2.4. As shown in this table, the null hypothesis is

Table 2.2 Copula selection and parameter estimation result for C-vine model Tree number

Parameters

Selected family

JML coeff.

SM coeff.

First

P13,2 P23,2

BB7 180

1.704 1.46

1.716 1.130

P3,5

Frank

−1.329

−1.198

P3,1

Frank

4.652

4.722

P13,4 P23,4

BB7 180

2.853 1.699

2.54 1.756

P2,5|3

Gaussian

−0.169

−0.165

P2,1|3

Joe 180

1.292

1.313

P2,33

Clayton

0.747

0.720

Third

P5,1|3,2

Joe 180

1.192

1.189

P5,4|3,2

Frank

0.935

0.756

Fourth

P1,4|3,2,5

Gaussian

1.203

Log-likelihood

75.47

Second

−0.140 75.050

2 Agricultural Risk Management Through Weather-Based Insurance …

21

rejected and the D-vine model is better for describing joint distribution. Therefore, we used the D-vine model to simulate the data. We used a large number of theoretical distributions (65) to represent wheat yield distribution using the EasyFit 5.5 software. The results of three tests for goodness of fit are presented in Table 2.5. According to Bokusheva (2010) and Goodwin (2012), Weibull distribution is suitable for yield, but in our study we found Wakeby to be more appropriate. Therefore, we used both the distributions and their properties for yield which can be written as: yeild ∼ WAK(ξ  308.31, α  1475.6, β  3.2585, γ  327.58, δ  −0.47761) yeild ∼ WEB(α  3.2723, β  960.82) where ξ is location parameter, α, β are scale parameters, and δ, γ are shape parameters in Wakeby distribution. In Weibull distribution, α and β are shape and scale parameters, respectively.

Fig. 2.1 C-vine tree structure

22

E. Pishbahar et al.

Table 2.3 Copula selection and parameter estimation result for D-vine model Tree number

Parameters

Selected family

JML coeff.

SM coeff.

First

P1,3

Frank

4.494

4.722

P13,4 P23,4

BB7

2.634 1.791

2.540 1.756

P4,2

Gumbel

2.189

2.106 −0.105

Second

P2,5

Gaussian

−0.105

P1,4|3

Joe 180

1.393

1.393

P3,2|4

Clayton 180

0.271

0.315

P4,5|2

Frank

1.021

1.121

Third

P1,2|3,4

Clayton 180

0.015

0.013

P3,5|4,2

Frank

−0.487

−0.601

Fourth

P1,5|3,4,2

Joe 180

1.203

Log-likelihood

74.25

Fig. 2.2 D-vine tree structure

1.168 74.17

2 Agricultural Risk Management Through Weather-Based Insurance …

23

Table 2.4 Vuong and Clarke tests to compare C-vine and D-vine models Test

Statistic

P-value

Vuong

−2.039

0.041

Clarke

3

0.00008

Table 2.5 Goodness of fit for wheat yield Suitable distribution

Chi-square

Anderson–Darling Kolmogorov–Smirnov

First candidate

Wakeby Statistic

0.075

0.131

0.058

P-value

0.994

–

0.999

Critical value (α  %5)

7.814

2.501

0.259

Statistic

0.321

0.308

0.112

P-value

0.956

–

0.858

Critical value (α  %5)

7.814

2.501

0.259

Second candidate

Weibull

Using D-vine, 10,000 observations for yield (and other weather indices) were simulated to convert data into a real form, and we computed the Wakeby and Weibull inverse cumulative distributions. According to Wakeby and Weibull distributions, the average simulated yield was about 871.68 and 863.41 kg/ha, respectively. The AIC and BIC statistics showed that ARIMA (1, 0, 0) was the optimal model. The forecasting yield amount for next year using ARIMA (1, 0, 0) was 871.73 kg/ha. Then, we calculated the critical value of the yield based on four coverage levels according to Eq. 2.11. The guaranteed price is about 10,500 Rial. Table 2.6 presents the amount of premia for Wakeby and Weibull distributions. The results of two distributions are similar. The probability of loss shows the probability of yc > y, that is, 45.83 with respect to forecasting yield according to Wakeby

Table 2.6 Computing premium amount for rainfed wheat in Miyaneh (Rial) Distribution Coverage level (%)

Critical values (kg/ha)

Loss probability

Average [max(yc − y), 0] (kg/ha)

Fair premium

Actual premium

Wakeby

100

871.73

45.83

113.42

1,190,917

1,323,241

90

784.56

35.91

77.93

818,283

909203.3

80

697/38

27.59

50.42

529488.4

588320.4

50

435/86

8.12

5.06

53130.49

59057.21

Weibull

100

871.73

51.49

119.98

1,259,860

1,399,844

90

784.56

40.17

80.12

841,282

934757.8

80

697/38

29.06

49.97

524703.2

583003.6

50

435/86

6.73

6.74

70824.58

78693.98

24

E. Pishbahar et al.

Table 2.7 Estimation of the parameter for Clayton, Gumbel, and Gaussian copulas Phenology stages

Variables

Gaussian

Clayton

Gumbel

Germination

Yield and T

−0.314

–

– 1.703

Tillering

Stem extension

Heading

Ripening

Yield and CRI

0.608

1.406

Yield and RH

0.404

0.721

1.36

–

– 1.24

−0.019

Yield and T Yield and CRI

0.3

0.48

Yield and RH

0.48

0.94

1.47

–

– 1.18

−0.28

Yield and T Yield and CRI

0.23

0.36

Yield and RH

0.39

0.69

1.34

–

– 1.07

−0.18

Yield and T Yield and CRI

0.11

0.15

Yield and RH

0.47

0.91

1.45

Yield and T

−0.19

–

–

Yield and CRI

−0.025

–

–

Yield and RH

0.25

0.29

1.19

Clayton and Gumbel parameters cannot be negative Table 2.8 Amount of trigger value and stop-loss Variable

Trigger value

CRI in germination

166.60

Stop-loss 5.30

RH in tillering

76.19

59.86

RH in stem extension

62.25

39.38

RH in heading

56.83

31.44

RH in ripening

29.23

46.01

U in harvest

10.00

18.00

Table 2.9 Liability insurance (Rial) Coverage level (%)

Critical values (kg/ha)

Liability insurance

100

871.73

9,153,201

90

784.56

8,237,881

80

697/38

7,322,561

50

435/86

4,576,600

distribution. The computed premium amount for this insurance scheme, at the 80% coverage level, is less than the current insurance amount of 860,000 Rial. In addition, as expected the premium amount reduces with decreasing coverage levels. According to Table 2.1, the parameter of the copula depends on Kendall’s τ . Therefore, we calculated Kendall’s τ for yield and weather variables in each phenology stage. The obtained copula parameters (for Clayton, Gumbel, and Gaussian) are given in Table 2.7. Note that faster wind speeds only affected yields

2 Agricultural Risk Management Through Weather-Based Insurance … Liability Insurance

Liability Insurance

L

L

i * −i i * −µ

µ=5.3

i*=166.6

CRI (mm)

i * −i i * −µ

µ=39.38

i*=29.23

L

L

i*=62.25

RH (%)

Liability Insurance

i * −i i * −µ

RH (%)

b) Indemnity function in tillering

i * −i i * −µ

µ=31.44

c) Indemnity function in stem extension

L

i*=76.19

µ=59.86

Liability Insurance

Liability Insurance

L

i * −i i * −µ

L

L

a) Indemnity function in germination

L

25

i*=56.83

RH (%)

d) Indemnity function in heading Liability Insurance

L

L

µ=46.01

RH (%)

e) Indemnity function in ripening

L

i * −i i * −µ

i*=10

µ=18

U (Knot)

f) Indemnity function in harvest

Fig. 2.3 Indemnity charts for wheat in phenology stages

in the harvest stage. Hence, there is no need to calculate the copula parameter for yield and this variable. The copula parameters’ results in Table 2.7 show that CRI had the strongest correlation with yield in the germination stage. In tillering and hibernation, stem extension, heading and ripening, RH had the largest effect on crop yields, so we used these variables to design the indemnity function. For the indemnity function, the amount of trigger value and stop-loss was determined based on our last information. According to our results, an increase in RH

26

E. Pishbahar et al.

and U in the ripening and harvest stages had a negative effect on yields. Therefore, trigger value and stop-loss were the minimum and maximum amount of these variables, respectively. However, we used the index deficiency method for other selected variables. The amount of trigger value and stop-loss is presented in Table 2.8. Liability insurance depends on past experience and the insurer’s policies which can be determined as part of production costs or production values. We calculated liability insurance as different percentages of critical yield values. The results are given in Table 2.9. The indemnity charts are given in Fig. 2.3. In Fig. 2.3, the charts for germination, tillering and hibernation, stem extension, and heading stages show the index deficiency method for indemnification. However, Fig. 2.3 exhibits the index increasing method for the ripening and harvest stages.

2.4 Conclusion and Recommendations Wheat as a strategic product has a special position in production and consumption in Iran, and East Azerbaijan Province and Miyaneh County have special places in the country’s wheat production. Like other agricultural products, the production of this crop is a risky business. One of the main risks in the agricultural sector is related to weather, especially in dryland farming. Traditional insurance is not efficient because of asymmetric information. Hence, this study designed weather-based crop insurance (WBCI) for rainfed wheat in Miyaneh County. This insurance system can solve the problem of asymmetric information and is a suitable tool for achieving sustainable development. According to our results, the WBCI premium is less than the current insurance premium; therefore, farmers can be encouraged to accept this insurance scheme. In addition, the government pays 65.6% of the current insurance premium. Less premium amounts can reduce such costs for the government. Our indemnification results show that CRI in germination, RH in tillering and hibernation, stem extension, heading and ripening, and U in the harvest stages had the strongest correlation with wheat yields. Weather indices change in a region over time, and every region has a special climate. Thus, such a study should be repeated by selecting regions that have similar weather. Further, successful methods adopted by developing countries should also be tried. Vine copula does not have limitations of high dimensions, so many weather indices can be used to get more reliable results.

References Aas K, Czado C, Frigessi A, Bakken H (2009) Pair-copula constructions of multiple dependence. Insur Math Econ 44(2):182–198 Ardestani H (2012) The feasibility of weather based crop insurance in the management of rainfed wheat. Master thesis. University of Tehran, Agricultural Faculty (in Persian)

2 Agricultural Risk Management Through Weather-Based Insurance …

27

Aziznasiri S (2011) Weather-based crop insurance as a viable instrument for agricultural risk management in Iran. Master thesis, Allameh Tabataba’i University, E.C.O. College of Insurance (in Persian) Bedford T, Cooke RM (2001) Probability density decomposition for conditionally dependent random variables modeled by vines. Ann Math Artif Intell 32(1–4):245–268 Bedford T, Cooke RM (2002) Vines: a new graphical model for dependent random variables. Ann Stat 30(4):1031–1068 Bokusheva R (2010) Measuring the dependence structure between yield and weather variables. Institute for Environmental Decisions, ETH Zurich Brechmann EC, Czado C (2011) Risk management with high-dimensional vine copulas: an analysis of the Euro Stoxx 50. Submitted for publication Brechmann EC, Czado C, Aas K (2010) Truncated regular vines and their applications. Can J Stat 40(1):68–85 Brechmann EC, Schepsmeier U (2013) Modeling dependence with C- and D-vinecopulas: the R-package CDVine. J Stat Softw 52(3):1–27 Chen S, Wilson WW, Larsen R, Dahl B (2013) Investing in agriculture as an asset class. Department of Agribusiness and Applied Economics Agricultural Experiment Station North Dakota State University Czado C, Brechmann EC, Gruber L (2013) Selection of vine copulas. In: Jaworski P, Durante F, Härdle W (eds) Copula in mathematical and quantitative finance. Lecture Notes in Statistics, vol 213. Springer, Berlin, Heidelberg Dißmann J, Brechmann EC, Czado C, Kurowicka D (2013) Selecting and estimating regular vine copulae and application to financial returns. Comput Stat Data Anal 59:52–69 Fischer M (2002) Tailoring copula-based multivariate generalized hyperbolic secant distributions to financial return data: an empirical investigation. Institute of Statistics and Econometrics University of Erlangen-Nurnberg, Lange Gasse 20, D-90403 Nurnberg, Germany Ghahremanzadeh M, Dashti GH, Afrasiyabi S, Hoseinzad J, Hayati B (2014) The feasibility of weather based crop insurance for rainfed wheat in Ahar County. Iran Agric Econ Develop Res 45(2):383–393 (in Persian) Goodwin BK (2012) Copula-based models of systemic risk in U.S. agriculture: implications for crop insurance and reinsurance contracts. The NBER conference on Insurance Markets and Catastrophe Risk in Boston Goodwin BK, Holt MT, Onel G, Prestemon JP (2011) Copula-based nonlinear models of spatial market linkages. Am J Agric Econ, in press Iran Agricultural Insurance Fund (2013) The report of Iran Agricultural Insurance Fund in recent years. Department of Planning & management (in Persian) Joe H (1996) Families of m-variate distributions with given margins and m(m − 1)/2 bivariate dependence parameters. In: Ruschendorf L, Schweizer B, Taylor MD (ed) Distributions with fixed marginals and related topics Karuaihe RN, Wang HH, Young DL (2006) Weather-based crop insurance contracts for African Countries. Contributed paper prepared for presentation at the International Association of Agricultural Economists Conference Kurowicka D, Cooke RM (2006) Uncertainty analysis with high dimensional dependence modelling. In: Wiley series in probability and statistics, Wiley, Chichester Nelsen RB (2006) An introduction to copulas, 2nd edn. Springer, New York Nourmohammadi GH, Siyadat A, Kashani A (2005) Agronomy. The 6th print. Shahid Chamran University of Ahvaz, Ahvaz, Iran (in Persian) Ofoghi R, Kianirad A, Aziznasiri S (2011) Agricultural insurance of climatic indices-based: an effective tool on agricultural risk management in Iran. Agric Insur (29):25–51 (in Persian) Sklar A (1959) Fonctions de répartition à n dimensions et leurs marges. Publications de l’Institut de Statistique de l’Université de Paris 8:229–231 Skees JR, Black JR, Barnett BJ (1997) Designing and rating an area-yield crop insurance contracts. Am J Agr Econ 79:430–438

28

E. Pishbahar et al.

The Ministry of Agriculture of Iran (2012) Iran Ministry of Agricultural Organization website. Available at: http://dbagri.maj.ir/cost/ostan.asp Turvey C, Belltawn BC (2009) Weather risk and the viability of weather insurance in western China. In: Conference of the American Agricultural Economics Association, Milwaukee, Wisconsin Zhu Y, Ghosh S, Goodwin B (2008) Modeling dependence in the design of whole farm insurance contract a copula-based approach. In: Contributed paper at the Annual Meeting of the AAEA, Orlando USA, July 27–29

Esmaeil Pishbahar is Associate Professor and Head of the Department of Agricultural Economics at University of Tabriz, Iran. He holds a B.Sc. in agricultural economics from University of Tabriz and a M.Sc. in agricultural economics from University of Tehran. He did his Ph.D. in science economics in the Departments of Economics and Management, University of Rennes 1, France. His areas of interest and research are applied econometrics, agricultural risk management and insurance, and international trade. His teaching areas are advanced econometrics, mathematical economics, and macroeconomics at under- and postgraduate levels. He has over 100 publications in journals and chapters in books. Sahar Abedi holds B.Sc. and M.Sc. in agricultural economics in the Department of Agricultural Economics, University of Tabriz. For her M.A. thesis, she aimed to have a better understanding of weather-based crop insurance premium for wheat crop. Her research interests lie at risk management and crop insurance. Ghader Dashti is Professor in the Department of Agricultural Economics at University of Tabriz. Currently, he is Dean of Faculty of Agriculture at University of Tabriz. He received both his M.Sc. and Ph.D., in agricultural economics, from University of Tehran. He currently teaches undergraduate and graduate courses in farm management, macroeconomics, and agricultural economics. His fields of expertise are food market policy, agricultural production management, productivity and efficiency, and climate changes. Ali KianiRad is Assistant Professor in agricultural economics at Agricultural Planning, Economics, and Rural Development Research Institute (APERDRI). He holds a B.Sc. in agricultural economics from Shahid Bahonar University of Kerman and a M.Sc. and Ph.D. in agricultural economics from University of Tehran. He teaches food policy, agricultural policy, and natural resource economics at both undergraduate and postgraduate levels. His areas of concentration are agricultural economics, agricultural insurance, and risk management.

Chapter 3

Determining Conditional Value at Risk (CVaR) for Food Industry Companies’ Stocks Portfolios in the Tehran Stock Market Sahar Abedi and Esmaeil Pishbahar

3.1 Introduction Investments are important for achieving sustainable growth and development; this also involves equipping long-term financial resources through financial markets. A strong and developed stock market for agricultural exchange not only increases the volume and quality of the investments but also contributes to the sustainable growth and development of this sector. Despite the numerous difficulties faced by Iranian producers in the traditional market, getting agricultural products on the stock market can be a solution. Although the supply of agricultural products on the stock market requires infrastructure and facilities, it leads to benefits for producers. In addition, supplying agricultural products on the stock market guarantees the sustainability of production and agricultural producers can also plan for the future by having sufficient capital at the right time. If the stock market has a supply and demand system, the quality of production will improve and consequently it will sustain future production. Since the quality of products is high on a stock market, prices will increase and will be clear and transparent. This will lead to secure incomes. However, all these effects depend on investors’ confidence in the agricultural products’ financial market. In recent years, the complexity of financial markets has led to their destabilization. Consequently, investors have been exposed to more financial risks than before, and they will increasingly demand an effective financial risk management instrument. VaR and CVaR are common measures of risk. VaR is a simple and popular tool which measures the worst expected losses in a portfolio’s value over a defined period at a given significance level, while CVaR as introduced by Artzner et al. (1999) S. Abedi · E. Pishbahar (B) Department of Agricultural Economics, University of Tabriz, Tabriz, Iran e-mail: [email protected] S. Abedi e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2019 M. Rashidghalam (ed.), Sustainable Agriculture and Agribusiness in Iran, Perspectives on Development in the Middle East and North Africa (MENA) Region, https://doi.org/10.1007/978-981-13-6283-5_3

29

30

S. Abedi and E. Pishbahar

is coherent and can indicate excess losses over the threshold VaR. An important element in computing VaR and CVaR is the distribution assumption for each return in a portfolio. It is assumed that returns are normally distributed. However, in the last two decades financial markets have experienced extreme risks because of currency crises, stock market crashes, and credit crises such as the Asian financial crisis in 1997, the US sub-prime crisis in 2007, and the EU debt crisis in 2009 (Ayusuk and Sriboonchitta 2014). Hence, the normal distribution assumption leads to an underestimation of risks. Therefore, searching for accurate VaR and CVaR models with respect to such extreme events leads to an application of the Extreme Value Theory (EVT) approach to estimate the tails of the distribution (see, e.g., Ayusuk and Sriboonchitta 2014; Bali 2003; Emmanouil and Nikos 2012; Liu 2011; Marimoutou et al. 2009; McNeil and Frey 2000). The EVT approach uses extreme values rather than the whole dataset so it can provide a better fit for heavy-tailed data. According to Emmanouil and Nikos (2012), the EVT approach considers the tails of distribution separately, offers a clear parametric form for each tail, and allows for asymmetry and extrapolation beyond the range of the data. Therefore, it can capture a more accurate estimate of a tail risk. Although EVT is a popular approach in risk management, it is important to note that it assumes that the data is independently and identically distributed (iid). This assumption implies that today’s events are not influenced by events that happened yesterday. However, according to Haugom et al. (2010) and Benterud et al. (2013), most financial return series have signs of stochastic volatility and fattailed distributions. To overcome this issue, McNeil and Frey (2000) suggested the GARCH-EVT method. In the GARCH-EVT method, a suitable GARCH type helps in estimating the volatility of the return series and EVT can be used for the GARCH standardized residuals. This method combines the advantages of GARCH and EVT so we can capture conditional heteroskedasticity and model extreme tail behavior. Bali and Neftci (2003), Bystrom (2005), Fernandez (2005), Chan and Gray (2006), Bhattacharyya and Ritolia (2008), Liu (2011), Emmanouil and Nikos (2012), and Ayusuk and Sriboonchitta (2014) found that the GARCH-EVT model can estimate VaR and CVaR more accurately as compared to the other parametric methods. When the volatility of the return series in a portfolio simultaneously affects each other, it is better to use the multivariate GARCH (M-GARCH) model that is a natural extension of the univariate GARCH model. According to the M-GARCH model, the conditional covariance matrix of the dependent variables has a flexible and dynamic structure. However, in this case the number of parameters is too large and they are not easy to estimate. Bollerslev (1990) proposed a constant conditional correlation GARCH (CCC-GARCH) model in which the conditional correlations are constant. In fact, the assumption of constant conditional correlation over time is not valid. Therefore, Engle (2002) introduced a generalized form of CCC-GARCH that is called the dynamic conditional correlation GARCH (DCC-GARCH) model. The DCC-GARCH model has the flexibility of the univariate GARCH model, and it is not as complex as the M-GARCH model. In addition, Engle (2002) showed that overall this model was the best when compared to several conditional covariances.

3 Determining Conditional Value at Risk (CVaR) …

31

A portfolio includes a set of assets. Since VaR is not sub-additive, the sum of individual VaRs is not equal to the total VaR of a portfolio (Artzner et al. 1999). Consequently, we can consider the portfolio as a multivariate set. Traditional studies have used multivariate normal distribution; however, as mentioned earlier, the returns are asymmetric. In this regard, Sklar (1959) introduced the copula method to produce a realistic and more flexible multivariate distribution. Therefore, a category of arbitrary marginal distributions can be linked with copulas to produce joint distribution. Although the simple copula functions are better than the other dependency structure measurement methods, Aas et al. (2009), Brechmann and Schepsmeier (2012), Dißmann et al. (2013), and Czado et al. (2014) show that they fail to model the dependency between the large number of variables flexibly. Many studies have also tried to increase the flexibility of multivariate copula models among which the Vine copulas are among the best-received ones (Czado et al. 2014). Vine copulas were proposed by Joe (1996) and developed in more detail by Bedford and Cooke (2001, 2002) and Kurowicka and Cooke (2006). Vine copulas allow various structural dependencies of pairs of variables to be modeled properly, in particular with regard to their symmetry, or lack thereof, the strength of the dependency, and tail dependencies (Czado et al. 2014). According to their flexible properties, the use of Vine copula models became popular in financial studies (see Aas and Berg 2009; Benterud et al. 2013; Brechmann and Czado 2011; Czado et al. 2011; Emmanouil and Nikos 2012; Heinen and Valdesogo 2009). The Tehran stock market for agricultural exchange was developed in September 2004. Different agricultural products are traded in spot markets or future markets. Food industry is one of the most important industries in this market. Hence, a modern risk management approach will help the food industry. In this study, we used a DCCGARCH-EVT-Vine copula approach to study the dependency structure across sugar and dairy portfolios. We collected weekly data from the Tehran stock market and computed the two portfolios’ risks using VaR and CVaR techniques. The rest of the paper is organized as follows. Section 3.2 gives the methodology while the results and discussion are presented in Sect. 3.3. Section 3.4 gives the summary and conclusions.

3.2 Methodology A. Extreme Value Theory (EVT ): EVT models the tail of a distribution. In fact, this theory enables us to analyze rare events. In general, the EVT approach embraces two methods known as ‘Block Maxima’ (BM) and ‘Peak Over Threshold’ (POT). The main difference between BM and POT is the identification of extreme events and the application of the principal distribution. The BM method is direct modeling of the maxima (or minima) from equally divided blocks. Fisher and Tippett (1928) showed that the limiting distribution of these extreme events is the generalized extreme value (GEV) distribution. The biggest criticism of the BM method is that only one maximum point is collected from one block. In other words, the BM method does not

32

S. Abedi and E. Pishbahar

use all the extreme information. Therefore, the POT method has become popular in the recent literature. POT considers the realizations that exceed a certain threshold. Let (X n ) be a sequence of iid random variables with unknown marginal F, and Fu be the distribution function of values of X that exceed a certain threshold u. Since available, estimating Fu is difficult. Balkema and de Haan (1974) and Pickands (1975) offer a solution to this problem. They show that for a large number of underlying distribution functions, Fu converges to a generalized Pareto distribution (GPD) (Liu 2011): ⎧   −1/ξ ⎨ 1 − 1 + ξ x−u (if ξ  0) β G ξ,u,β (x)  ) ⎩ 1 − e−( x−u β (if ξ  0)  [u, ∞], (if ξ ≥ 0) with x ∈ (3.1) [u, u − β/ξ ], (if ξ < 0) where G ξ,u,β is the GPD, ξ  1/α is the shape parameter, α is the tail index, and β is the scale parameter. These parameters can be estimated by the maximum likelihood (ML) method. The GPD embeds three distributions. For example, ξ  a −1 > 0 indicates ordinary Pareto distribution which is a heavy-tailed distribution; ξ  0 indicates an exponential distribution and for ξ  a −1 < 0, GPD corresponds to the Pareto II-type distribution. Financial literature shows that financial returns have a heavy-tailed distribution, thus distributions with shape parameter ξ > 0 are more suited (Gong et al. 2015). An important step in applying the POT method is the choice of an appropriate threshold u. According to Balkema and de Haan (1974) and Pickands (1975), u should be high enough so that the limiting distribution converges to GPD. On the other hand, we must have enough observations to estimate the parameters. McNeil and Frey (2000) set the threshold level at the 8th and 92nd percentiles of the data distribution for the lower and upper tails, respectively (Emmanouil and Nikos 2012). B. The DCC-GARCH Model: To satisfy the basic assumption of the EVT approach, we modeled the time-varying volatility of each portfolio using a DCC-GARCH model. A DCC-GARCH model is estimated in two steps. First, an ARMA-GARCH model for each series in a portfolio is estimated as: rt, j  μ j +

p  k1

φk, j rt−k, j + εt, j +

q 

θk, j εt−k, j

k1

εt, j  h t, j z t, j and z t, j ∼ iid(0, 1)

(3.2)

where rt is weekly returns, z t, j is a strong white noise, φk, j rt−k, j is the autoregressive q term, εt, j + θk, j εt−k, j is the moving average term, and h 2t is the conditional standard k1

deviation of εt that satisfies the recurrence equation:

3 Determining Conditional Value at Risk (CVaR) …

h 2t, j  ω j +

p  k1

2 ak, j εt−k, j +

33 q 

βk, j h 2t−k, j

(3.3)

k1

Equation 3.3 is a GARCH(p, q) process and if all βq  0, then Eq. 3.3 is an ARCH(p) process. In Eq. 3.3, a and β determine the strength of the previous error term and volatility, respectively. The second step in the DCC-GARCH estimation is using the standardized residuals of the univariate GARCH to compute the correlation matrix (Engle 2002). According to Bollerslev (1990), the conditional covariance matrix can be decomposed as:   (3.4) Ht  Dt Rt Dt Dt  diag h 2t, j where Rt is a dynamic conditional correlation matrix that must be a positive-definite matrix to invert the covariance matrix, Ht , and Dt are the diagonal matrices of conditional variance. Engle (2002) showed the proxy process to get the positivedefinite as:    Q t  Q + a z t−1 z t−1 − Q + b Q t−1 − Q   (1 − a − b)Q + az t−1 z t−1 + bQ t−1

(3.5)

where a and b are non-negative scalars and a + b < 1 to ensure stationary and positive definiteness of Q t , Q is the unconditional matrix of standardized residuals. a and b are the news and decay terms, respectively. They quantify the model’s sudden reaction to new information and the persistence after a market event. By Eq. 3.5, the correlation matrix is (Benterud et al. 2013): Rt  diag(Q t )−1/2 Q t diag(Q t )−1/2

(3.6)

We estimated the DCC-GARCH model using ML. Since the DCC-GARCH residuals are close to iid, we extracted the DCC-GARCH residuals and applied EVT to them. C. Vine Copula: We applied Vine copula to investigate the dependency structure 2 among portfolio returns. Let F : R → [0, 1] with R  R ∪ {−∞, +∞} be a joint distribution function, with margins F1 , . . . , Fd : R → [0, 1]. Then there exists a 2 copula function in the unit hypercube [0, 1]d , such that for all (x1 , . . . , xd ) ∈ R F(x1 , . . . , xd )  C(F1 (x1 ), . . . , Fd (xd ))

(3.7)

C can be obtained as:  C(u 1 , . . . , u d )  F F1−1 (u 1 ), . . . , F2−1 (u d )

(3.8)

34

S. Abedi and E. Pishbahar

where u 1 , . . . , u d ∈ [0, 1] and Fi−1 is the inverse distribution of the marginal. Given a joint CDF (Eq. 3.7), the density function f (x1 , . . . , xd ) can be obtained by the product of marginal densities, f (xi ), and copula densities, c(F1 (x1 ), . . . , Fd (xd )), as follows: ∂ d F(x1 , . . . , xd ) ∂ x1 . . . ∂ xd d ∂ C(F1 (x1 ), . . . , Fd (xd ))  ∂ x1 . . . ∂ xd d

 c(F1 (x1 ), . . . , Fd (xd )) f i (xi )

f (x1 , . . . , xd ) 

(3.9)

i1

We can select C from a large number of parametric families that have different tail dependencies. Vine copula decomposes the multivariate pdf into pair-copulas. We can select any of the pair-copulas among arbitrary families. Therefore, vines combine the advantages of multivariate copula modeling, which ultimately results in more flexibility in dependence modeling (Czado et al. 2014). Since the decomposition is not unique, and there are lots of pair-copula constructions, Bedford and Cooke (2001, 2002) introduced a graphical model, ‘regular Vine copula’ (R-Vine) to classify each of them. The R-Vine copula with d-elements is a set of linked trees,1 v  {T1 , . . . , Tn−1 }. Tree j has n + 1 − j nodes and n − j edges. Edges in tree j become nodes in tree j + 1, and two nodes in tree j + 1 can be joined by an edge if the corresponding edges in tree j share a node (proximity condition). The class of R-Vines indicates a large number of possible pair-copula decompositions. Generally, they can be grouped in three classes: R-Vines, C-Vines2 , and D-Vines.3 Following Aas et al. (2009), a joint, three-dimensional pdf can be decomposed as: f (x1 , x2 , x3 )  f 1 (x1 ) · f ( x2 |x1 ) · f ( x3 |x1 , x2 )

(3.10)

By Sklar’s theorem, we know that: f (x1 , x2 ) f 1 (x1 ) c12 (F1 (x1 ), F2 (x2 )) · f 1 (x1 ) · f 2 (x2 )  f 1 (x1 )  c12 (F1 (x1 ), F2 (x2 )) · f 2 (x2 )

f ( x2 |x1 ) 

1 Tree

(3.11)

is an acyclic connected graph with nodes and edges, and graph is a set of nodes N and a set of edges E which connect the nodes. 2 Canonical vine: C-Vine trees have a star structure. 3 Drawable vine: D-Vine trees have a path structure.

3 Determining Conditional Value at Risk (CVaR) …

35

and f ( x2 , x3 |x1 ) f ( x2 |x1 ) c 23|1 (F( x2 |x1 ), F( x3 |x1 )) · f ( x2 |x1 ) · f ( x3 |x1 )  f ( x2 |x1 )  c 23|1 (F( x2 |x1 ), F( x3 |x1 )) · f ( x3 |x1 )  c 23|1 (F( x2 |x1 ), F( x3 |x1 ))c13 (F1 (x1 ), F3 (x3 )) · f 3 (x3 )

f ( x3 |x1 , x2 ) 

(3.12)

with h( x|υ, θ )  F( x|υ)

     ∂C xυ j |υ− j F x|υ− j , F υ j υ− j θ    ∂ F υ j υ− j

(3.13)

where υ j is an arbitrary element of υ and υ− j denotes the (d − 1)-dimensional vector υ excluding υ j . Equation 3.10 can be rewritten as: Marginal

   f (x1 , x2 , x3 )  f 1 (x1 ) · f 2 (x2 ) · f 3 (x3 ) Unconditional pairs

   × c12 (F1 (x1 ), F2 (x2 )) · c13 (F1 (x1 ), F3 (x3 )) Conditional pairs

   × c 23|1 (F( x2 |x1 ), F( x3 |x1 ))

(3.14)

Since there are a significant number of R-Vines, Chollete et al. (2009), Emmanouil and Nikos (2012), Sriboonchitta et al. (2014), and Ayusuk and Sriboonchitta (2014) suggest that C-Vine copulas dominate the other dependence structures. C-Vines are characterized by a root node in each tree. A root node is connected to all the other nodes of the tree (Czado et al. 2014). The C-Vine density with the root nodes 1, …, d, can be determined as: f (x)  

d

k1

f k (xk )

d−1 d−i

c i,i+ j|1:(i−1)

i1 j1

   F( xi |x1 , . . . , xi−1 ), F x j+1 x1 , . . . , xi−1 θ i,i+ j|1:(i−1)

(3.15)

For using a C-Vine copula, it is necessary to determine the root node in each tree. Therefore, we need to compute the empirical Kendall’s Tau for all possible variable pairs and then select the spanning tree that maximizes the sum of absolute empirical Kendall’s τ . For each edge in the selected spanning tree, select a copula and estimate the parameters using joint maximum likelihood (JML). Applying Eq. 3.13, create

36

S. Abedi and E. Pishbahar

pseudo-observations. For tree 2 until tree d − 1, iterate the mentioned steps. We chose the optimal copula based on the minimum values of AIC and BIC. D. Value at Risk and Conditional Value at Risk: To compute risk estimates for each portfolio with M return series, we used VaR and CVaR. VaR is defined as the p-th quintile of distribution F as: VaR p  F −1 (1 − p)

(3.16)

where F −1 is the inverse of the distribution F. CVaR as a coherent measure indicates the excess loss over the threshold VaR. CVaR can be defined as (Liu 2011): CVaR p  E( r |r > VaR p )

(3.17)

To measure VaR and CVaR, we used the following steps: 1. Fitting a DCC-GARCH for each portfolio and extracting DCC-GARCH standardized residuals. 2. Determining the threshold for the upper and lower tails of residuals’ distribution and assuming excess residuals over the determined threshold following GPD. 3. Transforming the DCC-GARCH standardized residuals to copula data by empirical marginal distribution functions and fitting a C-Vine copula for them. 4. Generating 10,000 uniform random variables from the estimated C-Vine copula and converting them to standardized residuals using the suitable inverse distribution. 5. Using the simulation procedure to generate the dependent return series and simulating the portfolio return, rˆt , using: rˆt  ω1rˆt,1 + · · · + ω M rˆt,M

(3.18)

where ω j is portfolio weights, j  1, …, M. Computing VaR and CVaR by Eqs. 3.16 and 3.17, respectively.

3.3 Results and Discussion We considered two important food industry companies’ stock portfolios on the Tehran stock market (dairy and sugar). The selection of these industries was based on their high levels of trading interest and liquidity. The dairy portfolio’s stock dairy companies that we considered are Kalber Dairy Company (KD), the Pak Dairy Company (PD), the Khorasan Pegah Dairy Company (KPD), and the Isfahan Pegah Dairy Company (IPD). The sugar portfolio companies are Ghazvin Sugar Company (GS), the Sabet Khorasan Sugar Company (SKS), the Marvdasht Sugar Company (MS), and the Hegmatan Sugar Company (HS). We collected weekly data from the Tehran stock agricultural exchange market for the period 2006:M1–2016:M12. In total, there were

3 Determining Conditional Value at Risk (CVaR) …

37

Table 3.1 Data descriptive statistics for portfolio include ‘stocks dairy companies’ (dairy portfolio) Min

KD

PD

KPD

IPD

−62.12

−41.61

−43.65

−46.61

Mean

−0.066

Max

15.17

18.41

17.22

28.43

S.D.

3.81

2.51

2.99

3.03

0.0066

0.009

0.048

Skewness

−6.65

−8.29

−5.73

−5.24

Kurtosis

152.48

157.78

97.68

132.72

JB (p-value)

469,190.84 (0.00)

504,828.02 (0.00)

189,492.38 (0.00)

352,856.43 (0.00)

ADF unit-root test (p-value)

−21.51 (0.00)

−20.52 (0.00)

−20.75 (0.00)

−20.64 (0.00)

Table 3.2 Data descriptive statistics for portfolio include ‘stocks sugar companies’ (sugar portfolio) GS

SKS

MS

HS

−30.78

−62.29

−66.29

−63.87

Mean

−0.03

−0.13

−0.09

−0.14

Max

39.17

13.25

21.56

16.30

Min

S.D.

3.03

3.29

3.61

3.49

Skewness

1.04

−13.21

−12.58

−12.71

Kurtosis

85.91

255.74

232.45

299.71

JB (p-value)

143,299.89 (0.00)

1,345,323.4 (0.00)

1,110,006.8 (0.00)

1,847,562.5 (0.00)

ADF unit-root test (p-value)

−21.58 (0.00)

−21.02 (0.00)

−21.15 (0.00)

−21.01 (0.00)

500 weekly price observations for each series of dairy and sugar portfolios. The portt × 100. Tables 3.1 and 3.2 give the folio return series were generated by rt  log PPt−1 descriptive statistics for weekly returns in the dairy and sugar portfolios, respectively. As shown in Tables 3.1 and 3.2, the mean returns are close to zero, while the mean returns in the dairy portfolio are mostly positive. In both the portfolios, the maximum and minimum returns are extreme. The return series in both the portfolios are asymmetric, and most of them have negative skewness. All return series exhibit evidence of fat tails (Kurtosis > 3), and thus, they are leptokurtosis. In addition, the Jarque–Bera test rejected the null hypothesis of normality for each series in both the portfolios. The augmented Dickey–Fuller test rejected the unit-root assumption for all series, thus all return series were stationary. We modeled the time-varying volatility of each portfolio by DCC-GARCH. The DCC-GARCH results for the dairy and sugar portfolios are presented in Table 3.3. According to this table, in ARCH models all α’s are less than 0.1, so volatility is

38

S. Abedi and E. Pishbahar

Table 3.3 DCC-GARCH estimation results DCC-GARCH estimation results for the dairy portfolio KD

PD

KPD

IPD

ω

0.084** (0.032)

3.00E−06 (0.002)

0.004 (0.002)

1E−06 (7E−06)

ARCH(α)

0.025** (0.009)

0.001** (6.89E−04)

0.001* (9.37E−34)

0.002*** (3.56E−04)

GARCH(β)

0.922*** (0.001)

0.992*** (1.87E−04)

0.993*** (7.24E−04)

0.992*** (0.001)

DCC(a)

0.005** (0.003)

Log-likelihood

−3508.45

AIC

14.198

BIC

14.543

DCC(b)

0.984*** (0.004)

DCC-GARCH estimation results for the sugar portfolio GS

SKS

MS

HS

ω

0.005 (0.034)

0.155* (0.106)

0.029 (0.058)

0.013*** (7.41E−04)

ARCH(α)

0.004 (0.005)

0.042 (0.047)

0.002 (0.003)

0.044*** (7.47E−04)

GARCH(β)

0.99*** (0.008)

0.893*** (0.002)

0.976*** (0.006)

0.913*** (6.45E−04)

DCC(a)

0.068*** (0.013)

Log-likelihood

−3539.89

AIC

14.324

BIC

14.669

DCC(b)

0.915*** (0.016)

Note Standard error is in parentheses *, ** and *** mean significant levels at 10, 5, and 1%, respectively

not sensitive to market events, but all β values are above 0.9 in the GARCH models showing that volatility takes a long time to die. DCC(a) and DCC(b) are statistically significant and non-negative in both the cases. This indicates that the constant conditional correlation assumption is not supported. However, a + b < 1 is satisfied in both the models. The short-run persistence of special events on the dynamic conditional correlations is greatest for the sugar portfolio at 0.068, while the largest long-run persistence of events in the market to conditional correlations is 0.989 (0.005 + 0.984) for the dairy portfolio. Figures 3.1 and 3.2 show the dynamic correlations of the return series in dairy and sugar portfolios, respectively. According to Figs. 3.1 and 3.2, the dynamic correlation of returns in the sugar portfolio is time-varying and volatile while they are not more volatile and also all of them show positive correlation in the dairy portfolio. Box-plots represent a range of 70% observations of correlation. In the dairy portfolio, the largest and smallest

3 Determining Conditional Value at Risk (CVaR) …

39

Fig. 3.1 Dynamic conditional correlations of the return series in the dairy portfolio

correlation varying ranges are related to KD and PD and PD and KPD, respectively. The correlation between returns is relatively stable in the short-run. In Fig. 3.2, the dynamic correlation is only positive for SKS and HS. The dynamic correlation of MS and SKS has the largest varying range, while MS and HS have the smallest varying range. We applied the POT method and fit the GPD to the residuals’ series that exceed the threshold. We set the threshold at the 8th and 92nd percentiles of the residuals’ distribution. Table 3.4 shows the threshold values, u and the estimation results of the GPD parameters. There are 72 observations for the dairy portfolio residuals’ series and 79 observations for the sugar portfolio. The estimated parameters are mostly significant. The estimated tail index values range between 0.02 (PD, lower tail) and 0.97 (KD, lower tail) for the dairy portfolio, and between 0.51 (MS, upper tail) and 1.04 (GS, lower tail) for the sugar portfolio. A higher value of the estimated lower tail index is observed for KD in the dairy portfolio and for GS, MS and HS in the sugar portfolio. Therefore, their returns have a negative skew. This indicates that high losses are more likely than high rewards. For others, high positive returns are more likely. Since all the shape parameters are positive, we can say that all our returns have heavy-tailed distributions. This gives evidence of extreme events in the two markets. Our results support the idea of financial returns distribution.

40

S. Abedi and E. Pishbahar

Fig. 3.2 Dynamic conditional correlations of the return series in the sugar portfolio

We used the C-Vine copula to model the dependence structure among portfolio returns. First, we transformed the standardized residual series to uniform copula data. Then, we selected the tree structure of the C-Vine. In fact, we had to determine the root node for each tree. For different pairs, we calculated the empirical Kendall’s Tau, and we compared 12 types of C-Vines. Next, we selected the order of the variables in both the portfolios. The order in the dairy portfolio was: KPD, PD, KD, and IPD, and in the sugar portfolio it was KS, MS, GS, and HS. The selection of suitable copulas and estimation of the corresponding parameters’ results are presented in Tables 3.5 and 3.6 for the dairy and the sugar portfolios, respectively. We examined different types of copula functions for each pair and selected the best ones based on minimum AIC/BIC scores. Our results show that copula families are far from normality in the dairy portfolio and just one of them is Gaussian in the sugar portfolio. Most of the estimated copula parameters are significant. Copulas are from different families, and so they are difficult to compare. Therefore, we computed Kendall’s τ coefficient in which τ  0 indicates an independent structure. Our results show that the residuals are dependent in both the portfolios. With attention to first tree results in the dairy portfolio, KD and KPD and IPD and KPD adopted asymmetric copulas with lower tail dependencies, while PD and KPD had symmetric dependence. However, bivariate margins adopted symmetric copulas

3 Determining Conditional Value at Risk (CVaR) …

41

Table 3.4 Threshold values and ML GPD parameters estimation The dairy portfolio Series

Lower tail

Upper tail

u

ξ

β

U

ξ

β

KD

−1.19

0.97 (0.40)

0.39 (0.16)

1.26

0.63 (0.29)

0.43 (0.14)

PD

−1.01

0.02 (0.19)

0.92 (0.23)

1.02

0.30 (0.19)

0.71 (0.17)

KPD

−1.07

0.47 (0.27)

0.49 (0.15)

0.93

0.68 (0.27)

0.34 (0.10)

IPD

−0.78

0.39 (0.27)

0.65 (0.21)

0.79

0.52 (0.22)

0.56 (0.14)

u

ξ

Series

U

ξ

Series

GS

−0.71

1.04 (0.34)

0.31 (0.1)

0.75

0.97 (0.33)

0.27 (0.09)

SKS

−1.12

0.59 (0.22)

0.61 (0.15)

1.02

0.81 (0.29)

0.34 (0.11)

MS

−1.13

0.69 (0.22)

0.49 (0.12)

1.07

0.51 (0.21)

0.47 (0.12)

HS

−1.15

0.73 (0.24)

0.65 (0.17)

1.33

0.64 (0.27)

0.78 (0.23)

The sugar portfolio Series

Lower tail

Upper tail

Table 3.5 Copula selection and parameters’ estimation for the dairy portfolio Tree number

Edges

Selected family

Coeff.

Standard Lower and error upper tail dependence

Kendall’s τ

First

KD, KPD

Gumbel 180

1.53

1.28

(0, 0.42)

0.35

PD, KPD

Joe-Clayton (BB7)

1.38 0.46

0.005 0.02

(0.23, 0.34)

0.3

Second

Third

IPD, KPD

Joe

1.98

0.001

(0, 0.58)

0.35

KD, PD|KPD

Joe 180

1.36

0.15

(0.34, 0)

0.17

IPD, PD|KPD

t

0.24 2

0.13 0.001

(0.27, 0.27)

0.15

KD, IPD|PD, KPD

Joe 180

1.07

0.11

(0.09, 0)

0.04

Log-likelihood  88.42, AIC  −160.84, BIC  −142.62

42

S. Abedi and E. Pishbahar

Table 3.6 Copula selection and parameters’ estimation for the sugar portfolio Tree number

Edges

Selected family

Coeff.

Standard Lower and error upper tail dependence

Kendall’s τ

First

MS, KS

Joe-Frank (BB8)

2.27 0.89

0.07 0.001

(0, 0)

0.31

GS, KS

Joe-Frank (BB8)

2.89 0.84

0.64 0.002

(0, 0)

0.36

KS, HS

Joe-Frank (BB8)

2.21 0.96

0.67 0.001

(0, 0)

0.35

MS, HS|KS

Gaussian

0.15

0.08 0.002

(0, 0)

0.1

GS, MS|KS

t

0.096 2.65

0.6

(0.16, 0.16)

0.06

GS, HS|MS, KS

t

0.004 10.18

0.03 0.001

(0.006, 0.006)

0.01

Second

Third

Log-likelihood  63.1, AIC  −104.01, BIC  −77.94

Fig. 3.3 C-Vine tree structure for dairy portfolio

in the sugar portfolio. The C-Vine tree structures for the dairy and sugar portfolios are given in Figs. 3.3 and 3.4, respectively. The tree structures indicate the joint density function of the residuals’ series. Each edge includes two elements in which the first element shows the selected copula family and the second copula is Kendall’s τ . We used the C-Vine results by applying the Mote Carlo simulation for both portfolios to simulate 10,000 uniform copula observations. After the data generation process, we converted them into the returns form and then measured the risk for each equally weighted portfolio using VaR and CVaR. Table 3.7 shows the VaR and CVaR for both portfolios at two different significance levels (97.5 and 99%). The maximum possibility loss of the dairy portfolio’s value over a week is 3.4% at the 99% significance level, while for the sugar portfolio it is 4.6%. CVaR is computed as 5.1 and 7.1% at the 99% significance level, respectively, for the dairy and sugar portfolios. This implies that the expected loss of 5.1 and 7.1% will be exceeded the VaR at the 99% significance level. Therefore, the sugar portfolio is riskier than the dairy portfolio.

3 Determining Conditional Value at Risk (CVaR) …

43

Fig. 3.4 C-Vine tree structure for sugar portfolio Table 3.7 VaR and CVaR estimation in each portfolio

Dairy portfolio

Sugar portfolio

VaR0.99

0.034

0.046

VaR0.975

0.017

0.019

CVaR0.99

0.051

0.071

CVaR0.975

0.027

0.030

3.4 Summary and Conclusion This study investigated the portfolio risk for two important food industries (dairy and sugar) in Iran for the period 2006:M1–2015:M12. It applied the DCC-GARCH method to model the time-varying volatility in both the portfolios. Both portfolios’ returns were volatile as Engle (2002) showed they indicated that the DCCGARCH model was appropriate for time-varying modeling. We also extracted the DCC-GARCH residuals to investigate the EVT approach to model the tail behavior of residual distributions. Most of the returns had heavy-tailed distributions so the two markets experienced extreme price events. These results support the idea that financial returns’ distribution is fat-tailed. The study used the C-Vine copula to analyze the dependence of returns in both portfolios. The results show that all returns in the two portfolios were dependent and their copula families were far from normality. As a result, as Liu (2011), Emmanouil and Nikos (2012), and Ayusuk and Sriboonchitta (2014) show, applying the Vine copula provides more flexibility in describing joint distribution. Finally, we used the C-Vine copula by applying the Monte Carlo simulation to simulate the returns series to measure VaR and CVaR in both the portfolios. In the sugar portfolio, VaR and CVaR were more than they were in the dairy portfolio. As mentioned in the DCC-GARCH model’s results, the sugar portfolio was more volatile as compared to the dairy portfolio. Further, the EVT results show that most of the sugar portfolio’s returns exhibited a negative skew since the tail index for the lower tail is higher than the upper tail. Therefore, extreme losses are more likely than extreme rewards in the sugar portfolio. These statistical results show that the sugar portfolio is riskier

44

S. Abedi and E. Pishbahar

than the dairy portfolio. Consequently, we suggest that investors should pay more attention to the dairy portfolio.

References Aas K, Berg D (2009) Models for construction of multivariate dependence: a comparison study. Eur J Finance 15:639–659 Aas K, Czado C, Frigessi A, Bakken H (2009) Pair-copula constructions of multiple dependence. Insur Math Econ 44:182–198 Artzner P, Delbaen F, Eber J, Heath D (1999) Coherent measures of risk. Math Finance 9:203–228 Ayusuk A, Sriboonchitta S (2014) Risk analysis in Asian emerging markets using canonical vine copula and extreme value theory. Thai J Math 59–72 Bali TG (2003) An extreme value approach to estimating volatility and value at risk. J Bus 76:83–107 Bali TG, Neftci SN (2003) Disturbing external behavior of spot price dynamics. J Empir Finance 10:455–477 Balkema AA, de Haan L (1974) Residual life time at great age. Ann Probab 2(5):792–804 Bedford T, Cooke RM (2001) Probability density decomposition for conditionally dependent random variables modeled by vines. Ann Math Artif Intell 32:245–268 Bedford T, Cooke RM (2002) Vines: a new graphical model for dependent random variables. Ann Stat 30:1031–1068 Benterud JL, Haukaas MS, Huse PI (2013) Risk modeling using vine copulas. Modelling an energy company portfolio. Master thesis, Norwegian University of Science and Technology Bhattacharyya M, Ritolia G (2008) Conditional VaR using EVT-towards a planned margin scheme. Int Rev Financ Anal 17:382–395 Bollerslev T (1990) Modeling the coherence in short-run nominal exchange rates: a multivariate generalized ARCH model. Rev Econ Stat 72:498–505 Brechmann EC, Czado C (2011) Risk management with high-dimensional vine copulas: an analysis of the euro Stock 50. Working paper. Technische Universität München Brechmann EC, Schepsmeier U (2012) Modeling dependence with C- and D-vine copulas: the R-package CDVine. J Stat Softw (To appear) Bystrom H (2005) Extreme value theory and extremely large electricity price changes. Int Rev Econ Finance 14:41–55 Chan KF, Gray P (2006) Using extreme value theory to measure value-at-risk for daily electricity spot prices. Int J Forecast 22:283–300 Chollete L, Heinen A, Valdesogo A (2009) Modeling international financial returns with a multivariate regime-switching copula. J Financ Econometrics 7:437–480 Czado C, Schepsmeier U, Min A (2011) Maximum likelihood estimation of mixed C-vines with application to exchange rates. Stat Model (To appear) Czado C, Brechmann EC, Gruber L (2014) Selection of vine copulas. Technische Universitat Munchen Dißmann J, Brechmann EC, Czado C, Kurowicka D (2013) Selecting and estimating regular vine copula and application to financial returns. Comput Stat Data Anal 59:52–69 Emmanouil KN, Nikos N (2012) Extreme value theory and mixed canonical vine copulas on modeling energy price risks. Cass Business School, City University London Engle R (2002) Dynamic conditional correlation: a simple class of multivariate generalized autoregressive conditional heteroskedasticity models. J Bus Econ Stat 20:339–350 Fernandez V (2005) Risk management under extreme events. Int Rev Financ Anal 14:113–148 Fisher RA, Tippett LHC (1928) Limiting forms of the frequency distribution of the largest or smallest member of a sample. Proc Camb Philos Soc 24:180–290

3 Determining Conditional Value at Risk (CVaR) …

45

Gong X, Sriboonchitta S, Rahman S (2015) Modeling value at risk of agricultural commodity while accounting for seasonality and weather using extreme value theory. In: The international conference on Economics and Business Administration (EBA), Barcelona, Spain, 7–9 Apr 2015 Haugom E, Westgaard S, Solibakke P, Lien G (2010) Modeling day ahead Nord pool forward price volatility: realized volatility versus GARCH models. In 7th international conference on the European Energy Market (EEM), pp 1–9 Heinen A, Valdesogo A (2009) Asymmetric CAPM dependence for large dimensions: the canonical vine autoregressive copula model. Core discussion paper 2009069. Universite Catholique de Louvain, Center for Operations Research and Econometrics (CORE) Joe H (1996) Families of m-variate distributions with given margins and m(m-1)/2 bivariate dependence parameters. Institute of Mathematical Statistics, Hayward Kurowicka D, Cooke RM (2006) Uncertainty analysis with high dimensional dependence modelling. Wiley, Hoboken Liu J (2011) Extreme value theory and copula theory: a risk management application with energy futures. Ph.D. thesis, University of Victoria Marimoutou V, Raggad B, Trabesi A (2009) Extreme value theory and value at risk: application to oil market. Energy Econ 31:519–530 McNeil A, Frey R (2000) Estimation of tail-related risk measures for heteroscedastic financial time series: an extreme value approach. J Empir Finance 7:271–300 Pickands J (1975) Statistical inference using extreme order statistics. Ann Stat 3:119–131 Sklar A (1959) Fonctions de répartition à n dimensions et leurs marges. Publications de l’Institut de Statistique de l’Université de Paris 8:229–231 Sriboonchitta S, Liu J, Wiboonpongse A (2014) Vine copula-cross entropy evaluation of dependence structure and financial risk in agricultural commodity index returns. Modeling dependence in econometrics. Springer, Berlin, pp 275–287

Sahar Abedi holds B.Sc. and M.Sc. in Agricultural Economics from Department of Agricultural Economics, University of Tabriz. For her MA Thesis, she aimed to have a better understanding of weather-based crop insurance premium for wheat crop. Her research interest lies at Risk Management and Crop insurance. Esmaeil Pishbahar is Associate Professor and Head of Department of Agricultural Economics at University of Tabriz, Iran. He holds a B.Sc. in Agricultural Economics from University of Tabriz and a M.Sc. in Agricultural Economics from University of Tehran. He did his Ph.D. in Science Economics at departments of Economics and Management, University of Rennes1, France. His areas of interest and research are Applied Econometrics, Agricultural Risk Management and Insurance, and International Trade. His teaching area are Advanced Econometrics, Mathematical Economics, and Macroeconomics at under- and postgraduate levels. He has over 100 publications in journals and chapters in books.

Chapter 4

Assessing Climate Change Impacts on Land Use in Iran: The Spatial Fractional Multinomial Logit Modeling Approach Khadijeh Alefi and Mohammad Ghahremanzadeh

4.1 Introduction Researchers consider climate change as one of the main environmental problems of the twenty-first century. According to the Intergovernmental Panel on Climate Change’s (IPCC) fourth assessment report, the global mean surface temperature increased by 0.74 ± 0.18 °C in the last century and it is expected to increase by 1.1–6.0 °C in this century (Reidsma et al. 2010). Besides the increase in surface temperature, global patterns of rainfall have also changed. Both observations and model simulations show an increase in the global average mean precipitation and its variations; on average, annual precipitation has increased in wet areas and has decreased in dry areas. Precipitation inequalities have increased between different regions of the world (Chou and Lan 2012). Along with temperature and precipitation, other climatic variables have also changed. These changes are both a developmental and an environmental challenge and can be a fundamental threat to agricultural productivity, food security and development prospects (Di Falco 2014). Given the fundamental role of agriculture in human welfare, concerns have been raised about the potential effects of climate change on agriculture and its sub-sectors. Hence, many organizations consider the effects of climate change on agriculture and a substantial body of research is devoted to this subject (Adams et al. 1998). According to studies, agriculture is vulnerable to climate change which shows the necessity of implementing effective adaptation strategies for minimizing its damages on this sector (Niles et al. 2015). Producing crops which are consistent with new climatic conditions and changing cropping patterns are among the adaptation strategies that K. Alefi · M. Ghahremanzadeh (B) Department of Agricultural Economics, University of Tabriz, Tabriz, Iran e-mail: [email protected] K. Alefi e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2019 M. Rashidghalam (ed.), Sustainable Agriculture and Agribusiness in Iran, Perspectives on Development in the Middle East and North Africa (MENA) Region, https://doi.org/10.1007/978-981-13-6283-5_4

47

48

K. Alefi and M. Ghahremanzadeh

can be interpreted as land use changes. Agricultural conservation and providing stable food security rely on a thorough understanding of these changes and studying the factors affecting land allocation decisions for specific crops (Allen 2014). Existing studies that explore the effects of climate change on land allocation mostly use statistical methods. They often estimate limited dependent variable regression models. One of the recently considered models is the fractional multinomial logit model. Yang (2010) studied the effects of different market variables, commodity and conservation programs’ payments and county physical attributes on acreage allocation among crops including corn, soy legumes, wheat and hay and the conservation reserve programs in nine west states in the USA. He used panel data for the period 1996–2006 to estimate the fractional multinomial logit model. His estimation results suggest that crops’ profits, conservation reserve program (CRP), rental payment and the physical characteristics of the land collectively determined the acreage allocated among alternative crops. Mu and McCarl (2011) examined possible climate adaptations in the US land use and livestock context. They estimated fractional multinomial logit models using agronomic and weather data for 1978, 1992, 1997, 2002 and 2007. Their results confirm that as temperature and precipitation increase, producers respond by reducing crop land and increasing pasture land. Mudzonga (2011) studied factors that affect farmers’ adaptation to climate change in Chivi district of Zimbabwe using cross-sectional data for December 2010–January 2011 collected through a household survey. His results using the fractional multinomial logit model confirmed that farm household size, access to credit, farming experience and exposure to information about climate change all positively and significantly influenced farmers’ decisions to adapt to climate change in the district. Allen (2014) studied the land allocation determinants in Mali’s Koutiala production zone. He estimated the fractional multinomial logit econometric models for cotton, maze, sorghum, millets and secondary crops. According to his results, in villages with better access to markets, land was mostly allocated to millets. Kaminski et al. (2013) investigated land allocation changes considering technical changes in Israel’s 54 natural regions as a climate change adaption solution using a structural model. The main crops that they considered were vegetables, field crops, flowers and orchards. They estimated one spatial multinomial logit model for each structural equation. According to their results, long-term losses stemmed from potential reduction in yields driven by forecasted increases in temperature. Li et al. (2013) investigated the main factors of land use in China for the period 1988–2005 by assembling the GIS database. They used the spatial fractional multinomial logit model which explicitly takes into account spatial interactions between land allocation decisions. According to the results of their study, high value of urban land was a main driver of farmland development. On the other hand, increasing rural incomes were a major driver of conversion of farmland to forests and grassland. Cho et al. (2014) examined effective climatic and economic factors for land allocations between barley, corn, cotton, upland cotton, sorghum, soy legumes, winter wheat, durum wheat and spring wheat in US counties using the fractional multinomial logit model. They concluded that upland cotton, rice, sorghum, soy legumes and winter wheat were more likely chosen when the temperature increased. On the

4 Assessing Climate Change Impacts on Land Use in Iran …

49

other hand, barley, corn, spring wheat and durum wheat were less likely to be chosen when the temperature went down. Cho et al. (2015) used a spatial fractional multinomial logit model to estimate the drivers of transitions in US land use over the 48 US counties for 2002–2007 and 2007–2012. They considered transitions between different land uses. Their results show that climate change adaptation via change in land usage is an ongoing process but with spatial dependencies. Most of these studies employed limited dependent variable models, specifically the fractional multinomial logit model. Li et al. (2013) and Cho et al. (2015) imported spatial effects to the model and employed the spatial econometric method. Their results confirm that climate change affected the management of land allocation between different activities. This was more obvious in agricultural land. Iran is significantly dependent on agriculture as it provides 12% of its gross domestic production (GDP) and 21.2% of the employment. About 16 million hectares are devoted to agricultural activities (Statistical Center of Iran 2016). Considering the limited supply of land, there is no possibility of using more land for developing agriculture in the future. Hence, farmers are forced to allocate limited land to different competing activities. Land allocation among these activities is done to achieve the objectives of the agricultural sector and is in line with regional conditions. Among these conditions are climatic factors. These factors change profitability through changing yields, thus controlling farmers’ decisions. Iran is a vast country with a 1,648,195 km2 area located between latitudes 24° and 40°N, longitudes 44° and 64°E. It has high climatic diversity, and the climatic variables have high variability in various regions. This has led to cultivation of diverse crops. In recent years, climate change has altered the climatic variables in various areas which show the need for examining their impact on the agricultural land management. Large climatic diversity and the consequent diversity in agricultural activities provide an appropriate context for studying the ways in which climatic variables affect land allocations between different activities. However, not enough studies have been done in this field which shows that future research should focus on this. Annual crops form the main agricultural activity in Iran. They have been allocated about 12.2 million hectares of cultivation area and are classified into six groups—cereals, legumes, industrial crops, vegetables, cucurbits and forages. These groups’ cultivation areas and land shares are different in various regions. Weather differences are one of the reasons for this. This provides a suitable context for studying the impact of climatic variables on these groups’ land shares. Studying the effects of climate change on annual crop groups’ land share can help in making suitable policies to adapt to climate change by providing information on management and planning. Agriculture can provide food security in the future by producing sufficient food for people. So, this study examines the effects of climatic variables on land allocations between annual crop groups in Iran using the spatial fractional multinomial logit model. These crops are also important from the demand side as their raw or processed consumption includes about 7.2% of the expenditure of urban households in Iran (Central Bank of Iran 2016). The rest of this study is organized as follows. Next section reviews the recent literature. Section 4.2 provides methodology of the study which is followed by data

50

K. Alefi and M. Ghahremanzadeh

description in Sect. 4.3. Section 4.4 discusses the empirical results, and Sect. 4.5 summarizes and gives the conclusion.

4.2 Methodology This study investigates land allocation between annual crop groups based on Li et al.’s (2013) model. The basic assumption of this model is that land allocation between alternative activities is based on maximizing the present discounted value of the stream of expected net benefits from land (π ). Thus, a parcel of land i is allocated to activity j in period t if and only if πi jt  πikt (∀k  j) is satisfied. The profit term decomposes into a deterministic component (π¯ i jt ) and a random error term (εi jt ): πi jt  π¯ i jt + εi jt

(4.1)

where π¯ i jt represents the expected average net benefits of allocating a parcel of land in area i to land use j at time t, and εi jt shows the deviation from the average net benefit. εi jt is often assumed to follow a normal or type-I extreme value distribution. If it is assumed that εi jt follows the type-I extreme value distribution, a standard multinomial logit model of land allocation is derived. The multinomial logit model ignores the spatial interactions of land use on neighboring parcels that this can lead to biased (or inconsistent) estimates by inducing heteroskedastic errors. Taking into account the space effect, Eq. 4.1 is rewritten as (Li et al. 2013): πi jt 



ρ jt wim πi jt + π¯ i jt + εi jt

(4.2)

im

  where ρ jt is a spatial autoregressive parameter (ρ jt  < 1) and represents the degree to which the propensity to have land use j in one parcel is affected by the propensity to have land use j in neighboring areas (Li et al. 2013). wim shows the spatial contiguity matrix’s elements reflecting the spatial relationship between land areas i and m. According to spatial econometrics literature, the diagonal elements of the spatial contiguity matrix are 0 (wii  0). The specified model in Eq. 4.2 is known as the spatial autoregressive model that can be written in the stacked form as:  jt  ρ jt W  jt + X t−1 χ jt + ε jt

(4.3)

where  jt  (π1 jt , . . . , π N jt ) , X t−1  (x1t−1 , . . . , x N t−1 ) and ε jt  (ε1 jt , . . . , ε N jt ) . The reduced form of Eq. 4.3 can be written as  jt  (I N − ρ jt W )−1 X t−1 χ jt + (I N − ρ jt W )−1 ε jt , that I N shows the N-dimensional identity matrix. By expanding the (I N − ρ jt W )−1 term, the equation can be written as:  jt  X t−1 χ jt + ρ jt W X t−1 χ jt + ρ 2jt W 2 X t−1 χ jt + · · ·. If ψ jt is identified as ψ jt ≡ I N − ρ jt W , the variance–covariance matrix of  jt will be proportional to

4 Assessing Climate Change Impacts on Land Use in Iran …

51

[(ψ jt ) (ψ jt )]−1 . If σi2jt be the diagonal elements of the [(ψ jt ) (ψ jt )]−1 matrix and ∗∗ −1 ∗ new relations are defined as xi∗jt−1  xit−1 σi−1 jt and X jt−1  (ψ jt ) X jt−1 and εi jt follows distribution similar to the εi jt distribution in the standard multinomial logit model, the predicted fraction or share of land allocated to crop group j in i area at time t ( pi jt ) can be expressed as:  exp(xi∗∗jt−1 χ jt ) pi jt  E( si jt xi∗∗jt−1 )   ∗∗ j exp(x i jt−1 χ jt )

(4.4)

Equation 4.4 defines a spatial multinomial logit regression model. If the sample size (N) is large, it is not feasible to estimate the model using a traditional maximum likelihood method because the likelihood function involves a N-dimensional integration. So to estimate the spatial multinomial logit regression model, Li et al. (2013) employed a two-step estimation method that was first introduced by Klier and McMillen (2008) to estimate the binary spatial logit model. In the first step, the fractional multinomial logit model’s parameters are estimated using the maximum likelihood (ML), and in the next step, the spatial fractional multinomial logit model’s parameters are extracted using the generalized method of moments (GMM). Considering Eq. 4.4, the gradient terms can be extracted as:

∂ pi jt ∂ρkt

∂ pi jt  [δ(k  j) − pi jt ] pikt xi∗∗jt−1 ∂χkt xi∗∗jt−1 χkt  [δ(k  j) − pi jt ] pikt [(Hkt )i χkt − (kt )ii ] 2 σikt

(4.5) (4.6)

where δ(k  j) is an indicator function which equals 1 when k  j and zero otherwise. ∗∗ and kt  (I N −ρkt W )−1 W (I N −ρkt W )−1 (I N − Also Hkt  (I N −ρkt W )−1 W X kt−1 ρkt W  )−1 . Considering these equations, it will be kt  W if ρkt  0. In this case, (kt )ii  0 because wii  0. Hence, when ρkt  0∀k, t, Eqs. 4.5 and 4.6 can be written as: ∂ pi jt  [δ(k  j) − pi jt ] pikt xit−1 ∂χkt

(4.7)

∂ pi jt  [δ(k  j) − pi jt ] pikt (W X t−1 )n χkt ∂ρkt

(4.8)

Now if θt  (χt , ρt ) , where χt  (χ1t , . . . , χ J −1t ) and ρt  (ρ1t , . . . , ρ J −1t ), the gradient terms can be defined as gi jt  ∂ pi jt /∂θt . The generalized residuals can be extracted as: u i jt  si jt − pi jt

(4.9)

52

K. Alefi and M. Ghahremanzadeh

where si jt is the share of land use j in area i at time t. Linearizing the generalized residuals equation around the initial estimates (θt0  (χt0 , ρt0 ) ) u i jt ≈ u i0jt − gi jt (θt − θt0 ) can be extracted as: u i0jt + gi jt θt0 ≈ gi jt θt + u i jt

(4.10)

The spatial fractional multinomial logit model is estimated using Eqs. 4.7–4.10, and following the steps: first, Eq. 4.4 is estimated assuming ρt  0 (the standard fractional multinomial logit model) and using the maximum likelihood method. Employing the estimated parameters of this model (χˆ t0 ), primary estimates of matrix θt0 are extracted as θt0  (χˆ t0 , 0) . Then, using Eqs. 4.7–4.10, gradient terms (gi jt ), the generalized residuals (u i0jt ) and u i0jt + gi jt θt0 are calculated. Second, the G jt  (g1 jt , . . . , g N jt ) variables are regressed on instrumental vari

ables X t−1 , W X t−1 , . . . , W 5 X t−1 and their predicted values are calculated (G jt ). Finally, the [(u 011t + g11t θt0 ), . . . , (u 0N J −1t + g N J −1t θt0 )] variables are regressed on 







(G 1t , . . . , G J −1t ) variables. In this step, the estimated parameters are the same parameters of θt (the spatial fractional multinomial logit model’s parameters) which are expressed as θˆt  (χˆ t , ρˆt ) . After the model’s estimation, the marginal effects of the explanatory variables on expected alternative land shares can be calculated using Eq. 4.11 where ◦ is an element-by-element product operator (Cho et al. 2015):  ∂ pi jt  pi jt (χ jt /σi jt ◦ (I N − ρ jt W )−1 − χkt ρikt /σikt ◦ (I N − ρkt W )−1 ∂ xit−1 K (4.11) Considering that the spatial weight matrix (W ) is identified in different forms, it is necessary to decide the applicable matrix type. Following Li et al. (2013), we consider the spatial weight matrix as a row-standardized N × N first-order queen contiguity weight matrix.

4.3 Data Land allocation modeling can be done using aggregate or disaggregate data though disaggregated data is usually preferred. When disaggregated data is unavailable, economic agents’ behavior is estimated using aggregate or macro data (Wu and Li 2013). Considering the unavailability of disaggregated data in Iran, we used aggregate data at the level of the counties. Counties’ agronomic data was obtained from the Ministry of Agriculture, and climatic data was got from the Iran Meteorological Organization. Considering that an estimation of the spatial fractional multinomial logit model needs determining the spatial contiguity matrix, agronomic data and climatic data were put together with geographical data using the geographic information system (GIS). In contrast to agronomic data, climatic data is not according to counties and is collected

4 Assessing Climate Change Impacts on Land Use in Iran …

53

Fig. 4.1 Political divisions of the Iran’s counties

based on meteorological stations, so this data was calculated for each county based on meteorological stations’ geographical locations. In case the counties did not have meteorological stations, interpolation methods were employed for calculating their climatic data. Based on the available data, 336 counties were selected as the statistical population and agronomic data for 2007 and 2013 and agronomic and climatic data for 2006 and 2012 were used. Agronomic data has cultivation area under annual crop groups including cereals (wheat, barley, rice and maize), legumes (peas, legumes, lentils and other legumes), vegetables (potatoes, onions, tomatoes and other vegetables), industrial crops (cotton, tobacco, sugar beet, sugarcane, soybean, canola and other oilseeds), cucurbits (melons, watermelons, cucumbers and other cucurbits) and forage crops (alfalfa, clovers, corn and other forage crops) in hectares. Climatic data included mean monthly temperature (°C), total monthly precipitation (mm), mean monthly humidity (in percentage) and mean monthly wind speed (meters per second). Figure 4.1 shows the political divisions of the counties and their geographic proximity used for determining the spatial contiguity matrix.

4.4 Results The fractional dependent variables of our fractional models were generated by dividing each group’s cultivation area by the total cultivation area. Next, considering the fact that the original climatic data was monthly, this data was reformed as seasonal and annual data to provide alternative seasonal and annual models after which the

54

K. Alefi and M. Ghahremanzadeh

Table 4.1 Summary of estimated models and their information criteria The weather explanatory variables in the model Annually measures

2006–2007

2012–2013

BIC

BIC

AIC

AIC

924.06

714.12

947.14

737.20

Annually measures along with their standard deviation

1037.46

751.17

1059.28

773.00

Annually measures along with their coefficient of variation

1036.47

750.18

1060.09

773.81

Seasonally measures

1264.21

825.25

1283.35

844.38

Seasonally measures along with their standard deviation

1716.84

976.32

1739.07

994.73

Seasonally measures along with their coefficient of variation

1712.05

975.35

1739.09

994.75

best model was chosen. Since the primary model or the same fractional multinomial logit model estimation is done using the maximum likelihood estimators, the Akaike information criterion (AIC) and the Bayesian information criterion (BIC) value of each model were used for comparing the primary models. Based on Table 4.1, the fractional multinomial logit models with annual climatic variables were selected as the best models for the two periods. In the next step, we considered estimating the spatial fractional multinomial logit models. The results for the two time periods are reported in Table 4.2. As shown in Table 4.2, annual climatic variables and annual crop groups’ cultivation areas’ shares in 2006 and 2012 were considered as the explanatory variables for these crop groups’ share of cultivation area in 2007 and 2013, respectively. This was done based on farmers’ decision-making processes. According to literature, farmers’ expectations of future climate which are formed based on previous observations affect their decision-making processes about the cultivation of different crops. Meanwhile, traditional cropping patterns in each region effect farmers’ decisions. Thus, importing the data for land shares in the past can show the effects of these patterns. As shown in Table 4.2, spatial autoregressive parameters in the 2006–2007 model were more significant as compared to the 2012–2013 model. In the 2006–2007 model, spatial autoregressive parameters for all groups were significant except for cereals, while in the 2012–2013 model the parameters were significant only for industrial crops. Therefore, spatial autocorrelation in crop production decreased over time because the country had regular cropping patterns in the past and each region was more active in cultivating specific crops. This led to a high spatial correlation in crop production and producing similar crops in nearby regions. Over time, cultivation of traditional crops decreased and market demand and profitability had more of an impact on crop choices. As a result, crop production became scattered and the spatial correlation decreased. Based on Table 4.2, all significant spatial autoregressive parameters in the two periods were positive. These positive coefficients indicate that

4 Assessing Climate Change Impacts on Land Use in Iran …

55

Table 4.2 Estimates of land use transitions (2006–2007) model Variables

Vegetables Industrial crops

Cereals

Legumes

Temperature

0.01

Precipitation

−0.0008

Cucurbits

−0.01

0.01

0.07**

−0.005

0.0005

−0.0005

−0.0007

−0.003**

Humidity

0.01

0.01

0.01

0.01

0.01

Wind speed

−0.19

−0.26*

−0.15*

−0.21

−0.26*

Share of cereals (t − 1)

5.29***

3.37*

5.97***

5.62***

5.10***

Share of legumes (t − 1)

7.39**

3.32

4.71***

4.50**

11.69***

Share of vegetables (t − 1)

5.21***

−1.12

1.96*

2.43

2.70

Share of industrial crops (t − 1)

−0.62

2.05

1.84

6.68**

0.77

Share of cucurbits (t − 1)

2.41

1.80

2.14

5.72

3.99*

Share of forage crops (t − 1)

0.77

0.97

0.01

0.01

0.08

Constant

−4.02**

−1.85

−2.86***

−4.41***

−3.06**

Spatial lag (WX)

0.02***

0.006***

0.00005

0.02***

0.005***

Number of observations

336 Cereals

Legumes

Cucurbits

(2012–2013) model Variables

Vegetables Industrial crops

Temperature

0.04

−0.03

0.02

−0.18***

0.06

Precipitation

−0.001*

0.0006

0.0004

−0.0002

0.0005

Humidity

0.02*

−0.007

0.002

−0.02

−0.01

Wind speed

−0.35*

0.30

0.01

−0.08

0.01

Share of cereals (t − 1)

10.44***

5.21***

5.30***

8.57*

7.15**

Share of legumes (t − 1)

10.43***

8.08*

5.52***

11.12**

9.49**

Share of vegetables (t − 1)

14.63***

8.98***

3.93***

−2.05

8.43**

Share of industrial crops (t − 1)

7.35***

3.45

4.09**

7.50

7.46*

Share of cucurbits (t − 1)

12.22***

2.40

2.14

−7.52

12.19***

Share of forage crops (t − 1)

10.01***

4.67**

2.85***

3.87

8.44**

Constant

−12.20***

−3.78*

−3.86***

−4.27

−9.40***

Spatial lag (WX)

0.003

0.007***

0.00004

- 0.001

0.002

Number of observations

336

Note *, ** and *** indicate statistical significance at the levels 10, 5 and 1%, respectively

56

K. Alefi and M. Ghahremanzadeh

the increase in each crop group’s share of cultivation area in each county increased as compared to the crop group’s share of cultivation area in neighboring counties. As noted before, in 2012–2013 only industrial crops’ land share had a spatial correlation. Continuation of industrial crop production and its spatial autocorrelation occurred due to the unique features of industrial production. Unlike other crops, the production of industrial crops is purely for commercial reasons and these crops are used as inputs for processing in factories. This has led to industrial crop production being centralized around the factories leading to a positive spatial correlation with neighboring counties in industrial crop production. Another point to be mentioned here is the statistical insignificance of the spatial correlation of the area under cereal cultivation in the two periods of study. This happened because the cultivation of these crops was encouraged leading to widespread production in many regions as these are strategic crops. The spatial fractional multinomial logit model is nonlinear, and there are interactions between the equations. So one cannot make direct interpretations for its parameters, and it is necessary to calculate the marginal effects of the changes in the explanatory variables in different group’s shares. The marginal effects of the changes in the explanatory variables are calculated and reported in Table 4.3. It should be noted that all marginal effects are calculated in the mean of explanatory variables. Table 4.3 shows that sensitivity to changes in climatic variables increased over time. In the 2006–2007 period, temperature changes only affected the allocation of land share for legumes, while in 2012–2013, temperature changes affected all crop types, except industrial crops. This is because in recent years temperature increase has become faster and farmers’ awareness levels have also improved. This has led farmers to consider changes in climatic variables when taking decisions. Based on the reported results for 2006–2007, the increase in temperature in this period had a positive effect on legumes’ share of cultivated area. In 2012–2013, temperature increase had a negative effect on legumes’ cultivation area and a positive effect on vegetable, cereal and cucurbit cultivation areas. Different responses to temperature changes in the two periods are because the responses are dependent on temperature levels; an increase in temperature up to a specific level might encourage cultivation of some crops, while a higher increase can reduce cultivation and vice versa. Precipitation increase in 2006–2007 and 2012–2013 affected cucurbits and vegetables’ cultivation areas negatively, respectively. Changes in humidity in 2012–2013 affected the share of area under vegetables positively. Moreover, increase in wind speeds in 2012–2013 decreased vegetables’ share of cultivated area. Therefore, climate change and, as a result, changes in climatic variables affected the land allocated to these crops. Based on Table 4.3, in addition to climatic variables, annual crop groups’ share of cultivated area in the previous year also affected their share of cultivated area in the next year. Taking into consideration statistical significance, higher land allocated to cereals in 2006 increased the share of its cultivation area as also the share of land for legumes in 2007. A higher share in 2012 increased the share of areas for vegetables, industrial crops and legumes in 2013. Larger area for legume cultivation in 2006 increased cucurbits’ cultivation area in the next year. The increase in the share of land for vegetables in 2006 impacted the share of this area positively in 2007, and its

4 Assessing Climate Change Impacts on Land Use in Iran …

57

Table 4.3 Marginal effects on land use transitions in spatial multinomial logit model (2006–2007) Variables

Vegetables

Industrial crops

Cereals

Legumes

Cucurbits

Temperature

−0.0004

−0.004

−0.002

0.02***

−0.002

Precipitation

−0.00002

0.0001*

−0.000009

−0.0001

−0.0002***

Humidity

0.0003

0.00007

0.0002

0.003

0.0007

Wind speed

−0.003

−0.01

−0.006

−0.03

−0.01

Share of cereals (t − 1)

0.10

−0.10

1.00***

0.77***

0.07

Share of legumes (t − 1)

0.31

−0.11

0.42

0.35

0.72***

Share of vegetables (t − 1)

0.37***

−0.33

0.21

0.37

0.08

Share of industrial crops (t − 1)

−0.32**

−0.03

−0.09

1.78***

−0.15

Share of cucurbits (t − 1)

−0.03

−0.09

−0.09

1.31

0.12

Share of forage crops (t − 1)

0.06

0.09

−0.05

−0.04

−0.01

Variables

Vegetables

Industrial crops

Cereals

Legumes

Cucurbits

Temperature

0.001**

−0.01

0.01**

−0.003***

0.002**

Precipitation

−0.00008*** 0.0002

0.00008

−0.00001

0.00001

Humidity

0.001***

−0.003

0.002

−0.0004

−0.0005

Wind speed

−0.01***

0.13

−0.03

−0.004

−0.002

Share of cereals (t − 1)

0.22***

1.07*

1.08***

0.07

0.10

Share of legumes (t − 1)

0.17

2.19

0.68

0.09

0.15

Share of vegetables (t − 1)

0.33***

2.89***

−0.001

−0.15

0.11

Share of industrial crops (t − 1)

0.15***

0.56

0.89

0.08

0.17

Share of cucurbits (t − 1)

0.34***

0.40

0.29

−0.19

0.39**

Share of forage crops (t − 1)

0.23***

1.24

0.20

0.007

0.20

(2012–2013)

Note *, ** and *** indicate statistical significance at the levels 10, 5 and 1%, respectively

increase in 2012 had a positive effect on vegetables and industrial crops’ cultivation areas in 2013. A higher share for industrial crops in 2006 increased cultivation areas’ shares of vegetables and legumes in 2007, and its higher share in 2012 has increased only vegetables cultivation areas’ shares in 2013. A larger share for cucurbits’ cultivation in 2012 increased the area under cucurbit and vegetable cultivation in 2013. Moreover, the increase in forage crops’ share in 2012 had a positive effect on the area for vegetable cultivation in 2013.

58

K. Alefi and M. Ghahremanzadeh

4.5 Conclusion This study aimed to assess the impact of climate change on land allocation between annual crop groups in Iran’s counties. It estimated two spatial fractional multinomial logit models for two periods, 2006–2007 and 2012–2013. The results show that spatial autocorrelation parameters for the 2006–2007 period were more significant as compared to the 2012–2013 period. Therefore, the spatial correlation of crops’ land shares declined over time. It can also be concluded that spatial effects have an impact on the land allocated for different crop groups. Ignoring these effects can lead to incorrect results through biased parameters which shows the importance of using spatial models in similar studies. According to the calculated marginal effects, crops’ land shares were dependent on their land shares in the previous year. Moreover, annual crops’ cultivation areas’ responses to climate change increased over time. This is more obvious in temperature change because temperatures have changed faster in recent years and there is more awareness about these changes. Thus, considering the fact that climate change is happening, it is necessary to minimize the agriculture sector’s vulnerabilities to these changes by making accurate policies based on appropriate research in Iran. For this, it is essential to find out the country’s population size and the different nutritional needs of the population for the future. This can help determine the required production levels for different products and provide the necessary conditions for their production by adapting appropriate strategies on climate change. Hence, we recommend that research institutes should consider producing new heat-resistant varieties and the Ministry of Agriculture should support and promote these varieties. It is also recommended that future studies should consider land allocations in other agricultural sub-sectors (e.g., horticulture and livestock). Since this and other studies consider climate change as a factor affecting land allocation, we recommend that future studies consider climate change as a consequence of changes in land allocations for different crops.

References Adams RM, Hurd BH, Lenhart S, Leary N (1998) Effects of global climate change on agriculture: an interpretative review. Clim Res 11(1):19–30 Allen JE IV (2014) Determinants of land allocation in a multi-crop farming system: an application of the fractional multinomial logit model to agricultural households in Mali. In: 2014 annual meeting, July 27–29, 2014, Minneapolis, Minnesota 170175, Agricultural and Applied Economics Association Central Bank of Iran (2016). Available at https://www.cbi.ir/default_en.aspx Cho SJ, McCarl BA, Wu X (2014) Climate change adaptation and shifts in land use for major crops in the USA. In: 2014 annual meeting, July 27–29, Minneapolis, Minnesota (No. 170015). Agricultural and Applied Economics Association

4 Assessing Climate Change Impacts on Land Use in Iran …

59

Cho SJ, McCarl BA, Wu X (2015) Climate change adaptation via US land use transitions: a spatial econometric analysis. In: 2015 annual meeting, Jan 31–Feb 3, 2015, Atlanta, Georgia 196684, Southern Agricultural Economics Association Chou C, Lan CW (2012) Changes in the annual range of precipitation under global warming. J Clim 25(1):222–235 Di Falco S (2014) Adaptation to climate change in Sub-Saharan agriculture: assessing the evidence and rethinking the drivers. Eur Rev Agr Econ 41(3):405–430 Kaminski J, Kan I, Fleischer A (2013) A structural land-use analysis of agricultural adaptation to climate change: a proactive approach. Am J Agr Econ 95(1):70–93 Klier T, McMillen DP (2008) Clustering of auto supplier plants in the United States: generalized method of moments spatial logit for large samples. J Bus Econ Stat 26(4):460–471 Li M, Wu J, Deng X (2013) Identifying drivers of land use change in China: a spatial multinomial logit model analysis. Land Econ 89(4):632–654 Mu JH, McCarl BA (2011) Adaptation to climate change: land use and livestock management change in the US. Department of Agricultural Economics, Texas A&M University Mudzonga E (2011) Farmers’ adaptation to climate change in Chivi district of Zimbabwe. Trade and development studies centre. 3 Downie Avenue Belgravia, Harare, Zimbabwe Niles MT, Lubell M, Brown M (2015) How limiting factors drive agricultural adaptation to climate change. Agr Ecosyst Environ 200:178–185 Reidsma P, Ewert F, Lansink AO, Leemans R (2010) Adaptation to climate change and climate variability in European agriculture: the importance of farm level responses. Eur J Agron 32(1):91–102 Statistical Center of Iran (2016). Available at http://www.amar.org.ir Wu J, Li M (2013) Land use change and agricultural intensification: key research questions and innovative modeling approaches. A background paper submitted to the International Food Policy Research Institute. Final Report. Available at https://pim.cgiar.org/files/2013/12/Wu_Land_Use_ Change_and_Ag_Intensification.pdf Yang L (2010) Acreage allocation in the presence of various commodity and conservation programs: the case of conservation reserve program and crop production in the Midwest. Master of science dissertation. Iowa State University

Khadijeh Alefi is a lecturer in Department of Agricultural Economics at University of Tabriz. She received her B.Sc. and M.Sc. in Agricultural Economics from University of Tabriz and her Ph.D. in Farm Management from University of Tabriz. She teaches courses in Farm Management, Microeconomics and Macroeconomics at undergraduate level. Her research areas of interest are in Farm Management, Climate Change and Agricultural Production. Mohammad Ghahremanzadeh is an Associate Professor in Department of Agricultural Economics at University of Tabriz. He holds a B.Sc. from University of Tabriz and M.Sc. and Ph.D. from University of Tehran. He spent six months in Australia as Research Scholar to complete his thesis at University of Queensland. His fields of expertise include Agricultural Policy, Agricultural Price Analysis, Agricultural Insurance and Risk Management and (seasonal) time series modeling. He has co-supervised over 30 M.Sc. and 4 Ph.D. students.

Part II

Natural Resources and Environmental Economics

Chapter 5

Estimating the Non-use Values and Related Compensative Surplus of Arasbaran Forests in Iran: An Application of the Choice Experiment Method Maryam Haghjou, Babollah Hayati, Esmaeil Pishbahar and Morteza Molaei

5.1 Introduction When it comes to sustainable development and conservation of natural resources for future generations, preserving environmental resources, especially forest resources gets high priority. Today, we cannot ignore the effective role that forests play in preserving the balance in the ecosystem, controlling pollutants and producing oxygen, creating vitality, and keeping alive wildlife. All these make preserving forests one of the most important indicators of sustainable development in every country. However, the acceleration in economic development in recent decades has led to excessive pressure on the world’s environment and caused irreparable damage to nature. Hence, attention needs to be paid to the environmental effects of developmental patterns. The modern approach of the civilized world must be seen as moving from the “economic environment” to “environmental economics.” This approach emphasizes the need for comprehensive support and strengthening through interdisciplinary collaborations between natural and environmental experts on the one hand and specialists and elites from the economic sphere and policymakers on the other. M. Haghjou (B) · B. Hayati · E. Pishbahar Department of Agricultural Economics, University of Tabriz, Tabriz, Iran e-mail: [email protected] B. Hayati e-mail: [email protected] E. Pishbahar e-mail: [email protected] M. Molaei Department of Agricultural Economics, Urmia University, Urmia, Iran e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2019 M. Rashidghalam (ed.), Sustainable Agriculture and Agribusiness in Iran, Perspectives on Development in the Middle East and North Africa (MENA) Region, https://doi.org/10.1007/978-981-13-6283-5_5

63

64

M. Haghjou et al.

Estimating the economic value of environmental resources leads to the discovery of the demand curves for goods and services finally leading to the determination of the value that human beings attach to the environment. One essential thing that policymakers and planners need to focus on for preventive destructive uses of forests is carefully analyzing and evaluating these resources and moving toward sustainable development (Bateman and Willis 1999). Population growth has led to an increase in the demand for forest goods and services, which has led to an increase in their degradation. Statistics show that during 2005 and 2010, the European continent with the lowest population growth rates had the lowest rates of forest degradation (about 0.1%). On the other hand, the highest rate of destruction (around 0.8%) was on the African continent, (FAO 2010). Along with destruction of forest resources, the quality and quantity of services that come from the forests also reduces. Therefore, information about the benefits of forest resources will create incentives for protecting them. In other words, this will lead to a desire to pay for the protection of forests. According to the World Bank’s estimates, the current net value of losses from deforestation and the destruction of the Caspian forests in Iran (the loss of forest services and functions) is about $760 million and $147 million, respectively. In 2002 the value was equal to 0.8% of the country’s GDP (the World Bank 2005). Several classifications are used for evaluating the economic value of natural resources; however, to evaluate the characteristics of the Arasbaran forests, the economic value of the forest can be divided into: (a) use values and (b) non-use values. Use values can further be divided into: (1) direct-consumption value including the market value of forest products (for example, firewood) and (2) indirect-consumption value including the value of information functions (for example, recreational value, scientific research value, and aesthetic value) and three non-consumable values including habitat value (protection of plant and animal species) and regulation functions (regulation of gas, water, and soil). Forests’ non-use value consists of: existence value (which reflects the benefits that people receive without using a forest resource), bequest value (individual willingness to pay for maintaining or preserving a resource for the future) and option value (value of probable future use of forests resources for a person) (Heal et al. 2005; Pak et al. 2010; Pascual et al. 2010). Non-use value is a large part of environmental resources’ value including that of forests. Therefore, estimating this value is an important priority for the economic valuation of these resources. The total forest area in East Azerbaijan province is about 188,000 ha, of which about 164,000 ha belong to the Arasbaran forests of which, 148,000 ha have been reported as protection and conservation forests, and about 78,560 ha (about 56% of the area) are specified as protected forests. Arasbaran forests have 1,072 plant species and 97 species of trees. Since 1976 UNESCO designated it as one of the world’s largest conservation biospheres, it is one of the nine biosphere reserves in Iran. The Arasbaran area with its beautiful nature and pleasant landscape, and the existence of various historical buildings, has high potential for attracting tourists. Its medicinal plants can be of particular importance as another avenue of development in the region (Natural Resources Office of the East Azarbaijan Province 2003).

5 Estimating the Non-use Values and Related Compensative Surplus …

65

These forests are in critical condition due to excessive exploitation. They have been used for wood and coal mining over the past decades. The forests also have a sensitive and fragile condition and are in danger of destruction and demolition (Natural Resources Office of the East Azarbaijan Province 2003).1 Considering the importance of these forests for sustainable agriculture in Iran, we evaluate the existence value, bequest value, and option value of these forests. The rest of the paper is organized as follows. It first gives a brief literature review that focuses on the economic valuation of environmental services and forests. Section 5.3 describes the methodology and Sect. 5.4 discusses the results. The last section gives a conclusion.

5.2 Literature Review Considering the importance of the economic valuation of environmental services and forests, many studies have addressed this issue. Some studies have applied contingent valuation methods (CVM) to estimate the economic value of forests and other environmental services (Baral et al. 2008; Jahanshahi and Mousavi 2011; Khodaverdizadeh et al. 2008; Molaei 2009; Pattison 2009; Sattout et al. 2007; and Tao et al. 2012). Choice experiment is another method that has been used for environmental valuations (Cerda et al. 2013; Meyerhoff et al. 2009; Salehnia 2011; Taylor and Longo 2010; and Wallmo and Lew 2011) while other studies have used the contingent ranking approach (such as Garrod and Willis 1997 and Kumar and Kant, 2007). Some studies use the travel cost method for estimating the recreational value of environmental services (such as Chae et al. 2012; Hayati et al. 2011). Some others have used two valuation methods to compare the results of the valuation of the environment (for example, Bateman et al. 2006; Mogas et al. 2009; and Sayadi et al. 2005). This literature review shows that despite the different effects of the variables used in the studies, there are many factors that impact individuals’ willingness to pay for natural and environmental services including: (i) individual and socio-economic characteristics associated with each respondent, (ii) the characteristics of the environmental resource and, (iii) the bid price. One of the important parts of Arasbaran forests’ value is their non-used value. This is also one way of protecting and reviving natural resources, including forests through using popular contributions and motivating social tendencies to protect this valuable resource. Therefore, the contribution of our research is that it estimates the non-use value (option value, existence value, and bequest value) and also the compensative surplus (CS) of Arasbaran forests using the choice experiment approach.

1 According

to some informal sources, aerial maps show that 38% of the Arasbaran forests have been destroyed during the past 18 years due to human activities.

66

M. Haghjou et al.

5.3 Methodology Economic value is an instrumental value in maximizing the utility of human beings. It is human-oriented, meaning that it is based on human beings and on their preferences. It is possible to use market methods if there is a direct market for environmental goods and services. However, due to the lack of a suitable market for many environmental goods, it is often not possible to use these methods. Hence, one of the best methods for estimating non-market services is the stated preference method. The stated preference method has been developed in recent decades. This method directly measures the amount of willingness to pay. This preference approach relies on data obtained through direct questions posed to respondents on their preferences and includes several valuation techniques though a common feature of all these techniques is direct questioning of individuals about their possible choices in a hypothetical market. This approach involves contingent valuation techniques and multi-attribute valuation methods such as choice experiment and contingent rating. In a multi-attribute valuation method, an ecological resource is decomposed into its characteristics and attributes and values are estimated separately for each of its attributes. This study uses the choice experiment method. In this method, the respondents are asked to choose the desired option in each choice set according to their preferences. The choice experiment can create predictions of compatibility with welfare provided the existing status option is one of the choices so that responders can choose this if they are not interested in the alternative methods (Liu and Wirtz 2010). After estimating the models derived from the choice experiment, the implicit price of each attribute is the final rate of succession between the non-monetary and monetary characteristics. This can be calculated from the ratio of the non-monetary coefficient to its monetary coefficient:   β non−monetary (5.1) Marginal WTP  − β monetary The estimate of the compensative variation or the compensative surplus (w) which actually defines the difference in the utility for an individual with or without the existence of the specified attribute can be shown as:     1 Vi0 Vi1 w e − ln e (5.2) ln μ i∈C i∈C where μ shows the marginal utility of income and is the coefficient of the price factor in the estimated model (Liu and Wirtz 2010). After estimating WTP or implicit prices for each attribute, we used Krinsky and Robb’s (1986) approach to estimate the confidence intervals. This method is, in fact, a kind of parametric bootstrap and relies on the relative parameters of the desire to pay for a variety of statistical simulations and extracts. Ultimately, these extractions lead

5 Estimating the Non-use Values and Related Compensative Surplus …

67

to a distribution of the WTP ratio and thus a confidence interval for each parameter (Krinsky and Robb 1986). Designing the choice cards is the first and most important step in the choice experiment method. Here, first the main attributes of any studied resource and also the level of each attribute should be identified. Questions are designed based on the characteristics of the test cards. The attributes are practically selected after reviewing previous studies or interviewing a group of experts (called the “target group”). It is worth mentioning here that the bid price for the studied resources is one of the selected attributes and estimating the WTP for each attribute of the forest is possible through the monetary factor. Further, the level of each attribute is identified using exploratory studies, literature review, and interviews with the target group. We used the statistical design theory for a composition of the levels and for forming appropriate scenarios to present to the respondents. A complete factorial design is one of the available options in this method. However, due to the large number of compounds in this technique, we used the “partial factorial design” as an alternative method, which largely reduces the number of possible combinations. Table 5.1 shows selected attributes and their attribute levels. As shown in Table 5.1, the non-use value of Arasbaran forests is divided into three attributes (existence, options, and bequest value) with two levels, while the bid price has five selected levels. Hence, six alternatives and three choice sets were determined that were collected in one-trio-blocks. To design the cards, we used the SAS 9.2 software. Each choice set includes the relative importance and the maximum importance of these values besides one status quo option (their non-importance). Figure 5.1 shows a sample of the selected cards to estimate the value of the three non-use values of Arasbaran forests.

Table 5.1 Attributes and attribute levels Non-use value of Arasbaran forests

Bid price (rial)

Value

Existence value

Option value

Bequest value

10,000

Levels

Important

Important

Important

30,000

Not important

Not important

Not important

50,000

20,000

Question 1

40,000

attributes/ levels Existent value The value for the existence of the forest, not to use it option value The value of forests for future uses bequest value The value of forests for use of future generations WTP(rial)

Alternative A

Alternative B

Importance

Importance

Importance

Importance

Importance

Unimportant

50000

10000

Which one do you prefer the most? (1 is the best)

Fig. 5.1 An example of choice card used in calculation of non-use values

Alternative C None of these things are important to me and I am not willing to pay any price for it.

68

M. Haghjou et al.

The main assumption of the choice experiment method is the application of McFadden’s conditional logit regression model, but an important assumption of this specification is that selections from the choice sets should follow the “independence from irrelevant alternatives” (IIA) axiom. This means that the relative probabilities of the two options which are being selected should be unaffected by the introduction or removal of other alternatives (Hanely et al. 2002). This condition can be tested using the test introduced by Hausman and McFadden (1984) in which the model is estimated in an unrestricted way by all options and then one alternative is removed from the model and it is estimated again (restricted model); finally, the Hausman statistic is calculated using Eq. 5.3, which follows the χ 2 distribution and its significance can be confirmed via the calculated χ 2 table: 



ˆ  (νˆr − νˆ )−1 (β r − β) ˆ H∼0 χ 2 (m) H  (β r − β)

(5.3)

In case of any violation of the IIA hypothesis, the conditional logit model is no longer suitable for an analysis of the choice experiment. One of the alternative models in this condition is the mixed logit regression model. The mixed logit regression model is a highly flexible model that can approximate any random utility model. Its definition is based on the functional form of its choice probabilities. Any behavioral specification whose derived choice probabilities take this particular form is called a mixed logit model. This model has two forms or types, which are both the same theoretically and their results are also similar but they differ in the type of expression and formulation. Depending on the coefficients and the compound error of the problem, a researcher chooses one of these two forms. It should be noted that the Hausman test is used for determining the appropriate form. The two forms are: the compound error and random coefficients which is described as (Train 2003): Error components form: The random utility model is Uni  αi X ni + μi Z ni + εni , in which X ni and Z ni are vectors of explanatory variables, α  is vector of constant coefficients, μ is vector of random terms, and εni is the vector of error component which is distributed in iid extreme values. So, the random and unperceived section of the utility function is ηni  μi Z ni + εni . In this form and format of the error component, the possibility of a correlation between alternatives and individuals is created in the model, for example, as: Cov(ηni , ηn j )  E(μn Z ni + εni )(μn Z n j + εn j )    E(Z ni μn μn Z n j + Z ni μn εn j + εn j μn Z n j + εn j εn j )  W Znj + 0 + 0 + 0  Z ni

(5.4)

in which w is the covariance matrix of μ. This form is used in problems in which there is no correlation between alternatives and individuals (Train 2003). Random coefficient form: This form is the main form of the mixed logit model. To explain this model, we first review the concepts of the normal logit model. In the

5 Estimating the Non-use Values and Related Compensative Surplus …

69

normal logit model, if the coefficients of the utility function of the individual are known, then the probability of choosing the alternative i is: 

eβ xni Pni   β  x j j e

(5.5)

If, unlike the normal logit model, the coefficients are not constant, meaning that there is a certain coefficient for each individual and each alternative, it can be said that the utility function β is not deterministic and is random; but β follows a distribution with constant parameter θ . In this case, to obtain the probability of choosing alternative i and due to the variability of the coefficients, one other integral must be taken over Pni . In this condition, the probability of choosing alternative i will be:  Pni 

Pni (β)g(β|θ )dβ,

eVni (β) Pni (β)   J Vni (β) j1 e

(5.6)

where g(β|θ ) is the distribution function of β with parameter θ . Vni is the visible and non-random part of the utility function and its value depends on the coefficient β. If the utility function is a linear function of the coefficients, then Vni  βi X ni and the probability of choosing the mixed logit model is:  β  X ni

e  β  X g(β|θ )dβ Pni  (5.7) nj j e Clearly, in this model, the probability of choosing Pni is a function of the distribution parameter g(.) (Train 2003). Our study’s statistical population is approximately 334 people who were chosen randomly between visitors to the Arasbaran forests and also citizens of ten neighboring cities (within a radius of 250 km from the forests) from three adjacent provinces: West Azerbaijan, East Azerbaijan, and Ardabil in 2016. The sample size is calculated using the following formula presented by Orme (1998):   3 Nle  500 ≈ 167 (5.8) N  500 Nalt Nrep 3.3 where N lev is the highest number of levels per attribute, N alt is the alternative number in each choice set, and N rep is the number of questions that each respondent should answer.

70

M. Haghjou et al.

5.4 Results and Discussion Table 5.2 shows the main statistical characteristics of the study’s respondents. The mean of age variable represents a middle-aged population. A majority of the study’s subjects were married men and individuals with small families. The mean of annual gross income indicates a normal income for most of the respondents having less than one visit per year. The individuals’ friendly attitude toward Arasbaran forests (index of friendly attitude toward Arasbaran) consisted of ten comments to measure their friendly attitude toward the forests such as ignoring some utilities for safeguarding them. Each comment was evaluated through codes from 5 (very important) to 1 (not important). The mean of this variable indicates the relative importance of Arasbaran forests from the respondents’ point of view. A study of the respondents’ educational situation, measured by an index which is an ordinal variable identified as: 1  Illiterate, 2  Primary School, 3  Junior High School, 4  Senior High School, 5  Associated Diploma (AD), 6  B.Sc, 7  M.Sc, and 8  Ph.D. Mean of this variable states that most of the respondents had academic educational levels. The alternative specific constant (ASC) t is one of the measurable variables in the choice experiment. In ASC, it can be noted that most respondents regardless of the conditions and characteristics of the recovery of each option were willing to pay a price to implement any kind of plan to protect Arasbaran forests’ value. Table 5.3 shows the alternative specific constants’ frequency distribution for the non-usable features of Arasbaran forests. It is noticeable that 67% of the people considered these features important and were willing to pay for forage for themselves or for future generations, even if they did not use it. To derive the non-use value of Arasbaran forests, we estimated the conditional logit regression model. However, in view of what was mentioned earlier and due to the violation of the independence of the irrelevant alternatives (IIA) condition, the mixed logit regression model was replaced. The models were evaluated in two ways: standard and simple and interactive models. The hybrid model or a model with interactions was used for investigating the effect of some socio-economic characteristics and environmental indicators which are essential for respondents’ willingness to pay. The final hybrid model was obtained after initial investigations and variables

Table 5.2 Descriptive statistics of variables Max

Min

SD

Mean

Variable

6,000

250

740.761

5,343,000

Respondent’s income (rial)

71

23

7.700

40.396

Age of respondent

1

0

0.445

0.730

Gender (1  male, 0 otherwise)

8

4

1.010

5.860

Level of education

7

1

1.300

3.460

Family size

3

0

0.730

0.630

Number of annual visits

5

2

0.680

3.710

Index of respondents perspective on forests

5 Estimating the Non-use Values and Related Compensative Surplus …

71

Table 5.3 Frequency distribution of alternative specific constant (ASC) for non-use features among responses Option

Code

Frequencynumber

Relative frequency-percent

Importance and payment for non-use features

1

2007

67

Non-importance of non-use features

0

999

33

Total

–

3006

100

Table 5.4 Estimation of mixed logit model for non-use values Variable

Standard mixed logit regression

Mixed logit regression with interactions

Coefficient

Standard error

Coefficient

Standard error

ASC

0.3000***

0.124

0.260***

0.125

Price

−0.00004**

0.00003

−0.0011***

0.00012

Existence

0.3060***

0.099

0.293**

0.0101

Option

0.6101***

0.070

0.124***

0.071

Bequest

01.3100***

0.113

1.340***

0.0115

p × education

–

–

0.00037***

0.000020

p × income

–

–

0.000000037***

0.0000000122 0.000029 0.000029

p × Arasbaran

–

–

0.00014***

p × visit

–

–

0.0000762***

Log-likelihood: −1781.65 LR Chi2: 488.34***

Log-likelihood: −1734.38 LR Chi2: 553.81***

*p

< 0.1 < 0.05 *** p < 0.01 ** p

with high correlation coefficients with other variables were eliminated as were the variables that were not significant due to less variations. Finally, four effective variables—respondents’ level of education and income, index of individuals’ friendly attitude toward Arasbaran forests and number of annual visits to the forests were chosen in the model with their interactions. Table 5.4 summarizes the final results of the estimated mixed logit model for the non-use value of Arasbaran forests. It is noticeable that based on the results of Wald’s Chi-square statistic, both the models are statistically significant. The results of the likelihood ratio test for comparing the two models show that the computed Chi-square statistic with an amount of 54.94 was significant in less than 1% of the cases. Therefore, the model with the interactions is better than the standard one. Based on the results of the model, the positive and significant effect of ASC in both the models indicates that each option provided to the respondents regarding

72

M. Haghjou et al.

Table 5.5 Willingness to pay for extracting and ranking of non-use values ˙Importance of ˙Importance of ˙Importance of Values existent value option value bequest value Ind. monthly WTP (rial)

2,940

13,460

1,242

Ind. annual WTP (rial)

35,280

161,520

14,904

Total WTP (billion rial)

284.13 (6.76 × 106 USD)

1300.83 (3.09 × 107 USD)

120.03 (2.86 × 106 USD)

Ranking of values

2

1

3

the importance of the non-use features of Arasbaran forests significantly increased people’s utility and the respondents were willing to pay a price for the non-use value of Arasbaran forests. The negative and significant coefficient of the price variable indicates that with an increase in the amount of bid price, the respondents’ utility decreased. The positive and significant sign of all three non-use values shows individuals’ utility intensification for each of these characteristics. Moreover, according to the results of the interaction model, the positive and significant coefficients of the interaction of the price variable with other socio-economic and attitude variables indicate the positive effect of each of these factors on the respondents’ utility and willingness to pay. So it can be concluded that level of education, monthly income, number of annual visits to the forests, and friendly attitude toward Arasbaran forests significantly increased respondents’ WTP for the non-use value of Arasbaran forests. Our results confirm the findings of other studies like Bateman et al. (2006), Mogas et al. (2009), Tao et al. (2012), Sattout et al. (2007), and Sayadi et al. (2005). The results of extraction and ranking of individuals’ monthly and annual WTP and also the total WTP of the studied area for the non-use value of Arasbaran forests are presented in Table 5.5. It is worth noting that total WTP was calculated based on the statistical population of the three provinces which was about 80,53,684 in 2016. As mentioned earlier, the hybrid or the model with interactions was better than the standard one, so for calculating WTPs, the hybrid model’s results were applied to Eq. 5.1. The results show that the option value with 1,383.83 and bequest value with 120.03 billion rial had the most and least total annual WTP among respondents, respectively. Also, based on the results obtained from the model, the total non-use value of Arasbaran forests is about 1704. 99 billion rial, 76% of which is option value, 17% is existence value, and only 7% is the bequest value. A higher extracted value of the non-use features of Arasbaran forests show the need for protection of these valuable resources even if they are not used at the present time or are not used at all (the existence value). The results also show that option value is the most valuable non-use function of Arasbaran forests and has the highest share among these values. This could be based on the fact that due to the present lack of options for tourists and also lack of tourism facilities in the region people are willing to visit the forests in the future if their condition and the region’s situation gets any better so using the forests in the future is very important for the people. The lower amount of existence

5 Estimating the Non-use Values and Related Compensative Surplus …

73

Table 5.6 Estimating monthly compensative surplus of non-use values Existent value

Option value

Bequest value

2.7 × 106 rial (0.64 × 102 USD)

12.2 × 106 rial (2.90 × 102 USD)

1.1 × 106 rial (0.262 × 102 USD)

16.03 × 106 rial (3.82 × 102 USD)

Total compensative surplus of non-use values

value can be because of lack of information about its meaning and purpose in the respondents’ minds. The results of the compensative surplus (CS) calculations based on the estimated model and Eq. 5.2 are presented in Table 5.6. According to Table 5.6, the non-use value of Arasbaran forests could upgrade the SC of the respondents by about 16.06 × 106 rial (about 3.8 × 102 USD). Since this value is considerable and also because the ultimate goal of the government is increasing the utility and welfare of the community, implementing corrective measures and developing and protecting forests in addition to other economic benefits, can increase the satisfaction and welfare of the community.

5.5 Summary and Conclusion Forests as one of the most important renewable natural resources are the main and most valuable national resource for any country. Forests play an important role in different aspects including sustainability of life, preservation, and stability of the ecosystem, establishing a balanced and sustainable development of the environment, preserving biodiversity, providing raw material for industry and agriculture, job creation, controlling surface water and nutrition in the groundwater, preventing erosion and maintaining soil production, creating resorts, preventing climate change, providing wood and paper, pharmaceutical and industrial inputs, and helping in nature’s conservation. Therefore, forestry and forest conservation are important aspects of sustainable management and development. For protecting forests, improving and rehabilitating them are among the priorities of sustainable management. Therefore, the need for an economic valuation of forests, including their non-use value, is quite evident. Based on the findings of our study, we would like to make the following policy recommendations: • ASC’s results show that most respondents, regardless of the conditions and characteristics of each scenario, are willing to pay a premium to implement any plan to protect forests. This indicates the potential capital of public assistance in protecting Arasbaran forests and the need for proper planning to make use of this assistance.

74

M. Haghjou et al.

• A positive and significant relationship between price and income interaction with WTP suggests that improved incomes in forested areas can help in improving the environment. • The variable annual number of visits to the forests shows a positive relationship with the respondents’ WTP. This highlights the high potential of Arasbaran forests as an important tourist area because visitors have a high WTP for their conservation. • The positive impact of individuals’ level of education on their WTP was obvious in the estimated model. • Considering the positive relationship between respondents’ friendly attitude toward Arasbaran forests and WTP for protecting them, we can conclude that awareness and friendly beliefs about the forests have a direct impact on people’s WTP for the protection and conservation of forests. This result could be a sign that governmental training programs are needed for encouraging such beliefs. • The results show that the non-use value of Arasbaran forests is about 1704.99 billion rial annually, and if this non-use value is ignored in the calculations and in national accounting, then a large statistical error will occur and a large part of the forests’ economic value will be neglected. • The results of the compensatory surplus estimate show that the implementation of any conservation scheme in forests creates a significant increase in the respondents’ welfare and compensatory surplus. This is an interesting result for policymakers who care about increasing individual welfare. • The estimation results show that the option value is the most important among the non-use value of Arasbaran forests and accounts for the largest share. This could also be due to lack of other tourist sites and lack of tourism facilities in the region, as well as low awareness about the forests. Therefore, people will be more willing to visit these forests if their condition is improved; the availability of forests in the future is also very important for the people. • The existence value is perceived as less important which can be due to lack of information about its meaning in the minds of the respondents. Also, little importance given to the forests’ bequest value could be because this is not a tangible value for the respondents. • Our results show potential capital in the form of public assistance for protecting and recovering Arasbaran forests. This shows the importance of proper planning for using this assistance. • Increasing awareness and informing the respondents about the forests’ non-use value, in particular, their bequest and existence values, trying to induce the right to life and the existence of natural resources such as forests, as well as the right of future generations to this gift can effectively increase the respondents’ positive attitude toward the forests’ non-use value. This could be a key step in sustainable development. • Appropriate policies and employment programs in the region, along with supportive policies and environmental protection programs, can be considered as a pivot for improving forest resources.

5 Estimating the Non-use Values and Related Compensative Surplus …

75

• The inclusion of environmental education in the curriculum (from elementary to advanced levels) is important for infrastructure and indirect investments for protecting natural treasures, including forests.

References Baral NM, Stern J, Bhattarai R (2008) Contingent valuation of ecotourism in Annapurna conservation area, Nepal: implications for sustainable park finance and local development. Ecol Econ 66(2–3):218–227 Bateman IJ, Cole MA, Georgiou S, Hadley DJ (2006) Comparing contingent valuation and contingent ranking: a case study considering the benefits of urban river water quality improvements. J Environ Manage 79(3):221–231 Bateman IJ, Willis KG (1999) Valuing environmental preferences: theory and practice of the contingent valuation method in the US, EU, and developing countries. Oxford University Press, Oxford Cerda C, Ponce A, Zappi M (2013) Using choice experiments to understand public demand for the conservation of nature: a case study in a protected area of Chile. J Nat Conserv 21(3):143–153 Chae D, Wattage P, Pascoe S (2012) Recreational benefits from a marine protected area: a travel cost analysis of Lundy. Tour Manag 33(4):971–977 FAO (2010) Forest state of the world’s forests. Food and Agriculture Organization, Rome Garrod GD, Willis KG (1997) The non-use benefits of enhancing forest biodiversity: a contingent ranking study. Ecol Econ 21(1):45–61 Hayati B, Salehnia M, Hosseinzadeh J, Dashti GH (2011) Estimation of recreational value of Fadak park in Khoy city by individual cost method. First Conference on Urban Economics of Iran, Mashhad, Nov. 23–24 (in Persian) Hanely N, Wright R, Koop G (2002) Modelling recreation demand using choice experiments: rock climbing in Scotland. Environ Res Econ 22:449–466 Hausman J, McFadden D (1984) Specification tests for the multinomial logit model. J Jpn Int Econ 52(5):1219–1240 Heal GM, Barbier EB, Boyle KJ, Covich AP, Gloss SP, Hershner CH, Hoehn J, Pringle CM, Polasky S, Segerson K, Schrader-Ferchette K (2005) Valuing ecosystem services: toward better environmental decision-making. The National Academies Press, Washington, DC Jahanshahi D, Mousavi N (2011) The economic valuation of environmental amenities, Case Study: Yasouj waterfall. The first International Conference on Tourism Management and Sustainable Development, Marvdasht, 29–30 Sep (in Persian) Khodaverdizadeh M, Hayati B, Kavousi M (2008) Estimating the outdoor recreation value of Kandovan tourism village of East Azarbaijan with the use of contingent valuation method. J Environ Sci 4:43–54 (in Persian) Kumar S, Kant S (2007) Exploded logit modeling of stakeholders’ preferences for multiple forest values. Forest Policy Econ 9(5):516–526 Krinsky I, Robb AL (1986) On approximating the statistical properties of elasticities. Rev Econ Stat 68(4):715–719 Liu X, Wirtz KW (2010) Managing coastal area resources by stated choice experiments. Estuar Coast Shelf Sci 86(3):512–517 Meyerhoff J, Liebe U, Hartje V (2009) Benefits of biodiversity enhancement due to nature-oriented silviculture: evidence from two choice experiments in Germany. J Forest Econ 15(1/2):37–58 Mogas J, Riera P, Bennett J (2009) A comparison of contingent valuation and choice modeling with second-order interactions. J Forest Econ 12(1):5–30 Molaei M (2009) Economic-environmental valuation of Arasbaran forest ecosystem. Ph.D. Thesis, Faculty of Economics and Agricultural Development. University of Tehran (in Persian)

76

M. Haghjou et al.

Natural Resources Office of East Azarbaijan Province (2003) Protecting the Northern Arasbaran Forests (In Persian) Orme B (1998) Sample size issues for conjoint analysis studies. Sawtooth Software Inc., Provo Pak M, Turker MF, Ozturk A (2010) Total economic value of forest resources in Turkey. Afr J Agric Res 5(15):1908–1916 Pascual U, Muradian R, Brander LM, Gomez-Baggethun E, Martin-Lopez B, Verma M, Armsworth P, Christie M, Cornelissen H, Eppink F, Farley J, Loomis JB, Pearson L, Perrings C, Polasky S (2010) The economics of valuing ecosystem services and biodiversity. In: Kumar P (ed) The economics of ecosystems and biodiversity: ecological and economic foundations. Earthscan, London, pp 183–256 Pattison JK (2009) The non-market valuation of wetland restoration and retention in Manitob. M.Sc Thesis in Agricultural and Environmental Economics. University of Alberta, Canada Salehnia M (2011) Estimating willingness to pay for improvement in Lake Urmia’s environmental situation using choice experiment. Master Thesis, Department of Agricultural Economics, University of Tabriz (in Persian) Sattout EJ, Talhouk SN, Caligari PDS (2007) Economic value of cedar relics in Lebanon: an application of contingent valuation method for conservation. Ecol Econ 61:315–322 Sayadi S, Roa CG, Requena JC (2005) Ranking versus scale rating in conjoint analysis: evaluating landscapes in mountainous regions in southeastern Spain. Ecol Econ 55(4):539–550 Tao Z, Yan H, Zhan J (2012) Economic valuation of forest ecosystem services in Heshui watershed using contingent valuation method. Proced Environ Sci 13:2445–2450 Taylor T, Longo A (2010) Valuing algal bloom in the black sea coast of Bulgaria: a choice experiments approach. J Environ Manag 91(10):1963–1971 Train K (2003) Discrete choice methods with simulation. Cambridge University Press, Cambridge Wallmo K, Lew D (2011) Valuing improvements to threatened and endangered marine species: an application of stated preference choice experiments. J Environ Manag 92(7):1793–1801 World Bank (2005) Islamic Republic of Iran cost assessment of environmental degradation, Sector Not, No. 32043-IR. Available at http://documents.worldbank.org/curated/en/ 401941468284096627/pdf/320430IR.pdf

Maryam Haghjou is Lecturer in Department of Agricultural Economics at University of Tabriz. She holds her BS, M.Sc, and Ph.D. from Department of Agricultural Economics, University of Tabriz. She has taught a number of courses on natural resource economics and microeconomics. Her research and publication has focused on sustainable development, environmental management and natural resource management. She has publications in Journal of Agricultural Science and Technology (JAST). Babollah Hayati is Professor in Department of Agricultural Economics at University of Tabriz. He was Dean of Faculty of Agriculture during 2015–2018. He holds a B.Sc in agricultural economics at University of Tehran and M.Sc and Ph.D. in natural resource economics at Tarbiat Modares University. His areas of special interest are natural resource economics, sustainable development economics, and microeconomics. His recent publications have appeared in numerous journals including the Journal of Agricultural Science and Technology (JAST) and Engineering Sustainability. Esmaeil Pishbahar is Associate Professor and Head of Department of Agricultural Economics at University of Tabriz, Iran. He holds a B.Sc in Agricultural Economics from University of Tabriz and an M.Sc in agricultural economics from University of Tehran. He did his Ph.D. in science economics at departments of Economics and Management, University of Rennes1, France. His areas of interest and research are applied econometrics, agricultural risk management and insurance, and international trade. His teaching areas are advanced econometrics, mathematical economics, and

5 Estimating the Non-use Values and Related Compensative Surplus …

77

macroeconomics at under- and postgraduate levels. He has over 100 publications in journals and chapters in books. Morteza Molaei is Associate Professor in Department of Agricultural Economics at Urmia University. He is one of the pioneers of Department of Agricultural Economics at Urmia University. He was head of department during 2013–2017. He received his B.Sc degree in agricultural economics from University of Tabriz; M.Sc and Ph.D. degrees from University of Tehran. Dr. Molaei’s areas of expertise are environmental impact assessment, sustainable development, environmental efficiency and water resource economics. He teaches econometrics and environmental economics at postgraduate level. Currently, he works on a FAO project “Farm and Household Livelihood Survey Covering Selected Sub-basins within Lake Urmia Basin.”

Chapter 6

Optimized Reservoir Management for Meeting Conflicting Stakeholder Preferences: Methodological Innovations with Evidence from Iran Omid Zamani, Hemen Nader, Masoomeh Rashidghalam and Frank A. Ward

6.1 Introduction Water management has been a growing challenge in Iran since the 1990s. There are strong indications that Iran is facing not only a periodic ‘dry spell,’ but also a severe water crisis caused by low water productivity, high population growth, and a weakly informed management of water resources (Balali et al. 2011; Madani 2014). During the last decade, the water crises in the country have led to water resource management and water policies attracting the attention of scientists in different fields and also of policymakers (Foltz 2002; Madani 2014). Before developing policies and guidelines for water-related decision making for dams and reservoirs, there is a need to understand the elements of Iran’s current water crisis: (1) rapid population growth and unbalanced population distribution; (2) low productivity and efficiency of the O. Zamani (B) Department of Agricultural Market Analysis, Institute of Agricultural Economics, Christian-Albrechts-University Kiel, Olshausenstraße 40, 24118 Kiel, Germany e-mail: [email protected] H. Nader Department of Agricultural Economics, University of Zabol, Zabol, Iran e-mail: [email protected] M. Rashidghalam Department of Agricultural Economics, University of Tabriz, Tabriz, Iran e-mail: [email protected] F. A. Ward Department of Agricultural Economics and Agricultural Business, New Mexico State University, Las Cruces, NM 88003, USA e-mail: [email protected] O. Zamani Leibniz-Institut für Agrartechnik und Bioökonomie e.V. (ATB), Max-Eyth-Allee 100, 14469 Potsdam, Germany © Springer Nature Singapore Pte Ltd. 2019 M. Rashidghalam (ed.), Sustainable Agriculture and Agribusiness in Iran, Perspectives on Development in the Middle East and North Africa (MENA) Region, https://doi.org/10.1007/978-981-13-6283-5_6

79

80

O. Zamani et al.

Fig. 6.1 Geographical position of Mahabad

agricultural sector; and (3) challenges in management (Madani 2014). While only 15% of the historically irrigated land is now cultivated in Iran, up to 90% of the water resources are used by the agricultural sector, which has been characterized by low water productivity (Iran’s Ministry of Agriculture 2015; Iran’s Ministry of Energy 2015). To meet the growing demand for water by different sectors such as industry and municipalities along with addressing the country’s environmental problems what is also needed is that the water management practices be productively adjusted from crisis (reactive) management to preventive (proactive) management. This policy can help remedy the widespread negative externalities in these sectors (Madani 2014). Iran has 316 dams and ranks third worldwide in the number of dams; 132 dams are under construction (Iran’s Ministry of Energy 2016). In this regard, one of the challenging issues of water resources is water reservoir management. Mahabad Dam is an embankment dam located in West Azerbaijan Province in the northwest of Iran in the Urmia Lake Basin (Fig. 6.1. The dam is 700 m long with a maximum height of 47.5 m and a maximum volume of 1.5 million cubic meters. This dam plays a vital role in providing water for agriculture, industries, and municipal sector, and for generating electricity. Moreover, since it is close to Mahabad City, it is also an ideal place for recreation and leisure. However, a part of the Urmia Drainage Basin is facing a crisis due to some specific issues with Urmia Lake. This increases the importance of water resource management in this area.

6 Optimized Reservoir Management for Meeting Conflicting …

81

Like other essential commodities, water has a value which is based on its functions (in agriculture, recreational activities, power generation, and ecosystem services). Water’s value determines its opportunity costs for different activities and provides critical information for water allocation which is essential for water resource management (Dinar et al. 2000; Savenije and van der Zaag 2002). Water reservoir use is typically apportioned between different stakeholders with diverse interests and consequently different values. Reservoir management is a complex and multifaceted task which involves multiple stakeholders with different priorities and goals (Chung et al. 2014). Water allocation in Iran is usually managed by the government or by a community of private managers. Due to lack of water markets, the government is responsible for water reservoir supply and distribution systems, which are essential for decision making on public issues (Kiker et al. 2005). Thus, the management of water reservoirs is a complex problem which needs to take multiple objectives into account and involves many decision-making variables (Chung and Lee 2014). Since individuals might find it difficult to make choices in a complex decision-making environment, effective decision-making processes and methodologies are necessary for implementing policies on water distribution (McDaniels et al. 1999). In recent years, the analytical tools used for water management have improved considerably. So far, vast literature has used different approaches for assessing water reservoir distribution (see, e.g., Antunes et al. 2011; Rosso et al. 2014; ZamarrónMieza et al. 2017), but only a few of them discuss stakeholders’ competing priorities in using water from the reservoirs. FAHP and GP’s hybrid approach has some advantages which help overcome the weaknesses observed in applying the methods separately (Bertolini and Bevilacqua 2006; Schniederjans and Wilson 1991). Most importantly, ‘it is presented as an extension to consider additional criteria in decision making process’ (Badri 1999: 239). Therefore, in this study, we apply a hybrid approach using a multi-criteria analysis (MCA), stakeholder analysis (SA), and goal programming (GP) to allocate water reservoirs with the final aim of addressing conflicting demands between potential stakeholders and users to get water from the Mahabad reservoir in Iran’s West Azerbaijan Province. This study contributes to the literature by presenting a new approach and a more productive method of allocating Mahabad Dam’s water among competing stakeholders. The rest of the chapter is structured as follows. Section 6.2 introduces the methodology and the general structure of the model. Section 6.3 presents the main results of the study, and the final section provides a summary and concluding remarks.

6.2 Methodology Due to various stakeholders’ different objectives associated with reservoir systems, programming in water resource management is complex and needs a flexible and original mathematical method (Rani and Moreira 2010; Singh 2012). During recent decades, the mathematical modeling software has improved considerably to help

82 Fig. 6.2 General framework of the study

O. Zamani et al.

Identifying problem and stakeholders

Selection Criteria

Stakeholders Analysis Weighting the criteria

FAHP

Goal Programming

Water Allocation

Crop Land Use

resolve water resource problems (Singh 2012). These methods, in turn, have the potential, mostly unrealized to date, to provide important guidelines for management and planning. This study uses a hybrid model which considers both stakeholders’ priorities and technical constraints to allocate the water from the reservoir to different conflicting users. This approach is based on a combination of lexicographical goal programming (LGP) in which the priorities between different users are determined by a stakeholder analysis and a fuzzy analytic hierarchy process. The general framework of our approach is given in Fig. 6.2. First, we did a stakeholder analysis (SA) to identify the associated stakeholders and their criteria. This was followed by a calculation of water allocation’s priorities and augmented by additional criteria and sub-criteria. Finally, we used the results as input data to determine optimum water allocation among stakeholders and among alternative crop land uses.

6 Optimized Reservoir Management for Meeting Conflicting …

83

6.2.1 Stakeholder Analysis A stakeholder analysis (SA) is an approach that is followed for allocating resources to different organizations and people. It helps us understand the interests and objectives of various stakeholders using the environment, the trade-offs between different goals, and the costs and benefits of changes and interventions (Grimble et al. 1995). Environmental problems are multi-scale, uncertain, and complex which affect multiple individuals and agencies. Hence, environmental problems demand a decision-making process which is flexible to dynamic circumstances. In this regard, SA is embedded in environmental decision-making processes (see, e.g., Rosso et al. 2014; Reed 2008; Reed et al. 2009). Satisfactorily resolving conflicts among numerous stakeholders is an iterative and continuous process which involves at least the following steps: (1) identifying the stakeholders, (2) documenting their influence and interests, and (3) developing a good understanding of the goals and choices of certain key influential stakeholders. According to Mendelow (1981), one good method for classifying stakeholders’ interests is stakeholder mapping. This approach groups different stakeholders based on their influence and interests (Fig. 6.3). As shown in Fig. 6.3, it assigns different stakeholders to one of the four categories. Each stakeholder’s position on the grid shows the actions that a decision maker has to take: (i) high power/high interest stakeholders: Decision makers must make the greatest efforts to satisfy and fully engage them in the project, (ii) high power/low interest stakeholders: Decision makers should put in enough work to keep them satisfied, (iii) low power/high interest stakeholders: Decision makers should keep these stakeholders adequately informed and talk to them to ensure that there are no major issues, and (iv) low power/low interest stakeholders: It is better to monitor these stakeholders and not to overwhelm them with small issues. Table 6.1 lists the relevant stakeholders who play a role in the decision-making process for Mahabad Dam’s water demand and includes downstream farmers, agricultural organizations, Mahabad municipalities, and the Water Resource Authority (Table 6.2).

Fig. 6.3 Power/interest matrix

84

O. Zamani et al.

Table 6.1 TFN values Statement

TFN

Absolute

(7/2, 4, 9/2)

Very strong

(5/2, 3, 7/2)

Fairly strong

(3/2, 2, 5/2)

Weak

(2/3, 1, 3/2)

Equal

(1, 1, 1)

Source Tolga et al. (2005) Table 6.2 Survey of most relevant actors for the project Number

Stakeholder

Abbreviation

Level

1

Downstream farmers

DF

Regional

2

Agricultural organization

AO

Regional/provincial

3

Mahabad municipalities

MM

Regional

4

Water Resource Authority

WA

Regional/provincial

6.2.2 Hierarchical Analysis As described earlier, the first part of the methodology involves a hierarchical analysis. Theoretically, the analytic hierarchy process (AHP) is gaining knowledge related to economic, political, social, and technical matters (Saaty 2000; Zahedi 1986). AHP is an essential mechanism for multi-objective and multi-criteria decision making by pairwise judgments among different alternatives and criteria (Huang and Miller 2003). This methodology reflects the power of objective judgments with respect to issues so as to juxtapose different judgments for getting consistent expectations in the results. However, a subjective comparison cannot be precisely extracted using a mathematical point of view. It is extracted from uncertain possibilities. Therefore, FAHP could be a generalizable approach for handling well-known challenges that characterize hierarchical problems (Bozbura and Beskese 2007; Ertugrul and Karakasoglu 2009). On the other hand, subjective judging is considered as an inexact approach. Accordingly, to overcome this major drawback, FAHP has grown for hierarchical problems (Bozbura and Beskese 2007; Ertugrul and Karakasoglu 2009). This study uses Chang’s (2007) method for converting AHP to FAHP. Based on this methodology, each criterion is converted to a fuzzy set M i (li , mi , ui ). According to Chang’s methodology, the value of each criterion in a fuzzy triangular form will be li /l i , mi /mi and ui /ui . The membership function which is determined by comparing each peer should be done for each criterion and the intersections in the next step, as: 1 2 3 m , Mgi , Mgi , . . . Mgi Mgi

j  1, 2, . . . , m i  1, 2, . . . , n

(6.1)

6 Optimized Reservoir Management for Meeting Conflicting …

85

j

where Mgi are triangular fuzzy members and gi is the indicator of a goal (Chang 2007; Askin 2008). In our analysis, we follow Chang (2007) to formulate FAHP. An overview of the hierarchical structure of this study is given in Fig. 6.4. In Fig. 6.4, the main goal is at the highest level of the chart, followed by the criteria, sub-criteria, and finally some alternatives at lowest level of the structure. According to the current situation of the stakeholders, three criteria—economy, social, and environmental—were considered and each of them were linked to two sub-criteria with the alternatives agricultural, recreational–environmental, drinking, power generation, and flood control. After determining the criteria, sub-criteria, and stakeholder priorities, we used LGP to allocate water according to these assumptions. LGP was first introduced by Lee (1972) is a type of GP. To obtain optimum benefits when prioritizing a scheme, one should position positive and negative deviations according to their prominence. In this approach, the lower goals are considered after the higher goals have been attained (Chuang et al. 2007). A combination of these two models is formulated as:      − + (6.2) pk wk dk− , wk dk+ + pj dj , dj min Z  k

j1

Subject to:   k

 − + SFAHP,k xk + ds,FAHP − ds,FAHP  1 k  1, . . . 7

 

(6.3)

 − + Sl,k xk + ds,l − ds,l  Tk k  1, . . . , 7

(6.4)

k

where k (1,…,7) is the sub-criteria constraint, j (1,…,5) is the number of real goal constraints1 , pk are the priority targets extracted from the hierarchical structure, pj are the priority targets extracted from GP, W k are positive and negative deviations from the associated goals, S FAHP is the global weight of the alternatives and S l , T are the local weight values for each criterion, and x are the possible alternatives. The objective function of the model is to minimize the cost of water shortages summed over sectors. This objective shows that the number of alternatives with the highest weights should be combined. In other words, the selected alternatives automatically have the maximum weight in the FAHP process. Equations 6.3 and 6.4 are related to the constraints of a criterion or sub-criterion. It should be noted that deviation variables have a balancing role in this equation. The proposed evaluation tools that need to be applied to a real decision are given in Appendix A.

1 We

divided the goal constraints into two groups of sub-criteria—goal constraints and real goal constraints which refer to conventional goal constraints.

Fig. 6.4 Hierarchical structure in order to prioritize the use of dam’s water

86 O. Zamani et al.

6 Optimized Reservoir Management for Meeting Conflicting …

87

6.3 Data and the Case Study We used different data sources in our analysis including the regional Water Resource Authority, the Iranian Ministry of Energy, and Ministry of Agriculture, interviewing local farmers and a stakeholder analysis. Demands for drinking water, the reservoir’s water volume, capacity for electricity power generation, and the number of visitors to the dam over different months were obtained from the regional Water Resource Authority and the Iranian Ministry of Energy. Additionally, electricity power generation from the dam was estimated based on the data provided by the regional Water Resource Authority. Data for the agricultural sector, including crop prices, crop water requirements, crop yields and input requirements and input prices was partly collected through questionnaires administered to local farmers in the study area for crop year 2013–2014. Equations 6.4–6.30 in Appendix A present a detailed structure of our empirical model. This model captures the preferences of associated stakeholders, which are embedded in the goal equation. We used different weights for the stakeholders in the goal equation to estimate water distribution under different scenarios. In all, we considered seven goal constraints associated with stakeholder priorities, five regular goal constraints, and 13 technical constraints in the model.2

6.4 Results This section presents the results of the evaluation tool used for Mahabad Dam. An important part of the model is associated with implementing the hierarchical and stakeholder analysis as inputs for the second part. We designed a pairwise comparison matrix based on stakeholder priorities and following the hierarchical analysis process. Each stakeholder is defined with a value between 0 and 10 which indicates the level of power and interest to show its place in the power/interest matrix (Fig. 6.5); 0 indicates very low power/interest, and 10 indicates very high power/interest. By multiplying the level of power with the level of interest, we get the level of overall importance for each stakeholder. Once the overall importance is obtained, this value is normalized and the percentage importance of different stakeholders is defined (Table 6.3). In our survey, downstream farmers had the highest value of percentage importance. Agricultural organizations and Mahabad municipalities’ stakeholders had the least value. In the next step, we used stakeholders’ judgment expressed through SA to create a pairwise matrix and thus got the local weights of the criteria (Table 6.4). A comparison of all the criteria and sub-criteria highlights that the priority weight for ‘economy’ was the highest attribute followed by environmental and social aspects. Generally speaking, according to the stakeholders’ opinions and the defined sub-criteria, eco2 For

details, see Appendix A.

88

O. Zamani et al.

Fig. 6.5 Power/interest matrix for the stakeholder groups Note DF, AO, MM, and WA stand for downstream farmers, agricultural organization, Mahabad municipalities and Water Resource Authority, respectively Table 6.3 Importance of the stakeholders Number

Stakeholder

Percentage of importance

1

Downstream farmers

40

2

Agricultural organization

15

3

Mahabad municipalities

15

4

Water Resource Authority

30

Table 6.4 Local weights of criteria Criteria

Economy

Social

Environmental

Weights

0.45

0.23

0.32

Source Own calculation

nomic aspects were the main attributes in planning the allocation of water from the reservoir. After determining the priority weights for the criteria, the next step is calculating the sub-criteria’s weights (Table 6.5). Table 6.5 shows that the ‘environmental’ and ‘social’ attributes had equal weights in their sub-criteria. These results show that gross income was the most important sub-criterion in the ‘economy’ criteria for different alternatives. However, in the ‘social’ criteria, employment and population had an equal weight of 0.5. Similarly, in the ‘environmental’ criteria water salinity and water hardness were equally important among the alternatives.

6 Optimized Reservoir Management for Meeting Conflicting …

89

Table 6.5 Local weights of sub-criteria Criteria

Economy

Sub-criteria

Land use

Gross income

Social Employment

Population

Water salinity

Environmental Water hardness

Weights

0

1

0.5

0.5

0.5

0.5

Source Own calculation Table 6.6 Local weight of each alternative relevant to sub-criteria Alternative

Sub-criteria Land use

Gross income

Employment

Population

Water salinity

Water hardness

Agriculture 1.00

0.50

0.45

0.00

0.50

0.00

R-E

0.00

0.50

0.32

0.00

–

–

Drinking

–

–

–

0.66

0.50

1.00

P-G

–

–

–

0.34

–

–

F-C

–

–

0.23

0.00

–

–

Note R-E Recreational and environmental, P-G Power Generation, F-C Flood control

Table 6.6 illustrates the local weights of all the alternatives. The results in Table 6.6 are calculated based on the weights of alternative criteria and sub-criteria for each alternative. For instance, in the agricultural alternative, the most important subcriterion is land use. Similarly, for recreation, drinking, power generation, and flood control, the most important sub-criteria are gross income, water hardness, population, and employment, respectively. We also calculated the global weight of each item by multiplying the relevant local weights as:

Agriculture

0.45 × 0 × 1 + 0.45 × 1×0.5 + 0.23 × 0.5 × 0.45 + 0.23 × 0.5 × 0 + 0.32 × 0.5 × 0 + 0.32 × 0.5 × 0.5 × 0.5  0.356

Recreational and environmental (R-E)

0.45 × 1×0 + 0.45 × 1×0.5 + 0.23 × 0.5 × 0.32 + 0.5 × 0.23 × 0  0.261

Drinking

0.32 × 0.5 × 0.66 + 0.32 × 0.5 × 1 + 0.32 × 0.5 × 0.5  0.315

Power generation (P-G)

0.23 × 0.5 × 0.34  0.039

Flood control (F-C)

0.23 × 0.5 × 0.23 + 0.32 × 0.5 × 0  0.026

The results of the application show that the weight of the agricultural alternative (0.356) is more than other alternatives, followed by drinking (0.315), recreational–environmental (0.261), power generation (0.039), and flood control (0.026). These findings may be interpreted as stakeholders giving high priorities to agricultural and drinking purposes. Flood control had the least priority for stakeholders.

90

O. Zamani et al.

Table 6.7 displays the optimum land use and cropping patterns proposed by LGP considering the weights of the hierarchical structure and different priorities. Our findings show that wheat was the dominant crop in all the priorities followed by horticultural products, sugar beet, alfalfa, barley, and tomato crops, respectively. These results might be interpreted to show the importance of the agricultural sector for the stakeholders and also wheat’s role in providing food security. According to the results of eighth priority (priority for gross farm income), total farm land (10,135 ha) and farm gross income (315,005 million rial) is higher than in the other models. However, the results of tenth priority (water supply) indicate that total farm land and farm gross income is (water supply) for total farm land and farm gross income is 9,500 ha and 303,340 million rial, respectively. In all the priorities, the goals have unwanted deviations. The tenth and eighth priorities have the highest and lowest deviations from the goal of farm gross income, respectively. Additionally, the highest and lowest deviations from the water supply goal are given seventh and ninth priorities, respectively. Moreover, the maximum and minimum deviations from the entertainment–environmental goal are given the twelfth and ninth priorities at 38,568,000 and 10,478,000 m3 , respectively. Finally, the power generation goal is given the first and seventh priorities, providing the highest and lowest deviations from the goal by 12,443 and 4,261 MW, respectively. Table 6.7 shows the allocation of drinking water based on different priorities over 12 months. Our results suggest to allocate the highest volume of water for drinking purpose according to the ninth priority. As expected, this value in all the priorities is more during the summer season as compared to the rest of the year. Here, it should be mentioned that the three summer months have peak water consumption. The volume of drinking water being distributed in urban areas is at the same level as in the first to seventh priorities (goals of hierarchical structure) and in the eighth and tenth to twelfth priorities as well. We now discuss the results of water distribution for recreational purposes. Table 6.8 presents the volume of water distributed for recreational–environmental purposes. The model suggests that during April–July, the distribution of more and an increasing amount of water was done for achieving this goal. Therefore, we conclude that potentially spring is the best season in this area due to the large volume of water storage in the reservoir (Table 6.9). When the dam water is released for uses including agriculture, flood control, and drinking, the water released through the turbines for electricity generation, drinking, and flood control remains the same. The results of this model are shown in Table 6.10. As expected, the results for power generation show the highest production with respect to priorities for agriculture, flood control, and drinking. Our findings show that the possibility to generate electricity power is higher during April, May, and June due to abundant rainfall. Table 6.11 shows the amount of water required for avoiding floods. In this context, the volume of empty space needed to maintain the flood control capacity of the reservoir for achieving all the priorities is about 89 and 26 million cubic meters in April and May, respectively. This space for February and March is about 27 and 52 million meters, respectively. In this regard, during these four months, a certain

Flood control (m3 )

Power generation (megawatt)

Tank capacity (m3 )

Water drinking (m3 )

15,300,000

11,650

192,400,000

11,108,000

305,160

452

Tomato (ha)

Gross income (M Rials)

2,300

Orchards (ha)

9,422

1,320

Total land use (ha)

1,850

Alfalfa (ha)

850

Sugar beet (ha)

2,650

Barley (ha)

1–7

Priorities

Wheat (ha)

Variables

14,532,000

315,005

10,135

485

2,300

1,800

1,650

1,000

2,900

119,200,000

12,640

5,174,500,000

8

Table 6.7 Results of combination GP and FAHP model

303,340

9,500

500

2,300

1,500

1,600

900

2,700

119,200,000

14,210

172,500,000

17,562,000

9

289,300

8,950

450

2,300

1,350

1,500

850

2,500

119,200,000

14,310

186,200,000

13,245,000

10

280,805

8,650

350

2,300

1,250

1,550

800

2,400

119,200,000

19,832

156,300,000

13,764,000

11

280,805

8,650

350

2,300

1,250

1,550

800

2,400

182,200,000

13,470

185,300,000

13,764,000

12

6 Optimized Reservoir Management for Meeting Conflicting … 91

92

O. Zamani et al.

Table 6.8 Drinking water allocated in different months Month

Alternative 1–7

8

9

10

11

12

April

1,267,000

1,387,000

1,861,000

1,387,000

1,387,000

13,870,000

May

1,132,000

1,243,000

1,673,000

1,243,000

1,243,000

1,243,000

June

1,132,000

1,256,000

1,712,000

1,256,000

1,256,000

1,256,000

July

1,205,000

1,678,000

1,900,000

1,678,000

1,678,000

1,678,000

August

1,254,000

1,697,000

2,000,000

1,697,000

1,697,000

1,697,000

September

1,232,000

1,670,000

1,901,000

1,670,000

1,670,000

1,670,000 1,120,000

October

1,075,000

1,120,000

1,421,000

1,120,000

1,120,000

November

1,080,000

1,125,000

1,505,000

1,125,000

1,125,000

125,000

December

1,071,000

1,150,000

1,437,000

1,150,000

1,150,000

1,150,000

January

1,020,000

1,100,000

1,526,000

1,100,000

1,100,000

1,100,000

February

1,020,000

1,120,000

1,601,000

1,120,000

1,120,000

1,120,000

March

1,020,000

1,140,000

1,591,000

1,140,000

1,140,000

1,140,000

9

10

11

Table 6.9 Volume of reservoir Month

Alternative

April

165,560,000

May

176,710,000 183,650,000 183,650,000 183,650,000 183,650,000 194,171,000

June

153,850,000 163,760,000 163,760,000 163,760,000 163,760,000 174,385,000

July

127,930,000 117,730,000 117,730,000 117,730,000 117,730,000 148,093,000

1–7

8 17,520,000

17,520,000

17,520,000

12

17,520,000 186,156,000

August

86,360,000

97,740,000

97,740,000

97,740,000

97,740,000 116,236,000

September

59,040,000

69,740,000

69,740,000

69,740,000

69,740,000

89,204,000

October

85,690,000

68,900,000

68,900,000

68,900,000

68,900,000

68,769,000

November

51,460,000

61,540,000

61,540,000

61,540,000

61,540,000

61,246,000

December

52,010,000

62,870,000

62,870,000

62,870,000

62,870,000

62,601,000

January

61,950,000

71,430,000

71,430,000

71,430,000

71,430,000

71,295,000

76,430,000

86,270,000

86,270,000

86,270,000

86,270,000

89,943,000

116,730,000 118,930,000 118,930,000 118,930,000 118,930,000

273,000

February March

capacity of the reservoir should be empty. On the other hand, for the rest of the year, because more water is supplied and consumed, this may lead to water scarcity and so the reservoir does not need to remain empty.

6 Optimized Reservoir Management for Meeting Conflicting …

93

Table 6.10 Electricity production in different priorities (unit kW) Month

Alternative 1–7

8

9

10

11

12

8, 9, 11

April

1,231

1,313

1,431

2,074

1,327

1,313

2,413

May

1,252

1,351

1,474

2,085

1,379

1,351

2,451

June

1,146

1,246

1,351

1,984

1,361

1,246

2,346

July

1,047

1,174

1,287

1,954

1,182

1,174

2,274

August

1,056

1,155

1,267

1,943

1,189

1,155

2,255

990

890

983

1,924

901

890

2,190

September October

924

824

943

1,900

882

824

2,124

November

144

134

234

241

179

134

264

December

132

142

278

252

172

142

252

January

128

158

290

231

180

158

247

February

147

157

295

276

173

157

267

March

180

341

487

281

374

341

298

Table 6.11 Volume of empty space for flood control Month

Alternative 1–7

8

9

10

11

12

April

89,000,000

89,000,000

89,000,000

89,000,000

89,000,000

89,000,000

May

26,000,000

26,000,000

26,000,000

26,000,000

26,000,000

26,000,000

June

–

–

–

–

–

–

July

–

–

–

–

–

–

August

–

–

–

–

–

–

September

–

–

–

–

–

–

October

–

–

–

–

–

–

November

–

–

–

–

–

–

December

–

–

–

–

–

–

January

–

–

–

–

–

–

February

27,000,000

27,000,000

27,000,000

27,000,000

27,000,000

27,000,000

March

52,000,000

52,000,000

52,000,000

52,000,000

52,000,000

52,000,000

6.5 Conclusion and Discussion The main purpose of this paper was determining the optimal allocation of water from the reservoir for competing uses. We developed the hybrid lexicographical goal programming and fuzzy AHP models to optimally allocate Mahabad water reservoir’s water for consumption and non-consumption uses (downstream farmers, agricultural organizations, Mahabad municipalities, and the Water Resource Authority). In this

94

O. Zamani et al.

regard, three criteria and six sub-criteria including economy (land use and gross Income), social (employment and population), and environment (water salinity and water hardness) were considered. The five alternatives are agricultural, recreational–environmental, drinking, power generation, and flood control. In terms of the contributions of this study, the hybrid model offers a new approach of combining a multi-criteria analysis (MCA), stakeholder analysis (SA), and goal programming (GP) for allocating Mahabad Dam’s water among stakeholders and crop land use. A simple maximization of the discounted net present value over uses and time periods while respecting constraints on social justice, available water and existing infrastructure is another way of solving this problem that we plan to develop in future investigations. Our results show that the economic criteria had the highest weight (0.45). Hence, the government needs to pay more attention to this aspect. Agriculture at 0.36 had the highest weight among the other priorities. These results highlight that gross income is the most important sub-criterion in the economy criteria for different alternatives. However, in social criteria, employment and population have an equal weight of 0.5. Similarly, in the environmental criteria water salinity and water hardness are equally important criteria. The results of the optimum land use and cropping pattern showed that wheat had the main share of land use under all the priorities followed by horticultural products, sugar beet, alfalfa, barley, and tomato, respectively. However, since some crops such as sugar beet require a lot of water, it is recommended that these be substituted with alternative crops (maize, potatoes, and sunflower). The main results of this study focus on allocation of drinking water, on water distribution for recreational purposes and also on the amount of water required for avoiding floods. Our results show that during April–July, there is distribution of more and an increasing amount of water for recreational purposes. The results for flood control capacity show that the maximum level of the empty space that is needed for maintaining the flood control capacity of the reservoir is in April and March with volumes of about 89 and 52 million cubic meters, respectively. Our results also indicate that downstream farmers had the highest importance. Different studies (e.g., Madani 2014) have shown that Iranian farmers are using water very inefficiently. They practice traditional irrigation and rely on flooding irrigation which increases water wastage by about 65%. This problem can be solved by using new techniques and modern irrigation systems. Iranian farmers are also poor, and they cannot afford modern agricultural practices. Therefore, we suggest that the government finance the necessary technologies for new and efficient irrigation systems. We also suggest increasing greenhouse farming which leads to 90% less water usage because of which about 10 billion cubic meters of water can be saved annually. Another policy implication of our results is that the Ministry of Agriculture Jihad should move the crop production from high-water usage crops such as wheat and sugar beet to crops which are more appropriate for Iran’s dry climate such as canola and saffron. This, of course, might have a negative impact on the food security of the country in the short run.

6 Optimized Reservoir Management for Meeting Conflicting …

95

Appendix A The structure of the model Min Z 

7 

pk (wk dk− , wk dk+ ) +

k1

5 

p j (d − j )

(6.5)

j1

Subject to: w11 X agri + w12 X ent + d1− − d1+  T1

(6.5)

w21 X agri + w22 X ent + d2− − d2+  T2

(6.7)

w31 X agri + w32 X ent + w33 X con + d3− − d3+  T3

(6.8)

w41 X agri + w42 X ent + w43 X drin + w44 X p + d4− − d4+  T4

(6.9)

w51 X agri + w51 X drink + d5− − d5+  T5

(6.10)

w61 X agri + w62 X drink + d6− − d6+  T6

(6.11)

w71 X agri + w72 X ent + w73 X drink + w74 X p + w75 X con + d7− − d7+  T7 6 

(6.12)

a j X j + d8− − d8+  T8

(6.13)

drinki + d9− − d9+  T9

(6.14)

j1 12 

i1 12 

− + Pi + d10 − d10  T10

(6.15)

− + Fi + d11 − d11  T11 i  1, 2, 11, 12

(6.16)

i1 4  i1 12 

− + Si + d12 − d12  T12

(6.17)

i1 6 

X m ≤ LD

(6.18)

m1 6  m1

am j X m ≤ IR j

j  1, . . . , 12

(6.19)

96

O. Zamani et al.

W j ≤ drink j ≤ inf

(6.20)

j

f j ≤ con j

j  1, 2, 11, 12

(6.21)

Smin ≤ s1 ≤ Smax 6 

am j X j + drink j + c j − c j−1 − st1 ≤ flow j

(6.22) j  1, . . . , 12

(6.23)

j1 12 

Pj  α + β

j1

4 

12 

Sj

(6.24)

j1

IRmin ≤ X agri ≤ IRmax

(6.25)

X drink ≤ Wmax

(6.26)

X p ≤ Pmax

(6.27)

X con ≤ Cmax

(6.28)

Smin ≤ X ent ≤ Smax

(6.29)

Or  Orbound

(6.30)

− + Fi + d11 − d11  T11 i  1, 2, 11, 12

(6.31)

i1 12 

− + Si + d12 − d12  T12

(6.32)

i1 6 

X m ≤ LD

(6.33)

m1 6 

am j X m ≤ IR j

j  1, . . . , 12

(6.34)

m1

W j ≤ drink j ≤ inf j

f j ≤ con j

j  1, 2, 11, 12

Smin ≤ s1 ≤ Smax

(6.35) (6.36) (6.37)

6 Optimized Reservoir Management for Meeting Conflicting … 6 

am j X j + drink j + c j − c j−1 − st1 ≤ flow j

97

j  1, . . . , 12

(6.38)

j1 12  j1

Pj  α + β

12 

Sj

(6.39)

j1

IRmin ≤ X agri ≤ IRmax

(6.40)

X drink ≤ Wmax

(6.41)

X p ≤ Pmax

(6.42)

X con ≤ Cmax

(6.43)

Smin ≤ X ent ≤ Smax

(6.44)

Or  Orbound

(6.45)

where Eq. 6.20 represents the goal function of lexicographic goal programming, and Eqs. 6.21–6.27 show the goal constraints of sub-criterions and alternatives of structure of fuzzy analytic hierarchy process. In addition, Eqs. 6.28–6.32 are goal constraints of goal programming and—as mentioned above—T is the amount of each goal. Moreover, Eqs. 6.33–6.45 are systemic constraints, including relation 29 land available to farmers for 6 products, wheat, barley, sugar beet, alfalfa, tomatoes, and orchards; Eq. 6.34 is the maximum capacity of water transport channels for 12 months; Eq. 6.35 illustrated the maximum and minimum capacity of refineries for purification of drinking water and needs of urban and municipal drinking water monthly; the right side of Eq. 6.36 represents the maximum allowable volume of reservoir to control floods during months of April, May, and March. Equation 6.37 represents the maximum and minimum allowable volume of reservoir in order to meet the requirement of entertainment–environmental purpose. Equation 6.38 is the constraint of water uses and shows upper bound of urban drinking water which should be less than the volume of dam in upstream. In this equation c shows the storage transitive variables. Equation 6.39 presents constraint of power generation in which α and β are the estimated parameters, Pj is the generation of electricity in month j, and Sj is the output stream from the dam in different months. Equation 6.40 is the agricultural water supply constraint which is interpreted as the entire agriculture water requirement should be between the minimum and maximum transmission capacity of the canal water irrigation system. Equation 6.41 is urban water supply restrictions in which the whole need of urban drinking water should be smaller and equal to extreme capacity of refineries. Equation 6.42 presents constraint associated with power generation, where electricity generation should be smaller and equal to the maximum power generation electricity in powerhouse of Mahabad Dam. Equation 6.43 implies that reservoir capacity should not be higher than the maximum

98

O. Zamani et al.

capacity of the reservoir to control flooding. Equation 6.44 is another constraint associated with recreational–environment uses that show recreational and environmental requirements of the reservoir volume should be lower than reservoir volume and larger than minimum amount of water requirement in reservoir. Finally, in Eq. 6.45, because of delay in orchard products, acreage land use of these products is constant from the time of planting to fruiting and it considers current acreage land use.

References Antunes P, Karadzic V, Santos R, Beça P, Osann A (2011) Participatory multi-criteria analysis of irrigation management alternatives: the case of the Caia irrigation district, Portugal. Int J Agric Sustain 9(2):334–349 Askın Ö (2008) Analysis of selection criteria for manufacturing employees using fuzzy—AHP. J Bus 9(1):141–160 Badri MA (1999) Combining the analytic hierarchy process and goal programming for global facility location-allocation problem. Int J Prod Econ 62(3):237–248 Balali H, Khalilian S, Viaggi D, Bartolini F, Ahmadian M (2011) Groundwater balance and conservation under different water pricing and agricultural policy scenarios: a case study of the Hamadan-Bahar plain. Ecol Econ 70:863–872 Bertolini M, Bevilacqua M (2006) A combined goal programming AHP approach to maintenance selection problem. Reliab Eng Syst Saf 91(7):839–848 Bozbura FT, Beskese A (2007) Prioritization of organizational capital measurement indicators using fuzzy AHP. Int J Approximate Reasoning 44(2):124–147 Chang CT (2007) Multi-choice goal programming. Omega 35:389–396 Chuang TH, Hsiang HN, Chuang HJ (2007) A multiple-goal programming for nurse scheduling by AHP and simulated anneal. In: Proceedings of the 13th Asia Pacific management conference, Melbourne, Australia, pp 1056–1065 Chung E, Won K, Kim Y, Lee H (2014) Water resource vulnerability characteristics by district’s population size in a changing climate using subjective and objective weights. Sustainability 6:6141–6157 Dinar A, Rosegrant MW, Dick RM (2000) Water allocation mechanisms principles and examples. Policy Research Working Paper, 1779 Ertugrul I, Karakasoglu N (2009) Performance evaluation of Turkish cement firms with fuzzy analytic hierarchy process and TOPSIS methods. Expert Syst Appl 36:702–715 Foltz R (2002) Iran’s water crisis: cultural, political, and ethical dimensions. J Agric Environ Ethics 15:357–380 Grimble R, Chan MK, Aglionby J, Quan J (1995) Trees and Trade-offs: a stakeholder approach to natural resource management. (IIED) Int Inst Environ Dev Sustain Agric Rural Livelihoods Programme, GATEKEEPER SERIES No. SA52 Huang H, Miller GY (2003) Evaluation of swine odor management strategies in a fuzzy multicriteria decision environment. American Agricultural Economics Association Annual Meeting, Montreal, Canada Iran’s Ministry of Agriculture (2015) Annual report Iran’s Ministry of Energy (2015) Annual report Iran’s Ministry of Energy (2016) Annual report Kiker GA, Bridges TS, Varghese A, Seager TP, Linkov I (2005) Application of multicriteria decision analysis in environmental decision making. Integr Environ Manag 1:95–108 Lee SM (1972) Goal Programming for Decision Analysis. Auerbach Publishing, Boston, Philadelphia Pennsylvania

6 Optimized Reservoir Management for Meeting Conflicting …

99

Madani K (2014) Water management in Iran: what is causing the looming crisis? J Environ Stud Sci 1(14):315–328 McDaniels TL, Gregory RS, Fields D (1999) Democratizing risk management: successful public involvement in local water management decisions. Risk Anal 19(3):497–510 Mendelow A (1981) Environmental scanning: the impact of stakeholder concept. In: Proceedings of the second international conference on information systems. December 1981, Cambridge, Mass Rani D, Moreira MM (2010) Simulation-optimization modeling: a survey and potential application in reservoir. Water Resour Manage 24:1107–1138 Reed MS (2008) Stakeholder participation for environmental management: a literature review. Biol Cons 141(10):2417–2431 Reed MS, Graves A, Dandy N, Posthumus H, Hubacek K, Morris J, Prell C, Quinn CH, Stringer LC (2009) Who’s in and why? A typology of stakeholder analysis methods for natural resource management. J Environ Manage 90(5):1933–1949 Rosso M, Bottero M, Pomaric S, La Ferlita S, Comino E (2014) Integrating multicriteria evaluation and stakeholders analysis for assessing hydropower projects. Energy Policy 67:870–881 Saaty TL (2000) Fundamentals of decision making and priority theory, 2nd edn. RWS Publications, Pittsburgh, PA Savenije H, van der Zaag P (2002) Water as an economic good and demand management paradigm with pitfalls. Water International 27:98–104 Schniederjans MJ, Wilson RL (1991) Using the analytic hierarchy process and goal programming for information system project selection. Inf Manag 20:333–342 Singh A (2012) An overview of the optimization modelling applications. J Hydrol 466–467:167–182 Tolga E, Demircan ML, Kahraman C (2005) Operating system selection using fuzzy replacement analysis and analytic hierarchy process. Int J Prod Econ 97(1):89–117 Zahedi F (1986) The analytical hierarchy process: a survey of the method and its applications. Interfaces 16(4):96–108 Zamarrón-Mieza I, Yepes V, Moreno-Jiménez JM (2017) A systematic review of application of multi-criteria decision analysis for aging-dam management. J Clean Prod 147:217–230

Omid Zamani is a Ph.D. candidate in applied economics and agricultural economics (agricultural market analysis). He holds two M.Sc in agricultural economics (2012, Tarbiat Modares University, Iran) and in rural development (jointly by Ghent University, Belgium; Pretoria University, South Africa; Humboldt University of Berlin, Germany; Pisa University, Italy). His current research interests include behavioral economics, price transmission, and water resource policy. Hemen Nader holds a B.Sc in agricultural economics from University of Tabriz and an M.Sc in agricultural economics from Zabol University. His current research interests include agroeconomic modeling and water resource management. Masoomeh Rashidghalam is Researcher at University of Tabriz. She did her B.Sc and Ph.D. in Department of Agricultural Economics at University of Tabriz and holds an M.Sc from Tarbiat Modares University. She was Visiting Scholar in Sweden and South Korea (in Department of Economics at Jönköping International Business School (JIBS) and Department of Economics at Sogang University). She was Assistant in organization of ‘ADB Workshop on Urbanization in Asia’, Seoul. Her areas of expertise are agricultural production economics, farm management, poverty and inequality, child well-being, labor economics, urbanization, and technical efficiency measurement. She has a wide range of teaching experience in econometrics, agricultural production economics, and microeconomics. She has written a book: Measurement and Analysis of Performance of Industrial Crop Production: The Case of Iran’s Cotton and Sugar Beet Production, 2018, published by Springer. She has contributed different chapters to six books.

100

O. Zamani et al.

Frank A. Ward is Professor in the Department of Agricultural Economics and Agricultural Business at New Mexico State University and the Departmental Coordinator for M.Sc and Ph.D. in water science and management programs. He holds a B.Sc, M.Sc, and Ph.D. in economics from Colorado State University. His expertise is water policy. He is the author of numerous journal articles, research reports, and chapters. He has written two books: Valuing Nature with Travel Cost Models. 2000 with D. J. Beal published by Edward Elgar (UK) and Environmental and Natural Resource Economics by Prentice-Hall, scheduled for 2005. He teaches water resource economics and natural resource economics.

Chapter 7

Willingness to Pay for IPM: An Application of the Heckman-Copula Approach Esmaeil Pishbahar, Javad Hosseinzad, Sahar Abedi and Pariya Bageri

7.1 Introduction Nowadays, ensuring food security for a rapidly growing population leads to more use of production inputs. Hence, farmers increase the use of chemical pesticides and fertilizers to improve their yields. Their over and unbalanced use leads to an increase in production costs and water pollution and also decreases soil productivity. Further, it has negative effects on the environment and human health (Rasul and Thapa 2004). These negative effects are reported in many parts of the agriculture sector in Iran. Khuzestan province plays a noticeable role in Iran’s agriculture sector. Each year, farmers use lots of pesticides in different parts of this province which have negative and destructive effects. Hence, it is necessary to opt for sustainable operations. Entomologists believe that the use of pesticides leads to insects becoming resistant to them. Finding methods that reduce the use of pesticides and fertilizers in pest control is one of the main goals of sustainable agriculture, and IPM is one of the product protection systems that is coordinating sustainable agriculture development. This method includes recognizing the ecology of a system, maximizing the use of natural and agronomic controls and using pesticides as necessary. In fact, pest management is the minimum use of pesticides by increasing diversity in biological forms and balancing living organisms in a crop system (Baraki 2010). This method E. Pishbahar (B) · J. Hosseinzad · S. Abedi · P. Bageri Department of Agricultural Economics, University of Tabriz, Tabriz, Iran e-mail: [email protected] J. Hosseinzad e-mail: [email protected] S. Abedi e-mail: [email protected] P. Bageri e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2019 M. Rashidghalam (ed.), Sustainable Agriculture and Agribusiness in Iran, Perspectives on Development in the Middle East and North Africa (MENA) Region, https://doi.org/10.1007/978-981-13-6283-5_7

101

102

E. Pishbahar et al.

Biological method

Mixed cropping method

Integrated Pest Management

Chemical method

Mechanical method

Fig. 7.1 Integrated pest management components. Source Baraki (2010)

is based on both the traditional and chemical methods. Its components are shown in Fig. 7.1. In addition to insects, IPM emphasizes the management of weeds and diseases which is necessary for sustainable agriculture. Programs should also consider economic and social effects because IPM’s efficiency and effectiveness is almost dependent on these factors. IPM which matches sustainable agriculture is one of the best ways to protect products. Therefore, its application can decrease the destructive impact of chemical fertilizers on the environment. In recent years, the Iranian Plant Protection Organization has implemented sustainable operations in Iran to reduce the use of chemical fertilizers. Consequently, identifying WTP’s determinants for reducing pesticide risks on the environment will be helpful in extending IPM. According to economics and the dependent variable identity, Logit, Probit, Tobit and Heckman models are available for assessing the factors that effect WTP. Effective variables of the probability of WTP and their amount are the same in Logit and Probit models; this is a major problem of these models. Hence, it is better to restrict the dependent variables. The Tobit model can help solve sample-selection bias. Nevertheless, effective factors for decisions and their amounts are the same in the Tobit model as well. Heckman (1979) proposed a type-2 Tobit model to cope with this problem. A limitation of the Heckman model is its joint normality assumption in the maximum likelihood estimation method. However, WTP distribution is generally non-normal and skewed and platykurtic distributions are often a better fit (Genius and Strazzera 2011). Violation of the normality assumption leads to inconsistency. Zuehlke and Zeman (1991) show that any violation of the normal assumption leads to biased estimates. For this, semi-parametric or nonparametric approaches are used. According to Ahn and Powell (1993) and Cosslett (1993) in semi-parametric methods, the intercept in the outcome equation is unknown and estimating it is difficult. In addition, the selection of the bandwidth can affect the resulting estimates (Genius and Strazzera 2011), and if the distribution is correctly specified, then the parametric estimation leads to high efficiency (Hasebe and vijverberg 2012).

7 Willingness to Pay for IPM: An Application …

103

To solve the normality assumption problem, Lee (1982, 1983) proposed a twostep method. His approach converted the normality assumption to an assumption of a linear relationship between the error terms. This led to applying the standard normal distribution function’s inverse for the non-normal marginal function proposed by Smith (2003) to model the sample-selection method. Different studies have used the Heckman model for measuring WTP. See, for example, Amigues et al. (2002), Hoffmann and Kassouf (2005), Khodaverdizadeh et al. (2009), and Kong et al. (2014). These studies use the simple Heckman method that OLS estimates the outcome equation using the inverse Mills ratio. WTP’s distribution is not often normal and researchers have not paid attention to this fact. It seems a good point to use the Heckman-copula to achieve reliable results. Very few studies have used the copula approach in the sample-selection model. Genius and Strazzera (2011) applied Heckman-copula to data on female work and contingent valuation data for the recreational value of forests. They showed that the copula is an efficient approach. Hasebe and Vijverberg (2012) proposed a new approach to estimate sample-selection models which combines the Generalized Tukey Lambda (GTL) distributions with the copula. They showed that copula helped in applying best-fitting distributions for marginal. Due to the importance of sustainable agriculture, it is necessary to apply flexible approaches to achieve more reliable results. A few studies have considered the non-normality assumption. However, rejection of the normality assumption leads to inconsistency. We apply the Heckman-copula approach to measure effective factors for farmers’ WTP to avoid pesticides and their environmental risks. Data were collected through a questionnaire distributed to 180 farmers in the Khuzestan province of Iran in 2014. The rest of the paper is organized as follows. Section 7.2 includes a methodology. Results of the study are reported in Sect. 7.3. Section 7.4 gives a conclusion.

7.2 Methodology Following Cuyno et al. (2001), we divided the environment into five relevant impact categories (i  1, …, 5), that is, humans, animals, birds, aquatic species, and beneficial insects. In addition, we identified the level of risk to each category by individual pesticides’ active ingredients (j  1, 2, 3), that is, low-risk  1, moderate-risk  2 and high-risk  3. Therefore, there are 15 environmental classes (5×3) of environmental categories/risk levels for each pesticide. Then, we computed the WTP for avoiding pesticide risks. For this, we asked the farmers about the amount of WTP to avoid a given level of risk in each category. The respondents were asked to determine their WTP for each of the three risk levels (WTPj ) and the importance level to avoid a given level of risk with each of the five environmental categories (Importancei ), so WTP can be written as:

104

E. Pishbahar et al.

Importancei WTPi j  5 × WTP j i1 Importancei

(7.1)

where WTPij is the value of individual WTP for reducing the risks for each of the five environmental categories at each risk level. Since the value of a reduction in pesticide use that has high-risk for human is more than a corresponding reducing pesticide use that is high-risk to other categories, the three WTPj values are not equal among the five environmental categories (Brethour and Weersink 2001). Therefore, we computed WTP’s weighted average. We selected the weights 1, 4 and, 10 for low-risk, moderate-risk and high-risk, respectively. Paying attention to the features of the dependent variable (WTP), we used the sample-selection model to investigate WTP’s determinants. One of the most popular sample-selection models is the Heckman model or the type-2 Tobit model (Hasebe 2013). This model consists of two equations: the first equation is the “selection equation”:  Y1i 

1 if z i γ + vi ≥ 0 0 if z i γ + vi < 0

(7.2)

and the second equation is the “outcome equation” which is observable only for Y1i  1, that is:  Y2i 

xi β + εi if Y1i  1 0 if Y1i  0

(7.3)

Y1i is an indicator of selection (for WTP > 0, Y1i is equal to 1 and for WTP  0, Y1i is equal to 0), Y2i is the amount of WTP, z i and xi are vectors of independent variables, γ and β are vectors of unknown parameters and vi and εi are error terms with zero mean (Genius and Strazzera 2011). When vi and εi are dependent on each other, the OLS estimation of the outcome equation leads to a selectivity bias problem; so, Eqs. 7.2 and 7.3 can be estimated using the full information maximum likelihood method (FIML). The likelihood function of this model can be written as: ⎡  ⎤Y1i 0 ⎡ ⎤Y1i 1 −z  iγ ∞ N  ⎢ ⎢ ⎥ ⎥ L f v (v)dv⎦ f vε (v, εi )dv ⎦ ⎣ ⎣ i1

−∞

(7.4)

−z i γ

where f v is a univariate probability density function (pdf) of v and f vε is a joint pdf of v and ε. To apply the ML estimation, we have to determine the functional forms of f v and f vε . Generally, it is assumed that vi and εi are jointly normally distributed. As a result of this, Eq. 7.4 can be rewritten as:

7 Willingness to Pay for IPM: An Application …

105

  Y1i 1  N     

z − x β γ + (ρ/σ )(Y − x β) Y Y 0 2i 2i 1i i i   i (−z i γ ) L σ −1 φ 2 σ 1 − ρ i1 (7.5) where σ is the standard deviation of ε, ρ is the correlation coefficient between v and ε, φ(·) is pdf and (·) is cdf of the standard normal distribution. This equation is popular in literature, but empirical investigations show that distributions in the real world are rarely in this class; therefore, distributional mis-specification will lead to inconsistent estimates of parameters (Hasebe and Vijverberg 2012). Recent studies use the semiparametric approach to address this wrong assumption; however, this method is not very efficient. Lee (1982, 1983) introduced a new approach to transform non-normal data into random variables which are characterized by joint normal distribution. This approach means using the inverse of standard normal distribution and is a particular case of a copula function (Hasebe 2013). A copula is a multivariate distribution with given marginal distribution. In this function, we can use different kinds of marginal and it creates a more flexible and more realistic multivariate distribution. Let W j be a continuous random variable and its marginal distribution be F j  F(ω j )  Pr(W j ≤ ω j ) for j  1, 2. Then, joint distribution can be defined as F12  F(ω1 , ω2 )  Pr(W1 ≤ ω1 , W2 ≤ ω2 ). A copula function C(·) links two marginal distributions together to produce the joint distribution as: F(ω1 , ω2 )  C(F1 , F2 ; θ )

(7.6)

where θ is a copula parameter and shows the degree of dependence between the variables. The copula function’s properties are: 1. C(F1 , 0; θ )  C(0, F2 ; θ )  0 and C(F1 , F2 ; θ )  C(F2 , F1 ; θ ) 2. C(F1 , 1; θ )  F1 and C(1, F2 ; θ )  F2 3. Copula is a 2-increasing function, that is, ∂ 2 C/∂ F1 ∂ F2 ≥ 0 which ensures a non-negative pdf. Given a joint cdf the partial derivative of a joint CDF is obtained by the chain rule: ∂ ∂ ∂ F1 F12 (ω1 , ω2 )  C(F1 , F2 ; θ ) × ∂ω1 ∂ F1 ∂ω1

(7.7)

∂ F1 where ∂ω is a univariate pdf ( f 1 (ω1 )). Similarly, f 2 (ω2 |W1 ≤ ω1 ) is derived. Due to 1 rewriting Eq. 7.4 in terms of the copula, we can define the integral in Eq. 7.4 as:

∞ f vε (v, εi )dv  −z i γ

∂ (Fε (ε) − Fvε (−z i γ , ε))|εεi ∂ε

(7.8)

106

E. Pishbahar et al.

where Fε (·) is a univariate cdf of ε and Fvε (·) is a bivariate cdf of v and ε. Therefore, the likelihood function can be expressed as: L

N  i1

Fv (−z i γ )

Y1i 0

 1−

 Y1i 1 ∂ C(Fv (−z i γ ), Fε (εi ); θ × f ε (εi ) (7.9) ∂ Fε

The first part of Eq. 7.4 is the cdf of v, Fv (·). There are different kinds of copulas. The Gaussian copula is one of the more frequently used one: C(F1 , F2 ; θ )  2 (−1 (F1 ), −1 (F2 ); θ )

(7.10)

where 2 (·) is a cdf of a joint normal distribution and θ , −1 ≤ θ ≤ 1, is a dependence parameter and the correlation coefficient in the Gaussian copula (Hasebe and Vijverberg 2012). In fact, this copula was applied by Lee (1982, 1983). A Farlie–Gumbel–Morgenstern copula (known as FGM) and a Plackett copula are also used frequently (Hasebe 2013) (See Table 7.1 for details). The Archimedean copula families such as Ali–Mikhail–Haq (AMH), Clayton, Frank, Gumbel, and Joe are other popular copulas that can be obtained by generator functions (ϕ) as: C(F1 , F2 ; θ )  ϕ −1 (ϕ(F1 ) + ϕ(F2 ))

(7.11)

where ϕ : [0, 1] → [0, ∞] is a continuous strictly decreasing convex function such that ϕ(1)  0 and ϕ [−1] is the pseudo-inverse. This function is unique to each Archimedean copula. The partial derivative of this family is: ϕ  (F1 ) ∂ C(F1 , F2 ; θ )   ∂ F1 ϕ (C(F1 , F2 ; θ ))

(7.12)

where ϕ  (·) is the derivative of ϕ(·). The product copula is applied in cases of indifferent random variables. The most appealing feature of the copula is that one can see different dependent structures (Hasebe 2013). The Gaussian, Placket, FGM, and Frank copulas have symmetric dependence structures, that is, their upper and lower tails are the same. However, AHM, Clayton, Gumbel, and Joe are asymmetric. This property may be useful in copula selection. To implement the ML estimation, we have to select the marginal distributions of ε and v, and the appropriate copula form. Marginal distributions may or may not be the same. If the dependence structure is known, the selection of a suitable copula will be easy. However, we rarely have such information. Since copula families in Table 7.1 are not nested relative to each other, we can use AIC or BIC to choose the suitable copula. Otherwise, we can apply the Vuong test. As mentioned earlier, we can select marginal distribution from any univariate distribution. Hence, for the

− log(1 − (1 − t)θ )

1 − ((1 − F1 )θ + (1 − F2 )θ )

Joe

Note For Placket, r  1 + (θ − 1)(F1 + F2 ) θind is the value of θ if variables are independent Reference: Hasebe (2013)

− (1 − F1 ) (1 − F2 ) )

θ 1/θ

exp[−{(− log F1 )θ + 1 (− log F2 )θ } θ ]

Gumbel

θ

(− log(t))θ

− θ1 ln{1 +

Frank t)−1 − log( exp(−θ exp(−θ )−1 )

(t −θ − 1)/θ

(F1−θ + F2−θ − 1)−1/θ

Clayton

(e−θ F1 −1)(e−θ F2 −1) } e−θ −1

log( 1−θ (1−t) ) t

F1 F2 {1 − θ(1 − F1 )(1 − F2 )}−1

AMH

–

r−

Plackett

r 2 −4F1 F2 θ (θ −1) 2(θ −1)

–

F1 F2 {1 + θ(1 − F1 )(1 − F2 )} √

FGM

– –

F1F2

Product

ϕ(t)

Gaussian  (−1 (F ), −1 (F )) P 1 2

Functional form

Copula

Table 7.1 Properties of bivariate copula families

1

1≤θ ≤∞

0

0≤θ ≤∞

1

0

−1 ≤ θ ≤ 1

0≤θ 10. Our super-neutrality tests show that M2’s super-neutrality is rejected with respect to agricultural product both in real and nominal data. The estimated confidence interval of the nominal GDP equation includes one for the most part of the interval 0 < k < 30, while it is rejected for the rest of the range. Super-neutrality of M2 relative to real GDP cannot be excluded.

10.4 Summary and Conclusion According to our findings, we conclude that M2 with respect to the real variables GDP and agricultural production is neutral. This means that permanent changes in M2 in the long run do not have a permanent effect on Iran’s real agricultural production and real GDP. However, neutrality of M2 with respect to agricultural production at current prices is rejected because if we accept the existence of a structural break in

160

E. Pishbahar and Z. Rasouli

the data, neutrality of money with respect to agricultural product at current prices will be rejected with any doubt and if not, only for a small fraction of k estimated confidence intervals include one. The results for nominal GDP vary depending on the type of unit root test. Thus, assuming that there is no structural break in the data generating process, neutrality of M2 can be rejected, while assuming a structural break, its neutrality can be accepted for K > 10. The LRSN results also show that M2’s super-neutrality is confirmed only for real GDP. So, we can conclude that while permanent changes in the growth rate of money have no permanent effect on real GDP in the long run, they have a permanent effect on the real agricultural product. These findings have some policy implications for the government and for policymakers. Hence, policymakers should avoid monetary policies for increasing GDP. These monetary policies can impose burdens of high costs, and price increases in the coming periods. Second, the results of our superneutrality test show that super-neutrality does not hold for Iran’s agricultural sector. This suggests that the Iranian government should use policies other than monetary policies to support growth in its agricultural production.

References Barro RJ (1977) Unanticipated money growth and unemployment in the United States. Am Econ Rev 67(2):101–115 Barro RJ (1978) Unanticipated money, output, and price level in the United States. J Polit Econ 86(4):549–580 Barro RJ, Rush M (1980) Unanticipated money and economic activity. In: Fischer S (ed) Rational expectations and economic policy. University of Chicago Press Bullard J (1999) Testing long-run monetary neutrality propositions: lessons from the recent research. Fed Reserv Bank St Louis Rev 57–77 Chen CH, Steindl FG (1987) Anticipated monetary and fiscal policy effects on output. J Macroecon 9(2):255–274 Fisher ME, Seater JJ (1993) Long-run neutrality and super-neutrality in an ARIMA framework. Am Econ Rev 83(3):402–415 Gordon RJ (1982) Price inertia and policy ineffectiveness in the United States, 1890–1980. J Polit Econ 9(90):1087–1117 Jafari Samimi A, Erfani A (2004) Testing the long-run neutrality and super neutrality of money in Iran. J Econ Stud 67:117–138 (in Persian) Jafari Samimi A, Qhanbarzadeh Niear G (2009) Structural change and testing the hypotheses of rational expectations in Iran. J Econ Stud 88:47–67 (in Persian) Jalali Naini AR, Shiva R (1993) Monetary policy, rational expectations, output and inflation. The third Seminar of monetary and exchange policy, Monetary and Banking Research Institution (in Persian) Khataee M, Danehkar M (1994) Expected and unexpected monetary growth on GNP: case study of Iranian economy during 1971–1990. In: Fourth conference on monetary and foreign exchange rate policy, monetary and banking research institution (in Persian) Lucas R (1972) Expectation and the neutrality of money. J Econ Theory 4(2):103–124 Lucas R (1973) Some international evidence on output-inflation tradeoffs. Am Econ Rev 63(3):326–334

10 Testing for Neutrality and Super-neutrality of Money …

161

McGee RT, Stasiak RT (1985) Does anticipated monetary policy matter? Another look. J Money Credit Bank 17(1):16–27 Mishkin FS (1982a) Does anticipated monetary policy matter? An econometric investigation. J Polit Econ 90(1):22–51 Mishkin FS (1982b) Does anticipated aggregate demand policy matter? Further econometric results. Am Econ Rev Am Econ Assoc 72(4):789–802 Noriega A, Soria LM (2002) Structural breaks, orders of integration, and the neutrality hypothesis: further evidence. Department of Econometrics, Escuela de Economia, Universidad de Guanajuato. Available at: https://www.cass.city.ac.uk/__data/assets/pdf_file/0020/65342/neutbrelondon.pdf Plosser CI (1989) Understanding real business cycles. J Econ Perspect 3(3):51–77 Qadiminia N (1995) Hypothesis of rational expectations and monetary policy: a comparative study of the OPEC and fast-growing Southeast Asia countries. Master thesis in Economics, University of Allameh Tabatabai (in Persian) Sargent TJ, Wallace N (1975) Rational expectations, the optimal monetary instrument and the optimal money supply rate. J Polit Econ 83(2):241–254 Snowdon B, Vane HR, Wynarczyk P (1994) A modern guide to macroeconomics: an introduction to competing schools of thought. Edward Elgar Teshkini A, Shafiee A (2005) Monetary and financial variables of testing money neutrality. Q J Bus 35:125–152 (in Persian)

Esmaeil Pishbahar is Associate Professor and Head of the Department of Agricultural Economics at University of Tabriz, Iran. He holds a B.Sc. in agricultural economics from University of Tabriz and a M.Sc. in agricultural economics from University of Tehran. He did his Ph.D. in science economics at the Departments of Economics and Management, University of Rennes 1, France. His areas of interest and research are applied econometrics, agricultural risk management and insurance, and international trade. His teaching areas are advanced econometrics, mathematical economics, and macroeconomics at under- and postgraduate levels. He has over 100 publications in journals and chapters in books. Zahra Rasouli is Lecturer in the Department of Agricultural Economics at University of Tabriz. She holds her B.Sc., M.Sc., and Ph.D. in the Department of Agricultural Economics, University of Tabriz. Her fields of expertise and interest are agricultural market development, time series modeling, especially price seasonality and Markov switching modeling. She has a wide range of teaching experience in agricultural accounting, agricultural policy, and agricultural insurance at undergraduate level.

Chapter 11

Effects of Oil Prices and Exchange Rates on Imported Inputs’ Prices for the Livestock and Poultry Industry in Iran Esmaeil Pishbahar, Parisa Pakrooh and Mohammad Ghahremanzadeh

11.1 Introduction In Iran, the livestock and poultry industry is a subsector of agriculture. This subsector is considered to be the riskiest economic activity though it plays an important role in the life of the community. Consumption of about 12 kg red meat and 25 kg chicken meat per capita has made these products the most consumable animal protein in the Iranian food basket and has made the livestock and poultry industry the largest industry in the country (Hosseini 2016). Iran’s poultry and livestock industry faces various limitations in the production of inputs like lack of water resources and competition between agricultural products for human consumption and livestock feed (Kamalzadeh et al. 2009). Production units in the agricultural sector import their inputs. In 2014, the country needed 6,413,000 tons of corn since 70% of the poultry eats corn. This meant that the agricultural units had to import their inputs. Due to various constraints faced by the livestock and poultry subsector, many inputs such as soybean, corn, and barley are imported. On the other hand, in the current structure of the global economy, imports play an important role in determining economic development strategies and any changes in a country’s imports have a significant impact on its production, growth, and development process. Therefore, any factor that affects imports will affect production (Kamalabadi and Shahnooshi 2012).

E. Pishbahar (B) · P. Pakrooh · M. Ghahremanzadeh Department of Agricultural Economics, University of Tabriz, Tabriz, Iran e-mail: [email protected] P. Pakrooh e-mail: [email protected] M. Ghahremanzadeh e-mail: [email protected] © Springer Nature Singapore Pte Ltd. 2019 M. Rashidghalam (ed.), Sustainable Agriculture and Agribusiness in Iran, Perspectives on Development in the Middle East and North Africa (MENA) Region, https://doi.org/10.1007/978-981-13-6283-5_11

163

164

E. Pishbahar et al.

The petrochemical and agricultural sectors are the most dependent on oil. Hence, it can be argued that fluctuations in oil prices can affect agricultural commodity prices and inputs. Many researches have studied the impact of changes in oil prices on exchange rates and the price of agricultural inputs including Abbot et al. (2008), Gozgor and Kablamaci (2014), Hanson et al. (1993), Harri et al. (2009), Nazlioglu and Soytas (2011), Schnept (2008), and Trostle (2008). While a significant portion of the inputs for the livestock and poultry industry is imported, changes in oil prices and the exchange rate can also be factors that affect the poultry and livestock industry’s production. Meanwhile, fluctuations and shocks in oil prices due to the start of the Iraq–US wars in 2003, the beginning of the global financial crisis and the rise in global food prices and consequently currencies had a strong influence on global prices (Islamic Parliament Research Center 2009). According to the Food and Agriculture Organization (2010), in the last two decades agricultural policies have supported agricultural activities for increasing food production and using different sources of foreign exchange to reduce the losses caused by price fluctuations. Since 2004, when the price of crude oil and the US dollar started fluctuating impacting countries, policymakers have tried to support the rural sector and agricultural production (Harri et al. 2009). In Iran, over the past 10 years (2005–2014), about 70% corn, 26% barley, and 90% soybean have been imported. Therefore, it seems that one of the reasons for changes in animal feed prices is the high dependence on imported inputs (Javadi and Ghahremanzadeh 2016). Iran spends a large amount of foreign exchange in importing inputs for its livestock and poultry sector. Hence, it needs policies that increase the productivity of local inputs. The country is implementing policies formulated for agriculture sector also in the livestock and poultry industry which is leading to problems. Different researchers have focused on the problems being faced by the livestock and poultry industry in terms of inputs, and some of them have also focused on the role that policymakers play in the production process. The main focus of the policies is increasing production without concentrating on quality leading to socioeconomic problems. The rest of this paper is organized as follows: Section 11.2 gives a literature review on the relationship between oil, exchange rate, and inputs’ markets. Section 11.3 measures the static dependence between oil, exchange rate, and input prices in Iran. Section 11.4 gives the empirical results of the study, and Sect. 11.5 gives a conclusion.

11.2 Literature Review The oil market is one of the most important markets in the world which has an impact on other markets in other countries. There is a great deal of evidence on the globalization of markets through free flow of capital and international trade. Hence, the fluctuations in oil prices can transition to other markets and the degree of this transition increases with expanding communication systems and the high integration of markets with each other, for example, the agriculture market’s integration with other markets.

11 Effects of Oil Prices and Exchange Rates on Imported …

165

In Iran, as in other developing countries, there is high degree of instability in the macroeconomic variables, so in these countries, exchange rates, oil prices, and other important variables have more instability and fluctuations as compared to developed and industrial economies. These fluctuations cause an uncertain situation in the country (Doorandish et al. 2014). According to Hanson et al. (1993) the agricultural market is linked to global markets through intermediate inputs linked to other markets and also through trade. So international trade is important for the agricultural sector in most countries. A decrease in oil prices in world markets will lead to an increase in exchange rates, and the volume of imports of intermediate inputs to other markets will decrease. Assuming a direct and strong correlation between the agricultural and energy markets such as oil, it is expected that any changes in oil prices will also have an impact on other related markets. Figure 3.1 gives the relationship between the oil, exchange rate, and agriculture markets (Fig. 11.1). Several studies have examined the relationship between oil prices and exchange rates with the inputs’ markets (see Table 11.1). According to Chen et al. (2010), Gozgor and Kablamci (2014), Johnson et al. (2011), Khoung et al. (2013), Shams and Zareshenas (2014), and Turhan et al. (2014), the impact of the intensity of oil fluctuations on the exchange rate and commodity prices is not clear, but it seems that it has been greater after the Iraq–US war also called the ‘2000s Energy Crisis’ and the ‘Third Oil Crisis’ and after the beginning of the global financial crisis, which led to an increase in global prices of inputs. According to studies, the poultry and livestock industry in Iran has been both directly and indirectly affected by global fluctuations and crises. Therefore, the aim of this study is analyzing the relationship between oil prices, the US exchange rate, and input prices (corn, soybean, barley, and fish powder) for the livestock and poultry industry in Iran. The study uses monthly data for two periods—the pre- and post-crisis periods of 1995–2003 and 2003–2014. Data was collected from the Central Bank of Iran, Agricultural Jihad Organization, and the Livestock Supporting Company.

Exchange Rate Market

Oil Market

Agriculture Market

Fig. 11.1 Relationship between the oil market, exchange rate market, and agriculture market

166

E. Pishbahar et al.

Table 11.1 Summary of literature Author(s)

Aim

Conclusion

Shakibaei et al. (2008)

Relationship between oil price and US exchange rate, monthly data (1995–2006) for OPEC

The co-integration of the long-run relationship between oil prices and exchange rates was confirmed, and oil prices are the main source of exchange rate fluctuations; there is a long-term relationship between these two variables

Bazazan et al. (2009)

Relationship between oil price and US exchange rate, (1975–2008), Iran

There is a long-term relationship between the US dollar and the price of crude oil, and in the long run, oil prices are a factor affecting the exchange rate, and causality is not seen by the exchange rate to the price of oil

Bakhshoode et al. (2011)

Provide animal feed for the supply of animal protein products in Iran (1989–2004)

Shortage compensation of livestock protein needs to provide inputs and production factors needed for these products, which should be provided through production or imports

Kamalabadi and Shahnooshi (2012)

Investigates the spillover of import prices of soybean meal and poultry products from the world markets to domestic markets in Iran, (2001–2010)

Results have shown that the reduction or increase in global prices for soybean and fish powder has a positive effect on their domestic price changes and production units

Behrad Amin and Zamanian (2014)

Investigating the effect of US exchange rate uncertainty on Iran’s import demand, (1980–2012)

The exchange rate and import variables are co integrated and have a long-term relationship, but in the short run, there is no relationship between uncertainties in exchange rates and imports

Shavalpour et al. (2015)

Investigates the modeling of oil price shocks spillover to the crops market: wheat and soybean meal (2007–2014)

The results of this study show that there is a positive and significant risk spillover between crude oil and agricultural products such as wheat and soybean in Iran (continued)

11 Effects of Oil Prices and Exchange Rates on Imported …

167

Table 11.1 (continued) Author(s)

Aim

Conclusion

Hanson et al. (1993)

The impact of oil price shocks on the American agricultural sector for 20 year

Oil shocks have caused damage to the agricultural sector, so that the dairy, livestock, and poultry industry, and all of them were less productive than before, and the level of prices has increased, which has led to a reduction in the value added of the agricultural sector and ultimately lowered its revenues

Yazdanpanah (1994)

Oil prices and Iran’s agricultural policies (Relationship between imported wheat and oil prices), (1964–1991)

The results indicate that there is an indirect causality from oil prices to imported wheat. And also in this study, the history of past and present oil prices is valuable information for predicting wheat imports

Amano and Norden (1998)

The relationship between oil prices and real US exchange rate changes in Canada, (1972–1993)

The results showed a steady relationship between oil price shocks and real effective exchange rates. At the end, the study suggested that the price of oil is a good source for predicting the behavior of the exchange rate in the future

Baffes (2007)

The spillover of oil prices on other world commodities, (1960–2005)

The results of this study showed that oil price changes had an effect on food commodity indexes and elasticity of commodity had increased. As commodity prices continue to rise, commodity prices will remain high

Zheng and Reed (2008)

To estimate the impact of the world oil price on the price of Chinese agricultural commodities (corn, soybeans, and chicken meat), monthly data (2000–2007)

The results showed that the price of oil did not have a significant effect on the price of agricultural products during this period, but with the increase in oil prices during the 2006 and 2007 years, the costs of production and transportation of goods increased and the price level of food products increased, and if this trend continued, the price of oil will affect the price of agricultural products (continued)

168

E. Pishbahar et al.

Table 11.1 (continued) Author(s)

Aim

Conclusion

Harri et al. (2009)

Relationship between oil price and US exchange rate and agricultural commodities price, monthly data (2000–2008) in USA

The results show that the exchange rate plays an important role in the export and import of goods and services. Also, the price of corn, cotton, and soybeans is related to the price of oil, but the price of wheat is unrelated, and the exchange rate also plays the role of the interface between these price links

Chen et al. (2010)

Relationship between oil price and agricultural commodities price (corn, soybean, wheat and rice) in world, weekly data (1997–2008)

The results show that any changes in the price of oil will change the price of cereals and the intensity of this effect is higher in the period of 2005–2008

Johnson et al. (2011)

Analysis of the net effect of the exchange rate on input prices and agricultural sector prices in two monthly period (1997–2006) and (2007–2011) in USA

The results show that the correlation between the exchange rate and all variables have increased over time, and the increase in exchange rates has a negative effect on the prices of corn, wheat, and seeds. Also, with increasing dependence on imported inputs, the increase in exchange rates has had a negative effect on the price of imported inputs

Nazlioglu (2011)

Study of the nonlinear causal relationship between oil prices and agricultural commodity prices (corn, soybean, and wheat), weekly data (1994–2010) in Turkey

The results show a nonlinear and indirect relationship between oil prices and agricultural commodity prices. Understanding these results helps policymakers and farmers to choose the right policy tools

Nazlioglu and Soytas (2011)

Study of the relationship between oil prices and agricultural commodity prices (corn, soybean, cotton, sunflower, and wheat), monthly data (1994–2010) in Turkey

In the short term, agricultural commodity prices do not respond to oil price shocks, but in the long run, any decrease or increase in the price of oil or lira (Turkish currency) does not affect the price of agricultural goods, so prices for agricultural goods are neutral against oil price shocks (continued)

11 Effects of Oil Prices and Exchange Rates on Imported …

169

Table 11.1 (continued) Author(s)

Aim

Conclusion

Khoung et al. (2013)

Relationship between oil and exchange rate for major countries, daily data (2000–2011)

First, the appropriate functional form is selected and then interpreted. The results show that the functional form of T-copula is an appropriate form for correlating the form and there is a symmetrical and significant relationship between oil price and exchange rate, so with an increase in oil prices the exchange rate is reduced and this issue is more during the global financial crisis

Bakhat and Wurzburg (2013)

Relationship between oil price and cereals, sugar, vegetables, etc., monthly data (2000–2011) in Spain

The results confirm the link between oil prices and commodity prices, and increased consumption of biofuels has created a new relationship between oil and food prices (especially those that produce fossil fuels)

Sensoy et al. (2014)

A comparative analysis of the dynamic correlation between oil prices and exchange rates for G20 members, monthly data (2000–2008)

The results show a strong negative correlation between oil prices and exchange rates, which after the Iraq–US war and the European financial crisis between 2003 and 2008 led to a shift that increased the intensity of correlation

Gozgor and Kablamci (2014)

Relationship between oil price and exchange rate with agricultural commodity prices, monthly data (1990–2013) in Turkey

The results show that the global oil price and the exchange rate have a positive effect on the price of agricultural goods. In general, the results indicate the importance of the impact of world financial markets on agricultural markets, which require more attention

Shams and Zare shenas (2014)

Modeling relationship among oil price, exchange rate, and gold price, monthly data (2001–2008) for two period pre 2007 and post 2007 in world

Modeling the correlation between the variables with copula-GARCH approach. The results show that in both periods, the correlation between oil price, gold price, and exchange rate is significant, and identification of this correlation in economic policies is very important (continued)

170

E. Pishbahar et al.

Table 11.1 (continued) Author(s)

Aim

Conclusion

Boonyanuphong and Sriboonchitta (2014)

Analysis correlation between energy price, biofuel, and agricultural commodity prices (corn, soybean and sugar), daily data (2005–2013) in Thailand

In this study, the Vine copula method based on ARMA-GARCH has been used. According to the C-Vine model, the price of soybeans, corn, and sugar is rising by increasing in the oil and ethanol market

Brayek et al. (2015)

Analysis relationship between oil price and exchange rate for major countries in 3 period, pre-, during, and post-financial crisis, monthly data (2000–2014)

The results of DCC-MGARCH model showed that there was no dependence between oil prices and exchange rates before the crisis, but during the crisis, the price of oil and the exchange rate were correlated, as well as in the post-crisis period of this correlation

Bal and Rath (Bal and Rath 2015a, b)

Analysis of nonlinear relationship between oil price and exchange rate in India and China for monthly data (1994–2013)

The results of both countries have shown that oil prices have an effect on exchange rate regimes, and in India there is a indirect relationship between exchange rate and oil prices, which is not the case in China

11.3 Methodology We modeled the correlation between oil prices, exchange rate, and input prices for the livestock and poultry industry in Iran. Due to limitations of simple correlation coefficients, we used the copula functions because of high flexibility. We used the copula functions based on the ARIMA-MGARCH models (based on high flexibility to find the distribution of correlations and distribution without linear correlation assumptions) because of the abnormal marginal distribution of some variables. Before measuring the correlation of the variables, it was necessary to determine the status of the stationary, seasonal unit root and ARCH behaviors because of the monthly nature of the data. Next, if there is an ARCH behavior in the data series, we can apply the MGARCH models for obtaining the variables’ residuals. As we know, the GARCH and MGARCH models are applied for modeling fluctuations in various markets, so this study also applied the ARIMA-MGARCH models to consider the mean and conditional variance of the variables over time. Finally, the residuals obtained from the ARIMA-MGARCH of each equation were used to model the correlation of variables measured using the copulas functions with the residuals of the ARIMA-MGARCH models.

11 Effects of Oil Prices and Exchange Rates on Imported …

171

(I) Stationary and Seasonal Unit Root Stationary tests are done to prevent spurious regressions. This study applied the augmented Dickey–Fuller test (ADF) and the Kwiatkowski–Phillips–Schmidt–Shin (KPSS) test (Zivot 2006). An economic time series has four components: trend, cyclic movements, seasonal variations, and an irregular random component. The trend component, which indicates the incremental or decreasing change over time, can be attributed to a change in incomes, technology, and consumer tastes. Cyclic movements are related to business cycles and economic reversion over the years, while the seasonal variation component is related to repetitive fluctuations in the series throughout the year. Moreover, the behavior of a time series may be affected by irregular shocks caused by unusual events such as a war, financial crisis, or a famine (Lim and Mcaleer 2000). According to Beaulieu and Miron (1993), most economic time series have seasonal behavior and often include a random and seasonal process. Identifying the existence of a nonstationary stochastic process in time series is done using two ways of examining the sample autocorrelation function (SACF) and applying seasonal unit root tests. The purpose of the seasonal unit root tests is to assume the existence of a single root at a certain frequency without considering the lack of the root in other frequencies. In general, the regression equation of the hypothesis assumes that the roots of the seasonal and non-seasonal units are: (1 − L 12 )X t  α +

11 

δs Ds,t + Lt +

s1

12 

πi yi,t−1 +

i1

p 

φ j (1 − L 12 )X t− j + εt

j1

(11.1) where α is intercept, Dt,s are the monthly dummies, t is the trend, and p is the degree of generalization in Eq. 11.1 to provide the white noise. Equation 11.1 is estimated using the ordinary least squares (OLS) method, while the significance of the roots is evaluated by t and F statistics (Beaulieu and Miron 1993). (II) Fluctuation Modeling To model the fluctuations, after examining the status of the stationary and seasonal unit roots by applying the ARCH test, we first need to determine the presence or absence of the homoscedasticity characteristics of the residuals, thus modeling the fluctuations. We considered fluctuation modeling in some time series variables following Robert Engle’s (1982) ARCH model. Various models based on the ARCH model have been developed and proposed. The multivariate model of generalized autoregressive conditional heterogeneity (MGARCH) is one of the most important applications of this model. The general form of the MGARCH model is: yt  C x1 + εt 1/2

εt  Ht

νt

(11.2)

172

E. Pishbahar et al.

where yt is a m-vector of dependent variables, C is a m × k matrix of parameters, xt is a k-vector of explanatory variables that can include lags of the dependent variable, 1/2 Ht is the Cholesky factor of the time-varying conditional covariance matrix Ht , and νt is a m-vector of zero mean, unit variance, iid innovations. In the general framework, Ht is a matrix general form of the univariate GARCH model. MGARCH and GARCH models are widely used for modeling fluctuations. We can apply the ARMA-GARCH models for considering the mean and conditional variance of the variables over time. So, we can apply the obtained residuals from this model to examine the correlation structure of the variables. For example, Patton (2006) used the ARMA (p, q)-GARCH (1.1) models to estimate the marginal distribution of the dollar and yen. Similarly, Goorbergh (2004) and Jondeau and Rockinger (2006) applied the ARMA (p, q)-GARCH (1.1) models to obtain the structure of correlation in the stock market. The general form of the ARMA-GARCH models is: yt  α0 + α1 yt−1 + εt yt  μ + εt

(11.3)

 εt  z t h t , z t ∼ SkT (ν, γ )

(11.4)

2 h t  t + αεt−1 + βh t−1

(11.5)

Equation 11.3 presents the process of ARMA (p, 0). Equation 11.4 defines the residuals as a result of conditional variance and residuals of z t , and it is assumed that the residuals have a t-distribution with a ν degree of freedom and a γ bias parameter. Equation 11.5 shows the GARCH (1.1) process, in which β ≥ 0 and t , α ≥ 0 2 are necessary conditions for the positive conditional variance or h t > 0. The εt−1 term refers to the ARCH process, α refers to the stability of shocks in the short term, βh t−1 indicates the GARCH process, and β refers to long-term shocks. Therefore, (α + β) is the stability of shocks in the long run and this should be less than one (Boonyanuphong and Sriboonchitta 2014). MGARCH has a variety of models: (1) the Vech model, (2) the diagonal Vech model, (3) the BEKK model, (4) the diagonal BEKK, (5) the fixed conditional correlation model (CCC), (6) the dynamic conditional correlation model (DCC), and (7) the variable conditional correlation model (VCC), in which CCC and DCC models are more widely applied in economic studies due to their high flexibility. (III) Correlation and Copulas Different criteria are used for measuring the dependencies of the variables including Pearson, Spearman, and Kendall correlation coefficients. In the past, correlation coefficients have been used to explain dependencies, but recent research has shown that copulas are more useful because they have certain advantages: (1) They have a high flexibility in modeling and estimating marginal distribution, (2) they are constant during uniform transmissions, and (3) they provide information about the density and structure of the dependency (Shams and Zareshenas 2014).

11 Effects of Oil Prices and Exchange Rates on Imported …

173

Based on the Sklar’s theorem (1959), the joint structure of two (x, y) continuous random variables, Fx y (x, y) with their marginal distributions Fx (x) and Fy (y) can be obtained through the copulas. Copulas connect marginal distributions to the distribution function without limiting the distribution of the margins. In general, a copula is a joint distribution function with uniform marginal distributions of U  Fx (x) and V  Fy (y) as: c(u, v)  Pr[U ≤ u, V ≤ v]

(11.6)

The most important feature of the copulas is that they consider the structure of the tails. The structure of the tails is a criterion of probability which means the tails of two variables in the top or bottom of the distribution are the same (how the two variables are moved together up and down) and can be defined as:  C(u, u) (11.7) λ L  Limu→0 pr[X ≤ Fx−1 (u)Y ≤ Fy−1 (u)]  Limu→0 u  1 − 2u + C(u, u) λU  Limu→1 pr[X ≤ Fx−1 (u)Y ≤ Fy−1 (u)]  Limu→1 (11.8) 1−u Copula functions are divided into two categories—elliptical and Archimedean functions. Implicit copula functions have a definite form and measure the dependency of symmetric tails including normal distribution and student’s t-distribution. However, the functions of the Archimedean copula do not have a definite form and are produced by the generating function and include the functions of the Clayton, Gumble, Frank, and Joe functions as also a combination of these functions (Wu et al. 2012). Multivariate distributions have many limitations in the structure of the correlation between random variables. So, in this case there is a collection of other families of copulas for multivariate modes which are called Vine copulas (including R-Vine, CVine, and D-Vine models), and they are applied under certain conditions of marginal distribution for the construction of multivariate distribution (Boonyanuphong and Sriboonchitta 2014; Kiatmanaroch and Sriboonchitta 2014). The general functional form can be expressed as:  ∂C x,v|v− j (F(x v − j), F(v j |v − j)) (11.9) F(x|v)  ∂ F(v j |v − j) where ν is the conditional variable and C is the multivariate copula distribution function. The parameters of the model and the marginal distribution parameters are estimated using MLE. The MLE functions of R-Vine, C-Vine, and D-Vine copulas are expressed as (Dibmann et al. 2012): L(F, ν, B) 

N  k1

f 1:d (xk |F, ν, B)

(11.10)

174

E. Pishbahar et al.

   (1) 2  L  L + log c z i,k , z i,k ti,k , pi,k L cv (Bcv ) 

N  d−1  d−i 

(11.11)

 

 log ci,i+ j|1:(i−1) Fi|1:(i−1) , Fi+ j|1:(i−1)  Bi,i+ j|1:(i−1)

k1 i1 j1

(11.12) L Dv (B Dv ) 

d−1  d−i N  

 log c j, j+i|( j+1):( j+i−1)

k1 i1 j1



 F j|( j+1):( j+i−1) , F j+i|1( j+1):( j+i−1)  B j, j+i|( j+1):( j+i−1)

(11.13)

(IV) Vine Copulas Comparison Tests (1) The Vuong (1989) test: This test is used for comparing the non-nested models and is based on the likelihood ratio (relates to the ‘Kullback—Libler’ information criterion) which measures the distance between two statistical models. c1 and c2 are the density functions of two competing variables with β1 and β2 estimated parameters. To compute the Vuong test, we must calculate the total sum of standard logarithmic differences of the point-to-point likelihoods. The standard logarithmic difference of i  1, . . . , N , u i, j , j  1, 2 observations is:           m i  log c1 u i,1 , u i,2  Bˆ 1 − log c2 u i,1 , u i,2  Bˆ 2

(11.14)

The sum of the point-to-point likelihoods of standard logarithmic differences or ν as follows is: 1

N

ν n N i1

i1

mi

¯ (m i − m)

2

(11.15)

Vuong showed that ν has an asymptotic normal distribution. Model 1 is preferred

to Model 2 if at the α error level ν > − −1 1 − α2 . Similarly, if ν < − −1 1 − α2 ,

Model 2 is preferred to 1, and if |ν| ≤ − −1 1 − α2 , no decision is possible between the models, so the zero assumption is accepted and both the models are statistically equal. (2) The Clarke (2007) test: This test is also appropriate for comparing two nested models. The zero assumption in Clark’s test is the impossibility of statistical differentiation between the models. In case of statistical equivalence of two models, the exponential ratios are distributed uniformly around the zero distribution. The test statistic in this case will be: B

N  i1

1(0,∞)

(11.16)

11 Effects of Oil Prices and Exchange Rates on Imported …

175

where 1 is the index function proposed by Clark and has a binomial asymptotically distributed with N parameters and p  0.5. If it is not significantly different from the expected value N p  N2 , the first model is interpreted statistically as equivalent to the second model (Brechmann and Schepsmeier 2012).

11.4 Results and Discussion (I) Result of the Stationary and Seasonal Unit Root Tests The results of the stationary tests for variables in the pre- and post-crisis periods at the 5% level while the Dickey–Fuller stationary test showed that the zero hypothesis is based on the existence of the unit root for none of the variables accepted and all variables are not static at the (5%) level. So, we applied the test with one difference, which showed that the variables were static at first level. In the following, the results

Table 11.2 Result of stationary tests (1995–2014) Period

Pre-crisis

Logarithmic Optimal variables lag

ADF

Corn price

At first difference

At level

At first difference

12

−1.64

−8.36**

0.179**

0.050

Soybean price

3

−1.94

−7.18**

0.214**

0.060

Barley price

3

−2.23

−8.36**

0.139**

0.050

Fish powder price

2

−3.41

−8.61**

0.590**

0.036

Exchange rate

0

−2.64

−10.96**

2.140**

0.057

1

−1.71

−7.83**

1.080**

0.040

10

−1.64

−6.49**

0.156**

0.080

Soybean price

9

−1.63

−6.40**

0.183**

0.060

Barley price

5

−1.43

−6.12**

0.239**

0.060

Fish powder price

9

−1.68

−11.24**

0.282**

0.070

Exchange rate

1

−2.89

−14.41**

0.677**

0.030

Oil price

11

−2.02

−8.06**

0.126**

0.050

Oil price Post-crisis

At level

KPSS

Corn price

*p < 0.01; **p < 0.05; ***p < 0.1

176

E. Pishbahar et al.

Table 11.3 Result of seasonal unit root test (1995–2014) Period

Variable

NonSeasonal Seasonal Seasonal Seasonal Seasonal Seasonal seasonal 2 month 4 month 2.4 month 12 month 3 month 6 month root root root root root root root

Precrisis

Corn price (6)*

−1.75 (0.7)

−3.90 11.45 9.17 9.73 12.04 6.5 (0.00)*** (0.00)*** (0.00)*** (0.00)*** (0.00)*** (0.00)***

Soybean price (6)

−2.30 (0.3)

−3.20 8.11 5.17 (0.00)*** (0.00)*** (0.06)*

Barley price (5)

−2.82 (0.15)

−2.46 (0.1)*

11.18 10.85 7.45 12.58 6.71 (0.00)*** (0.00)*** (0.00)*** (0.00)*** (0.02)* *

Fish powder price (3)

−1.27 (0.87)

−2.62 (0.06)*

13.98 6.51 (0.00)*** (0.02)**

Exchange −1.83 rate (0) (0.63)

Postcrisis

8.65 7.14 7.37 (0.00)*** (0.01)** (0.00)***

7.68 6.35 16.32 (0.00)*** (0.02)** (0.00)***

−2.76 8.44 8.17 8.40 8.95 8.00 (0.04)** (0.00)*** (0.00)*** (0.00)*** (0.00)*** (0.00)***

Oil price (3)

−1.50 (0.87)

−2.70 (0.06)*

11.07 12.63 9.46 13.38 13.73 (0.00)*** (0.00)*** (0.00)*** (0.00)*** (0.00)***

Corn price (0)*

−2.40 (0.32)

−2.40 (0.1)*

7.11 12.97 8.97 13.62 6.41 (0.01)** (0.00)*** (0.00)*** (0.00)*** (0.01)**

Soybean price (0)

−1.67 (0.7)

−3.48 14.04 14.04 (0.01)** (0.00)*** (0.01)**

Barley price (0)

−2.33 (0.35)

−3.24 11.38 9.14 8.40 8.20 4.60 (0.02)** (0.00)*** (0.00)*** (0.00)*** (0.00)*** (0.02)**

Fish powder price (3)

−2.04 (0.47)

−2.96 7.85 15.32 (0.03)** (0.00)*** (0.02)**

Exchange −1.17 rate (1) (0.89) Oil price (2)

−2.05 (0.89)

7.17 9.25 14.80 (0.01)** (0.01)** (0.00)***

9.64 5.82 10.10 (0.01)** (0.04)** (0.00)***

−3.03 10.30 8.40 9.90 8.17 7.58 (0.03)** (0.00)*** (0.00)*** (0.00)*** (0.00)*** (0.00)*** −2.50 (0.1)*

10.35 4.41 (0.00)*** (0.01)**

6.63 9.73 7.60 (0.00)*** (0.00)*** (0.00)***

*p < 0.01; **p < 0.05; ***p < 0.1

of the KPSS stationary test are reported which show that the variables were static at the first level. According to the results in Tables 11.2 and 11.3, a comparison of the reported statistics shows that all variables had a non-seasonal unit root. Also, the statistical

11 Effects of Oil Prices and Exchange Rates on Imported …

177

Table 11.4 Result of ARIMA-MGARCH model for variables (1995–2014) Period

Pre-crisis

Post-crisis

Equation

ARMA (p,q)

ARCH

ϕ



α

β

Corn

0.005 (0.24)

0.0001 (0.13)

0.28 (0.00)***

0.71 (0.00)***

Soybean

0.01 (0.13)

0.0001 (0.00)***

0.51 (0.00)***

0.41 (0.00)***

Barley

0.01 (0.00)***

0.003 (0.13)

0.47 (0.00)***

0.4 (0.00)***

Fish powder

0.005 (0.4)

0.001 (0.04)**

0.32 (0.06)*

0.52 (0.00)***

Exchange rate

0.013 (0.28)

0.018 (0.00)***

–

–

Oil

0.021 (0.09)*

0.02 (0.00)***

–

–

Model

CCC-MGARCH

Corn

0.013 (0.00)***

0.002 (0.00)***

–

–

Soybean

0.01 (0.01)***

0.002 (0.00)***

–

–

Barley

0.005 (0.21)

0.0007 (0.00)***

0.97 (0.00)***

–

Fish powder

0.007 (0.04)**

0.001 (0.1)*

0.38 (0.00)***

–

Exchange rate

0.01 (0.02)**

0.002 (0.00)***

–

–

Oil

0.01 (0.3)

0.01 (0.00)***

–

–

Model

CCC-MGARCH

*p < 0.01; **p < 0.05; ***p < 0.1

significance of all the variables indicates that the zero hypothesis of the seasonal test was rejected and none of the variables had seasonal unit roots at any frequencies. Because the computational statistic of the seasonal test is greater than the critical value, the zero hypothesis which indicates a non-seasonal root is rejected for all variables in the pre- and post-crisis periods. We used the ARCH test to assess the existence of linear conditional heteroscedasticity of the variables. The results show that there was conditional heteroscedasticity behavior in the series in both the pre- and post-crisis periods. We applied the ARIMA-MGARCH models for obtaining the residuals of each variable. The estimated α and β coefficients of ARIMA-CCC-MGARCH models for the inputs in the pre- and post-crisis periods were significant and positive, and the sum

178

E. Pishbahar et al.

Table 11.5 Result of correlation with R-Vine model (1995–2014) Period

Variables Tree

Edge

Family

Precrisis

Corn

1

1, 3

Gaussian

1

2, 1

Gaussian

1|2, 3 1

Soybean

Barley

Postcrisis

First parameter

Second parameter

upper tail

Lower tail

Kendall tau

0.07

–

–

–

0.05

0.04

–

–

–

0.03

Gaussian

0.02

–

–

–

0.02

2, 1

Survival Gumble

1.08

–

0.10

–

0.07

1

4, 2

Survival Clayton

0.03

–

–

–

0.01

2|4, 1

Gaussian

−0.09

–

–

–

−0.06

1

1, 5

Frank

1.25

–

–

–

0.14

1

2, 1

Frank

0.54

–

–

–

0.05

1|2, 5

Frank

−0.81

–

–

–

−0.09

Fish powder

1

1, 8

Frank

0.79

–

–

–

0.10

1

2, 1

Frank

0.44

9.53

–

–

0.05

1|8, 2

Gaussian

0.06

–

–

–

0.04

Corn

1

1, 3

Clayton

0.55

–

–

0.28

0.21

1

3, 2

Frank

−1.20

–

–

–

−0.13

3|2, 1

Frank

−0.94

–

–

–

−0.10

1

4, 1

Frank

1.19

–

–

–

0.13

1

1, 2

Frank

−1.11

–

–

–

−0.13

1|2, 4

Clayton 90

−0.06

–

–

–

−0.03

1

1, 5

Clayton

1

2, 1

Frank

1|2, 5 1

1, 6

1

2, 1

Frank

−1.11

–

–

–

−0.13

2

1|6, 2

Gaussian

−0.30

–

–

–

−0.02

Soybean

Barley

Fish powder

0.38

–

–

0.16

−1.11

–

–

–

0.16

Survival Joe

1.18

–

–

0.20

0.10

Survival Joe

1.13

–

–

–

0.07

−0.13

Note Number of variables 1  Oil, 2  Exchange rate, 3  Corn, 4  Soybean, 5  Barley and 6  Fish powder *p < 0.01; ** p < 0.05; *** p < 0.1

of the two parameters was close to one (conditional variance convergence with longrun variance), and in following the residuals were obtained. The residuals of each equation had a net effect on the other variables because the past effects of each variable were taken with the ARIMA process. Hence, the residuals only included the effects of other variables, so these residuals can be used for measuring correlations. The result of ARIMA-CCC-MGARCH models of the variables is reported in Table 11.4. The correlations between the price of each input and the exchange rate and oil price were calculated using the three most commonly used forms of the Vine copulas (R-Vine, C-Vine, and D-Vine) models. We chose the R-Vine model using the Vuong and Clarke tests. Table 11.5 presents the correlation of oil prices and exchange rate with input prices. The correlation of oil prices with corn prices and the exchange rate in the first tree is reported by the Gaussian copula, and the Kendall value can be obtained with a positive correlation between these two variables, so any change in

11 Effects of Oil Prices and Exchange Rates on Imported …

179

average in the price of corn will be related to changes in oil prices. The correlations of oil prices and exchange rate are similar to the correction between corn prices and oil, as there is also a weak correlation between them and any increase in oil prices will increase the exchange rate. According to the second tree, oil prices were affected by the correlation between corn prices and the exchange rate, so any increase in oil prices led to an increase in corn prices and the exchange rate because the value of the copula parameter and the Kendall value of this tree are weak and positive. The second part of Table 11.5 gives the results of the R-Vine model for the post-crisis period. The prices of corn and oil during the post-crisis period showed greater and more positive values, and the Kendall Tau increased from 0.05 to 0.22. In this regard, the correlation between the exchange rate and the price of corn is stronger and more negative than before and Kendall Tau increased from 0.02 to −0.13. This is also seen for other inputs except fish powder.

11.5 Summary and Conclusion Corn, soybean, barley, and fish powder are important inputs for Iran’s livestock and poultry industry; they also account for a large volume of agricultural imports annually. One of the reasons for this is high demand for the production of all types of meat, other products, and also existing food security problems. As input imports are large, their prices are affected by shocks in the global market. This paper analyzed the relationship between oil prices and exchange rate and the livestock and poultry industry’s input prices. The results show that the correlation between oil prices and the exchange rate in the pre-crisis period changed in the post-crisis period. Therefore, based on Bakhat and Wurzburg (2013), Brayek et al. (2015), Boonyanuphong and Sriboonchitta (2014), Chen et al. (2010), Gozgor and Kablamci (2014), Harri et al. (2009); Johnson et al. (2011), Khoung et al. (2013), and Turhan et al. (2014), we can conclude that the correlation between oil prices and the exchange rate because of the Iraq–US war and the beginning of the global financial crisis had a greater effect on input prices such as corn and soybean over a period of time. According to our results, any shock in the global oil market will directly affect domestic input markets, as this will not be beneficial for the agricultural sector, especially the livestock and poultry industry. Since in Iran all inputs for the livestock and poultry industry are imported, we suggest that policymakers should regulate optimal import schedules. This can play an important role in saving foreign exchange and also provide appropriate food security. In general, domestic production and its relative prices which are less than world prices should be encouraged. Finally, we recommend that the government use invest oil revenues in the agricultural sector for domestically producing inputs which in turn will lead to sustainable development in the country.

180

E. Pishbahar et al.

References Abbot P, Hurt C, Tyner WE (2008) What’s driving food prices?. Farm Foundation, Issue Reports July Amano R, Van Norden S (1998) Oil prices and the rise and fall of the US real exchange rate. J Int Money Financ 17(2):299–316 Baffes J (2007) Oil spills on other commodities. The World Bank Policy Research Working Paper Development Prospect Group 4333:1–23 Bakhat M, Wurzburg K (2013) Price relationships of crude oil and food commodities. Working Papers fa06-2013, Economics for Energy Bal DP, Rath BN (2015a) Nonlinear causality between crude oil price and exchange rate: a comparative study of China and India. Energy Econ 51:149–156 Bakhshoode M, Shahnooshi N, Firooz Zare A, Doorandish A, Azinfar Y (2011) Investigating the supply of livestock and poultry food for the supply of animal protein products in Iran. J Agricult Econ 5(4):158–131 (in Persian) Bal DP, Rath BN (2015b) Nonlinear causality between crude oil price and exchange rate: a comparative study of China and India. Energy Econ 51:149–156 Bazazan F, Alienzhad Mehrabani F, Sidizad M (2009) Investigating the long-term relationship between crude oil prices and the real exchange rate of the US dollar. Quart J Energy Econ 6:117–193 (in Persian) Beaulieu JJ, Miron JA (1993) Seasonal unit roots in aggregate US data. Journal of Econometrics 55:305–328 Behrad Amin M, Zamanian G (2014) Exchange rate uncertainty impacts on Iran’s imports demand. Quart J Appl Econ St Iran (AESI) 3(12):129–148 Boonyanuphong P, Sriboonchitta S (2014) An analysis of interdependence among energy, biofuel and agricultural markets using vine copula model. In: Huynh VN, Kreinovich V, Sriboonchitta S (eds) Modeling dependence in econometrics. Advances in Intelligent Systems and Computing, vol 251. Springer, Cham Brayek AB, Sebai S, Naoui K (2015) A study of the interactive relationship between oil price and exchange rate: a Copula approach and DCC-MGARCH model. J Econ Asymmetries 12(2):173–189 Brechmann EC, Schepsmeier U (2012) Modelling dependence with C-vine and D-vine copulas: the r-package CD-vine. To appear in the J Stat Softw Chen ST, Kuo HI, Chen CC (2010) Modeling relationship between the oil price and global food price. Appl Energy 87(8):2517–2525 Clarke K (2007) A simple distribution-free test for nonnested hypotheses. Political Analysis 15(3):347–363 Dißmann J, Brechmann EC, Czado C, Kurowicka D (2012) Selecting and estimating regular Vine Copula and application to financial returns. Comput Stat Data Anal 59:52–69 Doorandish A, Shariat A, Arzande N (2014) The effect of exchange rate fluctuations on the index of agricultural industry accepted in the stock exchange. J Agric Econ Dev 28(2):184–177 (in Persian) Engle FR (1982) Autoregressive conditional heteroscedasticity with estimates of the variance of united kingdom inflation. Econometrica 50(4):987–1007 FAO (2010) Price volatility in agricultural markets, Policy Brief 12 Goorbergh RVD (2004) A copula-based autoregressive conditional dependence model of international stock markets. DNB Working Papers 022, Netherlands Central Bank, Research Department Gozgor G, Kablamaci B (2014) The linkage between oil and agricultural commodity prices in the light of the perceived global risk. Agric Econ 60(7):332–342 Hanson K, Robinson S, Schluter G (1993) Sectoral effects of a world oil price shock: economy wide linkages to the agricultural sector. J Agric Resour Econ 18(1):96–116 Harri R, Nalley L, Hudson D (2009) The relationship between oil, exchange rates, and commodity prices. J Agric Appl Econ 41(2):501–510

11 Effects of Oil Prices and Exchange Rates on Imported …

181

Hosseini A (2016) Monthly Poult Livest J 47:11–49 (in Persian) Islamic Parliament Research Center (2009) Department of infrastructure, Poultry and livestock, 250:1–25 Javadi A, Ghahremanzadeh M (2016) Analysis the fluctuation of poultry inputs industry market with DCC-MGARCH model. In: Tenth biennial conference of agricultural economics Iran May (2016), University of Kerman (in Persian) Johnson M, Anderson D, Bryant H, Herring A (2011) The net effect of exchange rates on agricultural inputs and outputs. In: Poster prepared for presentation at the agricultural & applied economics. Association’s 2011 AAEA & NAREA joint annual meeting, Pittsburgh, Pennsylvania, July 24–26. https://core.ac.uk/download/pdf/6502924.pdf Jondeau E, Rockinger M (2006) The copula-GARCH model of conditional dependencies: an international stock market applicatio. J Int Money Financ 25(5):827–853, Elsevier, August Kamalabadi H, Shahnoshi N (2012) Price transmission of imported poultry sector inputs from global markets to domestic markets: case study of soybean and fish powder. Agric Econ Dev 79 (in Persian) Kamalzadeh A, Rajabbaigy M, Moslehi H, Torkashvand R (2009) Poultry production systems in Iran. In: Book of proceedings 2nd mediterranean summit of WPSA 4:183–188 (in Persian) Khoung ND, Aloui R, Aissa BM (2013) Conditional dependence structure between oil prices and exchange rates: a copula-GARCH approach. J Int Money Finance (32(:719–738 Kiatmanaroch T, Sriboonchitta S (2014) Relationship between exchange rates, palm oil prices and crude oil prices: a Vine Copula based GARCH approach. In: Modeling Dependence in Econometrics, pp 399–413 Lim C, Mcaleer M (2000) A seasonal analysis of Asian tourist arrivals to Australia. Appl Econ 32:499–509 Nazlioglu S (2011) World oil and agricultural commodity prices: evidence from nonlinear causality. Energy Policy 39(5):2935–2943 Nazlioglu S, Soytas U (2011) World oil price and agriculture commodity price: evidence from an emerging market. Energy Econ 33(3):488–496 Patton AL (2006) Modelling asymmetric exchange rate dependence. Int Econ Rev 47:527–556 Schnept R (2008) High agricultural commodity prices: What are the issues? Congressional Research Service May 2008 Sensoy A, Turhan MI, Hacihasanoglu (2014) A comparative analysis of the dynamic relationship between oil prices and exchange rates. J Int Financ Mark I 32(C):397–414, Elsevier Shakibaei A, Aflatooni A, Nikbakht L (2008) Review of long-term relationship between exchange rate and oil prices in OPEC member countries. J Sci Dev 15(25):85–68 (in Persian) Shams S, Zareshenas M (2014) Copula approach for modeling oil and gold price and exchange rate co-movement in Iran. Int J Stat Appl 4(3):172–175 (in Persian) Shavalpour S, Jabbarzadeh A, Khanjarpanah H (2015) Modeling the spillover of oil shocks on crops market: the case of soybean and wheat. Growth Dev Rural Agric Econ 1(2):41–56 (in Persian) Sklar A (1959) Fonctions de répartition à n dimensions et leurs marges. Publ Inst Statist Univ Paris 8:229–231 Trostle R (2008) Global agricultural supply and demand: factors contributing to the recent increase in food commodity prices. Economic Research Service, United States Department of Agriculture Turhan MI, Sensoy A, Hacihasanoglu E (2014) A comparative analysis of the dynamic relationship between oil price and exchange rates. J Int Financ Mark Instit Money 32(c):397–414 Vuong QH (1989) Likelihood ratio tests for model selection and non-nested hypotheses. Econometrica 57(2):307–333 Wu Ch, Chung H, Chang Y (2012) The economic value of co-movement between oil price and exchange rate using Copula-based GARCH models. Energy Econ 34:270–282. Infrastructure Studies of the Islamic Consultative Assembly (2009) Status of Livestock and Poultry Industry Thematic code: 250:25–1 (the Office of Studies on the Context of the Islamic Council 2009) Yazdanpanah A (1994) Oil price and agriculture policy in Iran. Working paper for agriculture and rural development center Iowa State University

182

E. Pishbahar et al.

Zheng Q, Reed M (2008) Examining the Impact of the world crude oil price on Chinas agricultural commodity prices: The case of corn, soybean, and pork. Selected paper for presentation at the Southern Agricultural Economics Association Zivot E (2006) Unit-root and stationary tests. Unit root Lecture Washington

Esmaeil Pishbahar is Associate Professor and Head of Department of Agricultural Economics at University of Tabriz, Iran. He holds a B.Sc. in Agricultural Economics from University of Tabriz and a M.Sc. in Agricultural Economics from University of Tehran. He did his Ph.D. in Science Economics at departments of Economics and Management, University of Rennes 1, France. His areas of interest and research are Applied Econometrics, Agricultural Risk Management and Insurance, and International Trade. His teaching area are Advanced Econometrics, Mathematical Economics, and Macroeconomics at under- and postgraduate levels. He has over 100 publications in journals and chapters in books. Parisa Pakrooh is a Ph.D. candidate at the University of Tabriz, where she furthers her research on Natural Resource Economics and Game Theory. She has a Bachelor’s degree is in Agricultural Economics from University of Tabriz. She has a Master’s degree in Agricultural Policy from University of Tabriz. She has been researching Agricultural Policy and Natural Resource Economics. Mohammad Ghahremanzadeh is an Associate Professor in Department of Agricultural Economics at University of Tabriz. He holds a B.Sc. from University of Tabriz and M.Sc. and Ph.D. from University of Tehran. He spent six months in Australia as Research Scholar to complete his thesis at University of Queensland. His fields of expertise include Agricultural Policy, Agricultural Price Analysis, Agricultural Insurance and Risk Management and (seasonal) time series modeling. He has co-supervised over 30 M.Sc. and 4 Ph.D. students.

Index

A Agricultural insurance, 12 Agricultural land management, 49 Agricultural markets, 118 Agricultural price stabilization policies, 117 Agricultural producers, 29 Agricultural productivity, 12 Akaike Information Criterion (AIC), 54 Alternative specific constant, 70 Analytic hierarchy process, 136 Arasbaran forests, 64 Archimedean copula, 12, 106 Archimedean copula families, 108 Archimedean copulas, 14 ARIMA model, 18 Asian financial crisis, 30 Asymptotic critical values, 123 Asymptotic normal distribution, 174 Asymptotic variances, 122 Auxiliary regression, 123 B Bartlett estimator, 155 Bayesian Information Criterion (BIC), 54 BDS test, 121 Bequest value, 64 Bilateral price differentials, 124 Binary spatial logit model, 51 Biodiversity, 73 Bivariate copulas, 16 Block Maxima, 31

Business cycle, 151, 153 C Caspian forests, 64 Cause-and-effect relationships, 137 Central Bank of the Islamic Republic of Iran, 155 Choice cards, 67 Choice experiment, 66 Choice probabilities, 68 Clark’s test, 174 Climatic diversity, 49 Climatic variables, 49 Cointegration analysis, 121 Cointegration models, 119 Compensative surplus, 65, 66 Compensative variation, 66 Conditional correlation model, 172 Conditional variance, 172 Conservation reserve program, 48 Constant parameter, 69 Constant relative prices, 120 Contingent ranking approach, 65 Contingent rating, 66 Contingent valuation methods, 65 Copula modeling, 13 Covariance matrix, 33 Cropping patterns, 54 Cross-sectional data, 48 Customer loyalty, 136 Cyclic movements, 171

© Springer Nature Singapore Pte Ltd. 2019 M. Rashidghalam (ed.), Sustainable Agriculture and Agribusiness in Iran, Perspectives on Development in the Middle East and North Africa (MENA) Region, https://doi.org/10.1007/978-981-13-6283-5

183

184 D Decision-making process, 83 Decision-Making Trial and Evaluation Laboratory- Analytic Network Process, 136 Decision matrix, 141 Deforestation, 64 Degree of freedom, 108 Demand curves, 64 Density function, 14 Dependence modeling, 34 Dependence structure measurement methods, 13 Dependence structures, 35 Dependent variables, 30 Deterministic difference, 120 Developing countries, 13 Developing economies, 118 Development strategies, 163 DF-GLS unit root test, 124 Diagonal-Vech model, 172 Direct Impact Matrix, 137 Direct-consumption value, 64 Dynamic conditional correlation matrix, 33 Dynamic conditional correlation model, 172 E Economic shocks, 117 Economic value, 64 Effective decision-making processes, 81 Elliptical and Archimedean functions, 173 Elliptical copulas, 14 Endogenous variable, 151 Environmental economics, 63 Environmental problems, 47 Environmental valuations, 65 Equal distribution of incomes, 117 Error correction models, 119 Existence value, 64 Explanatory variables, 52 Exponential distribution, 32 Extreme values, 68 Extreme Value Theory (EVT), 30 F Factorial design, 67 Farlie–Gumbel–Morgenstern copula, 106 Farmland development, 48 Fat-tailed distributions, 30 Fexible multivariate stochastic models, 13 Financial markets, 29

Index Financial risks, 29 Flexible multivariate distribution, 13, 31 Food and Agriculture Organization, 164 Food security, 49 Foreign exchange, 164 Fractional dependent variables, 53 Fractional multinomial logit model, 48 Free flow of capital, 164 F-type test statistic, 121 Fuzzy analytic hierarchy process, 82 Fuzzy Delphi, 136 G Gaussian, 14 Gaussian copula, 106 Generalized Extreme Value (GEV) distribution, 31 Generalized method of moments, 51 Generalized pareto distribution, 32 Generalized Tukey Lambda (GTL), 103 Geographical markets, 118 Geographic information system, 52 Global economy, 163 Global financial crisis, 164 Global food prices, 164 Goal programming, 81 Granger Causality test, 119 Gross domestic production, 49 H Habitat value, 64 Hausman test, 68 Heavy-tailed distribution, 32 Hierarchical analysis, 84 Hierarchical problems, 84 Hierarchical structure, 85 Homoscedasticity, 171 Hypothesis of linearity, 121 Hypothetical market, 66 I Identity matrix, 50 Implicit prices, 66 Indemnity function, 18 Independence from irrelevant alternatives, 68 Independent and identical distribution, 121 Independently and identically distributed, 30 Index-based insurance, 12 Indirect-consumption value, 64 Information transmission, 117 Integrated fuzzy decision model, 137

Index Interest rate, 153 Intergovernmental Panel on Climate Change, 47 International trade, 164 Inverse Mills ratio, 103 Iranian Plant Protection Organization, 102 Iran Meteorological Organization, 52 Iran’s Ministry of Agriculture, 80 Iran’s Ministry of Energy, 80 Irregular shocks, 171 J Johansen test, 121 Jointly normally distributed, 104 Judge’s point-location model, 119 K Kendall correlation coefficients, 172 Kendall’s, 17 Keynesian school, 151 KPSS stationary test, 176 ’Kullback—Libler’ information criterion, 174 L Lag length, 155 Lexicographical goal programming, 82 Likelihood-ratio test, 108 Limited dependent variable models, 49 Limited dependent variable regression models, 48 Linear relationship, 103 Loading factor, 18 Local inputs, 164 Logistic distribution, 108 Logit-t models, 110 Long-run derivative, 153 Long-run price linkages, 121 Long-term financial resources, 29 M Marginal density functions, 14 Marginal distribution, 172 Marginal significance value, 125 Marginal utility, 66 Market integration, 117 Marketing decisions, 135 Market performance, 118 Maximum likelihood, 17 McCarthy’s 4Ps, 135 McFadden’s conditional logit regression model, 68 McGee-Stasiak approach, 152 Medicinal plants, 64 MGARCH models, 170

185 Mis-specification test, 121 Mixed logit regression model, 68 Monetary coefficient, 66 Monetary policy, 152 Money stock, 152 Monthly dummies, 171 Moving average term, 32 Multi-attribute valuation method, 66 Multi-criteria analysis, 81 Multi-criteria decision making, 84, 136 Multivariate copulas, 13 Multivariate data, 13 Multivariate distributions, 173 Multivariate-GARCH, 30 N Negative externalities, 80 Negative ideal solutions, 142 Net value of losses, 64 New Keynesians, 151 New macroeconomic theory, 152 Non-consumable values, 64 Non-homogeneous commodity, 120 Non-market services, 66 Non-monetary coefficient, 66 Non-normal marginal function, 103 Non-numerical criteria, 141 Non-parametric approaches, 102 Non-stationary tests, 124 Normal distribution assumption, The, 30 Normal logit model, 69 O Oil prices, 164 Opportunity costs, 81 Option value, 64 Ordinary Pareto distribution, 32 P Pair-copula, 17, 34 Pair copula constructions, 14 Panel data unit root tests, 119 Partial factorial design, 67 Peak Over Threshold, 31 Perfect market integration, 120 Pesticide risks, 103 Plackett copula, 106 Platykurtic distributions, 102 Point-to-point likelihoods, 174 Portmanteau test, 121 Positive-definite matrix, 33 Positive ideal solution, 143 Preserving environmental resources, 63 Price convergence, 119

186 Price factor, 66 Price signals, 118 Probability density function, 104 Probit models, 102 Probit-t model, 110 Production process, 164 Proximity condition, 16, 34 Pseudo-observations, 17 Q Quality differences, 120 R Random coefficient form, 68 Random utility model, 68 Ravallion’s model, 119 Recreational value, 65 Regular vine copula, 13, 16 Relative parameters, 66 Renewable natural resources, 73 Reservoir management, 81 Risk levels, 103 Risk premium, 120 Rural incomes, 48 R-vine, 16 S Sample autocorrelation function, 171 Seasonal unit root, 170 Seasonal variations, 171 Seemingly unrelated regression, 152 Selection criteria, 155 Sequential method, 17 Short-run price dynamics, 121 Simple copula functions, 31 Sklar’s theorem, 14 Slope parameter, 155 Smooth transition auto regressive, 122 Smooth transition vector error correction model, 119 Socioeconomic characteristics, 65 Socioeconomic problems, 164 Spatial autoregressive model, 50 Spatial autoregressive parameter, 50 Spatial contiguity matrix, 50 Spatial econometrics, 50 Spatial market integration, 117 Spatial price linkages, 119 Spatial trade, 120 Spatial weight matrix, 52 Stakeholder analysis, 81, 82

Index Stakeholder mapping, 83 Standard fractional multinomial logit model, 52 Standardized residuals, 33 Standard logarithmic differences, 174 Standard normal distribution, 105 Stated preference method, 66 Stationary series, 119 Statistical Center of Iran, 49 Statistical design theory, 67 Statistical methods, 48 Stochastic trends, 121 Stochastic volatility, 30 Stock market, 29 Structural breaks, 155 Structural equation, 48 Student-t copulas, 14 Super Decision software, 137 Super matrix, 137 Sustainability, 73 Sustainability of production, 29 Sustainable agriculture, 117 Sustainable development, 11, 63 Sustainable growth and development, 29 T Tail dependency, 14 Tail risk, 30 Taylor approximation, 123 Technical changes, 48 Technique for Order of Preference by Similarity to Ideal Solution, 136 Third Oil Crisis, 165 Three-dimensional joint density function, 14 Tobit model, 102 Total Impact Matrix, 138 Transaction costs, 12 Transportation costs, 120 Travel cost method, 65 Trigger values, 18 T-type statistic, 123 Two-step estimation method, 51 2000s Energy Crisis, 165 Type-2 Tobit model, 104 U Univariate GARCH model, 30 Univariate margins, 14 Utility function, 68 V Variable conditional correlation model, 172

Index Variance–covariance matrix, 50 Vuong statistic, 110 Vuong test, 106 W Water crises, 79 Water productivity, 80 Water Resource Authority, 87

187 Water resource management, 81 Weather-induced supply shocks, 117 Weighted super matrix, 140 White noise, 171 Z Zivot-Andrews test, 155