WORLD BANK TECHNICAL PAPER NO. 402 Work in progress W TPLI40 for public discussion March jqq3 Measuring the Impact of Climate Change on Indian Agriculture .1Ird11wi );, Robc,rtl I Ick/c/hin. Robert Ec e,,soli, .Jyofi Parikh,. - Iptrci Sa~-l Ixaz'i Kin//a(l; *(1111tW JA,usis,. .Stephw L/ ...O// RECENT WORLD BANK TECHNICAL PAPERS No. 327 Valdes and Schaeffer in collaboration with Martin, Surveillance of Agricultural Price and Trade Policies: A Handbookfor Paraguay No. 328 De Geyndt, Social Development and Absolute Poverty in Asia and Latin America No. 329 Mohan, editor, Bibliography of Publications: Technical Department, Africa Region, July 1987 to April 1996 No. 330 Echeverria, Trigo, and Byerlee, Institutional Change and Effective Financing of Agricultural Research in Latin America No. 331 Sharma, Darnhaug, Gilgan-Hunt, Grey, Okaru, and Rothberg, African Water Resources: Challenges and Opportunitiesfor Sustainable Development No. 332 Pohl, Djankov, and Anderson, Restructuring Large Industrial Firms in Central and Eastern Europe: An Empirical Analysis No. 333 Jha, Ranson, and Bobadilla, Measuring the Burden of Disease and the Cost-Effectiveness of Health Interventions: A Case Study in Guinea No. 334 Mosse and Sontheimer, Performance Monitoring Indicators Handbook No. 335 Kirmani and Le Moigne, Fostering Riparian Cooperation in International River Basins: The World Bank at Its Best in Development Diplomacy No. 336 Francis, with Akinwumi, Ngwu, Nkom, Odihi, Olomajeye, Okunmadewa, and Shehu, State, Community, and Local Development in Nigeria No. 337 Kerf and Smith, Privatizing Africa's Infrastructure: Promise and Change No. 338 Young, Measuring Economic Benefitsfor Water Invesrmenits and Policies No. 339 Andrews and Rashid, The Financing of Pension Systems in Central and Eastern Europe: An Overview of Major Trends and Their Determinants, 1990-1993 No. 340 Rutkowski, Changes in the Wage Structure during Economic Transition in Central and Eastern Europe No. 341 Goldstein, Preker, Adeyi, and Chellaraj, Trends in I-Iealth Status, Services, and Finance: The Transition in Central and Eastern Europe, Volurme I No. 342 Webster and Fidler, editors, Le secteur informel et les institutions de microfinancement en Afrique de l'Ouest No. 343 Kottelat and Whitten, Freshwater Biodiversity in Asia, with, Special Reference to Fish No. 344 Klugman and Schieber with Heleniak and Hon, A Survey of Health Reform in Central Asia No. 345 Industry and Mining Division, Industry and Energy Department, A Mining Strategyfor Latin America and the Caribbean No. 346 Psacharopoulos and Nguyen, The Role of Government and the Private Sector in Fighting Poverty No. 347 Stock and de Veen, Expanding Labor-based Methodsfor Road Works in Africa No. 348 Goldstein, Preker, Adeyi, and Chellaraj, Trends in Health Status, Services, and Finance: The Transition in Central and Eastern Europe, Volume II, Statistical Annex No. 349 Cummings, Dinar, and Olson, New Evaluation Proceduresfor a New Generation of Water-Related Projects No. 350 Buscaglia and Dakolias, Judicial Reform in Latin American Courts: The Experience in Argentina and Ecuador No. 351 Psacharopoulos, Morley, Fiszbein, Lee, and Wood, Poverty and Income Distribution in Latin America: The Story of the 1980s No. 352 Allison and Ringold, Labor Markets in Transition in Central and Eastern Europe, 1989-1995 No. 353 Ingco, Mitchell, and McCalla, Global Food Supply Prospects, A Background Paper Preparedfor the World Food Summit, Rome, November 1996 No. 354 Subramanian, Jagannathan, and Meinzen-Dick, User Organizations for Sustainable Water Services No. 355 Lambert, Srivastava, and Vietmeyer, Medicinal Plants: Rescuing a Global Heritage No. 356 Aryeetey, Heitige, Nissanke, and Steel, Financial Market Fragmentation and Reforms in Sub-Saharan Africa No. 357 Adamolekun, de Lusignan, and Atomate, editors, Civil Service Reform in Francophone Africa: Proceedings ofa Workshop Abidjan, January 23-26,1996 No. 358 Ayres, Busia, Dinar, Hirji, Lintner, McCalla, and Robelus, Integrated Lake and Reservoir Management: World Bank Approach and Experience (List continues on the inside back cover) WORLD BANK TECHNICAL PAPER NO. 402 Measuring the Impact of Climate Change on Indian Agriculture Ariel Dinar, Robert Mendelsohn, Robert Evenson, Jyoti Parikh, Apurva Sanghi, Kavi Kumar, James McKinsey, Stephen Lonergan The World Bank Washington, D.C. Copyright © 1998 The International Bank for Reconstruction and Development/THE WORLD BANK 1818 H Street, N.W. Washington, D.C. 20433, U.S.A. All rights reserved Manufactured in the United States of America First printing March 1998 Technical Papers are published to communicate the results of the Bank's work to the development community with the least possible delay. The typescript of this paper therefore has not been prepared in accordance with the procedures appropriate to formal printed texts, and the World Bank accepts no responsibility for errors. Some sources cited in this paper may be informal documents that are not readily available. The findings, interpretations, and conclusions expressed in this paper are entirely those of the author(s) and should not be attributed in any manner to the World Bank, to its affiliated organizations, or to members of its Board of Executive Directors or the countries they represent. The World Bank does not guarantee the accuracy of the data ncluded in this publication and accepts no responsibility for any consequence of their use. The boundaries, colors, denominations, and other information shown on any map in this volume do not imply on the part of the World Bank Group any judgment on the legal status of any territory or the endorsement or acceptance of such boundaries. The material in this publication is copyrighted. Requests for permission to reproduce portions of it should be sent to the Office of the Publisher at the address shown in the copyright notice above. The World Bank encourages dissemination of its work and will normally give permission promptly and, when the reproduction is for noncommercial purposes, without asking a fee. Permission to copy portions for classroom use is granted through the Copyright Clearance Center, Inc., Suite 910, 222 Rosewood Drive, Danvers, Massachusetts 01923, U.S.A. ISSN: 0253-7494 Ariel Dinar is a senior economist in the World Bank's Rural Development Department. Robert Mendelsohn and Robert Evenson are professors and James McKinsey is a visiting lecturer at Yale University. Jyoti Parikh is a professor and Kavi Kumar is a researcher at the Indira Gandhi Institute of Development Research in India. Apurva Sanghi is a researcher at the University of Chicago. Stephen Lonergan is a professor at the University of Victoria in British Columbia. Library of Congress Cataloging-in-Publication Data Measuring the impact of climate change on Indian agriculture / Ariel Dinar . .. [et al.]. p. cm. - (World Bank technical paper; no. 402) Includes bibliographical references. ISBN 0-8213-4192-8 1. Climate changes-India. 2. Crops and climate-India. 3. Agriculture-India. I. Dianr, Ariel, 1947- . II. Series. S600.7.C54M435 1998 338.1'4-dc2l 98-14309 CIP TABLE OF CONTENTS FOREWORD ....... ................................................................... vi ABSTRACT ....... .................................................................... vii ACKNOWLEDGMENT ......... ................................................................. ix 1 OVERVIEW .......................................................................... 1 INTRODUCTION .......................................................................... 1 THE ANALYTICAL FRAMEWORK .......................................................................... 2 SUMMARY OF THE STUDIES' RESULTS .......................................................................... 4 CONCLUSION .......................................................................... 6 REFERENCES .......................................................................... 6 2 MEASURING THE IMPACT OF AND THE ADAPTATION TO CLIMATE CHANGE IN AGRICULTURE AND OTHER SECTORS - LITERATURE REVIEW .......................................................................... 9 INTRODUCTION .......................................................................... 9 CLIMATE CHANGE ........................................................................... 9 STUDIES ......................................0.................................... 0 MEASURING CLIMATE CHANGE (MODELS) .................................. 12 IMPACT OF CLIMATE CHANGE .................................. 14 IMPACT ON NON-AGRICULTURE SECTORS .................................. 14 IMPACT ON AGRICULTURE .................................. 16 FOOD SECURITY .................................. 17 RESPONSES AND ADAPTATIONS TO CLIMATE CHANGE ................................................................... 18 RESPONSES AND ADAPTATIONS IN NON-AGRICULTURE SECTORS ........................................... 18 RESPONSES AND ADAPTATIONS IN AGRICULTURE ................................................................. 19 RESPONSES OF THE INTERNATIONAL TRADE AND POLICY COMMUNITY ........................................ 20 DISCUSSION .......................................................................... 22 RFERENCES CITED .......................................................................... 23 ADDITIONAL REFERENCES ONTOPIC ......................................................................... 29 3 CLIMATE WARMING AND INDIA .......................................................................... 33 INTRODUCTION ......................................................................... 33 THE GENERAL CASE OF CLIMATE WARMING ......................................................................... 33 THE GLOBAL WARMING CONTROVERSY .......................................................................... 36 GENERAL CIRCULATION MODELS ......................................................................... 36 CLIMATE CHANGE PROJECTIONS FOR INDIA ......................................................................... 38 TEMPERATURE ......................................................................... 38 PRECIPITATION ......................................................................... 40 EVAPORATION ......................................................................... 44 SOIL MOISTURE ......................................................................... 44 CONCLUSION .......................................................................... 44 APPENDIX A: FIGURES AND TABLES OF CLIMATE WARMING, USING A NORTH/SOUTH INDIA BREAKDOWN ......................................................................... 51 REFERENCES .......................................................................... 67 4 THE CLIMATE SENSITIVITY OF IND IAN AG RICULTURE ......................................................................... 69 INTRODUCTION ......................................................................... 69 CLIMATE CHANGE AND AGRICULTURE .................................................... ..................... 69 Two APPROACHES ......................................................................... 69 THE ECONOMIC FRAMEWORK ................................. 70 THE PRODUCTION FUNCTION APPROACH ................................. 70 THE RICARDIAN METHODOLOGY ................................ 71 COMBINING THE TWO APPROACHES ................................ 75 DEVELOPING COUNTRY MODIFICATIONS ................................ 75 iii DATA AND EMPIRICAL SPECIFICATIONS ........................................................ 76 UNITS OF ANALYSIS ........................................................ 76 THE DEPENDENT VARIABLE: NET REVENUE PER CROPPED HECTARE . . 77 CLIMATE VARIABLES: ESTIMATING A CLIMATE SURFACE . .................................................... 77 EDAPHIC AND OTHER CONTROL VARIABLES ........................................................ 90 EMPIRICAL RESULTS ........................................................ 90 RICARDIAN CLIMATE REGRESSIONS ........................................................ 90 INTERPRETATION OF CONTROL VARIABLES AND ROBUSTNESS ISSUES . . 92 SIMULATION OF NET REVENUE PER HECTARE ........................................................ 97 SIMULATION OF NET REVENUE PER HECTARE ........................................... 99 REGIONAL DISTRIBUTION OF SEASONAL IMPACTS ........................................ 99 OVERALL REGIONAL IMPACTS ..................................................... 107 IMPLICATIONS FOR GLOBAL WARMING-...................................................................................... 107 CONCLUSIONS AND POLICY IMPLICATIONS ......................................... 108 APPENDIX A: PRODUCTION FUNCTION ESTIMATES ........................................ 111 APPENDIX B: VARIABLES IN THE ORIGINAL DATA SET ....................................... 116 APPENDIX C: ERRORS IN ORIGINAL DATA SET ........................................ 125 APPENDIX D: CLIMATE VARIABLES .. ...................................... 129 APPENDIX E: EDAPHIC VARIABLES .. ...................................... 133 A. SOILTYPE ........................................ . 133 B. STORIE INDEX ........................................ 134 C. SOIL FERTILITY STATUS ........................................ 135 D. SOILPH ........................................ 136 E. AQUIFERS ........................................ 136 F. TOPSOIL DEPTH ........................................ 136 REFERENCES ......................................... 137 5 CLIMATE CHANGE IMPACTS ON INDIAN AGRICULTURE: THE RICARDIAN APPROACH ................. 141 INTRODUCTION ........................................................................ 141 CLIMATE CHANGE IMPACTS ON AGRICULTURE - METHODOLOGICAL BACKGROUND ........... 142 OBJECTIVES .............................................................................. 143 THE PRODUCTION FUNCTION APPROACH .................................................. 143 THE NET-REVENUE APPROACH ..................................................... i45 DATA AND MODEL SPECIFICATIONS ..................................................... 146 NET-REVENUE .............................................. ..6....... . 146 EDAPHIC VARIABLES ....................................... . 149 OTHER CONTROL VARIABLES ........................................ 49 CLIMATE VARIABLES........ .............................................................................. .. 149 ESTIMATING DISTRICT LEVEL CLIMATE AND WEATHER .................................. 150 MODEL SPECIFICATIONS ............................................. 151 ECONOMETRIC PROCEDURES ............................................. 152 EMPIRICAL ESTIMATION OF CLIMATE RESPONSE FUNCTION . ................................. 156 CONTROL VARIABLES .............................................. 156 IMPACT ESTIMATIONS .................................................... 156 YEARLY FLUCTUATIONS IN THE IMPACTS .................................................... . 157 IMPLICATIONS FOR GLOBAL CLIMATE CHANGE .................................................... 161 CLIMATE RESPONSE FUNCTION WITH YEARLY WEATHER INFLUENCE ....................................... 162 CONCLUSIONS AND POLICY GUIDELINES ............................... 162 APPENDIX A: CHARACTERISTICS OF METEOROLOGICAL STATIONS USED IN THE ANALYSES ....180 REFERENCES .............................. 183 6 TECHNOLOGY-CLIMATE INTERACTIONS: WAS THE GREEN REVOLUTION IN CLIMATE FRIENDLYZ185 INTRODUCTION ............................. 185 METHODOLOGY ............................. 185 iv THE MODEL ...................... 185 MODEL STRUCTURE ...................... 186 FUNCTIONAL FORM ...................... 187 ADAPTATION AND INTERACTION ...................... 188 DATA ...................... 189 VARIABLES ...................... 189 ECONOMETRIC ISSUES ...................... 191 ESTIMATES: TECHNOLOGY AND RELATED INFRASTRUCTURE .193 HYV - MULTIPLE CROPPING - IRRIGATION SYSTEM .193 CLIMATE CHANGE EFFECTS ON TECHNOLOGY AND INFRASTRUCTURE .194 ESTIMATES: NET REVENUE .196 CLIMATE CHANGE EFFECTS ON NET REVENUE .198 TECHNOLOGY AND INFRASTRUCTURE EFFECTS .199 SECONDARY IMPACTS ON CLIMATE AND TECHNOLOGY EFFECTS .201 CONCLUSIONS .203 REFERENCES .204 APPENDIX A: VARIABLES IN THE COMBINED INDIA AGRICULTURAL DATA SET .205 INTRODUCTION .205 COVERAGE .205 OUTPUTS .206 VARIABLE INPUTS .208 OTHER INPUTS .211 AGRO-CLIMATIC INPUTS .211 PUBLIC SECTOR INPUTS .213 SOCIOECONOMIC INPUTS .215 CLIMATE AND EDAPHIC VARIABLES .216 CLIMATE VARIABLES .217 EDAPHIC VARIABLES .219 APPENDIX B REGRESSION RESULTS .223 APPENDIX C ALTERNATE ESTIMATES OF CLIMATE EFFECTS UTILIZING SEASONAL CLIMATE VARIABLES .232 INTRODUCTION .........................................................a.......................... ................e........232 (NORMAL) TEMPERATURE RANGE VARIABLES ........................................................ 232 SEASONAL CLIMATE VARIABLES ........................................................ 234 ESTIMATES: TECHNOLOGY AND RELATED INFRASTRUCTURE .............................................. 235 HYV - MULTIPLE CROPPING - IRRIGATION SYSTEM ........................................................ 236 CLIMATE CHANGE EFFECTS ON TECHNOLOGY AND INFRASTRUCTURE ................................. 244 ESTIMATES: NET REVENUE ........................................................ 246 CLIMATE CHANGE EFFECTS ON NET REVENUE ........................................................ 253 TECHNOLOGY AND INFRASTRUCTURE EFFECTS ........................................................ 256 SECONDARY IMPACTS ON CLIMATE AND TECHNOLOGY EFFECTS ......................................... 258 APPENDIX D SEASONAL CLIMATE SPECIFICATION ........................................................ 260 SOWING AND HARVESTING SEASONS OF PRINCIPAL CROPS .................................................. 260 COMPARISON OF SEASONAL VS. MONTHLY CLIMATE SPECIFICATIONS ................................ 261 COMPARISON OF SEASONAL VS. MONTHLY SEMI-RICARDIAN MODELS ................................ 262 v FOREWORD The impact of climate change on agriculture in industrial countries has been studied by many scientists and economists. However, less is known about the economic impact of climate change on developing countries. With their warmer climates, labor-intensive low-capital practices, and alternative crop mixes, and with their less-developed market structures, developing countries are likely to respond differently to possible climate change scenarios than are industrial countries. A recent study applied several analytical frameworks, using data from India, to measure the climate responsiveness of the Indian agricultural sector. The results, reported here, indicate the potential for substantial private adaptation in developing countries. Alexander F. McCalla Director Rural Development Department vi ABSTRACT New scientific evidence made scientists more confident that greenhouse gases may lead to future climate change. Research on measuring the economic impacts climate change might cause has proceeded world-wide, but most of the empirical research has focused on developed countries. It has been commonly believed that developing countries are more vulnerable to climate change because of their reliance on low-capital agriculture. It has been assumed, but never tested, that low-capital agriculture would have more difficulty adapting to climate changes. Country-wide economic analyses have been completed only for the United States even though experts have extrapolated results to all countries. Agronomic studies of crop yield reductions support this wisdom implying large potential agricultural damages in India, for example. The vulnerability of low-capital agriculture to climate change, however, depends upon whether the affected farmers can adjust to changing climates. The recent research in the United States suggests that adaptation by private producers would reduce damages to agriculture from climate change, and carbon fertilization would actually lead to net agricultural benefits from climate change. The set of studies in this report, explores farm performance across climates in India. The goal of the study is to examine farm behavior and test if there is any evidence that farmers in developing countries, such as India, currently adjust to their local climates. The reported studies measure the climate sensitivity of low-capital agriculture. They test whether actual farm performance is as sensitive to climate as agronomic models predict assuming no adaptation. The studies also compare the climate sensitivity of low-capital farms against the results already calibrated for United States agriculture. The analyses feature the Ricardian approach, a cross sectional analysis of farm performance across different climate zones. The method uses an economic measure of farm performance: farm value or net farm income. Performance is compared across a large landscape where farms exist in different local climates. By regressing farm performance on long term climate, one can empirically measure long run climate sensitivity. Other important factors determining economic performance, such as access to markets and soil quality must also be included in the analysis. The approach carefully measures long run climate responses, and not short-run adjustments or weather effects. Although the method does not explicitly identify how farmers have adjusted, the measure of economic performance captures the consequences of all the adjustments farmers currently undertake in responding to their local climates. Each of the Ricardian studies emphasizes a different methodology--all leading to similar estimates of climate change impact. The pooled data analysis (Chapter 4) examines overall expected effects. The year by year analysis (Chapter 5) examines annual fluctuations in climate sensitivity, while controlling for annual prices and weather. The climate-technology study (Chapter 6) examines the interaction between endogenous technology and climate sensitivity in India and Brazil, respectively. The results indicate that existing farms are only mildly climate sensitive implying a substantial amount of adaptation. This adaptation is predicted to reduce potential warming damages by one-third to one-half. The analysis further suggests that the climate response vii functions for farmers in India and Brazil are similar to the estimated functions for United States farners. Low-capital agriculture appears to be no more climate sensitive than modem farms. These results suggest that warming alone will hurt agriculture in tropical (developing) countries relative to temperate countries. Damages from 8-12% are predicted by the Ricardian models. These results, however, do not include the effect of carbon fertilization. Carbon fertilization reduces the predicted damages in the agronomic models from 28% to 16%. Adding a 12% increase from carbon fertilization to the Ricardian estimates would drive the overall effects to near zero. The net results suggest that global warming will have only small effects on aggregate developing country agricultural sectors. The adaptation being measured in these Ricardian studies is largely private efforts by farmers to maximize net income given local environmental conditions. Each farmer is making different choices depending upon the conditions he/she faces. Because these subtle adaptations make farmers better off, we expect that farmers will engage in these activities as climate changes. These subtle adjustments reduce the overall sensitivity of agriculture to climate change. Technical change has been important to both India and Brazil over the years, substantially increasing productivity. However, agronomic research has not systematically focused on changing the climate sensitivity of crops. Investments in new technology have consequently not historically changed climate sensitivity in India or Brazil. This does not rule out the possibility of an important public research response to warming, it merely indicates that historic efforts have had no effect. Although aggregate agricultural sectors may not be at risk to climate change, individual farmers may still suffer large damages. Some areas will suffer from higher than average temperature changes and some areas may experience deleterious precipitation effects. The entire sector may not be affected because these effects will average out, but this does not protect local areas. Further, the aggregate sectors in developing countries may be less sensitive because important components of these sectors tend to lie in more temperate zones. Damages in marginal areas may have little impact on the aggregate because they contribute little to the aggregate outcome today. Poor people dependent on these local areas may be highly vulnerable to warming even when national agricultural impacts are minimal. viii ACKNOWLEDGMENT This report is a product of the research project "Measuring the Impact of Climate Change on Indian Agriculture" funded jointly by the World Bank Research Support Budget (RPO 680- 63) and the Electric Power Research Institute (EPRI), Palo Alto, California. Chapter 4 in this volume is based on Chapter 2 of Sanghi's 1997 University of Chicago Ph.D. Dissertation "Global Warming and Climate Change Sensitivity: Brazilian and Indian Agriculture." The work leading to Chapter 4 had benefited greatly from comments by Sherwin Rosen, George Tolley, Robert Evenson, and Participants in the Agricultural Economics Workshop and Natural Economics Seminar at the University of Chicago and Yale University, respectively. The help extended by Indian Meteorological Department in providing the climate and weather data sets used in Chapter 5, is gratefully acknowledged. The first author of Chapter 5 would like to acknowledge the extensive help received from Apurva Sanghi in handling various aspects of the work. Chapter 6 in this report is based largely on McKinsey's unpublished 1997 Yale Ph.D. Dissertation, "Climate and Technology Impacts in Indian Agriculture." Michele Rigaud prepared the first version of each chapter for publication, and Fulvia Toppin prepared the final version of each chapter, including typesetting of the entire report. Source note: Where not otherwise noted, source is author's data. ix 1 OVERVIEW Ariel Dinar and Robert Mendelsohn INTRODUCTION As scientists are more confident now that greenhouse gases will lead to future climate change there has been growing interest in understanding the economic impacts climate change might cause. Many observers are concerned that changes in climate will in turn lead to significant damages to both market and nonmarket sectors. In an effort to understand the entire picture of the effects of climate change, it is necessary to examine all sectors affected by climate change, although systems that are highly managed like agriculture may be less sensitive than systems that are managed less. Although several sectors have been studied, none have received more attention than agriculture. Research on this topic has proceeded world-wide, but most of the empirical research has focused on developed countries. Country-wide economic analyses have been completed only for the United States (Smith and Tirpak 1989 and Mendelsohn and Neumann 1998), but experts have extrapolated results to all countries (IPCC 1996b). In the United States, the initial studies suggested large negative agricultural effects in terms of crop yield reduction, loss of fertile soils, and increased cost of production (Smith and Tirpak 1989). More recent analyses, that have incorporated more up to date climate forecasts and adaptation, however, consistently find that American agriculture will be resilient to climate change (Crosson 1993; Kaiser et al. 1993; and Mendelsohn, Nordhaus, and Shaw 1994). The agricultural sectors of other developed countries in temperate climates are expected to react similarly. There have been many studies of climate change impact on agriculture in the United States and other developed countries (IPCC 1996a) but only two world-wide agricultural studies (Rosenzweig and Parry 1992 and Darwin et al. 1995). These world-wide analyses, however, have limited empirical evidence in developing countries. For example, Rosenzweig and Parry limit their inquiry to grains and Darwin et al. base their evaluation only on broad ecosystem types. However, it is not clear what effect climate change will have on agriculture in the rest of the world because, agricultural systems are different in developing countries. These agricultural systems may be less adaptable, and tropical and subtropical ecosystems may respond differently to climate change. Developing countries may be more vulnerable to climate change than developed countries because of the low-capital intensity of developing economies, the incomplete markets, the predominance of agriculture and other climate sensitive sectors, and their relatively warm baseline climates. However, empirical research in developing countries is limited so these hypotheses have yet to be tested. Further, no studies have measured what adaptation is likely to occur in developing countries. Recent research in the United States suggests that private adaptation is a critical component of climate change impacts (Mendelsohn and Neumann 1998). The absence of information about adaptation in developing countries consequently needs to be addressed. This report provides information on a series of associated analyses done on the agricultural system in India. The analyses utilize available information in the country to estimate the climate sensitivity of agriculture. Although we apply methodologies developed in the United States, careful attention is paid (see next section) to adapting these methods to developing country conditions. For example, the studies pay careful attention to technological development, family labor, and incomplete cost data. The analysis features the Ricardian approach which compares agricultural outcomes across farms under different climate conditions. It comprises background studies and Ricardian studies. Chapter 2 provides an extensive literature review that includes studies addressing also climate change impact on sectors other than agriculture. In Chapter 3 several Global Change Models (GCM) are employed to provide a range of temperature and rainfall values that may result from predicted changes in carbon dioxide levels in the future. Results from these GCM models are used later for simulation of Ricardian models' results. Each of the Ricardian studies emphasize a different methodology. The pooled data analysis (Chapter 4) examines long run response of farms to climate by Indian farming districts. The year by year analysis (Chapter 5) explains annual fluctuations in climate sensitivity by using regressions of a cross section of Indian agricultural districts for several years. This analysis measures how climate sensitivity varies from year to year in response to several variables, including prices, weather. The climate- technology study (Chapter 6) of Indian districts explores the role endogenous technology plays on farmer's climate sensitivity. The analysis examines whether technology has altered climate sensitivity and whether climate change might alter technical change.. In the next section, we briefly explain the methodology used in each chapter. In third section we summarize the overall results from India, and we conclude with some general policy observations and directions for continued research. THE ANALYTICAL FRAMEWORK The analyses in the studies rely upon the Ricardian method, an empirical approach that was developed by Mendelsohn et al. (1994). The Ricardian model examines a cross section of farms (in the case of India the unit analysis is a district) across the studied country. India is a country large enough so that farms face a variety of climates. By examining the economic performance of farms across different climates, one can estimate climate sensitivity. Economic performance is measured, in the different studies, through annual net revenues. These economic welfare measures include expected effects such as differences in crop productivity, but they also include less obvious effects such as impacts on costs of fertilizer, pesticides, and operations. The Ricardian analysis is a natural experiment, an experiment which occurs in nature and is not controlled by the researcher. One of the drawbacks of a natural experiment is that uncontrolled factors can bias the results. Bias will occur if the uncontrolled factor (such as land quality) is correlated with the variables of interest (in this case climate), affects net revenues, and is omitted from the analysis. In the Ricardian model, it is therefore important to try to measure and statistically control for variables which might affect farm value or net revenue and be 2 correlated with climate. The analyses consequently include measures of soils, market access, solar radiation, technology, household labor, and capital. However, in all cases, these measures are not perfect so that some component of these variables may still be affecting the results. This is the primary weakness of the Ricardian method and paradoxically the strength of the production function approach. The production function models are largely based on controlled experiments done in laboratory and field settings so they are not subject to these same problems. The most important advantage of this empirical cross sectional approach is that the measurements include private adaptation. Private adaptation entails changes that farmers make to adjust their operations to the environment they are in. Some of these adaptations increase productivity and some reduce costs. The issue this study addresses is whether there is evidence of adaptation of Indian farmers to changes in climate. If they do, the expectation is that they would change behavior in response to climate change. A valid criticism of the Ricardian approach is that it has historically assumed no price effects. Past studies have assumed that prices will not be affected by any change in the exogenous variables, namely climate. With the US studies (e.g., Mendelsohn et al., 1994), the Ricardian analyses were largely limited to a single time period so that prices were virtually identical across the sample. By assuming zero price effects, the Ricardian models tended to underestimate damages and overestimate benefits (Cline 1996 and Mendelsohn and Nordhaus 1996). However, this bias was calculated to be small in most relevant examples of climate change (Mendelsohn and Nordhaus 1996). In the multi-year India study (Chapter 5), a repeated cross-section of districts is utilized which permits exploration of the role output prices play. The results suggest that prices do not explain much of the intertemporal variation in net revenues and their omission does not appear to significantly bias the climate coefficients. Among the methodological and empirical difficulties addressed by the studies in this report we should mention several which have some more general implications: 1. Input prices are difficult to measure. Specifically, a great deal of the labor in developing country farms, such as in India, is provided by family members who are not paid competitive wages. We do not have a good measures of the amount of time the family members devote to farming. In order to control for household labor, dummy variables were included which identify farms which rely on household labor. Unfortunately, the farms which rely most heavily on family labor are also likely to be smaller, use more labor intensive technologies, and consume some or all of their production. It is consequently difficult to interpret the dummy variable. 2. Animal work is poorly priced. Although we have official prices for bullocks in India, these prices do not reflect the cost of keeping a bullock but rather simply the price of buying one. Since some areas grow bullock feed and others do not, we suspect that the cost of keeping a bullock might vary across India. Again, we proxy for the cost of bullocks by treating them as a fixed input and introducing bullocks per hectare as an independent variable. Many farms are subsistence. Not only do these farms depend solely on farm labor, but they are largely the sole consumers of their own output. Subsistence farms thus face different input prices (depending on family size and wealth) and different output prices (depending on personal 3 consumption and market access). The data from this study focuses on purchased inputs and sold outputs. We consequently believe the analysis captures only the market farm sector and does not represent subsistence farms. SUMMARY OF THE STUDIES' RESULTS Reviewing, in Chapter 2, a wide range of the existing studies on agricultural impacts of climate change, reveals a number of useful insights, some of which are also reported in the studies in Chapters 4-6. 1. The overall impacts of climate change on global agriculture, even assuming large local impacts, is expected to be small when trade is incorporated. 2. Carbon fertilization could offset the harmful impacts of climate change so that yields may be only marginally affected. 3. Adaptation is likely to mitigate some harmful effects so that with carbon fertilization, yields are likely to increase at least in developed countries. 4. Less is known about the ability of developing countries to adapt to climate change so certain climate scenarios may still cause regional disasters even if global production is not affected. 5. Only the major grains, which favor cool temperate zones, have been extensively studied so the effects of climate change on the remainder of agriculture types remains uncertain. Tlhe purpose of the study reported in Chapter 2 is to develop climate scenarios - based on the projections of several GCMs - to be used as input to an analysis of the impacts of potential climate warming on agriculture in India. The study uses the projections from three GCMs to develop projections of temperature and precipitation in India under a scenarion of doubling of CO2 from pre-industrial levels. Three models used, the Geophysical Fluid Dynamics Laboratory (GFDL), the United Kingdom Meteorological Office (UKMO), and the Goddard Institute for Space Studies (GISS) models, as a basis for assessing the impacts of climate warming on the region. The information produced by the GCMs indicates that the continued emission of trace gases into the earth's atmosphere will likely result in increases in both temperature and precipitation for India. While there will be significant spatial variation in the expected increases, data are presented for the country as a whole. Micro-scale modeling of climate systems is not advanced enough to make reasonable projections at a local scale, and the general projections must suffice. Solar radiation and evapo-transpiration likely will not change appreciably (or, at least, the models are inconsistent in their projections of these variables). Changes in soil moisture are unknown, since it depends on other factors besides the ones projected by the GCMs, including runoff, soil depth and percolation. The three Ricardian studies of India (Chapters 4, 5, and 6) produce consistent results of climate change impact on Indian agriculture. All three studies find Indian agriculture sensitive to warming. Specifically, the studies find that net revenues fall precipitously with warmer April's but also are sensitive to warmer January and July temperatures. Crop revenues increase with 4 October temperatures. Net revenues are also sensitive to precipitation, but the effects are smaller and offsetting. Wetter January's increase farm values and wetter April's reduce farm values. July and October effects are small. Because the effects across seasons are small and offsetting, changes in annual precipitation have little effect. The pooled analysis (Chapter 4) suggests that climate change will have an overall negative impact on Indian agriculture. A warming scenario of +2.0°C rise in mean temperature and a +7% increase in mean precipitation levels will create a 12% reduction in net revenues for the country as a whole. Rising temperature is damaging and increasing precipitation is beneficial. These effects will vary by season and region. There are regional impacts from warming even within India. Coastal and inland regions of Gujurat, Maharashtra, and Karnataka are most negatively affected. The high-value agricultural regions of Punjab, Haryana, and Western Uttar Pradesh show a small loss. The agriculturally low-value, hot and dry districts of Rajasthan and Central India are negatively impacted. Districts in many Eastern states (Andhra Pradesh, Orissa and West Bengal), however, benefit mildly from warming. These regional outcomes are largely caused by initial climate differences between regions. The repeated annual analysis (Chapter 5) measures a lower climate sensitivity than the results in Chapter 4 due to a different data set used. A warming scenario of +2.0°C rise in mean temperature and a +7% increase in mean precipitation levels will create an 8% reduction in net revenues for the country as a whole. The repeated analysis reveals also that estimated climate sensitivity varies from year-to-year. For example, annual marginal effect of temperature alone varies between -150 and +280 Rs/ha, while inclusion of weather reduces the variance to values between -100 and +100 Rs/ha. Although the average effects reported above continue to hold in most years, there are exceptions when warmer January and July temperatures appear to be beneficial. Combining effects across seasons, there are four years between 1970 and 1986, where warming appears beneficial (1974, 1976, 1978, and 1984). Neither annual weather nor annual prices can explain all of this intertemporal variation. The climate-technology analysis (Chapter 6) introduces endogenous technical change into the model. Technical change was measured by three variables, namely, intensity of modern high yielding varieties, intensity of multiple cropping, and irrigation intensity. It was found that technology and climate interact to affect net revenue in agriculture in India. Climate affects technical change: warmer areas generally have less irrigation and modern varieties but a little more multiple cropping. Wetter areas have less irrigation, modern varieties, and multiple cropping. These results are consistent with the general observation that the most significant technological improvements have come in areas which are more temperate. However, the overall effect is small so that warming is not expected to have a substantial impact on modernization. A simulation of a combined warming scenario of +2.0°C rise in mean temperature and a +7% increase in mean precipitation levels will create a 35% reduction in net revenues for the country as a whole. Also examined in Chapter 6 is the question whether technical change has altered climate sensitivity. It was found that higher levels of technology can help reduce sensitivity to warming but may increase damages from increased rainfall. However, the magnitude of these effects is small, so that technological change has not really affected the climate sensitivity of agriculture in India. 5 CONCLUSION Ricardian models were estimated for India and Brazil in order to determine the climate sensitivity of agriculture in both countries. The results of our Ricardian investigation of the climate sensitivity of Indian agriculture confirms that agriculture in both countries is sensitive to warmer temperatures. However, the analyses suggest that the climate response functions are not very different from the estimated function for the United States. The slightly more harmful effects found in India and Brazil are due to the warmer baseline conditions in these more tropical countries. The Ricardian model, which captures farmer adaptation, predicted much smaller damages to agriculture, compared with other approaches reported in the literature. The results suggest that farmer adaptation will mitigate from 40% to 60% of the potential damages from warming. In addition to farmer adaptation, there is also a possible research response to warming. However, the study of technical change indicates only a small interaction between climate and technical change in the past. These results suggest that overall, warming will hurt agriculture in India relative to temperate countries. However, with the mild climate scenarios predicted for the next century, carbon fertilization, and private adaptation, these effects are likely to be small. One important policy implication that emerges from our analysis, given the important role of private adaptation, is that governments should encourage private adaptation. Private adaptation is expected to be efficient and imposes no burden on the public budget. Measures may include development and dissemination of new technologies and practices. Although the analyses reported here provide some important initial insight into the climate sensitivity of a developing country economy, additional analyses are needed. For example, little is known about subsistence farming and what will happen to the poor families dependent on local climate conditions. Future biological research would probably have to be focused more specifically on warming for it to affect climate sensitivity. Agricultural studies should also be conducted in other regions of the world which have not yet been studied. REFERENCES Cline, W. 1996. "The Impact of Global Warming on Agriculture: Comment." American Economic Review 86: 1309-1312. Crosson, P. 1993. "Impacts of Climate Change on the Agriculture and Economy of the Missouri, Iowa, Nebraska and Kansas (MINK) Region." In Kaiser, H., and Drennen, T., eds., Agricultural Dimensions of Global Climate Change. Delroy Beach, Fla: St. Lucie Press. Darwin, R., M. Tigras, J. Lewandrowski, and A. Raneses. 1995. "World Agriculture and Climate Change: Economic Adaptations". U.S. Dept. of Agriculture, Economic Research Service, Publication # AER-703, Washington, D.C. IPCC (Intergovernmental Panel on Climate Change). 1996a. Watson, R., M. Zinyowera, R. Moss, and D. Dokken, eds., Climate Change 1995: Impacts, Adaptations, and Mitigation 6 of Climate Change: Scientific-Technical Analyses. Cambridge: Cambridge University Press. . 1996b. Bruce, J., H. Lee, and E. Haites, eds., Climate Change 1995: Economic and Social Dimensions of Climate Change. Cambridge: Cambridge University Press. Kaiser, H. M., Riha, S. J., Wilks, D. S., and Sampath, R. 1993. "Adaptation to Global Climate Change at the Farm Level." In Kaiser, H. M., and Drennen, T. E., eds., Agricultural Dimensions of Global Climate Change. Delray Beach, Fla: St. Lucie Press. Mendelsohn, R. and W. Nordhaus. 1996. "The Impact of Global Warrning on Agriculture: Reply." American Economic Review 86: 1312-1315. Mendelsohn, R. and Neumann, J. eds., 1998. The Economic Impact of Climate Change on the Economy of the United States. Cambridge: Cambridge University Press. Mendelsohn, R., W. Nordhaus, and D. Shaw. 1994. "The Impact of Global Warming on Agriculture: A Ricardian Analysis." American Economic Review 84: 753-771. Rosenzweig, C.and M. Parry. 1992. Climate Change and World Food Supply. Washington, D.C.: USEPA. Smith, J. and D. Tirpak. 1989. The Potential Effects of Global Climate Change on the United States: Report to Congress. EPA-230-05-89-050. U.S. Environmental Protection Agency, Washington D.C. 7 2 MEASURING THE IMPACT OF AND THE ADAPTATION TO CLIMATE CHANGE IN AGRICULTURE AND OTHER SECTORS - LITERATURE REVIEW Ariel Dinar and Heather Beach INTRODUCTION CLIMATE CHANGE In recent years, there has been growing interest in curbing the rapid rise of greenhouse gases in the atmosphere in order to control future climate change. Many observers are concerned that changes in climate will in turn lead to significant damages to both market and nonmarket sectors. In an effort to understand the entire picture of the effects of climate change, it is neces- sary to examine alLsectors affected by climate change. One such area examined is the social ef- fect of climate change focusing on several potentially affected sectors, among which are forestry and ecosystems, coastal zones, agriculture, fisheries water resources, and energy developments (Reilly and Thomas 1993). Toman, Firor, and Darmstadter suggest that while "the potential im- pacts of climate change are broad, some aspects of human society are more sensitive than others" (1996, 11). Further suggesting that systems that are highly managed like agriculture may be less sensitive than systems that are managed less. Although several sectors have been studied, none have received more attention than agri- culture. In the United States, the initial studies suggested large negative agricultural effects in terms of crop yield reduction, loss of fertile soils, and increased cost of production (Smith and Tirpak 1989). More recent analyses, that have incorporated more up to date climate forecasts and adaptation, however, consistently find that American agriculture will be resilient to climate change (Crosson 1993, Kaiser et al. 1993, and Mendelsohn, Nordhaus, and Shaw 1994). The ag- ricultural sectors of other developed countries in temperate climates are expected to react simi- larly. However, it is not clear what effect climate change will have on agriculture in the rest of the world because, agricultural systems are different in developing countries, these agricultural systems may be less adaptable, and tropical and subtropical ecosystems may respond differently to climate change. There has been considerable debate whether or not the steadily increasing levels of car- bon dioxide observed in the atmosphere will lead to climate change (Nierenberg 1995). Most atmospheric scientists concur that greenhouse gases will affect climate by raising temperatures and changing water cycles. The debate, at this point, is focused upon the magnitude of this change. The consensus of atmospheric scientists is presented in the Intergovernmental Panel on Climate Change (IPCC) report (1990) which predicts that world temperatures will increase by 1.5-4.5 degrees C by 2060 (from the doubling of greenhouse gases since 1880). The best guess of this group is a 2.5 degree C increase. An updated assessment, provided in the second IPCC report (Brack and Grubb 1996) suggests an increase in global mean surface air temperature of about 2 degree C by the year 2100 with a range of uncertainty of 1-3.5 degree C. These higher world temperatures will increase the hydrological cycle activity leading to a general increase in precipitation and evapotranspiration. What scientists have not been able to predict thus far is how this world wide change will manifest itself across the planet. There are no consensus predictions of temperature and precipitation changes at the regional or local level, al- though there are many individual models with detailed predictions (Barron 1995). The climate change any one area might receive as a result of global change could be quite different from the global average. In addition, policy implications for climate change must be viewed separately from other types of environmental pollution since the effects of global warming is an irreversible phenomenon (Mathews 1991). STUDIES Impact studies should focus on measuring climate response functions for specific areas rather than focusing on limited climate scenarios because the actual climate change that any area is likely to experience is not known at this time. For example, agricultural studies could be more effective if they measure how farm outputs and net revenues are affected by a range of tempera- ture and precipitation changes rather than a limited set of climate change scenarios. Agricultural studies must also consider the direct effect of carbon dioxide since it is clear that CO2 has a posi- tive influence on crop yields (Kimball 1983) and water efficiency (Woodward 1993). The mag- nitude of and circumstances in which this carbon fertilization effect takes place is debated (Bazzaz and Fajer 1992; Van de Geijn et al. 1993; Kimball et al. 1993; and Mooney and Koch 1994), but the importance of including this effect is unequivocal. Although a great deal has been learned about the link between energy systems, emissions, ambient greenhouse gas concentrations, and climate change, relatively little analysis has focused on what would happen if climate changed. Global warming may affect agriculture by directly altering yields, changing water availability, or affecting soils. Most empirical work undertaken has focused on the United States. This is especially true for economic analyses of the impact of climate change on agriculture. Existing studies include: USEPA (Adams et al. 1989, 1990, 1993; Rosenzweig and Parry 1993); USDA (Kaiser and Drennan 1993; Kaiser et al. 1993; Men- delsohn, Nordhaus, and Shaw 1994; Shaw et al. 1995); and DOE (Crosson 1993). The earlier studies in this group consistently predicted that there would be large negative impacts from cli- mate change in the form of reduced quantities qualities of yields of grain crops (See also Scim- melpfennig et al. 1996). In contrast, the more recent studies consistently predict that US agri- cultural systems will readily adapt to climate change, by introducing new technologies, new crop varieties, and cultivation practices, so that there will be minimal changes in yields and net prof- its. These results likely extend to other Organization for Economic Cooperation and Develop- ment (OECD) countries suggesting that agriculture in developed countries is not sensitive to cli- mate change. Less is known, however, about what effect climate change will have on the agricultural systems of developing countries. Kaiser and Crosson (1995) suggest that to be successful, the international context in which agriculture operates must be taken into account. Several studies have addressed the impact of climate change on the food supply and risk of hunger in developing countries, especially Africa (Parry 1992; Downing and Parry 1993; Rosenzweig and Parry 1993; and Parry and Rosenzweig 1994), and Egypt (Strzepek et al. 1994). A recent study by Ro- 10 senzweig and Iglesias 1994 has addressed responses of grain crops (maize, wheat, soybean, and rice) to climate change in various countries, including India and Brazil. Adaptation in these studies has been treated mainly by changing values of parameters on an ad-hoc basis. These studies suggest that if high amounts of adaptation are assumed, developing countries could adapt to climate change. The only problem would be isolated areas of subsistence farmers who could remain vulnerable to severe local climate effects. Although levels of adaptation are explored in these developing country studies, these studies do not model what adaptations are likely to occur. No attempt has been made to quantify how farmers have actually adapted to climates in devel- oping countries. Without solid adaptation estimates, the existing literature has probably overes- timated the impacts of climate change in developing countries. Further, like the early United States studies, the studies listed above include only grains, which capture only roughly half of the agricultural land in many developing countries (in particular 59% and 31% of the arable land in India and Brazil in 1993, respectively (FAO 1994)). A comprehensive empirical analysis of the impacts of climate change on other important crops in developing countries has yet to be accom- plished. In addition to these empirical studies that have attempted to predict the effect of climate scenarios on yields, there have been a handful of studies which have focused on the relevance of international trade. Reilly et al. (1994), and Kane, Reilly, and Tobey (1991) assume yield reduc- tion in different parts of the world and then predict what would happen to world agricultural prices. These studies, if combined with solid estimates of impacts in each region, could eventu- ally forecast what would happen to world agricultural prices. However, at the time these studies were completed, impacts in different regions were unknown, thus limiting the power of these in- ternational studies. Coupled with new estimates of yield changes, however, an updated version of the Reilly models could address concerns about price impacts that cannot be properly handled at the regional level. One way climate change may impact agriculture is by reducing the available supply of water for irrigation. Through either reduction in precipitation or increase in evapotranspiration available water for irrigation may be reduced. This may be especially important in developing countries because agriculture utilizes as much as 80% of water resources (WRI 1992; Xie et al. 1993). Water availability is a key factor in agricultural productivity, especially with developing countries located in arid and semi-arid regions. Further, developing countries may not be able to afford extensive manipulations of water systems so they may face limited options to adapt to water shortages (Frederiksen 1992). For example, a recent study by Fulglestvedt et al. (1994) provides a tally of country case studies on climate change, includes few studies on impact esti- mates, and even fewer studies on policy reaction in developing countries. Reviewing the full suite of studies on agricultural impacts of climate change reveals a number of useful insights: (1) The overall impacts of climate change on global agriculture, even assuming large local impacts, is expected to be small when trade is incorporated. (2) Carbon fertilization could offset the harmful impacts of climate change so that yields may be only marginally affected. 11 (3) Adaptation is likely to mitigate some harmful effects so that with carbon fertiliza- tion, yields are likely to increase at least in developed countries. (4) Less is known about the ability of developing countries to adapt to climate change so certain climate scenarios may still cause regional disasters even if global pro- duction is not affected. (5) Only the major grains, which favor cool temperate zones, have been extensively studied so the effects of climate change on the remainder of agriculture types re- mains uncertain. MEASURING CLIMATE CHANGE (MODELS) Agricultural impact studies must draw a link between climate change and farming be- cause of the direct relationship between climate and agricultural production processes. The tra- ditional approach has been to turn to a General Circulation Model (GCM) for a forecast of the climate change associated with a doubling of greenhouse gases. Each GCM model produces a specific forecast which can then be used to predict yield and farm revenue changes. This climate scenario approach is very limiting. First, there is tremendous variation across models as to what will happen in each location. Although the analyst can proceed by choosing specific forecasts, it is impossible to capture the range of impacts from a limited set of model runs. Second, the cli- mates associated with a doubling of greenhouse gases represent one distant moment in time (around 2060). Between now and then, and continuing after, climates will gradually be changing and so the earth will experience a full range of different climates. Rather then be limited to a handful of specific scenarios, impact studies should develop response functions. As climate sce- narios change, these response functions can be used to predict what will happen at each location and time period. In contrast, the climate scenario approach becomes outdated the moment one moves away from the specific changes being analyzed. There are two main approaches to measuring climate impacts on agriculture in the lit- erature: an agronomic production function approach and a Ricardian model. The agronomic production function approach begins with a crop simulation model and predicts changes in yield in response to climate. The yield changes are then either extrapolated to an aggregate effect as with Rosenzweig and Parry (1993), or they are introduced into an economic model as in Adams et al. (1989, 1990, 1993) or Crosson (1993). The economic models in turn estimate aggregate damages to the agricultural sector. An alternative approach is to empirically estimate the direct impact of climate on agricultural net revenues, using the Ricardian model (Mendelsohn et al. 1994; and Shaw et al. 1995). Both approaches have strengths and weaknesses and tend to com- plement each other. The agronomic production function approach has the strength of being tied closely to carefully controlled experiments where specific climate or C02 levels are varied holding all other variables constant. This eliminates one of the potential problems with the Ricardian method, that climate variables may be correlated with other omitted variables resulting in biased estimates. In order to handle the agronomic production function approach properly, farmer ad- aptations should be included in the modeling. Simulations should be run with a variety of differ- ent farm methods such as varying planting times, crop varieties, harvests dates, and tilling and ir- rigation methods. The researcher would then be able to determine which activity would maxi- 12 mize profit and then, could trace out actual yields and net revenues for different climates. In practice, this is too expensive, and studies using this methodology either do not incorporate ad- aptation at all, or at best explore a limited number of alternative farming methods. One of the limitations of the agronomic production function approach, therefore, is that it fails to properly account for adaptation and therefore usually overestimates negative impacts. Another distinction that is often made between models is whether or not they include price effects. The Ricardian model explicitly assumes away price changes. In contrast, some of the economic models using the agronomic production function approach have emphasized price changes as part of their results (Adams et al. 1989, 1990, 1993). Given that agricultural markets are worldwide, this use of domestic studies to predict prices is questionable. Agricultural prices can only be reliably predicted using global models. However, predictions of global yields are generally quite poor in quality because there are not sufficient measurements of both climate and yields at all sites. Attempts such as Rosenzweig and Parry (1994), Darwin et al. (1994), and Reilly et al. (1994) represent best efforts to date to measure global prices. However, each of these studies suffers from inaccurate measurements of climate-induced supply changes, espe- cially in developing countries. For example, Darwin (1994) and Darwin et al. (1994) use a global model that includes 8 world regions (US, Canada, European Community, Japan, China and several other east Asia Countries, some Southeast Asia countries, Australia and New Zea- land, and the rest of the world). These studies use a Computerized General Equilibrium (CGE) model that aggregates information on land and climatic resource changes (based on a Geographic Information System (GIS)) and changes in climate that are predicted by GCMs. Although com- prehensive, the CGE model requires detailed knowledge of land and agricultural uses which can- not be accurately measured. Further, the authors have no way of reliably predicting how these large land areas would actually react to climate change. By providing sound measurements of impacts in developing countries, our results should complement these global models and help them generate more reliable global estimates. The agronomic production function approach begins with the basic relationship between climate and crop production. Through agronomic experiments, agronomists have calibrated models which predict the yield of specific crops depending on weather patterns. These simula- tion models have historically been used to predict changes in yields for specific crops (Adams et al. 1989 or Rosenzweig and Parry 1993). The outcome from these simulations is then fed through an economic model of farmer behavior which in turn leads to a partial equilibrium model of the farm sector. This agronomic production function approach is attractive in its close col- laboration with agronomic science. It can also be carefully linked with hydrologic conditions. Finally, the production function approach is the only current method capable of including carbon dioxide fertilization (provided appropriate agronomic models). Innes and Kane provide a discussion on agricultural impacts of global warming. They have generated a list of variables which they feel need to be included in the complex modeling to predict the potential effects of greenhouse gas accumulation on agricultural production and pric- ing (1995, 747): (1) Effects of greenhouse gases on climate itself, including ambient temperature, pre- cipitation, ENSO cycles, and climatic events such as cyclones and hurricanes. Such effects on the climate system may result in a complicated mosaic of changes 13 in climatic conditions with a wide geographical variance and strong localized ef- fects driven primarily by precipitation patterns. (2) Physical responses of agricultural systems to greenhouse gases and climate change. Such responses include the so-called C02 - fertilization effect and po- tential changes in pest problems from warmer and more humid conditions. (3) Behavioral responses to climate change for given technologies and institutions. Such adaptations may include not only farmer shifts in cropping patterns, planting dates, tillage practices, irrigation techniques, and other management approaches, but also regional adaptation in, for example, water delivery systems. (4) Technological change, which may include both exogenous improvements in agri- cultural productivity and changes in technology induced by climatic changes. (5) Equilibrium effects in international agricultural markets. Innes and Kane indicate that no analysis, thus far, has successfully incorporated all of these variables. IMPACT OF CLIMATE CHANGE IMPACT ON NON-AGRICULTURE SECTORS To understand the broader implications and effects of climate change, it is necessary to examine non-agriculture sectors. This sections reviews the literature examining climate change impact on: ecosystems, forestry, animal production, fisheries, water resources, and energy de- velopments. Reilly and Thomas (1993) conduct a discussion of studies that examine these non- agriculture sectors. There is a general theme found in a majority of these sectors, a lack of com- prehensive models examining the individual sectors and few combinations with a climate change model. A majority of these sectors include a complex interaction of variables which have hin- dered attempts at forecasting changes, even with the exclusion of the climate change component. Both man made regional changes and global climate changes impact natural ecosystems (plants and fish populations), land use, water quality, desertification, air quality, and human health and livelihood. Reilly and Thomas (1993) indicate that the most widely examined effects of climate change upon ecosystems is done on a global or regional scale. For example, Smith (1994) esti- mates the impact regional and global climate change have on deltaic ecosystems in the Aral Sea basin. Walker (1994) provides a discussion of two ecosystem models that are now being in- cluded as part of GCMs in a first attempt to combine atmosphere-biosphere models. Studies on ecosystem change are impeded by lack of integration of the various ecological response models, notwithstanding, some attempts to incorporate the nutrient cycling component of ecosystem models into the gap-phase models in forests have been made (Pastor and Post 1988). The FAG (1991) reports that based on the current predictions, global climate change on forest resources are negative with few positive effects. Climate change might cause forest species to either occupy a greater land area or it could also force a species to occupy smaller areas suitable for its growth (FAG 1991). Perhaps the main effect suggested by the FAG would be the restriction of natural ranges, reduced genetic variability and the possible extinction of some species. 14 The literature suggests that the effects of climate change towards animal production will follow the general trend of unequal distribution of changes. An FAO report (1 991) indicates that there will be both positive and negative impacts. The example used is that higher temperatures would reduce the need for housing livestock during winter periods, would decrease maintenance requirements and increase the productivity of winter pastures, restricting the need for feed con- centrates. However, higher temperatures would produce negative impacts in areas where climate change involves reduction in rainfall or increased evapotranspiration. Adverse affects to rainfed forage and fodder production, and availability of crop residues would be caused by lower rain- fall. In addition, perhaps the greatest impact would be on pastoral families who would migrate to arable areas to secure their livelihood (FAO 1991). The migration of human population due to climate change is another non-agricultural sector that is difficult to anticipate without being able to predict how sectors might adapt to climate change (Lonergan 1994). The generalization of the possible impacts of climate change upon fisheries is limited by the complexity of the variables upon which fisheries are influenced: physical, chemical, and biological processes (Reilly and Thomas 1993). Climate change has the potential to effect any and all of these variables. It is reported by the FAO (1991) that as a whole, global marine fish production will not be severely impacted by climate change, with areas of high productivity pos- sibly shifting polewards. However, individual fish species may suffer and with the variabilities of ocean currents, planning and management is likely to cause problems for individual countries (FAO 1991). In addition, Reilly and Thomas (1993) point out that the ability to make predic- tions is further limited because there is no combination of a comprehensive ocean model and climate change models. It is most probable that inland water fisheries will be affected by droughts and floods similarly to land resources (FAO 1991). Studies have shown fresh water resources to be very sensitive to climate changes (Gleick 1989; FAO 1991; Lonergan 1991; Lonergan and Kavanagh 1991). Like the other non- agriculture sectors, water resources are difficult to do future change models on because of the complexities of global and regional climate systems. Gleick (1986) suggests that the foremost limitations to using GCMs to assess changes in water availability are their coarse resolution and their simplified hydrologic parameterizations. These limitations have created an increase of re- search to develop other methods for evaluating the water resource effects caused by clirnate change (Gleick 1986; Nemec and Schaake; 1982; EPA 1984). FAO reports simulations have suggested that "a 25 percent rainfall decrease and a 5 percent evapotranspiration increase, could reduce the irrigated area by 75 percent" (1991, 8). The FAO further reports that there could be a significant alteration of the balance between water supplies and water needs in major irrigation areas of arid zones due to climate change, many of these areas are located in developing coun- tries. Energy development h,as been examined in the context of how urbanization and energy use effects global warming. Parikh and Shukla (1995) conducted a cross-national study of de- veloping countries and the impact of energy use and climate change in economic development. Their conclusions can be summarized as: a positive correlation between greenhouse emissions and countries' urbanization levels; aggregate energy use rises with urbanization; and disaggre- gate energy use suggest that the sectoral and fuel use shifts accompanying urbanization have greenhouse augmenting potential. Parikh and Shukla determined that cities of developing coun- 15 tries are facing more serious consequences as a result of local rather than global pollution. Also, developing countries are likely to encounter different energy efficient technological alternatives in the future than developed countries (Parikh and Shukla 1995). In addition to the study of energy use in developing countries, Lonergan and Young (1989) conducted an assessment of the effects of climate warming on energy developments in an arctic climate, Canada's North. The study concluded that: GCMs provide estimates that are not suitable for assessing local impacts because they do not capture the variability of climate; the disappearance of permafrost do to increased temperatures will impact existing and planned pipe- lines and buildings; and in the area being studied, there will be an extreme increase of precipita- tion which will be of concern in terms of future design and expected impacts (Lonergan and Young 1989). Blitzer et al. (1991) suggest that the possibilities of greenhouse gases emission reductions should be discussed using country-level models with attention to structural detail, and conducted a study modeling Egypt. Parikh and Gokarn (1993) conducted a study on climate change and India's energy policies. Their paper presents an analysis of C02 emissions in the Indian econ- omy and addresses the implications of alternative policies to reduce them. This study is unique in that the analysis is based on flows of energy in the economy of India using a 60 sector input- output model instead of the standard of examining energy supply structure and end-uses of energy. IMPACT ON AGRICULTURE Analysts have already determined that climate change will affect different areas to differ- ent extents. The impacts are dependent on the enormity and distribution changes of the weather variables, including current climate and environmental characteristics and the agricultural struc- ture of the different areas (Bacsi et al. 1991). Decker and Achutuni (1990) suggest that while climate change will alter the way agricultural enterprises are managed, one should not conclude that the climate change will have negative impacts on agriculture and cause a decrease in levels of productivity. However, the importance of examining climate change from a long term per- spective is the uncertainty it introduces. There have been many studies conducted examining the impact of climate change on: US Agriculture (Kaiser et al. 1995; Kaiser and Crosson 1995; Mendelsohn 1996) World Agriculture (Kane et al. 1992; Karim et al 1994; Reilly 1995; Sonka 1992) Brazil (Mendelsohn 1996; Sanghi et al. 1997) China (Jin et al. 1994) Egypt (Strzepek et al. 1994) India (Dadhwal 1989; Mendel- sohn 1996; Rao and Sinha 1994) Pakistan (Qureshi and Iglesias 1994) Southern Africa (Schulze et al. 1993) Southeast Asia (Parry et al. 1992; ) Sri Lanka (Vidanage and Abeygunawardena 1994) to name a few. Kaiser et al. (1995) examined potential agronomic effects of several climate change sce- narios on grain (corn, wheat, and soybeans) farming in the US. The agronomic results indicated that a mild climate change scenario had a minimal negative effect on all crop yields. Corn and soybean yields were negatively affected by the more severe (hotter) climate-change scenario. Southern states were more negatively affected than northern states. Economic results suggest that crop prices are quite sensitive to the rate and form of the assumed climate-change scenario. 16 Under all climate change scenarios, corn and wheat area and production decline over the 100 year analysis, and soybean area and production increased over the years. Sanghi et al. (1997) estimate the impact of climate change on agriculture in Brazil using a Ricardian Approach. Impact was estimated using agricultural yields, and land values. Although there are varying regional consequences, the net impact is negative. The Center-West region is most negatively impacted and the South region benefits mildly from warming. Dadhwal (1989) summarizes the results of a study on the effect of temperature on wheat in India. The results of this study follow a simple approach to quantify the effect of temperature on wheat in the field by sowing the crop on different dates, testing across the large seasonal changes in temperature for the different thermal regimes found in India. Jin et al. (1994) examined the effects of climate change on rice production in southern China. Some of their results found that the rice yields, effected by climate change alone, de- creased as a result of higher temperature, which in turn shortened the growing cycle of rice and caused water stress in some regions. However, when the rice yields were examined by climate change with physiological C02 effects, the rainfed rice yields increased in the northern and east- ern'sites in the study. Under all of the doubled C02 scenarios, the temperature increased and' could result in an extension of the growing season for rice in Southern China. However, the in- creased temperature had a negative impact on yields and shortened the simulated lifecycle of rice. Antle (1 995) bases his research on the generally agreed upon theory that the poorest countries' agriculture's are likely to be the most vulnerable to, and least capable of adapting to, climate change or other environmental disruptions. Antle discusses the climate change impacts on tropical agriculture. He suggests that increased atmospheric C02 can raise plant productivity through C02 fertilization, and can cause changes in temperature, rainfall, solar radiation, and wind patterns that can impact either positively or negatively on plant and animal productivity. Mendelsohn (1 996) suggests that the impact of climate change research now indicates that there may be small benefits in OECD countries and only mild damnages in developing countries. FOOD SECURITY Global food supplies have been affected for years by climate changes effecting agricul- ture. Areas suffering from drought are unable to produce crops. Thus far, studies examining climate changes on agriculture suggest that since the effects will be spread globally, that even though some areas will have significant reduction in agriculture capabilities, other areas may be able to balance these negative impacts with increased production. The concept of food security has developed from these studies which examine climate change and agriculture and more re- cently, climate change and non-agricultural sectors. A few studies have examined the implica- tions for global food security (Gleick 1989; Oram 1989; Peterson; Parry 1990; Pimentel 1991; Downing 1993; Rosenzweig et al. 1993; Rosenzweig and Parry 1993). Pimentel's work is based on global warming in conjunction with rapid population growth with the expectation that together they will have negative impacts on natural resources and .food production (1991). Pimentel suggests several main changes in the agricultural ecosystem which will have a major impact on food production: temperature rise, rainfall changes, and pest attack. 17 However, Pimentel suggests that the negative effects of global warming on agriculture can be offset to a certain degree with sound agricultural adaptation and management techniques, thus further limiting the impact on the world's food production capabilities. Parry (1990) devotes an entire chapter of his book Climate Change and World Agricul- ture on the subject, "Implications for Global Food Security." Parry emphasizes that the infor- mation currently available is extremely limited and the analyses do not take into account changes in technology and management that would alter any potential effects from climate change. In addition, the studies presented in this chapter focus only on a single scenario of climate change and a single response of yield to that change. Parry cites the working group on food security from the 1988 Toronto Conference on The Changing Atmosphere to indicate the need for addi- tional modeling and examination of the global food sector: "While averaged global food sup- plies may not be seriously threatened, unless appropriate action is taken to anticipate climate change and adapt to it, serious regional and year-to-year food shortages may result, with particu- lar impact on vulnerable groups" (1990, 105). Parry's conclusions, based on preliminary results of the limited scenarios, suggest that while there may be only minor interruption of global food supplies, the increase in food prices could seriously influence the ability of food-deficit countries to pay for importing. Rosenzweig et al. (1993) conduct an extensive review of climate change and its impact on the World's food supply. The aim of this study was to provide a global assessment of the potential effects of climate change on crop yields, world food supply, and regions vulnerable to food deficits. It is noted by the authors that it is critical to conduct research in determining how countries, with particular attention to developing countries, can and will respond to reduced yields and increased costs of food. The concept of food security is based on direct and indirect effects of climate change to the agricultural system, which includes not only changes to crop production, but also the hydro- logic cycle. Changes to animal production and fisheries will also impact the diet of the world's population. The area of non-agriculture sectors effected by climate change needs additional at- tention and focus by scientists to produce the same level of results being accomplished with the future agricultural assessments. RESPONSES AND ADAPTATIONS To CLIMATE CHANGE RESPONSES AND ADAPTATIONS IN NON-AGRICULTURE SECTORS There have been few studies completed examining what the responses and adaptations in non-agriculture sectors have been previously and should be in the future. Three studies that ad- dress appropriate responses and adaptations on the subject are FAO (1991, Parry (1990) and Downing (1993). FAO (1991) reports that the greatest risk appears to be in semi-arid areas that are already vulnerable to drought, or in low-lying coastal areas and deltas of developing countries. The FAO suggests that responses to climatic changes in these vulnerable areas would involve inodifica- tions in cropping patterns, forestry, livestock, and fishery production. In addition, out-migration from affected coastal and pastoral areas would be expected. 18 Parry (1990) found that substantial increases in the need for and costs of irrigation are likely to occur in order to substitute for moisture losses due to increased evapotranspiration. Ad- ditional costs are likely to be needed to control the spread of subtropical weed species into cur- rent major cereal-producing regions and increase of pests in the warmer, more humid climates. Downing (1993) indicates that the ultimate effect of climate change will depend to a great extent on adaptive responses at several scales. Downing conducted a simple model of household food security in eastern Kenya and found that the potential of adaptive responses in the sub- humid zone marginally suitable for maize. Downing suggests that there are two main ways to increase household food security, increase yields by applying fertilizer and adopt soil conserva- tion measures. An additional strategy would be to increase the length of food storage from the current average of 3 months of consumption to a target of 6 months. RESPONSES AND ADAPTATIONS IN AGRICULTURE Many studies have been completed examining the responses and adaptations in agricul- ture previously and indicating future responses. These analyses range in suggestions from shift- ing the actual land use where crops are grown to genetically changing and breeding crop species for resistance. A few of these studies are discussed below. Decker and Achutuni provide a list of five research areas which, through their research, appear to have the highest priority in need to respond to sustain agriculture under conditions of global warming (1990, 436-437): (1) The design of simulation models for crop response which accounts for both cli- matic change and the direct effects of increase C02 concentration. (2) The use of genetic engineering to develop stress-resistant cultivars. (3) Development of econometric models which evaluate the impacts of climate change on national farm policy alternatives and the development of foreign markets. (4) Development of soil water models for investigating the feasibility for the move- ment of cropping zones. (5) Simulation studies using climatic scenarios for determining the need for and prof- itability of irrigation. Antle (1995) argues that to conduct analysis of agricultural adaptation, particularly in de- veloping countries or regions of the world, information on the rate of climate change is critical. However, estimating this dynamic is so intensive with GCMs, resolution would have to be sub- stantially reduced, diminishing the value of the analysis. Darwin et al. (1994) conducted a study linking economic activities to land resources de- termined by climate. Despite negative impacts in some regions, Darwin et al.'s results indicate that climate change will have a relatively small impact on the long-term ability of global agri- cultural resources to meet future world food demands. However, these results are dependent on the ability to shift crop production to new locations. Unfortunately, many of these shifts would have to occur across international borders, which in some regions, and between some countries 19 would be difficult to accomplish. Parry (1990) also found that there would be a need to switch crop locations to land uses that show a greater increase in productivity potential. Kaiser and Crosson (1995) believe that the conclusion of Rosenzweig et al. (1993) that agricultural production in developed countries in the temperate zone actually might be increased by climate changes resulting from an equivalent doubling of atmospheric C02 to be applicable to the United States if there are ample options for technological adaptation and an incentive to adopt the appropriate technologies. In addition, they found that maintaining and steregthening the international trading system to be crucial in easing the impacts of climate change on not only agriculture in the United States but around the world as well. Rao et al. (1989) found that in India, technological innovations involving genetic alteina- tions in conjunction with input management practices and a better understanding of the process of adaptation has resulted in moderate advances in the productivity and stability of sorghum and millet cultivated during the rainy season. While these changes have been accomplished only during the rainy season, the lessons of adaptation demonstrates and can be generalized to show how manipulations of -genotype and environment have resulted in accelerated growth rates and stability in production which can be applied to similar areas of climate. In addition, analyzing areas in similar situations show the technological potential for a change in productivity and sta- bility. Srivastava (1989) found that breeding strategies are increasingly being directed toward the development of horizontal or generalized resistance. He indicates that barley and wheat cul- tivars with improved yield stability that yield satisfactorily under stress and respond to favorable conditions with substantially higher yields have been developed. Virmani's (1989) analysis found that improved cropping systems can cope best with ;ain- fall variability when they are applied simultaneously with an efficient set of land and water man- agement techniques. Jin et al. (1994) studied the effects of climate change on rice production in Southern China and ran some scenarios of possible strategies for adaptation to climate change. The re- sulting analysis showed that creating a new cultivar created a higher yield in five out of seven sites. Changing the planting dates of the currently used cultivars caused increased rice yields in the northern sites, but not in the southern sites. Combining the changing of both the cultivars and the planting dates significantly increased the rice yields at six of the seven site locations. RESPONSES OF THE INTERNATIONAL TRADE AND POLICY COMMUNITY While a large portion of the debate taking place is over how climate change occurs and at what rate is it occurring, another area under consideration is how to best deal with the phenome- non, whose responsibility is climate change reduction and at what level will policies be most ef- fective, state, international, or globally? Many authors have conducted economic impact studies on international trade at the global and regional levels (Reilly, Hohmann, and Kane 1993; Reilly and Hohmann 1993). While other authors have focused on international policy responses to cli- mate change (Larson and Tobey 1994; Reilly 1995; Drennen 1993; Stewart 1992; and Viscusi 1992). 20 Reilly and Hohmann (1992) investigate the agricultural effects of climate change recog- nizing that effects will simultaneously occur worldwide. The authors found that there are still significant sources of error due largely to the underlying uncertainties in many variables (e.g. the climate scenarios, agronomic factors, and increased competition for land or water due to in- creased demand from other sectors caused by climate change). In addition, like other studies, Reilly and Hohmann note that the models currently used in the analyses do not included techno- logical change, population growth, or other changes that may occur simultaneously with eco- nomic growth and development. Larson and Tobey (1994) analyze the implications of alternative policy responses to un- certain climate change within the context of a simple dynamic economic model. Larson and To- bey focus on the policy question of how best to respond in such an uncertain environment given the uncertainty surrounding predictions of future global climate change. Stewart (1992) examines two approaches to global change policy, comprehensive and market-based. In this examination, Stewart points out that there is a need to integrate both the physical and social science in addressing all of the relevant components of the complex global system. In addition, he notes that there are three basic kinds of questions to answer before one can examine the policy prescriptions (1992, 26-27): (1) To what extent and when will global change occur? What would be the impacts of global change, and their costs and benefits? What further scientific research is needed to resolve remaining uncertainties? (2) What are the costs and benefits of measures to limit or adapt to global change? In light of these costs and benefits, what actions, if any, are warranted now? What is the appropriate combination of measures to limit net emissions of trace gases contributing to global change and measure to adapt to any adverse effects of global change? (3) If limitation efforts are warranted - a big "if' that can only be decided based on a careful look at the costs, benefits, and uncertainties - how should they be de- signed? Should they be narrowly focused on specific activities, or comprehensive to match the global system? Should they employ traditional command-and- control regulations, or make use of market-based economic-incentive tools? In the World Resources Institute publication Greenhouse Warming: Negotiating a Global Regime (1991), 9 authors present a range in discussion on the global response to green- house warming. The chapters discuss such topics as: Lessons from "the Ozone Hole"; Beyond Vienna and Montreal - Multilateral Agreements on Greenhouse Gases; Elements of a Framework Treaty on Climate Change; A Proposed Structure for an International Convention on Climate Change; Alternative Legal and Institutional Approaches to Global Change; Managing the Tran- sition to a Global Warming Regime or What to Do til the Treaty Comes; and Negotiating a Re- gime to Control Global Warming. This document is a comprehensive approach in how to deal "globally" with climate change - through a global regime and treaties. The material presented in this document was an effort to increase the interest of the widest possible choices to be consid- ered at several global conferences on climate change. 21 Reilly et al. (1993) conducted a study based on the agricultural production, price and economic welfare implications for 32 separate geographic regions computed for 9 scenarios us- ing 3 different GCMs, estimated with and without the direct effects of carbon dioxide on plant growth, and with different levels of adaptation. The authors major conclusions were similar to other impact studies done on agricultural production, economic welfare losses tend to be more severe in developing countries, major agricultural exporters will gain significantly if global agli- cultural prices rise, and the CO2 fertilization effect offsets losses due to climate change alone. Reilly et al. offer six policy relevant conclusions (1993, 34-35): (1) Based on current analyses, climate change effects on global agriculture appear manageable and possibly beneficial for an equilibrium doubled trace gas climate, but the possibility of more severe effects cannot be ruled out until more is known about the nature of transient climate and economic models are designed that can better consider adjustment costs. (2) International trade is an important risk pooling mechanism. (3) Significant relative change in agricultural productivity is likely whether the net effect on global agricultural production potential is negative or positive. Attempts by countries to protect domestic agricultural producers could interfere with inter- national trade and such trade interventions would create additional losses. (4) The difficulty of predicting climate change at relevant scales and time frames for agricultural decision-making suggests that regional experiment stations should continue to play an important role in evaluating and selecting crops and agricul- tural production strategies. Considerations of climate change place greater em- phasis on evaluating broader measures of success such as profitability rather than narrower measures such as yield. (5) It is difficult to predict with confidence the direction of economic impact for specific areas; flexibility is a key to minimizing the cost of adjustment. For de- veloping countries, increased education and economic development that includes development of manufacturing appears to increase flexibility and serve as an in- surance against climatic conditions that may turn unfavorable for agriculture. (6) Subsistence agricultural systems are most at risk because they cannot avail them- selves of the risk pooling value of markets. Education, economic development, and integration of these areas into national commodity and labor markets appear to be successful insurance policies against the possibility that agricultural condi- tions could degrade substantially in some areas. DISCUSSION A great deal of effort has been put into estimating the extent of climate change under future scenarios in various parts of the world. A growing amount of effort is being invested in estimating the impact of climate change on various sectors in various countries. Like in the story about the person who lost a coin and looks for it under the light, most of the work reported in the literature is concentrated in developed countries. 22 Given the great deal of controversy surrounding the issue of climate change-different interpretations of the evidence, different interests of policy makers, and different evaluation of intervention options-complicates the measures that might be appropriate in dealing with such phenomena. Toman (1997) suggests a policy envelope for climate change decision making. It exists on a comprehensive framework that takes into account: * Risks and costs (of impact and control) * Long-term considerations * Adaptation considerations * Regional and international cooperation * Distributional impacts With these components in mind one can already suggest a direction for future work in preparing the background work, and the future scenarios and responses. This report focuses on two aspects from the above list-adaptation and distribution. The various studies reviewed in this report suggest that there is much more flexibility introduced into the policy framework if adaptation is considered. Distributional impacts are another important pan of impact of and responses to climate change. Distributional impacts both within affected individuals and sectors, and among countries regions, indicate the need to introduce, both into re- search and policy, spatial effects and considerations. Distributional issues also prevail between developed and developing countries. Given the relatively lack of information and sufficient economic analysis on impact of and adaptation to climate change in developing countries, it is probably one of the most impor- tant tasks to initiate empirical research in developing countries, who might be subject to negative impact of climate change in the future. REFERENCES CITED Adams, R., Fleming, R., Chang, C., McCarl, B., and Rosenzweig, C. 1993. "A Reassessment of the Economic Effects of Global Climate Change on U.S. Agriculture." Unpublished: September. ., Glyer, D., and McCarl, B. 1989. "The Economic Effects of Climate Change on U.S. Agriculture: A Preliminary Assessment." In Smith, J., and Tirpak, D., eds., The Poten- tial Effects of Global Climate Change on the United States. Washington, D.C.: USEPA. ., Rosenzweig, C., Peart, R., Ritchie, J., McCarl, B., Glyer, D., Curry, B., Jones, J., Boote, K., and Allen, H. 1990. "Global Climate Change and U.S. Agriculture." Nature. 345 (6272, May): 219-224. Antle, J. M. 1995. "Climate Change and Agriculture in Developing Countries." American Jour- nal ofAgricultural Economics. 77 (August): 741-746. Bacsi, Z., Thornton, P. K., and Dent, J. B. 1991. "Impacts of Future Climate Change on Hun- garian Crop Production: An Application of Crop Growth Simulation Models." Agricul- tural Systems. 37: 435-450. 23 Barron, E. J. 1995. "Advances in Predicting Global Warming." The Bridge (National Academy of Engineering). 25 (2, Summer): 10-15. Bazzaz, F. A., and Fajer, E. D. 1992. "Plant Life in a C02 Rich World." Scientific American. 1821. Brack, D. and M. Grubb. 1996. Climate Change, "A Summary of the Second Assessment Report of the IPCC." FEEM (Fondazione ENI Enrico Mattei, Milano Italy) newsletter, 3, 1996. Crosson, P. 1993. "Impacts of Climate Change on the Agriculture and Economy of the Missouri, Iowa, Nebraska and Kansas (MINK) Region." In Kaiser, H., and Drennen, T., eds., Agri- cultural Dimensions of Global Climate Change. Cambridge: Cambridge University Press. Dadhwal, V. K. 1989. "Effect of Temperature on Wheat in India." In Climate and Food Secu- rity. International Symposium on Climate Variability and Food Security in Developing Countries 5-9 February 1987. Manila, Philippines: International Rice Research Institute and Washington, D.C.: American Association for the Advancement of Science. Darwin, R. 1994. "Shifting Uses for Natural Resources in Changing Climate." World Resources Review, FAO Agrostat 1961-1990. . Rome ., Lewandrowski, J., Tsigas, M., and Raneses, A. 1994. "Shifting Uses for Natural Re- sources in a Changing Climate." World Resource Review. 6 (4): 559-569. Decker, W. L., and Achutuni, V. R. 1990. "Agricultural Production and Agricultural Policy Op- tions in Response to Global Warming." World Resource Review. 2 (4): 418-440. Downing, T. E. 1993. "The Effects of Climate Change on Agriculture and Food Security." Re- newable Energy. 3 (4/5): 491-497. ., and Parry, M. 1993. "Long-run Global Change: Implication for Agricultural Sus- tainability." pp.137-162. In Agricultural Sustainability, Growth, and Poverty Alleviation: Issues and Policies. Washington, D. C.: International Food Policy Research Institute, and Feldafing Germany: Food and Agriculture Development Centre. Epstein, P. R. 1994. "Health and Global Change: Human Pathogens and Agricultural Pests." Lecture Notes. Harvard School of Public Health FAO (Food and Agriculture Organization of the United Nations) 1990. "Climate Change and Agriculture, Forestry and Fisheries." Second World Climate Conference. International Conference Centre, Geneva, Switzerland: FAO Position Paper. 29 October to 7 Novem- ber. . 1994. "FAO Year Book, Production." FAO Statistics Series. (117). Rome: FAO. Frederiksen, H. 1992. "Water Resources Institutions." World Bank Technical Paper. (191). Washington, D.C.: The World Bank. Fuglestvedt, J., Hanisch, T., Isaksen, I., Selrod, R., Strand, J., and Torvanger, A. 1994. "A Re- view of Country Case Studies on Climate Change." Working Paper. (7). Washington, D.C.: Global Environment Facility, UNDP, UNEP, and The World Bank. 24 Gleick, P. 1989. "Global Climatic Change, Water Resources, and Food Security." In Climate and Food Security. International Symposium on Climate Variability and Food Security in Developing Countries 5-9 February 1987. Manila, Philippines: International Rice Re- search Institute and Washington, D.C.: American Association for the Advancement of Science. Innes, R., and Kane, S. 1995. "Agricultural Impacts of Global Warming: Discussion." American Journal ofAgricultural Economics. 77 (August): 747-750. Jin, Z., Ge, D., Chen, H., and Fang, J. 1994. "Effects of Climate Change on Rice Production and Strategies for Adaptation in Southern China." In Rosenzweig, C., and Iglesias, A., eds., Implications of Climate Change for International Agriculture: Crop Modeling Study. Washington, D.C.: United States Environmental Protection Agency, Office of Policy, Planning and Evaluation, Climate Change Division, Adaptation Branch. Kaiser, H. M., and Crosson, P. 1995. "Implications of Climate Change for U.S. Agriculture." American Journal ofAgricultural Economics. 77 (August): 734-740. _____., Drennen, T., and eds. 1991. Agricultural Dimensions of GlobalClimate Change. Delray Beach, FL: St. Lucie Press. ______., Riha, S. J., Wilks, D. S., and Sampath, R. 1993. "Adaptation to Global Climate Change at the Farm Level." In Kaiser, H. M., and Drennen, T. E., eds., Agricultural Di- mensions of Global Climate Change. Delray Beach, FL: St. Lucie Press. ______., Riha, S. J., Wilks, D. S., and Sampath, R. 1995. "Potential Implications of Climate Change for U.S. Agriculture." ERS Staff Paper. (9522). Washington, D C.: United States Department of Agriculture, Economic Research Service, Natural Resources and Envi- ronment Division. Kane, S. M., Reilly, J. M., and Tobey, J. 1992. "A Sensitivity Analysis of the Implications of Climate Change for World Agriculture." In Reilly, J. M., and Anderson, M., eds. Eco- nomic Issues in Global Climate Change: Agriculture, Forestry, and Natural Resources. Boulder, CO: Westview Press. ______., Reilly, J., and Tobey, J. 1991. "Climate Change: Economic Implications for World Agriculture." Agricultural Economic Report. (647). Washington, D.C.: United States Department of Agriculture, Economic Research Service. Karim, Z., Ahmed, M., Hussain, S. G., and Rashid, K. B. 1994. "Impact of Climate Change on the Production of Modem Rice in Bangladesh." In Rosenzweig, C., and Iglesias, A., eds., Implications of Climate Change for International Agriculture: Crop Modeling 'Study. Washington, D.C.: United States Environmental Protection Agency, Office of Policy, Planning and Evaluation, Climate Change Division, Adaptation Branch. Kimball, B. A. 1983. "Carbon Dioxide and Agricultural Yield: An Assemblage and Analysis of 430 Prior Observations." Agronomy Journal. 75: 779-788. _____., Mauney, J. R., Nakayama, F. S., and Idso, S. B. 1993. "Effects of Elevated C02 anc. Climate Variables on Plants." Journal of Soil and Water Conservation. 48: 9-14. 25 Lonergan, S. 1991. "Climate Warming, Water Resources and Geopolitical Conflict: A Study of Nations Dependent on the Nile, Litani and Jordan River Systems." Extra-Mural Paper. (55). Ottawa, Canada: Operational Research and Analysis Establishment, Department of National Defence. . 1994. "Environmental Change and Regional Security in Southeast Asia." Operational Research and Analysis Directorate of Strategic Analysis, Project Report, (659). Ottawa, Canada: Department of National Defence. ., and Kavanagh, B. 1991. "Climate Change, Water Resources and Security in the Middle East." Global Environmental Change (September): 272-290. ., and Young, K. 1989. "An Assessment of the Effects of Climate Warming on Energy Developments in the Mackenzie River Valley and Delta, Canadian Arctic." Energy Ex- ploration and Exploitation. 7 (5): 359-381. Mathews, J. T. 1991. "Introduction and Overview." In Greenhouse Warming: Negotiating a Global Regime. Washington, D. C.: World Resources Institute. Mendelsohn, R. 1996. "The Market Impacts of Climate Change Across Nations: A Comparison of Effects in the United States, Brazil, and India." Environmetrics. Sao Paulo, Brazil. July. ., Nordhaus, W. D., and Shaw, D. 1994. "The Impact of Global Warming on Agricul- ture: A Ricardian Analysis." The American Economic Review. 84 (4, September): 753- 771. Mooney, H. A., and Koch, W. 1994. "The Impact of Rising C02 Concentrations on the Terres- trial Biosphere." Ambio. 23 (1): 74-76. Nemec, J., and Schaake, J. 1982. "Sensitivity of Water Resource Systems to Climate Variation." Hydrologic Science. 27 (3): 327-343. Nierenberg, W. A. 1995. "Progress and Problems: A Decade of Research on Global Warming." The Bridge (National Academy of Engineering). 25 (2, Summer): 4-9. Oram, P. A. 1989. "Sensitivity of Agricultural Production to Climatic Change, An Update." In Climate and Food Security. International Symposium on Climate Variability and Food Security in Developing Countries 5-9 February 1987. Manila, Philippines: International Rice Research Institute and Washington, D.C.: American Association for the Advance- ment of Science. Parikh, J. K., and Gokarn, S. 1993. "Climate Change and India's Energy Policy Options: New Perspectives on Sectoral C02 Emissions and Incremental Costs." Global Environmental Change (September): 276-291. ., and Shukla, V. 1995. "Urbanization, Energy Use and Greenhouse Effects in Eco- nomic Development: Results from a Cross-National Study of Developing Countries." RP. (95-196). Bombay, India: Indira Gandhi Institute of Development Research. Parry, M. L. 1990. Climate Change and World Agriculture. London Enland: Earthscan Publica- tions Ltd. 26 ., Blantran de Rozari, M., Chong, A. L., Panich, S., eds., 1992. "The Potential Socio- Economic Effects of Climate Change in South-East Asia." Nairobi: United Nations Envi- ronment Programme. ., and Rosenzweig, C. 1994. "Food Supply and Risk of Hunger, Health and Climate Change." The LANCET. London: 23-25. Pastor, J., and Post, W. 1988. "Response of Northern Forests to C02 Induced Climate Change." Nature. 334: 55-58. Peterson, D. F. 1989. "Irrigation as a Factor in Food Security." In Climate and Food Security. International Symposium on Climate Variability and Food Security in Developing Coun- tries 5-9 February 1987. Manila, Philippines: International Rice Research Institute and Washington, D.C.: American Association for the Advancement of Science. Pimentel, D. 1991. "Global Warming, Population Growth, and Natural Resources for Food Pro- duction." Society and Natural Resources. 4: 347-363. Qureshi, A., and Iglesias, A. 1994. "Implications of Global Climate Change for Agriculte in Pakistan: Impacts on Simulated Wheat Production." In Rosenzweig, C., and Iglesias, A., eds., Implications of Climate Change for International Agriculture: Crop Modeling Study. Washington, D.C.: United States Environmental Protection Agency, Office of Policy, Planning and Evaluation, Climate Change Division, Adaptation Branch. Rao, N. G., P., Rao, G. R., K., and Acharya, H. S. 1989. "Yield Stability of Sorghum and Millet Across Climates." In Climate and Food Security. International Symposium on Climate Variability and Food Security in Developing Countries 5-9 February 1987. Manila, Phil- ippines: International Rice Research Institute and Washington, D.C.: American Associa- tion for the Advancement of Science. Reilly, J. 1995. "Climate Change and Global Agriculture: Recent Findings and Issues." Ameri- can Journal ofAgricultural Economics. 77 (August): 727-733. ____ .,I and Thomas, C. 1993. Toward Economic Evaluation of Climate Change Impacts: A Review and Evaluation of Studies of the Impact of Climate Change. Cambridge, MA: Center for Energy and Environmental Policy Research, MIT. Rosenzweig, C., and Iglesias, A. 1994. "The Use of Crop Models for International Climate Change Impact Assessment: Study Design, Methodology, and Conclusions." In Ro- senzweig, C., and Iglesias, A., eds. Implications of Climate Change for International Ag- riculture: Crop Modeling Study. Washington, D.C.: United States Environmental Pro- tection Agency, Office of Policy, Planning and Evaluation, Climate Change Division, Adaptation Branch. ., and Parry, M. L. 1993. "Potential Impacts of Climate Change on World Food Sup- ply: A Summary of a Recent International Study." In Kaiser, H. M., and Drennen, T. E., eds., Agricultural Dimensions of Global Climate Change. Delray Beach, Fla: St. Lucie Press. 27 ., Parry, M. L., Fischer, G., and Frohberg, K. 1993. "Climate Change and World Food Supply." Research Report. (3). Oxford, England: Environmental Change Unit, Univer- sity of Oxford. Sanghi, A., Alves, D., Evenson, R., and Mendelsohn, R. 1997. "Global Warming Impacts on Brazilian Agriculture: Estimates of the Ricardian Model." Economia Aplicada Janeiro- Marco. Schimmelpfennig, D., J. Lewandrowski, J. Reily, M. Tsigas, and I. Parry. 1996. "Agricultural Adaptation to Climate Change Issues of Longrun Sustainability." An Economic Research Service Report. United States Department of Agriculture Agricultural Economic Report Number 740, Washington, DC, June. Schulze, R. E., Kiker, G. A., and Kunz, R. P. 1993. "Global Climate Change and Agricultural Productivity in Southern Africa." Global Environmental Change (December): 330-349. Shaw, D., Mendelsohn, R., and Nordhaus, W. 1994. The Impact of Climate Variation on Agri- culture. New Haven, CT: Yale FES. Smith, D. R. 1994. "Change and Variability in Climate and Ecosystem Decline in Aral Sea Ba- sin Deltas." Post-Soviet Geography. 35 (3): 142-165. Sonka, S. T. 1992. "Evaluating Socioeconomic Assessments of the Effect of Climate Change on Agriculture." In Reilly, J. M., and Anderson, M., eds., Economic Issues in Global Cli- mate Change: Agriculture, Forestry, and Natural Resources. Boulder, CO: Westview Press. Srivastava, J. P. 1989. "Development of Crop Germplasm with Improved Tolerance for Envi- ronmental and Biotic Stresses." In Climate and Food Security. International Symposium on Climate Variability and Food Security in Developing Countries 5-9 February 1987. Manila, Philippines: International Rice Research Institute and Washington, D.C.: Ameri- can Association for the Advancement of Science. Strzepek, K. M., Onyeji, S. C., Saleh, M., and Yates, D. N. 1994. "An Assessment of Integrated Climate Change Impacts on Egypt." Working Paper. (94-48). Laxenburg, Austria: Inter- national Institute for Applied Systems Analysis. ., Smith, J. B., and eds. 1989. As Climate Changes: International Impacts and Impli- cations - Executive Summary. Cambridge, MA: Cambridge University Press for the United States Environmental Protection Agency. Toman, M. A., Firor, J., and Darmstadter, J. 1996. "Climate Change and Its Consequences." Re- sources (124, Summer): 10-13. Toman, M. A. 1977. "A framework for Climate Change Policy." Resources (127, Spring): 11-14. USEPA 1984. Potential Climatic Impacts of Increasing Atmospheric C02 with Emphasis on Water Availability and Hydrology in the US. Washington, DC: USEPA, Strategic Studies Staff. Office of Policy, Planning and Evaluation. Vidanage, S. P., and Abeygunawardena, P. 1994. "An Economic Assessment of Global Warm- ing on Agriculture: The Case of Paddy Production in Sri Lanka." MARGA. 13 (1): 33-44. 28 Virnani, S. M. 1989. "Cropping Systems Strategies for Coping with Climatic Fluctuations." In Climate and Food Security. International Symposium on Climate Variability and Food Security in Developing Countries 5-9 February 1987. Manila, Philippines: International Rice Research Institute and Washington, D.C.: American Association for the Advance- ment of Science. Walker, B. H. 1994. "Landscape to Regional-Scale Responses of Terrestrial Ecosystems to Global Change." Ambio. 23 (1, February): 67-73. Woodward, IF. I. 1993. "Leaf Responses to the Environment and Extrapolation to Larger Scales." In Solomon, A., and Shugart, H., eds., Vegetation Dynamics and Global Change. New York, NY: Chapman and Hall. WRI (Water Resources Institute) 1992. World Resources 1992-1993. Oxford, England: Oxford University Press. Xie, M., Kuffner, U., and Le Moigne, G. 1993. "Using Water Efficiently." World Bank Techni- cal Paper. (205). Washington, D.C.: The World Bank. ADDITIONAL REFERENCES ON ToPIc Antle, J. M. 1994. "Climate Change and Agriculture in Developing Countries." Allied Social Sciene Association. Washington, D.C. January. Arrhenius, E., and Waltz, T. W. 1990. "The Greenhouse Effect: Implications for Economic De- velopment." World Bank Discussion Papers. (78). Washington, D.C.: The World Bank. Arrow, K. J., Cline, W. R., Maler, K., Munasinghe, M., and Stiglitz, J. E. 1995. "Intertemporal Equity, Discounting, and Economic Efficiency." In Munasinghe, M., ed., Global Climate Change: Economic and Policy Issues, World Bank Environment Paper Number 12. Washington, D.C.: The World Bank. Atkinson, D., and ed. 1993. "Global Climatic Change: Its Implications for Crop Protection." British Crop Protection Council Monograph. (56). Surrey, England: British Crop Protec- tion Council. Baranzini, A., Chesney, M., and Morisset, J. 1995. "Uncertainty and Global Warming: An Op- tion-Pricing Approach to Policy." Policy Research Working Paper. (1417). Washington, D.C.: The World Bank. Benedick, R. E. 1991. "Lessons from the Ozone Hole." In Greenhouse Warming: Negotiating a Global Regime. Washington, D. C.: World Resources Institute. Chattopadhyay, D., and Parikh, J. K. 1993. "C02 Emissions Reduction From the Power System in India." Natural Resources Forum (November): 251-261. Chayes, A. 1991. "Managing the Transition to a Global Warming Regime or What to Do til the Treaty Comes." In Greenhouse Warming: Negotiating a Global Regime. Washington, D. C.: World Resources Institute. Crosson, P. 1994. "Degradation of Resources as a Threat to Sustainable Agriculture. First World Congress of Professionals in Agronomy." Santiago, Chile. 5-8, September. 29 Darwin, R., Tsigas, M., Lewandrowski, J., and Raneses, A. 1995. "World Agriculture and Cli- mate Change: Economic Adaptations." Agricultural Economic Report. (703). Washing- ton, D.C.: United States Department of Agriculture, Economic Research Service. Downing, T. E., and Parry, M. L. 1991. "Long-run Global Change: Implications for Agricul- tural Sustainability." In Vosti, S. A., ed., Agricultural Sustainability, Growth, and Pov- erty Alleviation: Issues and Policies - Proceedings of the Conference heldfrom 23 to 27 September 1991 in Feldafing, Germany. Washington, D.C.: International Food Policy Research Institute (IFPRI). Drennan, T. E. 1993. "After RIO: The Status of Climate Change Negotiations." In Kaiser, H. M., and Drennen, T. E., eds. Agricultural Dimensions of Global Climate Change. Delray Beach, FL: St. Lucie Press. Easterling, W. A., Rosenberg, N. J., McKenney, M. S., Jones, C. A., Dyke, P. T., and Williams, J. R. 1992. "Preparing the Erosion Productivity Impact Calculator (EPIC) Model to Simulate Crop Response to Climate Change and the Direct Effects of C02." Agricultural and Forest Meteorology. 59: 17-34. Easterling, W. E., and Sonka, S. T. 1989. "Economic Adjustments to Climatic Variability." In Climate and Food Security. International Symposium on Climate Variability and Food Security in Developing Countries 5-9 February 1987. Manila, Philippines: International Rice Research Institute and Washington, D.C.: American Association for the Advance- ment of Science. Femandes de Siqueira, 0. J., Farias, J. R., B., and Aguiar Sans, L. M. 1994. "Potential Effects of Global Climate Change for Brazilian Agriculture: Applied Simulation Studies for Wheat, Maize, and Soybeans." In Rosenzweig, C., and Iglesias, A., eds., Implications of Climate Change for International Agriculture: Crop Afodeling Study. Washington, D.C.: United States Environmental Protection Agency, Office of Policy, Planning and Evaluation, Climate Change Division, Adaptation Branch. Fernando, C. S., Fitzgerald, K. B., Kleindorfer, P. R., and Munasinghe, M. 1995. "Financial Global Environmental Programs: Efficient Approaches to Cooperation and Institutional Design." In Munasinghe, M., ed., Global Climate Change: Economic and Policy Issues, World Bank Environment Paper Number 12. Washington, D.C.: The World Bank. Gangadhar Rao, D., and Sinha, S. K. 1994. "Impact of Climate Change on Simulated Wheat Production in India." In Rosenzweig, C., and Iglesias, A., eds., Implications of Climate Change for International Agriculture: Crop Modeling Study. Washington, D.C.: United States Environmental Protection Agency, Office of Policy, Planning and Evaluation, Climate Change Division, Adaptation Branch. Gleick, P. 1986. "Methods for Evaluating the Regional Hydrologic Impacts of Global Climate Changes." Journal of Hydrology. 88: 97-116. . 1989. "Climate Change and International Politics: Problems Facing Developing Countries." Ambio. 18 (6): 333-339. Hansen, L. 1991. "Farmer Response to Changes in Climate: The Case of Corn Production." The Journal ofAgricultural Economics Research. 43 (4, Fall): 18-. 30 Jones, C. A., Dyke, P. T., Williams, J. R., Kiniry, J. R., Benson, V. W., and Griggs, R. H. 1991. "EPIC: An Operational Model for Evaluation of Agricultural Sustainability." Agricul- tural Systems. 37: 341-350. Joyce, L. A., Fosberg, M. A., and Comanor, J. M. 1990. "Climate Change and America's For- ests." USDA Forest Service General Technical Report. (RM-187). Washington, DC: USDA. Larson, B. A., and Tobey, J. A. 1994. "Uncertain Climate Change and the International Policy Response." Ecological Economics. I1: 77-84. Lonergan, S., Difrancesco, R., and Woo, M. 1993. "Climate Change and Transportation in Northern Canada: An Integrated Impact Assessment." Climatic Change. 24: 331-351. Minami, K. 1990. "Agricultural Ecosystems and the Greenhouse Effect." In Agriculture and the Greenhouse Effect: National Institute of Agro-Environmental Sciences. Mount, T. 1992. "Global Climate Change: Effects on Agriculture." In Reilly, J. M., and Ander- son, M., eds., Economic Issues in Global Climate Change: Agriculture, Forestry, and Natural Resources. Boulder, CO: Westview Press. Munasinghe, M., Meier, P., Hoel, M., Hong, S. W., and Aaheim, A. 1995. "Applicability of Techniques of Cost-Benefit Analysis to Climate Change." In Munasinghe, M., ed., Global Climate Change: Economic and Policy Issues, World Bank Environment Paper Number 12. Washington, D.C.: The World Bank. Nitze, W. A. 1991. "A Proposed Structure for an International Convention on Climate Change." In Greenhouse Warming: Negotiating a Global Regime. Washington, D. C.: World Re- sources Institute. NOAA 1994. "El Nino and Climate Prediction." Reports to the Nation On Our Changing Planet. (Spring, No. 3). Silver Spring, MD: University Corporation for Atmospheric Re- search. Offutt, S. 1993. "Bridging the Gap Between Policymakers and Academics." In Kaiser, H. M., and Drennen, T. E., eds., Agricultural Dimensions of Global Climate Change. Delray Beach, FL: St. Lucie Press. Parikh, J. K. 1994. "North-South Cooperation in Climate Change Through Joint Implementa- tion." RP. (94-142). Bombay, India: Indira Gandhi Institute of Development Research. . 1995. "North-South Issues for Climate Change." RP. (95-173). Bombay, India: In- dira Gandhi Institute of Development Research. Reilly, J., Hohmann, N., and Kane, S. 1994. "Climate Change and Agricultural Trade: Who Benefits, Who Loses?" Global Environmental Change. 4 (1): 24-36. ., and Hohmann, N. 1993. "Climate Change and Agriculture: The Role of Interna- tional Trade." AEA Papers and Proceedings. 83 (2, May): 306 - 312. _____., Hohmann, N., and Kane, S. 1993. Climate Change and Agriculture: Global and Re- gional Effects Using an Economic Model of International Trade. Cambridge, MA: Center for Energy and Environmental Policy Research, MIT. 31 Richardson, E. L. 1991. "Elements of a Framework Treaty on Climate Change." In Greenhouse Warming: Negotiating a Global Regime. Washington, D. C.: World Resources Institute. Riebsame, W. E. 1989. "Assessing the Social Implications of Climate Fluctuation Through Cli- mate Impact Studies." In Climate and Food Security. International Symposium on Cli- mate Variability and Food Security in Developing Countries 5-9 February 1987. Manila, Philippines: International Rice Research Institute and Washington, D.C.: American As- sociation for the Advancement of Science. Robertson, T., Benson, V. W., and Williams, J. R. 1987. "Impacts of Climate Change on Yields and Erosion for Selected Crops in the Southern United States." Symposium on Climate Change in the Southern United States: Future Impacts and Present Policy Issues. New Orleans, LA. 28-29, May. Sanders, G. E., Dixon, J., and Cobb, A. H. 1993. "Will Increasing Ozone Pollution Associated with Global Climate Change Alter Crop Tolerance to Herbicides?" In Atkinson, D., eds., Global Climatic Change: Its Implications for Crop Protection. Surrey, England: British Crop Protection Council. Sebenius, J. K. 1991. "Negotiating a Regime to Control Global Warming." In Greenhouse Warming: Negotiating a Global Regime. Washington, D. C.: World Resources Institute. Seino, H. 1990. "Global Warming and Its Effect on Food Production." In Agriculture and the Greenhouse Effect: National Institute of Agro-Environmental Sciences. Sivakumar, M. V., K. 1989. "Some Strategies to Cope with Drought in the Sahelian Zone." In Climate and Food Security. International Symposium on Climate Variability and Food Security in Developing Countries 5-9 February 1987. Manila, Philippines: International Rice Research Institute and Washington, D.C.: American Association for the Advance- ment of Science. Stewart, R. B. 1992. "Comprehensive and Market-Based Approaches to Global-Change Policy." In Reilly, J. M., and Anderson, M., eds., Economic Issues in Global Climate Change: Agriculture, Forestry, and Natural Resources. Boulder, CO: Westview Press. Thacher, P. S. 1991. "Alternative Legal and Institutional Approaches to Global Change." In Greenhouse Warming: Negotiating a Global Regime. Washington, D. C.: World Re- sources Institute. Viscusi, W. K. 1992. "Implications of Global-Change Uncertainties: Agricultural and Natural Resource Policies." In Reilly, J. M., and Anderson, M., eds., Economic Issues in Global Climate Change: Agriculture, Forestry, and Natural Resources. Boulder, CO: Westview Press. Williams, J. R., Jones, C. A., Kiniry, J. R., and Spanel, D. A. 1989. "The EPIC Crop Growth Model." Transactions of the ASAE. 32 (2, March-April): 497-511. Wirth, D. A., and Lashof, D. A. 1991. "Beyond Vienna and Montreal - Multilateral Agreements on Greenhouse Gases." In Greenhouse Warming: Negotiating a Global Regime. Wash- ington, D. C.: World Resources Institute. 32 3 CLIMATE WARMING AND INDIA Steve Lonergan INTRODUCTION The purpose of this study is to develop climate scenarios - based on the projections of general circulation modes or GCMs of the earth's atmosphere - to be used as input to an analysis of the impacts of potential climate warming on agriculture in India. The study uses the projections from three GCMs to develop projections of temperature and precipitation in india under a doubling of CO2 from pre-industrial levels. The chapter is divided into five sections. Section one is the short introduction. The second section outlines the general characteristics of climate warming - or the "enlianced greenhouse effect," and tries to clarify key aspects of the controversy over whether climate warming is occuring. Section three discusses the general circulation models (GCMs) which are used to project changes in climate under various levels of carbon dioxide in the atmosphere. These GCMs were used to generate the results presented in this paper. The fourth section highlights the key changes in climate - notably temperature, precipitation and evapo-transpiration - which are expected to occur under a doubling of CO2 in the atmosphere. These projections focus on India, and are presented for the country as a whole and for a north-south division. The final section summarizes the conclusions of the paper. THE GENERAL CASE OF CLIMATE WARMING Possibly the greatest environmental threat society will fact over the next century is that of global warming, with the potential to disrupt natural and social systems throughout the world. This threat is the result of the anthropogenic emission of certain gases, most notably CO. CHi, CFCs, NO and water vapor, into the atmosphere, contributing to a general process known as the "greenhouse effect." The term greenhouse gas has been applied to atmospheric gases that are relatively transparent to incoming short wave solar radiation but which absorb the long wave radiation from the surface of the earth and reemit it downward, warming the surface of the earth and the lower atmosphere. The primary concern to date has been with C02, released from the burning of coal and other carbon-based fuels and the burning and decay of the world's forests. As depicted in Figure 3.1, there is a strong relationship between atmospheric levels of CO2 and temperature. Since 1958, when measurements of CO2 in the atmosphere began, its concentration has increased from 315 to over 360 parts per million (ppm). A doubling of CO2 from pre-industrial levels - expected sometime next century - could result in a general warming of the earth's surface, although to what extent is not known, since the present concentrations of CO2 in the atmosphere are far greater than at any previous time. Concentrations of other greenhouse gases, currently at much lower concentrations than CO2, but potentially more potent, are increasing even more rapidly. Methane, which is emitted from wetlands, rice paddies, livestock and warming permafrost, is increasing at a rate of one percent per year (compared to 0.4% per year f&l C02). CFCs have been increasing at five percent per year, although recent international agreements have resulted in a significant decline in the global emissions of CFCs.I Table 3.1 lists the major greenhouse gases, their sources and their characteristics. Figure 3.1: Relationship between atmospheric CO2 concentrations and temperature 280 -260 1~ 240 4-'~~~~~~~~~~~~~~~~~~~~~~- c9 2.5 220 0 40 B0 t0 6200 -2.5 180 c1 E -5.0 .c 7.5- -10.0 0 40 80 120 160 Age ( Thousand years before present) Source: Barnola et al. (1987) The theory of a general warming of the world's climate has gained widespread acceptance over the past two decades, despite the reluctance of some governments to agree to emission reduction requirements. In addition, there have been numerous studies of the potential impacts of global warming, by sector (such as for health, agriculture, water and tourism) and by region (including India; ADB, 1994). Most of these studies incorporate scenarios of climate warming which are drawn from the output of large-scale computer models of the atmosphere, known as general circulation modes, or GCMs. The present study is based on the projections of three of these GCMs. I There remains some concern that CFC production in developing countries, and, particularly, China and India, will offset any reductions in emissions from industrialized countries. There is international disagreement at present over the level of financial support promised to developing nations to ensure reductions in CFC emissions will not negatively affect economic growth. To date, as of late 1996, Thailand is the only developing country to voluntarily restrict CFC emissions. 34 When addressing climate warming and its potential impacts, there are two issues wvhich must be considered. First, the GCMs provide medium or long term projections of climate under various concentrations of greenhouse gases in the atmosphere. The most common projection. and the one that will be used in this study, is of a doubling of the CO2 concentration in the atmosphere from pre-industrial levels (termed the 2 x CO2 scenario). At present emission rates - and this is a restrictive assumption - a doubling of CO2 is expected to occur in the middle of the next century. With this doubling, the models project, the average levels of temperature, precipitation and other climatic variables will change as well. A second, and potentially more important issue, is the short term variability that might accompany a general warming trend. Hurricanes and droughts would be more severe and more frequent; storm surges along the coasts would be higher, and the frequency and duration of periods of abnormally high temperatures would increase. In addition, much of the work on climate change and its impacts to date has been on the change in climate over time; when a CO2 doubling will occur and what effects a gradual increase in temperature might have on physical and natural systems. Of greater concern, however, may be climate changes over space. This may be even more relevant when considering short-term climate variability (and volatility), as the economic loss due to natural hazards is much higher in developing countries. Table 3.1: Major greenhouse gases and their characteristics Atmospheric Annual Relative Current Principle Concentration Increase Lifespan Efficiency Contribution Source Gas (ppm) (%) (years) (%) CO2 360 0.4 1 57 coal, oil, fossil fuel (44) natural gas biological (13) deforestation CFCs .000225 5 75 - 111 15,000 25 foams, coolants CH4 1.675 1 11 25 12 wetlands, rice, livestock NOX 0.31 0.2 150 230 6 fossil fuel, deforestation Source: Compiled from Houghton, et al. (1990) 35 THE GLOBAL WARMING CONTROVERSY Over the past few years, there has been a number of articles published in the popular press which criticize the projections of global warming. For the most part, the criticismns focas on the inability of the GCMs to accurately represent the complexities of our climate system. Many of these articles have been stimulated by the concern on the part of some western governments of the cost of imposing CO2 emission standards unilaterally. These have been estimated to be as high as 3% of GNP for the U.S. under a scenario of 20% reduction in 1988 levels of CO2 emissions by the year 2010 (Manne and Richels, 1990). Despite this controversy, what should be recognized is that the major sources of greenhouse gas emissions are known and that the role of these gases in influencing climate is well understood and accepted. The controversy surrounding the general global warming which may accompany the rapid increase in greenhouse gas emissions is based on our inability to accurately predict the effect these gases will have on climate in the presence of other atmospheric processes. Climate variables, such as temperature, precipitation and solar radiation have been projected under different concentrations of greenhouse gases in the atmosphere with the general circulation models. While these models are quite sophisticated, they are not able to capture all of the complexities of our climate system. In many cases, we simply do not know what the buffering capacity of clouds will be or the effect the oceans will have on mitigating the expected global warming. The following, however, are accepted. First, anthropogenic releases of greenhouse gases are increasing at a constant rate. Second, these gases, in the absence of other changes, will result in a general global warming. Third, there will be regional variation in the amount of warming (and changes in other climate parameters, such as precipitation). And fourth, the consequences of such warming could be very great, depending on the ability of systems to adapt to these potential changes. What is not accepted is how other earth systems, such as the oceans and clouds, will act to alter the expected warming or how natural variations in climate may mitigate many of the expected changes. GENERAL CIRCULATION MODELS Most of the predictions of climate warming discussed in both the scientific and popular literature are the result of output from three-dimensional, numerical models of the earth's atmosphere known as general circulation models or GCMs. As was noted above, these are numerical weather prediction models that simulate climate processes over long periods of time. They are generally based on five prognostic variables: temperature, humidity, surface pressure and two dimensions of wind, and are used for controlled runs, perturbed usually by changes in CO2 until they reach an equilibrium level. Generally, runs are done on a doubling of atmospheric CO2 concentrations from pre-industrial levels. The first GCM studies were undertaken by Manabe and Weathereld (1975) at the Geophysical Fluid Dynamics Laboratory (GFDL) in Princeton, New Jersey. More recent studies have been undertaken at the National Centre for Atmospheric Research (NCAR) in Boulder, Colorado, the Goddard Institute for Space Studies (GISS), the United Kingdom Meteorological Office (UKMO), and Oregon State University (OSU). A more complete description of these models can be found in Dickinson (1986). The most recent model to appear has been developed by the Canadian Climate Centre (CCC), a 36 branch of the Atmospheric Environment Service within Environment Canada (however, data from the CCC model is only available for a North American "window"). The model projections are made by grid cells which are roughly 400km by 400km. All of the models show remarkable consistency with observed temperature and precipitation when run at present levels of CO2 (see Figure 3.2), but their projections - particularly in regards to precipitation - differ considerably under a 2 x CO2 scenario. This study of India used output from three of the models, the GFDL, the UKMO, and the GISS models, as a basis for assessing the impacts of climate warming on the region. Despite the consistencies shown by the models, there are a number of problems which should be noted. First, they do not possess a degree of variability that is apparent in the real climate (Katz, 1988). This lack of variability is also a problem with the climate projections, since many analysts feel the greatest impact from climate warming may result from a greater magnitude and frequency of extreme events (extended droughts, more severe storms) than from gradual increases in mean temperature and precipitation. The GCMs are not able to project such changes in extreme events. Second, the models cannot provide information about temperature and precipitation changes at a small enough spatial scale to be useful for detailed impact studies. Accordingly, this analysis was done for all of India, along with an extension to the "north" and the "south" of the country. Third, the models are inadequate in treating one or more of five important feedback mechanisms in the climate system, including: water vapour; snow and sea ice; cloud cover; cloud radiative properties; and the ocean-atmosphere interface. The difficulties incorporating these feedback mechanisms into the climate models, and their potential importance in mitigating any tendency towards global warming, has contributed to the controversy surrounding the issue. Figure 3.2: Simulation runs of GCMs versus observed data 310- Observed (JAEGER) 300 - GISS // n ~ ~~~~ GFDL R30 290 ~~~~~~~~~~~~~~~~~....UKMO 2.5 deg Z 240 230 220 250 - -90 -80 .70 -60 -50 -40 -30 -20 -10 0 10 20 30 40 50 60 70 80 90 Latitude Source: Houghton et al. (1990) 37 CLIMATE CHANGE PROJECTIONS FOR INDIA The following figures and tables present monthly temperature, precipitation and soiar radiation projections for India under a 1 x CO2 scenario (pre-industrial levels) and a 2 x CO, scenario, from three different GCMs. The GCMs used were from the GFDL, the GISS and the UKMO. Model outputs for the globe were obtained from the National Center for Atmospheric Research in Boulder, Colorado. Point assignments of climate projections from each of the three models were used. The points denote the center of the grid cell on which the projections are based (the size of the cells varies according to the model). The GFDL Model consists of 12 cells, the GISS model consists of 7 cells, and UKMO model consists of 13 cells. TEMPERATURE Figures 3.3 - 3.11 and Tables 3.2 - 3.4 present temperature, precipitation and solar radiation data by the three GCMs, by month, quarter and year. The UKMO and GFDL models project temperature increases of between 16.2 and 23.5% from the 1xCO2 levels, respectively, with the GISS model projecting an increase of approximately 10% (Figure 3.6). Typical of GCM projections of temperature change under a doubling of C02, there is reasonable consistency in the results: all models project increases in temperature for all months of the year. (Figures 3.4, 3.5, and 3.8a). This is true despite the different spatial scales in which the GCMs operate. In absolute terms, temperatures are expected in increase between 2.33 C° and 4.78 CO over the entire country. Figures 3.9a, 3.10a and 3.11a show the monthly differences between the temperature scenarios by GCM. The UKMO model - which records the lowest pre-industrial temperature levels - projects the greatest increase in temperatures of 4.78 C°, an increase of over 23%. Complete data are presented in Tables 3.2 - 3.4, by GCM. 38 Figure 3.3: Climate change and Indiac - projections from three GCMs 35 30 25 20 UKM.!_ ii 0,- 5 . Temp ;Precip - Solar climatic variable Figure 3.4: Temperature levels projected by the GCMs for India-pre-industrial levels 35.00 - -- , ,, ,. X ' '"' 30.00 - ------- -#- - -- 25.00 - ---- -.-- - 4, 15.0 --- - - --- - - -- - - -- -- --;,- - ----- --GFDL 1 0 .0 0 '---- ]S------- ''-'- - ' - ^- - -- - - - -------- ---------.. . - - . - -- - - -- - -- - - , - - - -.GISS . UMKO 0.00 Jan Feb Mar Apr May June July Aug Sept Oct Nov Dec month 39 Figure 3.5: Temperature levels projected by the GCMs for India: 2 x CO2 scenario 30 0. Jan Feb Mar Apr May June July Aug Sept Oct Nov Dec month As noted above, the increase in temperature corresponding to a doubling of CO2 in the atmosphere ranges from 2.3 to 4.8 Co. On a monthly basis, the increase in temperature could range from a low of 1 C° in October (GFDL model) to a high of 6 C° in January (UKMO model). It should be noted that all of these figures are likely at the upper range of projected increases; recent studies have shown that the GCMs are probably too high in their general projections of temperature. For comparison purposes, the average monthly temperature for Delhi is depicted as a "base case" (Figure 3.4). The projections are for India as a whole, and weather and climate are very location specific. In addition, differences between the recorded temperatures and the model outputs reflect both the complexity of the atmospheric system and our inability to accurately capture this complexity. PRECIPITATION Figures 3.3, 3.6, 3.7, and 3.8a provide the base data for the GCM precipitation projections. Unlike the temperature projections presented above, the precipitation projections show little consistency across models, even when averaged over the entire country. This is primarily due to the complexity of precipitation patterns and the inability of the GCMs to adequately capture this complexity. The GFDL model projects the greatest annual precipitation increase (and percent increase), with most of this increase occuring during the late summer (July/August/September). On the other hand, the GISS model projects the greatest absolute level of precipitation under both scenarios; over 50% higher than either the UKMO or GFDL models (this is consistent with precipitation projections in other regions of the world, where the GISS model invariably yields the highest amounts). The variability in projections is apparent in 40 Figures 3.7 and 3.8. Additional data on the GCM runs are provided in Figures 3.9b, 3.1Ob and 3.1 lb, and Tables 3.2 - 3.4. The variability in precipitation projections across models is characteristic of the degree of our knowledge about precipitation dynamics and the large regional variation in precipitation that occurs over quite small areas. The models exhibit much higher levels of precipitation during the winter months than occur naturally (Figure 3.6), and somewhat lower levels in the monsoon months. One of the models - the GISS - even seems to project the monsoon at the wrong time of year. Nevertheless, it is important to note that all the models project increases in precipitation. This consistency is somewhat unusual with the GCMs; in regions such as Southeast Asia and the Middle East, the projections are much more variable regarding precipitation. For the purposes of modeling agricultural yields under conditions of climate warming, it is likely best to select one model to use as a base model, and incorporate the projections from that model. As noted previously, the GCMs often differ greatly in their precipitation estimates, and one must be careful to use them as alternative scenarios rather than select a single model output for analysis. Figure 3.6: Precipitation levels projected by the GCMs for India: 1 x CO2 scenario 3__GFDL : 8 - -. - .GISS --.+UMKO -K - Base 6 ---------------- -------- .---- ---- - ----- -------------- >----- --;;:------------------- ---:- 2 ----_----l 4 ;.........i-*X/:- iL 3 - - . .. -------------- -----,.- 2 -- -- H 2: a-_---_-_-----___ tws. ~--' '''X Je - Jan Feb Mar Apr May June July Aug Sept Oct Nov Dec month 41 Figure 3.7: Precipitation levels projected by the GCMs for India: 2 x CO2 scenario 42 Figure 3.8: Projected temperature and precipitation changes for India from a 1 x C02 to a 2 x C02 scenario, by month a Projected Temperature Change 7 05 .i 4- 3 =---- a 2 -------- GFDL 1 , --. -- .- - -** GISS UKMO 0 Jan Feb Mar Apr May June July Aug Sept Oct Nov Dec b Projected Precipitation Change 3.00 2.50 2.00 1.50 E 1. °° t / / -- + 0GFDL 0.50 ~~~~~~~~~~~~~____GISS 0.00 ______UKMO Jan Feb Mar Apr May June July Aug Sept Oct Nov Dec 43 EVAPORATION What influence will these expected temperature and precipitation levels have on water resources in the region? One of the most important will be the impact on evapo-transpiration, which will affect water supply. Although detailed data on surface water availability and use are not readily accessible for India (at least not at a level that would be appropriate for this modelling), rough estimates of the change in evaporation due to climate warming were made to illustrate the potential impact on water availability. Two natural factors may affect the amount of water available for human consumption in the future; precipitation and evaporation. As a rough approximation, changes in evaporation resulting from climate warming were approximated uising a combination of energy-budget and mass transfer approaches based on the Penman cquation (Penman, 1948). The key variable is solar radiation, and the models vary with respect to projections in evapo-transpiration, just as they do with solar radiation. On average, evapo- transpiration will be within 5% of present levels, higher or lower depending upon the particular GCM one is using. SOIL MOISTURE Soil moisture may be an important variable affecting crop growth, along with the distribution of precipitation and the depth of the soil. The GCMs used in this study do not project changes in soil moisture, since this factor is determined not only by precipitation, but by runoff, percolation, evaporation and rainfall distribution. Any attempt at estimating soil moisture at the broad spatial level used by the GCMs would invariably yield erroneous results, and hence projections are not included here. It is important, however, to note that precipitation is expected to increase according to all the models. Depending on the intensity of rainfall, this may also increase soil erosion, and have the effect of reducing agricultural productivity. CONCLUSION The information presented above indicates that the continued emission of trace gases into the earth's atmosphere will likely result in increases in both temperature and precipitation for India. While there will be significant spatial variation in the expected increases, data are presented for the country as a whole. Micro-scale modeling of climate systems is not advanced enough to make reasonable projections at a local scale, and the general projections must suffice. However, for the purposes of this study, the country was arbitrarily divided into "north" and "south" to determine whether there was a significant difference in the projections in these two regions. The results are presented graphically in the Appendix. While agricultural yield is a function of many variables, it is relevant to note that temperature and precipitation in India will likely increase under conditions of global warming, while solar radiation and evapo-transpiration likely will not change appreciably (or, at least, the models are inconsistent in their projections of these variables). Changes in soil moisture are unknown, since it depends on other factors besides the ones projected by the models, including runoff, soil depth and percolation. 44 Figure 3.9: Changes in temperature and precipitation between a 1 x C02 and a 2 x C02 scenario for India, GISS model a Temperature - GISS Model 35E 30 ' a 25'-'o ~'20 co U, 101 5 ~lx C02 -~ - -2x C02 Jan Feb Mar Apr May June July Aug Sept Oct Nov Dec b Precipitation - GISS Model 7600 f\ 6.00 _* 4.00- E 3.00 2.00 1.00 1x CC2 - - - 2x C02 Jan Feb Mar Apr May June July Aug Sept Oct Nov Dec 45 Figure 3.10: Changes in temperature and precipitation between a 1 x C02 and a 2 x C02 scenario for India, GFDL model a Temperature - GFDL Model 35. 30'-- 25A 'A 15' 101 5 -1x C02 - 2x c02 Jan Feb Mar Apr May June July Aug Sept Oct Nov Dec b Precipitation - GFDL Model 8.00 7.00- / 6.00 . 5.00 E E4.00' 3.00 2.00, 1~x C02 1.001 - - - 2x C02 Jan Feb Mar Apr May June July Aug Sept Oct Nov Dec 46 Figure 3.11: Changes in temperature and precipitation between a 1 x C02 and a 2 x C02 scenario for India, UKMO model a Temperature - UKMO Model 35 30' *15 10 0 1x C02 5- -- 2xC02 Jan Feb Mar Apr May June July Aug Sept Oct Nov Dec b Precipitation - UKMO Model 7.00 6.00- 5.001 4.00- E 3.00- 2.00 1.00 1x C02 00 - 2x C02 Jan Feb Mar Apr May June July Aug Sept Oct Nov Dec 47 Table 3.2: Temperature and precipitation projections (GISS model) GISS MODEL Temperature (C) Precip (mm/day) Solar Rad (W/m2) 1x C02 2x C02 Change lx C02 2x C02 Change lx C02 2x C02 Ratio Annual Avg 21.88 25.43 3.55 4.78 5.21 0.43 229.52 231.85 1.010 Jan Avg 16.41 20.27 3.86 4.97 5.81 0.84 150.71 153.71 1.020 Feb Avg 17.67 21.31 3.64 4.44 5.29 0.84 179.86 184.43 1.025 Mar Avg 19.23 22.74 3.51 4.17 4.37 0.20 228.71 227.57 0.995 Apr Avg 21.20 25.09 3.89 4.84 4.51 -0.33 259.71 265.86 1.024 May Avg 23.49 26.34 2.86 5.60 6.69 1.09 273.86 272.86 0.996 June Avg 25.30 28.03 2.73 5.14 5.53 0.39 285.57 284.43 0.996 July Avg 27.20 30.36 3.16 4.81 4.77 -0.04 297.86 294.14 0.988 Aug Avg 27.74 31.07 3.33 4.14 4.24 0.10 290.14 291.86 1.006 Sept Avg 25.76 29.54 3.79 3.56 3.80 0.24 265.43 267.57 1.008 Oct Avg 22.53 26.36 3.83 5.04 5.19 0.14 212.71 219.14 1.030 Nov Avg 19.17 23.16 3.99 5.49 6.44 0.96 164.86 170.00 1.031 Dec Avg 16.84 20.90 4.06 5.14 5.90 0.76 144.86 150.57 1.039 1st Quarter (JFM) 17.77 21.44 3.67 4.53 5.16 0.63 186.43 188.57 1.011 2nd Quarter (AMJ) 23.33 26.49 3.16 5.20 5.58 0.38 273.05 274.38 1.005 3rd Quarter (JAS) 26.90 30.32 3.42 4.17 4.27 0.10 284.48 284.52 1.000 4th Quarter (OND) 19.51 23.47 3.96 5.22 5.84 0.62 174.14 179.90 1.033 Table 3.3: Temperature and precipitation projections (GFDL model) GFDL MODEL Te perature (C) Prcip (mm,dayJ Solar Rad (W/m2) lx C02 2x C02 Change lx C02 2x C02 Change lx C02 2x C02 Ratio Annual Avg. 23.20 25.53 2.33 2.52 3.33 0.81 175.70 170.10 0.968 Jan 15.46 18.56 3.10 2.32 1.73 -0.58 134.75 142.00 1.054 Feb 17.81 21.16 3.35 1.99 1.33 -0.66 160.25 171.92 1.073 Mar 20.85 24.82 3.97 1.28 1.23 -0.04 198.00 200.08 1.011 Apr 25.32 28.90 3.58 1.08 1.04 -0.04 219.33 221.08 1.008 May 26.83 30.43 3.59 2.54 3.22 0.68 208.08 205.67 0.988 June 28.93 30.75 1.83 3.38 5.48 2.09 201.83 183.50 0.909 July 27.93 29.05 1.12 4.88 7.34 2.46 183.17 168.00 0.917 Aug 26.63 27.98 1.35 5.57 7.82 2.25 167.67 157.00 0.936 Sept 26.73 28.58 1.85 2.23 3.79 1.56 181.33 166.08 0.916 Oct 25.12 26.21 1.09 1.02 2.39 1.38 178.08 157.75 0.886 Nov 20.20 21.98 1.78 1.63 2.43 0.79 148.58 138.50 0.932 Dec 16.59 17.92 1.33 2.28 2.15 -0.13 127.33 129.58 1.018 1st Quarter (JFM) 18.04 21.51 3.47 1.86 1.43 -0.43 164.33 171.33 1.043 2nd Quarter (AMJ) 27.03 30.03 3.00 2.34 3.24 0.91 209.75 203.42 0.970 3rd Quarter (JAS) 27.10 28.54 1.44 4.23 6.32 2.09 177.39 163.69 0.923 4th Quarter (OND) 20.64 22.03 1.40 1.64 2.32 0.68 151.33 141.94 0.938 Table 3.4: Temperature and precipitation projections (UKMO model) UKMO MODEL I Tempera ture (C) Precip (mm/day) Solar Rad W/m2 lx C02 2x C02 Change lx C02 2x C02 Change 1x C02 2x C02 Ratio Annual Avg. 20.31 25.10 4.78 2.75 3.24 0.49 244.34 248.62 1.017 Jan Avg 11.17 17.18 6.02 2.08 1.45 -0.63 180.77 194.46 1.076 Feb Avg 13.61 19.42 5.82 1.35 1.19 -0.16 225.77 235.23 1.042 Mar Avg 18.41 23.93 5.52 1.62 1.65 0.04 258.54 266.54 1.031 Apr Avg 21.98 26.52 4.55 2.10 2.85 0.75 281.77 284.46 1.010 May Avg 24.65 29.23 4.58 2.95 3.82 0.88 297.38 298.77 1.005 June Avg 26.72 31.25 4.53 3.08 4.29 1.21 311.46 304.00 0.976 July Avg 28.08 32.04 3.96 2.97 4.58 1.62 288.00 280.08 0.972 Aug Avg 25.90 29.20 3.30 4.48 6.45 1.98 251.31 249.77 0.994 Sept Avg 24.12 27.95 3.83 3.83 4.42 0.59 258.62 260.08 1.006 Oct Avg 20.65 25.63 4.98 3.02 3.59 0.58 233.23 238.62 1.023 Nov Avg 16.01 20.98 4.98 3.04 2.69 -0.35 180.31 192.15 1.066 Dec Avg 12.49 17.84 5.35 2.49 1.92 -0.58 164.92 179.23 1.087 1 st Quarter (JFM) 14.39 20.18 5.78 1.68 1.43 -0.25 221.69 232.08 1.047 2nd Quarter (AMJ) 24.45 29.00 4.55 2.71 3.66 0.95 296.87 295.74 0.996 3rd Quarter (JAS) 26.03 29.73 3.70 3.76 5.15 1.39 265.971 263.31 0.990 4th Quarter (OND) 16.38 21.48 5.10 2.85 2.73 -0.12 192.82 203.33 1.055 APPENDIX: FIGURES AND TABLES OF CLIMATE WARMING, USING A NORTH/SouTH INDIA BREAKDOwN Figure 3A.1: Annual average temperature and precipitation projections for northern India for two GCMs Average Annual Temperature Northern India 30 - 28-- 26-_ 24-- 20 10 GFDL GFDL UKM0 UKMO lx 2x lx 2x C02 C02 C02 C02 Annual Average Precipitation Nothern India 3.00 - 2.50- 2.00 XX 1.50- E 1.00.- 0.50- 0.00 GFDL GFDL UKMO UKMO Ix 2x Ix 2x C02 C02 C02 C02 51 Figure 3A.2: Temperature projections under two scenarios for northern India Temperature -IX C02 Northern India 35 30 ,25 .3~~~~~~~~~~~~~ 5 20 10 ~GFDL S - - UKMO Jan Feb Mar Apr May June July Aug Sept Oct Nov Dec Temperature - 2X C02 Northern India 35- 30 . .25 0t 15 10 GFDL 5 ~~~~~~~~~~--UKMO 5 Jan Feb Mar Apr May June July Aug Sept Oct Nov Dec 52 Figure 3A.3: Precipitation projections under two scenarios for northern India Precipitation - IX C02 Northern India 4.50 4.00 3.50 3.00 2.50 200 - - / 1.50 0.50 - GFDL 0.00 I I I I I I I I I I - - UKMO Jan Feb Mar Apr May June July Aug Sept Oct Nov Dec Precipitation - 2X C02 Northern India 8.00- 7.00 6.00 5.00 ~-4.00 3.00 00, 1 2.00 1.00 ~ ~~~~~~~~~~~~~~~GFDL 1.00 - -uw Jan Feb Mar Apr May June July Aug Sept Oct Nov Dec 53 Figure 3A.4: Projected increases in temperature and precipitation under two GCMs, by month, for northern India Predicted Temperature Increase Northern India 7 o- I I I I I I I I I I 5 .5 2 7D 'j - GFDL 1.50 / 5 - - UKMO Jan Feb Mar Apr May June July Aug Sept Oct Nov Dec Predicted Precipitation Increase Northern India 3.50 3.00 2.50 ~200 p150~~~~~~~~~~~~ --UKMO 0.50 Jan Feb Mar Apr May June July Aug Sept Oct Nov Dec 54 Figure 3A.5: UKMO projections for northern India Temperature - UKMO Model Northern India 35 30 \\0 4.00 - _~~~~~~~0. 25 'A a3.0 - 10 5 1X C02 . -' 2xC02 Jan Feb Mar Apr May June July Aug Sept Oct Nov Dec Precipitation - UKMO Model Northern India 4.00 3.50 3.00 2.50 2.00 1.50 1.00 0.50 2)(~~~~~~~~~~~~~~~~~~~~~ C02 0.50 --2x Jan Feb Mar Apr May June July Aug Sept Oct Nov Dec 55 Figure 3A.6: GFDL projections for northern India Temperature - GFDL Model Northern India 35 10~~~~~~~~ 1Ix C02 --2x C02 15 Jan Feb Mar Apr May June July Aug Sept Oct Nov Dec Precipitation - GFDL Model Northern India 8.00 7.00 6.00 5.00 -4.00 3.00. 2.00 -xC0 Jan Feb Mar Apr May June July Aug Sept Oct Nov Dec 56 Table 3A.1: UKMO model - northern India UKMO MODEL - NORTHERN INDIA _, Temperature (C) Precip (mm/day) Solar Rad (W/m2) lx C02 2x C02 Difference 1x C02 2x C02 Difference Ix C02 2x C02 Ratio Annual Avg. 17.98 23.25 5.28 1.64 2.03 1.28 246.76 251.02 1.036 Jan Avg 5.90 12.31 6.42 1.40 1.20 0.77 174.22 185.56 1.083 Feb Avg 9.17 15.06 5.90 0.92 1.12 1.04 219.44 227.11 1.046 Mar Avg 15.31 21.06 5.74 1.40 1.37 1.01 249.11 258.78 1.072 Apr Avg 19.57 24.73 5.15 1.56 1.94 1.39 280.56 287.89 1.059 May Avg 23.01 28.48 5.47 1.78 2.41 1.48 308.56 312.33 1.022 June Avg 26.23 31.77 5.52 1.71 2.34 1.54 329.11 325.44 0.990 July Avg 28.71 33.19 4.45 1.46 2.83 2.07 306.33 297.67 0.980 Aug Avg 26.06 29.34 3.29 2.60 3.59 1.38 267.67 262.44 0.979 Sept Avg 23.33 27.38 4.06 1.80 2.43 1.30 266.33 262.78 0.994 Oct Avg 18.38 24.00 5.63 1.68 1.94 1.71 232.44 237.78 1.033 Nov Avg 12.28 18.01 5.73 1.71 1.66 0.90 172.78 185.22 1.077 Dec Avg 7.76 13.73 5.98 1.70 1.52 0.80 154.56 169.22 1.077 1st Quarter (JFM) 10.13 16.14 6.02 1.24 1.23 0.94 214.26 223.81 1.067 2nd Quarter (AMJ) 22.94 28.33 5.38 1.68 2.23 1.47 306.07 308.56 1.024 3rd Quarter (JAS) 26.03 29.97 3.93 1.95 2.95 1.58 280.11 274.30 0.984 4th Quarter (OND) 12.80 18.58 5.78 1.70 1.71 1.14 186.59 197.41 1.062 Table 3A.2: GFDL model northern India GFDL MODEL - NORTHERN INDIA l l Temperature (C) Precip (mm/day) Solar Rad (W/m2) lx C02 2x C02 Difference Ix C02 2x C02 Difference lx C02 2x C02 Ratio Annual Avg. 22.25 24.91 2.65 1.97 2.76 1.60 171.99 166.13 0.976 Jan 12.13 15.98 3.84 2.24 1.69 0.73 121.50 127.50 1.046 Feb 15.00 18.89 3.88 2.18 1.14 0.53 144.25 156.75 1.094 Mar 18.39 23.35 4.96 1.11 1.26 1.20 187.63 185.25 0.989 Apr 24.40 28.59 4.18 1.00 1.09 1.20 209.25 210.00 1.005 May 26.74 30.89 4.16 1.99 2.19 0.98 208.25 205.25 0.984 June 29.75 31.88 2.13 1.64 4.39 2.68 214.00 193.13 0.901 July 29.05 29.94 0.90 3.73 6.94 2.68 192.38 173.25 0.900 Aug 27.53 28.54 1.01 4.45 7.15 1.62 176.13 160.13 0.914 Sept 27.35 29.15 1.80 1.29 2.89 3.08 187.13 168.25 0.900 Oct 24.81 26.05 1.23 0.90 1.70 2.60 168.38 154.88 0.923 Nov 18.31 20.38 2.06 1.18 1.48 1.30 139.25 133.63 0.964 Dec 13.58 15.26 1.68 1.98 1.28 0.66 115.75 125.50 1.095 lst Quarter (JFM) 15.17 19.40 4.23 1.84 1.36 0.82 151.13 156.50 1.043 2nd Quarter (AMJ) 26.96 30.45 3.49 1.54 2.55 1.62 210.50 202.79 0.963 3rd Quarter (JAS) 27.98 29.21 1.24 3.15 5.66 2.46 185.21 167.21 0.905 4th Quarter (OND) 18.90 20.56 1.66 1.35 1.48 1.52 141.13 138.00 0.994 Figure 3A.7: Annual average temperature and precipitation projections for southern India for two GCMs Average Annual Temperature Southern India 29 28 " 27 26 4.0 400 25- 24 23 UKMO 1x C02 UKMO 2x C02 GFDL lx C02 GFDL 2x C02 Average Annual Precipitation Southern India 4.50- 4.00 3.50 3.00 2.50 1.50 1.00 0.50 UKMOIxCO2 UKMO 2x C02 GFDLIx C02 GFDL 2x C02 59 Figure 3A.8: Temperature projections under two scenarios for southern India Temperature - 1X C02 Southern India 32 30 28 2a 120 - 16 UKMO 14 - - GFDL 12 10 Jan Feb Mar Apr May June July Aug Sept Oct Nov Dec Temperature - 2X C02 Southern India 35 30 a 2 20 - 15 ' UKMO - - GFDL Jan Feb Mar Apr May June July Aug Sept Oct Nov Dec 60 Figure 3A.9: Precipitation projections under two scenarios for southern India Precipitation - 1X C02 Southem India 7.00 6.00 5.00 . * 4.00- < / ;\^ _ 4.00 3.00 2.00 '' - - GFDL 0.00- Jan Feb Mar Apr May June July Aug Sept Oct Nov Dec Precipitation - 2X C02 Southem India 10.00 900 800 7.00 6.00 4.500 3.00 2.00 - / IJr UKMO 1.CO0_0_ . # - - GFDL 0.00 l l l I l l l l l I Jan Feb Mar Apr May June July Aug Sept Oct Nov Dec 61 Figure 3A.10: Projected increases in temperature and precipitation under two GCMs, by month, for southern India Predicted Temperature Increase Southem India 7 6 5 4 - 2 -. - - ~UKMO 1 .E'~% - ~GFDL 0 I I Jan Feb Mar Apr May June July Aug Sept Oct Nov Dec Predicted Precipitation Increase Southem India 3.00 2.50 2.00 11.50 0.50 - UKMO - - GFDL Jan Feb Mar Apr May June July Aug Sept Oct Nov Dec 62 Figure 3A.11: UKMO projections for southern India Temperature - UKMO Model Southem India 35 30 'a25 h20 15 1x C02 - - 2x C02 Jan Feb Mar Apr May June July Aug Sept Oct Nov Dec Precipitation - UKMO Model Southem India 10.00 9.00f 800 7.00 ~600 0 -5.00 I4.00 F 3.00 l~~~~~~~~~~~~~~~~~x C02 2.00 - xO Jan Feb Mar Apr May June July Aug Sept Oct Nov Dec 63 Figure 3A.12: GFDL projections for southern India Temperature - GFDL Model Southem India 35 30 - ~ -- .125 20 5 21x C02 15- . - 2xC02 Jan Feb Mar Apr May June July Aug Sept Oct Nov Dec Precipitation - GFDL Model Southern India 9 7 S J5 64 Table 3A.3: UKMO model - southern India UKMO MODEL SOUTHERN INDIA Temperature (C_ Precip (mm/day) Solar Rad (W/m2) ______________ 1x C02 2x C02 Difference lx C02 2x C02 Difference Ix C02 2x C02 Ratio Annual Avg. 24.96 28.88 3.91 3.63 4.39 1.29 250.44 250.91 1.013 Jan Avg 17.94 23.99 6.07 2.46 1.49 0.62 196.00 209.13 1.073 Feb Avg 20.04 25.98 5.95 1.61 1.14 0.70 242.63 252.38 1.043 Mar Avg 24.26 29.64 5.36 1.58 1.80 1.22 284.75 286.50 1.006 Apr Avg 27.58 31.03 3.45 2.36 3.75 1.97 298.88 289.75 0.970 MayAvg 29.15 31.95 2.78 4.10 5.26 1.50 293.75 289.00 0.990 June Avg 29.40 32.25 2.84 4.36 6.24 1.59 295.25 280.13 0.950 July Avg 30.00 32.33 2.31 3.94 6.46 2.13 283.25 260.25 0.930 Aug Avg 28.10 30.43 2.34 5.93 9.05 1.64 247.38 244.25 0.995 SeptAvg 27.35 30.28 2.93 5.54 6.30 1.19 259.75 263.88 1.021 Oct Avg 25.11 29.14 4.02 4.35 5.36 1.40 236.75 240.75 1.016 Nov Avg 21.50 25.65 4.14 4.21 3.61 0.87 189.88 202.38 1.069 Dec Avg 19.15 23.93 4.77 3.18 2.26 0.64 177.00 192.50 1.093 1st Quarter (JFM) 20.75 26.53 5.79 1.88 1.48 0.85 241.13 249.33 1.040 2nd Quarter (AMJ) 28.71 31.74 3.02 3.61 5.08 1.69 295.96 286.29 0.970 3rd Quarter (JAS) 28.48 31.01 2.53 5.13 7.27 1.65 263.46 256.13 0.982 4th Quarter (OND) 21.92 26.24 4.31 3.91 3.75 0.97 201.21 211.88 1.059 Table 3A.4: FGDL model - southern India GFDL MODEL - SOUTHERN INDIA l l_ l Temperature (C) Precip (mm/day) Solar Rad (W/m 1 x C02 2x C02 Difference 1x C02 2x C02 Difference lx C02 2x C02 Ratio Annual Avg. 25.50 27.25 1.74 2.77 3.75 1.51 186.61 179.04 0.963 Jan 20.59 22.64 2.07 2.07 1.39 0.63 156.57 166.14 1.069 Feb 22.50 25.37 2.87 1.47 1.33 0.79 185.29 193.71 1.043 Mar 25.29 28.17 2.89 1.26 1.06 0.98 216.00 221.86 1.031 Apr 28.01 30.99 2.98 0.93 0.89 1.08 235.57 237.14 1.004 May 28.29 31.16 2.91 2.91 4.07 1.48 214.00 210.71 0.987 June 29.23 30.24 0.99 4.70 6.44 1.98 197.86 177.14 0.911 July 27.51 28.50 0.97 5.53 7.67 1.66 179.14 162.86 0.921 Aug 26.24 27.61 1.3t 6:30 8.60 1.46 163.71 152.29 0.936 Sept 27.06 28.53 1.48 2.86 4.81 2.15 183.71 166.71 0.916 Oct 26.73 27.27 0.53 1.11 3.00 2.80 194.14 167.00 0.861 Nov 23.51 24.79 1.28 1.84 3.13 1.77 165.86 149.14 0.897 Dec 21.10 21.67 0.56 2.27 2.56 1.32 147.43 149.14 0.983 1st Quarter (JFM) 22.79 25.40 2.61 1.60 1.26 0.80 185.95 193.90 1.048 2nd Quarter (AMJ) 28.51 30.80 2.29 2.85 3.80 1.51 215.81 208.33 0.968 3rd Quarter (JAS) 26.94 28.21 1.27 4.90 7.03 1.76 175.52 160.62 0.924 4th Quarter (OND) 23.78 24.58 0.79 1.74 2.90 1.96 169.14 155.10 0.914 REFERENCES ADB (Asian Development Bank). 1994. "Climate Change in Asia." Manila. Barnola, J.M., D. Raynaud, Y.S. Korotkevich and C. Lorius, 1987. "Vostok Ice Core Provides 160,000-year Record of Atmospheric C02." Nature 329: 408-413. Houghton, J.T., G.J. Jenkins and J.J. Ephraums eds. 1990. Climate Change: The IPCC Scientific Assessment. Cambridge: Cambridge Univ. Press. Manabe, S. and R.T. Wetherald, 1975. The effects of doubling the CO2 Concentration on the Climate of a General Circulation Model. Journal of Atmospheric Science, 32: 3-15. 67 4 THE CLIMATE SENSITIVITY OF INDIAN AGRICULTURE Apurva Sanghi, Robert Mendelsohn and Ariel Dinar INTRODUCTION CLIMATE CHANGE AND AGRICULTURE Since recognition of the potential for future climate change (IPCC, 1990), efforts have been underway to estimate the economic impact of projected changes in climate on important sectors such as agriculture. Most impact studies have focused on agriculture in the developed world (Callway et al., 1982; Adams et al., 1988, U.S. EPA, 1989, Easterling et al., 1993; Adams et al., 1996). Few of these studies, however, fully account for farmer-adaptation to changing climates. Some studies do allow changes in fertilizer application, irrigation, or cultivars, but adaptation is limited at best. A recent study looks at farmer-adapted responses to climate change in the agricultural sector as a whole in the United States (Mendelsohn et al., 1994). Given the inherent global nature of the problem, it is essential to discern the impact of climate change on agriculture in developing countries as well. This is especially important for countries where agriculture is a key component of GDP, and which lie in tropical and sub- tropical zones likely to be adversely affected by climate change (IPCC, 1990b; NAS, 1992). Two such developing countries are Brazil and India, with large land masses, differing climate zones, and varying cropping patterns. Agriculture constituted 30% of 1994 Indian GNP, and 10% of 1994 Brazilian GNP (World Bank, 1996). This paper provides the first detailed estimates of the impact of potential climate change on the agricultural sector in India.' The analysis computes the impacts of changes in temperature and precipitation on agricultural net revenue per cropped hectare. We utilize district-level agricultural, climate, and edaphic data for 271 districts, for the period 1966-1986, to examine farmer-adapted responses to climate variations across the country. A wide variety of crops, including cereals, pulses, oilseeds, fibre crops, and other non-food grain crops such as sugarcane tobacco, are considered. Two APPROACHES Two distinct approaches have emerged in the literature to estimate the impact of climate change on agricultural activity: the traditional "Production Function" approach and, more recently, the "Ricardian" approach. The production function approach takes an underlying production function and varies the relevant environmental input variables to estimate the impact of these inputs on production yield (Callway et al., 1982; Decker et al., 1986; Adams et al., 1988, ' A recent study does examine the impact of climate change on agriculture in both developed and developing countries (Rosenzweig and Parry, 1993). However, the focus is primarily on food grains. Tropical and sub-tropical crops that may benefit from warming are not studied. This exclusion of less heat-sensitive, warmer weather crops will tend to overstate damages. 1990; Adams, 1989; Rind et al., 1990; Rosenzweig and Parry, 1993). The production function approach is attractive in its close collaboration with agronomic science. However, although this approach isolates the impact of environmental change, it does not fully account for the myriad of adaptations that farmers may make in response to varying environmental conditions. Thus, the traditional production function approach may have an inherent bias in that it tends to overestimate the damage from climate change by failing to incorporate economic substitutions by farmers as environmental conditions change. The Ricardian approach, instead of looking at the yields of specific crops, examines how climate in different places affects the net rent or value of farmland, (Mendelsohn et al., 1994). This approach takes into account both the direct impacts of climate on yields of different crops as well as the indirect substitution to other activities, introduction of new land uses, and other potential adaptations to different climates. Part 2 presents the production-function and Ricardian frameworks. Part 3 outlines data sources, definitions, and empirical specifications. Part 4 presents the empirical results and addresses robustness issues. Part 5 analyzes, and interprets in detail, the climate coefficients and the seasonal and regional impacts. Part 6 focuses on implications for potential global warming. Part 7 concludes, and outlines policy implications and future work. THE ECONOMIC FRAMEWORK THE PRODUCTION FUNCTION APPROACH The agronomic production function approach begins with the basic relationship between climate and crop production. Through experiments, agronomists have calibrated models which predict the yield of specific crops depending on weather patterns. These simulation models have historically been used to predict changes in yields for specific crops (Adams et al. 1989, Rosenzweig and Parry 1993). The outcome from these simulations is then incorporated into an economic model of farmer behavior, which in turn leads to a partial equilibrium model of the farm sector. The production function approach is closely linked to agronomic science and hydrological conditions. It is also the only current method capable of including carbon dioxide fertilization (provided appropriate agronomic models) in the analysis.2 The production function model is a structural model which predicts two basic phenomenon: yields (Q1) and crop shares (H.). Based on previous research (e.g., Rosenzweig and Iglesias 1994) the yield per hectare of each crop is likely to be a function of climate, soils, and other inputs. For a set of well-behaved production fimctions, (3) Qi = Qi (K;,E), i =1,..,n 2 Higher concentrations of CO2 in the atmosphere will spur plant growth as carbon is a plant nutrient (Kimball, 1982, Strain and Cure, 1985, Wittwer, 1986). IPCC states that "...studies carried out on small-scale experimental stands of vegetation, under optirnal conditions of water and nutrient supply, suggest potential increases in photosynthesis of 20% to 40% when CO2 is doubled" (IPCC, 1994, p.18). 70 where, K; = [ 1 Ku,... K;J ] is a vector of all purchased inputs in the production of good i; K1.j is the purchased input j ( j = 1,...J ) in the production of good i, and E=[EIE, ...Em] is a vector of exogenous environmental inputs such as temperature, precipitation, and soils which is common to a production site. By comparing the yields in different locations, one can estimate the actual response of yields to climate: (4) Qi = Q; (Ki,E)+uli where Qi is the predicted yield for crop i, and ul; an error term. The mix or the frequency of any crop being planted is the result of the relative profitability of each crop and aggregate constraints such as water availability. The share of a given crop can be modeled as: (5) Hi = Hi (E, Z, Qi) + u2 where u2; is an error term. The fraction of hectares of each crop (Hi) in an area is expected to be a function of climate, market access (Z), and crop yields per hectare. Because Qi includes the error terms uli, the predicted values of Qi should be used in the crop share model. For a given farm, assuming profit maximizing farners, the objective is: (6) maxN = ZHi. P *Q;-Q wij Kij) subject to the physical conditions facing the farm (soil, climate, water). Where N is the per unit of land net income as estimated by the empirical production function approach, H; is the share of crop i, Pi is market price for crop i, Kii is the amount of input j used on crop i, and wj is the unit cost of input j. Thus, by estimating the profit maximizing level of inputs and yields for each crop, one could predict the net revenue from each crop. Combining this information with the predicted crop mix, one could estimate net revenues per hectare in each region. Aggregating net impacts across locations would then yield an aggregate net impact. THE RICARDIAN METHODOLOGY The main drawback of the production function approach is its weakness in modeling the myriad of adjustments that farmers make in response to varying environmental conditions. Farmers are likely to respond to changing climate and other environmental factors by varying, among other things, the crop mix, planting and harvest dates, irrigation scheduling, and application of fertilizers and pesticides, thereby mitigating potential harmful effects of climate change. 71 Consider Figure 4.1 which plots response of more than one crop against temperature. As the figure demonstrates, the optimum range to grow crop A is [T1-T2]. As temperature increases, farmers will substitute crop A for crop B which prefers warmer temperatures (optimal range [T2-T3] ). The same argument carries over to crops B and C: for even hotter temperatures, it will be optimal to switch from crop B to C in the temperature range [T3-T4]. Thus the true sensitivity of crop response to temperature change is the envelope (value function) of the individual crop (structural) responses as is depicted by the solid curve in Figure 4.1. Ignoring adaptation would clearly overstate damages. Figure 4.1: Envelope crop response curves to changes in temperature I Crop Response Crop B Crop C TI T2 T3 T4 > Temperature Adaptations, however, are costly, and these costs are economic damages. However, since adaptations are voluntary, farmers will choose to undertake them only if they are beneficial. Thus, to fully take into account both the costs and benefits of adaptation, the relevant dependent variable should be net revenues or land values (capitalized net revenues), and not yields. The Ricardian approach, therefore, measures damages as reductions in net revenues or land values. Assume a set of well-behaved production functions of the form: (7) Qi = Qi (K; ,E), i =1,..,n where, as before, Ki = [ Kul,... Kij,... Kii ] is a vector of all purchased inputs in the production of good i; Kij is the purchased inputj (j = 1,.. .J ) in the production of good i, and E = [ Ej,...Em,...EM] is a vector of exogenous environmental inputs such as temperature, precipitation, and soils which is common to a production site. Given a set of factor prices wj for Kj, E, and Q, cost minimization leads to a cost function: 72 (8) Ci= Ci( Qi,W E ) where C; is the cost of production of good i and w = [w,,...w1,...w, ] is the vector of factor prices. Assume a set of utility maximizing consumers with well behaved utility functions and linear budget constraints who take prices as given. This leads to a system of inverse demand functions for outputs i =1, ...,n: (9) Pi = D- (Q ,. Qi .Q., 'Y) where Pi and Qi are respectively the price and quantity of good i and Y is the aggregate income. Given market prices, profit maximization on a given site yields: (I0) max P1Qi - Ci-(Qi,w,E)-pL LI Qi where PL is the annual cost or rent of land at that site and Ci() is the cost function of all purchased inputs other than land. Ci() is defined as C1 + PL Li Perfect competition in the land market will drive profits to zero: (11) PiQ* - C*(Qi*,w,E)-pL Li* = 0 If use i is the best use for the land given E and R, the observed market rent on the land will be equal to the annual net profits from the production of good i. Solving (5) for PL gives land rent per hectare to be equal to net revenue per hectare: (12) PL = [PiQ* - C;*(Qi*,w,E)] / Li The present value of the stream of current and future revenues gives land value: (13) VL = JpLe- dt J[PQi -C- (Qi,w,E)]e-r /Lidt 0 0 The issue to be analyzed is the impact of exogenous changes in environmental variables on net economic welfare. Consider an environmental change from the environmental state A to B, which leads environmental inputs to change from EA to E.. The change in annual welfare from this environmental change is given by: 73 QB Q, AW=W(EB)-W(EA)= fZD-' (Qj)dQj-ZCi-(Qi ,W,EA)- fZD-' (Qj)dQj-FCj_(Qj,w,EB) 0 0 If market prices are unchanged as a result of the change in E, then the above equation reduces to: (14) AW= W(EB)-W(EA) = PQB -XCi-(Qi,w,EA) -[PQA -Ci (Q I WEB)] Substituting (12) into (14) gives: (15) AW = W(EB) -W(EA) =(pLBLB -pLALA) where PLA and LA are at EA and PLB and LB are at EB. The present value of this welfare change is thus: (16) fAWe&dt =E(VLBLB - VLALA) 0 1 The Ricardian model takes the form of either (12) or (16) depending on whether the dependent variable is annual net revenues or capitalized net revenues (farm values). The value of the change in the environmental variable is captured exactly by the change in land values across differing environmental conditions. Cross-sectional observations, where normal climate and edaphic factors vary, can hence be utilized to estimate farmer-adapted climate impacts on production and land rents. Due to imperfect land markets and weak documentation of agricultural farm values in India, in this paper we estimate (12), using annual net revenues as the independent variable (Sanghi, 1997 estimates a version of (16) for Brazil using agricultural farm values as the dependent variable). The Ricardian approach is most reliable for estimating impacts of small changes in climate because of the assumption of unchanged relative crop prices. If relative prices change as a result of the impact of climate change on aggregate supply, the Ricardian analysis underestimates or overestimates the impacts, depending on whether the supply of that commodity increases or decreases. However, partial equilibrium welfare estimates provide a reasonable approximation of welfare changes estimated from general equilibrium models (Kokoski and Smith, 1987; Mendelsohn et al., 1996). Though not explicitly modeled in this paper, it is reasonable to assume that, due to moderating effects of international trade, aggregate world supply will not change by much. Therefore, the Ricardian estimates provide reliable measures of changes in economic welfare. Since the analysis makes forecasts based on current fanning practices, it does not capture future changes affecting agriculture such as technical change and carbon dioxide fertilization: potentially increased plant growth from higher concentrations of CO2 Even though a crucial assumption made in CO2 fertilization is that all other inputs are applied at an optimal level, ignoring CO2 fertilization may overestimate damages and underestimate gains. 74 'COMBINING THE Two APPROACHES In principle, estimates from the Ricardian and production function approaches would yield identical estimates of farm net revenue as a function of climate, from which welfare estimates could be made from simulating different climate outcomes. As mentioned earlier, the agronomic production function approach explains differences in yields on the basis of varying environmental conditions. These yield response functions are then incorporated in economic models, giving changes in income. The Ricardian approach directly explains differences in income. However, in practice, due to limitations in fully accounting for adaptation, the production function approach tends to overestimate damages. Nevertheless, the two models can be used to cross check each other, as the empirical yield models by crops can provide some insight into the reduced-form envelope function of the Ricardian approach. We would like to have estimated the underlying production functions for the individual crops (that form the envelope in Figure 4. 1), and then compute changes in net income employing the production function approach as described earlier. Applying the production function approach would, obviously, require estimating the production function with its appropriate inputs. However, no crop-specific inputs are reported in the censuses. This poses a problem in estimating the crop- specific production function.3 Any estimation of crop-specific production functions solely on the basis of climate and edaphic variables will be biased and inaccurate. For this reason, welfare estimates are made utilizing the Ricardian climate coefficients. The production function approach is restricted to demonstrating that climate variables have a significant effect on district-level production yields. These estimates are reported in Appendix A. DEVELOPING COUNTRY MODIFICATIONS Given differences in agricultural sectors in developing and developed countries, methodologies used to date to estimate climate change impacts in developed countries have to be modified. This section highlights some of the major differences between developed and developing countries' agricultural sectors, which are taken into account in the methodology uatilized in this analysis. (i) A key characteristic of developing country agricultural sectors is the large share of non-hired household labor in agriculture. Given the absence of formal markets for self-employed/household labor, ignoring non-hired labor underestimates input costs, and hence overestimates net revenues, the key dependent variable. To account for household/non-hired labor, we use as a proxy the number of cultivators (self- employed males who list their primary job classification as cultivators), and include it on the right hand side of the regression equation. Doing this enables us to compute an implicit shadow wage for self-employed labor. (ii) Prices of crucial inputs are often controlled by developing country governments, as is the case in India. For example, tractor prices do not vary across India for the We tried two procedures to allocate inputs to crops. First, inputs were allocated according to the area planted under Rajasthan Maharashtra Madhya Pradesh Andhra Pradesh Orissa Gujarat Uttar Pradesh Haryana Tamil Nadu Karnataka West Bengal Punjab Temperature (oC) T (D~~~~~~~~~~~N N. 5 WB Bihar , Rajasthan ________ I~~~~~~ 9 Maharashtra -- s . I~~~~~~~~~~~~~~~~~~~~~~~ _ Madhya Pradesh .6 Orissa Gujarat . Uttar Pradesh Haryana _ Tamil Nadu _ Karnataka - I -~~~~~~~~~~~~ L ~ ~ ~~~~~~~~~~~~~ .. . .... ... . . . ........ West Bengal . . - - ... Punjab . ...;......... .. ... .. .... Temperature (oC) Bihar _ ~~~~~~~~~~~3. Rajasthan _ ^ Maharashtra EL Madhya Pradesh Andhra Pradesh Orissa Gujarat Uttar Pradesh Haryana Tamil Nadu Karnataka West Bengal Punjab Temperature (oC) Bihar Rajasthan -w 0 Maharashtra -w O Madhya Pradesh I.i Andhra Pradesh Orissa Gujarat Uttar Pradesh Haryana Tamil Nadu Karnataka West Bengal Punjab Precipitation (mm) Bihar Rajasthan w Maharashtra Madhya Pradesh Andhra Pradesh Orissa Gujarat Uttar Pradesh Haryana Tamil Nadu Karnataka West Bengal Punjab Precipitation (mm) Bihar Rajasthan = Maharashtra Madhya Pradesh Madhya Pradesh Andhra Pradesh Orissa Gujarat Uttar Pradesh Haryana Tamil Nadu Karnataka West Bengal Punjab Precipitation (mm) Bihar Rajasthan Maharashtra i Madhya Pradesh Andhra Pradesh Orissa Gujarat Uttar Pradesh Haryana Tamil Nadu Karnataka West Bengal Punjab Precipitation (mm) m Bihar o m U' . "3~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~00 Rajasthan Maharashtra ..2 . N~~~~~~~~~~~~~~~. Madhya Pradesh v~~~~~~~~~~~~~ Andhra Pradesh Orissa Gujarat Uttar Pradesh Haryana Tamil Nadu Karnataka West Bengal Punjab EDAPHIC AND OTHER CONTROL VARIABLES Edaphic variables (soil slope, type, texture, top soil depth) vary significantly over the districts, and hence need to be controlled for in order to isolate climate from other effects. Appendix D details the edaphic variables. Care is taken not to include soil variables that might be "hidden climate indicators" i.e. variables that may be correlated to temperature and precipitation, such as the Storie Index C, which in part is based on climatic variables. Population density and literacy rates are also included to control for urban influences on agricultural rent. District latitude and mean altitude can be considered as proxies for day length and the diurnal cycle respectively. EMPICAL RESULTS RICARDIAN CLIMATE REGRESSIONS The data is pooled and district level net revenues per hectare are regressed on climate, edaphic, geographic and control variables as defined above to estimate the best-use value function across different districts. Yearly dummies are used to control for year to year variations in weather, prices, and other variables affecting agriculture over the 20 year period. The regression results are presented in Table 4.3. The independent variables include the linear and quadratic temperature and precipitation terms for the four seasonal months; the four corresponding rainfall-temperature interaction terns; edaphic and geographic variables (soil type, slope, and latitude); shadow inputs (bullocks, tractors, and self-employed labor); and other control variables (fraction of area planted under HYV, population density, and literacy proportion). The quadratic and interaction climate terms are included to capture second-order effects of climate on crop yields (and hence net revenues). The signs and magnitudes of the higher order terms determine if the response function is U- shaped or hill- shaped. Each observation is weighted by the area in cropland in each district (acreage weights).9 The linear, quadratic, and interaction climate terms are demeaned. In the preliminary runs, altitude was included to control for the diurnal cycle. However, district-level altitude is highly correlated with district-level October temperature and thus turns out to be a hidden indicator of climate. For this reason, altitude is excluded from the final rWgressions. The soil variables used in the final analysis are district-level soil types and top soils. In earlier runs, all the soil types, and other soil variables such as soil PH and soil slopes were used, but were found to be highly (statistically) insignificant. Soil types and top soils that had coefficients of similar magnitude and signs were combined for ease of handling. 9 The justification for using acreage weights is that the data are at the district, and not at the farn level. Larger districts include more farms, resulting in lower measurement errors. Since the aim is to treat farms (not districts) as having comparable information, larger districts are given a higher weight. 90 Table 4.3: Results of pooled analysis Independent Variable Parameter Independent Variable Parameter Independent Parameter Variable Intercept 4659.3 (Apr rain)x(Apr temp) 8.21 dmyr68 291.85 (8.92) (11.59) (5.37) Jan temperature -132.67 (Jul rain)x(Jul temp) -0.21 dmyr69 288.73 (-3.38) (-1.97) (5.38) April temperature -371.74 (Oct rain)x(Oct temp) 3.01 dmyr70 410.9 (-16.71) (5.83) (7.75) July temperature -103.26 Soil 1 193.55 dmyr7l 365.48 (-2.84) (9.28) (6.93) October temperature 486 Soil 2 221.68 dmyr72 354.38 (7.35) (8.59) (6.74) Jan temperature square -39.25 Soil 3 -153.56 dmyr73 611.14 (-11.4) (-4.39) (11.8) April temperature square 80.34 Soil 4 13.31 dmyr74 562.95 (12.48) (0.41) (10.77) July temperature square 35.02 Soil 5 -10.06 dmyr75 519.47 (4.62) (-0.22) (10.07) October temperature square -68.08 Soil 6 81.32 dmyr76 350.85 (-6.77) (1.56) (6.77) Jan precipitation 18.53 Cultivators/ha -26.95 dmyr77 430.96 (6.11) (-0.78) (8.4) April precipitation -14.38 Bulls/ha 49.73 dmyr78 337.96 (-8) (1.16) (6.61) July precipitation -0.41 Tractors/ha 28681 dmyr80 329.72 (-2.11) (8.98) (6.46) October precipitation 2.28 Population Density 13.55 dmyr81 206.03 (2.23) (2.16) (4.06) January precipitation square -0.16 Literacy 769.94 dmyr82 160.51 (-1.57) (6.85) (3.13) April precipitation square 0.28 Fraction of area under 137.31 dmyr83 307.47 (10.58) HYV (1.87) (6.05) July precipitation square 0.01 Latitude -174.43 dmyr84 155.17 (3.89) (-7.83) (2.96) October precipitation square -0.04 dmyr66 376.62 dmyr85 79.72 (-7.34) (6.77) (1.52) (Jan rain)x(Jan temp) -3.62 dmyr67 541.24 dmyr86 -12.08 (-4.57) (9.92) (-0.23) Note: Dependent Variable: Net revenue per hectare Number of Observations: 5690 Adj R2 =0.44 Weight=Gross cropped area )=t-value 91 INTERPRETATION OF CONTROL VARIABLES AND ROBUSTNESS ISSUES Inclusion of interaction terms in the regression precludes the direct interpretation of the climate coefficients estimated in the regression'". The discussion and interpretation of the climate coefficients is therefore investigated in detail in the next section. This section examines the coefficients on the control variables, and emphasizes some robustness issues. The control variables in the pooled regression behave as expected. The dummy year excluded was for 1979. Net revenues per hectare were the lowest for 1979 relative to other years. The yearly dummies included in the regression capture this effect and are thereby significantly positive. Population density and literacy have a positive impact on net revenues as expected, because of proximity to markets and other urban influences. The coefficient on the number of cultivators, though statistically insignificant, has a negative impact on net revenues. This can be interpreted as farms' constraints on hiring more efficient farm labor especially during peak seasons. The shadow price of both bullocks and tractors is positive, tractors contributing significantly more than bullocks. The coefficient on fraction of area under HYV is positive, reflecting the positive effect on net revenues through increased productivity as a result of the Green Revolution. As a robustness check, the model was estimated using two other procedures. Since we are interested in the long-run impact of climate on net revenues, one approach is to regress averaged district-level net revenues per hectare on the average of the independent variables over the 20 year time period. Another is to estimate the model independently for each year over the 20 year period (1966-1986). Regression coefficients from these procedures alongwith the relevant estimates from the pooled regression in Table 4.3 are presented in Table 4.4. '° If y-a(x-xm)2+b(x-xm)+e, then the coefficient 'b' of the linear term reflects the marginal value of the variable x evaluated at its mean xm, while the quadratic term reflects the change in the marginal effects as one moves away from the mean. However, if y=a(x-xm)2+b(x-xm)+c(x-xm).(z-zm)+e, then the marginal value of x evaluated at its mean is b+c(z-zm). 92 Table 4.4: Comparison of pooled, averaged, and repeated cross-sectional regression results Pool Average 1966 1967 1968 1969 1970 1971 1972 1973 1974 Intercept 4659.3 5386.38 3064.57 6635.49 3310.29 5309.92 6335.23 9761.61 8148.08 9364.17 7700.53 (8.92) (2.8) (1.31) (2.95) (1.53) (2.76) (2.96) (4.22) (3.7) (4.27) (3.4) Jan temperature -132.67 -185.4 -356.83 -361.33 -338.35 -383.58 -350.78 -427.77 -564.03 -253.95 -377.78 (-3.38) (-1.28) (-2.01) (-2.13) (-2.08) (-2.63) (-2.18) (-2.45) (-3.38) (-1.54) (-2.19) April temperature -371.74 -304.76 -36.19 -160.44 -48.31 -111.16 -227.26 -319.49 -79.54 -389.39 -165.12 (-16.71) (-3.61) (-0.35) (-1.6) (-0.5) (-1.28) (-2.35) (-3.05) (-0.81) (-3.94) (-1.6) July temperature -103.26 -123.96 -374.71 -263.84 -349.65 -307.61 -164.82 -80.88 -161.47 -83.21 -196.32 (-2.84) (-0.92) (-2.39) (-1.74) (-2.41) (-2.34) (-1.14) (-0.51) (-1.06) (-0.55) (-1.22) October temperature 486 504.5 718.68 770.08 515.84 568.17 478.89 505.77 370.72 504.6 467.94 (7.35) (2.06) (2.48) (2.77) (1.94) (2.37) (1.79) (1.73) (1.33) (1.82) (1.61) Jan temperature square -39.25 -39.36 -11.07 -37.64 -11.71 -20.68 -27.12 -48.51 -30.04 -69.03 -53.25 (-11.4) (-3.1) (-0.72) (-2.55) (-0.83) (-1.64) (-1.93) (-3.17) (-2.06) (4.79) (-3.49) April temperature 80.34 80.94 67.42 60.81 72.47 84.57 74.39 80.32 86.12 58.62 85.84 square (12.48) (3.41) (2.31) (2.17) (2.74) (3.56) (2.8) (2.79) (3.16) (2.14) (3.05) July temperature square 35.02 42.18 31.68 34.83 35.09 24.48 13.89 37.89 22.49 58.28 74.1 (4.62) (1.51) (0.94) (1.08) (1.13) (0.88) (0.45) (1.12) (0.69) (1.81) (2.21) Octobet lemperature -68.08 -60.4S -59.87 -5732 -37.77 46.87 4.37 -32.64 -63.11 -5138 -103.8 square (-6.77) (-1.63) (-1.31) (-1.32) (-0.91) (-1.26) (-0.11) (-0.72) (-1.47) (.1.19) (-2.34) Jan precipitation 18.53 16.48 18.44 20.9 2.39 13.4 -3.89 -1.7 -5.49 16.63 2.62 (6.11) (1.47) (1.39) (1.64) (0.19) (1.21) (-0.32) (-0.13) (-0.43) (1.3) (0.2) April precipitation -14,38 -16.57 -12.74 -8.31 -13.77 -19.21 -19.52 -19.95 -22.25 -7.57 -14.65 (-8) (-2.49) (-1.59) (-1. 1) (-1.9) (-2.96) (-2.65) (-2.46) (-2.9) (01) (-1.9) July precipitation -0.41 -0.01 0.03 0.34 0.46 -0.01 0.55 0.72 0.52 -1.23 -0.88 /K-2.11) (-0.01) (0.04) (0.4) (0.56) (-0.02) (0.68) (0.82) (0.62) (-1.5) (-1.03) October precipitation 2.28 3.41 9.3 1.36 11.4 7.54 8.38 3.45 12.55 -1.74 3.35 (2.23) (0.9) (1.99) (0.31) (2.66) (1.99) (1.95) (0.73) (2.81) (-0.4) (0.74) January precipitation -0.16 -0.2 0.04 -0.27 -0.29 -0.42 -0.25 -0.36 0.51 -0.38 0.4 square (-1.57) (-0.54) (0.08) (-0.63) (-0.71) (-1.14) (-0.61) (-0.8 1) (1.19) (.0.89) (0.89) April precipitation 0.28 0.32 0.28 0.17 0.34 0.46 0.44 0.33 0.37 0.2 0.23 square (10.58) (3.22) (2.45) (1.48) (3.13) (4.8) (4.04) (2.78) (3.29) (1.73) (2.02) July precipitation 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 square (3.89) (1.33) (1.3) (-0.05) (0.99) (0.61) (1.52) (1.13) (2.43) (0.35) (1.13) October precipitation -0.04 -0.04 -0.01 -0.03 -0.04 -0.05 -0.04 -0.04 -0.05 -0.03 -0.05 square (-7.34) (-2.29) (-0.55) (-I.49) (1.73) (-2.51) (-2.04) (-4.7) (-2.1) (-1.51) (-2.16) (Jan rain)x(Jan temp) -3.62 -3.58 -3.13 -7.92 -3.42 -3.36 -2.28 5.3 0.15 -9.56 -4.65 (-4.57) (-1.23) (-0.89) (-2.37) (-1.07) (-1.17) (-0.7) (-1.5) (0.04) (-2.89) (-1.35) (Apr rain)x(Apr temp) 8.21 8.31 6.91 3.99 9.05 10.31 11.59 7.84 10.83 5.55 6.32 _(11.59) (3.2) (2.2) (1.32) (3.17) (4.02) (4.01) (2.49) (3.59) (1.85) (2.06) Table 4.4 Comparison of pooled, averaged, and repeated cross-sectional regression results (cont.) Pool Average 1966 1967 1968 1969 1970 1971 1972 1973 1974 (Jul rain)x(Jul temp) -0.21 -0.02 0.09 -0.45 0.1 -0.07 0.12 0.05 0.11 -0.57 -0.36 (-1.97) (-0.05) (0.I8) (-0.97) (0.23) (-0.18) (0.26) (0.11) (0.24) (-1.25) (-0.77) (Oct rain)x(Oct temp) 3.01 2.86 0.09 4.46 -0.23 1.42 0.32 4.25 2.88 4.61 6.81 (5.83) (1.51) (0.04) (2.05) (-0.11) (0.76) (0.15) (1.84) (1.3) (2.11) (3.01) Soil 1 193.55 168.83 192.2 188.08 261 222.31 209.09 309.05 263.72 143.34 187.34 __________________ (9.28) (2.2) (2.11) (2.15) (3.14) (2.96) (2.49) (3.35) (3) (1.64) (2.08) Soil 2 221.68 211.28 256.5 250.56 240.55 188.33 292.26 268.03 117.9 173.2 50.55 (8.59) (2.23) (2.23) (2.28) (2.31) (2.01) (2.76) (2.31) (1.06) (1.57) (0.45) Soil 3 -153.56 -115.86 -304.82 -143.43 -120.49 -113.44 -45.41 -29.3 -219.07 -173.48 -285.7 (-4.39) (-0.9) (-2.01) (-0.99) (-0.85) (-0.91) (-0.32) (-0.19) (-1.49) (-1.18) (-1.87) Soil 4 13.31 47.82 -127.55 51.14 23.03 -136.93 -88.79 -236.58 -358.31 -237.59 -247.31 (0.41) (0.4) (-0.85) (0.37) (0.18) (-1.17) (-0.67) (-1.64) (-2.56) (-1.73) (-1.75) Soil 5 -10.06 -7.64 47.15 98.34 138.92 71.49 -116.65 26.83 -55.42 69.69 5.91 (-0.22) (-0.05) (0.23) (0.49) (0.72) (0.42) (-0.62) (0.13) (-0.28) (0.35) (0.03) Soil 6 81.32 70.31 39.17 162.05 87.01 -24.82 -139.95 99.07 49.89 240.32 223.95 (1.56) (0.37) (0.17) (0.72) (0.41) (-0.13) (-0.66) (0.43) (0.22) (1.09) (0.96) Cultivators/hectare -26.95 79.33 546.47 410.41 140 78.1 33.73 166.55 230.85 256.61 338.91 (-0.78) (0.56) (2.75) (2.29) (0.91) (0.62) (0.26) (1.32) (1.8) (1.84) (2.29) Bulls/hectare 49.73 -77.13 416.33 -79.06 125.05 75.13 73.43 89.12 24.74 183.93 -58.93 (1.16) (-0.44) (-1.76) (-0.34) (0.63) (0.45) (0.43) (0.53) (0.17) (1.17) (-0.37) Tractors/hectare 28681 62670 400742 273264 168561 98942 103381 90520 80406 111999 59625 (8.98) (3.44) (4.16) (3.47) (2.9) (2.34) (2.36) (2.15) (2.25) (3.26) (1.92) Population Density 13.55 9.15 2.53 44.93 4.34 30.99 13.66 15.73 7.25 58.4 44.68 (2.16) (0.39) (0.08) (1.44) (-0.14) (1.21) (0.45) (0.48) (0.25) (2.03) (1.58) Literacy 769.94 529.15 305.57 -672.36 250.32 464.41 -93.99 19.56 561.09 -123.79 167.04 (6.85) (1.22) (0.56) (-1.24) (0.48) (1.01) (-0.18) (0.04) (1.06) (-0.24) (0.32) Fraction of area under 137.31 474.81 1018.66 447.68 899.19 314.94 859.14 640.82 364.1 295.99 1308.36 HYV (1.87) (1.25) (0.33) (0.39) (1.05) (0.81) (1.49) (1.17) (0.76) (0.7) (3.45) Latitude -174.43 -195.29 -107 -245.21 -124.81 -203.51 -226.59 -388.01 -320.41 -354.22 -297.46 (-7.83) (-2.38) (-1.07) (-2.55) (-1.35) (-2.47) (-2.47) (-3.91) (-3.39) (-3.77) (-3.06) Adj, R-square 0.44 0.51 0.49 0.52 0.53 0.50 0.48 0.49 0.55 0.54 0.50 Number of Observations 5690 270 270 270 270 270 270 270 270 270 270 Table 4.4: Comparison of pooled, averaged, and repeated cross-sectional regression results (cont.) 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 Intercept 190.96 6784.76 8600.45 2056.72 4533.7 7455.42 4324.81 1699.56 6512.18 4650.96 5052.68 2593.47 (0.08) (2.98) (4.29) (0.99) (2.25) (3.07) (2.05) (0.8) (2.9) (1.79) (1.94) (0.87) Jan temperature 189.32 -33.22 -58.44 395.25 91.95 -246.98 -44.58 72.27 -129.52 -307.03 -364.32 8.29 (1.01) (-0.19) (-0.38) (2.49) (0.6) (-1.38) (-0.28) (0.46) (-0.76) (-1.56) (-1.83) (0.04) April temperature -372.88 -411.2 454.46 -619.19 -500.66 406.78 -313.99 -359.66 -364.25 -370.73 -208.23 -495 (-3.45) (4.08) (-5.16) (-6.77) (-5.5) (-3.89) (-3.43) (-3.94) (-3.7) (-3.33) (-1.85) (-3.92) July temperature 36.26 132.15 150.98 375.66 98.62 -175.82 -100.07 -32.49 75.64 -500.45 413.61 -193.93 (0.21) (0.82) (1.06) (2.54) (0.68) (-1.05) (-0.68) (-0.22) (0.48) (-2.73) (-2.23) (-0.93) October temperature -52.15 316.09 305.14 -29.84 313.58 769.38 535.69 347.65 460.4 1238.26 827.35 610.98 (-0.16) (1.07) (1.17) (-0.11) (1.2) (2.56) (2.01) (1.3) (1.6) (3.72) (2.47) (1.63) Jan temperature square -24.29 -52.18 -73.32 40.2 -49.21 -50.49 -28.33 -26.43 -54.7 -30.71 -30.44 -36.98 (-1.48) (-3.47) (-5.5) (-2.91) (-3.67) (-3.08) (-2.02) (-1.84) (-3.68) (-1.8) (-1.76) (-1.87) April temperature square 57.65 55.5 117.04 67.76 50.12 86.48 65.08 54.95 94.15 98.76 112.57 110.56 (1.89) (1.98) (4.69) (2.6) (1.99) (2.95) (2.52) (2.12) (3.41) (3.1) (3.5) (3.09) July temperature square -31.29 51.41 50.53 29.9 33.25 69.2 32.99 16.73 7.64 87.06 77.45 75.36 (-0.86) (1.53) (1.71) (0.99) (1.13) (1.99) (1.1) (0.55) (0.24) (2.34) (2.08) (1.8) October temperature 49.8 -45.93 -98.71 0.63 -31.94 -100.03 -38.56 -33. I 8 -17.63 -78.45 -147.24 -101.35 square (-1.04) (-1.05) (-2.53) (0.02) (-0.82) (-2.18) (-0.96) (-0.83) (-0.41) (-1.56) (-2.91) (-1.77) Jan precipitation 6.66 19.94 26.5 23.53 11.86 31.92 27.39 23.62 16.49 35.65 37.54 47.77 (0.45) (1.48) (2.22) (1.92) (0.99) (2.31) (2.26) (1.95) (1.27) (2.36) (2.47) (2.81) April precipitation -17.72 -12.32 -6.33 -22.96 -1.92 -23.7 -17.91 -21.26 -13.65 -16.7 -21.95 -31.37 (-2.08) (-1.56) (-0.9) (-3.16) (-0.27) (-2.9) (-2.45) (-2.9) (-1.72) (-1.82) (-2.39) (-3.05) July precipitation -1.62 -0.3 -0.51 0.36 -0.5 0.55 0.55 -0.02 0.59 -0.22 -0.86 0.04 (-1.74) (-0.36) (-0.68) (0.46) (-0.65) (0.63) (0.72) (-0.03) (0.7) (-0.23) (-0.89) (0.04) October precipitation 6.1 -3.97 -5.42 -0.59 -6.68 0.4 -0.44 3.38 5.6 0.33 6.34 8.3 (1.25) (-0.87) (-1.34) (-0.14) (-1.64) (0.09) (-0.11) (0.82) (1.29) (0.07) (1.25) (1.49) January precipitation -0.1 0.43 -0.18 -0.24 -0.06 -0.43 -0.18 -0.61 -0.64 -0.27 0 -0.24 square (-0.2) (0.95) (-0.44) (-0.58) (-0.15) (-0.91) (-0.42) (-1.44) (-1.42) (-0.51) (-0.01) (-0.4) April precipitation square 0.3 0.24 0.29 0.36 0.14 0.35 0.32 0.3 0.39 0.41 0.35 0.4 (2.34) (2.05) (2.77) (3.29) (1.31) (2.86) (2.96) (2.75) (3.25) (3.04) (2.61) (2.7) July precipitation square 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 (-0.66) (0.23) (-055) (1.2) (-0.01) (1.33) (0.68) (1.34) (0.67) (1.95) (2.19) (2.82) October precipitation -0.01 -0.05 -0.03 -0.02 -0.05 -0.04 -0.03 -0.05 -0.08 -0.08 -0.06 -0.04 square (-0.5) (-2.16) (-1.38) (-0.93) (-2.52) (-1.77) (-1.42) (-2.17) (-3.64) (-3.05) (-2.24) (-1.34) (Jan rain)x(Jan temp) -1.67 -2.42 -5.32 0.93 -3.68 -7.82 -2.73 -3.27 4.95 -1.56 -0.59 0.1 (-0.44) (-0.7) (-1.73) (0.29) (-1.19) (-2.14) (-0.86) (-1.01) (-1.45) (-0.39) (-0.15) (0.02) (Apr ra.n)x(Apr temp) 9.44 3.68 9.24 8.47 4.47 7.18 6.96 6.71 11.43 9.12 9.02 12.37 (2.84) (1.19) (3.36) (2.96) (1.6) (2.24) (2.45) (2.33) (3.72) (2.59) (2.53) (3.12) Table 4.4: Comparison of pooled, averaged, and repeated cross-sectional regression results (cont.) 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 (Jul rain)x(Jul temp) -1.33 -0.35 -0.44 0.38 -0.37 0.2 -0.15 -0.1 -0.42 0.43 0.26 1.14 (-2.62) (-0.74) (-1.06) (0.89) (-0.87) (0.41) (-0.36) (-0.23) (-0.9) (0.8) (0.49) (1.89) (Oct rain)x(Oct temp) -1.24 4.84 2.96 -0.07 4.69 3.88 2.43 1.7 2.3 7.19 5.63 -0.21 (-0.51) (2.14) (1.47) ((-0.04) (2.31) (1.66) (1.18) (0.82) (I 04) (2.78) (2.15) (-0.07) Soil 1 86.23 9.13 - 5.44 124.02 1.69 56.19 166.85 210.15 225.71 225.24 209.87 177.1 (0.87) (0.1) (-0.19) (1.47) (0.02) (0.59) (1 96) (2.45) (2.46) (2.11) (1.95) (1.46) Soil 2 232.71 41.41 127.82 72.12 83.05 220.41 181.09 207.17 341.24 178.66 250.92 271.13 (1.87) (0.36) (1.24) (0.69) (0.83) (1.89) (1.76) (2.02) (3.1) (1.41) (1.98) (1.91) Soil 3 -42.86 -107.32 -151.86 -85.35 -178.89 -239.77 -143.78 -94.82 -40.93 -205.47 -159.72 -180.79 (-0.26) (-0.7) (-1.1) (-0.6) (-1.29) (-1.49) (-1.02) (-0.66) (-0.27) (-1.17) (-0.9) (-0.91) Soil 4 82.43 261.72 137.27 114.14 181.99 36.37 95.4 147.11 285.6 244.41 271.74 269.3 (0.53) (1.83) (1.1) (0.87) (1.45) (0.25) (0.74) (1.11) (2.05) (1.5) (1.67) (1.47) Soil 5 202.05 -30.56 -61.57 -32.9 -24.25 -9.29 -29.33 4.45 49.89 -41.42 -200.1 46.01 (0.92) (-0.15) (-0.34) (-0.17) (-0.14) (-0.05) (-0.17) (0.03) (0.27) (-0.18) (-0.86) (-0.17) Soil 6 300.54 175.23 128.36 170.53 48.65 58.48 -16.71 101.62 64.58 89.56 -82.1 43.05 (1.22) (0.76) (0.62) (0.8) (0.25) (0.26) (-0.08) (0.52) (0.3) (0.34) (-0.31) (0.14) Cultivators/hectare -171.81 -132.78 243.58 -37.73 38.81 -36.14 112.71 -97.21 -40.98 -266.96 -146.77 -275.17 _____________________ (-0.95) (-0.8) (1.54) (-0.23) (0.24) (-0.2) (0.69) (-0.62) (-0.25) (-1.64) (-0.93) (-1.72) Bulls/hectare 933.06 -105.19 -51.81 114.27 -265.64 39.24 -215.59 -169.39 -543.36 -390.8 459.34 -364.42 (4.66) (-0.58) (-0.3) (0.61) (-.49) (0.18) (-1.08) (-0.9) (-2.46) (-1.64) (-1.93) (-1.45) Tractors/hectare 19790 59948 59400 28455 39529 68137 53559 35719 28826 27208 26635 42931 (0.61) (2.25) (2.67) (1.42) (2.37) (4.04) (3.98) (3.03) (2.45) (2.46) (2.58) (4.06) Population Density 24.15 44.58 20.43 -22.46 19.71 8.58 -8.63 22.74 -29.5 14.33 4.57 -34.94 _ (0.79) (1.57) (0.84) (-0.8) (0.8) (0.31) (-0.36) (0.95) (-1.22) (0.55) (0.17) (-1.22) Literacy 1219.15 667.11 606.06 1101.6 663.49 914.05 188.98 521.63 685.16 802.45 754.32 893.68 (2.14) (1.28) (1.32) (2.38) (1.52) (1.85) (0.44) (1.23) (1.52) (1.56) (1.49) (1.63) Fraction of area under HYV 830.83 1336.02 966.13 994.4 622.34 -111.39 551.4 199.56 -25.01 -204.87 -166.9 -248.03 (2) (3.57) (3.24) (3.17) (2.04) (-0.33) (1.93) (0.73) (-0.09) (-0.73) (-0.61) (-0.82) Latitude 11.93 -256.49 -324.18 -48.92 -156.27 -284.42 -149.65 -31.55 215.29 -156.36 -169.79 -60.09 (0.11) (-2.64) (-3.77) (-0.55) (-1.81) (-2.74) (-1.67) (-0.35) (-2.24) (-1.41) (-1.53) (-0.47) Adj, R-square 0.40 0.50 0.56 0.54 0.48 0.43 0.46 0.43 0.45 0.46 0.44 0.44 Number of Observations 270 270 270 270 270 270 270 270 270 270 270 270 Comparing the coefficients row by row, it is readily observed that the model is remarkably robust across all three econometric procedures. The climate coefficients retain their signs (though their magnitude varies), with very few exceptions. Some of the climate coefficients change signs in some years. For example, though January temperature coefficients are for the most part negative, they are positive in 1975, 1978, 1979, and 1982. This is most likely due to the effect of weather on annual net revenues. Coefficients on the control variables are also markedly stable. The coefficient on fraction of area under HYV is positive and significant for the 1974-1979 period, the period when the Green Revolution was at its peak. Since the yearly dummies in the pooled regression control for weather and other distortions, we base the impacts in this analysis, on the regression coefficients from the pooled version." SIMULATION OF IMPACTS AND INTERPRETATION OF CLIMATE COEFFICIENTS SIMULATION OF NET REVENUE PER HECTARE As the regression results indicate, climate has a highly non-linear and significant impact on net revenues. The quadratic and interaction climate terms are significant, capturing underlying nonlinearities. As mentioned in the previous section, even though the climate variables are demeaned, the interaction terms make the interpretation of the marginal effects at the mean slightly complicated. We thereby compute total, and partial temperature and precipitation effects by season, for a +2°C rise in temperature and a +7% increase in rainfall. The change in the dependent variable (agricultural net revenues per hectare) is simulated utilizing estimated regression coefficients from the pooled analysis (Table 4.3) for each of the 271 districts for the 1966-1986 period. District level changes in net revenue per hectare are then aggregated to get a measure of the net impact for the country as a whole. The change in net revenue per hectare of a given climate scenario in year y is given by: 271 (18) GWy=E[nrevhay(Td +AT.,,Pd +AP,) nrevhai(Td,Pd)] d=1 where: y=1966. ..1986, (Td, Pd) describes the climate (temperature and precipitation) for district d, (Td + AT,,, Pd + APJ ) describes the new climate under a simulated climate scenario, nrevha Y(Td, Pd) = predicted value of the net revenue per hectare for district d in year y, nrevhay (Td + ATc., Pd + APcs) = forecasted value of net revenues under a climate scenario for district d and year yl2 Yearly changes in the net revenue per hectare are correspondingly averaged over the 20 year period (1966-86) to yield an average net impact. 1 Impacts from the averaged procedure are very similar to those from the pooled one. 12 If the resulting loss from a climate change for any individual district was more than that district's (predicted) net revenue, then the loss was limited to its current net revenue. 97 (b) Seasonal Temperature and Precipitation Effects The change in net revenue per hectare is calculated for the benchmark warming scenario of a +2.0°C rise in mean temperature and a +7% increase in mean precipitation levels for each of the four months (IPCC, 1996).'3 Table 4.5 presents these impacts by season. Table 4.4: Change in net revenue per hectare (1980 Rs.) seasonal temperature and precipitation effects January April July October Total Temperature Effects (+2.0°C) -408.1 -414.1 -66.4 +699.7 -207.9 Precipitation Effects (+7%) +28.6 -13.7 +0.30 -0.80 +14.4 Note: Average net revenue per hectare (in 1980 Rs.)=1424.7 Overall, a rise in temperatures is harmful, reducing net revenue per hectare, whereas an increase in precipitation levels is beneficial, increasing net revenue per hectare. However, net revenues are much more sensitive to temperatures than to precipitation. Consequently, the negative temperature effects exceed the positive precipitation effects, leading to an overall reduction in net revenue per hectare. As observed from the table, there is significant seasonal variation in both temperature and precipitation effects. Seasonal impacts of a +2.0°C rise in temperature are as follows: The change in net revenue per hectare in October, the post-summer harvesting season and the planting season for winter crops, is positive. However, the change in net revenue per hectare is negative in January, the winter growing season; in April, the summer planting season; and in July, the post-monsoon summer growing season. April and January effects are the most negative, and outweigh the beneficial October effect, leading to an overall reduction in net revenues per hectare. October temperature effects may be positive because warmer temperatures during this harvesting season may facilitate the ripening process and ensure optimal crop production. This finding is consistent with the U.S. and Brazilian results (Mendelsohn et al., 1994; Sanghi, 1997). Negative January temperature effects could be the result of heat sensitivity of winter crops (such as wheat and winter maize) to even incremental increases in temperature."4 Furthermore, because of the shorter hibernation period as a result of a warmer winter, there can be an increased incidence of pests and insects such as rice stemborers, leaf and plant hoppers, blast, and tungro. Such infestation has a damaging impact on agricultural activity."5 This finding is also consistent with U.S. and Brazilian results mentioned above. Negative April and July temperature effects 13 The partial effect of, say, January temperatures, is calculated by simulating an increase in January temperature only, holding all else constant in equation (18). 14 This finding is consistent with detailed studies on wheat production in India (Rosenzweig and Parry, 1993; Rao and Sinha, 1994). " A pioneering global study by Cramer (1967) estimates losses due to insects and pests to be as high as 34.4% of potential production before harvest. Way (1976) quantifies actual pest losses to the order of 35% reduction of the total rice crop in India. Later studies report loss estimates of 23.7% and 18.3% of potential production in East and Southeast Asia, and in the Central Luzon region of the Philippines (Ahrens et al., 1 983; Litsinger et al., 1987). 98 are, in all probability, the result of increased heat stress during already hot planting and growing seasons. Although minor relative to temperature effects, there is notable seasonal variation in precipitation effects as well. The January effect is the most positive, reflecting the potential benefit of increased moisture to winter crops. The April effect is the most negative, perhaps because higher rainfall during this planting season could adversely affect seedling establishment and growth. The October effect is mildly negative, indicating that increased rainfall during harvesting is likely to be harmful. The impact of increased precipitation in July is marginal, since most of the country receives ample rainfall during this time. REGIONAL DISTRIBUTION OF SEASONAL IMPACTS Given the broad effects outlined above, the regional variations in temperature and precipitation impacts in each season are now discussed. Figures 4.11-4.14 exhibits the change in net revenue per hectare regionally and seasonally from a uniform increase of +2.0°C temperature across the four months. As seen from the figure, although the temperature increase is uniform, the spatial distribution of impacts is not. In the (winter) month of January, the relatively warm Southern-Central peninsula is the most negatively impacted from the temperature increase. Cooler northern districts of Punjab, Haryana, and Western Uttar Pradesh benefit from warming. This could be because these regions can shift to higher (warmer) value activities, which the Southern-Central peninsula may not be able to, since it is much warmer. April effects are the most negative for the cooler regions of Punjab, Haryana, and western Uttar Pradesh, the main temperate wheat growing area of India. This probably reflects the affinity of wheat for cooler temperatures. A counter-intuitive result is that in April, the hottest districts (in the Central-Southern peninsula) show a gain. This could be because these are the regions that primarily grow jowar and bajra, which are both heat loving crops. However, since these are low value activities (compared to rice, wheat, and maize), the overall loss in net revenue is still negative from an increase in April temperatures. The distribution of July temperature effects is mostly neutral. The distribution of October temperature effects is almost uniformly beneficial, since a warmer harvesting season is expected to facilitate expeditious (summer) crop harvesting. The regional distribution of the partial temperature (+2.0°C ) and precipitation (+7%) effects are portrayed in Figures 4.11-4.14. The combined net impact (+2.0°C and +7%P) is portrayed in Figure 4.15. From a comparison of Figures 4.11-4.14 and 4.15-4.16, it can be seen that the minor precipitation effects do not alter the spatial patterns of the overall regional impacts. Therefore, overall regional impacts shown in Figures 15-16 are mainly a function of the regional temperature effects shown in Figures 4.11-4.14. 99 Change in Net Revenue/Ha (1980 Rs.) . I9 Bihar M. Rajasthan. Maharashtra Madhya Pradesh rA Andhra Pradesh Orissa, _ Gujarat Uttar Pradesh Haryana Tamil Nadu Karnataka West Bengal Punjab Change in Net Revenue/Ha (1980 Rs.) 2 I 1. ~~I I I . Bihar Rajasthan Maharashtra E Madhya Pradesh Andhra Pradesh Orissa Gujarat Uttar Pradesh Haryana Tamil Nadu Kamataka West Bengal Punjab _ Change in Net RevenuelHa (1980 Rs.) Bihar a Rajasthan 2 Maharashtra Madhya Pradesh Andhra Pradesho orissa 4. G3ujarat . Uftar Pradesh Haryana Tamil Nadu Karnataka_ West Bengal Punjab ,Change in Net Revenue/Ha (1980 Rs.) Bihar Rajasthan Maharashtra Madhya Pradesh - Andhra Pradesh + Orissa WN~~~~~~~~~~~~~~~~~r Gujarat Uttar Pradesh Haryana Tamil Nadu Karnataka West Bengal Punjab Change in Not RevenuelHa (1980 Rs.) Bihar Rajasthan Maharashtra Madhya Pradesh 0 Andhra Pradesh Orissa Gujarat Uttar Pradesh Haryana Tamil Nadu Karnataka West Bengal Punjab Change in Net Revenue/Ha (1980 Rs.) .~~ . ~I I S Bihar Rajasthan _ . Maharashtra Madhya Pradesh . IX, Andhra Pradesh |i Orissa j Gujarat Uttar Pradesh Haryana Tamil Nadu Karnataka West Bengal Punjab Change in Net RevenuelHa (1980 Rs.) t Bihar Rajasthan Maharashtra Madhya Pradesh h Andhra Pradesh Orissa _ 00 Gujarat Uttar Pradesh Haryana Tamil Nadu Karnataka West Bengal Punjab OVERALL REGIONAL IMPACTS Displaying the spatial temperature effects provides useful insights into overall regional impacts displayed in Figures 4.15 and 4.16. Aggregating the temperature effects in Figures 4.11- 4.14, it is readily observed that the coastal and inland regions of Gujurat, Maharashtra, and Karnataka are most adversely affected as the harmful January and April effects overcome the beneficial October and (mildly positive) July effects. The high value wheat (winter crop) growing regions of Punjab, Haryana, and Western Uttar Pradesh are also damaged, but not as much as the western districts, because of potential substitution to higher value activities (such as summer rice, for example) in currently cooler regions. The agriculturally low value, hot and dry districts of Rajasthan and Central India are also negatively impacted. However, not all warming is harmful. Eastern districts of Andhra Pradesh, Orissa and West Bengal benefit mildly from warming. Overall, the net impact for the country is negative. IMPLICATIONS FOR GLOBAL WARMING The above findings have implications for potential impact of global warming on Indian agriculture. As described in Section 5, the regression coefficients are applied to forecast changes in aggregate net revenues (netrevenueperhectare totalcroppedarea) for selected climate scenarios. Given the inherent uncertainty in predicting precise country forecasts, we consider a range of climate scenarios, including IPCC's best-guess estimate of a 2.0°C rise in temperature and a 7% increase in rainfall (IPCC, 1996). Although forecasts for India employing General Circulation Models (GCMs) were made for this project, they are believed to be too high in their general projections, especially for rainfall (Lonergan, 1998). For a benchmark doubling of carbon from pre-industrial levels (expected to occur sometime in the next century), these models predict as much as a +288% increase in precipitation levels for the month of October in Southern India, and as high as a +6.420C rise in winter temperatures for Northern India. One GCM even predicts the monsoon at the wrong time of the year. For these reasons, GCM forecasts are not used to compute impacts. Though the impacts from uniform warming scenarios shown below include all farmer adaptations, they do not consider (damage reducing) CO2 fertilization effects. In order to express the impact on the 1994 Indian economy in 1996 US dollars, net impacts were converted from 1980 Rupees to 1990 U.S. dollars by using official exchange rates (Source: International Financial Statistics Yearbook, International Monetary Fund, Washington D.C.,1996). Net impacts were then converted to 1996 U.S dollars using a GDP deflator (Source: Budget of the United States Government, Historical Tables, Fiscal Year 1998). Table 5.6 shows the impacts for a range of climate scenarios in 1996 U.S. dollars, and as a percentage change in net revenues. For benchmark warming, the losses are a 12.3% reduction in net revenues. The response function with respect to temperature is hill-shaped, with higher temperatures reducing net revenues. Although increased precipitation has a moderating impact on temperature effects, overall losses continue to be dominated by negative temperature effects. Compared to impacts forecasted for India by calibrating U.S. climate response functions (0.31% loss GDP for the (+2.50C, +7%mm) scenario in Mendelsohn, 1996), the losses calculated here are much higher (12.3% reduction in net revenues, the relevant dependent variable, for the (+2.0°C, +7%mm) scenario). This difference makes evident potential errors in using climate 107 response functions from one region to make forecasts about others, given fundamental economic, technological, and ecological differences, including key differences in adaptability of agricultural systems. To quantify these differences, as a future project, we plan to use calibrated response functions from Brazil, India, and the U.S. to estimate impacts for each of these countries from response functions derived from the other two. Table 4.5: Impacts in 1996 billion dollars Impacts +0.00C +1.00C +2.0°C +3.50C +0% 0 -1.06 -1.94 -2.54 (-8.76) (16.03) (-21) +7% +0.13 -0.77 -1.49 -1.91 (+1.07) (-6.36) (-12.30) (-15.78) +14% +0.27 -0.46 -1.03 -1.25 (+2.23) (-3.80) (-8.51) (-10.33) Note: Total net revenue in 1996 billion $ = 12.10. Numbers in parentheses are % change in net revenue. CONCLUSIONS AND POLICY IMPLICATIONS This report provides the first detailed estimates of the impact of climate change on India's agricultural sector. The analysis utilizes the Ricardian methodology that captures farmer adaptation to varying environmental factors. The methodology is modified to take into account key differences between agricultural sectors in developed and developing countries. We analyze data from 271 districts for the period 1966-1986 to examine the farmer- adapted responses to climate variations across the country. A wide variety of crops that include cereals, pulses, oilseeds, fibre crops, and other non-food grain crops such as sugarcane and tobacco are considered. Using a pooled analysis, district-level net revenues per hectare are regressed on climate, edaphic, geographic and control variables to estimate a best-use value function across different districts. Yearly dummies are used to control for annual variations in weather, prices, and other factors affecting agriculture over the 20-year period. The pattern of estimated climate coefficients is remarkably robust and consistent over three econometric specifications, one of which includes twenty independent cross-sections (Table 4.4). Findings indicate that climate change will have a overall negative impact on Indian agriculture, with varying seasonal and regional implications. Applying IPCC's uniform benchmark warming scenario of +2.0°C rise in mean temperature and a +7% increase in mean precipitation levels, we calculate the impact to be a 12.3% reduction in net revenues for the country as a whole (Table 4.6). We also compute separately the effects of the rise in temperature (+2.0°C), as well as the rise in precipitation (+7%). We find that a rise in temperature is damaging, whereas an increase in precipitation levels is beneficial. However, the positive precipitation effect is dwarfed by the negative temperature effect (Figures 4.15-4.16). These temperature and precipitation effects are further broken down into seasonal and regional impacts. Seasonal impacts (Table 4.5) of a +2.0°C rise in temperature are as follows: The change in net revenues per hectare in October, the post-summer harvesting season and the planting season for winter crops, is positive. However, the change in net revenues per hectare is negative 108 in January, the winter growing season; in April, the summer planting season; and in July, the post-monsoon summer growing season. April and January effects are the most negative, and outweigh the beneficial October effect, leading to an overall reduction in net revenues per hectare. October temperature effects may be positive because warmer temperatures during this harvesting season may facilitate the ripening process and ensure optimal crop production. Negative January temperature effects could be the result of heat sensitivity of winter crops (such as wheat and winter maize) to even incremental increases in temperature, as well as potential crop losses from a higher incidence of pest infestations. Negative April and July temperature effects are, in all probability, the result of increased heat stress during already hot planting and growing seasons. Although minor, the seasonal impacts of a 7% increase in precipitation levels (Table 4.5) are as follows: the January effect is the most positive, reflecting the potential benefit of increased moisture to winter crops. The April effect is the most negative, perhaps because higher rainfall during this planting season could adversely affect seedling establishment and growth. The October effect is mildly negative, indicating that increased rainfall during harvesting is likely to be harmful. The impact of increased precipitation in July is marginal, since most of the country receives ample rainfall during this time. There is significant regional variation in the impacts from a +2.0°C rise in mean temperature and a +7% increase in mean precipitation levels (Figure 4.17). Coastal and inland regions of Gujurat, Maharashtra, and Karnataka are most negatively affected. High-value agricultural regions of Punjab, Haryana, and Western Uttar Pradesh show a small loss. The magnitude of the small losses in these regions suggests that agriculture will shift to potentially more valuable summer crops. The agriculturally low value, hot and dry districts of Rajasthan and Central India are negatively impacted. However, not all warming is harmful. Eastern districts of Andhra Pradesh, Orissa and West Bengal benefit mildly from warming. Policy implications arising from the findings in the paper include the need to develop increased heat tolerance in high-value temperature sensitive crops. Furthermore, minimizing run- offs to capture benefits from increased rainfall will be a beneficial strategy, particularly for winter crops. The potential for increased pest infestations as a result of warmer winter climates calls for research into mitigation strategies. Finally, since the potential to substitute to other activities is lower, subsistence farming (not included in this analysis) is likely to be as, if not more, adversely affected as the commercial sector by warming. Thus, appropriate mitigation strategies, such as technical and financial support, may need to be in place. Multilateral efforts to contain and mitigate potential global warming are already underway at the international level. It remains to be seen whether the Framework Convention on Climate Change (FCCC), the international treaty on climate change, will be successful in realizing its objective of getting nations to voluntarily restrict their greenhouse gas emissions at 1990 levels by the turn of the century. While the analysis in this paper may be broadly applicable to some countries in South Asia with comparable agricultural sectors, detailed research on the impact of climate change on agriculture in other developing countries is required. Furthermore, research on other climate sensitive sectors such as energy and forestry will be necessary to unmask more fully-the economic impacts of possible climate change in developing countries. 109 APPENDIX A: PRODUCTION FUNCTION ESTIMATES This Appendix reports the regression results of estimating an (incomplete) production function for the five major Indian crops: rice, wheat, maize, jowar, and bajra. For each crop, three different specifications were tried: linear-quadratic, log-linear, and log-log. Regression results from the specification that gave the best fit are reported below in Tables 4A. 1-4A.5. Observations were pooled for the period 1966--1986 for all the regressions with 1979 as the reference year. As was the case in the net revenues regressions, soil types and top soils that had coefficients of similar magnitude and signs in preliminary runs were combined for ease of handling. The results below indicate that climate has a significant and non-linear impact on crop yields. Table 4A.1: Yield regressions for rice Independent Variable Parameter Independent Variable Parameter Intercept 1.62 dmyr69 0.24 (4.58) (5.83) Jan temperature 0.17 dmyr70 0.34 (17.09) (8.35) April temperature -0.22 dmyr71 0.31 (-25.26) (7.58) July temperature 0.24 dmyr72 0.06 (14.22) (1.35) October temperature -0.21 dmyr73 0.4 l _________________________ (-11.59) (9.82) Jan precipitation -0.01 dmyr74 0.19 (-4.08) (4.53) April precipitation 0.01 dmyr75 0.49 _ (1.55) (12.02) July precipitation -0.01 dmyr76 0.41 (-9.35) (10.05) October precipitation 0.01 dmyr77 0.56 __________________________ _ =(3.85) ____ (13.78) Soil 1 0.17 dmyr78 0.56 (10.77) (13.92) Soil 2 -0.16 dmyr80 0.56 (-5.46) (13.89) Soil 3 -0.04 dmyr81 0.56 (-1.23) (13.74) Soil 4 0.05 dmyr82 0.45 (1.46) (11.08) dmyr66 -0.09 dmyr83 0.67 (-2.05) (16.78) dmyr67 0.19 dmyr84 0.59 __________________________ (4.56) _ (14.36) drnyr68 0.12 dmyr85 0.6 (2.98) (14.8) dmyr86 0.51 1(12.38) Note: Adjusted R square = 0.45 Number of Observations=5390 Dependent Variable=log(yield per hectare-rice) (= t-value l 111 Table 4A.2: Yield regressions for wheat Independent Variable Parameter (x1 05) Independent Parameter Variable (x105) Intercept 694882.1 Soil 5 -19754.5 (2.07) (-5.26) Jan temperature -81876.7 dmyr66 -37135.9 (-9.35) _(-8.5) April temperature 35090.9 dmyr67 -20893.9 (1.97) (-4.82) July temperature 149178.9 dmyr68 -16309.4 (8.36) (-3-74) October temperature 194300 dmyr69 -12820.5 (-5.47) (-2.96) Jan temperature square 1762.1 dmyr7O -600.4 _ _ ~~~~~(8.93) (-0.14) April temperature square -705.3 dmyr7l 4188 (-2.37) _(0.97) _ July temperature square -2853 dmyr72 -4093 (-9.01) (-0.95) October temperature square 4404.1 dmyr73 -9210.5 (6.65) (-2.15) Jan precipitation -109.9 dmyr74 5072.2 (-0.4) (1.17) April precipitation 829.6 dmyr75 17586.5 (4.71) (4.1) July precipitation 242.6 dmyr76 9986.9 (10.91) (2.31) October precipitation -58.9 dmyr77 18958.8 (-0.63) (4.43) January precipitation square -29.6 dmyr78 20992.7 (-5.72) (4.92) April precipitation square -9.06 dmyr8O 13356.9 (-6.37) (3.13) July precipitation square -0.28 dmyr8l 20010.1 (-9.29) (4.72) October precipitation square -1.3 dmyr82 31862.7 (-3.52) (7.47) Soil 1 23374 dmyr83 42895.6 (12.18) (10.19) Soil 2 -7786.6 dmyr84 42226.8 (-4.03) (9.79) Soil 3 -16028.7 dmyr85 49152.4 (-3.8) (11.41) Soil 4 -14778.5 dmyr86 45549.6 (-4.51) (10.53) Note: Adjusted R square = 0.58 Number of Observations=5690 Dependent Variable=log(yield per hectare_wheat) 0= t-va]ue 112 Table 4A.3: Yield regressions for maize Independent Variable Parameter Independent Parameter Variable Intercept 6.62 dmyr69 -0.12 (7.76) (-2.67) log(Jan temperature) 2.71 dmyr7O 0.14 (12.89) (3.13) log(April temperature) -3.67 dmyr7l -0.28 (-11.65) (-6.17) log(July temperature) 5.18 dmyr72 -0.02 (11.15) (-0.45) log(October temperature) -5.86 dmyr73 -0.1 (-11.4) (-2.12) log(Jan precipitation) -0.16 dmyr74 -0.16 (-13.51) (-3.48) log(April precipitation) 0.04 dmyr75 0.1 (4.13) (2.21) log(July precipitation) 0.01 dmyr76 0.08 (2.32) (1.65) log(October precipitation) -0.02 dmyr77 0.01 (-0.94) (0.19) Soil I -0.01 dmyr78 0.08 (-0.16) (1.86) Soil 2 -0.04 dmyr80 0.08 (-1.3) (1.76) Soil 3 -0.08 dmyr8l 0.17 (-1.47) (3.8) Soil 4 0.02 diyr82 0.16 ___________________ (0.4) (3.49) dmyr66 -0.09 dmyr83 0.31 ==___________________ (-1.97) (6.96) dmyr67 -0.01 dmyr84 0.31 (-0.21) _ _ (6.92) dmyr68 -0.18 dmyr85 0.17 (-3.84) (3.66) dmyr86 0.16 (3.55) Note: Adjusted R square = 0.25 Number of Observations=5138 Dependent Variable=log(yield per hectare_maize) ( t-value 113 Table 4A.4: Yield regressions for jowar Independent Variable Parameter Independent Variable Parameter Intercept -2.45 dmyr69 -0.01 (-4.61) (-0.23) Jan temperature 0.09 dmyr70 0.03 (6.11) (0.51) April temperature 0.06 dmyr71 -0.14 (4.51) (-2.44) July temperature 0.01 dmyr72 -0.10 (0.03) (-1.78) October temperature -0.08 dmyr73 0.08 (-3.43) (1.51) Jan precipitation 0.02 dmyr74 0.07 (9.67) (1.27) April precipitation 0.01 dmyr75 0.11 (3.24) (2.04) July precipitation 0.01 dmyr76 0.17 (8.44) (3.02) October precipitation -0.01 dmyr77 0.22 (-0.35) (3.99) Soil 1 0.01 dmyr78 0.23 (0.36) (4.1) Soil 2 -0.19 dmyr80 0.12 (-8.02) (2.08) Soil 3 -0.59 dmyr8l 0.25 (-11.62) (4.55) Soil 4 -0.15 dmyr82 0.14 (-3.83) (2.51) Soil 5 -0.11 dmyr83 0.37 (-2.34) (6.73) dmyr66 -0.10 dmyr84 0.41 (-1.79) (7.27) dmyr67 0.01 dmyr85 0.16 (0.25) (2.89) dmyr68 -0.10 dmyr86 0.15 (-1.81) (2.68) Note: Adjusted R square = 0.29 Number of Observations=4663 Dependent Variable=log(yield per hectarejowar) (= t-value 114 Table 4A.5: Yield regressions for bajra Independent Variable Parameter Independent Parameter Variable _ Intercept -13.78 Soil 5 -0.01 (-4.1) (-0.06) Jan temperature 0.27 dmyr66 0.04 (2.7) (0.78) April temperature -0.58 dmyr67 0.03 (-2.81) (0.48) July temperature 0.77 dmyr68 -0.01 (4.14) (-0.07) October temperature 0.79 dmyr69 0.08 (1.85) (1.48) Jan temperature square 0.01 dmyr70 0.29 (-1.93) (5.48) April temperature square 0.01 dmyr71 0.04 (2.16) (0.72) July temperature square -0.01 dmyr72 0.03 (-3.92) (0.48) October temperature square -0.02 dmyr73 0.12 (-1.95) _ (2.24) Jan precipitation 0.02 dmyr74 0.04 (5.3) (0.74) April precipitation -0.01 dmyr75 0.12 (-2.45) (2.27) July precipitation -0.01 dmyr76 0.16 (-3.41) (3.09) October precipitation -0.01 dmyr77 0.12 (-4.23) (2.29) January precipitation square -0.01 dmyr78 0.12 (-1.11) (2.26) April precipitation square -0.01 dmyr80 0.10 (-1.01) (1.81) July precipitation square 0.01 dmyr81 0.10 (2.38) (1.86) October precipitation square 0.01 dmyr82 0.07 (3.96) (1.32) Soil 1 0.01 dmyr83 0.18 (0.41) (3.37) Soil 2 -0.05 dmyr84 0.17 (-1.69) (3.27) Soil 3 -0.05 dmyr85 0.08 (-1.05) (1.46) Soil 4 -0.15 dmyr86 0.13 (-6.31) (2.39) Note: Adjusted R square = 0.10 Number of Observations=5690 Dependent Variable=yield per hectare_bajra (= t-value 115 APPENDix B: VARIABLES IN THE ORIGINAL DATA SET The original data set was created by James McKinsey and Robert Evenson between 1980 and 1990, and has been used in numerous studies of production and productivity in Indian agriculture. The data set contains observations for each of the variables for the agricultural years 1957/58 through 1987/87. The agricultural year 1957/58 is denoted by 1957. Unless noted, all variables are expressed as annual flows or average annual stocks or average annual levels. This appendix describes in detail the economic variables used in the study as described in the original data set. The next appendix describes the errors that were discovered and corrected for in the original data set. DATA SOURCE ABBREVIATIONS: API: Agricultural Prices in India. APPCI: Area and Production of Principal Crops in India, GOI. ASI: Agricultural Situation in India. AWI: Agricultural Wages in India. BFS: Bulletin of Food Statistics. BRS: Basic Road Statistics. CSRS: Crop and Season Reports of the various States. CTOI: Climatological Tables of Observatories in India. DCD: Department of Community Development. DES: Directorate of Economics and Statistics. DOM: Directorate of Marketing. EPW: Economic and Political Weekly. FAI: Fertilizer Association of India. FHPPI: Farm Harvest Prices of Principal crops in India. FS: Fertilizer Statistics (published by FAI). GOI: Government of India. IAS: Indian Agricultural Statistics. ICAR: Indian Council of Agriculture and Research. ICRISAT: International Crop Research Institute... ISA: Indian Science Abstracts. JIW: Journal of Income and Wealth. 116 LS: Livestock Census. MOA: Ministry of Agriculture. PSA: Primary Census Abstract (reprinted in SAS). SAS: Statistical Abstracts of India. WB: World Bank. Table 4B.1: Description of variables in the data set Variables Source(s) Available by Interpolated/ Month Year Distiict State Constructed (1) Coverage . State . . District . . Year = . (2) Outputs Ax, Ay ('000 ha) APPCI ('54- No Yes Yes Yes No x=major crops '70); y=minor crops SAS; CSRS; ASI('70s- '80s) Qx, Qy ('000 tons) " " Yes Yes No Px, Py (Rs/quintal) FHPPI .. .. Yes Yes No (DES) Ix ('000 ha) SAS; .. .. Yes Yes No CSRS; ICRISAT HYVx ('000 ha) c " . Yes Yes No (3) Variable Inputs _ RURPOP PSA; No Yes Yes Yes No SAS; Census ('5 1, '61, '71, '81) AGLABOR " .. .. Yes Yes Inter. CULTIVAT c " .. Yes Yes Inter. QLABOR . . ? Yes Const. WAGE AWI (DES) Yes ". Yes Yes No NITRO_TP FS .. .. Yes Yes P205_TP Const. (Rs/ton) K20 TP 117 Table 4B.1: Description of variables in the data set (cont.) Variables Source(S) Available By Interpolated/ Month Year District State Constructed QBULLOCK LS (I &2) 6 Yes Yes Inter. ('56, '6 1, '66, '72, '77); QTRACTOR "4 Yes Yes Inter. PBULLOCK API (DES) " Yes Yes Const. PTRACTOR It Yes Yes Const. (4) Other Inputs Literacy PSA; No Yes Yes Yes Inter. SAS; Census ('51, '61, '71, '81) Population Density PSA; No Yes Yes Yes Inter. SAS; Census ('51, '61, '71, '81) Note: "Intermediate" variables have been omitted from this table. 118 The data set covers 271 districts within thirteen states of India. Table 4B.2 lists the districts and states: Table 4B.2: Districts in the data set District State District State 24 Parganas West Bengal Basti Uttar Pradesh Adilabad Andhra Pradesh Bathinda Punjab Agra Uttar Pradesh Belgaum Kamataka Ahmadnagar Maharashtra Bellary Karnataka Ahmedabad Gujarat Betul Madhya Pradesh Ajmer Rajasthan Bhagalpur Bihar Akola Maharashtra Bhandara Maharashtra Aligarh Uttar Pradesh Bharatpur Rajasthan Allahabad Uttar Pradesh Bharuch Gujarat Alwar Rajasthan Bhavnagar Gujarat Ambala Haryana Bhilwara Rajasthan Amravati Maharashtra Bhind Madhya Pradesh Amreli Gujarat Bid Maharashtra Amritsar Punjab Bidar Karnataka Anantapur Andhra Pradesh Bijapur Karnataka Aurangabad Maharashtra Bijnor Uttar Pradesh Azamgarh Uttar Pradesh Bikaner Rajasthan Bahraich Uttar Pradesh Bilaspur Madhya Pradesh Balaghat Madhya Pradesh Birbhum West Bengal Balangir Orissa Budaun Uttar Pradesh Baleshwar Orissa Bulandshahr Uttar Pradesh Ballia Uttar Pradesh Buldana MahaTashtra Banas-Kantha Gujarat Bundi Rajasthan Banda Uttar Pradesh Champaran Bihar Banglore Kamataka Chandrapur Maharashtra Bankura West Bengal Chengalpattu Tamil Nadu Banswara Rajasthan Chhatapur Madhya Pradesh Barabanki Uttar Pradesh Chhindwara Madhya Pradesh Barddhaman West Bengal Chikmagalur Kamataka Bareilly Uttar Pradesh Chitradurga Kamataka Banner Rajasthan Chittaurgarh Rajasthan Bastar Madhya Pradesh Chittoor Andhra Pradesh (cont.) 119 Table 4B.2: Districts in the data set (cont.) District State District State Churu Rajasthan Hoshangabad Madhya Pradesh Coimbatore Tamil Nadu Hoshiarpur Punjab Cuddapah Andhra Pradesh Hugli West Bengal Cuttack Orissa Hyderabad Andhra Pradesh Dakshin Kannad Karnataka Indore Madhya Pradesh Damnoh Madhya Pradesh Jabalpur Madhya Pradesh Darbhanga Bihar Jaipur Rajasthan Darjiling West Bengal Jaisaltner Rajasthan Datia Madhya Pradesh Jalandhar Punjab Dehradun Uttar Pradesh Jalaun Uttar Pradesh Deoria Uttar Pradesh Jalgaon Maharashtra Dewas Madhya Pradesh Jalor Rajasthan Dhanbad Bihar Jalpaiguri West Bengal Dhar Madhya Pradesh Jamnagar Gujarat Dharwad Karnataka Jaunpur Uttar Pradesh Dhenkanal Orissa Jhabua Madhya Pradesh Dhule Maharashtra Jhalawar Rajasthan Dumka Bihar Jhansi Uttar Pradesh Dungarpur Rajasthan Jhunjhunun Rajasthan Durg Madhya Pradesh Jodhpur Rajasthan East Godavari Andhra Pradesh Junagarh Gujarat East Nimar Madhya Pradesh Kachchh Gujarat Etah Uttar Pradesh Kalahandi Orissa Etawah Uttar Pradesh Kanniyakumari Tamil Nadu Faizabad Uttar Pradesh Kanpur Uttar Pradesh Farrukhabad Uttar Pradesh Kapurthala Punjab Fatehpur Uttar Pradesh Karimnagar Andhra Pradesh Firozpur Punjab Kamnal Haryana Ganganagar Rajasthan Kendujhar Orissa Ganjam Orissa Khammam Andhra Pradesh Gaya Bihar Kheda Gujarat Ghazipur Uttar Pradesh Kheri UKtar Pradesh Gonda Uttar Pradesh Koch-Bihar West Bengal Gorakhpur Uttar Pradesh Kodagu Kamataka Gulbarga Karnataka Kolar Kamataka Guna Madhya Pradesh Kolhapur Maharashtra Guntur Andhra Pradesh Koraput Orissa Gurdaspur Punjab Kota Rajasthan Gurgaon Haryana Krishna Andhra Pradesh Gwalior Madhya Pradesh Kurnool Andhra Pradesh Hamirpur Uttar Pradesh Lucknow UKtar Pradesh Haora Wes Bengal Ludhiana Punjab Hardoi Uttar Pradesh Madurai Tamil Nadu Hassan Karnataka Mahbubnagar Andhra Pradesh Hazaribag Bihar Mahendragarh Haryana (cont.) 120 Table 4B.2: Districts in the data set (cont.) District State District State Hissar Haryana Mahesana Gujarat Hoshangabad Madhya Pradesh Mainpuri Uttar Pradesh Maldah West Bengal Raisen Madhya Pradesh Mandla _ Madhya Pradesh Rajgarh Madhya Pradesh Mandsaur Madhya Pradesh Rajkot Gujarat Mandya Karnataka Ramanathapuram Tamil Nadu Mathura Uttar Pradesh Rampur Uftar Pradesh Mayurbhanj Orissa Ranchi Bihar Medak Andhra Pradesh Ratlan Madhya Pradesh Medinipur West Bengal Ratnagiri Maharashtra Meerut Uttar Pradesh Rewa Madhya Pradesh Mirzapur Uttar Pradesh Rohtak Haryana Moradabad Uttar Pradesh Sabar-Kantha Gujarat Morena Madhya Pradesh Sagar Madhya Pradesh Munger Bihar Saharanpur Uttar Pradesh Murshidabad West Bengal Saharsa Bihar Muzaffarnagar Uttar Pradesh Salem Tamil Nadu Muzzaffarpur Bihar Saambalpur Orissa Mysore Kamataka Sangli Maharashtra Nadia West Bengal Sangrur Punjab Nagaur Rajasthan Saran Bihar Nagpur Maharashtra Satara Maharashtra Nainital Uttar Pradesh Satna Madhya Pradesh Nalgonda Andhra Pradesh Sawai Madhopur Rajasthan Nanded Maharashtra Sehore Madhya Pradesh Narsimhapur Madhya Pradesh Seoni Madhya Pradesh Nashik Maharashtra Shahabad Bihar Nellore Andhra Pradesh Shahdol Madhya Pradesh Nilgiri Tamil Nadu Shahjahanpur Uttar Pradesh Nizamabad Andhra Pradesh Shajapur Madhya Pradesh North Arcot Tamil Nadu Shimoga Karnataka Osmanabad Maharashtra Shivpuri Madhya Pradesh Palamu Bihar Sidhi Madhya Pradesh Pali Rajasthan Sikar Rajasthan Panch-Mahals Gujarat Singhbhum Bihar Panna Madhya Pradesh Sirohi Rajasthan Parbhani Maharashtra Sitapur Uttar Pradesh Patiala Punjab Solapur Maharashtra Patna Bihar South Arcot Tamil Nadu Phulabani Orissa Srikakulam Andhra Pradesh Pilibhit Uttar Pradesh Sultanpur Uttar Pradesh Pratapgarh Uttar Pradesh Sundargarh Orissa Pune Maharashtra Surat Guj arat Puri Orissa Surendranagar Gujarat Purnea Bihar Surguja Madhya Pradesh (cont.) 121 Table 4.B.2: Districts in the data set (cont.) District State Puruliya West Bengal Rae-Bareli Uttar Pradesh Raichur Karnataka Raigarh Madhya Pradesh Raigarh Maharashtra Thane Maharashtra Thanjavur Tamil Nadu The-Dangs Gujarat Tikamgarh Madhya Pradesh Tiruchchirappalli Tamil Nadu Tirunelveli-Kattabo Tamil Nadu Tonk Rajasthan Tumkur Kamataka Udaipur Rajasthan Ujjain Madhya Pradesh Unnao Uttar Pradesh Uttar Kannad Kamataka Vadodara Gujarat Valsad Gujarat Varanasi Uttar Pradesh Vidisha Madhya Pradesh Visakhapatnam Andhra Pradesh Warangal Andhra Pradesh Wardha Maharashtra West Dinajpur West Bengal West Nimar Madhya Pradesh Yavatrnal Maharashtra Definitions Of Variables: (1) Coverage: STATE= 2 digit state code DISTRICT= 2 digit district code YEAR=Years used in the data set (1956/57--1987/88) Kerala and Assarn are the major agricultural states absent from the data set. Also absent, but less important agriculturally, are the minor states and Union Territories in the Northeastern part of India, as well as the far-northern states of Himachal Pradesh and Jammu & Kashmir. The 271 districts and 13 states constitute the three primary northern wheat and northern rice producing states (Haryana, Punjab, Uttar Pradesh), two northwestern bajra-producing states (Gujarat and Rajasthan), three Eastern states (Bihar, Orissa, and West Bengal), and the Semi- Arid Tropics States as specified by ICRISAT. 122 Any changes occurring in district boundaries have been accounted for. Original district boundaries have been preserved by consolidating new districts into their 'parent' districts (which is why the actual number of modem-day districts is considerably larger than 271). These changes have occurred, for example, because of the division of the former Punjab into Punjab, Haryana and Himachal Pradesh; the division of some districts into two or more smaller districts in many states (especially in Bihar); or the transfer of parts of one district to another. Using district latitudes and longitudes, the distance of a district's center to the nearest shoreline was calculated by using Geographical Information Systems software (MAPINFO). (2) Outputs: x= major crops: BAJRA, JOWAR, MAIZE, RICE, WHEAT y=minor crops: BAR (barley), COTN (cotton), GNUT (groundnut), GRAM, JUTE, OPULS (other pulses), POTAT (potato), RAGI, TUR, RMSEED (rapeseed and mustard), SESA (sesame), SOY, SUGAR, SUNFL, TOBAC (tobacco) Ax, Ay=Area Planted ('000 ha) Qx, Qy=Production ('000 tons) Px,Py=Farm Harvest Price (Rs/quintal) Ix=Area irrigated under the crop x ('000 ha) HYVx=Area planted to HYV of crop x ('000 ha) (3) Variable Inputs: AGLABOR=Number of rural males whose primary job classification is agricultural labor CULTIVAT=Number of rural males whose primary job classification is cultivation (note: both AGLABOR and CULTIVAT are stock variables). QLABOR=Wtd. sum of AGLABOR and CULTIVAT=(AGLABOR+CULTIVAT)*(# of days worked in the state by farm workers) WAGE=Wtd. (by month) annual labor cost. Wages of a male ploughman were recorded; if not available then wages of a male field laborer or male "Other Agricultural Labor" were selected. June and August were weighted more heavily than other months because of the high intensity of field work during those months. NITRO-TP; P205-TP; K20-TP=Prices of fertilizers (nitrogen, phosphorus and potassium) in Rs/ton. Prices of fertilizers are strictly controlled by the GOI, so the only cross- section price variation arises from the cost of transportation from the railhead to the field. Prices are based on reported maximum sale prices of common fertilizer compounds adjusted for the proportion of the nutrient present in each compound. QBULLOCK=Number of castrated (male) cattle over the age of 3 years which are used in rural areas for work only. QTRACTOR=Number of four-wheel machines (not tracked or walk behind two-wheeled ones). PBULLOCK=(0.5)*(bullock price). The 0.5 represents the substantial annual flow of expenses entailed in breeding, raising and feeding bullocks, as well as the necessary rate of return on their ownership. PTRACTOR=Average tractor prices (controlling for depreciation) using the prices of Eicher 24-HP tractors and Escort tractors. 123 (4) Other Variables: LITERACY= Proportion of rural males who are classified as literate (defined as "the ability to read and write in any language"). Census enumerators, beginning with the 1971 census, were required to observe each individual's ability to read and write before classifying him or her to be literate. As is true for all census variables, values for the inter-censal years were obtained by linear interpolation. Literacy rates change so slowly and so regularly that this procedure seems amply justified. POPDEN=Population Density was calculated by dividing the population as per the population censuses by the area in each district. Like the LITERACY variable, values for the inter-censal variables were linearly interpolated. 124 APPENDIX C: ERRoRs IN ORIGINAL DATA SET Over the course of the project, various errors were discovered in the original data set and corrected for. Presented below is a list of errors corrected for, to caution researchers who have/are using the original data set. Some of the errors are huge, and can substantially modify the results. For example, the price of sugar in 1980 is entered as Rs.71693.96, whereas the actual price is Rs. 380.86. Various such errors were discovered and amended. Other errors include mistakes in appropriate transformation of units. For example, for 1984 onwards, cotton production is reported in '000 bales but the price of cotton is reported in Rs./ton. In most cases, the errors were rectified by tallying them with the original census variables. In a small number of cases, the errors were approximated by using information from neighboring years or districts. 1) ABARLEY: Area planted under barley for districts in Madhya Pradesh for the years 1966-1983 was missing in the original data set. 2) QBARLEY: Barley production for districts in Madhya Pradesh for the years 1966-1983 was missing in the original data set. 3) APOTATO: Area planted under potato for districts in Madhya Pradesh for the years 1966-1983 was missing in the original data set. 4) QPOTATO: Potato production for districts in Madhya Pradesh for the years 1966-1983 was missing in the original data set. 5) ATUR: With the exception of Bihar and Orissa, area planted under tur is missing for 17 year period (1966-1983) in the original data set. 6) QTUR: With the exception of Bihar and Orissa, production of tur is missing for thel7 year period (1966-1983) in the original data set. 7) PTUR: With the exception of Bihar and Orissa, price of tur is missing for the 17 year period (1966-1983) in the original data set. 8) ATUR: For the 20 year period 1966-1986, the variable ATUR is incorrectly recorded for the state of Orissa 9) QJUTE: For all the districts in the data set, and for all the years, the quantity of jute is recorded is recorded incorrectly. 10) For the year 1966, the following variables have been recorded incorrectly in the following districts: District Variable All districts in Orissa agram Dhar qrice Thane agnut Thane qgnut Banglore qiowar Banglore ajowar Kolar qiowar Farrukhabad apotato Kota arice Jhalawar qrice 125 11) ARICE: For the year 1969 area planted under rice has been recorded incorrectly for Kota. 12) In 1972, the following variables are recorded incorrectly in the original data set. District Variable Ludhiana ptobac Jalor pwheat 13) In 1974, the following variables are recorded incorrectly in the data set. District Variable South Arcot psesamum Banas-Kantha qsesamum 14) In 1976, the following variables are recorded incorrectly in the data set. District Variable Anantpur qrice Bidar other pulses Faizabad qsesamum 15) In 1978, the following variables are recorded incorrectly in the data set: District Variable_ Hoshangabad pcotton Mahendragarh -qjowarJ 16) In 1979, the following variables were recorded incorrectly: District Variable All districts in Gujurat agramn Datia qrice Bhatinda awheat Sangrur awheat Patiala awheat Banas-Kantha abarley Banas-Kantha qbarley Kachchh qbarley Kachchh abarley Mahensa abarley Mahensa qbarley Ahembdabad abarley Ahembdabad qbarley 126 17) In 1980, the following variables were recorded incorrectly: District Variable All districts in Gujurat agram Champaran psugar Adilabad qwheat Chikmagalur abajra Chikmagalur qbajra Banas-Kantha abarley Banas-Kantha qbarley Kachchh abarley Kachchh qbarley Mahensa abarley Mahensa qbarley Ahembdabad abarley Ahembdabad qbarley 18) In 1981, the following variables were recorded incorrectly: District Variable All districts in Gujurat agram All districts in Bihar armseed All districts in Bihar qrmseed Agra agram Banas-Kantha abarley Kachchh abarley Mahensa abarley Ahembdabad abarley Banas-Kantha qbarley Kachchh qbarley Mahensa qbarley Ahembdabad qbarley 19) In 1982, the following variables were recorded incorrectly: District Variable Banas-Kantha abarley Kachchh abarley Mahensa abarley Ahembdabad abarley Banas-Kantha qbarley Kachchh qbarley Mahensa qbarley Ahembdabad qbarley 127 20) In 1983, the following variables were recorded incorrectly: District Variable Banas-Kantha abarley Kachchh abarley Mahensa abarley Ahembdabad abarley Banas-Kantha qbarley Kachchh qbarley Mahensa qbarley Ahembdabad qbarley 21) In 1984, the following variables were recorded incorrectly: District Variable Nellore hyvrice South Arcot 1 qrice all districts jqcotton Li~1 22) QCOTTON: For years >=1984, production of cotton has been recorded incorrectly. 23) In 1985, the price of gram is recorded incorrectly in the original data set for Bankura (West Bengal). 128 APPENDIX D: CLIMATE VARIABLES Climate variables are made available from the Food and Agricultural Organization (FAO). The climate variables come from 160 weather stations well scattered across India (Table 4D. 1). The climatological data includes monthly maximum and minimum temperature, mean daily relative humidity, sunshine hours, windspeed, precipitation and calculated values for reference evapotranspiration and effective rainfall. Table (4D. 1) presents an alphabetical list of weather stations with their geographical coordinates. Table 4D.1: List of weather stations Station District State I Union Territory Altitude Latitude Longitude _ _ _ _ _ _ ___(in) (0,1/) (0,/) Agra Agra Uttar Pradesh 169 27.1 78.02 Ahmadabad Ahmadabad Gujarat 55 23.04 72.38 Ahmadnagar Ahmadnagar Maharashtra 657 19.05 74.48 Ajmer Ajmer Rajasthan 486 26.27 74.37 Akola Akola Maharashtra 282 20.42 77.02 Alibag Raigarh Maharashtra 7 18.38 72.52 Aligarh Aligarh Uttar Pradesh 187 27.53 78.04 Allababad Allahabad Uttar Pradesh 98 25.27 81.44 Ambala Ambala Haryana 272 30.23 76.46 Amini (N/A) (Island) 4 11.07 72.44 Amraoti Amravati Maharashtra 370 20.56 77.47 Amritsar Amritsar Punjab 234 31.38 74.52 Angul Dhenkanal Orissa 139 20.5 85.06 Asansol Barddhaman West Bengal 126 23.41 86.58 Aurangabad Aurangabad Maharashtra 581 19.53 75.2 Bahraich Bahraich Uttar Pradesh 124 27.34 81.36 Balasore Baleshwar Orissa 20 21.31 86.56 Balehonnur Chikmagalur Kamataka 889 13.22 75.27 Bangalore Banglore Kamataka 921 12.58 77.35 Bareilly Bareilly Uttar Pradesh 173 28.22 79.24 Barmer Barmer Rajasthan 194 25.45 71.23 Baroda Vadodara Gujarat 34 22.18 73.15 Belgaum Belgaum Karnataka 753 15.51 74.32 Bellary Bellary Kamataka . 449 15.09 76.51 Berhampore Murshidabad West Bengal 19 24.08 88.16 Bhaunagar (Aero) Bhavnagar Gujarat 11 21.45 72.11 Bhopal (Bairagarh) Sehore Madhya Pradesh 523 23.17 77.21 Bhuj (Rudramata) Kachchh Gujarat 80 23.15 69.4 Bidar Bidar Kamataka 664 17.55 77.32 Bijapur Bijapur Karnataka 594 16.49 75.43 Bikaner Bikaner Rajasthan 224 28 73.18 Bombay Greater Bombay Maharashtra 11 18.54 72.49 Burdwan Barddhaman West Bengal 32 23.14 87.51 Calcutta (Dum Dum) 24 Parganas West Bengal 6 22.39 88.26 Chaibasa Singhbhum Bihar 226 22.33 85.49 129 Table 4D.1: List of weather stations (cont.) Station District State I Union Territory Altitude Latitude Longitude (m) (0,/) (0,/) Chandbali Baleshwar Orissa 6 20.47 86.44 Chandrapur Chandrapur Maharashtra 193 19.58 79.18 Cherrapunji (N/A) Assam 1313 25.15 91.44 Chitradurga Chitradurga Karnataka 733 14.14 76.26 Cochin Ernakulam Kerela 3 9.58 76.14 Coimbatore Coimbatore Tamil Nadu 409 11 78.58 Coonoor Nilgiri Tamil Nadu 1747 11.21 76.48 Cuddalore South Arcot Tamil Nadu 12 11.46 79.46 Cuddapah Cuddapah Andhra Pradesh 130 14.29 78.5 Cuttack Cuttack Orissa 27 20.28 85.56 Daltonganj Palamu Bihar 221 24.03 84.04 Darbhanga Darbhanga Bihar 49 26.1 85.54 Darjeeling Dariling West Bengal 2127 27.03 88.16 Dehra Dun Dehradun Uttar Pradesh 682 30.19 78.02 Dhambad Dhanbad Bihar 257 23.47 86.26 Dhubri (N/A) Assam 35 26.01 89.58 Dibrugarh (N/A) Assarn 110 27.29 95.01 Dohad Panch-Mahals Gujarat 333 22.5 74.16 Dumka Dumka Bihar 149 24.16 87.15 Dwarka Jamnagar Gujarat 11 22.22 69.05 Fatehpor Fatehpur Uttar Pradesh 114 25.56 80.5 Gadag Dharwad Karnataka 650 15.25 75.38 Ganganagar Ganganagar Rajasthan 177 29.55 73.53 Gauhati (N/A) Assam 54 26.06 91.35 Gaya Gaya Bihar 116 24.45 84.57 Gonda Gonda Uttar Pradesh 110 27.08 81.58 Gopalpur Ganjam Orissa 17 19.16 84.53 Gorakhpur Gorakhpur Uttar Pradesh 77 26.45 83.22 Gulbarga Gulbarga Kamataka 458 17.21 76.51 Guna Guna Madhya Pradesh 478 24.39 77.19 Gwalior Gwalior Madhya Pradesh 207 26.14 78.15 Hanamkonda Adilabad Andhra Pradesh 269 19.01 79.34 Hassan Hassan Karnataka 960 13 76.09 Hazaribagh Hazaribag Bihar 611 23.59 85.22 Hissar Hissar Haryana 221 29.1 75.44 Honavar Uttar_Kannad Karnataka 29 14.17 74.27 Hoshangabad Sehore Madhya Pradesh 302 22.46 77.46 Hyderabad Hyderabad Andhra Pradesh 545 17.27 78.28 Indore Indore Madhya Pradesh 567 22.43 75.48 Jabalpur Jabalpur Madhya Pradesh 393 23.12 79.57 Jagdalpur Bastar Madhya Pradesh 553 19.05 82.02 Jaipur (Sanganer) Jaipur Rajasthan 390 26.49 75.48 Jalgaon Jaigaon Maharashtra 201 21.03 75.34 Jalpaiguri Jalpaiguri West Bengal 83 26.32 88.43 Jammu (N/A) Jammu & Kashmir 366 32.4 74.5 Jamnagar (Aero) Jamnagar Gujarat 20 22.27 70.02 Jamshedpur Singhbhum Bihar 129 22.49 86.11 130 Table 4D.1: List of weather stations (cont.) Station District State / Union Territory Altitude Latitude Longitude _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ (in) (0 /1) (0/) Jhalawar Jhalawar Rajasthan 321 24.32 76.1 Jhansi Jhansi Uttar Pradesh 251 25.27 78.35 Jodhpur Jodhpur Rajasthan 224 26.18 73.01 Kakinada East Godavari Andhra Pradesh 8 16.57 82.14 Kalimpong Darjiling West Bengal 1209 27.04 88.28 Kalingapatam Srikakulam Andhra Pradesh 6 18.2 84.08 Kanker Bastar Madhya Pradesh 402 20.16 81.29 Kanpur Air Fid Kanpur Uttar Pradesh 126 26.26 80.22 Khandwa East Nimar Madhya Pradesh 318 21.5 76.22 Kodaikanal Madurai Tamil Nadu 2343 10.14 77.28 Kota Kota Rajasthan 257 25.11 75.51 Kozhikode (Calicut) (N/A) Kerela 5 11.15 75.47 Krishnanagar Nadia West Bengal 15 23.24 88.31 KuMool Kurnool Andhra Pradesh 281 15.5 78.04 Leh (N/A) Jammu & Kashmir 3514 34.09 77.34 Lucknow Lucknow Uttar Pradesh 111 26.52 80.56 Ludhiana Ludhiana Punjab 247 30.56 75.52 Lumding (NIA) Assam 149 25.45 93.11 Machilipatam Krishna Andhra Pradesh 3 16.12 81.09 Madras Chengalpattu Tamil Nadu 16 13 80.11 (Minambakkam) _ _ Madurai Madurai Tamil Nadu 133 9.55 78.07 Mahabaleshwar Satara Maharashtra 1382 17.56 73.4 Mainpuri Mainpuri Uttar Pradesh 157 27.14 79.03 Malda Maldah West Bengal 31 25.02 88.08 Malegaon Nashik Maharashtra 437 20.33 74.32 Mangalore Dakshin Kannad Karnataka 22 12.52 74.51 Marmugao North Goa Goa 62 15.25 73.47 Mercara Kodagu Karnataka 1152 12.25 75.44 Midnapore Medinipur West Bengal 45 22.25 87.19 Minicoy (N/A) (Island) 2 8.18 73 Miraj (Sangli) Sangli Maharashtra 554 16.49 74.41 Motihari Champaran Bihar 66 26.4 84.55 Mount Abu Sirohi Rajasthan 1195 24.36 72.43 Mowgong Hamirpur Uttar Pradesh 229 25.04 79.27 Mukteswar (Kumaon) Almora Uttar Pradesh 2311 29.28 79.39 Mussoorie Tehri Garwahal Uttar Pradesh 2042 30.27 78.05 Mysore Mandya Kamataka 767 12.18 76.42 Nagappattinam Thanjavur Tamil Nadu 9 10.46 79.51 Nellore Nellore Andhra Pradesh 20 14.27 79.59 New Delhi-Safdarjang (NIA) New Delhi 216 28.35 77.12 Nimach Mandsaur Madhya Pradesh 496 24.28 74.54 Nizamabad Nizamabad Andhra Pradesh 381 18.4 78.06 Pachmarhi Hoshangabad Madhya Pradesh 1075 22.28 78.26 Pamban (N/A) (Island) 11 9.16 79.18 131 Table 4D.1: List of weather stations (cont.) Station District State / Union Territory Altitude Latitude (0,) Longitude __________________ _ _ (in) (0,/) Patna Patna Bihar 53 25.37 85.1 Pendra Bilaspur Madhya Pradesh 625 22.46 81.54 Phalodi Jodhpur Rajasthan 234 27.08 72.22 Poona Pune Maharashtra 559 18.32 73.51 Puri Puri Orissa 6 19.48 85.49 Purnea Bhagalpur Bihar 38 25.16 87.28 Raichor Raichor Karnataka 400 16.12 77.21 Raipur Raipur Madhya Pradesh 298 21.14 81.39 Rajkot Rajkot Gujarat 138 22.18 70.47 Ranchi Ranchi Bihar 655 23.26 85.24 Rentachintala Guntur Andhra Pradesh 106 16.33 79.33 Roorkee Saharanpur Uftar Pradesh 274 29.51 77.53 Sabaur Bhagalpur Bihar 37 25.14 87.04 Sagar Sagar Madhya Pradesh 551 23.51 78.45 Sagar Island (NIA) (Island) 3 21.39 88.03 Salem Salem Tamil Nadu 278 11.39 78.1 Sambalpur Sambalpur Orissa 148 21.28 83.58 Satna Satna Madhya Pradesh 317 24.34 80.5 Seoni Seoni Madhya Pradesh 619 22.05 79.33 Shillong (N/A) Meghalaya 1598 25.34 91.53 Sholapur Solapur Maharashtra 479 17.4 75.54 Sibsagar (N/A) Assam 97 26.59 94.38 Silchar (N/A) Assam 29 24.49 92.48 Simla (N/A) Himachal Pradesh 2202 31.06 77.1 Srinagar (N/A) Jammu & Kashmir 1586 34.05 74.5 Surat Surat Gujarat 12 21.12 72.5 Tezpur (NIA) Assam 79 26.37 92.47 Tiruchchirapalli Tiruchchirappalli Tamil Nadu 88 10.46 78.43 Trivandrum Thiruvanenthapuram Kerela 64 8.29 76.57 Umaria Shahdol Madhya Pradesh 459 23.32 80.53 Varanasi (Babatpur) Varanasi Uttar Pradesh 85 25.27 82.52 Vellore North Arcot Ambedkar Tamil Nadu 214 12.55 79.09 Veraval Junagarh Gujarat 8 20.54 70.22 Vishakhapatnam Visakhapatnam Andhra Pradesh 3 17.43 83.14 As with the districts, the distance to sea of these weather stations was calculated by using Geographical Information Systems Software (MAPINFO) for India. For the purposes of this study, normal temperatures (°C) and precipitation (mm) for the months of January, April, July, and October were used to predict district-level climates. The climate variables are 30 year normns (1931-1960). 132 APPENDIX E: EDAPHIC VARIABLES Two types of sources" were used to collect edaphic data: tables and figures. The figures display, by color and pattern, the value of some variable for areas which may be smaller than districts, and which often span district boundaries. Most of the time, districts contain areas of more than one value, such as more than one soil type within the district, or regions of different slope within the district, or areas with different Ph within the district, etc. This was handled in the data set in two distinct ways. 1. In constructing the variables for soil type, the proportion of each district's area under each soil type was estimated. The current data set contains nineteen soil type dummy variables, one for each type. The value of a soil type dummy variable is 1 for a district if that type is one of the two predominant soil types in the district [that is, if that type's proportion is one of the two highest proportions in the district]. This implies, obviously, that a district will have two soil type dummies with the value of 1 (except for the few districts all of whose soil is of one type), and thus that one cannot interpret the dummies' coefficients in the customary way. 2. In constructing all of the other variables derived from "color and pattern" maps [namely, Ph, the depth of aquifers, topsoil depth, and slope] the dominant value (which covered more of the district's area than did any other value) was selected. This approach does not distinguish between districts which are entirely covered by one value [say, a Ph of 7; or a moderately steep slope], on the one hand, as against districts with several values, one of which covers slightly more area than do any others [say, 19% of the soil has a Ph of 5,19% has a Ph of 6, 24% has a Ph of 7, 19% with Ph of 8, and 19% with Ph of 9; or 20% of the district's area is flat, 41% is moderately steep, and 39% is very steep], on the other hand. A possible avenue for further research is to enrich the data set in this way. SOIL TYPE 1. variable names of the form DMSnn: a) for example, DMS02, DMS03, DMS04, DMS 19, DMS20; b) DM denotes a dummy variable; c) S denotes soil type d. nn ranges from 02 to 20, for "traditional" soil types 2. source: visual inspection of soil maps for each State: a) S. P. Raychaudhuri et al, Soils of India (New Delhi: Indian Council of Agricultural Research, 1963); b) soil maps 3. types a. 01 not used b. 02 Laterite c. 03 Red and Yellow d. 04 Shallow Black e. 05 Medium Black f. 06 Deep Black 16 This appendix was compiled by James McKinsey, Stonehill College. We acknowledge his efforts in gathering the soil data as described in this appendix. 133 g. 07 Mixed Red and Black h. 08 Coastal Alluvial i. 09 Deltaic Alluvium j. 10 Calcerous k. 11 Gray Brown 1. 12 Desert m. 13 Tarai n. 14 Black (Karail) o. 15 Saline and Alkaline p. 16 Alluvial River q. 17 Skeletal r. 18 Saline and Deltaic S. 19 Red t. 20 Red and Gravely STORIE INDEX 1. variable names and components a. STRA * measuring the character of the soil profile * values range from 0.65 to 1.00, where a higher value represents a more favorable or more productive rating. b. STRB measuring topography, texture and structure values range from 0.65 to 1.00, where a higher value represents a more favorable or more productive rating. c. STRC * measuring the degree of climatic suitability, salinity, stoniness and the tendency to erode * values range from 0.65 to 1.00, where a higher value represents a more favorable or more productive rating. d. STORIE - product of STRA, STRB and STRC * thus the values of STORIE could range from as low as 0.274625 to 1.00. 1) 0.274625 = 0.653, and thus is the theoretical minimum 2) the actual minimum value in any district is 0. * the combined Storie index is designed as an overall measure of soil productivity 2. source a. K. B. Shome and S. P. Raychaudhuri, Rating of Soils of India Proceedings, b. National Institute of Sciences of India, vol. 26, Part A, 1960 c. method adapted from R. E. Storie, Transactions, Fourth International d. Conference of Soil Science, 1950 134 SOIL FERTILITY STATUS 1. variable names and components a. N i. nitrogen fertility level ii. values include 1 for low, 2 for medium and 3 for high b. P i. phosphorous fertility level ii. values include 1 for low, 2 for medium and 3 for high c. K i. potassium fertility level ii. values include 1 for low, 2 for medium and 3 for high d. twelve fertility class groups i. dummy variables ii. each involves a combination of fertility level of N, P and K iii. the groups, and the levels of N, P and K, are listed in the following table: Variable N Level P Level K Level DMFO 1 Low Low Low DMFO2 Low Low High DMFO3 Low Low Medium DMFO4 Low Medium Low DMFO5 Low Medium High DMFO6 Low Medium Medium DMFO7 Medium Low Low DMFO8 Medium Low Medium DMFO9 Medium Low High DMFIO Medium Medium Low DMF 11 Medium Medium Medium DMF12 Medium Medium High 2. Source a. originally compiled by A. B. Ghosh and Rehanul Hasan, Indian Agricultural Research Institute, New Delhi, based on the results of soil tests carried out, and data provided by, state and regional soil testing laboratories b. data presented by M. Velayuthan and A. B. Ghosh, Proceedings, Fertilizer Association of India National Seminar on Strategies for Achieving Fertilizer Consumption Targets and Improving Fertilizer Use Efficiency, 1981 c. published in various annual editions of Fertilizer Statistics of India, published by the Fertilizer Association of India. 135 SOIL PH 1. source: National Atlas of India, vol. 1, plate 59 2. Variables: a. a series of dummy variables DMPH5: strongly alkali 4.5 150 meters thick 3. [obviously this is not exhaustive, in the sense that major areas of the nation are above no aquifers at all, so for many districts none of these dummy variables has the value of one.] TOPSOIL DEPTH 1. Source: National Atlas of India, vol. 1, Plate 50: "Depth of Soil, All-India" 2. Variables: a. DMTS1: dummy variable = 1 if topsoil is 0 - 25 cm. thick b. DMTS2: dummy variable = 1 if topsoil is 25-50 cm. thick c. DMTS3: dummy variable = 1 if topsoil is 50 - 100 cm. thick d. DMTS4: dummy variable = 1 if topsoil is 100 - 300cm. thick e. DMTS5: dummy variable = 1 if topsoil is > 300 cm. thick 136 REFERENCES Adams, Richard. 1989. "Global Climate Change and Agriculture: An Economic Perspective." American Journal ofAgricultural Economics 71(5), pp.1272-79. Adams, R, B. McCarl, D. Dudek, and D. Glyer. 1988. "Implications of Global Climate Change for Western Agriculture." Western Journal ofAgricultural Economics, 13(2), pp. 348-56. Adams, R., C. Rosenzweig, R. Pearl, J. Ritchie., B. Mc Carl, D. Glyer, B. Curry, J. James, K. Boote, H. Allen, 1990. "Global Climate Change and U.S. Agriculture." Nature, 17 345(6272), pp. 219-24. Adams, R., R. Fleming, C. Change, B. McCarl, and C. Rosenzweig. 1993. "A Reassessment of the Economic Effects of Global Climate Change on U.S. Agriculture." Unpublished. Ahrens, C., Cramer, H. H., Mogk, M. and Peschel, H. 1983. "Economic Impact of Crop Losses." In: Procedings of 10th International Congress of Plant Protection. British Crop Protection Council, Croydon, UK. Pp. 65-73. Callaway, J., F. Cronin, J. Currie, and J. Tawil. 1982. "An Analysis of Methods and Models for Assessing the Direct and Indirect Economic Impacts of CO2 Induced Environmental Changes in the Agricultural Sector of the U.S. Economy," Pacific Northwest. Laboratory (Richland, WA), 1982. Working Paper No. PNL-4384. Chambers, R., and R.E Lopez. 1984. "Equity Formation and the On-farm, Off-farm Labor Choice," University of Maryland, Dept. of Agricultural and Resource Economics, College Park, Md. Processed. Cramer, H. H. 1992. Plant Protection and World Crop Production. Pflanzenschutz-Nachricten. Crop Protection Advisory Department of Farbenfabriken Bayer AG-Leverkusen. 1967. Cline, W.R. The Economics of Global Warming. Institute for International Economics. Washington, D.C. Decker, W., V. Jones, and R. Achtuni. 1986. The Impact of Climate Change from Increased Atmospheric Carbon dioxide on American Agriculture. DOE/NBB-0077. Washington, DC: U.S. Department of Energy. Dinar, A., and D. Zilberman. 1991. "The Economics of Resource-Conservation, Pollution- Reduction Technology Selection: The Case of Irrigation Water." Resources and Energy, 13 323-348. Easterling, W., M. McKenney, N. Rosenberg, and K. Lemon 1991. Processes for identifying regional influences of and responses to increasing atmospheric C02 and climate change- -The MINK project--Report IIB, DOEIRL/01830T-H8. Washington, DC: U.S. Department of Energy. Evenson, Robert E., and Antonio Flavio Dias Avila. 1994. "Total Factor Productivity Growth in Brazilian Agriculture and the Role of Agricultural Research: An Analysis by Sector and Agro-Ecological Zones." Economic Growth Center, Yale University and EMBRAPA, Brasilia, Brazil. 137 Intergovernmental Panel on Climate Change. 1990a Scientific Assessment of Climate Change: Report prepared by for IPCC by Working Group 1. New York: World Meteorological Organization and United Nations Environmental Program. January. Intergovernmental Panel on Climate Change. 1990b. Potential Impacts of Climate Change: Report prepared by for IPCC by Working Group 2. New York: World Meteorological Organization and United Nations Environmental Program. January. Kimball, B. 1982. "Carbon Dioxide and Agricultural Yield." Agronomy Journal, 75: 779-788. Kokoski, M. F., and V. K Smith. 1987. "A General Equilibrium Analysis of Partial Equilibrium Welfare Measures: The Case of Climate Change." American Economic Review. 77:331-41. Litsinger, J.A., Canapi, B.L., Bandong, J.P., dela Cruz, C.G., Apostol, RF, Pantua, P.C., Lumaban, M.D., Alviola A.L., Raymundo, F., Libetario, E.M., Loevinsohn, M.E., and Joshi, R.C. 1987. Rice Crop Loss from Insect Pests in Wetland and Dryland Environments with Emphasis on the Philippines. Entomology Department, International Rice Research Institute, Los Banos, Philippines (mimeographed). Lonergan, Steve. 1998. "Climate Warming and India." Chapter 3 in this report. Mendelsohn, Robert. 1996. "The Market Impacts of Climate Change Across Nations: A Comparison of Effects in the United States, Brazil, and India." Paper presented at the Environmetrics Conference, Sao Paulo, Brazil. July. Mendelsohn, Robert, William Nordhaus, and Dai Gee Shaw. 1994. "The Impact of Global Warming on Agriculture: A Ricardian Analysis." American Economic Review, Vol. 84, Number 4, 88: 753-771. September. Mendelsohn, Robert and William Nordhaus. 1996. "The Impact of Global Warming on Agriculture: Reply." American Economic Review, 1312-1315. December. Myneni, R. B., C. D. Keeling, C. J. Tucker, G. Asrar & R.R Nemani. 1997. "Increased Plant Growth in the Northern High Latitudes from 1981 to 1991". Nature 17 (386) No. 6626, April. National Academy of Sciences. 1992. Policy Implications of Greenhouse Warming: Mitigation, Adaptation, and the Science Base. National Academy Press, Washington D.C. Parry, Martin. 1990. Climate Change and WorldAgriculture. London: Earthscan. Rao, D. G. and S. K. Sinha. 1994. "Impact of Climate Change on simulated Wheat Production in India.". in C. Rosenzweig and A. Iglesias (eds.) Implications of Climate Change for International Agriculture: Crop Modeling Study. U.S. EPA, June. Raval, A., and V. Ramanathan. 1989. "Observational Determination of the Greenhouse Effect." Nature 342: 758-61. July. Rind, D., R. Goldberg, J. Hansen, C. Rosenzweig, and R. Ruedy. 1990. "Potential Evapotransporation and the Likelihood of Future Drought." Journal of Geophysical Research 95 (D7): 9983-10004. 138 Rosen, Sherwin. 1974. "Hedonic Prices and Implicit Markets: Product Differentiation in Pure Competition." Journal of Political Economy, 82 (1): 34-55. Rosenzweig, C., and M. L Parry. "Potential Impact of Climate Chanige on World Food Supply: A summary of Recent International Study." 1993. In Agricultural Dimensions of Global Climate Change ed. H. Kaiser and T. Drennen. Rosenzweig, C., and M. L Parry. 1994. "Potential Impact of Climate Change on World Food Supply." Nature, 367 (6459): 133-38. January. Sanghi, Apurva. 1997. "Global Warming and Climate Sensitivity: Brazilian & Indian Agriculture." Forthcoming Ph.D. dissertation, Department of Economics, University of Chicago. Silva, G. L. S. P., J. R. Vincente, and D. V. Caser. 1993. "Mundanca tecnologicae produtividade do milho e da soja no Brasil." R. Bras. Econ. 47(20): 281-303, April/June. U.S. Environmental Protection Agency (EPA). 1989. "The Potential Effects of Global Climate Change on The United States: Appendix C, Agriculture." Volume 1. May. Way, M. J. 1976. "Entomology and the World Food Situation." Bulletin of the Entomological Society ofAmerica 22: 125-129. Wittwer, S. 1986. "Worldwide Status and History of CO2 Enrichment: An Overview." C02 Enrichment of Greenhouse Crops, pp. 3-15, Florida. 139 I 5 CLIMATE CHANGE IMPACTS ON INDIAN AGRICULTURE: THE RICARDIAN APPROACH K.S.Kavi Kumar and Jyoti Parikh INTRODUCTION The present debate on global climate change problem is centered around the costs of initiating policy actions to control greenhouse gas (GHG) emissions, and the associated benefits of such actions in terms of reduced impacts. Though there has been a lot of research carried out on the costs side of the above debate, the benefits side received relatively less attention, mainly due to the complexity and the uncertainty associated. The knowledge about impacts of climate change becomes very crucial to formulate appropriate abatement policies at international and national levels. Research carried out at global level in this area has revealed that the climate change impacts may not be uniform across countries and there could be some losers and some gainers. According to these studies the countries in the higher latitudes (mostly developed countries) would have relatively less severe impacts due to climate change, than the countries in the lower latitudes (mostly developing countries) (Rosenzweig and Parry, 1994; Reilly, 1994). Such disparity assumes greater significance if one views at the responsibilities of various countries for the climate change problem in the first place. The Northern countries with per capita GHG emissions far above those of the Southern countries, have contributed more towards perpetuating the climate change problem so far, and are likely to get benefited also due to climate change. With such background, it becomes very crucial to understand exactly how the developing countries are going to get affected due to climate change. Research in this area, so far has concentrated mainly on developed countries and the need of the hour is to get reliable estimates for developing countries. With its huge population and relatively high dependence on climate sensitive sector like, agriculture, India provides an ideal case study. The present study focuses on climate change impacts on Indian agriculture, recognizing the significance of the same in India's economy, and its vulnerability to potential changes in climate. In the coming decades, Indian agriculture is bound to face a number of challenges which include the pressure from increasing population, changing scenario of world trade in agriculture, and also the potential changes in the global climate. Some of the characteristics of Indian agriculture are: * this climate sensitive sector still accounts for more than 30% of India's GDP * more than 60% of population, directly or indirectly, depend on agriculture * despite significant progresses made in food grain production, the per capita net availability of food grains still fluctuates between 400 to 500 gms/day, which is way below to the international standards * with a weight of 57% in the consumer price index, the food prices have close link with the inflation and any adverse shock on agriculture could have cumulative effects on the economy. Before outlining the objectives of this study, the methodological background of estimating climate change impacts on agriculture is discussed. CLIMATE CHANGE IMPACTS ON AGRICULTURE - METHODOLOGICAL BACKGROUND Two broad methodologies have evolved so far in the literature for assessing the climate change impacts on agriculture: (a) production function approach, and (b) ricardian approach. The production function approach estimates the changes in yield directly from crop-response models. This approach provides a detailed understanding of the physical, biological and economic responses and adjustments. However, given the site specific nature of the approach, one has to draw broad inferences while aggregating from a relatively few sites and crops to large areas and diverse production systems. Also, using this approach it would become rather difficult to model farm level adaptations, because in theory innumerable number of such adaptation possibilities are available for the farmer to respond to changes in the climate. On the positive side, the production function approach is the only approach which can estimate in a satisfactory manner the potential beneficial effects due to increase in CO2 concentration levels in the atmosphere. There is scientific evidence that CO2 increases plant growth and yields, even at farm level conditions (Senft, 1995). Studies by Kimball (1983), Cure and Acock (1986), Kimball (1985) and Woodward (1993) have indicated that the potential increase in yields of C3 crops (most of cereal crops, excluding Sorghum and Corn) could be 30% following doubling of atmospheric CO2 concentration. A number of national and global level studies have followed the production function approach to estimate the impacts due to climate change. These include, Adams et.al. (1988, 1990), Kane et.al. (1991), Rielly et.al. (1994), Rosenzweig and Iglesias (1994) and Rosenzweig and Parry (1994). For India, Rao and Sinha (1994) have used this approach to assess climate change impacts on wheat yield. More recently, Kumar and Parikh (1996) have estimated the climate change impacts on Rice and Wheat crops and assessed the associated economic and welfare implications using an applied general equilibrium framework. Ricardian approach, developed by Mendelsohn and others (1994), uses the cross-section evidence on current production in various regions, and attempts to draw implications with regard to how the cooler regions could adapt the practices of warm regions if climate becomes warmer. The Ricardian approach, in its original form as developed by Mendelsohn and others (1994), relates cross-sectional climate differences to differences in agricultural land values. However, as used in this study, other variations of this original approach are possible. Ricardian approach provides direct evidence on how profit-maximizing farmers have responded to various climatic conditions. On the positive side, this approach takes the farm-level adaptations into consideration while assessing impacts. However, a major limitation of the approach is that it assumes the prices to be in equilibrium, and in case of large climate change the crop prices could change for prolonged periods. Under such circumstances, the Ricardian estimates would be either over or under estimating the climate change impacts, depending on how the prices change. Also, this approach can not include in its estimates the possible positive impacts due to CO2 fertilization. 142 There have been a few other studies estimating the climate change impacts on agriculture, which can not be strictly classified under either of these two broad methodologies. These include studies by Leemans and Solomon (1993) and Kaiser et.al. (1993). OBJECTIVES The general objective of this study is to asses the climate change impact on Indian agriculture. The specific objectives of the study are: * to apply the Ricardian approach with appropriate modifications, to account for specific characteristics of Indian agriculture, to assess functional relation between farm level net-revenue and climate using the cross-sectional observations; to assess the regional and seasonal distribution of impacts due to plausible climate change scenarios; * to assess the influence of yearly weather on the functional relation estimated between farm level net-revenue and long term climate. The structure of this report is as follows. In the next section, the methodology and main findings of production function approach followed by Kumar and Parikh (1996) are described. The net-revenue methodology, and its link with the traditional Ricardian methodology are discussed in the following section. This is followed by discussion on the data and the model specifications. The next section discusses the estimation of the functional relation between farm level net-revenue and long term climate, estimated using net-revenue methodology. This section also discusses the seasonal and regional distribution of impacts due to changes in climate. This is followed in the next two sections by discussion on implications of the results for global climate change, and the influence of yearly weather on functional relation between net-revenue and climate. The final section provides the conclusions and policy guidelines. THE PRODUCTION FUNCTION APPROACH In this section, we briefly discuss the methodological aspects related to production function approach for assessing climate change impacts on agriculture. Since production function approach serves as an alternative way of assessing the climate change impacts on agriculture, the discussion in this section would provide a back-drop for the approach followed in the present study, namely Ricardian approach. Specifically, this section would describe the approach followed in Kumar and Parikh (1996), and the major findings obtained therein. There are two broad stages in the study, namely the physical impacts assessment and the economic implications of such physical impacts. The future climate change scenarios have been developed using results from the equilibrium experiments of general circulation models (GCMs), along with the observed climate. For assessing the physical impacts of climate change on agriculture, the study has followed crop simulation modeling approach. Crop simulation models are among the most powerful tools for analysing the interactions of the crop-management-climate-soil system. Generally carefully calibrated models can be used to assess the situations that do not exist at present. Impact analyses 143 of future climate change are of this type. The crop model should be sensitive to major weather variables (namely, precipitation, temperature, solar radiation and sunshine) and to the CO2 concentration in the climate. Erosion, Productivity and Impact Calculator (EPIC) (Jones et.al., 1991), one of the widely used crop model, has been used in the study for assessing physical impacts of climate change. After calibrating the crop model at a specific site, the site specific base yield of a particular crop is assessed. The climate change scenarios are then applied to modify the climate parameters, and the simulations are carried out to estimate the future yield of the crop. Since the crop models do not accurately reproduce the observed crop yields due to a variety of limitations in the input data and crop model formulation, the percentage deviations of future yield, with respect to base yield are used for further analysis. To translate the physical impacts - which can be viewed as gradually occurring supply shocks to the production sectors - into economic terms, one has to consider the consequences of such shocks on both producer and consumer surpluses. Such an approach would allow estimation of the true social welfare loss/gain of the impacts of climate change on agricultural output. Also, it is important to consider the consequences on different sections of the economy in the present as well as the future time periods. A natural framework satisfying these concerns is a dynamic applied general equilibrium model. The Agriculture, Growth and Redistribution of Income Model (AGRIM), developed by Narayana, Parikh and Srinivasan (Narayana et.al., 1991) is used for assessing the socio-economic implications of climate change. In the economic modeling framework, the imposed yield changes would trigger changes in production levels, and also prices. The price changes would result in a host of other changes, such as movement of labor and capital between various sectors. The yield changes imposed on the system are assumed to occur gradually, i.e., though the climate change induced yield changes start occurring in the yearl990, they reach their full impact in the year 2060. This allows the economic agents to adjust their behaviour over this time period. Thus, even though specific farm level adaptations have not been taken into consideration, possible adaptations by the economic agents have been modeled. The final objective of the economic modeling exercise is to assess the welfare implications, in terms of equivalent incomes and population proportions in various expenditure classes of the economy. The major findings of the study are that wheat crop, grown generally in the winter season of the year, is likely to be affected more than rice crop following climate change; CO2 fertilization effects seem to dramatically reduce the negative impacts due to climate change. From the welfare angle, the study showed that substantial number of people move from higher income classes to lower income classes as a result of climate change induced shocks, and the social welfare is adversely affected. As wheat crop is expected to get affected more adversely than rice, one possible inference could be that the wheat growing regions, such as Haryana, Punjab, and Uttar Pradesh, may experience more severe negative impacts, than predominantly rice growing states such as West Bengal and Orissa. 144 T HE NET-REVENUE APPROACH Absence of well-functioning land markets in the Indian context makes it rather difficult to apply the Ricardian approach in strict sense. However, as shown below, and argued by Sanghi et.al. (1998), one can develop an alternative, namely, 'Net-Revenue' approach. Consider profit maximizing farmers producing outputs Y, using variable inputs X, and quasi-fixed factors (such as, land, irrigation facilities etc.) F. Let these farmers be facing climate factors C, soil characteristics S, technologies T, and infrastructure I. For a given output (PY) and input (P.) prices, the short-run (maximized) profits can be expressed as, II' = 1 (Py, P,x F, I, T, C, S) Using Hotelling's lemma we can obtain the output supply and factor demand equations: Y' = ar17aPy ; X' = afflaPX and using these the variable profit function can be written as, lv =PyY* -PxX* = fI(Py,Px FIT C, S) In the above equation, the farmers have completely adapted to the climate variables C, by adjusting their short and long-term inputs such as, F, I, and T, that they are facing and have arrived at profit maximizing mix of outputs and inputs. In equilibrium, short-term profits, flv can be measured and expressed as net revenues per unit of land, ([v / L). The net revenues can then be considered to incorporate the optimizing choices of outputs and inputs by farmers. This implies that, the climate impacts ( = (arI,/L)I8C ) calculated from the above equation represents the responses of farmers to climate, soil and other factors. Following Mendelsohn et.al., (1994), one can argue that the prices Py and P. are 'equilibrium' prices and in a cross-section differ only by transport costs (which would continue to exist in future). Similarly, arguments can be made for the elimination of technology and infrastructure variables. Refer Sanghi et.al. (1996) for more details. As mentioned in the beginning of this section, lack of land prices does not allow one to estimate Ricardian estimates in strict sense. However, land prices are presumably based on expected future net revenue. And, in the present study, the current net revenues are considered as a proxy for expected future net revenues. To the extent that this is true, the estimates provided in this study are similar to those of Ricardian approach. Also, to account for the specific characteristics of Indian agriculture, certain modifications have been introduced in the 'traditional' Ricardian methodology. The key characteristics of Indian (and also other developing countries) agriculture are: a) government administered prices for crucial agricultural inputs such as tractors and bullocks; b) introduction of high yielding variety (HYV) seeds in phased manner across the nation, which has led to differences between regional observations; c) participation of household labor in agricultural activities, leading to underestimation of input costs. 145 To address these issues, appropriate proxies have been included in the list of independent variables while estimating the model. DATA AND MODEL SPECIFICATIONS To test the above net-revenue model, district level cross-sectional data of India has been considered in this study. The districts are considered as units of analysis because most of the data, with the exception of climate (see below for more details about this), is available at district level, and substantial variations in climatic, edaphic and economic factors exist among districts. Most of the agricultural data used in the study was obtained from a data set compiled from various publications by James McKinsey and Robert Evenson at Yale University, USA. This data set includes time-series data on various agricultural variables for 271 districts covering 13 major states of India, from 1957 to 1987. Though the data set as such includes many variables, this study has used for each district the crop-wise production, area, price; area under High Yielding Varieties (HYV); fertilizer input and its price; agricultural labor and their wage rate; fraction of cultivators in the total number of people engaged in agricultural activities; quantities and prices of bullocks and tractors; and dummies representing various soil characteristics. A number of corrections have been made in the original data set after cross checking the same with the available records. The corrections have been made mainly on area, production and prices of various crops. Additional data on various census variables, such as, population density and literacy proportion, has been added to the above data set. Similarly, data on districts altitude, estimated using figures and data on various meteorological stations across India, has also been added to the original data set. For predicting the district level climate, this altitude data has been used. It has also been used to proxy solar energy in the main net-revenue regressions. In the following sections, we discuss in detail about the use of edaphic and other control variables in the study; and the estimation of net-revenue and district level climate variables. NET-REVENUE District level net-revenue per hectare, which has been used as the dependent variable in the main regressions, is defined as, Net Revenue per hectare = [(Value of production of crops) - (Expenditure on fertilizer and hired labor)] / (Total area under crops) The study has considered 20 crops while estimating the net-revenue. These 20 crops consist five major crops, namely, Bajra, Jowar, Maize, Rice and Wheat; and 15 minor crops, including Barley, Cotton, Groundnut, Gram, Soybean, and Sugarcane. While assessing the value of production of crops, the farm harvest prices have been used. These prices have been considered appropriate for the analysis in this study, because based on these prices only farmers form their expectations and behave. As Ricardian approach, and the net-revenue approach used here, assume that 'equilibrium' prices exist, it is important to estimate the net revenues at local prices; and the study has used district level prices only in this study. In the net revenue 146 calculations the expenditure on family labor has not been included, because appropriate wage rate for the same is difficult to obtain. Using the wage rate of agricultural labor would overestimate the expenditure on family labor, as the agricultural labor are typically hired during peak demand periods and thus the wages tend to be high. However, in the main regressions an independent variable has been included to this extent, namely, fraction of cultivators in the total number of people engaged in the agricultural activities. Similarly, In the case of bullocks and tractors, due to govermment's policies the official prices do not vary much across districts in India, and hence using these prices for calculating the rental costs of these inputs may not be appropriate. However, the quantities of these crucial inputs vary across districts and to capture their influence the study has used the district level per hectare quantities of bullocks and tractors as independent variables in the net-revenue regressions. While calculating the revenue and costs, all prices and wages have been deflated using the agricultural GDP deflator, and are expressed in 1980 rupee terms. Figure 5.1 provides the regional distribution of net-revenue per hectare, averaged over 20 year time period. 147 Figure 5.1: District level net-revenues per hectare (1980 Rs.) Punjab W.Bengal Karnataka Tamil Haryana Uttar Pradesh Gujarat Orissa Nadu Andhra Madhya Maharastra Rajastan Bihar Pradesh Pradesh Note:;..uu, it n. i i Note: Values on horizontal axis are districts' codes. EDAPHIC VARIABLES In the original data set, no quantitative data on edaphic variables has been recorded, because no such data at district level exists in India. However, inclusion of edaphic variables is very crucial to the analysis, as districts differ substantially in the soil characteristics they possess and thus have differences among the crops that could be grown. Though various other soil characteristics have been considered in the initial runs, in the final model specification, two sets of soil dummies, one representing the soil types and another the top soil depth classes, have been used. In the data set, there are 19 soil types and five top soil depth classes. However, these have been 'clubbed' in the main regressions, for ease of handling. For clubbing these dummies, the criteria followed is that the dummies with similar influence on the net-revenue can be combined. Thus, first regressions with all the dummies included was run and then using the magnitude and signs of their coefficients as guiding elements appropriate dummies have been clubbed. The idea behind including these dummies has been only to control for edaphic variation across districts in the main regressions, and hence direct interpretation of the soil dummy coefficients is not appropriate in the analyses. Also, the soil variables that could have climate information in-built in them have not been included in the model. OTHER CONTROL VARIABLES Other than fraction of cultivators, per hectare bullocks and tractors, as defined above, few other variables have been used as independent variables to control for the cross-sectional variations that could result in variations in the dependent variable. These variables including, population density and literacy proportion, are expected to proxy urban development. The spread of 'green-revolution' across different districts is captured by the control variable, fraction of area under high yielding varieties (HYV). Altitude has also been included as an independent variable to act as a proxy for solar energy. CLIMATE VARIABLES The appropriate climate variables that are suitable for the present study are the 'normal' climate variables. The normal climate is defined in climatology as 30 year average climate. As against the yearly weather, the normal climate is considered appropriate for the Ricardian approach because it is against these climate variables that farmers have made adaptations. However, the present study using net-revenue approach, does depend on yearly weather along with the normal climate. Keeping this in view, and also the data available, the following two analyses have been carried out: a) Analysis I : Estimating functional relation between net-revenues and the normal climate, where normal climate is the average temperature and precipitation observed over the period 1960-'80; b) Analysis II: Estimating functional relation between net-revenues and the normal climate, in the presence of yearly weather deviations from this normal climate. Here, the normal climate corresponds to the period 1960-1980, and the yearly weather data is for 149 the same period. The focus here is to see whether inclusion of yearly weather influences in the estimation procedure brings down the fluctuations in the results. The data for Analysis I and II is based on a new data set provided by Indian Meteorological Departrnent (IMD), Pune. This data set corresponds to 120 meteorological stations across India. Each year's daily average temperature and monthly total rainfall (for all 12 months) for the period 1960 to 1980 are available. For Analysis I mentioned above, simply the average temperature and precipitation over the 20 year period has been used and considered as 'normal' climate. Note that this is not exactly 'normal' climate in Climatologist's sense, because it is the average of 20 years data only. For Analysis II, the normal climate of the period 1960-1980, along with the deviation of each year's weather from this climate has been used. There is one old data set available from Food and Agriculture Organization (FAO) for climate data. This data set covers the time period 1930-1960; and has climate data for about 160 meteorological stations across India. This data set has been used in a parallel study by Sanghi et.al (I 997). There are a number differences between the IMD and FAO data sets: e they correspond to different time periods e the geographical coverage of stations under the two data sets is different e the same station under two data sets has different climate. Figure 5.1 provides a comparison between the two data sets; and Appendix A gives the coverage under IMD data set and its correspondence with the FAO data. ESTIMATING DISTRICT LEVEL CLIMATE AND WEATHER As mentioned above the climate (and weather) data is available at meteorological station level, whereas the rest of the data is available at district level. To bring the climate (and weather) data from station level to district level, a spatial statistical analysis is carried out, which examines the determinants of the climate of each district. For doing this, it is assumed that all the meteorological stations within 600 miles radius from the geographic center of the district provide some useful information about that district's climate. The choice of 600 mile radius is arbitrary; and the intention was to draw as many stations as possible into the circle, so that the estimates do not depend too heavily on any one station. A climate surface in the vicinity of the district is then estimated by running a weighted regression across all meteorological stations within 600 miles. The stations which are nearer to the district center presumably contain more information, than the stations that are far away. Hence the inverse of the square root of a station's distance' from the districts center is used as the weight. Separate regressions have been run for each district, as each district's set of meteorological stations and the corresponding distances would be different. I The distance between any two points on the Earth's surface with known latitude and longitude, can be calculated using the formula: cos D = sin a sin j + cos a cos L cos L where, D = angle subtended at the Earth's center by the arc distance between points A and B a, p = latitudes of points A and B respectively L = difference in longitude of points A and B. Using the Earth's radius and D, the distance between A and B can be estimated. 150 The dependent variables are monthly temperatures and precipitations, and the independent variables include latitude, longitude, altitude and shoreline distance. Based on the evidence from US and Brazil studies (Mendelsohn et.al., 1994; Sanghi et.al., 1996), a second order polynomial had been fitted over these four basic variables, including interactive terms. Thus the climate regressions included 14 variables and a constant term. Under analysis I, a total of 6504 (= 12x2x271) regressions were run, to assess district-level normal temperature and precipitation using the normal climate available at meteorological stations. Where as for analysis II, a total of 71544 (=I lx12x2x271) regressions were carried out to assess district level yearly weather for the period 1970 to 1980 (i.e., for 11 years), using the yearly weather data available at meteorological stations. A sample regression is shown in Table 5.1. Though in the final analysis the temperature and precipitation corresponding to only four months was used, estimations were carried for all the twelve months for each district. This is expected to be useful for any future analyses. Using the regression coefficients, the predicted value of climatic variables for the geographic center of the district is estimated. These predicted climate (and weather) variables are used as independent variables in the main net-revenue regression. This complicated procedure for assessing district level climate (and weather) has been adapted to obtain accurate estimates. Figures 5.2 and 5.3 show the district level seasonal distribution of temperature and rainfall. For comparison, the district level seasonal distribution of temperature as estimated using FAO data has been shown in Figure 5.4. Because of the inherent differences between the two data sets, the district level climate estimations are also different. MODEL SPECIFICATIONS The following are the common specifications for models used under Analysis I and II: * net-revenue per hectare has been used as the dependent variable * normal temperature and precipitation of months January, April, July and October - representing the four seasons - have been considered to represent climate 3 linear and quadratic the above climate terms have been used as independent variables, as the inherent functional relation is expected to be non-linear * climate interaction terms have also been used as independent variables * all the climate terms have been 'demeaned' * all the observations have been weighted by the total area under the crops considered in the study2 control variables included are soil dummies, fraction of cultivators, bullocks, tractors, fraction of HYV, population density, literacy proportion, and altitude For Analysis II, along with the above specifications, eight more independent variables are included to account for the deviation of each year's weather from the normal climate. 2 As the data is at the district level and not at the farm level, the extent of measurement error in each district's observation is not known. Assuming that larger districts contain more number of farms and hence would have less measurement error, higher weights are provided to larger districts by using acreage as weight. However, such specification has inherent bias associated with it. A particular district could be large in size, but may not have many farms due to climatic conditions prevailing; and hence acreage weight would be misleading. 151 ECONOMETRIC PROCEDURES For Analysis I, three econometric procedures have been applied for checking the robustness of the estimates. These are: * Pooled Procedure: Observations corresponding to the study period, 1966 to 1986, have been pooled and yearly dummies have been added to control for year-to-year variation in weather, prices etc. for running a single regression. * Averaged Procedure: Averaged values of dependent and independent variables over the period 1966 to 1986 have been used for running a single regression. * Repeated Cross-sectional Procedure: For each year a separate regression is run. Since the overall objective is to assess the long-run impact of climate on net-revenue, the first two procedures are assumed to provide appropriate estimates; on the other hand, as the dependent variable in the present study, namely, net-revenue, is more an yearly phenomena, the third procedure is expected to capture the yearly fluctuations in the estimates. For Analysis II, the period of analysis has been restricted to 1970 to 1980. In this analysis, the objective was to assess the influence of yearly weather on the climate coefficients. Towards meeting this objective, two procedures have been applied: * Pooled Procedure: For the period 1970 to 1980, all the observations have been pooled, once without weather terms and once with weather terms, to run a single regression. Each time, yearly dummies have been included to capture the year-to- year variations. Without weather terms included, the year dummies would account for the year-to-year influences of weather, prices and other variables affecting agriculture. With weather terms included in the model, the year dummies would account for the year-to-year influences due to prices and other variables affecting agriculture. * Repeated Cross-sectional Procedure: For each year a separate regression is run. From the pooled estimates, one would know to what extent the inclusion of weather terms has influenced the estimation of climate coefficients. Whereas from the repeated cross-sectional procedure, it is possible to see whether the inclusion of weather terms has been useful in explaining the year-to-year fluctuations of climate coefficient estimations. If weather were to be the cause behind year-to-year fluctuations, one would expect that climate coefficients would get stabilized with the inclusion of weather terms in the regression. 152 S a, bA-4 00 la C 0 w _ _a Punjab G Punjab G West Bengal T West Bengal _t Karnataka Kamnataka Tamil Nadu Tamnil Nadu Haryana c Haryana Uttar Pradesh Uttar Pradesh Gujarat j l | I | | llr Gujarat Orissa 1 l l l E I IS Orissa Andhra Pradesh Andhra Pradesh Madhya Pradesh Madhya Pradesh Maharashtra Maharashtra Rajasthan Rajasthan Bihar Bihar s. I - Nc c CA C A 0 U I ~~~~~~~o - 0 ues oIIou nonou Punja| Punjab est Bengal |est Bengal a Karnataka Karnataka TainilNadu Tamil Nadu Haryana . Haryana > Uttar Pradesh C| Uttar Pradesh Gujarat Gujarat | Orissa l g l l C | | Orissa L Andhra Pradesh Andhra Pradesh .* Madhya Pradesh Madhya Pradeshj Maharashtra Mahamashtra Rajasthan Rajasthan Bihar Bihar ______________________________ Biha 0 500 1000 1500 0 20 40 60 80 Punjab Punjab West Bengal to West Bengal C. Karnataka Karnataka 04 Tarnil Nadu Tamil Nadu I Haryana C- Haryana O c~~~~~~~~~~~ Uttar Pradesh Uttar Pradesh GruIjasra t l Gujarat Orissa Ois Andhra Pradesh Andhra Pradesh Madhya Pradesh Madhya Pradesh Maharashtra Maharashtra Rajasthan Rajasthan Bihar Bihar 0 50 100 150 200 250 0 50 100 150 200 250 300 Punjab h Punjab West Bengal West Bengal Karnataka Karnataka Tamil Nadu Tamil Nadu Haryana 0 Uttar Pradesh CD Uttar Pradesh _3- Gujarat Gujarat Orissa Orissa Andhra Pradesh Andhra Pradesh Madhya Pradesh Madhya Pradesh Maharashtra Maharashtra Rajasthan Rajasthan Bihar Bihar Figure 5.4: State level seasonal temperatures as per FAO data Degrees Celcius Januaryciu 35 JaDegeesrylci5 April 30 30 25 25 20 20- 15 is 10 10 5 5 0 0 OF. 0 30 July 28 October. ~ . 29 27 5 0. 5 2972 26 28 2 25 ~~~~~~~~~~~~~~~~~~~~~22 24 ~~~~~~~~~~~~~~~~~~~~~~~24 23 20 1 1 1 1 1 r~~~~~~~~~~~~~~~2 25~~~~~~zE 22 24 21~~~~~~~~~~~~~~~~~~~~~~ EMiPRICAL ESTIMATION OF CLIMATE RESPONSE FUNCTION In this section, the results of the analysis I are presented and the characteristics of climate response function estimated are discussed. Table 5.2 gives the coefficients estimated using pooled and averaged procedures; while Table 5.3 provides the coefficients estimated using repeated cross-sectional procedure. Note that, though the repeated cross-sectional procedure has been carried-out for all the years from 1966 to 1986, only the results corresponding to years from 1970 to 1980 have been presented in Table 5.3. As could be seen from Table 5.2, the two estimation procedures, namely pooled and averaged, provided remarkably close estimates. The estimates from repeated cross-sectional procedure have also retained to a large extent the significance levels and signs of the coefficients. For the rest of the discussion here, the estimates from the pooled procedure are considered. CONTROL VARIABLES As the net-revenues for the year 1979 were lowest among all the years, the dummy for this year has been omitted. Accordingly, the dummy coefficients for all the years are significantly positive. Fraction of cultivators has positive effect on net-revenue, implying that more of household labor is beneficial. As expected other control variables, population density, literacy proportion, bullocks and tractors have positive influence on the dependent variable. IMPACT ESTIMATIONS Using the above coefficients estimated under pooled procedure, the possible impacts on district level net-revenue per hectare are calculated. The climate coefficients indicate that there are significant influences on net-revenue, and that they are highly non-linear. As direct interpretation of climate coefficients is not easy because of the inclusion of climate interaction terms, the changes in the dependent variable are assessed under a scenario with 2.0°C temperature rise and 7% increase in precipitation. For each district, to get a measure of possible change in net-revenue per hectare, the dependent variable is estimated once with base climate and once with the modified climate. To get a measure of national level change in net-revenue, the district level estimates are summed up. While calculating the seasonal effects, the changes in climate are applied to the appropriate month; and similarly while estimating the partial effects due to temperature alone and precipitation alone, the changes are applied to the appropriate climate variable keeping the other climate variable at its base value. Thus while estimating the partial effect due to temperature alone, the precipitation variable is kept at its base value. The national level estimates indicate that: * January, April and July seasons have negative influence on net-revenue; whereas October effects are positive * Negative temperature effects have major share in the overall effects, which are negative. Though precipitation contributed positively, it gets nullified by the large negative effect due to temperature rise 156 The general trend of these results is in tune with those obtained by Sanghi et.al. (1997) using the FAO climate data set. The regional distribution of partial effects due to temperature and precipitation are shown in Figure 5.5a and 5.5b respectively; while Figure 5.6 shows the regional distribution of overall effects. As it is clear from these Figures, the trend of effects observed under partial temperature change is same as that observed under overall change in climate. The regional distribution of overall effects shows that the sub-tropical regions, namely, the states of Haryana, Punjab, and Western Uttar Pradesh get adversely affected along with few western coastal regions and few in the south India. In comparison, the eastern states, West Bengal, Orissa and to some extent Bihar enjoy beneficial effects due to changes in climate. YEARLY FLUCTUATIONS IN THE IMPACTS As discussed earlier, the net-revenue is more of a yearly phenomena and accordingly the estimates from repeated cross-sectional procedure are expected to provide an idea of how the climate coefficient estimates vary from year to year. Figure 5.8 gives the estimates from year- wise regression analysis. Here, the impacts due to four scenarios are shown, of which scenario II corresponds to the 2.0°C temperature rise and 7% precipitation increase. Discussion on other scenarios is made in the next section. As could be seen from the figure, the estimates vary substantially from year to year. There could be a number of reasons for these fluctuations in the impacts, including weather, prices and other variables influencing agriculture. In order to see to what extent year-to-year variation in weather causes these fluctuations in impacts, in the later section includes the weather variables in the regression and estimates the climate coefficients. 157 Figure 5.5a: Impacts on district level net-revenues per hectare due to +2.0°C temperature rise Punjab VWst Karnataa Tarril Haryana Ulttar Pradesh Gujarat Orissa i Ardhra Madhya Pradesh Maharastra Rajastran Bihar Note: Values on horizontal axis are districts' loe s- I i I a 9 El fi E E la f IiR S fi ! 3 S S a 11 _ Fs I 9 6 o s a i s I i a a la a e i 0 II la I a 1! a E Note: Values on horizontal axis are districts' codes. Figure 5.5b: Impacts on district level net-revenues per hectare due to +7.0% increase in rainfall " Punjab V%bst Kamataka Tamil Haryana Uttar Pradesh Gujarat Orssa Bengal Nadu I~~~~~~~~~~~~~~~~~~~~~~~~~~ o f s...... ........ . .....; l.s.l. ............... A.... .. q ........... I Andhra Madhya Maharastra Rajastan Bihar Pradesh Pradesh Go a a 1! a I a a a g El I a a if a is F. El a F. a if i a g i 5e a 5 a a a 9 a a5 9 E a 1 3 5 r. is I -, e 5 5 F a F a I a S Note: Values on horizontal axis are districts' codes. Figure 5.6: Impacts on district level net-revenues per hectare due to +2.OoC and +7% rainfall scenario g Punjab West Bengal Kamataka Tamil Haryana Uttar Pradesh Gujarat Orssa rN t Nadu cc Andhra Madhya Pradesh Maharastra Rajastan Bihar Pradesh 1S Note: Values on horizontal axis are districts' codes. IMPLICATIONS FOR GLOBAL CLIMATE CHANGE Earlier IPCC estimates have put the possible changes in the global mean temperature to lie between 1.5 to 4.5°C, following a doubling of atmospheric concentration of CO2 equivalent (IPCC, 1990). The best guess estimate of the above range was placed at 2.5°C. The latest estimates (IPCC, 1996), however indicate a lower temperature change following CO2 doubling. The differences are due to the differences in the structure of underlying models, and the latest estimates are based on more advanced coupled ocean general circulation models (GCMs) and are expected to be more reliable. As GCM scenarios at the district level details are not easy to estimate, the present study has considered the following four scenarios for assessing the possible impacts due to climate change at state and national levels: a) Scenario I : +1.0°C temperature change and 0% increase in precipitation b) Scenario II: +2.0°C temperature change and 7% increase in precipitation. c) Scenario III: +2.5°C temperature change and 15% increase in precipitation. d) Scenario IV: +3.5°C temperature change and 15% increase in precipitation. For calculating the overall impacts due to a particular climate change scenario, the district level per hectare estimates are first converted to district level total estimates by multiplying the per hectare estimates by the total cropped area of the district. The district level total effects are then aggregated to state and national level. Tables 5.4 and 5.5 present the results of above scenarios for Analysis I. The break-up of results for all the seasons and for temperature and precipitation are given in Table 5.4. From the results one can observe that the temperature response function is of inverted 'U' shape. That is, with higher climate change, more losses would occur. However, the response function of precipitation appear to be of 'U' shape. The overall response function is of inverted 'U' shape. The overall impacts estimated using IMD data set appear to be lower than those estimated using FAO data set. One possible explanation for this is that, the response function obtained using IMD data is more nearer to peak of the inverted 'U' curve, than the one estimated using FAO data. However, results from both the data sets concur that the overall response function would be of inverted 'U' shape. Figures 5.9 and 5.10 provide the regional distribution of impacts due to the above climate change scenarios, for results from using IMD data set and FAO data set respectively. The general pattern followed in both the graphs is similar, except for a few differences, notable among them being the case of sub-tropical belt, i.e., Haryana, Punjab and UP. These three states are in sub- tropical climates and winter crop, wheat is the main crop grown there. With results from many studies, including those by Rao and Sinha (1994) and Kumar and Parikh (1996), showing that climate change would have relatively higher negative impacts on wheat crop production, the regional distribution of impacts as obtained under Figure 5.9 appears to be more in tune with the conventional wisdom. 161 CLIMATE RESPONSE FUNCTION WITH YEARLY WEATHER INFLUENCE As discussed above, one of the possible reasons for the yearly fluctuations in impacts observed in Figure 5.4 could be due to the fluctuations in the yearly weather. In this section the results from Analysis II are presented, where along with the independent variables included in Analysis I, the yearly weather deviation tenrs are added to the list of independent variables. The time period has been restricted to 1 1 years, from 1970 to 1980. First to assess to what extent the coefficients of climate terms would get disturbed due to the presence of weather terms, pooled analysis has been conducted, once without weather terms and once with weather terms. The results are presented in Table 5.6. From Table 5.6 it is evident that though the weather variables are significant, they do not drastically change the climate coefficients. The weather variables themselves are also significant, indicating that they do have some influence. All the control variables retain their signs and significance. To assess to what extent the variability of climate coefficients has reduced with the inclusion of weather terms in the model, repeated cross-sectional analysis has been carried out for the period 1970 to 1980. The results from this are presented in Table 5.7, and the results are compared with those obtained under Analysis I in Figure 5.11 for the impacts under scenario II - i.e., 2.0°C temperature rise and 7% increase in precipitation. The hypotheses was that, if the weather were the cause behind the yearly fluctuations in impacts, then inclusion of yearly weather deviation would bring down the fluctuations. From Figure 5.11 it appears that the above hypotheses may not be holding. Though the variation in impacts has come down under Analysis II, compared to that under Analysis I, the fluctuations are still significant. This indicates that there are influences other than weather, that dictate the movement of net-revenue from year to year. One possible explanation is that the relative prices between crops is not constant as assumed under Ricardian approach. Another explanation could be fluctuations in the government policies. C ONCLUSIONS AND POLICY GUIDELINES This study estimated the relationship between farm level net-revenue and climate variables using the cross-sectional evidence in India. The district level data has been used for the analysis. Using the observed reactions of the farmers, the study tried to understand how they have adapted to different climatic conditions across India. Using data from 1966 to 1986, in Analysis I, the cross-sectional evidence between net-revenue and long-term normal climate variables has been explored, while controlling for a host of other variables which could have influence on net-revenue. The climate data used for this analysis has been obtained from Indian Meteorological Department, and the results are compared with those arrived at using an older climate data set available from FAO. The underlying relation between net-revenue and climate variables appear to be non- linear. The seasonal variation in impacts shows that but for October, all other seasons would have negative influence on net-revenue. The positive influence of October is not sufficient to counter the negative influence of all other seasons, and hence the overall impact is negative. The 162 impacts due to temperature change are negative and are much more than the positive impacts due to precipitation change. The results show that an inverted 'U' shape relation exists between temperature and net-revenue; and that a 'U' shape relation exists between net-revenue and rainfall. The coefficients are remarkably stable across three different econometric procedures, namely, pooled, averaged and repeated cross-sectional procedures. The impacts due to a set of climate change scenarios have been estimated using the functional forms obtained from the above analyses. The results from this study show that the losses expected from the climate change on Indian agriculture would be more than similar estimates made in USA and Brazil. The regional distribution of impacts seem to concur with the conventional wisdom. In an attempt to explain the yearly fluctuations in the climate coefficients, in Analysis II, the study has included the weather deviation terms in the cross-sectional regression. The hypotheses being, if weather were to explain the yearly fluctuations in the climate coefficients, inclusion of weather deviation terms would make the climate coefficients stable over years. However, the study showed that inclusion of yearly weather deviations terms has not brought down the fluctuations in the estimates in substantial manner. This indicates that there could be influences other than weather, such as prices and government policies. The climate variables retained their significance even in the presence of yearly weather terms, thus proving the robustness of the estimates. The estimates provided in this study could be considered as only a beginning of an attempt to quantify some of the impacts and the costs, despite adaptation, that Indian agriculture is likely to face in the event of climate change. One need to be careful to use them for policy purposes. They should be used in a very conservative manner. They are in no way accurate enough to judge courses of action for more than a billion people for years to come. Evidence from the literature on risk avoidance should be taken into consideration to understand how many billions of dollars are spent to avoid possible risk in, say nuclear power plants. Such knowledge would give a criteria to compare with for understanding how much the food security is of worth for billions of people including those outside India. 163 Table 5.1: District climate interpolation - Guntur, Andhra Pradesh Independent Temperature Rainfall Variable January October January October Intercept 61.731 47.357 3470.9* -4043.1 Altitude -2.05E-03 1.35E-04 -0.15406 0.32095 Latitude -3.1441* 2.2618 55.342* -126.22* Longitude 6.88E-04 -1.0203 -101.64* 126.01 Distance to sea 8.56E-02* 2.36E-02 -0.40994 3.1553 Latitude x longitude 5.16E-02 -2.49E-02* -0.82535* 1.4656* Latitude x altitude 2.36E-04* -2.66E-05 -3.72E-04 7.08E-03 Altitude x longitude -1.04E-04 -6.59E-05 1.99E-03 -5.75E-03 Alitude squared 4.40E-07 -4.91E-07 1.48E-05* 3.57E-05* Latitude squared -4.23E-02* -1.26E-02* 9.58E-02 0.11844 Longitude squared -6.66E-03 9.55E-03 0.75089* -0.89006 Distance to sea squared -1.79E-05 -1.81E-05* -2.73E-04 1.88E-03* Distance x altitude -4.06E-07 5.71E-08 1.44E-06 6.79E-06 Distance x latitude -1.OOE-04 1.88E-05 2.69E-02* -1.28E-02 Distance x longitude -9.29E-04* -1.83E-04 1.08E-03 -5.04E-02* Adjusted R2 0.9952 0.997 0.7509 0.8891 Number of observations 79 79 79 79 Statistically significant at the 5% significance level 164 Table 5.2: Net revenue regression for the period 1966-1986, pooled and averaged procedure - analysis I Variable Name (Abbreviation) Pooled Procedure Averaged Procedure Coefficient T-ratio Coefficient T-ratio January Temperature (JanT) -94.872 -6.808 -99.544 -2.054 April Temperature (AprT) -174.035 -11.961 -137.744 -2.623 July Temperature (JulT) -140.771 -5.203 -149.422 -1.598 October Temperature (OctT) 457.583 13.055 468.182 3.880 January Temperature Square (JanT2) -7.645 -3.386 -7.824 -1.010 April Temperature Square (AprT2) 32.481 8.362 31.977 2.343 July Temperature Square (Ju1T2) -20.587 -2.581 -23.309 -0.833 October Temperature Square (OctT2) -56.070 -12.113 -56.433 -3.545 January Rain (JanR) 7.454 4.391 7.476 1.281 April Rain (AprR) -4.502 -8.908 -4.335 -2.503 July Rain (JulR) -0.446 -3.861 -0.321 -0.801 October Rain (OctR) 6.370 10.782 6.771 3.334 January Rain Square (JanR2) -0.656 -12.354 -0.604 -3.310 April Rain Square (AprR2) 0.041 9.394 0.040 2.649 July Rain Square (JulR2) 0.001 3.094 0.001 1.176 October Rain Square (OctR2) 0.004 0.928 0.003 0.183 Jan. Temperature X January Rain (Janlnt) -7.687 -11.871 -7.045 -3.153 April Temperature X April Rain (AprInt) 3.608 17.857 3.541 5.031 July Temperature X July Rain (JulInt) -0.325 -4.294 -0.249 -0.944 Oct. Temperature X October Rain (Octlnt) -0.535 -2.084 -0.689 -0.785 Soil Type Dummy (Soill) 104.833 5.698 80.121 1.268 Soil Type Dummy (Soil2) 315.436 13.074 314.670 3.821 Soil Type Dummy (Soil3) -243.170 -7.694 -208.299 -1.901 Soil Type Dummy (Soil4) 73.963 2.818 62.200 0.693 Top Soil Depth Dummy (Soil5) -40.264 -0.850 -56.176 -0.342 Top Soil Depth Dummy (Soil6) 111.073 2.140 97.401 0.543 165 Table 5.2: Net revenue regression for the period 1966-1986: pooled and averaged procedure - analysis I (cont.) Variable Name (Abbreviation) Pooled Procedure Averaged Procedure Coefficient T-ratio Coefficient T-ratio Cultivators per Ha. (Cultiv) 76.030 2.288 230.736 1.632 No. of Bullocks per Hectare (Bullock) 23.127 0.552 -99.847 -0.627 No. of Tractors per Hectare (Tractor) 24449.370 7.583 60717.790 3.396 Population Density (Popden) 33.866 4.834 44.290 1.752 Literacy Proportion (Litprop) 720.208 6.958 582.393 1.559 Fraction of HYV (Hyvfr) -172.055 -2.762 -254.218 -0.839 Altitude (Alt) -0.299 -5.070 -0.243 -1.205 Year Dummy (dy66) 226.647 4.711 Year Dummy (dy67) 398.713 8.485 Year Dummy (dy68) 172.021 3.650 Year Dummy (dy69) 174.145 3.757 Year Dummy (dy7O) 303.212 6.651 Year Dummy (dy7l) 259.206 5.710 Year Dummy (dy72) 260.161 5.742 Year Dummy (dy73) 528.277 11.930 Year Dummy (dy74) 496.972 10.962 Year Dummy (dy75) 455.348 10.272 Year Dummy (dy76) 321.300 7.174 Year Dummy (dy77) 384.584 8.748 Year Dummy (dy78) 312.276 7.150 Year Dummy (dy8O) 292.776 6.717 Year Dummy (dy81) 179.790 4.187 Year Dummy (dy82) 145.872 3.345 Year Dummy (dy83) 287.872 6.733 Year Dummy (dy84) 134.542 2.982 Year Dummy (dy85) 85.294 1.878 Year Dummy (dy86) -44.700 -0.977 Constant (Const) 630.311 7.722 858.006 3.234 Adjusted R-square 0.5331 0.6582 166 Table 5.3: Net revenue regressions for the period 1970 to 1974: repeated cross-sectional procedure - analysis I 1970 1971 1972 1973 1974 JanT -149.380 -2.666 -112.240 -1.859 -221.630 -3.817 -35.902 -0.601 -132.380 -2.173 AprT -235.590 -3.672 -139.640 -2.015 84.282 1.269 -132.040 -1.923 153.320 2.170 JulT -146.920 -1.342 =140.430 -1.168 -231.110 -1.999 -189.620 -1.616 -160.770 -1 =331 OctT 605.930 4.261 486.150 3.124 303.900 2.045 284.550 1.891 183.770 1.211 JanT2 -7.681 -0.868 -5.477 -0.583 -3.075 -0.336 -17.610 -1.910 -14.199 -1.435 AprT2 21.342 1.292 14.738 0.831 -3.983 -0.239 -0.036 -0.002 15.475 0.905 JulT2 -16.332 -0.521 -38.276 -1.120 18.093 0.542 17.595 0.511 60.774 1.686 OctT2 -73.584 -3.940 -50.412 -2.476 -35.359 -1.806 -41.022 -2.051 -23.945 -1.188 JanR 1.163 0.171 1.744 0.235 7.523 1.060 -3.734 -0.510 -9.954 -1.353 AprR -4.559 -2.195 -5.551 -2.409 -7.486 -3.449 -3.641 -1.670 -5.248 -2.364 JuIR -0.773 -1.631 -0.557 -1.079 -0.613 -1.259 -1.017 -2.060 0.031 0.061 OctR 10.371 4.354 8.938 3.429 12.638 5.051 5.761 2.282 7.319 2.856 JanR2 -0.432 -2.037 -0.449 -1.917 -0.207 -0.936 -0.541 -2.444 0.430 -1.844 AprR2 0.042 2.335 0.036 1.839 0.033 1.749 0.024 1.242 0.034 1.812 JuIR2 0.001 1.474 0.000 0.392 0.000 0.190 0.001 1.354 0.002 1.689 OctR2 -0.019 -0.982 0.000 -0.016 0.000 -0.006 0.029 1.459 0.040 1.929 JanInt -8.182 -3.203 -7.751 -2.769 -5.150 -1.940 -10.225 -3.892 -9.458 -3.332 Aprlnt 3.370 3.927 2.709 2.945 1.787 2.057 2.689 3.086 3.134 3.598 Jullnt -0.164 -0.541 -0.444 -1.331 -0.358 -1.105 -0.111 -0.336 0.166 0.507 Octlnt -1.185 -1.111 -0.310 -0.263 1.259 1.132 -0.482 -0.422 0.998 0.868 Soill 89.741 1.249 251.270 3.172 172.680 2.238 177.240 2.287 137.770 1.746 Soil2 543.290 5.644 529.830 4.945 288.060 2.765 394.000 3.740 149.260 1.414 Soil3 -162.240 -1.291 -72.056 -0.527 -213.180 -1.603 -90.009 -0.659 -243.480 -1.733 Soil4 -71.678 -0.685 76.507 0.667 50.541 0.451 79.354 0.712 96.176 0.877 SoilS -326.110 -1.691 -83.554 -0.394 -226.580 -1.084 -72.990 -0.345 -48.897 0.213 Soil6 -243.140 -1.167 85.959 0.373 44.206 0.195 225.720 0.992 220.270 0.904 Cultiv -37.792 -0.378 70.544 0.719 157.330 1.532 308.990 2.626 502.260 4.011 Bullock 96.199 0.608 -27.491 -0.171 2.487 0.017 96.116 0.621 -207.750 -1.341 Tractor 59764.00 1.425 65193.00 1.621 56360.00 1.640 88554.00 2.534 69863.00 2.122 Popden 72.660 2.079 69.759 1.812 41.710 1.186 123.470 3.616 80.593 2.449 Litprop 34.990 0.073 149.190 0.294 452.050 0.940 -589.610 -1.204 -80.727 -0.168 Hyvfr -568.560 -1.041 173.410 0.334 471.630 1.081 42.034 0.106 985.170 2.757 Alt -0.393 -1.703 -0.564 -2.213 -0.638 -2.631 -0.021 -0.083 -0.428 -1.699 Const 1473.700 4.901 920.020 2.798 1055.500 3.306 868.860 2.654 636.570 1.896 Adj.R2 0.6152 0.597 0.5933 Table 5.3: Net revenue regressions for the period 1975 to 1980: repeated cross-sectional procedure - analysis I (cont.) 1975 1976 1977 1978 1979 1980 JanT -107.250 -1.618 -28.937 -0.489 -25.840 -0.431 155.790 2.705 6.535 0.118 -49.608 -0.805 AprT -53.765 -0.733 -238.840 -3.756 -275.100 -4.399 -418.300 -6.864 -259.230 -4.322 -159.330 -2.392 JulT -198.320 -1.525 -97.079 -0.844 -247.210 -2.144 195.780 1.748 -118.450 -1.097 -140.610 -1.174 OctT 422.240 2.534 378.100 2.557 385.910 2.602 348.890 2.416 362.240 2.599 480.880 3.164 JanT2 -9.878 -0.914 -13.136 -1.366 -17.060 -1.830 -6.756 -0.744 -7.867 -0.890 -6.593 -0.635 AprT2 4.423 0.235 55.910 3.335 59.762 3.553 49.532 3.011 29.247 1.831 16.563 0.925 Ju1T2 -53.517 -1.372 31.047 0.896 40.508 1.193 -71.140 -2.175 -9.499 -0.300 -70.777 -1.934 OctT2 -51.205 -2.322 -57.033 -2.931 -66.498 -3.396 -40.212 -2.087 -48.465 -2.636 -54.614 -2.685 JanR 6.881 0.844 -0.209 -0.029 8.790 1.230 17.050 2.464 2.786 0.414 11.458 1.516 AprR -2.176 -0.894 -7.413 -3.466 -3.726 -1.755 -5.190 -2.508 -0.594 -0.297 -5.263 -2.316 JuIR -0.977 -1.787 -0.195 -0.400 -1.249 -2.597 -0.045 -0.095 -0.416 -0.896 -0.275 -0.532 OctR 4.053 1.437 4.692 1.877 3.600 1.445 1.065 0.435 -0.483 -0.204 4.288 1.641 JanR 2 -0.935 -3.689 -0.493 -2.129 -0.625 -2.717 -0.643 -2.914 -0.663 -3.110 -0.872 -3.587 AprR2 0.029 1.384 0.034 1.847 0.034 1.865 0.031 1.656 0.022 1.236 0.040 2.078 JuIR2 0.000 0.247 0.001 1.471 0.001 0.696 0.000 0.262 0.001 1.348 0.000 0.012 OctR2 0.030 1.307 0.038 1.876 0.032 1.613 0.027 1.352 0.033 1.728 0.009 0.432 JanInt -13.837 -4.443 -6.036 -2.147 -4.769 -1.747 -3.966 -1.493 -7.371 -2.886 -9.581 -3.183 Aprlnt 3.251 3.357 4.037 4.715 3.890 4.573 3.808 4.471 3.267 3.937 3.315 3.619 Julint -0.893 -2.437 0.279 0.859 0.072 0.223 -0.324 -1.026 -0.055 -0.180 -0.714 -2.112 OctInt -0.156 -0.126 -2.375 -2.186 -1.348 -1.245 -3.754 -3.544 -1.463 -1.429 -1.157 -1.020 Soill -24.814 -0.285 -57.573 -0.727 -32.276 -0.411 37.744 0.489 -63.087 -0.846 -4.044 -0.049 Soil2 378.790 3.235 196.120 1.904 366.770 3.471 258.290 2.564 163.910 1.719 286.010 2.732 Soil3 -20.996 -0.140 -254.330 -1.879 -302.520 -2.186 -267.020 -1.968 -270.530 -2.071 -255.760 -1.778 Soil4 75.221 0.608 173.170 1.546 170.310 1.626 18.329 0.170 136.070 1.299 -40.725 -0.349 Soil5 176.250 0.741 -144.530 -0.677 -106.600 -0.490 -124.360 -0.589 -1.481 -0.009 -10.226 -0.056 Soil6 309.650 1.205 162.120 0.698 180.060 0.764 100.170 0.440 83.967 0.446 131.380 0.632 Cultiv 23.287 0.146 267.180 1.715 414.790 2.350 136.790 0.705 176.810 0.893 286.660 1.233 Bullock 750.800 3.975 -193.860 -1.130 2.670 0.015 37.815 0.197 -238.380 -1.240 134.290 0.594 Tractor 16778.00 0.488 53422.00 1.982 56818.00 2.271 28578.00 1.367 39062.00 2.277 77547.00 4.739 Popden 12.289 0.341 80.828 2.546 30.684 1.017 -0.097 -0.003 52.689 1.789 44.184 1.424 Litprop 1310.900 2.485 551.730 1.198 526.780 1.146 1286.600 2.905 657.650 1.591 1122.800 2.488 Hyvfr 315.320 0.823 396.050 1.213 270.410 0.943 157.300 0.558 77.071 0.287 -735.830 -2.682 Alt -0.127 -0.443 -0.193 -0.753 -0.111 -0.432 -0.374 -1.527 -0.304 -1.287 -0.200 -0.777 Const 112.380 0.302 676.860 2.056 782.400 2.343 864.050 2.654 531.570 1.838 657.630 2.073 Adj.R 0.4819 0.6329 0.604 0.6437 0.5563 0.5365 Note:Under each year, the first column provides the coefficient and the second one the corresponding t-ratio 168 Table 5.4: Break-up of impacts under various climate change scenarios Climate Impacts Scenario January April July October Temperature Rainfall Total I -29.563 -18.546 -46.507 90.436 -4.182 0 -4.182 if -38.106 -18.355 -62.439 104.018 -18.172 3.265 -14.882 III -47.746 -14.889 -76.790 114.339 -32.414 7.264 -25.852 IV -57.274 -10.676 -95.656 120.754 -50.190 7.264 -42.852 Table 5.5: Possible impacts due to different climate change scenarios Pooled Analysis Averaged Analysis Repeated Cross-Sectional Analysis I II III IV I II III IV I II III IV Total Impacts -5.495 -15.031 -23.905 -43.269 -4.182 -14.880 -25.086 -42.852 -5.509 -13.999 -23.157 -39.642 As Percent of Total Revenue -3.176 -8.685 -13.813 -24.991 -2.410 -8.574 -14.453 -24.689 -3.085 -7.821 -13.061 -22.403 As Percent of AGDP -0.745 -2.037 -3.241 -5.865 -0.567 -2.017 -3.401 -5.809 -0.747 -1.898 -3.139 -5.352 As Percent of GDP -0.236 -0.644 -1.026 -1.857 -0.179 -0.639 -1.077 -1.839 -0.236 -0.601 -0.994 -1.695 Note: Total impacts are in billions of rupees (1980 prices); AGDP represents agricultural GDP; both AGDP and GDP correspond to the 1990 economy. Table 5.6: Net revenue regression for the period 1970-1980, pooled procedure with and without weather terms - analysis II Variable Name (Abbreviation) Without Weather Terms With Weather Terms Coefficient T-ratio Coefficient T-ratio January Temperature (JanT) -56.245 -2.981 -51.588 -2.705 April Temperature (AprT) -178.624 -8.735 -183.949 -8.739 July Temperature (JulT) -127.858 -3.471 -124.959 -3.356 October Temperature (OctT) 388.030 8.181 381.156 7.970 January Temperature Square (JanT2) -9.626 -3.197 -10.017 -3.324 April Temperature Square (AprT2) 23.671 4.449 25.187 4.669 July Temperature Square (JulT2) -12.874 -1.183 -12.341 -1.119 October Temperature Square (OctT2) -49.815 -7.941 -53.151 -8.142 January Rain (JanR) 5.036 2.192 5.168 2.252 April Rain (AprR) -4.677 -6.822 -4.442 -6.406 July Rain (JuIR) -0.591 -3.789 -0.637 -4.077 October Rain (OctR) 5.286 6.592 5.266 6.570 January Rain Square (JanR2) -0.606 -8.438 -0.598 -8.337 April Rain Square (AprR2) 0.033 5.587 0.034 5.674 July Rain Square (JulR2) 0.001 2.050 0.001 1.983 October Rain Square (OctR2) 0.020 3.167 0.020 3.157 January Temperature X January Rain (JanInt) -8.088 -9.339 -8.019 -9.230 April Temperature X April Rain (AprInt) 3.225 11.703 3.289 11.892 July Temperature X July Rain (JulInt) -0.293 -2.841 -0.298 -2.888 October Temperature X October Rain (Octlnt) -0.865 -2.473 -0.939 -2.645 January Rain Development (JanDevR) -0.114 -0.218 April Rain Development (AprDevR) 0.029 0.223 July Rain Development (JulDevR) 0.134 2.265 October Rain Development (OctDevR) -0.610 -4.022 January Temperature Development (JanDevT) -42.951 -3.162 April Temperature Development (AprDevT) 15.626 0.987 July Temperature Development (JulDevT) 29.779 1.627 OctoberTemperature Development (OctDevT) 10.254 0.612 170 Table 5.6: Net revenue regression for the period 1970-1980: pooled procedure with and without weather terms - analysis II (cont.) Variable Name (Abbreviation) Without Weather Terms With Weather Terms Coefficient T-ratio Coefficient T-ratio Soil Type Dummy (Soill) 62.909 2.523 59.182 2.378 Soil Type Dummy (Soil2) 335.556 10.147 334.239 10.142 Soil Type Dummy (Soil3) -216.848 -5.011 -218.035 -5.047 Soil Type Dummny (Soil4) 69.964 1.985 69.095 1.962 Top Soil Depth Dummy (SoilS) -78.395 -1.202 -74.526 -1.145 Top Soil Depth Dummy (Soil6) 125.031 1.759 126.274 1.782 Fraction of Cultivators (Cultiv) 89.017 2.120 92.602 2.211 Number of Bullocks per Hectare (Bullock) 106.468 2.001 108.278 2.036 Number of Tractors per Hectare (Tractor) 39279.390 5.418 40480.100 5.579 Population Density (Popden) 54.322 5.303 53.539 5.238 Literacy Proportion (Litprop) 578.029 3.902 560.647 3.771 Fraction of HYV (Hyvfr) 60.295 0.585 61.158 0.591 Altitude (Alt) -0.330 -4.118 -0.329 -4.112 Year Dummy (dy7O) 357.037 7.641 360.141 7.190 Year Dummy (dy7l) 308.242 6.668 296.661 5.443 Year Dummy (dy72) 302.992 6.632 310.717 6.265 Year Dummy (dy73) 561.541 12.682 542.441 9.682 Year Dummy (dy74) 525.985 11.683 541.288 10.427 Year Dummy (dy75) 477.014 10.878 491.465 9.098 Year Dummy (dy76) 334.455 7.581 379.796 7.738 Year Dummy (dy77) 393.051 9.089 395.817 8.318 Year Dummy (dy78) 315.564 7.367 340.790 6.816 Year Dummy (dy8O) 288.656 6.752 326.925 6.914 Constant (Const) 484.052 4.478 488.322 4.439 Adjusted R-square 0.5411 0.5785 171 Table 5.7: Net revenue regressions for the period 1970 to 1973: repeated cross-sectional procedure - analysis II 1970 1971 1972 1973 JanT -177.000 -2.090 -32.505 -0.428 -245.780 -3.143 -72.421 -0.797 AprT -187.450 -2.205 -30.044 -0.346 95.128 1.084 -141.810 -1.569 JuIT -253.970 -1.824 -60.618 -0.379 -339.070 -2.551 -224.980 -1.599 OctT 663.490 3.785 290.240 1.365 386.780 2.100 178.470 1.096 JanT2 -19.642 -1.939 -7.086 -0.662 5.959 0.551 -13.538 -1.196 AprT2 49.064 2.422 29.253 1.322 -11.061 -0.534 -25.322 -1.179 julT2 24.897 0.673 -2.494 -0.061 14.165 0.349 3.944 0.097 OctT2 -6.384 -0.201 -31.325 -0.797 2.647 0.080 -6.200 -0.177 JanR 1.223 0.124 10.211 1.189 -18.095 -1.093 -1.784 -0.169 AprR -8.650 -2.330 -4.106 -1.350 -6.196 -2.424 -0.425 -0.137 JuIR -0.960 -1.966 0.132 0.226 -0.287 -0.463 -0.329 -0.573 OctR 14.098 4.973 7.900 2.291 13.458 4.246 4.978 1.392 JanR2 -0.377 -1.785 -0.375 -1.555 0.052 0.196 -0.448 -1.549 AprR2 0.064 3.474 0.045 2.131 0.018 0.893 0.011 0.588 JuIR2 0.001 0.'764 0.000 -0.038 0.001 0.952 0.002 2.317 OctR2 -0.066 -3.468 0.001 0.032 -0.004 -0.196 0.025 1.198 Janlnt -5.543 -2.103 -3.869 -1.335 -2.899 -0.956 -8.636 -2.067 Aprlnt 3.902 4.170 2.765 2.728 1.794 1.804 1.985 2.144 JulInt -0.047 -0.141 -0.390 -1.052 -0.196 -0.501 0.079 0.226 Octlnt 0.885 0.650 1.346 0.776 -0.968 -0.717 -2.212 -1.476 JanDevR 0.540 0.099 11.272 2.759 27.321 1.415 33.411 3.321 AprDevR -1.541 -1.350 4.669 1.231 4.393 1.532 8.528 2.725 JulDevR 0.178 0.271 -2.539 -2.865 -0.260 -0.209 -0.003 -0.003 OctDevR -3.746 -1.688 5.639 2.220 0.622 0.344 -0.200 -0.081 JanDevT -388.960 -2.872 -102.230 -0.525 -44.026 -0.214 133.600 0.841 AprDevT. 621.050 4.425 85.917 0.723 -75.074 -0.470 85.133 0.721 JulDevT -387.180 -2.667 -15.927 -0.093 -160.050 -0.921 245.310 1.607 OctDevT -338.620 -1.691 451350 2.290 386.990 2.204 -382,910 -1.510 Soill 171.050 2.481 222.630 2.793 193.290 2.434 125.470 1.571 MMiI2 516.720 5.773 417.470 3.914 199.120 1.857 310.140 2.969 Soil3 -179.060 -1.508 -114.710 -0.857 -131.640 -0.931 -168.460 -1.233 Soi14 -52.679 -0.538 58.675 0.529 13.829 0.118 108.240 0.976 Soil5 -160.820 -0.903 85.275 0.421 -197.690 -0.942 -13.938 -0.068 Soil6 -90.981 -0.473 221.450 1.010 49.230 0.217 290.900 1.307 Cultiv 43.156 0.464 67.092 0.705 199.880 1.935 352.450 2.998 Bullock 33.043 0.208 -120.570 -0.735 -51.214 -0.347 105.960 0.676 Tractor 98280.000 2.412 121640.000 2.969 64648.000 1.854 107500.0 3.092 Popden 91.977 2.746 94.650 2.473 58.785 1.558 164.900 4.751 Litprop -632.750 -1.285 -859.850 -1.514 189.480 0.364 -1432.900 -2.551 Hyvfr 462.100 0.845 556.770 1.105 922.420 1.974 345.480 0.810 Alt -0.307 -1.414 -0.553 -2.272 -0.719 -2.934 -0.074 -0.300 Const 1429.000 4.768 1275.000 3.528 626.480 1.367 1447.400 3.344 Adj.R2 0.6786 0.6456 0.6486 0.6232 172 Table 5.7: Net revenue regressions for the period 1974 to 1977: repeated cross-sectional procedure - analysis II (cont.) 1974 1975 1976 1977 JanT -132.130 -1.555 -81.938 -0.836 142.600 1.438 -74.549 -0.879 AprT 100.030 1.225 -71.433 -0.768 -238.450 -2.797 -264.22 -3.161 JuIT -169.880 -1.042 -179.450 -0.853 27.615 0.167 -245.99 -1.633 OctT 247.940 1.258 214.010 0.704 90.056 0.416 466.610 2.618 JanT2 -10.383 -0.825 -17.561 -1.146 -43.845 -3.775 -54.921 -5.461 AprT2 45.112 2.214 4.806 0.195 78.043 4.044 56.718 3.178 JulT2 23.365 0.523 18.157 0.379 72.627 2.035 110.480 2.557 OctT2 -49.558 -1.540 -26.531 -0.542 -66.340 -2.126 -80.830 -3.549 JanR 5.703 0.324 4.975 0.391 -16.025 -1.389 -7.427 -0.961 AprR -2.726 -0.688 0.101 0.028 -7.955 -3.608 0.042 0.013 JulR -0.599 -0.921 -0.913 -1.307 -0.957 -1.699 -1.978 -3.956 OctR 9.831 3.323 1.799 0.550 7.537 2.607 4.918 1.775 JanR2 -0.181 -0.741 -0.757 -2.677 -0.553 -2.267 -0.689 -2.824 AprR2 0.038 1.989 0.023 0.955 0.047 2.721 0.034 1.933 JulR2 0.001 1.113 0.001 0.599 0.000 -0.011 0.001 0.609 OctR2 0.017 0.818 0.025 1.007 -0.022 -1.067 -0.025 -1.258 Janlnt -2.710 -0.895 -9.014 -2.305 -10.324 -2.974 -13.284 -4.139 Aprlnt 4.002 4.258 2.482 2.209 3.678 4.395 2.661 3.182 JulInt -0.227 -0.622 -0.407 -0.926 0.050 0.156 0.265 0.843 OctInt -1.009 -0.689 1.424 0.973 -1.242 -1.016 0.045 0.029 JanDevR -1.715 -0.089 -5.469 -0.727 -2.833 -0.222 0.437 0.263 AprDevR 5.038 1.747 2.985 1.748 0.435 0.602 2.666 1.226 JulDevR -0.216 -0.621 0.211 0.326 -1.180 -1.716 0.590 1.028 OctDevR -3.198 -1.988 -2.613 -1.425 6.166 4.434 -6.817 -6.399 JanDevT -160.360 -1.008 -153.990 -0.768 -514.580 -4.023 -566.33 -3.403 AprDevT 58.679 0.450 208.820 1.704 347.550 2.939 498.030 3.671 JulDevT 341.130 3.042 -273.110 -1.281 -241.300 -1.807 -120.76 -0.896 OctDevT 213.010 1.483 -10.956 -0.061 -204.080 -1.231 -367.15 -3.218 Soil 155.480 2.036 -20.743 -0.227 -43.679 -0.569 -89.882 -1.207 Soil2 175.280 1.728 357.180 2.957 247.950 2.639 343.410 3.485 Soil3 -149.250 -1.089 -174.310 -1.095 -195.340 -1.505 -295.31 -2.211 Soil4 144.890 1.346 101.860 0.821 247.940 2.363 140.970 1.354 SoilS 182.110 0.814 203.820 0.859 -80.242 -0.410 -41.880 -0.208 Soil6 407.610 1.729 370.380 1.447 204.690 0.967 258.620 1.189 Cultiv 528.330 4.417 91.476 0.572 246.850 1.671 531.890 3.196 Bullock -252.670 -1.554 842.300 4.248 -267.910 -1.674 -45.716 -0.269 Tractor 81870.000 2.563 21557.000 0.628 53410.000 2.165 56223.0 2.411 Popden 88.303 2.536 39.669 1.067 155.540 4.892 59.212 2.006 Litprop -290.860 -0.574 790.960 1.317 -258.740 -0.586 -267.71 -0.588 Hyvfr 1320.300 3.591 484.240 1.191 947.190 2.913 547.270 1.988 Alt -0.296 -1.229 -0.095 -0.328 -0.074 -0.315 -0.081 -0.339 Const 749.710 1.440 489.970 0.976 252.380 0.719 1078.70 3.076 0.4973 0.6989 0.6709 173 Table 5.7: Net revenue regressions for the period 1974 to 1977: repeated cross-sectional procedure - analysis II (cont.) 1978 1979 1980 JanT 138.550 1.493 53.537 0.563 12.315 0.156 AprT -536.640 -5.596 -497.200 -4.295 -373.000 -3.950 JuIT 66.040 0.427 -114.260 -0.749 -126.740 -0.950 OctT 432.640 2.408 439.930 1.908 516.720 3.033 JanT2 -19.464 -1.980 -10.302 -1.001 -3.768 -0.329 AprT2 89.893 4.311 48.552 2.749 38.794 1.933 JuIT2 -44.630 -1.116 37.916 0.943 -21.708 -0.545 OctT2 -48.125 -1.474 -69.976 -2.044 -124.500 -3.694 JanR 15.301 1.734 -2.230 -0.287 12.668 1.272 AprR -3.139 -1.106 1.162 0.268 -14.690 -3.572 JuIR -0.969 -1.641 -1.145 -1.572 -0.262 -0.414 OctR 0.146 0.048 -0.079 -0.022 2.305 0.760 JanR2 -0.644 -2.831 -0.549 -2.257 -0.615 -2.394 AprR2 0.037 1.820 0.027 1.458 0.050 2.540 JulR2 -0.001 -0.562 0.001 1.119 0.000 -0.225 OctR2 0.014 0.642 0.025 1.263 0.001 0.061 Janlnt -1.375 -0.448 -5.957 -1.580 -7.753 -2.278 AprInt 4.198 4.263 3.665 4.158 4.276 4.458 JulInt -0.459 -1.296 0.004 0.011 -0.587 -1.640 OctInt -3.165 -2.221 -1.691 -1.277 -2.234 -1.575 JanDevR 1.736 0.278 -0.861 -0.145 -22.650 -1.674 AprDevR -1.415 -0.725 1.607 0.285 14.030 3.266 JulDevR 0.782 0.918 0.230 0.237 -0.001 -0.002 OctDevR -0.292 -0.113 1.138 0.450 -1.681 -1.259 JanDevT -470.660 -2.462 -294.360 -1.607 -414.720 -2.415 AprDevT 574.340 4.349 365.520 2.154 252.650 1.501 JulDevT 51.785 0.293 177.990 1.145 335.060 1.721 OctDevT -211.210 -1.204 -267.000 -1.381 -235.440 -1.616 Soil 85.397 1.069 -97.013 -1.279 -28.871 -0.353 Soil2 292.770 2.899 200.980 2.053 259.060 2.472 Soil3 -291.750 -2.123 -315.130 -2.270 -286.810 -1.982 Soil4 159.230 1.401 231.820 2.144 -45.784 -0.386 SoilS -2.490 -0.012 72.617 0.432 7.028 0.038 Soil6 150.080 0.672 138.840 0.743 151,280 0.740 Cultiv 114.290 0.574 174.460 0.861 373.100 1.603 Bullock 32.699 0.171 -351.920 -1.762 111,070 0.499 Tractor 26542.000 1.298 44400.000 2.604 80645.000 4.980 Popden -20.605 -0.649 50.200 1.699 56.673 1.833 Litprop 1191.800 2.461 -60.833 -0.133 661.950 1.382 Hyvfr 348.830 1.205 453.760 1.617 -313.260 -1.062 Alt -0.231 -0.948 -0.332 -1.397 -0.222 -0.879 Const 1011.000 2.787 494.130 1.420 411.460 1.015 Adj.R2 0.6667 0.5755 0.5678 Note: Under each year, the first column provides the coefficient and the second one the corresponding t-ratio. 174 Figure 5.7: Comparison of IMD and FAO normal temperature (Bareilly Station) 35 30- 25 20 28 151 ~ ~ ~ ~ ~ ~ ~ ~~A lO-- 15 10 1 2 3 4 5 6 7 8 9 10 11 12 35 - 30- 25 20 ] 15 10 5 1 2 3 4 5 6 7 8 9 10 11 12 175 1980 Rs. CJ* CD CJ o o 0 °0oS 1966 I| i 1967 1968 1969 - 1970 -o 0 1971 1972 -.a 1973 1974 1975 1976 18978 8 1979 1980 1981 1982 UU 1983 19840 1985 19860 = == I I 0 0 00 0O -.S.S q._ aO O O Andhra Pradesh. Haryana Madhya Pradesh Maharastra. Kamataka Cb Punjab Tamil Nadu Uttar-Pradesh Bihar fA Gujarat a Rajasthan West Bengal i mndi I lo 0 0 (1a %e C) O 0 0 0 0 CDe Andhra Pradesh Haryn Madhya Pradesh Maharashtra Karnataka ~~~~~~~~~~~~~~~~~~~~~~~~~Tamil Nad_o 00 Uttar Pradesh India~~~~~~~~ Bihar = l Gujarat _ o Rajasthhan Orissad West Bengai India ~ U 00~'.'? OY ... , Y. ,.- ....... iMM.fi.e cy . w. SyS.S m SEg eD El 3 sc M Figure 5.11: Repeated cross sectional procedure 2000 *-,...., ._..... _ 1500 Go 1000.- _ NRIHa.(Obs) --m-NR/Ha.(Pred.) 500 -*-Scenario 11 1 1974 \7VD 1976 1978 10 -500 2000 en 1500 _ _ l X 1000. _ NR/Ha.(Obs) cn ~~~~~~~~~~~~~~~~NR/Ha.(Pred.) 500 -*-Scenario 11 1 !70 1971 1~~ 19'74 7V I 9'S 1#7 1978 1 0 -500 179 APPENDiX A: CHARACTERISTICS OF METEOROLOGICAL STATIONS USED IN THE ANALYSES Table 5.A.1: Meteorological stations' data Station Number Station Name Latitude Longitude Altitude DSEA 160 Stn Set (deg.min.) (deg.min.) (meters) (miles) (FAO data set) 42027 Srinagar 34.05 74.50 1587.00 801.03 yes 42056 Jammu 32.40 74.50 367.00 715.20 yes 42062 Dharamashala 32.16 76.23 1211.00 710.00 no 42071 Amritsar 31.38 74.52 234.00 646.03 yes 42083 Simla 31.06 77.10 2202.00 660.06 yes 42099 Ludhiana 30.56 75.52 247.00 630.69 yes 42101 Patiala 30.20 76.28 251.00 612.48 no 42103 Ambala 30.23 76.46 272.00 607.06 yes 42111 Dehradun 30.19 78.02 682.00 636.98 yes 42123 Ganganagar 29.55 73.53 177.00 519.24 yes 42131 Hissar 29.10 75.44 221.00 533.26 yes 42147 Mukteswar 29.28 79.39 2311.00 653.56 yes 42165 Bikaner 28.00 73.18 224.00 396.29 yes 42170 Churu 28.15 74.55 291.00 420,00 no 42182 New-Delhi 28.35 77.12 216.00 508.64 yes 42189 Bareilly 28.22 79.24 173.00 680.88 yes 42260 Agra 27.10 78.02 169.00 461.64 yes 42273 Bahraich 27.34 81.36 124.00 417.30 yes 42309 North-Lakhimpu 27.14 94.07 102.00 340.00 no 42314 Dibrugarh 27.29 95.01 111.00 392.23 yes 42328 Jaisalmer 26.54 70.55 242.00 281.42 no 42339 Jodhpur 26.18 73.01 217.00 275.60 yes 42343 Ajmer 26.27 74.37 486.00 320.72 yes 42348 Jaipur 26.49 75.48 390.00 370.61 yes 42361 Gwalior 26.14 78.15 207.00 437.68 yes 42369 Lucknow 26.45 80.53 128.00 538.35 yes 42379 Gorakhpur 26.45 83.22 77.00 292.09 yes 42410 Gauhati 26.06 91.35 54.00 225.37 yes 42435 Barmer 25.45 71.23 194.00 197.29 yes 42452 Kota 25.09 75.51 274.00 275.95 yes 42463 Jhansi 25.27 78.35 251.00 421.34 yes 42475 Allahabad 25.27 81.44 98.00 433.90 yes 42483 Varanasi (Bhu) 25.18 83.01 76.00 385.30 yes 42492 Patna 25.36 85.06 60.00 163.36 yes 42515 Cherrapunji 25.15 91.44 1313.00 173.37 yes 42516 Shillong 25.34 91.53 1598.00 198.88 yes 42542 Udaipur(Dabok) 24.37 73.53 514.00 180.00 no 42559 Guna 24.39 77.19 478.00 323.52 yes 42571 Satna 24.34 80.50 317.00 453.03 yes 42577 Sidhi 24.25 81.52 272.00 390.00 no 42587 Dattonganj 24.03 84.04 221.00 246.42 no 42591 Gaya 24.45 84.57 116.00 174.38 yes 42620 Silchar 24.55 92.59 97.00 162.63 yes 180 Table 5.A.1: Meteorological stations' data (cont.) Station Number Station Name Latitude Longtitude Altitude DSEA 160 Stn Set (deg.min.) (deg.min.) (meters) (miles) (FAO data set) 42623 Imphal 24.46 93.54 781.00 200.00 no 42631 Maliya 23.15 68.51 21.00 19.88 no 42634 Bhuj 23.15 69.40 80.00 28.89 yes 42647 Ahmedabad 23.04 72.38 55.00 58.95 yes 42667 Bhopal 23.17 77.21 523.00 301.65 yes 42675 Jabalpur 23.12 79.57 393.00 463.64 yes 42701 Ranchi 23.19 85.19 652.00 171.48 yes 42704 Asansol 23.28 86.26 74.00 91.27 yes 42708 Sri Niketan 23.39 87.42 59.00 120.00 no 42724 Agartala 23.53 91.15 16.00 50.00 no 42730 Okha 22.29 69.07 7.00 5.00 no 42737 Rajkot 22.18 70.47 138.00 49.48 yes 42754 Indore 22.43 75.48 567.00 194.47 yes 42763 Hoshangabad 22.46 77.46 302.00 318.65 yes 42779 Pendra 22.46 81.54 625.00 337.28 yes 42807 Calcutta 22.39 88.27 6.00 61.18 yes 42809 Calcutta-Dumdu 22.36 88.30 5.00 61.18 no 42830 Porbandar 21.39 69.40 7.00 5.00 no 42867 Nagpur 21.06 79.03 309.00 360.00 no 42875 Raipur 21.14 81.39 298.00 248.24 yes 42886 Jharsuguda 21.55 84.05 230.00 190.G0 no 42895 Balasore 21.31 86.56 20.00 10.05 yes 42909 Veraval 20.54 70.22 8.00 5.64 yes 42916 Daman 20.25 72.51 12.00 8.00 no 42920 Ozar 20.08 73.55 608.00 70.00 no 42934 Akola 20.42 77.04 309.00 278.65 yes 42971 Bhubaneshwar 20.15 85.50 46.00 30.00 no 43001 Dahanu 19.58 72.43 5.00 5.00 no 43003 Bombay-Santacr 19.07 72.51 15.00 9.00 yes 43014 Aurangabad 19.51 75.24 579.00 177.01 yes 43041 Jagdalpur 19.05 82.02 553.00 126.04 yes 43049 Gopalpur 19.16 84.53 17.00 2.55 yes 43053 Puri 19.48 85.49 6.00 4.91 yes 43057 Bombay 18.54 72.49 11.00 10.00 no 43063 Pune 18.32 73.51 559.00 70.82 yes 43081 Nizamabad 18.40 78.06 381.00 345.79 yes 43086 Ramgundam 18.46 79.26 156.00 230.00 no 43105 Kalingapatnam 18.20 84.08 6.00 1.00 yes 43110 Ratnagiri 16.59 73.20 67.00 7.00 no 43111 Mahabaleswar 17.56 73.40 1382.00 51.43 yes 43117 Sholapur 17.40 75.54 479.00 186.12 yes 43128 Hyderabad 17.27 78.28 545.00 174.41 yes 43149 Vishakhapatnam 17.43 83.14 3.00 3.59 yes 43181 Gannavaram 16.32 80.48 24.00 40.00 no 181 Table 5.A.1: Meteorological stations' data (cont.) Station Number Station Name Latitude Longitude Altitude DSEA 160 Stn Set (deg.min.) (deg.min.) (meters) (miles) (FAO data set) 43185 Musulipatnam 16.11 81.08 3.00 2.13 yes 43189 Kakinada 16.57 82.14 8.00 2.83 yes 43192 Panjim 15.29 73.49 60.00 6.00 no 43196 Mormugao 15.25 73.47 62.00 4.00 yes 43198 Belgaum 15.51 74.37 747.00 60.94 yes 43201 Gadag 15.25 75.38 650.00 116.78 yes 43213 Kumool 15.50 78.04 281.00 128.33 yes 43220 Baptata 15.54 80.28 6.00 10.00 no 43225 Karwar 14.47 74.08 4.00 5.00 no 43226 Honavar 14.17 74.27 26.00 3.10 yes 43233 Chitradurga 14.14 76.26 733.00 124.51 yes 43237 Anantpur 14.41 77.37 350.00 170.00 no 43245 Nellore 14.27 79.59 20.00 12.13 yes 43257 Agumbe 13.30 75.06 659.00 30.00 no 43279 Madras (Meenam 13.00 80.11 16.00 3.38 yes 43284 Mangalore 12.55 74.53 102.00 1.00 yes 43295 Bangalore 12.58 77.35 921.00 171.43 yes 43311 Amini 11.07 72.44 4.00 1.00 yes 43314 Kozhikode 11.15 75.47 5.00 7.31 yes 43321 Coimbatore 11.02 77.03 400.00 74.34 yes 43329 Cuddalore 11.46 79.46 12.00 1.03 yes 43333 Port-Blair 11.40 92.43 79.00 5.00 no 43339 Kodaikanal 10.14 77.28 2343.00 84.12 yes 43344 Tiruchirapalli 10.46 78.43 88.00 179.66 yes 43346 Karaikal 10.55 79.50 7.00 4.00 no 43347 Nagapattinam 10.46 79.51 9.00 3.70 yes 43348 Adirampattinam 10.20 79.23 6.00 i0.00 no 43361 Tondi 9.44 79.02 5.00 5.00 no 43363 Pamban 9.16 79.18 11.00 1.00 yes 43369 Minicoy 8.18 73.00 2.00 1.00 yes 43371 Trivandrum 8.29 76.57 64.00 6.83 yes 43377 Kanyakumari 8.05 77.30 37.00 4.00 no 43379 Tuticosin S.4 78.09 4.00 7.00 no 182 RNEFERENCES Adams, R., B.McCarl, D.Dudek, and D.Glyer. 1988. "Implications of Global Climate Change for Western Agriculture." Western Journal ofAgricultural Economics, 13(2): 348-56. Adams, R., C.Rosenzweig, R.Pearl, J.Ritchie, B.McCarl, D.Glyer, B.Curry, J.James, K.Boote, and H.Allen. 1990. "Global Climate Change and U.S. Agriculture". Nature 345: 219-23. Cure, J.D., and B.Acock. "Crop Responses to Carbon Dioxide Doubling: A literature survey". Agriculture and Forest Meteorology, 38: 127-145, 1986. Intergovernmental Panel on Climate Change (IPCC). 1990. Scientific assessment of climate change: Report prepared by working group 1. New York: World Meteorological Organization and United Nations Environmental Program. _.__ * Climate Change 1995. 1996. The science of climate change. World Meteorological Organization of the United Nations Environment Program, Cambridge: Cambridge University Press. Jones, C. A., P. T. Dike, J. R. Williams, J. R. Kiniry, V. W. Benson, and R. H. Griggs. 1991. "EPIC: An Operational Model For Evaluation of Agricultural Sustainability". Agricultural Systems, 37: 1991. 341-350. Kaiser, H. M., S. J. Riha, D. S. Wilks, D. G. Rossiter, and R. Sampath. 1993. "A farm-level analysis of economic and agronomic impacts of gradual warming". American Journal of Agricultural Economics, 75: 387-398. Kane, S., J. Reilly, and J. Tobey. 1991. Climate Change: Economic Implications for World Agriculture, AER-647, U.S. Department of Agriculture. Kimball, B. A. 1983. "Carbon Dioxide and Agricultural Yield: An Assemblage and Analysis of 430 prior Observations". Agronomy Journal, 75: 779-788. 1985. "Adaptation of Vegetation and Management Practices to a Higher Carbon Dioxide World," in B. R. Strain and J. D. Currie, eds. Direct Effects of Increasing Carbon Dioxide on Vegetation. U.S. Department of Energy, Washington D.C. Kumar, K. S. Kavi, and Jyoti Parikh. "Potential Impacts Of Global Climate Change on Indian Agriculture", Communicated to Global Environmental Change, 1996. Leemans, R., and A. M. Solomon. 1993. "Potential Response and Redistribution of Crops Under a Doubled CO2 climate". Climate Research, 3: 79-96. Mendelsohn, R., W. Nordhaus, and D. G. Shaw. 1994. "The Impact of Global Warming on Agriculture: A Ricardian Analysis." American Economic Review, 84 (88): 753-771. Narayana, N. S. S ., K. S. Parikh, and T. N. Srinivasan. 1991. Agriculture, Growth and Redistribution of Income. New Delhi: North-Holland/Allied Publishers. 183 Rao, D. G. and S. K. Sinha. 1994. "Impact of Climate Change on Simulated Wheat Production in India," in C.Rosenzweig and A.Iglesias, eds. Implications of Climate Change for International Agriculture: Crop Modelling Study. Washington D.C.: US EPA. Reilly, J. 1994. "Crops and climate change". Nature, 367, pp. 118-119. Reilly, J., N. Hohmann, and S. Kane. 1994. "Climate Change and Agricultural Trade: Who Benefits, Who Loses?". Global Environmental Change, 4(1): 24-36. Rosenzweig, C. and A. Iglesias, eds. 1994. Implications of Climate Change for International Agriculture: Crop Modeling Study, Washington D.C.: U.S. Environmental Protection Agency. Rosenzweig, C. and M. L. Parry. 1994. "Potential Impact of Climate Change on World Food Supply." Nature, 367: 133-138. Sanghi, A., D. Alves, R. Evenson, and R. Mendelsohn. 1996. "Global Warming Impacts on Indian Agriculture: Estimates of the Ricardian Model." Economia Aplicada, 1(1): 7-33. Sanghi, A., R. Mendelsohn, and A. Dinar. 1998. "The climate sensitivity of Indian agriculture," Chapter 4 in this report. Senft, D. 1995. "FACE-ing the Future." Agricultural Research, 43(4): 4-7. UN Framework Convention on Climate Change (FCCC). 1992. Published by IUCc/UNEP (Information Unit Climate Change). Woodward, F. I. 1993. "Leaf responses to the environment and extrapolation to larger scales," in A. M. Solomon and H. H. Shugard, eds. Vegetation Dynamics and Global Change. International Institute for Applied Systems Analysis, New York: Chapman & Hall. 184 6 TECHNOLOGY-CLIMATE INTERACTIONS: WAS THE GREEN REVOLUTION IN INDIA CLIMATE FRIENDLY? James W. McKinsey, Jr. and Robert E. Evenson INTRODUCTION Estimates of the impact of a rise in normal temperatures and of increases in rainfall levels for different regions in India are reported in this chapter. Technology-climate interactions are incorporated into the estimates, enabling an assessment of the climate "friendliness" of the Green Revolution in Indian agriculture. The analytic model underlying our estimation specification is related to the "Ricardian" model developed by Mendelsohn, et al. (1994). In this chapter we report estimates of a "net revenue" specification of the model (which is related to, but not the same as, the Ricardian specification), in which we include variables characterizing technological change. The second section reviews the methodology underlying the specifications. The third section summarizes the data set. Sections 4 and 5 report estimates of the technology specification and climate effects thereon. Sections 6 and 7 report estimates of the net revenue specification and climate effects thereon. Section 8 reports estimates of technology effects, and section 9 reports computed "'secondary" effects of technology on the climate effects. METHODOLOGY THE MODEL We begin with a very general characterization of agricultural production using the transformation function (1) G(Y,X,H,E,C,T,I,W)=0 where Y is a vector of outputs (with prices Py); X is a vector of inputs (with prices Px); H is a vector of land allocations; E is a vector of edaphic factors; C is a vector of normal climate factors (temperature, rainfall, etc.); T is a vector characterizing available technology; I is a vector characterizing infrastructure and institutions; and W is a vector of current weather (expressed as departures from normal). The maximized variable profits function from (1) is (2) fI* = Y*Py - X*Px = P*(Py, Px, H, E, C, T, I, W) One can apply the Shephard-Hotelling lemma to (2) to obtain the output supply - factor demand system a*/aPy Y =Y(Py, Px, H, E, C, T, I, W) (3) arI*/aPx = X = X(Py, Px, H, E, C, T, ', W) Our interest is in variable profit per unit of land, which can be expressed as (4) rI */H = rI **(Py, Px, E, C, T, I, W) and which we shall call "net revenue per hectare", or simply "net revenue" (NR; measured by the variable LNOFPKRE)'. We do not observe land values, but we do observe net revenue per hectare of land; thus we estimate a version of (4) rather than (5). As described in the following section, we extend (4) in two dimensions over previous work (e.g., Sangbi et al., 1998): i. We treat T as endogenously determined by adding three equations for T to the model: (5) T = T(E, C, I*) ii. We estimate technology-climate interactions terms in the net revenue equation (4)2. MODEL STRUCTURE Given the importance of the technology specification (6) to this paper, we want to develop this further. Over the period of this study (1970/71 through 1987/88: popularly known as the "Green Revolution" period in Indian agricultural history), Indian agriculture was realizing rapid technological gains. It was characterized by three major activities: 1) the development and diffusion of "High Yielding Varieties" of cereal grains, especially wheat and rice (which we measure as WHYV, the proportion of area under the five major food crops which are planted to high yielding varieties); 2) the expansion of multiple-cropped area, i.e., area cropped more than once during a year (which we measure as Gross Cropped Area divided by Net Cropped Area, GCANCA, in which NCA is area cropped at least once during the year, and GCA is total cropped area during the year); and 3) the expansion of area under irrigation (which we measure as net irrigated area divided by net cropped area, NIANCA). It is clear, however, that these three activities cannot be treated as exogenous variables in equation (4). They must be modeled as endogenous variables and treated accordingly. Our basic econometric model is then a four equation model with the following structure: 'The Ricardian methodology for measuring climate effects treats (4) as one of the land use revenues capitalized into land values. 2 A bit of nomenclature: we call approaches and models which are almost identical to the Ricardian models, except that the dependent variable is net revenue per hectare instead of land price or land value, "Semi-Ricardian". Our approach goes beyond the Ricardian or Semi-Ricardian in that we explicitly model the process of technological change in agriculture; we call this approach "Supra-Ricardian". 186 i. Technology and related infrastructure (the first three equations) WH{YV = fi (GCANCA, NIANCA, STRES5, EXT, E, C) NIANCA = f2 (IRR57, WHYV, E, C) GCANCA= f3 (GCANCA57, WHYV, NIANCA, E, C) 2. Net Revenue (the fourth equation, following equation (4) above) LNOFPKRE = f4 (LNCSTCLT, LNCSTBUL, I, STRES5, EXT, WHYV, NIANCA, GCANCA, E, C, WHYVxC, NIANCAxC, GCANCAxC) These four variables - WHYV, NIANCA, GCANCA and LNOFPKRE - are the endogenous variables in the model. The first three are designed to capture the main features of the green revolution, where the area planted to HYVs is determined by multiple-cropping, irrigation, public agricultural research (STRES5) and extension (EXT), and edaphic (E) and climatic (C) variables. Irrigation intensity is driven by its 1957 level (IRR57), the adoption of HYVs, and edaphic and climatic variables. Multiple-cropping is driven by its 1957 level (GCANCA57), HYV adoption, irrigation intensity, and edaphic and climatic variables. T'he first three equations (the adoption of modem varieties, the index of multiple- cropping, and irrigation intensity) are estimated in a two-stage least squares framework. The predicted values of each of these three variables, based on the results of the 2SLS procedure, are then computed. Then (the logarithm of) net revenue per cropped hectare, the fourth equation, is determined by the logarithms of the imputed value of family labor per hectare (LNCSTCLT) and bullock labor per hectare (LNCSTBUL)3 research and extension, infrastructural variables (I), and the edaphic and climate variables. In addition, net revenue per hectare is also determined by the three endogenous technology variables and their interactions with climate variables. This equation is estimated by weighted Ordinary Least Squares. FUNCTIONAL FoRM The models described thus far impose no restrictions on the functional form to be estimated. We use a very general form, with quadratic and interaction terms in order to capture non-linearities and moderating impacts of technology on climate effects and vice versa. Modem varieties, multiple-cropping and irrigation intensity are estimated in the 2SLS system with linear and quadratic terms. Unpriced and poorly-measured bullock and cultivator data appear in the right-hand-side of our net revenue regression equation; that prompts us to cast those terms, as well as the dependent variable, net revenue, in logarithmic form, as a Cobb-Douglas. The climate variables already appear in quadratic (flexible) form, so the logarithmic status is not inconsistent with them. The equation is weighted by net cropped area, to control for excessive influence on the estimates by larger districts. 3 These variables are generally not marketed and priced in India. Thus this specification treats them in a production function framework. 187 ADAPTATION AND INTERACTION Expressions such as (4) allow for farmer adaptation to climate change. This adaptation includes investments in farm level irrigation and drainage as well as changes in farm practices including cropping patterns. There is, as well, potential adaptation by the organizations producing technology and infrastructure for farmers. These organizations include private firms who conduct R&D to develop improved factors to be supplied to the agricultural sector, and the public sector agricultural research and extension organizations who also provide improved technology to agriculture. It also includes public sector units providing, and maintaining, infrastructure. Implicitly, this suggests that there may be important climate-technology and climate- infrastructure interactions in (4). We cannot observe expected interactions, and there is a serious question whether actual interactions, estimated from (4), would be good predictors of future interactions. Some might argue that since climate has changed little over the past 25 years or so, the (public and private) inventors of technology and the investors in infrastructure have not responded to climate change. However, the underlying premise of the estimates obtained from cross-section data where climate varies over locations, is that these do measure responses to climate. If technology and infrastructure are exogenous to farmer decisions one may argue that climate-technology and climate-infrastructure interactions estimated from (4) are reasonable estimates of future effects. We know that researchers do respond to climate conditions. Plant breeders continually seek genetic traits to change the length of growing seasons and to endow plants with host plant tolerance (HPT) to cold and warm temperatures, to drought and flood stress, and related climate effects. Their motivation for seeking to incorporate these traits in crop varieties is to allow superior genetic material (e.g., the semi-dwarf wheat and rice) to overcome climate and edaphic barriers to their "migration" to new areas. This cross-section motive is likely to be a good proxy for a time series motive, that is, to respond to a long-term rise in temperature. Suppose, for example, that we have two regions (1 and 2) which differ in temperature (tl, t2) and edaphic factors (El, E2). Suppose that a rise in temperature damages crops in both regions if crops do not migrate. If t, rises to the former level of t2, region 1 can minimize the damage if the crops suited to region 2 migrate to region 1. This cannot happen if there are edaphic barriers to such migration. Plant breeders can mitigate these edaphic barrier effects through HPT breeding. They can also mitigate the non-migration effects through HPT breeding for high temperature tolerance. These interactions enable us to estimate two dimensions of the technology-climate relationship. In system (5) we estimate the effects of climate on the production and diffusion of technology. (That is, we can capture dT/dC.) From these estimates we can ask: Will an increase in normal temperatures and rainfall be "friendly" to the development and diffusion of agricultural technology? From the estimation of (4) with interaction terms we can ask the following question: What is the impact of an increase in rainfall and temperature on net revenue (including the effects operating through technology)? (7) dNR/dC = aNR/aC + aNR/aT * aT/aC 188 We can also capture the secondary impacts of technology on the climate effects: (8) d(dNR/dC)/dT which enables us to ask whether climate effects on net revenue per hectare were altered by changes in these variables during the period of the Green Revolution. This will tell us whether this technology was "friendly" to projected temperature and rainfall effects or "unfriendly" to them. DATA VARIABLES The dependent variable in a Semi-Ricardian or Supra-Ricardian study is "out-of-pocket" net revenue per hectare (hereafter simply "net revenue"). To compute net revenue first sum the products of all crop outputs multiplied by their farm harvest prices, then subtract the cost of purchased inputs (which are fertilizer, tractors, and hired agricultural labor), then divide the entire quantity by net cropped area. In this study we go one step further, and take the natural logarithm of the just-computed net revenue per hectare, as described above, yielding the variable LNOFPKRE. Two important inputs are excluded from this computation: bullocks and family labor. Bullocks are primarily used by their owners (with little renting in or out), as obviously is true for fiamily labor; these inputs are treated as quasi-fixed factors, and the logarithms of the imputed values of these inputs, per hectare, are included as independent variables. All variables are described in detail in Appendix A4 The variables are measured at the district level, covering nearly all (271) of the districts in the 13 major crop-producing states of India. Table 6.1 displays the variable names, descriptions and summary statistics of the major variables used. 4 Appendix A contains definitions, sources, discussions of transformations, and other information about the inputs, crop outputs, technology and infrastructure, prices, edaphic and climatic variables. 189 Table 6.1: Variables in the Supra-Ricardian framework Variable Standard Name Description Mean Deviation A. Endogenous Variables LNOFPKRE Log of Out-of-Pocket Net Revenue 6.95 .80 WHYV Modem (High-Yielding) Variety Use .28 .25 GCANCA Multiple-Cropping Index 1.24 .20 NIANCA Index of Irrigation Intensity .29 .24 B. Exogenous Variables Edaphic Variables (E) DMSnn Soil Type Dummy Variables; nn: 03 to 21 -- -- DMpHn Soil pH Dummy Variables; n: 5, 6, 8 and 9 -- -- DMTSn Topsoil Depth; n: 1, 2 and 3 -- -- Geographic Variables DMSLPn Slope Dummy Variables; n: 1 to 4 DMSEA Dummy: 1 if District is on the seacoast -- -- DMSEANEI Dummy: 1 if District abuts one on seacoast -- -- ALT Altitude of District's weather station, meters 296.89 316.36 (Normal) Climate Variables (C) JANMDT January Normal Temperature Midpoint 18.55 3.69 JANMDTSQ January Normal Temp. Midpoint Squared 357.86 138.29 JANRN January Normal Rainfall 12.38 13.14 JANRNSQ January Normal Rainfall Squared 325.75 725.31 APRMDT April Normal Temperature Midpoint 29.59 2.35 APRMDTSQ April Normal Temp. Midpoint Squared 881.23 125.86 APRRN April Normal Rainfall 16.19 22.65 APRRNSQ April Normal Rainfall Squared 775.13 2355.60 JULMDT July Normal Temperature Midpoint 28.34 2.55 JULMDTSQ July Nornal Temp. Midpoint Squared 809.90 134.44 JULRN July Normal Rainfall 245.77 234.42 JULRNSQ July Normal Rainfall Squared 115342.86 430432.59 OCTMDT October Normal Temperature Midpoint 26.05 1.88 OCTMDTSQ October Normal Temp. Midpoint Squared 682.01 90.17 OCTRN October Normal Rainfall 66.64 66.03 OCTRNSQ OctoberNormal Rainfall Squared 8800.36 15815.10 JURNCV Coef. of Variation, June Rain, 1957 - 1987 .61 .21 JARNCV Coef. of Var., July/Aug Rain, 1957 - 1987 .35 .21 190 Table 6.1: Variables in the Supra-Ricardian framework (cont.) Variable Standard Name Description Mean Deviation (Current) Weather Variables (W) JUNERAIN Actual Rainfall in June 131.46 121.98 JUAURAIN Actual Rainfall in July and August 561.55 364.05 YEARRAIN Actual Annual Rainfall 1088.33 602.14 Technology and Infrastructural Variables (T, l) WHYVNEW Proportion of Area under HYVs .29 .25 GCANCA Multiple-Cropping Index 1.24 .21 NIANCA Irrigation Intensity .29 .24 STRES5 Cumulated Stock of Agricultural Research 36.30 43.15 EXT Index of Extension Activity 7.39 5.65 LITERACY Literacy Rate, Adult Rural Males .37 .11 RELWAGE Ratio of Rural Factory Wage to Farm Wage 1.20 .60 Interaction Terms JANMDTRN January Temperature times Rainfall 213.90 231.26 APRMDTRN April Temperature times Rainfall 464.13 597.08 JULMDTRN July Temperature times Rainfall 6711.02 5267.12 OCTMDTRN October Temperature times Rainfall 1741.94 1754.22 Other DMYRnn Year Dummies; nn: 71 to 87 -- -- IRR57 Value of NIANCA in 1957 .18 .18 GCANCA57 Value of GCANCA in 1957 1.15 .15 This study applies to the years 1970/71 through 1987/88. By 1970 the use of modem high-yielding varieties of several crops had become established in nearly every district, and partly in concert with the expansion of HYV there was substantial new investment in irrigation, fertilizer distribution, research and extension activities, and so forth. The 1970s and 1980s also were the years during which the interaction of technology and infrastructure with climate was more marked and more important. The data set includes a number of interactions between climate variables, on the one hand, and some of the technological variables, on the other hand: those interactions, and the moderating influence of the technology and infrastructure investments on climatic effects which the interactions embody, are in fact among the primary foci of this study. ECONOMETRIC ISSUES Obvious econometric issues in this study involve, first, the existence of heteroscedasticity; second, the high degree of multicollinearity which necessarily inheres in data of the sort used in this study; and third, specification issues of variable inclusion or exclusion. Heteroscedasticity is prevalent in this study. But even in its presence, least squares estimators of coefficients are consistent; thus the sample sizes in the thousands are sufficient to 191 justify use of OLS. However, the standard errors of the coefficients estimated by OLS are not consistent, so the reported t-statistics would be somewhat deficient. The standard errors are estimated by White's consistent estimator5 of the least squares covariance matrix, and the resultant estimated t-ratios are consistent. The problems of multicollinearity are likely to be severe6, especially given the inclusion of so many climate terms, and their squares. Future work will include the computation of condition indices to determine just how severe the multicollinearity is. The inclusion of the climate terms is crucial and the squared terms are necessary to allow for nonlinearities in climate effects. Because of the likely rampant multicollinearity, one should use caution in interpreting, and in using, any individual coefficient estimate: its true value may substantially differ from its estimated value, and the variable may be a valid, important regressor even if the estimated t-ratio is below the customary critical value. But of crucial importance to later Sections, the computations of estimated effects of climate change, using all the estimated coefficients (whether significant or not), are likely to be valid, for any mis-estimation of the value of one coefficient is likely to be compensated for in the estimation of the values of the coefficients of the other collinear variables; thus the joint impact of all the variables together is probably much more accurate. Finally is the issue of variable inclusion or exclusion. In the 1994 Ricardian estimates of Mendelsohn et al., (1994) a cross-section of land values was regressed on climate (C), edaphic (E) and infrastructure (I) variables. Prices and technology variables were excluded on the grounds that prices did not vary in cross-section for specific commodities7 and that technology was "equally" accessible to all farmers in the United States8. The price question is less important in the Ricardian model because future expected prices are relevant. In a semi-Ricardian (net revenue) specification, prices must enter because in any given year the price of a particular commodity may be unusually high or low. Technology is more difficult. A large number of agricultural productivity studies have measured significant differences in cross-section productivity levels which are at least partly due to edaphic and climate differences. More importantly, the studies have also measured time series differences in rate of change of partial or total factor productivity change for different regions (Evenson & Hufftnan (U.S.), Avila & Evenson (Brazil), McKinsey & Evenson (India)). These differences have persisted over long periods of time and have been related to cross-section (and time-series) differences in investments in regionally oriented agricultural research programs. One might argue (as Mendelsohn et al., (1994) do) that regional differences in the productivity growth are likely to "converge" over time as technology from the leading regions is diffused to the laggard regions. If so, the regional productivity differences would not be capitalized into land values. Yet this is quite unlikely, given the nature of agricultural technology which is highly location-specific: studies of agricultural research indicate that regions with little 5 See, for example, Kathleen Segerson and Bruce L. Dixon (June 1996). 6 See again Dixon and Segerson, op. cit. Or that transport-related differentials would continue in the future. s Obviously, this precluded climate-technology interaction estimates. 192 or no research effort targeted to their particular climate and edaphic conditions remain laggard regions. And even if productivity were to converge over time, such that current productivity differences would not be capitalized into differences in land values, current net revenues would still reflect existing productivity differences. In a similar way, variables measuring current weather are important in net revenue specifications, but not in land value specifications where current weather gets averaged into climate. It is extremely difficult to construct meaningful weather indices. Rainfall affects production in a non-linear way and its effect is sensitive to the timing of planting, flowering, etc. ESTIMATES: TECHNOLOGY AND RELATED INFRASTRUCTURE In the Supra-Ricardian framework we treat the three technology variables as endogenous variables, whose predicted values then contribute to the determination of net revenue per hectare. In this Section we report estimates of the determinants of these variables, in a two-stage least squares framework. Then in Section five we report computed climate effects on technology and related infrastructure. HYV - MULTIPLE CROPPING - IRRIGATION SYSTEM The proportion of area sown to modem varieties (WVHYV), a measure of multiple- cropping (GCANCA), and irrigation intensity (NIANCA) were estimated by two-stage least squares. Appendix Table 6B. 1 displays the variables used as instruments in the 2SLS system; an asterisk following a variable name denotes that that variable also appears in one or more second- stage regression equations, and/or the net revenue equation. These instruments include fundamental climatic and edaphic variables, as well as two technology variables and a number of price ratios proxying institutional factors. Appendix Tables 6B.2 through 6B.4 then display the regression results from the second stage; -WHYV in Table 6B.2, GCANCA in Table 6B.3, and NIANCA in Table 6B.4. A number of striking results emerge. First is the degree to which this system captures the modeled behavior. Grossly, all three second-stage regressions have highly significant F-statistics, and adjusted R2s. For each of the three regressions, we tested the null hypotheses that the rainfall variables taken as a group did not significantly influence the technology variables, that the temperature variables taken as a group did not significantly influence the technology variables, and that the climate variables (that is, the temperature and rainfall variables combined) taken as a group did not significantly influence the technology variables. The results of the F-tests of sets of excluded variables are reported in Appendix Table 6B.5. All three of the null hypotheses were rejected in all three of these regressions: as groups, the climate variables do significantly influence the adoption of modem varieties, irrigation intensity, and the extent of multiple-cropping. But even more important than the general goodness-of-fit of these regression equations, and the significance of groups of variables, are the patterns revealed within each equation. 193 Slope significantly influences both multiple-cropping and, to a lesser extent, irrigation intensity9 And irrigation intensity tends to be higher in districts above aquifers which are geologically thickest"0. The second-stage variables exercise an important influence on each other: the coefficients of both GCANCA and NIANCA on WHYV are significantly positive, as are the coefficients of both WHYV and NIANCA on GCANCA and the coefficient of WHYV on NIANCA. That is to say, the adoption of modem high-yielding varieties, multiple-cropping and irrigation are mutually-reinforcing. The adoption of modem varieties also responds favorably to greater extension activity; perhaps surprisingly, though, additional state-level agricultural research does not significantly increase the adoption of modem varieties. There is considerable inertia in this behavior: both the extent of multiple-cropping and irrigation intensity are highest in those districts in which such activity was largest in 1957". CLIMATE CHANGE EFFECTS ON TECHNOLOGY AND INFRASTRUCTURE Based on the regression results reported in the previous Section, and the actual district- level values of the climate and technology variables, one can compute the predicted effects on high-yielding variety use, multiple cropping, and irrigation intensity of changes in normal temperature and rainfall. The motivation for these computations is the by-now familiar predictions of global warming; we make no effort to validate or calibrate the predicted changes, but merely use the familiar predictions to drive a simulation. A number of possible values for temperature and rainfall change have been proposed; no scenario has received unanimous agreement. We predict the effects of a one degree Celsius temperature increase, and a three percent rainfall increase, values which some models predict could be achieved from a generation to a century from now. The impacts of different changes could easily be scaled"2. These predicted effects are presented in Table 6.2, in which appear first the temperature effects on the adoption of modem varieties, the extent of multiple-cropping, and irrigation intensity. Next appear the rainfall effects on WHYV, GCANCA and NIANCA. 9 This may reflect the importance of drainage to avoid waterlogging or soil salinity; it may reflect the geological requirements for proper functioning of a canal system. '° This does not measure the annual water depth within the aquifer, but rather a long term geological potential. Farmers may respond to this in their cropping choices; farmers and probably governments also respond in their irrigation investments. '" There is no related variable for the use of modern varieties, because no such modern variety existed before the onset of the Green Revolution in the middle 1 960s. 12 Most global climate change scenarios also posit an increase in atmospheric (and thus soil-based) carbon - usually, in fact, the climate change is initiated by an increase in atmospheric CO2. Nearly every experimental crop model predicts higher crop yields associated with increases in available carbon. This study deals with changes in temperature and rainfall,, but not with changes in carbon; thus the actual effects on crop output (and thus on net revenue) of climate change, taking into account carbon changes as well as temperature, rainfall and okta changes, would almost certainly be more beneficial than the results of this study predict. 194 Predicted impacts are computed for each district; they are presented in Table 6.2 as state-wide and national averages". We first consider the impact of a one degree Celsius rise on the technology variables. The national average temperature impact on HYV adoptions is negative but relatively small. It is most negative in Gujarat and is actually positive in a belt of states stretching from Kamataka on the southwest coast, through Andhra Pradesh, Orissa, and West Bengal moving up the east coast. The temperature impact on multiple-cropping is positive on average, although some regions (notably Karnataka, and to a lesser extent the entire nation from the Deccan plateau south: Madhya Pradesh, Maharashtra, Orissa, Andhra Pradesh and Tamil Nadu) suffer a negative impact. The temperature impacts on irrigation are uniformly negative. The effects of an increase in rainfall on HYV adoption and irrigation intensity are negative but small throughout India except in the Punjab. Similarly, increased rainfall has negative and relatively small effects on multiple-cropping in every state. Thus, these estimates generally show an increase in temperature is "unfriendly" to HYV adoption and to irrigation. A temperature increase is on balance friendly to multiple-cropping. Increased rainfall is unfriendly to all three technology variables. The regional impact estimates show that these potential temperature and rainfall effects are least unfriendly in the states where the early Green Revolution gains in wheat were made. 13 And see Appendix C, in which are displayed the temperature, rainfall and cloud cover effects by season, revealing interesting alternating patterns. 195 Table 6.2: Climate effects on technology and related infrastructure (Supra-Ricardian Model, 1970/71 through 1987/88) Temperature Rainfall ____ _HYV GCA NIA HY GCA NIA India -.0075 . 0035 -.0358 -.0093 -.0100 -.0054 1:Andhra Pradesh .0051 -.0024 -.0298 -.0157 -.0121 -.0099 2:Haryana -.0108 . 0287 -.0462 -.0019 -.0083 -.0003 3:Madhya Pradesh -.0134 -.0056 -.0432 -.0077 -.0092 -.0046 4:Madhya Pradesh -.0072 -.0063 -.0341 -.0129 -.0107 -.0082 5:Karyana .0018 -.0193 -.0076 -.0176 -.0128 -.0114 6:Punjab -.0123 .0339 -.0428 .0004 -.0079 .0011 7:Tamil Nadu -.0155 -.0008 -.0331 -.0118 -.0150 -.0068 8:Uttar Pradesh -.0077 .0123 -.0412 -.0051 -.0086 -.0024 9:Bihar -.0047 .0055 -.0036 -.0069 -.0098 -.0036 10:Gujarat -.0223 .0075 -.0544 -.0104 -.0093 -.0062 14:Rajasthan -.0133 .0160 -.0456 -.0072 -.0081 -.0039 15:Orissa .0114 -.0045 -.0158 -.0125 -.0116 -.0075 17:WestBengal .0017 -.0146 -.0126 -.0121 -.0070 Note: Temperature Effect: Percentage change due to a 1° Celsius temperature increase. Rainfall Effect: Percentage change due to a 3% rainfall increase. ESTIMATES: NET REVENUE Appendix Table 6B.6 displays the results of the regression of net revenue on edaphic, climatic, and geographic variables, the predicted values of the technology and infrastructure variables from the two-stage system described above, interactions between climate and technology or between climate and infrastructure, and dummy variables for time". This equation also fits the data very well: the adjusted R2 is nearly 0.6, and the F-statistic is highly significant. Appendix Table 6B.7 presents results of F-tests of the null hypotheses that the set of edaphic variables, the set of temperature variables, the set of rainfall variables, the set of all climate variables, the set of technology and related infrastructure variables, the set of technology interactions, and the set of year dummies do not influence net revenue. All of these null hypotheses were rejected: each set of variables does significantly influence net revenue; most of the variables in fact by themselves significantly influence net revenue. In addition to their role in the second-stage equations reported in the previous section, the edaphic variables are important determinants of net revenue: fourteen of the nineteen soil type '4 The climate-technology interactions involve temperature squared and rainfall squared, rather than simple temperature or rainfall, in order to capture nonlinearities in the interactions. 196 dummies have significant coefficients, and soil of neutral pH'5 contributes more to net revenue than either acidic or base soil. The coefficient on the predicted value of modem varieties (WHYVPRDK) is positive and significant while the coefficient on the predicted value of multiple-cropping (GCNCPRDK) is positive and not significant. The coefficient on the predicted value of irrigation intensity (NINCPRDK) is negative and significant. All those variables are interacted with climate terms, as discussed below, and half of the interactions' coefficients are significant. Research and extension conatribute to net revenue (the coefficient of extension on net revenue, although negative, is quite small and insignificant, but extension significantly increased the adoption of modem varieties; the coefficient of research on WHYV is significantly positive in both the WHYV and the net revenue regressions). Higher bullock costs reduce net revenue, as does an increase in the ratio of off-farm wages relative to agricultural wages, which probably denotes a decline in the quality of available farm workers as off-farm opportunities attract more and more of the best and most-highly-skilled laborers. Sixteen of the seventeen year dummies are positive and significant; the dummy for 1970, the first year in the sample, is omitted, so these dummies are picking up omitted time trends and price index effects. Current weather, and its timing, also obviously influences current net revenue: given a normal seasonal rainfall, higher rainfall in July and August (the variable JUAURAIN) will increase net revenuel6. The climate variables represent long-term averages or norms, to which farmers respond in their decisions about cropping patterns, input use, investment in technology and infrastructure, and so forth. This model displays quite rich (normal) temperature and rainfall effects on net revenue. The squared and "raw" terms are usually of the opposite sign. The impacts of temperature and rainfall differ by month. Interestingly a higher coefficient of variation of rainfall contributes to net revenue17. A key focus of this study is the interaction of climate with technology, infrastructure, and geographic variables, beyond the so-called "purely climate" variables. six such interactions each month are included. The interactions of temperature squared and rainfall squared with the predicted values of modem varieties, multiple-cropping and irrigation intensity are complex, yielding significant coefficients in some months. i5 In terms of the data, of pH 7, whose variable (DMPH7) is the omitted dummy. 16 Probably occurring during crucial maturation phases of many important crops in most states; actual June rain (holding constant the level of normal seasonal rainfall) had a negative but insignificant coefficient, perhaps reflecting the difficulty in planting when the ground is too wet. " This may reflect monsoon timing: a higher coefficient of variation may indicate that the rains were spread more evenly across the two months. 197 CLIMATE CHANGE EFFECTS ON NET REVENUE Based on the regression results reported in the previous section, the actual district-level values of the climate and technology variables, and the computed climate change effects on the technology variables, one can similarly compute the predicted effects on net revenue of changes in normal temperature and normal rainfall. Those effects (and their sum) are presented in Table 6.3. Temperature and rainfall affect net revenue via three avenues. The first could be called direct, operating via the temperature and rainfall terms, their squares, and the temperature-rainfall interaction term in the net revenue equation. The second avenue could be called local, beginning with the temperature and rainfall effects on WHYV, GCANCA and NIANCA and operating through their terms in the net revenue equation. The third avenue, meandering, operates through the terms in the net revenue equation which capture the interactions between climate and the predicted values of WHYV, GCANCA and NIANCA. Thus one cannot simply glance at the direct climate terms in the net revenue equation to perceive the temperature and rainfall effects; one must compute carefully the effects through all three avenues, and use in the appropriate places in the computations the values of the climate terms and the technology variables. The predicted temperature and rainfall effects on net revenue are high compared to estimates for U.S. and Brazilian agriculture: on average over all the districts in this study, a one degree increase in normal temperature would reduce net revenue by three and one-third percent"8. The differences in estimated effects might be due to the fact that these estimates are based on a net revenue framework, while other estimates are based on reasonably well-defined land prices; other differences include the time period involved, the use of a panel of cross-sections here, and the inclusion of endogenous technology variables. Appendix C reports estimates based on seasonal climate variables that are closer to estimates for the United States and Brazil. The States whose net revenue is most adversely affected by an increase in temperature are Maharashtra, Haryana and Tamil Nadu, while the States least adversely affected are Karnataka, West Bengal, Orissa and Gujarat (where in fact higher temperatures would increase net revenue). The predicted rainfall effects on net revenue are negative in all states, are on average nearly twice as large as the temperature effects, and are especially large and negative in Tamil Nadu"9, Andhra Pradesh, Orissa and Karnataka. In all States except Gujarat and Kamataka higher temperature and rainfall together are predicted to decrease net revenue: in four of the thirteen States the combined temperature and rainfall effects would in fact decrease net revenue by nearly one-fifth or more, and on average in the nation as a whole the combined climate effects would decrease net revenue by nearly one-tenth. " Recall that net revenue is crop revenue minus purchased inputs. Nationwide, net revenue averages far less than twenty percent of crop revenue: thus a 3% temperature impact on net revenue amounts to less than 1% impact on crop revenue. Impacts on net revenue are appropriate to maintain approximate comparability with impacts on land value in Ricardian systems, but considering impacts on crop revenue as well offers important perspective. 19 Tamil Nadu poses difficulties in estimation and calculation of effects based on rainfall and irrigation; those difficulties likely arise from its position on the Eastern edge of the tip of the Indian sub-continent, the influence of its ghats, and -- linked to those two -- the dramatic variation and amounts of rainfall the state receives from both monsoon systems. The extremely large predicted climate effects in Tamil Nadu are unlikely to be accurate. 198 Table 6.3: Climate effects on net revenue (Supra-Ricardian Model, 1970/1 through 1987/88) Temperature Rainfall Sum of Temperature and Rainfall INDIA -.0330 -.0624 -0.0954 1 :Andhra Pradesh -.0425 -.1665 -0.2090 2:Haryana -.1681 -.0243 -0.1924 3:Madhya Pradesh -.0695 -.0185 -0.0880 4:Maharashtra -.1883 -.0773 -0.2656 5:Karnataka .1715 -.1403 0.0312 6:Punjab -.0835 -.0108 -0.0943 7:Tamil Nadu -.1340 -.2348 -0.3688 8:Uttar Pradesh -.0453 -.0181 -0.0634 9:Bihar -.0145 -.0340 -0.0485 10:Gujarat .0328 -.0023 0.0305 14:Rajasthan -.0124 -.0037 -0.0161 15:Orissa .0478 -.1536 -0.1058 17:West Bengal 1.0734 -.0999 -0.0265 Note: Temperature Effect: Percentage change due to a 10 Celsius temperature increase. Rainfall Effect: Percentage change due to a 3% rainfall increase. TECHNOLOGY AND INFRASTRUCTURE EFFECTS As discussed above, our specification in equation (4) includes important technology and related infrastructure variables which are central to India's Green Revolution experience. Section four presented two stage least squares estimates of NIANCA, GCANCA and WHYV, in which each of the ithree variables was determined inter alia by one or both of the other two. And Section five presented estimates of Net Revenue, in which all three of the technology and related infrastructure variables, as well as their interactions with climate variables, appeared on the right- hand side. Table 6.4 presents the computed effects of an increase in any of the three technology and related infrastructure variables on the other two and on Net Revenue. The computations reveal the total effects, involving direct terms, indirect effects through a third variable, and (for effects on Net Revenue) interaction terms20. The broad message of Table 6.4 is that each of the 20 For example, the direct effect of WHYV on Net Revenue derives from the predicted WHYV term (which is named WHYVPRDX) in the net revenue equation. The indirect effects - in calculus, requiring the use of the chain rule - arise from the impact of WHYV on predicted GCANCA and on predicted NIANCA, and subsequently the effects of predicted GCANCA and predicted NIANCA on Net Revenue. And the interaction effects arise from the terms in which predicted WHYV is multiplied by various climate variables - in calculus, requiring the use of the product rule. 199 technology and related infrastructure variables encourages and reinforces the others, and contributes to Net Revenue. The HYV effects are consistently positive. A one percent increase in the proportion of crops planted to modem varieties would induce an increase in multiple-cropping of about one-ninth of one percent nationwide, nearly one-quarter of one percent in Haryana, two-firths of one percent in the Punjab. The effects of increased modem variety adoption are considerably larger on Net Revenue, especially in the States along the northeastern rim of the country (the Punjab, Haryana and Western Uttar Pradesh, which were the early beneficiaries of the Green Revolution in wheat, as well as Eastern Uttar Pradesh, Bihar and West Bengal, and in Rajasthan and Gujarat), and smaller but still consistently positive on irrigation intensity. The effects on irrigation are not difficult to understand, given the responsiveness of nearly all modern varieties to an assured2' supply of water: as farmers use more and more modern varieties, the payoff to irrigation increases, inducing more investment (privately and by both state and Central governments) in irrigation capacity and facilities. And the HYV effects on Net Revenue are also easy to understand: to have been selected and released, the modem varieties will already have been shown to offer yield increases in excess of their additional input requirements. The multiple-cropping effects on modern varieties, irrigation intensity and net revenue are positive and rather small: on average for the entire nation, a one percent increase in multiple- cropping would induce an increase in the use of modern varieties by nearly one-fifth of one percent, an increase in irrigation intensity of only four hundredths of one percent, and an increase in Net Revenue of about three-fifths of one percent. Surprisingly, an increase in multiple- cropping would reduce net revenue in Haryana, the Punjab, and to a lesser extent in Tamil Nadu and West Bengal. An increase in irrigation intensity would increase the adoption of modern varieties, for essentially the reason discussed above: with more assured water availability, the payoff to the adoption of HYVs is much higher. The effect of an increase in NIANCA on multiple-cropping and on Net Revenue is larger, with an increase in irrigation intensity of one percent increasing net revenue on average nearly one percent. Effects in Table 6.4 are expressed in quasi-elasticity form. For example, the upper-left entry indicates that, on average for the nation as a whole, a one percent increase in the proportion of crops sown to modern high-yielding varieties would tend to increase multiple cropping by 0.1141 percent, which is about one-ninth as much. 21 Assured not only in quantity, but at least as importantly in timing. 200 Table 6.4: Technology and infrastructure effects (Supra-Ricardian Model, 1970/71 through 1987/88) Effect of: HYV GCA NIA on: GCA NIA NR HYV NIA NR HYV GCA NR India .1141 .0601 1.0960 .1824 .0416 0.6023 .1127 .1546 0.9604 1:Andhra Pradesh .0978 .0566 0.8690 .1564 .0364 0.9017 .0986 .1455 1.2538 2:Haryana .2353 .1123 1.5720 .3761 .0956 -1.4792 .2589 .2889 1.7052 3:Madhya Pradesh .0527 .0222 0.9188 .0843 .0085 0.8103 .0230 .0571 0.9457 4:Maharashtra .0749 .0168 0.8714 .1197 .0080 0.7361 .0217 .0433 1.1006 5:Kamataka .0595 .0309 0.8543 .0951 .0130 1.5471 .0353 .0795 0.7578 6:Punjab .4190 .1903 1.6187 .6698 .2364 -1.8535 .6403 .4894 1.4382 7:Tamil Nadu .1415 .0768 09138 .2262 .0707 -0.5613 .1915 .1976 1.1642 8:Uttar Pradesh .1659 .1083 1.2247 .2651 .0735 -0.0759 .1992 .2786 1.0710 9:Bihar .1571 .0644 1.0576 .2512 .0497 0.6254 .1346 .1656 0.6704 10:Gujarat .0592 .0308 1.2041 .0946 .0126 2.5879 .0342 .0793 0.1552 14:Rajasthan .0535 .0400 1.4951 .0855 .0129 0.6171 .0349 .1030 1.1016 15:Orissa .0756 .0541 0.6954 .1208 .0161 0.8548 .0435 .1392 0.5494 17:West Bengal .1347 .0591 1.2435 .2153 .0389 1.3450 .1054 .1520 0.8345 SECONDARY IMPACTS ON CLIMATE AND TECHNOLOGY EFFECTS We have seen from Table 6.2 and the associated discussion, and from Table 6.3 and its associated discussion, the predicted effects of changes in climate variables on the three technology and related infrastructure variables, and on Net Revenue. One of the strengths of our approach, integrating climate and edaphic variables with technological and infrastructural variables, is that we can also compute the predicted impact of changes in technology and related infrastructural variables on the already-reported climate or other effects. In simple terms, these secondary impacts on, say, the temperature effects simply measure the extent to which changes in technology and infrastructure - over which policy-makers exercise some influence - might modify or ameliorate the effect of temperature changes on Indian agriculture. Table 6.5 presents the impacts of increases in the technology and related infrastructure variables on the effect of higher temperature on Net Revenue; Table 6.6 presents the impacts of increases in the technology and related infrastructure variables on the effect of higher rainfall on Net Revenue. 201 Table 6.5: Secondary impacts on the temperature effects on net revenue (Supra-Ricardian Model, 1970/1 through 1987/88) .____________ Research Extension WHYV GCANCA NIANCA India .00000419 .0004 .0308 -.0182 .1756 1: Andhra Pradesh .00000490 .0004 .0360 .0018 .2084 2: Haryana .00000603 .0005 .0443 -.0897 .1932 3: Madhya Pradesh .00000295 .0002 .0217 -.0249 .1766 4: Maharashtra .00000331 .0003 .0243 .0139 .1813 5: Karnataka .00000344 .0003 .0253 .0374 .1706 6: Punjab .00000652 .0006 .0479 -.0932 .1826 7: Tamil Nadu .00000723 .0006 .0532 -.0034 .2163 8: Uttar Pradesh .00000426 .0004 .0313 -.0503 .1727 9: Bihar .00000387 .0003 .0284 -.0218 .1645 10: Gujarat .00000419 .0004 .0308 .0283 .1494 14: Rajasthan .00000460 .0004 .0338 -.0405 .1649 15: Orissa .00000368 .0003 .0271 .0083 .1755 17: West Bengal .00000390 .0003 .0287 -.0012 .1523 Table 6.6: Secondary impacts on the rainfall effects on net revenue (Supra-Ricardian Model, 1970/71 through 1987/88) Research Extension WHYV GCANCA NIANCA India -.00000127 -.000010 -.0093 -.0257 -.0091 1: Andhra Pradesh -.00000054 -.000046 -.0040 -.0136 -.0011 2: Haryana -.00000269 -.000200 -.0198 -.0488 -.0244 3: Madhya Pradesh -.00000140 -.000100 -.0103 -.0269 -.0120 4: Maharashtra -.00000066 -.000056 -.0048 -.0158 -.0035 5: Karnataka -.00000012 -.000010 -.0008 -.0095 .0056 6: Punjab -.00000321 -.000300 -.0236 -.0580 -.0292 7: Tamil Nadu -00000331 -.000300 -.0243 -.0663 -.0220 8: Uttar Pradesh -.00000179 -.000200 -.0132 -.0333 -.0158 9: Bihar -.00000186 -.000200 -.0137 -.0363 -.0147 10: Gujarat -.00000018 -.000015 -.0013 -.0048 -.0008 14: Rajasthan -.00000058 -.000049 -.0043 -.0112 -.0050 15: Orissa -.00000133 -.000100 -.0098 -.0289 -.0079 17: West Bengal -.00000065 -.000055 -.0048 -.0191 .0010 202 In broad strokes, an increase in the technology and related infrastructure variables tends to worsen the (already negative) rainfall effects on net revenue somewhat, and except for multiple-cropping tends to improve the (negative) temperature effects on net revenue. In both cases, the research and extension impacts are smaller than the modem variety, multiple-cropping and irrigation impacts. The largest impact is that of an increase in irrigation on the temperature effect, 0.1756: this may represent the possibility of shifting the growing season to less-hot months with assured water from irrigation infrastructure. CONCLUSIONS We have constructed a four-equation model of the Indian agricultural sector which is rich and powerful. We begin by modeling the processes of technological and infrastructural change which have characterized India's Green Revolution: the adoption of modem high-yielding varieties; the expansion of multiple-cropping; and the expansion of irrigation. These are estimated in a 2SLS system in which the independent variables include one or both of the other technology variables, climatic and edaphic variables, and in some of the equations base-year values of the dependent variables or public investment in agricultural technology. Instruments in the 2SLS system include private investment in agricultural technology, measures of infrastructure (including price ratios), many climatic and edaphic and geographic variables, and regional and time dummies. Many others have studied the Green Revolution, and the accompanying changes in productivity, in India and elsewhere. But never before has it been studied in a framework in which the three primary technological variables which characterize the processes of change (namely, WHYV, GCANCA and NIANCA) are directly modeled, in a system which includes detailed edaphic and climatic variables as well as the more familiar public and private investment variables, prices, etc. The system is well-specified. Tests reveal that each group of variables, as well as most individual variables, significantly affect the three variables whose expansion characterizes India's Green Revolution experience, in ways which are consistent with prior studies. We then estimate our fourth equation, the logarithm of net revenue per hectare, modeled as dependent upon climate, edaphic and geographic variables, the predicted values of WHYV, GCANCA and NIANCA obtained from the 2SLS system, interactions between these variables and climate variables, and additional public investment and infrastructure variables. As before, this fourth equation is also quite well-specified: tests reveal that each group of variables, as well as most individual variables, significantly affect the net revenue. We then computed effects of climate change, and of continued increases in technology and related infrastructure, on the process of technological change and on net revenue per hectare. Our broad findings are summarized below: Climate affects technology development and diffusion; conversely, technology development and diffusion affects the impacts of climate on productivity in India. Technology development and diffusion, as well as climate, also affect net revenue in agriculture in India. First consider the effects of climate on the development and diffusion of technology. Estimates indicate small but negative effects of a rise in rainfall on all three indices of technology 203 (modem varieties, irrigation intensity and multiple-cropping). The effects of an increase in temperature are mixed: a small negative effect on the adoption of modem varieties, a somewhat larger negative effect on irrigation intensity, and a small positive effect on multiple-cropping. Thus potential climate change is not substantially adverse to the development and diffusion of technology of the type realized in India since 1965. Next, consider the estimates of the secondary impacts of technology on the climate effects on net revenue. These show that all indices of technology worsen (that is, make more negative) the estimated negative impacts of an increase in rainfall on farm production. They also show that all but one of the indices of technology improve (that is, make less negative) the effects of an increase in temperature. Thus technology is temperature friendly and rainfall unfriendly. Finally, the net revenue estimates show that the technology variables significantly affect crop production and that many technology - climate interactions are statistically significant. The exclusion of technology variables probably leads to some bias in predictions of climate impacts on crop production. REFERENCES Mendelsohn, R., Nordhaus, w. and Shaw, D. 1994. "The Impact of Global Warming on Agriculture: A Ricardian Analysis" American Economic Review 84: 753-771. Sanghi, A., R. Mendelsohn and A. Dinar. 1998. "The Climate Sensivity of Indian Agriculture." Chapter 4 in this report. 204 APPENDIX A: VARIABLES IN THE COMBINED INDIA AGRICULTURAL DATA SET INTRODUCTION This Appendix describes the variables which are used in this study: their definitions, units, sources, any transformations which they underwent, and any special treatment which they required. These variables come from two distinct data sets: one, which is sometimes called the "original" data set, which was created between 1980 and 1990, and has been used in numerous studies of production and productivity in Indian agriculture; the other, created in 1996, which added edaphic and climatic variables. Not every variable from either of the sets is used in this study; this Appendix covers only those which are used. COVERAGE The data set covers nearly all the districts (for a total of 271 districts) within thirteen of the States of India: Andhra Pradesh Madhya Pradesh Rajasthan Bihar Maharashtra Tamil Nadu Gujarat Orissa Uttar Pradesh Haryana Punjab West Bengal Karnataka These thirteen States constitute the three primary Northern Wheat and Northern Rice producing states (viz., Haryana, Punjab and Uttar Pradesh), two Northwestern Bajra-producing states (Gujarat and Rajasthan), three Eastern states (Bihar, Orissa and West Bengal), as well as all of the Semi-Arid Tropics States as specified by ICRISAT. The major agricultural states which are absent from the data set are Kerala, at the southern tip of the subcontinent and the Eastern state of Assam; also absent, but less important agriculturally, are the minor states and Union Territories in the Northeastern part of India, as well as the far-northern states of Himachal Pradesh, Jammu and Kashmir. During the period covered by the data set, there have been numerous adjustments in the boundaries (and even existence!) of some of the districts. These changes have occurred, for example, upon the division of the former Punjab into Punjab, Haryana and Himachal Pradesh; in the division of certain districts into two or more smaller districts in many places (especially in Bihar); or in the transfer of parts of one district to another. Insofar as possible, the data set preserves the original district boundaries: where districts have been broken up, values for the resultant districts in later years have been summed to yield values appropriate to a "shadow"consolidated district22. (However, the data set treats Haryana's districts as though they 22 This obviously means that the actual number of modem-day districts covered is considerably larger than 271, because many current districts have been "consolidated" into the larger districts from which they had emerged. 205 always belonged to a State named Haryana even though, before 1966, they were part of the original Punjab.) Some districts which currently exist will not appear in the data set, therefore, because they have been combined with other districts to create aggregations which approximate historical boundaries. Other current districts may not appear for other reasons, primarily because of dearth of agricultural activity (e.g., Bombay, a few Himalayan districts of northwestern Uttar Pradesh, or a few desert districts of Rajasthan) but occasionally because very little data is available for them. Each district is assigned a unique identification code in the data set, composed of a two- digit State code (in the variable STATE) and a two-digit district code (in the variable DISTRICT). Some State and district codes contain leading zeros. The State and district codes have been combined into a three- or four-digit code named ID. In addition, the variable STNAME contains the namne of each state, or its abbreviation. Part VI contains a list of districts and their identification codes. The data set contains observations for each of the variables for the agricultural years 1957/58 through 1987/87 The agricultural year 1957/58 is denoted by 1957 in the variable YEAR in the data set; the agricultural year 1983/84 is denoted by 1983; and so forth. With the exception of three of the rainfall variables (which are clearly identified to refer only to a few specified months during the given year) and the wage variable (which refers to a daily wage), all variables are expressed as annual flows or average annual stocks or average annual levels or amounts. OUTPUTS The data set contains data pertaining to five "major" and thirteen "minor" crops, enumerated below: Major Crops BAJRA JOWAR MAIZE RICE WHEAT Minor Crops BAR (Barley) COTN (Cotton) GNUT (Groundnut) GRAM JUTE OPULS ("Other" Pulses, other than Gram) POTAT (Potato) RMSEED (Rapeseed and Mustard) 206 SESA (Sesamum) SOY (Soybeans) SUGAR SUNFL (Sunflower) TOBAC (Tobacco) For each of the so-called minor crops the data set includes: Area Planted (1000 hectares; A followed by crop code) Production (1000 tonnes; Q followed by crop code) Farm Harvest Price (Rupees per quintal; P followed by crop code). For the five major crops, the data includes the three variables listed above plus: Area irrigated under the crop (1000 hectares; I followed by the crop code) Area planted to HYV in each crop (1000 hectares; H followed by an abbreviated crop code). The primary sources of data on Area and Production include: Area and Production of Principal Crops in India, GOI Crop and Season Reports of the various States Statistical Abstracts of the various States Agricultural Situation in India, GOI. Beginning in 1954 and extending until the late 1960's, the Directorate of Economics and Statistics published Area and Production in two Parts: Part I contained All-India and Statewide data, while Part II (Detailed Tables) contained District-level data. Typically each issue of Part I would cover three years or so, while Part II would appear less frequently and cover a longer time span. But no Part II has been published for twenty years. Therefore, recently, the most convenient source for Area and Production data has been the monthly Agricultural Situation. This creates two small problems. The first is practical: One must cull through twelve issues each year, finding usually no more than three or four crops' data presented in any one issue. The second, and far more significant, problem is substantive: the district-level estimates of Area and Production presented in Agricultural Situation are called "Final" estimates, and usually are the first estimates to be published. But so-called "Final" estimates are still subject to change, to be superseded by what are called "Revised" estimates. No such changes are reported in Agricultural Situation so there is no way to know whether such revisions have even been made without consulting sources other than Agricultural Situation. Seldom are the revisions large, however, so this data set relies heavily upon estimates from Agricultural Situation throughout much of the 1970's and 1980's. Whenever it was possible to gain access to the Statistical Abstracts and/or Crop and Season Reports of any States for any of the years covered by the data, those sources were used for the estimates of Area and Production. In addition, those sources were especially valuable in 207 providing data, for the major crops, on Area Irrigated under each crop and Area planted to HYVs in each of the crops, although Agricultural Situation has begun to include that data as well. District-level Farm Harvest Prices are easily available from Farm Harvest Prices of Principal Crops in India, published every four years or so by the Directorate of Economics and Statistics. The prices are reported in Rupees per quintal. (Both wholesale and retail prices of all crops are also available, published regularly in Bulletin on Food Statistics, in Agricultural Prices in India, and elsewhere. Retail prices would be appropriate, for example, in a study of consumption behavior or poverty. Wholesale prices would be of interest, for example, in studying government grain procurement policies or interstate food movements. The obvious prices of most interest to this study are Farm Harvest Prices, because it is on the basis of those prices, or farmers' expectations of their future values, that farmers determine their behavior. But the data set also includes the wholesale prices of most crops, as well as a computed weighted and aggregated average Relative Price variable, RELPRICE, whose weights are the share of the crop in total area in the district, computed as the Farm Harvest Prices divided by the Wholesale Prices. Relative Price is one of the institutional variables which help to determine the levels of investment in infrastructure and public goods.) VARIABLE INPUTS The data set includes three categories of variable inputs: labor, fertilizer, and power. The variables relating to labor include: Rural Population: (the total population of the district, male and female, residing in areas classified as rural: RURPOP) Agricultural Labor: (the number of rural males whose primary job classification is agricultural labor: AGLABOR) Cultivators: (the number of rural males whose primary job classification is Cultivators: CULTIVAT) Total Farm Labor: (a weighted sum of Agricultural Labor and Cultivators: QLABOR) Wages: (weighted annual labor cost: WAGE) Factory Earnings: (weighted annual earnings in a rural factory: FACTEARN) The first three variables are obtained from the decennial population census, which reports the job classifications of all persons enumerated as well as many population totals. The population census has been conducted in India for more than a century and is widely deemed to be highly accurate. Census results are published in an extensive series of volumes for each state; the district-level values of the rural population and job classifications are reported in the Primary Census Abstract, and are reprinted frequently in Statistical Abstracts as well as many other sources. The data set is based on the reported values for the census years 1951, 1961, 1971 and 1981; the values of RURPOP, AGLABOR and CULTIVAT for the other years in the data set are linear interpolations (for 1956 through 1960, 1962 through 1970, and 1972 through 1980) and linear extrapolations (1982 through 1987) of the reported data. Interpolating population values is probably benign: such variables change in relatively regular and consistent ways. The numbers of agricultural laborers and cultivators often change substantially within a decade, so 208 linear interpolations between census years may mask more volatile behavior. Unfortunately, however, the values of the population variables are not measured during any inter-censal years, so no better data could exist. The Rural Population values appear in the data set exactly as they had been recorded. The Cultivator and Agricultural Labor values, however, measure a stock: the number of people who claim those activities as their primary job. The economically appropriate variable is a flow: the amount of labor performed during the year by such workers. The number of Agricultural Laborers and Cultivators are added, and their sum is multiplied by the average number of days worked in the State by farm workers (as obtained from various Farm Management Surveys; see the table on the subsequent page) in order to compute the appropriate flow of labor services variable: QLABOR. # of days worked State by farm workers Andhra Pradesh 230 Bihar Guj arat 215 Haryana 244 Karnataka 217 Madhya Pradesh 239 Maharashtra 240 Orissa Punjab 244 Rajasthan 215 Tamil Nadu 293 Uttar Pradesh 210 West Bengal Agricultural Wages are obtained from Agricultural Wages in India, published by the Directorate of Economics and Statistics every two or three years, reporting daily wages and normal daily working hours for each of the twelve months from reporting centers in most districts for different farming activities. Wages are reported separately for men, women and children for some activities. Whenever possible, the wages of a male ploughman were recorded; if a district did not record such a wage, the wages of a male field laborer or male "Other Agricultural Labour" were selected instead. An average annual wage was constructed from the monthly wages, weighting June and August more heavily than other months because of the intensity in those months of field work in most cropping patterns and most states. Factory Earnings measure the average annual earnings of an unskilled laborer working in a rural "factory" (which could include establishments employing as few as two persons). This variable not only measures the opportunity cost of working on one's own farm, but captures to an extent some of the supply conditions of the local rural labor market. We compute a Relative Wage (simply by dividing annual farm earnings by average annual factory earnings), which is 209 one of the institutional variables which help to determine the levels of investment in infrastructure and public goods. The variables relating to fertilizer include the quantities of nitrogen, phosphorous and potassium fertilizers (in tonnes: denoted NITRO_TQ, P205_TQ and K20_TQ) and the prices of the three fertilizers (in Rupees per tonne of nutrient: NITRO_TP, P205_TP, K20_TP). The fertilizer data source is Fertilizer Statistics, published annually by The Fertilizer Association of India. Quantity data is given by district, by nutrient, and often by season; only yearly data is included in the set. Prices of fertilizers are strictly controlled by the Central Government, so the only cross-section price variation arises from the cost of transportation from the railhead to the field; the prices of the nutrients in the data set, therefore, exhibit no cross-section variation, but are based on reported maximum sale prices of common fertilizer compounds adjusted for the proportion of the nutrient present in each compound. Prices are not reported for all nutrients for all years; prices for intervening years are estimated based on movements of the fertilizer wholesale price index during those years. Farm (draft) power is obtained from two primary sources: bullocks and tractors. The quantities of both are enumerated in the quinquennial Livestock Census. The results of each Livestock Census are published in two Parts: Part I contains All-India and statewide data, while Part II contains district-level data. Part II has been published for the Censuses of 1956, 1961, 1966, 1972 (the census which had been scheduled to occur in 1971, according to the former sequence, had to be postponed to 1972) and 1977, but district-level data is not yet available for the census of 1982. (The publication backlog seems to be increasing, and since the 1977 district- level data were not released until December, 1987, it is unlikely that 1982 district-level data will be available within in the next four or five years.) Bullocks (QBULLOCK), as recorded for the data set, refer to castrated (male) cattle, over the age of 3 years, which are used in rural areas for work only. Tractors (QTRACTOR) are four- wheel (not tracked, nor walk-behind two-wheeled) machines. The numbers of bullocks and tractors in the inter-censal years (1957-1960, 1962-1965, 1967-1971, and 1973-1976) are estimated by linear interpolation. For years after 1977, for which no district-level data have yet been published, the data set contains estimates computed by extrapolating the 1982 observations at a rate equal to the percentage change in the state values from 1977 to 1982. Tractor prices do not vary across India: a single tractor price therefore appears for all districts in any given year. The tractor price is constructed as follows: The price index for Agricultural Machinery and Transport Equipment from 1954 through 1985 was compared to observed prices for Eicher 24-horsepower tractors during selected months from 1978 to 1987. (The Eicher prices were collected by P. C. Bansil of the Techno-Economic Research Centre, New Delhi). Movements in the price index mirrored movements in the Eicher tractor prices almost perfectly. So the Eicher price series was extended back to 1956, on the basis of proportional changes in the Agricultural Machinery and Transport Equipment price index. Eicher commands more than 50% of the market of tractors in the 1 to 25 horsepower range, which is the largest segment of the tractor market in India, but larger tractors command a higher price, so the Eicher 24-horsepower tractor's share in the total value of tractors is smaller than its share in the 210 number of tractors. So the average price of a tractor would be larger than the price of an Eicher 24-horsepower (an "average".?) tractor. To adjust for that, the estimated Eicher price series was multiplied by 1.66 (based on data showing the difference in prices for Escort tractors of various horsepower ratings in the early 1970's), producing a tractor price series which is consistent with both the movements of the price index and independent data on the prices of actual tractors. The resulting tractor price series was finally multiplied by one-fourth to derive an annual tractor cost variable (PTRACTOR). The value of one-fourth, or 25%, represents both the depreciation and debt service on the investment, as well as the rate of return which is required for tractors to be bought in the first place. Thus the annual tractor cost variable represents a sort of shadow rental cost of a tractor, in the appropriate flow form. The data set contains three bullock prices, reflecting the physical differences in bullocks in different parts of India. Each price series is based on retail price indices reported in various issues of Agricultural Prices in India, published by the Directorate of Economics and Statistics, in which bullocks are identified by state (e.g., Bihar, Gujarat, Haryana, and Uttar Pradesh). The so-called Haryana price was applied to bullocks in Haryana and Punjab; the Gujarat price was applied to bullocks in Gujarat, and the more prevalent Uttar Pradesh price was applied to bullocks in all other states. Rental fees for bullocks are very difficult to obtain. The annual bullock cost variable (PBULLOCK) was obtained by multiplying each bullock price by 0.50, representing both the substantial annual flow of expenses entailed in breeding, raising and feeding bullocks, as well as the necessary rate of return on their ownership. In closing the discussion of the variable inputs, it is interesting to note that the values of these prices and quantities are "realistic" in the sense that they imply input cost shares which are consistent with the range of cost shares obtained in earlier research. OTHER INPUTS The data set contains additional "inputs" which cannot be considered to be subject to the control of farmers in the short run. Some of this class of inputs, such as rainfall, are for all practical purposes beyond the influence of any human agency. And some, such as certain forms of irrigation, and perhaps literacy, can be influenced by farmers' decisions and behavior only over a substantially long period of time. Others, such as research and extension, are in part the result of governmental decisions, possibly in response to a diffuse and highly-lagged "demand" from farmers which is as much political as economic. Although not variable in the traditional sense, these "other" inputs do significantly influence agricultural output and productivity. These "other" inputs can be classified as members of three subgroups: Agro-climatic, Public, and Socioeconomic. AGRO-CLIMATIC INPUTS The inputs which are classified as agro-climatic pertain to the most basic agricultural inputs: soil and water. Two of them measure the use of land: Gross Cropped Area (GCA) and Net Cropped Area (NCA). Net Cropped Area is the total geographic area on which a crop has been planted at least once during the year. Gross Cropped Area is the total area planted to crops during all the growing seasons of the year; if any land has been doublecropped it will appear only once in Net Cropped Area, but twice in Gross Cropped Area. Both GCA and NCA are 211 measured in units of 1000 hectares23. From these two variables is computed a third, GCANCA, the ratio of total acreage during all seasons of the year to land planted only once; a measure of multiple-cropping. (See Section VI below for a discussion of more detailed edaphic variables from the second data set). Water is supplied in two ways: naturally, as Rainfall, and artificially, as irrigation. Data relating to irrigation are reported in several forms: area irrigated by source (e.g., by canal or tank or tubewell), area irrigated under certain crops, or total areas irrigated. The data set includes two variables of this last form: Net Irrigated Area (NIA) measures the total geographic area which has received irrigation (from any source) during the year, and Gross Irrigated Area (GIA) measures the total area under crops which has received irrigation during all the growing seasons of the year. As was true for NCA and GCA, if any irrigated land has been double-cropped it will appear only once in Net Irrigated Area, but twice in Gross Irrigated Area; again, the variables are measured in units of 1000 hectares24. (From NIA and NCA we compute the variable named NIANCA, the ratio of net irrigated area to net cropped area, which measures irrigation intensity.) Estimates of Gross and Net Cropped Area and of Gross and Net Irrigated Area are available from the annual Indian Agricultural Statistics which are published in two volumes: Volume I presents All-India and Statewide data, while Volume II contains District-wide data. This data is also available in most states' Crop and Season Reports and Statistical Abstracts, and has been published in the Agricultural Situation in India since the early 1980's. Crop-specific irrigated area is also reported in Fertilizer Statistics. Rainfall is measured daily in most districts in India at so-called "meteorological observatories" established by the India Meteorological Department. The district data are aggregated into approximately three dozen so-called "sub-divisions", which range from parts of a State (such as Coastal Karnataka, North Interior Karnataka and South Interior Kamataka) to an entire State (such as Orissa or Punjab). The monthly sub-divisional data are then published in a number of sources, including Agricultural Situation in India. Annual sub-divisional data are reprinted in many sources, most conveniently in Fertilizer Statistics. District-level (that is, non- aggregated) data are also published in some states' Crop and Season Reports, Statistical Abstracts, and in some specialized meteorological publications such as the occasional Climatological Tables of Observatories in India; a number of states augment the India Meteorological Department's data collection (and publication) with data collected by their own means. The data set contains five rainfall variables, four of which measure actual precipitation during parts or all of the agricultural year. The first, YEARRAIN, is the total rainfall in the given year: it is the sum of the rainfall in each of the twelve months. The other two rainfall variables measure rainfall in only one or a few months, at periods crucial to crop production: rainfall in June, at the beginning of the monsoon in most states (JUTNERAIN), and in July and August, 23 We also compute a variable which measures the intensity of double-cropping: GCANCA, which is GCA divided by NCA. It obviously ranges from a value of one, at which no double-cropping occurs, upwards. 24 And, as before, we compute a variable which measures the intensity of irrigation: NIANCA, which is NIA divided by NIA. It ranges from a value of zero, at which no irrigation occurs, upwards. 212 during the remainder of the monsoon in most parts of India (JUAURAIN). The fourth rainfall variable is a dummy variable (DROUGHT) which takes the value of one for those districts which are designated as "drought-prone" by the ICAR; it has no time-series variation, and does not measure whether a drought, by whatever definition, had occurred in any given year, but instead simply denotes those districts which historically have had low levels and high variability of rainfall. (See Section VI below for a discussion of more detailed normal rainfall variables from the second data set.) PUBLIC SECTOR INPUTS The public sector provides physical infrastructure which facilitates agricultural production. One of the most important inputs into agriculture provided by the public sector is research results, in the forn of new seeds, or improved implement design, or improved management practices, or any number of other forms. Research activities are undertaken by all of the states as well as by numerous Central schemes and projects, focussing on practically every crop grown in India as well as many inputs and all of the basic agricultural sciences. The specification of a valid and appropriate research variable is difficult, for a number of familiar reasons. Budget data are seldom available in a form which allows the separation of the accounts of research units from their parent organizations; even the unsatisfactory budget data that exists is flawed in that it is seldom obvious how to separate the current from the capital, and the researchers from the other staff. Even if one could confidently measure staff and expenditures, it is difficult to measure research output, especially if one recognizes the problems posed by quality differences, the almost-stochastic nature of most research efforts, and well-known vintage issues. In constructing the research variables for this data set, therefore, special efforts have been made to address those problems as fully as possible. The research variables are based on three sets of data. First is the indigenous State agricultural research expenditures series, covering the years 1953 through 1971, which was reported in R. Mohan, D. Jha and R. Evenson, "The Indian Agricultural Research System" Economic and Political Weekly (vol VIII, # 13, 31 March 1973). Second is a data set which contains the number of articles reporting research results which were abstracted in Indian Science Abstracts from 1950 through 1979. This data set provides crop- specific (for the crops wheat, rice, maize, jowar, bajra, cotton, sugar, for "other" crops, and for "general" agricultural research) and State-specific (including Delhi) data measuring the output of the research activity; the editorial authority exercised by the abstractors in imposing and enforcing quality thresholds for inclusion in the Indian Science Abstracts makes this set particularly useful. Third is recent state budget information regarding research spending, especially at the State Agricultural Universities during the late 1970's and the 1980's. These three were combined to create commodity-specific expenditures data series for each of the states from 1950 through 1983, multiplying each year's research expenditures by the ratio of the number of publications abstracted for that commodity in that state to the total number of commodity-level (that is, not "general") publications in the state. In addition, for each state a "general" expenditures data series was created by multiplying the year's research expenditures by the ratio of "general" abstracts to the total publications. (This procedure obviously uses the proportion of abstracted publications in each crop to allocate the total research effort, as measured by expenditures, among the various commodities.) 213 For each state and each commodity a research "stock" variable was then defined by cumulating past research activity utilizing several patterns of time-shape "inverted V" weights as first used in Evenson (1968). The inverted V has three regions. The first, sloping upward, refers to the number of years between the first appearance of the research result and its full effect, during which the research outcome has successively-greater impact. In this region more-recent results are multiplied by smaller fractions, while moderately-distant results are multiplied by larger fractions, until at the top of the upward-sloping region the weights become one. The second region, a horizontal plateau, refers to the number of years during which the research output can continue to contribute at "full strength", during which time the weights remain equal to one. The third region, sloping downward, represents a sort of "decay" in the research contribution, due perhaps to biological changes or merely being supplanted by later, superior discoveries. In this region earlier (that is, more distant in the past) research contributions are multiplied by successively-smaller weights. The data set contains six measures of public research, which differ in the time pattern of the three sets of weights. The table below lists the six statewide crop research variables, followed by the number of years specified in their upward-sloping, horizontal and downward-sloping regions, respectively: STRESI: 333 STRES4: 666 STRES2: 336 STRES5: 999 STRES3: 366 STRES6: 699 (In order not to lose early observations because of the lengthy lag structure, research activity in years prior to 1950 was set equal to one half of the activity in 1950.) Finally, each of the STRES variables was weighted by the share of the crop in the total value of output, summed across districts, and by the Gross Cropped Area planted to that crop in the state. The data set also includes a variable measuring private research activity (PRIVRES), which has increased in importance markedly during the past two decades. This variable is based on data collected by Prof. Carl Pray, measuring research spending by private firms in the seed, fertilizer and machinery industries. From this expenditure data three research stock variables were constructed using a linear five-year lag structure with no decay: the stock was defined as one-fifth of the previous year's spending plus two-fifths of the spending two years ago plus three-fifths of the spending three years ago plus four-fifths of the spending four years ago plus the sum of all spending five years ago and earlier. The lag structure obviously reflects the time required for are search program to produce economically meaningful results: from inception of spending to invention to innovation to manufacture to marketing to full diffusion. From the three input-specific private research stocks was then created the variable PRIVRES, measuring the local contribution (or potential) of this private research knowledge within each district, by adding the year's stock of seed research multiplied by the district's input share of land, plus the year's stock of fertilizer research multiplied by the district's input share of fertilizer, plus the year's stock of machinery research multiplied by the district's input share of bullocks and tractors. 214 The extension variable (EXT) is based on three sets of information. The first is data measuring the size of the extension service staff in 1975, 1980, 1983 and 1986 in each state, based on surveys by the World Bank. The second is the number of villages in each state. And the third is data published in various years' annual Reports of the Department of Community Development of the Ministry of Agriculture (during some years the Ministry of Food, Agriculture, Community Development and Cooperation), covering the years 1955 through 1972,. which report the number of Community Development Blocks in each state which were classified as Stage I, Stage II or Stage III. A Stage III block (strictly speaking, the blocks were called "post- Stage II") is the most advanced, not only denoting more contemporary extension activity but also resulting from the success of past and current extension activity. The expectation was that a block would remain in Stage I for about five years, and in Stage II for another five years, so to some extent the variability in Stages reflects the staggered onset of extension activity in the various blocks. (By the middle 1970's practically 100% of all blocks had progressed beyond Stage II). The staffing data were interpolated to obtain estimates for the years 1976 through 1979, 1981 and 1982. Then the staffing data were divided by the number of villages (in units of hundreds) to obtain a measure of the number of extension workers per hundred villages, interpreted as an indicator of extension presence. This variable was then extended backward, from 1975 to 1956, as follows: First, the Stage data were combined into a single weighted variable by multiplying the number of Stage I blocks by two-fifths, adding the number of Stage II blocks multiplied by four-fifths, adding the number of Stage III blocks, and dividing the final sum by the total number of blocks. (The coefficients 0.4 and 0.8 are admittedly a bit arbitrary; they were chosen to reflect the lower intensity of extension activity in the earlier Stages.) The resulting quotient is necessarily a positive fraction, which can be interpreted as the degree to which the extension effort has reached the norm. Under the assumption that the extension effort had reached the norm by 1975, the weighted Stage variable would equal one for 1975 (because all blocks are assumed to have reached Stage III, the numerator contains only one term, the number of Stage III blocks, and the denominator equals the numerator, since all the blocks are in Stage III) and the 1975 staffing level can be taken to represent the "normal" staffing. Thus for years before 1975 the estimated extension variable is computed as the product of the 1975 staffing levels times the weighted Stage coefficient, interpreted as the level of staffing which would prevail at the particular "sub- norm" level of extension activity which the pattern of Stages discloses. SOCIOECONOMIC INPUTS The variable LITERACY, obtained from the decennial population census, measures the proportion of rural males who are classified as literate, which is defined as "the ability to read and write in any language". Census enumerators, beginning with the 1971 Census, were required to observe each individual's ability to read and write before classifying him or her to be literate. As is true for all census variables, values for the inter-censal years were obtained by linear interpolation. Literacy rates change so slowly and so regularly that this procedure seems amply justified. 215 CLIMATIC AND EDAPHIC VARIABLES These additional variables extend the original data set in two dimensions. First are included climatic variables which were not available in the original data set (which contained only the three actual rainfall variables, measuring current weather). This data set adds monthly normal minimum and maximum temperatures (from which can be computed the monthly normal temperature range or the monthly normal temperature midpoint) for all twelve months; normal monthly rainfall (again for all twelve months) and the coefficient of variation of actual rainfall; and for some of the districts still other monthly climate variables. These climatic variables are described below. The second dimension includes edaphic variables which were not included in the original data set (which contains only the area under particular crops). This additional data set includes soil type, soil pH, the depth of aquifers, and topsoil depth; in addition, this data set includes geographic variables measuring latitude, longitude, altitude, slope, and proximity to the ocean. These additional land variables are described below. Some of the climatic and edaphic variables are multiplied by each other, and/or by other variables, capturing the interactions of the phenomena which those variables represent. Notable among those interactions are the maximum temperature squared, rainfall times maximum temperature, rainfall squared, and both rainfall and maximum temperature times the the predicted values of modern varieties, multiple-cropping and irrigation intensity. When each variable is described in the subsequent sections, its source is identified. In general, there were two types of sources used: tables and maps. Data from tables offers no problems. A few preliminary words about obtaining numeric data from maps, however, are in order. First, two general kinds of maps were used. Some (e.g., the maximum temperature map mentioned below in b)i)) are small, all-India maps which exhibit isotherms (or other isobars) weaving across the nation, usually at intervals of 2.50 C. To determine the air temperature for a district for which a direct temperature observation is not reported in any tables, we superimposed copies of these maps, made on transparencies, over maps showing district boundaries, and interpolated temperatures at one-half degree intervals between the isotherms. If an isotherm split a district substantially down the middle, that temperature was assigned as a "reference value" to that district. Otherwise, we assigned to the district the value of the half-degree interpolated isotherm which most nearly approached the middle of the district. We then used these "reference value" temperatures to determine which neighboring district, for which we did have complete temperature data from tables, was climatologically most like the district without temperature data, and thus whose temperature data we would assign to the district in question. The other kind of map displays, by color and pattern, the value of some variable for areas which may be smaller than districts, and which often span district boundaries. Most of the time, districts contain areas of more than one value, such as more than one soil type within the district, or regions of different slope within the district, or areas with different pH within the district, etc. This was handled in the data set in two distinct ways. 1. In constructing the variables for soil type (using the maps cited below), we estimated the proportion of each district's area under each soil type. The current data set contains 216 nineteen soil type dummy variables, one for each type. The value of a soil type dummy variable is one for a district if that type is one of the two predominant soil types in the district (that is, if that type's proportion is one of the two highest proportions in the district). This implies, obviously, that a district will have two soil type dummies with the value of one (except for the few districts all of whose soil is of one type), and thus that one cannot interpret the dummies' coefficients in the customary way. 2. In constructing all of the other variables derived from "color and pattern" maps (namely, pH, the depth of aquifers, topsoil depth, and slope) we simply selected the value which seemed predominant, which covered more of the district's area than did any other value. This approach does not distinguish between districts which are entirely covered by one value (say, a pH of 7; or a moderately steep slope), on the one hand, as against districts with several values, one of which covers slightly more area than do any others (say, 19% of the soil has a pH of 5, 19% has a pH of 6, 24% has a pH of 7, 19% with pH of 8, and 19% with pH of 9; or 20% of the district's area is flat, 41% is moderately steep, and 39% is very steep), on the other hand. A possible avenue for further research is to enrich the data set in this way. CLIMATIC VARIABLES 1. Air Temperature a. Monthly Normal Maximum i. variable names of the form __ _ x t 1) for example, janxt, febxt, marxt, ..., decxt 2) - - - is a three-character abbreviation of the month: jan, feb, etc. 3) x refers to maximum and t of course refers to temperature ii. sources 1) Climatological Tables of Meteorological Observatories in India a) 235 observatories' observations are reported i) 35 observatories are in states (or even countries: Pakistan, for example) not included in our data set ii) this leaves 200 observatories in relevant districts b) 32 of these remaining observatories are duplicates i) in a district which had two or more observatories (1) we chose the observatory which was located in the district headquarters, if present (2) otherwise closest to the geographic center ii) so the Climatological Tables... provided 168 distinct districts' observations (1) 130 of them, identified by the value of 1 for the dummy variable DMPEN, are full 30-year norms based on data from 1931 through 1960 (2) thus the other 38 of them, identified by the value of 1 for the dummy variable DMCLMTAB, are norms based on less than 30 years of observations, from 217 observatories which began recording temperature data after 1931. 2) temperatuire maps a) for remaining 102 districts (for which we had no temperature data from the Climatological Tables...), we assigned temperature data identical to a "neighboring" district (for which we did have temperature data from the Climatological Tables... ) b) We determined which adjacent district to use by visual inspection of a temperature map, pairing districts with same normal mean daily maximum temperature i) maps from Agroclimatic Atlas of India, India Meteorological Department, Pune, P.DGM.118(N) / 1000-1987 (DSKII) (1) first published in 1978 (2) subsequently reprinted many times ii) Plate 21: July mean daily maximum temperature b. Monthly Normal Minimum i. variable names of the form ___n n t 1) where the n refers to minimum 2) for example: jannt, febnt, marnt, ..., decnt ii. same sources and procedures as for maximum temperature, above c. Monthly Normal Temperature Range i. variable name of form_ t r g 1) where the t r g refers to temperature range 2) for example: jantrg, febtrg, martrg, ..., dectrg ii. computed, for each month, as ( _ _ x t - _ _ _ n t) d. Monthly Normal Temperature Midpoint i. variable name of form ___ md t 1) where the m d t refers to midpoint 2) for example: janmdt, febmdt, marmdt, ..., decmdt ii. computed, for each month, as x t + n t) / 2 2. Rainfall a. Normal Monthly Rainfall i. variable names of the form ___n r n 1) for example: janrn, febrn, marrn, ..., decm 2) again, _ _ _ is a three-character abbreviation of the month 3) r n denotes "rainfall norm" ii. sources 1) Climatological Tables... a) normal rainfall data is available for the same districts for which we obtained normal temperature data b) similar pattern of 30-year and shorter norms 218 2) Monthly & Annual Rainfall & Number of Rainy Days, 1901 - 1950 a) Indian Meteorological Department, in several volumes b) some 3000 rainguage stations across the subcontinent i) obviously then most districts have more than one rainguage station ii) as before, we used the rainguage station in the district headquarters, if it existed c) We computed 30-year normals (1921 - 1950) for districts for which no data was available in the Climatological Tables... d) thus, obviously, we did not have to resort to neighboring districts to obtain normal rainfall data for all districts b. Variability of actual rainfall i. coefficients of variation computed from the actual rainfall data ii. JURNCV: coefficient of variation of June rainfall, 1957 through 1987 iii. JARNCV: coefficient of variation of July & August rain, 1957 - 1987 3. Oktas of cloud cover a. variables of the form ___n ok i. as always, --_ is the three-character month abbreviation ii. for example, janok, febok, marok, ..., decok b. measures the extent to which clouds obscure the sun c. values range from to d. source: Climatological Tables.... EDAPHIC VARIABLES 1. Soil Type a. variable names of the form DMSnn i. for example, DMS02, DMS03, DMS04, ..., DMS19, DMS20 ii. DM denotes a dummy variable iii. S denotes soil type iv. nn ranges from 02 to 20, for "traditional" soil types b2.source: visual inspection of soil maps for each State i. S. P. Raychaudhuri et alia, Soils of India (New Delhi: Indian Council of Agricultural Research, 1963) ii. recall the discussion, in the Introductory section of this Appendix, of the methodology by which these variables were created c. types 01 not used 02 Laterite 03 Red and Yellow 04 Shallow Black 219 05 Medium Black 06 Deep Black 07 Mixed Red and Black 08 Coastal Alluvial 09 Deltaic Alluvium 10 Calcereous 11 Gray Brown 12 Desert 13 Tarai 14 Black (Karail) 15 Saline and Alkaline 16 Alluvial River 17 Skeletal 18 Saline and Deltaic 19 Red 20 Red and Gravelly 21 2. Soil pH a. source: National Atlas of India, vol 1, plate 59 b. variables: i. a series of dummy variables 1) DMPH5: strongly alkali: 4.5 < pH < 5.5 2) DMPH6: slightly alkali: 5.5 < pH < 6.5 3) DMPH7: neutral 6.5 150 meters thick c. (obviously this is not exhaustive, in the sense that major areas of the nation are above no aquifers at all, so for many districts none of these dummy variables has the value of one.) 4. Topsoil depth a. source: National Atlas of India, vol. 1, Plate 50: "Depth of Soil, All-India" b. variables: i: DMTS1: dummy variable, = I if topsoil is 0 - 25 cm. thick ii. DMTS2: dummy variable, = 1 if topsoil is 25 - 50 cm. thick iii. DMTS3: dummy variable, = 1 if topsoil is 50 - 100 cm. thick iv. DMTS4: dummy variable, = I if topsoil is 100 - 300 cm. thick v. DMTS5: dummy variable, = 1 if topsoil is > 300 cm. thick 5. Geography a. variable names i. Latitude 1) LATDEG the degree portion of the latitude (North) 2) LATMIN the minutes portion of the latitude ii. Longitude 1) LONGDEG the degree portion of the longitude (east of Greenwich) 2) LONGMIN the minutes portion of the longitude iii. Altitude 1) ALTITUDE 2) the altitude of the meteorological observatory, in meters, above mean sea level b. sources i. Climatological Tables... ii. for districts which have no observatories within their boundaries 1) we assigned Latitude and Longitude by visual inspection of detailed state-level maps 2) we estimated the latitude and longitude of the district headquarters iii. those districts, though, lack Altitude data 6. Slope a. source: National Atlas of India, vol. 1, several plates i. plate 44: All-India ii. plate 45: Northern India 221 iii. plate 46: Western India iv. plate 47: Central India v. plate 48: Eastern India vi. plate 49: Southern India b. variables: L. a series of dummy variables included in the data set 1) DMSL1: flat: less than 10 meters / km. 2) DMSL2: flat: 10 to 20 meters / km. 3) DMSL3: gentle slope: 20 to 80 meters / km. 4) DMSL4: moderately steep slope: 80 to 150 meters / km. 5) DMSL5: steep slope: > 150 meters / km. ii. the maps had more gradations, which exist in the data set but from which the preceding variables were constructed: DMSLla: flat (<10), < 100 meters above mean sea level DMSLIb: flat (<10), 100 to 500 meters above mean sea level DMSLIc: flat (<10), > 500 meters above mean sea level DMSL2a: flat (10-20), < 100 meters above mean sea level DMSL2b: flat (10-20), 100 to 500 meters above mean sea level DMSL2c: flat (10-20), > 500 meters above mean sea level DMSL3a: gentle (20-80), < 100 meters above mean sea level DMSL3b: gentle (20-80), 100 to 500 meters above mean sea level DMSL3c: gentle (20-80), > 500 meters above mean sea level DMSL5 originally was: steep slope: 150 to 300 meters / km. DMSL6: steep slope: 300 to 600 meters / km. DMSL7: very steep slope: > 600 meters / km. iii. No districts in the data set were steeper than 300 meters / km., so levels 6 and 7 were superfluous. 7. Proximity to the ocean a. variables i. DMSEA: value is one if the district is on the seacoast ii. DMSEANEI: value is one if the district is not itself on the seacoast, but borders another district which is on the seacoast; that is, if the district is in the "second tier" inland b. source: easily constructed by visually inspecting a district-boundary map 222 APPENDIX B: REGRESSION REsuLTs Table 6B.1: List of Instruments for WHYV, GCANCA and NIANCA (Supra-Ricardian model, 1970/71 through 1987/88) Variable Standard Name Description Mean Deviation Edaphic Variables *DMSnn Soil Type Dummies -- DMTSn Topsoil Depth Dummies -- Geographic Variables *AGROBn Agroeconomic Region Dummies -- DMlAQn Aquifer Depth Dummies -- -- Institutional Variables PRWTWG Price Ratio: Wheat to Wage 28.52 14.41 PRRCWG Price Ratio: Rice to Wage 28.60 17.76 PRMZWG Price Ratio: Maize to Wage 20.86 9.77 PRJWWG Price Ratio: Jowar to Wage 19.63 13.19 PRBJWG Price Ratio: Bajra to Wage 19.09 12.34 PRFRWG Price Ratio: Fertilizer to Wage 683.69 301.06 PRTRWG Price Ratio: Tractor to Wage 2309.40 811.41 POPDENI Population Density: 3948.41 2717.01 Technology Variables EXT Extension PRIVRES Private Research 247.94 187.08 LITERACY Male Rural Literacy rate .37 .11 Climate Variables (as defined in Appendix A and in Table 6.3) * JANMDT,* JANMDTSQ, * JANRN~~,* JANRNSQ, * JANMDTRN * APRMDT,* APRMDTSQ, * APRR~N,* APRRNSQ, * APRMDTRN * JULMDT,* JULMDTSQ, * JULRN, * JULRNSQ, * JULMDTRN * OCTMDT,* OCTMDTSQ, * OCTRN, * OCTRNSQ, * OCTMDTRN 223 Table 61B2: Second stage regrsion of modern varieties (WHYV) (Supra-Ricardian Model, 1970/71 through 1987/88) Multiple R .83346 R Square .69465 Adjusted R Square .69173 Standard Error .13390 Analysis of Variance: DF Sum of Squares Mean Square Regression 43 183.22723 4.2610984 Residuals 4492 80.54089 .0179299 F = 237.65387 Signif F = .0000 …---------- Variables in the Equation ------------------ Variable B SE B Beta T Sig T GCANCA .357154 .075073 .292757 4.757 .0000 NIANCA .738063 .062316 .709577 11.844 .0000 STRES5 .000136 .000177 .021037 .772 .4404 EXT .011476 .000776 .260476 14.790 .0000 DMS03 .002540 .013125 .001881 .194 .8466 DMS04 -.007558 .011651 -.005889 -.649 .5166 DMS05 .042430 .008296 .057862 5.114 .0000 DMS06 .016136 .008695 .019592 1.856 .0636 DMS07 -.001629 .009299 -.001757 -.175 .8610 DMS08 .040662 .017034 .042854 2.387 .0170 DMS09 .036131 .018201 .021988 1.985 .0472 DMS10 .033140 .025236 .016534 1.313 .1892 DMS11 .048609 .020029 .029581 2.427 .0153 DMS12 .055735 .018096 .041287 3.080 .0021 DMS13 -.290194 .022068 -.190360 -13.150 .0000 DMS14 -.055462 .014052 -.041084 -3.947 .0001 DMS15 -.075280 .012060 -.084823 -6.242 .0000 DMS16 .052262 .007149 .103660 7.310 .0000 DMS17 .024859 .019893 .010762 1.250 .2115 DMS18 -.261465 .029374 -.092609 -8.901 .0000 DMS19 -.071279 .008601 -.119033 -8.287 .0000 DMS20 .118241 .018723 .058990 6.315 .0000 DMS21 .267520 .022617 .175487 11.828 .0000 JANMDT .093802 .015983 1.386612 5.869 .0000 JANMDTSQ -.002110 .000402 -1.168254 -5.250 .0000 JANRN -.000751 .001318 -.039501 -.570 .5688 JANRNSQ 9.89963339E-05 1.2586E-05 .287959 7.866 .0000 JANMDTRN -.000218 6.2448E-05 -.201555 -3.491 .0005 APRM4DT -.181868 .033151 -1.722097 -5.486 .0000 APRMDTSQ .003237 .000567 1.638528 5.713 .0000 APRRN .007816 .002042 .709219 3.827 .0001 APRRNSQ -1.50380489E-05 4.6485E-06 -.142133 -3.235 .0012 224 Table 6B.2: Second stage regression of modern varieties(WHYV) (cont.) Variable B SE B Beta T Sig T APRMDTRN -.000158 5.9340E-05 -.377586 -2.664 .0077 JULMDT .200461 .045647 2.041751 4.392 .0000 JULMDTSQ -.003297 .000813 -1.766686 -4.055 .0001 JULRN .000565 .000164 .530478 3.446 .0006 JULRNSQ -5.88238133E-08 2.2397E-08 -.101628 -2.626 .0087 JULMDTRN-1.43573001E-05 5.7564E-06 -.302688 -2.494 .0127 OCTMDT -.087417 .050046 -.653945 -1.747 .0808 OCTMDTSQ .000831 .000935 .296904 .888 .3745 OCTRN -.006904 .001134 -1.823514 -6.086 .0000 OCTRNSQ -2.96119591E-06 7.4229E-07 -.187707 -3.989 .0001 OCTMDTRN .000277 4.3940E-05 1.939733 6.295 .0000 (Constant) -.282918 .395738 -.715 .4747 Table 6B.3: Second stage regression of multiple-cropping (GCANCA (Supra-Ricardian Model, 1970/71 through 1987/88) Multiple R .75632 R Square .57202 Adjusted R Square .56763 Standard Error .13308 Analysis of Variance: DF Sum of Squares Mean Square Regression 46 106.25928 2.3099844 Residuals 4489 79.50351 .0177107 F = 130.42845 Signif F = .0000 …V---------------- Variables in the Equation - Variable B SE B Beta T Sig T GCANCA57 .377395 .072195 .273130 5.227 .0000 WHYVNEW .233151 .037748 .284436 6.177 .0000 NIANCA .186687 .085350 .218962 2.187 .0288 DMSLP1 -.222332 .079963 -.537994 -2.780 .0055 DMSLP2 -.263840 .077798 -.547231 -3.391 .0007 DMSLP3 -.309631 .096053 -.502036 -3.224 .0013 DMSLP4 .003765 .102212 .003904 .037 .9706 DMS03 -.065678 .031961 -.059353 -2.055 .0399 DMS04 -.018027 .018577 -.017137 -.970 .3319 DMS05 .008785 .009415 .014616 .933 .3508 DMS06 .047887 .010892 .070935 4.397 .0000 DMS07 -.003753 .011360 -.004940 -.330 .7411 DMS08 -.106929 .020690 -.137485 -5.168 .0000 225 Table 6B.3: Second stage regression of multiple-cropping (GCANCA) (cont.) Variable B SE B Beta T Sig T DMS10 .031258 .028509 .019025 1.096 .2730 DMS11 -.132306 .022050 -.098226 -6.000 .0000 DMS12 .052284 .019974 .047249 2.618 .0089 DMS13 .005629 .032748 .004505 .172 .8635 DMS14 .015973 .016649 .014435 .959 .3374 DMS15 -.017206 .014650 -.023651 -1.174 .2403 DMS16 .001827 .009475 .004421 .193 .8471 DMS17 .069544 .027366 .036730 2.541 .0111 DMS18 .270371 .028586 .116828 9.458 .0000 DMS19 .028059 .008770 .057166 3.199 .0014 DMS20 -.008698 .026683 -.005294 -.326 .7445 DMS21 -.044489 .031543 -.035603 -1.410 .1585 JANMDT .070836 .019834 1.277451 3.572 .0004 JANMDTSQ -.002380 .000533 -1.607883 -4.462 .0000 JANRN .001382 .001533 .088625 .901 .3676 JANRNSQ 6.27504887E-05 1.8917E-05 .222677 3.317 .0009 JANMDTRN -.000316 7.5886E-05 -.356233 -4.162 .0000 APRMDT -.023535 .037061 -.271876 -.635 .5254 APRMDTSQ-5.23051397E-05 .000630 -.032299 -.083 .9339 APRRN -.000348 .002880 -.038496 -.121 .9039 APRRNSQ -3.44414657E-06 6.1910E-06 -.039713 -.556 .5780 APRMDTRN 6.22573446E-05 8.2954E-05 .181406 .751 .4530 JULMDT .170149 .056273 2.114218 3.024 .0025 JULMDTSQ -.003390 .000989 -2.216250 -3.427 .0006 JULRN -.000296 .000236 -.339023 -1.254 .2097 JULRNSQ 1.11617210E-07 3.7958E-08 .235256 2.941 .0033 JULMDTRN 6.12137530E-06 7.5253E-06 .157442 .813 .4160 OCTMDT -.070827 .053375 -.646389 -1.327 .1846 OCTMDTSQ .001586 .000996 .691561 1.593 .1113 OCTRN -.007233 .001545 -2.330586 -4.682 .0000 OCTRNSQ -3.83635353E-06 1.0744E-06 -.296675 -3.571 .0004 OCTMDTRN .000309 6.3457E-05 2.639487 4.862 .0000 (Constant) -.117862 .503849 -.234 .8151 226 Table 6B.4: Second Stage Regression of Irrigation Intensity (NIANCA) (Supra-Ricardian Model, 1970/71 through 1987/88) Multiple R .91081 R Square .82957 Adjusted R Square .82774 Standard Error .09001 Analysis of Variance: DF Sum of Squares Mean Square Regression 48 176.93205 3.6860843 Residuals 4487 36.35033 .0081013 F = 455.00160 Signif F = .0000 …------------- Variables in the Equation ------------------ Variable B SE B Beta T Sig T IRR57 .607678 .039646 .455961 15.328 .0000 WHYVNEW .324140 .013507 .337153 23.998 .0000 DMSLP1 .000940 .056591 .001940 .017 .9867 DMSLP2 -.121120 .060933 -.214187 -1.988 .0469 DMSLP3 -.133937 .064869 -.185156 -2.065 .0390 DMSLP4 .092177 .093152 .081502 .990 .3225 DMAQ1 .011689 .007550 .014998 1.548 .1216 DMAQ2 -.016622 .009135 -.031045 -1.820 .0689 DMAQ3 .031595 .008186 .036257 3.860 .0001 DMS03 -.090025 .030794 -.069365 -2.923 .0035 DMS04 -.054744 .016915 -.044371 -3.236 .0012 DMS05 -.004834 .006798 -.006856 -.711 .4771 DMS06 .016641 .007905 .021017 2.105 .0353 DMS07 -.027274 .009410 -.030609 -2.898 .0038 DMSO8 -.043401 .014476 -.047578 -2.998 .0027 DMS09 -.046483 .011528 -.029423 -4.032 .0001 DMS10 -.042864 .021094 -.022243 -2.032 .0422 DMS11 -.022515 .017738 -.014252 -1.269 .2044 DMS12 -.131475 .0-14065 -.101301 -9.348 .0000 DMS13 .193763 .016890 .132206 11.472 .0000 DMS14 .056663 .011440 .043659 4.953 .0000 DMS15 .041244 .007509 .048337 5.492 .0000 DMS16 -.018724 .007442 -.038629 -2.516 .0119 DMS17 .058669 .022247 .026419 2.637 .0084 DMS18 .069127 .020466 .025467 3.378 .0007 DMS19 .005167 .007203 .008975 .717 .4732 DMS20 .022416 .023211 .011632 .966 .3342 DMS21 -.047832 .023625 -.032636 -2.025 .0430 JANMDT -.075087 .014990 -1.154514 -5.009 .0000 JANMDTSQ .001802 .000428 1.037849 4.205 .0000 JANRN .002441 .000963 .133489 2.535 .0113 227 Table 6B.4: Second stage regression of irrigation intensity (NIANCA) (cont.) Variable B SE B Beta T Sig T JANRNSQ -4.40948615E-05 1.6107E-05 -.133411 -2.738 .0062 JANMDTRN-2.34731078E-05 5.1270E-05 -.022571 -.458 .6471 APRMDT .203175 .026003 2.001085 7.814 .0000 APRMDTSQ -.003398 .000446 -1.789259 -7.628 .0000 APRRN .001231 .002390 .116163 .515 .6066 APRRNSQ -1.62990113E-05 5.4190E-06 -.160236 -3.008 .0026 APRMDTRN-4.44743028E-05 6.5099E-05 -.110488 -.683 .4945 JULMDT -.295957 .039746 -3.135410 -7.446 .0000 JULMDTSQ .005195 .000707 2.896080 7.352 .0000 JULRN -.001103 .000162 -1.077133 -6.798 .0000 JULRNSQ 1.38994064E-07 2.8499E-08 .249777 4.877 .0000 JULMDTRN 3.17318491E-05 5.0643E-06 .695845 6.266 .0000 OCTMDT .082512 .040323 .642036 2.046 .0408 OCTMDTSQ -.001284 .000766 -.477208 -1.677 .0937 OCTRN .003635 .001313 .998629 2.769 .0056 OCTRNSQ -8.48787353E-07 8.8791E-07 -.055964 -.956 .3392 OCTMDTRN -.000134 5.4693E-05 -.979306 -2.455 .0141 (Constant) .824730 .370357 2.227 .0260 Table 6B.5: F-Tests of exclusion of groups of variables Excluded WHYV GCANCA NIANCA Variables Regression Regression Regression Temperature F = 7.64521 F = 26.55391 F = 141.65357 (M = 12) Rainfall F= 48.33569 F = 870.91645 F = 482.25801 (M = 12) _ _ Temperature & Rain F = 18.59580 F = 16.58001 F = 130.18181 (M = 20) l Note: Entries in Table 6B.5 are the F-statistics on the null hypotheses that the excluded variables specified in the left column do not influence the given technology/infrastructure variable. Depending upon the number of variables excluded in each group (given as M = i in the left column), the degrees of freedom of these F-tests range from 12 to 20 for the numerator and more than 4000 for the denominator; thus the 5% critical value of the F statistic is about 1.83; the 1% critical value is about 2.32. 228 Table 6B.6: Regression of the logarithm of net revenue (LNOFPKRE) (Supra-Ricardian Model, 1970/71 itrough 1987/88) Multiple R .78082 R Square .60968 Adjusted R Square .59977 Standard Error 11.27194 Analysis of Variance DF Sum of Squares Mean Square Regression 100 781161.51146 7811.61511 Residual 3936 500094.70342 127.05658 F = 61.48139 Signif F = .0000 -------------- Variables in the Equation ------------------ Variable B SE B Beta T Sig T DMS03 .056995 .053022 .015449 1.075 .2825 DMS04 -.022639 .051602 -.005374 -.439 .6609 DMSOS -.136949 .039221 -.069474 -3.492 .0005 DMS06 -.137941 .040012 -.059337 -3.448 .0006 DMS07 -.002828 .040420 -9.040E-04 -.070 .9442 DMS08 .215560 .082023 .064469 2.628 .0086 DMS09 -.229059 .070459 -.052726 -3.251 .0012 DMS10 -.721042 .116297 -.110921 -6.200 .0000 DMS11 -.266854 .085911 -.052064 -3.106 .0019 DMS12 -.408711 .074279 -.128383 -5.502 .0000 DMS13 .726649 .114239 .110821 6.361 .0000 DMS14 -.306930 .072960 -.067544 -4.207 .0000 DMS15 .090879 .057281 .027618 1.587 .1127 DMS16 -.072123 .035097 -.043765 -2.055 .0399 DMS17 -.378172 .075430 -.058659 -5.014 .0000 DMS18 .181294 .132379 .022327 1.370 .1709 DMS19 -.127367 .037370 -.064565 -3.408 .0007 DMS20 -.230053 .080763 -.034807 -2.848 .0044 DMS21 -.415604 .104479 -.081832 -3.978 .0001 DMPH5 -.087309 .044646 -.042833 -1.956 .0506 DMPH6 -.082979 .039210 -.041612 -2.116 .0344 DMPH8 -.189117 .038805 -.110127 -4.873 .0000 DMPH9 -.351368 .044711 -.162234 -7.859 .0000 JANMDT .127274 .084063 .569203 1.514 .1301 JANMDTSQ -.003552 .002651 -.605100 -1.340 .1805 JANRN .059188 .010767 .879678 5.497 .0000 JANRNSQ .001392 2.4704E-04 1.002708 5.634 .0000 JANMDTRN -.004232 5.2131E-04 -1.208540 -8.119 .0000 APRMDT -.165682 .153778 -.407409 -1.077 .2814 APRMDTSQ .001998 .003222 .277708 .620 .5352 APRRN .033379 .012428 .810168 2.686 .0073 APRRNSQ -2.42956E-04 8.4262E-05 -.513442 -2.883 .0040 APRMDTRN -4.16256E-04 3.6987E-04 -.287228 -1.125 .2605 JULMDT .393804 .204485 1.173100 1.926 .0542 JULMDTSQ .007152 .003604 1.157598 1.985 .0473 JULRN .005808 .001133 1.637535 5.125 .0000 229 Table 6B.6: Regression of the logarithm of net revenue (LNOFPKRE) (cont.) Variable B SE B Beta T Sig TJULRNSQ 3.02613E-06 1.3168E-06 1.768303 2.298 .0216 JULMDTRN -2.18016E-04 3.8814E-05 -1.330956 -5.617 .0000 OCTMDT .830156 .225847 1.572793 3.676 .0002 OCTMDTSQ -.030153 .005132 -2.846942 -5.875 .0000 OCTRN -.029832 .006642 -2.353529 -4.491 .0000 OCTRNSQ 3.71860E-05 1.0508E-05 .690590 3.539 .0004 OCTMDTRN .001230 2.5186E-04 2.617094 4.885 .0000 JURNCV .307770 .084846 .081715 3.627 .0003 JARNCV .550359 .081982 .122917 6.713 .0000 JUNERAIN 2.30215E-05 1.2469E-04 .003477 .185 .8535 JUAURAIN 2.34096E-04 7.7756E-05 .085834 3.011 .0026 YEARRAIN 3.95782E-05 5.6573E-05 .023636 .700 .4842 DMSEA .044959 .060905 .018225 .738 .4605 DMSEANEI .131669 .039196 .058391 3.359 .0008 LNCSTCLT .088365 .025583 .098499 3.454 .0006 LNCSTBUL -.045104 .020480 -.051291 -2.202 .0277 RELWAGEK -.105498 .026671 -.076487 -3.955 .0001 GCNCPRDX 2.320813 1.513060 .508790 1.534 .1251 EXT -.001490 .004236 -.010294 -.352 .7250 WHYVPRDX 1.737278 1.022308 .493028 1.699 .0893 STRES5 .002314 2.8632E-04 .118217 8.080 .0000 NINCPRDX -3.796297 1.417844 -1.032130 -2.678 .0074 JAMDSQHX 6.47420E-04 .001463 .065089 .443 .6581 JAMDSQGX -.001685 .001714 -.313307 -.983 .3256 JAMDSQNX .005567 .001817 .534606 3.064 .0022 APMDSQHX -.004586 .001154 -1.092933 -3.973 .0001 APMDSQGX 2.34319E-04 .002139 .049470 .110 .9128 APMDSQNX .005238 .001680 1.187119 3.118 .0018 JUMDSQHX .004363 .001201 1.155292 3.634 .0003 JUMDSQGX -.015282 .001791 -4.702900 -8.532 .0000 JUMDSQNX .012216 .001644 3.144438 7.430 .0000 OCMDSQHX -5.29263E-04 .002051 -.103950 -.258 .7964 OCMDSQGX .016836 .003339 3.135448 5.043 .0000 OCMDSQNX -.019089 .002955 -3.576864 -6.459 .0000 JARNSQHX -1.11001E-04 1.5172E-04 -.049372 -.732 .4644 JARNSQGX -8.28585E-04 2.4846E-04 -.772173 -3.335 .0009 JARNSQNX -6.41624E-05 1.7826E-04 -.026867 -.360 .7189 APRNSQHX 5.71100E-06 7.5223E-05 .003227 .076 .9395 APRNSQGX 7.29124E-05 6.0648E-05 .200157 1.202 .2293 APRNSQNX 1.01295E-04 8.6140E-05 .052133 1.176 .2397 JURNSQHX 4.64521E-07 2.7251E-07 .053208 1.705 .0883 JURNSQGX -3.32555E-06 1.2816E-06 -2.173201 -2.595 .0095 JURNSQNX 1.54934E-06 1.2672E-06 .134677 1.223 .2215 OCRNSQHX -5.61231E-06 7.0628E-06 -.047969 -.795 .4269 OCRNSQGX -3.74039E-05 9.4436E-06 -.904898 -3.961 .0001 OCRNSQNX 6.04045E-06 1.1142E-05 .056100 .542 .5878 ALT 1.81160E-04 1.3479E-04 .057730 1.344 .1790 DMYR71 .002039 .047913 5.828E-04 .043 .9661 230 Table 6B.6: Regression of the logarithm of net revenue (LNOFPKRE) (cont.) Variable B SE B Beta T Sig T DMYR72 .120618 .049685 .034072 2.428 .0152 DMYR73 .577850 .049345 .166707 11.711 .0000 DMYR74 .620182 .050498 .175841 12.281 .0000 DMYR75 .488680 .049782 .140511 9.816 .0000 DMYR76 .339487 .051720 .096642 6.564 .0000 DMYR77 .504268 .052922 .145091 9.528 .0000 DMYR78 .397011 .053030 .114114 7.487 .0000 DMYR79 .199352 .055501 .056621 3.592 .0003 DMYR80 .576301 .057131 .164354 10.087 .0000 DMYR81 .555922 .061011 .159137 9.112 .0000 DMYR82 .585726 .065561 .167167 8.934 .0000 DMYR83 .783142 .069882 .227598 11.207 .0000 DMYR84 .682641 .072775 .191490 9.380 .0000 DMYR85 .648057 .074738 .180197 8.671 .0000 DMYR86 .574351 .077147 .158605 7.445 .0000 DMYR87 .700880 .079386 .186410 8.829 .0000 (Constant) -12.005503 3.161783 -3.797 .0001 Table 6B.7: F-Tests of exclusion of groups of variables - LNOFPKRE regression Excluded Variables F-Statistic Edaphic (M =23) F = 20.17573 l Temperature (M = 12) F = 20.89333 Temperature and Interactions (M = 24) F = 16.13143 Rainfall (M = 12) F =23.12331 Rainfall and Interactions (M = 24) F = 14.71552 Temperature & Rain (M = 20) F = 21.46841 Temperature, Rain and Interactions (M = 44) F =15.71037 Technology and Related Infrastructure (M = 30) F = 24.89907 Technology - Climate Interactions (M = 24) F = 12.51359 Year Dummies (M = 17) F = 26.34394 Note: Entries in Table 6B.7 are the F-statistics on the null hypotheses that the excluded variables specified in the left column do not influence Net Revenue. Depending upon the number of variables excluded in each group (given as M = i in the left column), the degrees of freedom of these F-tests are anywhere from 12 to 44 for the numerator and more than 4000 for the denominator; thus the 5% critical value of the F statistic would range from about 1.5 to about 2.1; correspondingly, the 1% critical value of the F statistic would range from about 1.8 to about 2.8. 231 APPENDIX C: ALTERNATE ESTIMATES OF CLIMATE EFFECTS UTILITIZING SEASONAL CLIMATE VARIABLES INTRODUCTION An earlier stage of work on this project led to the development of alternative specifications of the normal climate variables. This specification differed in two dimensions from that reported in the body of this paper (which was designed to conform to the simple "Ricardian" estimates of Sanghi et al., 1997). The first was to include maximum temperatures and the maximum - minimum temperature range. The second was to utilize cropping season weights to define "seasonal" normal temperature and climate variables. The seasonal weights are open to the criticism that they may be "endogenous"; i.e., that they reflect farmers' behavioral responses to climate. In the simple Ricardian estimates of Sanghi et al. where technology is ignored the strategy of using four arbitrary months to characterize climate worked reasonably well. When the model is expanded to incorporate technology the four-month specification appears to be less stable than the seasonal specification. Hence we report results here for comparative purposes. The advantage of the seasonal weights is that we do not expect all months to have equal effects. For example, many parts of India ahve single cropping seasons - wet versus dry or warm versus cool, etc. A single linear and quadratic term (i.e., with only two coefficients) may not capture the differences between cropping seasons which vary by region. The question of the endogeneity of these weights is really a question of profit-maximizing behavior of farmers. If cropping seasons are optimizing responses to prices, technology and normal climate they can be regarded to be functions of exogenous variables. (NORMAL) TEMPERATURE RANGE VARIABLES Most studies of the climate impact on agriculture use only one temperature variable, measured over several months. Though often not specified precisely, it is usually (and incorrectly) termed an "average" temperature: incorrectly, because it is almost always constructed as the average of (thus the midpoint between) the monthly normal minimum temperature and the monthly normal maximum temperature, rather than a true average of all the normal temperatures experienced at all hours through the month. But even if the average were constructed properly, this practice of collapsing all temperature data into a single variable discards very useful information, and will almost certainly render simulations of responses to long-term climate change inferred from cross-sectional differences in so-called average temperatures biased and unstable. In the study reported in this Appendix we use two normal temperature variables instead of only one: the normal maximum temperature and normal temperature range, which is defined as the difference between the normal maximum and minimum temperatures. Agronomically, if a change in temperature is going to harm Indian (crop) agriculture, the maximum temperature will be the culprit, and distinctly not the temperature midpoint. And the difference between maximum temperature and the so-called average temperature can be so large and so variable in parts of India that only small differences in so-called average temperature might appear, masking 232 substantial differences in the maximum temperature. So it is important that maximum temperature be included in the regression equations. To put the point a different way: the temperature midpoint, or the so-called average temperature, varies from 12.8344 (in the maturation season, in Rajasthan) to 31.7237 (in the sowing season, in Orissa). The correlation between state-wise maximum temperature and temperature midpoint is 0.96, which might seem to indicate that either variable can interchangeably be used. But there is a wide distribution of normal maximum temperatures corresponding to the districts and seasons (or months) whose temperature range falls in any interval. Those ranges of normal maximum temperatures overlap substantially: Given that, in the current cross-section, District B's temperature midpoint is one degree higher than District A's temperature midpoint, the maximum temperature in District B could be several degrees higher or even lower than the maximum temperature in District A, and simply knowing the difference in their temperature midpoints tells us nothing about any differences in their normal maximum temperatures. In fact, many districts with higher temperature midpoints than others have a lower maximum temperature, and thus are under less current temperature stress. We want to simulate (the effects of, and farmers' responses to) an increase in temperature over time by looking at current cross-section differences in temperature. As explained above, the appropriate temperature variable is maximum temperature, but we cannot infer much of anything about cross-sectional (let alone future!) differences in maximum temperature from cross- sectional differences in temperature midpoint. Thus using cross-section differences in temperature midpoints to infer differences in (either current or future) maximum temperatures is biased and imprecise, at best: the maximum temperature itself must be included in the regression equation and in all the resulting effects computations, so that current cross-section differences in maximum temperatures (as well as farmers' current responses to those cross-sectional differences) can be used, directly and validly, to simulate long-term changes in maximum temperature (and farmers' responses over time to any such long-term climate changes). For our second temperature variable we choose to use temperature range, rather than minimum temperature (which would convey the same information), because the former is a more appropriate framework within which to consider the effects of temperature changes. Few if any global climate change scenarios explicitly treat the issue of what would happen to minimum temperatures; in the absence of any compelling reason to believe otherwise, it is reasonable to assume that global climate change would cause both the minimum and the maximum temperatures to increase by the same amount25. This implies that the temperature range would not change. Now in computing the impact of a temperature change, it makes no sense to have two terms each season -- both the minimum and the maximum temperatures -- changing in response to the single phenomenon that temperatures are rising. By casting the model in terms of the maximum temperature and the temperature range, we capture the increase in temperature, while we also maintain the richness of the data with two variables. 25 Or possibly by the same proportion, but the difference between saying that they both increase by one degree Celsius vs. that they both increase by, say, three percent is quite small. 233 SEASONAL CLIMATE VARIABLES One important transformation in the data is the creation of seasonal climate variables. Most Ricardian studies, including all those in this volume and the Supra-Ricardian study reported in the body of this paper, choose four evenly spaced months for climate variables: often, January, April, July and October96. The stated intent is to capture seasonal climatic effects throughout the year, although in some cases those may be the only four months available27. Data for the important climate variables in India -- maximum temperature, temperature range, rainfall, and cloud cover -- are available for each of the twelve months. The values of these variables within any given month differ widely across India, as do the cropping patterns and cropping seasons. That is to say, the planting season for a given crop in one state may differ by four or five months from the planting season for the same crop in another state; further, the sowing (or harvesting) season for one crop in one state may differ by many months from the sowing (or harvesting) season of another crop in that state28. Rather than arbitrarily choose four months from the twelve available, and rather than lose the important information which cropping calendars provide, this study is based on seasonal climate variables rather than on monthly climate variables. Three climatic seasons are defined: sowing, maturation, and harvesting. The sowing and harvesting seasons for a district are a weighted average of the sowing and harvesting seasons for the five major food crops: rice, wheat, maize, jowar and bajra. For example, to construct the sowing season maximum temperature variable for a district, we first computed the average nornal maximum temperature, in that district, during the months of the rice sowing season in that district's state. We did the same for the wheat sowing season, for the maize sowing season, and for jowar and bajra. Then we constructed a weighted average of the five crop-based sowing season maximum temperatures, using the share of area under each crop as weights. The procedure for the harvest season maximum temperature variable was the same, using of course each state's harvest months instead. And the procedures for temperature range, rainfall and oktas were similar as well. The maturation season is defined as the middle third of the interval from the beginning of the sowing season to the end of the harvesting season. Appendix D contains comparisons of regressions using the seasonal climate variables here defined, vs. regressions equivalent except for the use of four monthly climate variables. The comparisons are semi-Ricardian: no technology or infrastructural variables appear. The seasonal specifications are superior not only because of the theoretical reasons discussed above (namely, that the seasonal climate variables capture crop calendar information, and that the seasonal specification includes months appropriate for each given district rather than four arbitrarily chosen months imposed uniformly across the nation) but also as seen in Appendix D the seasonal 26 For which it could be argued, for example in North America, that in April the planting might begin, that in July nearly every summer crop is growing, that by October the harvesting should be well underway, and that December represents either the maturation experience of winter crops or the effects of frost on weed and insect control or of snow on next season's soil moisture content. 27 Even if they are the only months available, the use for the entire area of the study of only four months, spaced over the year, may serve the purpose well if the entire area of the study is sufficiently homogeneous in tenns of cropping seasons. 28 See Appendix D for a table displaying the cropping seasons of the major crops in each State. 234 specification is superior in terms of the consistency of the computed climate effects (especially the temperature effects) across states. There is still considerable cross-sectional variation in the seasonal climate variables, especially rainfall; the state-wise means of the six major seasonal climate variables are displayed below in Table 6C.la (normal maximum temperature and normal rainfall) and in Table 6C.lb (normal temperature range and normal oktas of cloud cover). Table 6C.la: Means of seasonal climate variables: maximum temperature and rainfall Seasonal Maximum Temperature Seasonal Rainfall SOWX MATXT HARXT SOWRN1 MATRNI HARRNI _ _ _ _ _ _ _ _ _ _ _ _ T I 1 1 India 31.9168 25.9508 28.1703 125.4206 130.0125 67.7689 Andhra Pradesh 34.8194 33.3519 32.8285 79.4097 94.0577 138.0382 Haryana 32.9257 18.5846 25.0597 70.2200 46.8836 16.2340 Madhya Pradesh 30.9787 27.7285 30.0951 139.0140 113.4026 39.0939 Maharastra 27.8751 23.2293 26.3539 282.7803 128.2241 16.7087 Karnataka 28.4505 24.5282 26.4117 213.9141 183.5560 71.9228 Punjab 31.5863 26.2758 34.7293 49.0225 69.2831 20.7722 Tamil Nadu 32.1970 27.9427 26.5515 102.0830 126.6152 100.6109 Uttar Pradesh 31.7724 27.7984 30.6421 64.0392 103.1376 42.5272 Bihar 31.8867 28.7979 32.7372 142.0795 235.3772 184.8334 Gujarat 31.9103 19.5360 21.2827 138.8117 60.6682 6.5967 Rajasthan 34.3675 15.7135 17.8448 68.2971 30.5153 6.7726 Orissa 36.9648 30.4839 31.0263 127.9972 328.0940 177.0605 West Bengal 33.5293 31.3290 30.4340 120.3099 285.3750 198.5125 ESTIMATES: TECHNOLOGY AND RELATED INFRASTRUCTURE Recall that in the Supra-Ricardian framework we treat the three technology variables (the use of modem varieties, irrigation intensity and multiple-cropping) as endogenous variables, whose predicted values then contribute to the determination of net revenue per hectare. In this Section we report estimates of the determinants of these variables, in a two-stage least squares fiamework. Then in the next Section we report computed climate effects on technology and related infrastructure. 235 Table 6C.lb: Means of seasonal climate variables: temperature range and Oktas Seasonal Temperature Range Seasonal Cloud Cover (Oktas) SOWTR MATTR HARTR SOWOKI MATOK HAROK 1 1 1 1 1 India 10.2370 8.5968 10.9071 4.3093 4.1253 2.9320 Andhra Pradesh 10.0475 9.7746 8.3156 4.6233 5.0553 6.0997 Haryana 12.9440 8.9384 10.4906 2.5807 2.0279 1.1753 Madhya Pradesh 10.3878 11.1489 14.4372 4.7274 3.9616 2.5719 Maharashtra 7.4549 7.5450 12.2927 6.0890 4.1782 2.0043 Karnataka 7.9985 6.7725 9.1874 6.2053 5.8455 4.2788 Punjab 14.7541 12.5580 15.8348 1.9663 2.8732 1.6549 Tamil Nadu 8.3840 7.2662 7.4497 5.7079 5.0726 4.2446 Uttar Pradesh 11.8617 10.6232 13.0831 2.5260 3.4288 1.8594 Bihar 9.5451 7.4634 8.8069 4.1308 5.3137 4.8124 Gujarat 8.0186 6.3123 10.2572 4.8153 2.6433 1.0618 Rajasthan 11.2709 5.7583 8.1286 3.8902 1.9670 1.1171 Orissa 10.4822 5.7740 7.4696 5.2171 6.9747 5.4721 West Bengal 11.1471 7.0478 7.9376 3.7219 5.9979 4.7859 LI - MULTIPLE CROPPING - IRRIGATION SYSTEM The proportion of area sown to modem varieties (WTHYV), a measure of multiple- cropping (GCANCA), and irrigation intensity (NIANCA) were estimated by two-stage least squares. Table 6C.2 displays the variables used as instruments in the 2SLS system; an asterisk following a variable name denotes that that variable also appears in one or more second-stage regression equations, and/or the net revenue equation. These instruments include fundamental climatic and edaphic variables, as well as two technology variables and a number of price ratios proxying institutional factors. Tables 6C.3 through 6C.5 then display the regression results from the second stage; WHYV in Table 6C.3, GCANCA in Table 6C.4, and NIANCA in Table 6C.5. A number of striking results emerge. First is the degree to which this system captures the modeled behavior. Grossly, all three second-stage regressions have highly significant F-statistics, and adjusted R2s which range from 0.59 to 0.85. For each of the three regressions, we tested the null hypotheses that the rainfall variables taken as a group did not significantly influence the technology variables, that the temperature variables taken as a group did not significantly influence the technology variables, and that the climate variables (that is, the temperature and rainfall variables combined) taken as a group did not significantly influence the technology variables. The results of the F-tests of sets of excluded variables are reported in Table 6C.6. All three of the null hypotheses were rejected in 236 all three of these regressions: as groups, the climate variables do significantly influence the adoption of modern varieties, the extent of multiple-cropping, and irrigation intensity. But even more important than the general goodness-of-fit of these regression equations, and the significance of groups of variables, are the patterns revealed within each equation. The coefficients on six or seven of the seven agroeconomic region dummies, and on either twelve or fifteen of the nineteen soil type dummies, are significant29. Slope significantly influences irrigation intensity30 but not multiple-cropping. And irrigation intensity tends to be higher in districts above aquifers which are geologically thickest3". The second-stage variables exercise an important influence on each other: the coefficients of both GCANCA and NIANCA on WHYV are significantly positive; as are the coefficients of both WHYV and NIANCA on GCANCA and the coefficient of WHYV on NIANCA. That is to say, the adoption of modem high-yielding varieties, multiple-cropping and irrigation are mutually-reinforcing. The adoption of modern varieties also responds favorably to greater extension activity; perhaps surprisingly, though, additional state-level agricultural research does not significantly increase the adoption of modern varieties. There is considerable inertia in this behavior: both the extent of multiple-cropping and irrigation intensity are highest in those districts in which such activity was largest in 195732. 29 In neither the WHYV nor the NIANCA equation is the coefficient on AGROB6 significantly different from zero; once the coefficient has a negative sign, once positive. 30 This may reflect the importance of drainage to avoid waterlogging or soil salinity; it may reflect the geological requirements for proper functioning of a canal system. 3 This does not measure the annual water depth within the aquifer, but rather a long term geological potential. Farmers may respond to this in their cropping choices; farmers and probably governments also respond in their irrigation investments. 32 There is no related variable for the use of modem varieties, because no such modern variety existed before the onset of the Green Revolution in the middle 1960s. 237 Table 6C.2: List of instruments for WHYV, GCANCA and NIANCA (Seasonal Supra-Ricardian Model, 1970/71 through 1987/88) Variable Standard Name Description Mean Deviation Edaphic Variables * DMSnn Soil Type Dummies - - DMTSn Topsoil Depth Dummies STRA Storie A: character of the soil profile 82.48 4.36 STRB Storie B: topography, texture, structure 88.39 7.23 STRC Storie C: salinity, climatic suitability, erodability 82.48 6.49 Geographic Variables * AGROBn Agroecononic Region Dumnies - - DMAQn Aquifer Depth Dummies - - Institutional Variables PRWTWG Price Ratio: Wheat to Wage 28.52 14.41 PRRCWG Price Ratio: Rice to Wage 28.60 17.76 PRMZWG Price Ratio: Maize to Wage 20.86 9.77 PRJWWG Price Ratio: Jowar to Wage 19.63 13.19 PRBJWG Price Ratio: Bajra to Wage 19.09 12.34 PRFRWG Price Ratio: Fertilizer to Wage 683.69 301.06 PRTRWG Price Ratio: Tractor to Wage 2309.40 811.41 POPDEN1 Population Density: 3948.41 2717.01 Technology Variables PRIVRES Private Research 247.94 187.08 LITERACY Male Rural Literacy rate .37 .11 Climate Variables (as defined in Appendix B and in Table 6B.1) * SOWXT1 * SOWXrlSQ * SOWRN1 * SOWRN1SQ * SOWXT1RN * MATXT1 * MATXTlSQ * MATRN1 * MATRN1SQ * MATXT1RN * HARXT1 * HARXT1SQ * HARRN1 * HARRN1SQ * HARXT1RN 238 Table 6C.3: Second stage regression of modern varieties (WHYV) (Seasonal Supra-Ricardian Model, 1970/71 through 1987/88) Multiple R .84331 R Square .71118 Adjusted R Square .70838 Standard Error .12904 Analysis of Variance: DF Sum of Squares Mean Square Regression 45 190.60630 4.2356956 Residuals 4649 77.40926 .0166507 F = 254.38492 Signif F = .0000 ------------------ Variables in the Equation ------------------ Variable B SE B Beta T Sig T GCANCA .307627 .043664 .253495 7.045 .0000 NIANCA .680667 .032664 .659120 20.838 .0000 STRES5 .000139 .000152 .024447 .912 .3617 EXT .011441 .000677 .260217 16.903 .0000 AGROB1 .033334 .014025 .030187 2.377 .0175 AGROB2 -.125362 .030126 -.100506 -4.161 .0000 AGROB3 -.058120 .013451 -.071144 -4.321 .0000 AGROB4 -.057650 .011483 -.115212 -5.021 .0000 AGROB5 .062355 .018212 .060123 3.424 .0006 AGROB6 -.012877 .013247 -.019376 -.972 .3311 AGROB7 -.071478 .021040 -.057726 -3.397 .0007 DMS03 .034400 .013131 .026541 2.620 .0088 DMS04 -.010029 .011680 -.007738 -.859 .3906 DMS05 .076805 .007870 .105194 9.759 .0000 DMS06 .031945 .008021 .039103 3.983 .0001 DMS07 .020784 .009085 .022222 2.288 .0222 DMS08 -.026146 .012428 -.027304 -2.104 .0354 DMS09 -.006688 .014582 -.004029 -.459 .6465 DMS10 -.002967 .023812 -.001465 -.125 .9008 DMS11 .082018 .021682 .049407 3.783 .0002 DMS12 -.020910 .013926 -.016764 -1.501 .1333 DMS13 -.132632 .015649 -.086129 -8.475 .0000 DMS14 -.083213 .013004 -.061030 -6.399 .0000 DMS15 -.058156 .010260 -.064943 -5.668 .0000 DMS16 .069466 .007234 .138456 9.603 .0000 DMS17 .015308 .020943 .006559 .731 .4649 DMS18 -.213001 .025426 -.074659 -8.377 .0000 DMS19 -.042767 .008908 -.072323 -4.801 .0000 DMS20 .119588 .017363 .059050 6.888 .0000 DMS21 .247663 .020872 .160828 11.866 .0000 239 Table 6C.3: Second stage regression of modern varieties (WHYV) (cont.) (Seasonal Supra-Ricardian Model, 1970/71 through 1987/88) Variable B SE B Beta T Sig T SOWXT1 -.047284 .009889 -.571475 -4.781 .0000 SOWXT1SQ .000729 .000153 .548439 4.778 .0000 SOWRN1 -.000342 .000301 -.210549 -1.138 .2552 SOWRNlSQ-1.66367627E-07 6.9280E-08 -.122731 -2.401 .0164 SOWXT1RN 2.98864445E-05 8.4443E-06 .484225 3.539 .0004 MATXT1 .031145 .005044 .877787 6.174 .0000 MATXT1SQ -.000489 8.1288E-05 -.597553 -6.011 .0000 MATRN1 -.000267 .000285 -.142633 -.938 .3481 MATRN1SQ 1.38146584E-07 9.2523E-08 .046088 1.493 .1355 MATXT1RN 1.53320826E-06 9.2947E-06 .024058 .165 .8690 HARXT1 -.050837 .005432 -1.399910 -9.359 .0000 HARXT1SQ .000937 8.4973E-05 1.124229 11.024 .0000 HARRN1 .000602 .000372 .210397 1.619 .1056 HARRNlSQ-1.12173086E-07 2.4333E-07 -.014344 -.461 .6448 HARXTlRN-1.95967163E-05 1.0734E-05 -.207222 -1.826 .0680 (Constant) .522724 .166049 3.148 .0017 Table 6C.4: Second stage regression of multiple cropping (GCANCA) (Seasonal Supra-Ricardian Model, 1970/71 tirough 1987/88) Multiple R .77336 R Square .59809 Adjusted R Square .59394 Standard Error .12441 Analysis of Variance: DF Sum of Squares Mean Square Regression 48 107.01720 2.2295251 Residuals 4646 71.91493 .0154789 F = 144.03649 Signif F = .0000 ------------------ Variables in the Equation ------------------ Variable B SE B Beta T Sig T GCANCA57 .381691 .036234 .277712 10.534 .0000 WHYVNEW .275614 .025884 .334468 10.648 .0000 NIANCA .098844 .040834 .116154 2.421 .0155 DMSLP1 .071175 .078187 .172870 .910 .3627 240 Table 6C.4: Second stage regression of multiple cropping (GCANCA) (cont.) Variable B SE B Beta T Sig T DMSLP2 .020761 .076208 .043331 .272 .7853 DMSLP3 .033498 .078041 .055677 .429 .6678 DMSLP4 .170439 .109421 .180864 1.558 .1194 DMS03 -.015676 .015219 -.014677 -1.030 .3031 DMS04 .036112 .015209 .033812 2.374 .0176 DMS05 -.021918 .010062 -.036430 -2.178 .0294 DMS06 .001582 .009056 .002350 .175 .8613 DMS07 -.029221 .009996 -.037913 -2.923 .0035 DMS08 -.058410 .015207 -.074020 -3.841 .0001 DMS09 -.050500 .014687 -.036917 -3.439 .0006 DMS10 .095620 .023465 .057297 4.075 .0000 DMS11 .048599 .027299 .035527 1.780 .0751 DMS12 -.002948 .015659 -.002868 -.188 .8507 DMS13 .091849 .016475 .072381 5.575 .0000 DMS14 .032304 .013171 .028751 2.453 .0142 DMS15 -.025041 .010672 -.033934 -2.346 .0190 DMS16 -.001320 .008165 -.003192 -.162 .8716 DMS17 -.021727 .022379 -.011297 -.971 .3316 DMS18 .213185 .025692 .090680 8.298 .0000 DMS19 .054043 .008284 .110909 6.524 .0000 DMS20 -.016090 .020713 -.009641 -.777 .4373 DMS21 -.041430 .021991 -.032649 -1.884 .0596 AGROB1 -.054427 .014016 -.059815 -3.883 .0001 AGROB2 -.151715 .030114 -.147608 -5.038 .0000 AGROB3 -.040815 .013362 -.060629 -3.055 .0023 AGROB4 -.072686 .011229 -.176281 -6.473 .0000 AGROB5 -.214874 .014623 -.251425 -14.695 .0000 AGROB6 -.112961 .013604 -.206276 -8.303 .0000 AGROB7 -.108151 .024184 -.105994 -4.472 .0000 SOWXT1 .043482 .013748 .637744 3.163 .0016 SOWXT1SQ -.000699 .000223 -.637951 -3.129 .0018 SOWRN1 -.000788 .000543 -.587800 -1.451 .1468 SOWRN1SQ 2.34341245E-07 1.2344E-07 .209791 1.898 .0577 SOWXT1RN 1.37449868E-05 1.5139E-05 .270253 .908 .3640 MATXT1 .000923 .005955 .031561 .155 .8769 MATXT1SQ 9.23301170E-06 .000101 .013703 .092 .9269 MATRN1 .001206 .000415 .781618 2.903 .0037 MATRN1SQ-1.38682103E-07 8.8016E-08 -.056146 -1.576 .1152 MATXT1RN-2.64236313E-05 1.3191E-05 -.503164 -2.003 .0452 HARXT1 -.010345 .006535 -.345695 -1.583 .1135 HARXT1SQ .000200 .000108 .291898 1.854 .0638 HARRN1 -.000294 .000558 -.124585 -.527 .5984 HARRN1SQ 5.81148921E-07 2.5216E-07 .090181 2.305 .0212 HARXT1RN-7.11900123E-06 1.7506E-05 -.091354 -.407 .6843 (Constant) .145277 .176192 .825 .4097 241 Table 6C.5: Second stage regression of irrigation intensity (Seasonal Supra-Ricardian Model, 1970/71 through 1987/88) Multiple R .92280 R Square .85156 Adjusted R Square .84996 Standard Error .08526 Analysis of Variance: DF Sum of Squares Mean Square Regression 50 193.66011 3.8732022 Residuals 4644 33.75719 .0072690 F = 532.83907 Signif F .0000 --------V--------- Variables in the Equation -- Variable B SE B Beta T Sig T IRR57 .745007 .024082 .555960 30.937 .0000 WHYVNEW .320394 .012394 .330868 25.851 .0000 DMSLP1 -.150492 .056412 -.311043 -2.668 .0077 DMSLP2 -.235273 .055060 -.417866 -4.273 .0000 DMSLP3 -.243622 .053418 -.344580 -4.561 .0000 DMSLP4 -.201977 .080731 -.182390 -2.502 .0124 DMAQ1 .010877 .006393 .014141 1.701 .0889 DMAQ2 .014509 .005800 .026799 2.501 .0124 DMAQ3 .036886 .007265 .042537 5.077 .0000 DMS03 -.019474 .011098 -.015516 -1.755 .0794 DMS04 -.029165 .010978 -.023238 -2.657 .0079 DMS05 -.029956 .007280 -.042370 -4.115 .0000 DMS06 .016773 .006453 .021202 2.599 .0094 DMS07 -.007098 .006890 -.007837 -1.030 .3030 DMS08 -.051066 .011035 -.055070 -4.628 .0000 DMS09 -.047315 .010345 -.029434 -4.573 .0000 DMS10 -.052963 .015832 -.027007 -3.345 .0008 DMS11 -.125462 .019036 -.078048 -6.591 .0000 DMS12 -.067498 .010703 -.055884 -6.306 .0000 DMS13 .083194 .010846 .055790 7.670 .0000 DMS14 .064895 .008689 .049152 7.468 .0000 DMS15 .043998 .007045 .050738 6.245 .0000 DMS16 -.020143 .005578 -.041460 -3.611 .0003 DMS17 .010536 .015530 .004662 .678 .4976 DMS18 .056925 .018037 .020605 3.156 .0016 DMS19 -.005673 .005783 -.009906 -.981 .3267 DMS20 -.016687 .014836 -.008509 -1.125 .2607 242 Table 6C.5: Second stage regression of irrigation intensity (cont.) Variable B SE B Beta T Sig T DMS21 .020128 .015612 .013498 1.289 .1974 AGROB1 -.025563 .009554 -.023907 -2.676 .0075 AGROB2 .088847 .020363 .073560 4.363 .0000 AGROB3 .040550 .009240 .051260 4.389 .0000 AGROB4 .021023 .007540 .043387 2.788 .0053 AGROB5 -.036529 .010704 -.036373 -3.413 .0006 AGROB6 .007476 .009072 .011617 .824 .4100 AGROB7 .047768 .016626 .039839 2.873 .0041 SOWXT1 .036231 .009450 .452201 3.834 .0001 SOWXT1SQ -.000604 .000152 -.469461 -3.967 .0001 SOWRN1 -.002409 .000352 -1.529429 -6.847 .0000 SOWRN1SQ 6.03110265E-07 7.8799E-08 .459464 7.654 .0000 SOWXT1RN 6.19190596E-05 9.9217E-06 1.036017 6.241 .0000 MATXT1 -.039183 .003710 -1.140457 -10.562 .0000 MATXT1SQ .000594 6.3898E-05 .750042 9.294 .0000 MATRN1 .001392 .000269 .767824 5.184 .0000 MATRN1SQ 1.17076740E-07 6.0084E-08 .040336 1.949 .0514 MATXTlRN-5.09914616E-05 8.5788E-06 -.826286 -5.944 .0000 HARXT1 .047641 .004131 1.354770 11.533 .0000 HARXT1SQ -.000778 6.9216E-05 -.964553 -11.243 .0000 HARRN1 -.001426 .000378 -.514398 -3.776 .0002 HARRN1SQ 1.09153369E-07 1.7617E-07 .014414 .620 .5356 HARXT1RN 4.38773168E-05 1.1826E-05 .479140 3.710 .0002 (Constant) -.344438 .120651 -2.855 .0043 Table 6C.6: F-Tests of exclusion of groups of variables (Seasonal climate terms) Excluded WHYV GCANCA NIANCA Variables Regression Regression Regression Temperature F = 41.12268 F = 34.37395 F = 42.05588 (M = 9) Rainfall F =4.38067 F = 13.99238 F = 53.77791 (M = 9) Temperature and Rain F =28.24379 F = 31.86579 F = 42.88782 (M= 15) Note: Entries in Table 6C.6 are the F-statistics on the null hypotheses that the excluded variables specified in the left column do not influence the given technology/infrastructure variable. Depending upon the number of variables excluded in each group (given as M = i in the left column), the degrees of freedom of these F-tests range from 9 to 15 for the numerator and more than 4000 for the deno-minator; thus the 5% critical value of the F statistic is about 1.83; the 1% critical value is about 2.32. 243 The role of the climate variables in determining these three is especially interesting. Fifteen climate variables appear in each of these three equations, five for each of the three seasons: normal maximum temperature, normal maximum temperature squared, normal rainfall, normal rainfall squared, and the interaction of maximum normal temperature and normal rainfall. Of the fifteen, the coefficients of nine are significantly different from zero in the WHYV equation; the coefficients of seven are significantly different from zero in the GCANCA equation, and the coefficients of fourteen in the NIANCA equation. In both the WHYV and the NIANCA equations, which exhibit a better fit with the climate variables, one can see three important alternating patterns in the temperature variables: first, the sign on maximum maturation season temperature is the opposite of the sign on both maximum sowing season and maximum harvesting season temperature; further, the sign on each temperature squared term is the opposite of the sign on the same season's temperature term, denoting diminishing returns to the temperature effects; and third, the signs on the temperature terms in the WHYV equation are the opposite of the signs on the corresponding temperature terms in the NIANCA equation. (The computed net effect of temperature and rainfall on the adoption of modem varieties, multiple- cropping and irrigation are reported subsequently.) CLIMATE CHANGE EFFECTS ON TECHNOLOGY AND INFRASTRUCTURE Based on the regression results reported in the previous section, and the actual district- level values of the climate and technology variables, one can compute the predicted effects on high-yielding variety use, multiple cropping, and irrigation intensity of changes in normal temperature and rainfall. The motivation for these computations is the by-now familiar predictions of global warming; we make no effort to validate or calibrate the predicted changes, but merely use the familiar predictions to drive a simulation. A number of possible values for temperature and rainfall change have been proposed; no scenario has received unanimous agreement. We predict the effects of a one degree Celsius temperature increase, and a three percent rainfall increase, values which some models predict could be achieved from a generation to a century from now. The impacts of different changes could easily be scaled. These predicted effects are presented in Table 6C.7 in which appear first the temperature effects on the adoption of modern varieties, the extent of multiple-cropping, and irrigation intensity. Next appear the rainfall effects on WHYV, GCANCA and NIANCA. Predicted impacts are computed for each district; they are presented in Table 6C.7 as state-wide and national averages". 3 And see Appendix C, in which are displayed the temperature, rainfall and cloud cover effects by season, revealing interesting alternating patterns. 244 Table 6C.7: Climate effects on technology and related infrastructure (Seasonal Supra-Ricardian Model, 1970/71 through 1987/88) Temperature Rainfall HYV GCA NIA HV GCA NIA India .0108 .0006 .0005 .0012 .0057 .0001 1:Andhra Pradesh .0140 .0008 .0018 -.0005 -.0023 -.0006 2:Haryana .0064 -.0009 -.0073 .0011 .0007 .0002 3:Madhya Pradesh .0150 .0053 .0027 .0013 .0039 .0000 4:Maharashtra .0281 .0108 .0187 .0061 .0459 .0017 5:Karnataka .0154 .0059 .0104 .0025 .0159 .0004 6:Punjab .0135 .0055 -.0116 .0003 .0003 -.0001 7:Tamil Nadu .0063 -.0004 .0040 -.0000 -.0007 -.0006 8:Uttar Pradesh .0079 .0009 -.0053 .0001 .0003 -.0004 9:Bihar .0130 -.0019 -.0018 .0006 -.0010 -.0000 10:Gujarat .0079 -.0014 .0036 .0024 .0031 .0005 .14:Rajasthan -.0014 -.0072 -.0034 .0010 .0006 .0001 15:Orissa .0042 -.0128 -.0117 .0015 .0004 .0004 17:West Bengal .0045 -.0064 -.0029 -.0005 -.0001 -.0005 Note: Temperature Effect: Percentage change due to a 1° Celsius temperature increase. Rainfall Effect: Percentage change due to a 3% rainfall increase. Looking only at the national averages, the temperature effects on multiple-cropping and irrigation, anid the rainfall effects on all three are quite small: only one-half of a percent change in multiple-cropping in response to a 1% change in rainfall, one-tenth of a percent change in modem varieties in response to a 1% change in rainfall, and less than a tenth of a percent change in the others. But these averages hide considerable local variation, with a few states experiencing negative effects and a few experiencing effects as large as 4%. There is some consistency in the patterns of the effects across states: Maharashtra always exhibits the largest (positive) impact of rainfall on WHYV, GCANCA and NIANCA; Karnataka, Gujarat, Madhya Pradesh and Orissa nearly always also enjoy relatively large positive rainfall impacts. These states form a triangle across the Deccan in the middle of the country, wider at the West and tapering towards Orissa on the East. Andhra Pradesh, Tamil Nadu, and to some extent Bihar and West Bengal exhibit the largest negative impact of a change in rainfall; these states, along with Orissa, make up the Eastern edge of the country, with Andhra and Tamil at the southern end of that ribbon. The pattern of positive temperature impacts on multiple-cropping and on irrigation intensity is roughly similar, with Maharashtra and Karnataka exhibiting the largest positive effects, but with no similarities in the statewide patterns of negative temperature effects. The temperature effect on the adoption of modern varieties is somewhat stronger than the others. On average across the nation, a one percent increase in temperature would increase the adoption of modern varieties by slightly more than one percent, for an elasticity of approximately one. This result is more consistent than the others: only one state, Rajasthan, exhibits a negative 245 effect (one-seventh of one percent) while the other states' effects range from 0.42% to 2.81% (again in Maharashtra). From these results it is clear that climate change does not inhibit the process of technological change and infrastructural investment upon which India's Green Revolution has been built. ESTIMATES: NET REVENUE Table 6C.8 displays the results of the regression of net revenue on edaphic, climatic, and geographic variables, the predicted values of the technology and infrastructure variables from the two-stage system described above, interactions between climate and technology or between climate and infrastructure, and dummy variables for time and agroclimatic regions. This equation also fits the data very well: the adjusted R2 is nearly two-thirds, and the F- statistic is highly significant at 63.0875. Table 6C.9 presents results of F-tests of the null hypotheses that the set of edaphic variables, the set of temperature variables, the set of rainfall variables, the set of cloud cover variables, the set of all climate variables, the set of temperature range variables, the set of technology and related infrastructure variables, the set of technology interactions, and the set of year dummies do not influence net revenue. All of these null hypotheses are rejected: each set of variables does significantly influence net revenue; most of the variables in fact by themselves significantly influence net revenue. In addition to their role in the second-stage equations reported in a previous section, the edaphic variables are important determinants of net revenue: twelve of the nineteen soil type dummies have significant coefficients, and soil of neutral pH34 contributes more to net revenue than either acidic or base soil. The geographic variables are also important, with again six of seven agroeconomic region dummies significant. Districts on the seacoast, and especially their neighbors, or "second-tier" districts, have significantly higher net revenue than inland districts35, and districts lying at higher altitudes36 have lower net revenue. Each of the seasonal temperatures is interacted with latitude (a measure, among other things, of day length and of the angular incidence of solar radiation), but none of the three coefficients is significantly different from zero. 34 In terms of the data, of pH 7, whose variable (DMPH7) is the omitted dummy. 3 Due perhaps to a moderating influence of the ocean on annual weather, and perhaps to some degree on the fertility of delta regions which the soil type dummies do not pick up. 36 With concomitant changes in cropping patterns, inter alia. 246 Table 6C.8: Regression of the logarithm of net revenue (LNOFPKRE) (Seasonal Supra-Ricardian Model, 1970/71 through 1987/88) Multiple JR .81205 R Square .65942 Adjusted R Square .64897 Standard Error 10.97377 Analysis of Variance DF Sum of Squares Mean Square Regression 125 949652.91536 7597.22332 Residual 4073 490485.28979 120.42359 F = 63.08750 Signif F = .0000 -------- Variables in the Equation ------------------ Variable B SE B Beta T Sig T DMS03 .054552 .072459 .014624 .753 .4516 DMS04 .228057 .060825 .051111 3.749 .0002 DMS05 -.071670 .044797 -.035014 -1.600 .1097 DMS06 -.011457 .040655 -.004780 -.282 .7781 DMS07 -.007442 .046600 -.002248 -.160 .8731 DMS08 .088634 .078633 .025045 1.127 .2597 DMS09 -.128747 .069981 -.027979 -1.840 .0659 DMS10 -.807814 .119605 -.117262 -6.754 .0000 DMS11 .249808 .088832 .046002 2.812 .0049 DMS12 -.373038 .074994 -.126955 -4.974 .0000 DMS13 .112156 .108050 .016140 1.038 .2993 DMS14 .234469 .081655 .048710 2.871 .0041 DMS15 .211475 .055585 .060724 3.805 .0001 DMS16 -.242892 .039046 -.141758 -6.221 .0000 DMS17 -.264839 .083592 -.038763 -3.168 .0015 DMS18 .052901 .145101 .006147 .365 .7154 DMS19 -.090477 .051810 -.044261 -1.746 .0808 DMS20 -.202641 .086673 -.028930 -2.338 .0194 DMS21 -.488113 .121020 -.090713 -4.033 .0001 STORIE -.173900 .127633 -.019854 -1.362 .1731 DMPH5 -.096539 .046371 -.046213 -2.082 .0374 DMPH6 -.064780 .040572 -.031-286 -1.597 .1104 DMPH8 -.238242 .038463 -.134278 -6.194 .0000 DMPH9 -.226464 .045791 -.101747 -4.946 .0000 SOWXT1 .205087 .121377 .669816 1.690 .0912 SOWXT1SQ -.001860 .001116 -.393458 -1.666 .0957 SOWTR1 -.757177 .142564 -2.034670 -5.311 .0000 SOWXT1LT -9.54530E-04 7.2639E-04 -.204919 -1.314 .1889 SOWRN1 .001485 .003372 .260431 .440 .6598 SOWRN1SQ -7.21408E-07 4.6254E-07 -.169252 -1.560 .1189 247 Table 6C.8: Regression of the logarithm of net revenue (LNOFPKRE) (cont.) (Seasonal Supra-Ricardian Model, 1970/71 through 1987/88) Variable B SE B Beta T Sig T SOWXT1RN -1.11040E-04 5.9287E-05 -.494205 -1.873 .0612 MATXT1 -.222658 .076107 -2.218927 -2.926 .0035 MATXT1SQ -5.58911E-04 8.1490E-04 -.226554 -.686 .4928 MATTR1 -.122923 .165009 -.456976 -.745 .4563 MATXT1LT 7.42638E-05 .002245 .019218 .033 .9736 MATRN1 -.005293 .003738 -.799679 -1.416 .1568 MATRN1SQ 3.65133E-06 9.2339E-07 .337507 3.954 .0001 MATXT1RN 3.95982E-04 6.0763E-05 1.800304 6.517 .0000 HARXT1 .101693 .066253 .996446 1.535 .1249 HARXT1SQ -5.70657E-04 7.5346E-04 -.226169 -.757 .4489 HARTR1 .378414 .122537 1.764279 3.088 .0020 HARXT1LT 6.17475E-04 .002213 .175804 .279 .7802 HARRN1 -.002058 .005507 -.206151 -.374 .7086 HARRNlSQ -5.35116E-06 2.7653E-06 -.182831 -1.935 .0530 HARXT1RN -1.23765E-04 1.0975E-04 -.385230 -1.128 .2595 SOWOK1 -.191758 .166651 -.343683 -1.151 .2499 MATOK1 -.402711 .180774 -.867486 -2.228 .0260 HAROK1 .950843 .217260 2.006206 4.377 .0000 SOWXT1OK .004516 .004792 .263935 .942 .3460 MATXT1OK .010274 .005728 .764239 1.794 .0729 HARXT1OK -.022267 .006642 -1.544941 -3.352 .0008 SOWRN1DR .001415 4.0515E-04 .214677 3.493 .0005 MATRN1DR -.001314 5.7537E-04 -.142207 -2.284 .0224 HARRN1DR 7.64760E-04 7.3235E-04 .043570 1.044 .2964 JURNCV -.216387 .095226 -.060302 -2.272 .0231 JARNCV .403147 .082513 .091799 4.886 .0000 JUNERAIN -8.36467E-05 1.2053E-04 -.012126 -.694 .4877 JUAURAIN 1.44034E-04 7.4578E-05 .050938 1.931 .0535 YEARRAIN 2.70187E-05 5.4392E-05 .015666 .497 .6194 DMSEA .138790 .066972 .053254 2.072 .0383 DMSEANEI .202542 .041775 .087420 4.848 .0000 AGROBI -.215229 .084060 -.055054 -2.560 .0105 AGROB2 -.274846 .143101 -.084329 -1.921 .0548 AGROB3 -.053689 .091621 -.018752 -.586 .5579 AGROB4 -.221600 .078835 -.131924 -2.811 .0050 AGROB5 -.521666 .121442 -.148867 -4.296 .0000 AGROB6 -.650403 .092375 -.304840 -7.041 .0000 AGROB7 -.596806 .114039 -.112721 -5.233 .0000 LNCSTCLT .005070 .025429 .005481 .199 .8420 LNCSTBUL -.118732 .019439 -.150294 -6.108 .0000 RELWAGEK -.088414 .027528 -.061431 -3.212 .0013 GCNCPRDK -1.829905 2.051189 -.355564 -.892 .3724 EXT .001536 .004681 .010220 .328 .7428 248 Table 6C.8: Regression of the logarithm of net revenue (LNOFPKRE) (cont.) (Seasonal Supra-Ricardian Model, 1970/71 through 1987/88) / Variable B SE B Beta T Sig T WHYVPRDK 2.529281 1.830873 .664703 1.381 .1672 STRES5 .001459 2.6241E-04 .076885 5.558 .0000 NINCPRDK -2.929354 2.288324 -.793983 -1.280 .2006 SOWXTHYK -.058570 .064543 -.505594 -.907 .3642 SOWXTNIK .249874 .079263 2.223949 3.152 .0016 MATXTHYK .106064 .052784 .812056 2.009 .0446 MATXTNIK -.067679 .055675 -.524482 -1.216 .2242 HARXTHYK -.123782 .046304 -1.079055 -2.673 .0075 HARXTNIK -.144289 .054606 -1.283127 -2.642 .0083 SOWRNHYK 4.48243E-04 .001021 .020325 .439 .6607 SOWRNNIK .002693 .002153 .088081 1.251 .2111 MATRNHYK .003306 .001898 .171094 1.742 .0816 MATRNNIK 5.99302E-04 .002697 .031402 .222 .8241 HARRNHYK -.001082 .002229 -.044604 -.485 .6274 HARRNNIK .004914 .002701 .227481 1.819 .0690 SOWXTGCK -.198333 .075241 -1.467736 -2.636 .0084 SOWRNGCK .001556 .002039 .308685 .763 .4453 MATXTGCK .161058 .045832 2.191050 3.514 .0004 MATRNGCK -.007328 .003061 -1.421274 -2.394 .0167 HARRNGCK .004602 .004017 .616189 1.146 .2520 SOWTRHYK .103307 .095853 .346481 1.078 .2812 SOWTRGCK .780809 .131154 3.618342 5.953 .0000 SOWTRNIK -.430069 .115872 -1.536378 -3.712 .0002 MATTRHYK -.103027 .114465 -.290828 -.900 .3681 MATTRGCK .021311 .148659 .107180 .143 .8860 MATTRNIK .168471 .119246 .500618 1.413 .1578 HARTRHYK .069185 .063511 .238009 1.089 .2761 HARTRGCK -.352068 .116554 -2.195637 -3.021 .0025 HARTRNIK .300911 .080020 1.073437 3.760 .0002 GCANCA57 -.403485 .157103 -.068925 -2.568 .0103 IRR57 -.115579 .219248 -.024608 -.527 .5981 ALT -3.39930E-04 1.0025E-04 -.104415 -3.391 .0007 SOWXT1DR -.006635 .003318 -.131493 -2.000 .0456 MATXT1DR .013793 .011305 .206662 1.220 .2225 HARXT1DR -.009143 .011551 -.137737 -.792 .4287 DMYR71 -.024890 .045575 -.006876 -.546 .5850 DMYR72 .098980 .047431 .027051 2.087 .0370 DMYR73 .612971 .047082 .171189 13.019 .0000 DMYR74 .632374 .048542 .173171 13.027 .0000 DMYR75 .547989 .048000 .152391 11.417 .0000 DMYR76 .408188 .050087 .112479 8.150 .0000 249 Table 6C.8: Regression of the logarithm of net revenue (LNOFPKRE) (cont.) (Seasonal Supra-Ricardian Model, 1970/71 through 1987/88) Variable B SE B Beta T Sig T DMYR77 .579901 .051624 .161470 11.233 .0000 DMYR78 .453281 .052042 .126118 8.710 .0000 DMYR79 .239891 .054975 .065854 4.364 .0000 DMYR80 .642767 .056769 .177554 11.323 .0000 DMYR81 .636479 .061539 .176443 10.343 .0000 DMYR82 .645755 .066709 .178519 9.680 .0000 DMYR83 .846677 .071939 .238286 11.769 .0000 DMYR84 .777526 .075392 .211324 10.313 .0000 DMYR85 .738356 .077672 .199203 9.506 .0000 DMYR86 .693740 .080356 .186054 8.633 .0000 DMYR87 .754168 .082400 .194070 9.152 .0000 (Constant) 12.650747 2.843905 4.448 .0000 250 Table 6C.9: F-Tests of exclusion of groups of variables Excluded Variables Net Revenue Regression Edaphic (M = 24) F = 13.66114 Temperature (M = 38) F = 17.16124 Rainfall (M = 26) F = 7.83944 Oktas (Cloud Cover) (M = 6) F = 8.42512 Temperature Rain and Oktas (M = 70) F = 15.18449 Temperature Range (M = 12) F = 12.98978 Technology and Related Infrastructure(M = 33) F = 17.95629 Technology - Climate Interactions (M = 26) F = 8.40978 Year Dummies (M = 17) F 31.10041 Note: Entries in Table 6C.9 are the F-statistics on the null hypotheses that the excluded variables specified in the left column do not influence Net Revenue. Depending upon the number of variables excluded in each group (given as M = i in the left column), the degrees of freedom of these F-tests are anywhere from 6 to 70 for the numerator and more than 4000 for the denominator; thus the 5% critical value of the F statistic would range from about 1.5 to about 2.1; correspondingly, the 1% critical value of the F statistic would range from about 1.8 to about 2.8. The coefficients on the predicted values of modem varieties (WHYVPRDK), multiple- cropping (GCNCPRDK) and irrigation intensity (NINCPRDK) are not significantly different from zero; but those variables are interacted with climate terms, as discussed below, and nearly half of the interactions' coefficients are significant. Research and extension contribute to net revenue (the coefficient of extension on net revenue is not significant, though positive, but it significantly increased the adoption of modem varieties; the coefficient of research on WHYV was insignificant, but it appears significantly positive in the net revenue regression). Higher bullock costs reduce net revenue, as does an increase in off-farm wages relative to agricultural wages, which probably denotes a decline in the quality of available farm workers as off-farm opportunities attract more and more of the best and most-highly-skilled laborers. Sixteen of the seventeen year dummies are positive and significant; the dummy for 1970, the first year in the sample, is omitted, so these dummies are picking up omitted time trends and price index effects. Current weather, and its timing, also obviously influences current net revenue: given a normal seasonal rainfall, higher rainfall in July and August (the variable JUAURAIN) will increase net revenue. 251 The climate variables represent long-term averages or norms, to which farmers respond in their decisions about cropping patterns, input use, investment in technology and infrastructure, and so forth. This model displays quite rich (normal) temperature and rainfall effects on net revenue. Looking first at the twentyfour purely climate terrns, we find that the coefficients of thirteen are significantly different from zero. The squared and "raw" terms are usually of the opposite sign. The impact of maximum temperature, temperature range, and rainfall differ by season: four of the eight categories differ in sign between the sowing and the maturation seasons; four of eight differ in sign between the sowing and harvesting seasons; and six of the eight differ between the maturation and the harvesting seasons. In particular, higher maximum temperature in the sowing season increases net revenue at a decreasing rate: perhaps seed germination and night-time temperatures are important here. Given the value of maximum sowing season temperature, a larger temperature range (which is to say, a lower minimum temperature) decreases net revenue. This surely also partakes of issues of germination and fragility of young plants. The coefficient of the interaction of sowing season maximum temperature and rainfall is significantly negative. In the maturation season, higher maximum temperature would decrease net revenue, likely reflecting temperature stress. The coefficient of rainfall during this season is not significant (though negative), but the coefficient on rainfall squared is positive and significant, and rainfall during this season ameliorates the temperature damage, as evidenced by a positive and significant coefficient on the interaction between maximum temperature and rainfall (in the variable MATXTIRN). Increased cloud cover in the maturation season (given the level of rainfall) reduces net revenue, as would be expected, depriving the growing plants of sunlight to support photosynthesis; yet it also ameliorates the temperature damage (as seen in the positive and significant coefficient on the variable MATXTIOK). The okta pattern is reversed in the harvest season: additional cloud cover increases net revenue, but worsens the temperature impact. Higher maximum temperature (though barely below significance threshold) increases net revenue, no doubt by extending somewhat the growing season; interestingly, a greater temperature range in the harvest season also increases net revenue. Interestingly a higher coefficient of variation of June rainfall will decrease net revenue while a higher coefficient of variation of July and August rainfall contributes to net revenue. A key focus of this study is the interaction of climate with technology, infrastructure, and geographic variables", beyond the so-called "purely climate" variables. Eleven such interactions each season are included. About a third of the nation's districts have been identified as "drought-prone", indicating not only low levels of normal rainfall but also poor irrigation potential, probably high rates of evapotranspiration, and so forth. For each season, normal temperature and rainfall have been interacted with a dummy whose value is one for each such drought-prone district. As one would expect, higher normal rainfall in the sowing season will increase net revenue in the drought- prone districts, allowing a crop to become established in such adverse situations (although higher 252 rainfall in the maturation season is harmful), while higher temperatures in the sowing season, inhibiting crop establishment, would reduce net revenue in those districts. The interactions of maximum temperature, temperature range and rainfall with the predicted values of modem varieties, multiple-cropping and irrigation intensity are complex, yielding significant coefficients in some seasons. HYVs interact positively (and significantly) with maturation season temperature and rainfall but negatively with harvest temperature. Multiple-cropping interacts positively (and significantly) with maturation season maximum temperature and sowing season temperature range, and negatively with sowing season maximum temperature and maturation season rainfall. And irrigation interacts positively with sowing season maximum temperature and harvest season rainfall, and negatively with harvest season maximum temperature. CLIMATE CHANGE EFFECTS ON NET REVENUE Based on the regression results reported in the previous section, the actual district-level values of the climate and technology variables, and the computed climate change effects on the technology variables, one can similarly compute the predicted effects on net revenue of changes in normal temperature, normal rainfall and oktas of cloud cover. Those effects (and their sum) are presented in Table 6C. 10. Temperature and rainfall affect net revenue via three avenues. The first could be called direct or express, operating via the temperature and rainfall terms, their squares, and the temperature-rainfall interaction term in the net revenue equation. The second avenue could be called local, beginning with the temperature and rainfall effects on WHYV, GCANCA and NIANCA and operating through their terms in the net revenue equation. The third avenue, meandering, operates through the terms in the net revenue equation which capture the interactions between climate and the predicted values of WHYV, GCANCA and NIANCA. Thus one cannot simply glance at the direct climate terms in the net revenue equation to perceive the temperature and rainfall effects; one must compute carefully the effects through all three avenues, and use in the appropriate places in the computations the values of the climate terms and the technology variables. The predicted temperature effects on net revenue are a bit high compared to previous estimates for U.S. and Brazilian agriculture: on average over all the districts in this study, a one degree Celsius increase in normal temperature would reduce net revenue by approximately four percent. The differences in estimated effects might be due to the fact that these estimates are based on a net revenue framework, while other estimates are based on reasonably well-defined land prices; other differences include the time period involved, the use of a panel of cross- sections here, the use of seasonal climate variables, the inclusion of additional climate terms (in particular, temperature range and oktas of cloud cover), the inclusion of technology and infrastructure variables, and of course the inherent differences between the Indian and the U.S. or Brazilian agricultural environments. There also seems to be more variation within India, as the sub-continent contains widely varying agricultural environments. There iCVII-l displays a clear geographic pattern to the temperature impacts on net revenue: the average State-wise change in net revenue ranges from an increase of 1.43% in the Punjab to a decrease of 11.7% in Andhra Pradesh. 253 Table 6C.10: Climate effects on net revenue (Seasonal Supra-Ricardian Model, 1970/71 through 1987/88) Maximum Rainfall Sum of Oktas Climate Temperature Temperature Sum and Rainfall India -.0397 .0666 .0269 .0283 .0552 1:Andhra Pradesh -.1170 .0041 -.1129 .1305 .0176 2:Haryana .0116 .0252 .0378 -.0119 .0249 3:Madhya Pradesh -.0390 -.0203 -.0593 .0111 -.0482 4:Maharashtra -.0366 .3664 .3298 -.0189 .3109 5:Karnataka -.0492 .1759 .1267 .0508 .1775 6:Punjab .0143 .0075 .0218 -.0241 -.0023 7:Tamil Nadu -.0802 .0485 -.0317 .0894 .0577 8:Uttar Pradesh -.0168 .0334 .0166 .0047 .0213 9:Bihar -.0735 .0405 -.0330 .0706 .0376 10:Gujarat -.0275 .0349 .0074 -.0282 -.0208 14:Rajasthan -.0068 .0169 .0101 -.0077 .0024 15:Orissa -.0327 .1000 .0673 .0946 .1619 17:West Bengal -.0737 .1108 .0371 .0883 .1254 Note: Temperature Effect: Percentage change due to a 1 Celsius temperature increase. Rainfall Effect: Percentage change due to a 3% rainfall increase. Okta Effect: Percentage change due to a 3% increase in cloud cover. The States whose net revenue is most adversely affected by an increase in temperature are, in order, Andhra Pradesh, Tamil Nadu, Bihar and West Bengal, all suffering a decline in net revenue in excess of seven percent, while the States least adversely affected are, in order, the Punjab, Haryana, Rajasthan and Uttar Pradesh, the first two benefitting and the latter two harmed by less than two percent. The predicted rainfall effects on net revenue are moderate, slightly larger than, and of the opposite sign to, the predicted temperature effects. There is considerable variation among the States, but little clear geographic pattern-. In all States except Madhya Pradesh higher rainfall is predicted to increase net revenue. and in nine of the thirteen States the combined temperature and rainfall effects would increase net revenue. The predicted effect on net revenue of an increase in cloud cover is, surprisingly, to increase net revenue on a National average and for eight of the States; the exceptions are five States in the north-central and western parts of the nation-. Thus the combined temperature, rainfall and okta effects yield an increase in net revenue in ten of the thirteen states. 254 Most global climate change scenarios refer simply to increases in normal temperature; some refer to changes in normal average temperature. As discussed earlier, this study captures a richer specification, including both normal maximum temperature and the temperature range (the normal maximum temperature minus the normal minimum temperature). The literature offers essentially no guidance on the possible change in temperature range in response to global climate change. If maximum and minimum temperatures were to change by the same amount, there would be no change in the temperature range. That may be the implicit assumption underlying nearly every global climate model's silence on this issue; it was the conscious, explicit justification for our excluding changes in the temperature range from the computations upon which Table C-VII-1 was based, and the associated discussion of climate change effects on net revenue. An alternative possibility is that normal minimum and maximurm temperatures change not by the same amount, but by the same proportion. Under that scenario, one could calculate that one degree Celsius is 3.133% of the nationwide average maximum temperature in the sowing season, 3.853% in the maturation season, and 3.55% in the harvest season. One could next calculate the corresponding proportionally-equal changes in minimum temperature (0.6790 in the sowing season, 0.669° in maturation, and 0.6130 in the harvest season) and then the increases in temperature range (.321°, .33 1 and .387°). Of course, the Kelvin scale, which has a true ("absolute") zero, may be appropriate for this computation, in which case a one degree Celsius (or Kelvin} increase in maximum temperature is a 0.33% change, implying changes in the temperature range which are only about one tenth as large as in the previous paragraph: 0.0332° in the sowing season, 0.0284° in the maturation season, and 0.0358° in the harvest season. Nevertheless, it is of interest to calculate temperature range effects in order to learn what would happen were the range in fact to change. Given that both the normal maximum temperature and the normal temperature range are included in the regression equation, an increase in the temperature range implies either that the minimum temperature decreases, holding constant the maximum temperature, or that the minimum temperature increases by less than does the maximum temperature. From another point of view, a larger range implies that, for a given maximum temperature, the minimum temperature is lower than it otherwise would have been. The agronomic basis for a temperature range effect then is the impact of the minimum temperature on germination or growth of crops. A larger temperature range would be beneficial, for example, if, given hot daytime temperatures, cooler nights and mornings fostered more complete germination, or reduced the temperature stress on new seedlings or on maturing plants. Alternatively, a larger temperature range might be harmful if the cooler nights inhibited growth or the formation of grains, and so forth. Table 6C. 11 displays the effects on net revenue of a one degree Celsius increase in the temperature range -- as discussed above, actual changes in the range, if there be any, are likely to be smaller, but these computed effects can be scaled downward accordingly. Appendix D reports the temperature range effects by season. It is apparent from Table 6C. 11 that increases in the temperature range are beneficial: a one degree increase in the nationwide average temperature range would increase net revenue by 255 more than eight percent; an ipcrXease of 0.30 (based on the "proportional Censius" discussion above) would increase net revenue by about 2.5%, while a 0.030 increase (based on the "proportional Kelvin" discussion above) would increase net revenue by about one quarter of one percent. There is a strong geographic pattern to these effects: the largest increases in net revenue would occur (in order) in the Punjab, Haryana, Uttar Pradesh, West Bengal, Orissa and Bihar: these states are the northern-most and eastern-most in our data set. Table 6C.11: Temperature range effects on net revenue Temperature Range India .0862 1 :Andhra Pradesh .0484 2:Haryana .2307 3:Madhya Pradesh .0308 4:Maharashtra .0121 5 :Karnataka .0204 6:Punj ab .2818 7:Tamil Nadu .0712 8:Uttar Pradesh .1553 9:Bihar .1195 1 0:GuJarat .0167 14:Rajasthan .0562 15 :Orissa .1335 17:West Bengal .1537 TECHNOLOGY AND INFRASTRUCTURE EFFECTS As discussed above, our specification in equation (7b) includes important technology and related infrastructure variables which are central to India's Green Revolution experience. Section four presented two stage least squares estimates of NIANCA, GCANCA and WHYV, in which each of the three variables was determined inter alia by one or both of the other two. And Section six presented estimates of Net Revenue, in which all three of the technology and related infrastructure variables, as well as their interactions with climate variables, appeared on the right- hand side. 256 Table 6C.12: Technology and infrastructure effects (Supra-Ricardian Model, 1970/71 through 198a/88) Effect of: HYV GCA NIA on: GCA NIA NR HW NIA NR HYV GCA NR India .0626 .6559 .1891 3.3800 3.0955 .7798 1.0273 .0634 .0652 I:Andhra Pradesh .0802 .4334 .1116 1.9595 .5130 .3232 1.0189 .0767 -.0919 2:Haryana .0931 .3458 .1008 1.2416 .3794 -.1017 .7896 .0966 -.0203 3:Madhya Pradesh .0471 .8367 .1898 3.9996 2.2394 1.0307 .5807 .0302 .1591 4:Maharashtra .0677 1.6865 .2510 2.4852 1.9166 1.3255 .5504 .0250 .2932 5:Karnataka .0760 .8173 .2328 2.1846 1.0870 1.2733 .6798 .0449 .0619 6:Punjab .1207 .2826 .0956 .8122 .2122 .2653 .9114 .1454 -.1363 7:Tamil Nadu .1013 .3199 .1674 1.0034 .9328 .7778 .8365 .1060 .0148 8:UttarPradesh .0562 .1884 .1748 2.2212 .3133 .7804 1.7204 .1048 -.0209 9:Bihar .0607 .4796 .1890 3.0893 1.0371 .7359 .8448 .0612 -.0304 10:Gujarat .0740 .6741 .1922 2.0000 .7678 .5304 .7007 .0469 .3091 14:Rajasthan .0394 .4901 .1957 12.2518 21.7892 -.0099 1.7183 .0468 .1408 15:Orissa .0384 .4556 .2412 3.4718 1.0431 1.0394 1.2607 .0429 -.0157 17:West Bengal .0499 .7508 .2393 2.0837 1.3646 1.3952 .9335 .0623 -.1046 Note: Effects expressed in elasticity form. Table 6C.12 presents the computed effects of an increase in any of the three technology and related infrastructure variables on the other two and on Net Revenue. The computations reveal the total effects, involving direct terms, indirect effects through a third variable, and (for effects on Net Revenue) interaction terms. The broad message of Table 6C. 12 is that each of the technology and related infrastructure variables encourages and reinforces the others, and contributes to Net Revenue. The HYV effects are consistently positive, yet inelastic (except for the effect of modem varieties on irrigation intensity in Maharasthra, which is positive and elastic). A one percent increase in the proportion of crops planted to modem varieties would induce a minute increase in multiple-cropping: six one-hundredths of one percent nationwide, nearly one-tenth of one percent or more in Haryana, the Punjab and Tamil Nadu. The effects of increased modem variety adoption are somewhat larger (and considerably more consistent) on Net Revenue, and larger still on irrigation intensity. The effects on irrigation are not difficult to understand, given the responsiveness of nearly all modem varieites to an assured supply of water: as farmers use more and more modem varieties, the payoff to irrigation increases, inducing more investment (privately and by both state and Central governments) in irrigation capacity and facilities. And the HYV effects on Net Revenue, somewhat smaller, are also easier to understand: to have been selected and released, the modem varieties will already have been shown to offer yield increases in excess of their additional input requirements. The multiple-cropping effects are substantially larger (except for two relatively small and negative effects of multiple-cropping on Net Revenue, in the two adjacent states of (surprisingly) Haryana and Rajasthan): on average for the entire nation, a one percent increase in multiple- cropping would induce an increase in the use of modem varieties and in irrigation intensity by 257 more than three percent, and an increase in Net Revenue of more than three-quarters of one percent. The computed effects of an increase in multiple-cropping in Rajasthan are anomolous and improbable: an increase in WHYV of more than twelve percent, and an increase in irrigation intensity of more than twenty percent! And Rajasthan is one of the two states in which the predicted effect of an increase in multiple-cropping on Net Revenue is negative. An increase in irrigation intensity would increase the adoption of modem varieties, with an elasticity slightly exceeding one, for essentially the reason discussed above: with more assured water availability, the payoff to the adoption of HYVs is much higher. The effect of an increase in NIANCA on multiple-cropping and on Net Revenue is much smaller. SECONDARY IMPACTS ON CLIMATE AND TECHNOLOGY EFFECTS We have seen from Table 6C.1 1 and the associated discussion, and from Table 6C.12 and its associated discussion, the predicted effects of changes in climate variables on the three technology and related infrastructure variables, and on Net Revenue. One of the strengths of our approach, integrating climate and edaphic variables with technological and infrastructural variables, is that we can also compute the predicted impact of changes in technology and related infrastructural variables on the already-reported climate or other effects. In simple terns, these secondary impacts on, say, the temperature effects simply measure the extent to which changes in technology and infrastructure -- over which policy-makers exercise some influence -- might modify or ameliorate the effect of temperature changes on Indian agriculture. Table 6C. 13 presents the impacts of increases in the technology and related infrastructure variables on the effect of higher temperature on Net Revenue; Table 6C. 14 presents the impacts of increases in the technology and related infrastructure variables on the effect of higher rainfall on Net Revenue. Table 6C.13: Secondary impacts on the temperature effects on net revenue (Seasonal Supra-Ricardian Model, 1970/71 through 1987/88) | Research Extension WHYV GCANCA NIANCA India -.0036 -.0618 -.1026 -.2963 -.1628 1: Andhra Pradesh -.0021 -.0317 -.0841 -.1839 .0012 2: Haryana -.0017 -.0252 -.0852 -.1282 .0232 3: Madhya Pradesh -.0031 -.1194 -.1836 -.3427 -.9598 4: Maharashtra -.0039 -.0465 -.1048 -.2298 -.1818 5: Karnataka -.0029 -.0425 -.0948 -.2078 -.1104 6: Punjab -.0009 -.0251 -.0828 -.0934 .0492 7: Tamil Nadu -.0045 -.0374 -.0858 -.1098 .0276 8: Uttar Pradesh -.0013 -.0461 -.0846 -.2046 .0655 9: Bihar -.0019 -0184 -.0828 -.2783 .0346 10: Gujarat -.0063 -.0598 -.0823 -.1887 -.0317 14: Rajasthan -.0098 -.1526 -.0832 -.9699 .0323 15: Orissa -.0087 -.0419 -.0855 -.3002 -.0571 17: West Bengal -.0012 -.0108 -.0857 -.1938 -.0114 258 Table 6C.14: Secondary impacts on the rainfall effects on net revenue (Seasonal Supra-Ricardian Model, 1970/71 through 1987/88) Research Extension WHYV GCANCA NIANCA India .0001 .0012 .0023 .0085 .0092 1: Andhra Pradesh .0001 .0010 .0028 .0047 .0103 2: Haryana .0001 .0009 .0030 .0030 .0099 3: Madhya Pradesh .0000 -.0011 -.0005 .0098 .0023 4: Maharashtra .0001 .0011 .0021 .0059 .0080 5: Karnataka .0001 .0011 .0024 .0052 .0088 6: Punjab .0000 .0010 .0035 .0023 .0104 7: Tamil Nadu .0002 .0013 .0030 .0024 .0101 8: Uttar Pradesh .0000 .0016 .0030 .0056 .0124 9: Bihar .0001 .0005 .0029 .0780 .0101 10: Gujarat .0002 .0020 .0028 .0046 .0093 14: Rajasthan .0003 .0054 .0027 .0320 .0122 15: Orissa .0002 .0012 .0027 .0085 .0105 17: West Bengal .0000 .0003 .0028 .0049 .0100 In broad strokes, an increase in the technology and related infrastructure variables tends to worsen the (already negative) temperature effects on net revenue somewhat, and tends to improve the (already positive) rainfall effects on net revenue ever so slightly. In both cases, the research and extension impacts are smaller than the modem variety, multiple-cropping and irrigation impacts. 259 APPENDIX D: SEASONAL CLIMATE SPECIFICATION SOWING AND HARVESTING SEASONS OF PRINCIPAL CROPS The creation of seasonal climate variables, based on monthly values of the normal climate variables and on the sowing and harvesting seasons of principal crops, was described in Section III.B. of the text. This Appendix presents the cropping seasons for the five major food crops in India. Table 6D. 1 displays the sowing seasons of the five crops; Table 6D.2 displays the harvesting seasons. Table 6D.1: Sowing Seasons of Principal Crops Rice Jowar Bajra Maize Wheat (Autumn) (Kharit) 1: Andhra Pradesh March-May June-Oct June-Aug June-Aug --- 2: Haryana May-Aug June-July June-Aug June-July Oct-Dec 3: Madhya Pradesh I OJun- I 5Aug 1 OJun-7Aug 15Jun-8Aug June-July Oct-Nov 4: Maharashtra June-July June-July June-July June-July Oct-Nov 5: Karnataka May-Aug May-July June-Aug April-July Sept-Nov 6: Punjab May-Aug June-July June-Aug June-July Oct-Dec 7: Tamil Nadu May-Nov June-Oct May-Dec --- --- 8: Uttar Pradesh May-July June-July June-Aug June-July Oct-Dec 9: Bihar May-July April-Aug April-Aug June-July Oct-Dec 10: Gujarat June-Aug June-July June-July June-July Oct-Nov 14: Raj asthan July-Aug June-Aug June-July June-July Oct-Dec 15: Orissa May-June June-July --- June-July Oct-Nov 17: West Bengal March-June ________April-June Oct-Dec -- No appreciable amount of that crop is grown in the given State. 260 Table 6D.2: Harvesting Seasons of Principal Crops Rice Jowar Bajra Maize Wheat (Autumn) (Kharif) 1: Andhra Pradesh June-Sept Jan-April Sept-Dec Sept-Oct --- 2: Haryana Sept-Nov Sept-Nov Sept-Nov Sept-Nov April-May 3: Madhya Pradesh 15Sept- 1 5Dec Nov- I Jan 1 OOct-Dec 1 5Aug-Sept I 5Feb-April 4: Maharashtra Oct-Dec Nov-Jan Oct-Nov Aug-Nov Feb-March 5: Kamataka Sept-Dec Oct-Dec Sept-Jan July-Nov Jan-March 6: Punjab Sept-Nov Sept-Nov Sept-Nov Sept-Nov April-May 7: Tamil Nadu Sept-Feb Sept-Jan Sept-March ___ 8: Uttar Pradesh Sept-Nov Oct-Dec Sept-Nov Aug-Oct March-May 9: Bihar Sept-Oct Sept-Dec Sept-Dec Sept-Oct Feb-May 10: Gujarat Oct-Dec Nov-Dec Sept-Nov Sept-Nov Feb-March 14: Rajasthan Oct-Dec Oct-Dec Sept-Nov Sept-Nov March-May 15: Orissa Sept-Oct Sept-Oct ___ Sept-Oct March-April 17: West Bengal July-Nov ___ __ July-Aug March-April -- No appreciable amount of that crop is grown in the given state. Note: Cropping calendars, from which these Tables are constructed, are available in many places. One convenient source is a set of tables printed in most of the recent issues of Area and Production of Principal Crops in India, published by the Directorate of Economics and Statistics, Department of Agriculture and Co-operation, Ministry of Agriculture of the Government of India. COMPARISON OF SEASONAL VS. MONTHLY CLIMATE SPECIFICATIONS As discussed in the text, we argue that the seasonal specification is preferable to a monthly specification on theoretical grounds: * the choice of any four months is arbitrary to begin with; * given the wide differences across India in sowing and harvesting seasons, no single choice of four months would adequately capture the important farming activities and biological processes occurmng over the year in all states or districts; and * the monthly specification simply fails to exploit the important information which is contained in cropping calendars. We believe that these arguments are compelling on their own. Yet the case for a seasonal specification can be bolstered by its comparison with the three possible specifications which involve four equally-spaced months. This section presents a comparison of seasonal vs. four- monthly semi-Ricardian specifications (that is, models which contain climatic, edaphic, and geographic variables). The comparisons are based on the estimated regressions from each specification, and the climate effects computed from those regression results. 261 COMPARISON OF SEASONAL VS. MONTHLY SEMI-RICARDIAN MODELS Table 6D.3 presents results of the four semi-Ricardian regressions on the logarithm of net revenue per hectare: using seasonal variables (the first two columns of Table 6D.3), using the months of January, April, July and October (using the third and fourth columns of Table 6D.3), using the months of February, May, August and November (using the fifth and sixth columns of Table 6D.3); and using the months of March, June, September and December (using the seventh and eighth columns of Table 6D.3). Finally, Table 6D.4 displays the estimated temperature and rainfall effects on net revenue computed from each of the four specifications. The major conclusions to be drawn from Table 6D.4 are, first, that the temperature and rainfall effects in the semi-Ricardian formulation are worse (that is, are more damaging to net revenue) than the temperature and rainfall effects presented in the text for the supra-Ricardian specification: on average across the country, the semi-Ricardian temperature effect is to reduce net revenue by nearly 14%, rather than reduce it by less than 4%; the rainfall effect is to reduce net revenue by almost 2% rather than increase it. This underscores the potential for bias in computed climate effects if technology and related infrastructure are excluded from the model for India in the post-Green Revolution period. 262 Table 6D.3: Semi-ricardian regressions of net revenue (comparison of seasonal vs. monthly climate variables) Seasonal JanAprJulOct FebMayAugNov MarJunSepDec DMS03 0.047391 DMS03 -0.064591 DMS03 0.138235 DMS03 -0.095791 DMS04 0.156287 DMS04 0.194134 DMS04 0.428703 DMS04 0.332627 DMS05 -0.057968 DMS05 -0.029151 DMS05 -0.051054 DMS05 -0.062284 DMS06 0.028143 DMS06 -0.094722 DMS06 0.122821 DMS06 0.05262 DMS07 0.006123 DMS07 -0.057414 DMS07 0.054629 DMS07 0.226469 DMS08 0.305105 DMS08 0.266848 DMS08 0.001536 DMS08 0.536485 DMS09 -0.125437 DMS09 0.279877 DMS09 0.101456 DMS09 0.516419 DMS10 -0.687224 DMS10 -0.726315 DMS10 -0.544645 DMS10 -0.636275 DMS11 0.418193 DMS11 0.136947 DMS11 0.449822 DMS11 -0.143429 DMS12 -0.541489 DMS12 0.253248 DMS12 0.116274 DMS12 0.206286 DMS13 0.195525 DMS13 -0.67164 DMS13 -0.02402 DMS13 -0.302366 DMS14 0.165627 DMS14 2.07E-04 DMS14 -0.051572 DMS14 0.300579 DMS15 0.225131 DMS15 0.303233 DMS15 0.390857 DMS15 0.546584 DMS16 -0.029486 DMS16 0.061276 DMS16 0.125834 DMS16 0.108507 DMS17 -0.343428 DMS17 -0.25561 DMS17 -0.017997 DMS17 -0.169563 DMS18 0.361077 DMS18 0.38236 DMS18 0.413322 DMS18 0.602686 DMS19 0.062551 DMS19 -0.02162 DMS19 -0.034553 DMS19 0.048573 DMS20 -0.143958 DMS20 -0.422014 DMS20 -0.08253 DMS20 0.110646 DMS21 0.022633 DMS21 0.199175 DMS21 0.010773 DMS21 0.018309 STORIE -0.030066 STORIE -0.261257 STORIE -0.10075 STORIE -0.055537 DMPH5 -0.119451 DMPH5 0.06547 DMPH5 -0.123651 DMPH5 -0.154866 DMPH6 -0.12982 DMPH6 -0.058365 DMPH6 -0.161722 DMPH6 -0.199935 DMPH8 -0.300347 DMPH8 -0.245566 DMPH8 -0.315937 DMPH8 -0.255513 DMPH9 -0.20063 DMPH9 -0.356711 DMPH9 -0.350658 DMPH9 -0.333695 SOWXT1 -0.216174 JANXT 0.994395 FEBXT 0.432457 MARXT 0.515218 SOWXT1SQ 0.001405 JANXTSQ -0.011564 FEBXTSQ -0.004406 MARXTSQ -0.004235 SOWXT1 LT -7.84E-04 JANXTLT -0.01723 FEBXTLT -0.016651 MARXTLT -0.010791 SOWRN 1 -5.11 E-04 JANRN 0.08124 FEBRN -0.036341 MARRN -0.038739 SOWRN1SQ -3.47E-08 JANRNSQ -1.98E-05 FEBRNSQ -3.39E-04 MARRNSQ -2.58E-05 SOWXT1 RN -7.62E-06 JANXTRN -0.004164 FEBXTRN 0.001541 MARXTRN 0.001645 MATXT1 0.079894 APRXT -0.049427 MAYXT 0.095467 JUNXT 0.086496 MATXT1SQ -6.76E-04 APRXTSQ -0.006813 MAYXTSQ -0.002573 JUNXTSQ -0.002972 MATXT1 LT -0.003441 APRXTLT 0.018733 MAYXTLT 0.002695 JUNXTLT -0.001062 MATRN1 -0.00271 APRRN 0.174133 MAYRN 0.002219 JUNRN 0.010013 MATRN1SQ -1.30E-06 APRRNSQ -2.45E-04 MAYRNSQ -8.64E-06 JUNRNSQ 2.1 OE-06 MATXT1 RN 1.46E-04 APRXTRN -0.003951 MAYXTRN 4.05E-05 JUNXTRN -4.05E-04 HARXT1 0.041876 JULXT -0.055961 AUGXT -0.654834 SEPXT 0.68286 HARXT1SQ -0.001382 JULXTSQ 6.43E-04 AUGXTSQ 1.26E-04 SEPXTSQ -0.013985 HARXT1 LT 0.002816 JULXTLT -0.009168 AUGXTLT 0.009234 SEPXTLT 0.006774 HARRN1 -0.004965 JULRN -0.001634 AUGRN -0.008168 SEPRN -0.006182 HARRN1SQ 2.88E-06 JULRNSQ 1.08E-07 AUGRNSQ 1.40E-06 SEPRNSQ -2.43E-08 HARXT1RN 9.18E-05 JULXTRN 5.47E-05 AUGXTRN 2.47E-04 SEPXTRN 2.39E-04 263 Table 6D.3: Semi-ricardian regressions of net revenue (cont.) (comparison of seasonal vs. monthly climate variables) Seasonal JanAprJulOct FebMayAugNov MarJunSepDec OCTXT 0.245338 NOVXT -0.866645 DECXT -0.818222 OCTXTSQ -0.001604 NOVXTSQ 0.017906 DECXTSQ 0.013041 NOVXTLT 0.003153 DECXTLT 0.005284 OCTRN -0.059728 NOVRN 0.008971 DECRN 0.032982 OCTRNSQ 6.42E-06 NOVRNSQ 4.65E-06 DECRNSQ 4.26E-05 OCTXTRN 0.001801 NOVXTRN -2.91 E-04 DECXTRN -0.001635 SOWOK1 0.063237 JANOK 1.568514 FEBOK 0.645663 MAROK 0.650107 MATOK1 -0.613005 APROK -1.604114 MAYOK -0.393134 JUNOK -1.0706 HAROK1 0.849798 JULOK -0.457994 AUGOK -1.281817 SEPOK -0.059318 OCTOK -0.214595 NOVOK -0.735821 DECOK -0.105595 SOWXT1 OK -0.002229 JANXTOK -0.050723 FEBXTOK -0.019029 MARXTOK -0.024373 MATXT1OK 0.016289 APRXTOK 0.043819 MAYXTOK 0.01277 JUNXTOK 0.032293 HARXT1 OK -0.02389 JULXTOK 0.009781 AUGXTOK 0.035992 SEPXTOK -0.006199 OCTXTOK 0.006886 NOVXTOK 0.022761 DECXTOK 0.013919 SOWRN1 DR 0.001216 JANRNDR -0.012945 FEBRNDR 0.013665 MARRNDR 0.031834 MATRN1DR -0.001416 APRRNDR -0.012208 MAYRNDR -0.008901 JUNRNDR 0.001414 HARRN1DR 0.002668 JULRNDR 9.72E-04 AUGRNDR 0.001317 SEPRNDR -0.001291 OCTRNDR 0.0019 NOVRNDR -0.005993 DECRNDR -0.006612 JURNCV -0.302145 JURNCV 0.09904 JURNCV 0.464227 JURNCV 0.176198 JARNCV 0.372398 JARNCV 0.513689 JARNCV 0.128412 JARNCV 0.511188 JUNERAIN -1.28E-05 JUNERAIN -4.47E-05 JUNERAIN 8.09E-05 JUNERAIN 6.28E-05 JUAURAIN 2.18E-04 JUAURAIN 2.18E-04 JUAURAIN 3.39E-04 JUAURAIN 3.OOE-04 YEARRAIN 1.39E-06 YEARRAIN 5.1 OE-05 YEARRAIN 1.25E-05 YEARRAIN -4.57E-06 DMSEA 0.121358 DMSEA -0.158873 DMSEA 0.01148 DMSEA -0.363871 DMSEANEI 0.27414 DMSEANEI 0.129393 DMSEANEI 0.258619 DMSEANEI 0.05273 AGROB1 -0.237214 AGROB1 -0.806599 AGROB1 -0.626631 AGROB1 -0.456057 AGROB2 -0.259659 AGROB2 -0.811866 AGROB2 -0.829606 AGROB2 -0.905527 AGROB3 -0.016507 AGROB3 0.240472 AGROB3 0.417542 AGROB3 0.12071 AGROB4 -0.222141 AGROB4 -0.173391 AGROB4 -0.127879 AGROB4 -0.336791 AGROB5 -0.585149 AGROB5 -0.234677 AGROB5 0.16182 AGROB5 -0.849395 AGROB6 -0.618018 AGROB6 -0.617847 AGROB6 -0.500026 AGROB6 -1.015301 AGROB7 -0.480066 AGROB7 -0.837377 AGROB7 -1.323785 AGROB7 -1.216763 LNCSTCLT 0.082129 LNCSTCLT 0.05652 LNCSTCLT 0.055137 LNCSTCLT 0.04223 LNCSTBUL -0.116711 LNCSTBUL 0.010581 LNCSTBUL 0.096032 LNCSTBUL 0.043916 RELWAGEK -0.065124 RELWAGEK -0.251109 RELWAGEK -0.250618 RELWAGEK -0.244593 SOWXT1 DR -0.001309 JANXTDR 0.063108 FEBXTDR 0.158244 MARXTDR -0.046007 MATXT1DR 0.016493 APRXTDR 0.025242 MAYXTDR -0.101918 JUNXTDR 0.16101 HARXT1DR -0.023725 JULXTDR 0.134036 AUGXTDR 0.153761 SEPXTDR -0.21705 OCTXTDR -0.219077 NOVXTDR -0.189673 DECXTDR 0.078489 264 Table 6D.3 Semi-Ricardian Regressions of Net Revenue (cont.) (comparison of seasonal vs. monthly climate variables) Seasonal JanAprJulOct FebMayAugNov MarJunSepDec ALT -6.OOE-04 ALT -3.74E-04 ALT -3.52E-04 ALT -2.80E-04 DMYR71 -0.024095 DMYR71 0.026187 DMYR71 0.026542 DMYR71 0.026913 DMYR72 0.113087 DMYR72 0.11735 DMYR72 0.114339 DMYR72 0.112715 DMYR73 0.620249 DMYR73 0.620794 DMYR73 0.576533 DMYR73 0.614414 DMYR74 0.639548 DMYR74 0.700235 DMYR74 0.643728 DMYR74 0.687008 DMYR75 0.544241 DMYR75 0.618913 DMYR75 0.55726 DMYR75 0.60862 DMYR76 0.440597 DMYR76 0.427818 DMYR76 0.343059 DMYR76 0.408834 DMYR77 0.636301 DMYR77 0.587674 DMYR77 0.489512 DMYR77 0.568517 DMYR78 0.542495 DMYR78 0.534823 DMYR78 0.450097 DMYR78 0.517973 DMYR79 0.360747 DMYR79 0.34254 DMYR79 0.27588 DMYR79 0.335741 DMYR8O 0.756199 DMYR8O 0.726411 DMYR80 0.641845 DMYR80 0.70927 DMYR81 0.781418 DMYR81 0.707195 DMYR81 0.609599 DMYR81 0.691418 DMYR82 0.81268 DMYR82 0.752096 DMYR82 0.662783 DMYR82 0.740612 DMYR83 1.042098 DMYR83 1.017263 DMYR83 0.929758 DMYR83 1.007165 DMYR84 1.018682 DMYR84 0.986503 DMYR84 0.884251 DMYR84 0.966064 DMYR85 0.990557 DMYR85 0.949377 DMYR85 0.838518 DMYR85 0.929903 DMYR86 0.947856 DMYR86 0.90282 DMYR86 0.795348 DMYR86 0.88126 DMYR87 1.007276 DMYR87 1.084669 DMYR87 0.99966 DMYR87 1.085213 SOWTR1 0.147266 JANTR 0.037028 FEBTR 0.200087 MARTR 0.132593 MATTR1 -0.115227 APRTR 0.048441 MAYTR -0.054387 JUNTR -0.131788 HARTR1 0.027549 JULTR -0.048389 AUGTR 0.139375 SEPTR 0.141348 OCTTR -0.092935 NOVTR -0.211682 DECTR -0.176978 adj Rsq 0.60104 adj Rsq 0.62789 adj Rsq 0.62124 adj Rsq 0.62452 F 69.74246 F 53.83778 F 51.57613 F 52.57683 Note: Coefficients which appear in bold style are statistically-significant. The second mnajor conclusion is that, especially with respect to the temperature effects, the seasonal specification is far more stable: the standard deviation of all districts' computed temperature effect is 0.363, which is anywhere from one half to only one fourth as large as the standard deviation of the temperature effects in the monthly specifications. 265 Table 6D.4a: Comparison of temperature effects, semi-ricardian model Seasonal JaApJulOc FeMayAuNo MarJunSeDe India -.1397 -.0742 -.0720 -.1269 I:Andhra Pradesh -.2065 .0598 -.0592 -.0710 2:Haryana -.1346 -.2009 -.1686 -.3384 3:Madhya Pradesh -.1510 -.1247 -.0688 -.0616 4:Maharashtra -.1186 -.0372 -.0096 -.0259 5:Karnataka -.1317 .0924 -.0146 .0121 6:Punjab -.1662 -.2139 -.2212 -.3667 7:Tarnil Nadu -.1398 -.0950 -.0699 -.1813 8:Uttar Pradesh -.1410 -.1488 -.1381 -.2265 9:Bihar -.1558 -.1269 -.1032 -.1867 10:Gujarat -.1099 -.1650 .0181 -.0713 14:Rajasthan -.1107 -.1313 -.1399 -.2597 15:Orissa -.1134 .0379 -.0137 -.058 17:West Bengal -.1284 -.0762 -.0558 -.1147 St.Dev. .0361 .1504 .0876 .1360 Table 6D.4b: Comparison of rainfall effects, semi-ricardian model Seasonal JaApJulOc FeMayAuNo MarJunSeDe India -.0011 -.0066 .0172 .0102 1:Andhra Pradesh .0007 .0057 .028 .0107 2:Haryana -.0022 -.0005 .0047 .0134 3:Madhya Pradesh -.0009 -.0137 .0191 .0049 4:Maharashtra -.0015 -.0032 .0278 .0089 5:Karnataka -.0005 -.0129 .0243 .0202 6:Punjab -.0015 .0070 .0030 .0000 7:Tamil Nadu -.0002 -.0053 .0146 .0228 8:Uttar Pradesh -.0012 .0002 .0079 .0076 9:Bihar .0001 -.0084 .0113 .0093 10:Gujarat -.0019 .0091 .0275 .0160 14:Rajasthan -.0032 .0002 .0190 .0124 15:Orissa -.0010 -.0101 .0138 .0014 17:West Bengal -.0002 -.0055 .0092 .0068 St. Dev. .0017 .0220 .0122 .0131 266 Distributors of COLOMBIA GERMANY IRELAND MEXICO PHILIPPINES Murdi-Prensa Sacelona Infoenlace Ltda. NO-Veldag Goverment Supplies Agency INFOTEC Intemational Bookource Center Inc. Consell ide Cent. 391 W orld Bank CarreraNob.51-21 PoppelsdorerAllee55 OgigantSolathair Av. San Femando No.37 1127-A Antipolo St, Barangay, Venezuela 08009 Barcelona Apartado Aereo 34270 53115 Bonn 4-5 Harcourt Road Col. Toiello Guerra Maoali City Tel: (343)488-3492 Publications Sfantaf do Beogota, D.C. Tel: (49 228) 949020 Dublin 2 14050 Mexico, DF Tel: (632)896 6501; 85; 6507 Fax: (343) 487-7659 Pri ces and cred it terms, ,yrefrom Tel: (57 1) 285-2798 Fax: (49 228) 217492 Tot: (353 1) 661-31 11 Tel: (5205)624-2806 Fan: (63 2) 896 1741 E-mail: borcelona@mundiprensa.es country to countryr. ConsurYt your Fax: (57 1) 280-2798 URL: httpJunwouno-verlag.do Fax: (353 1) 475-2670 Fax: (52 5) 624-2822 raditribountory CrCoEnsVuRlEmal:uovelouelromE-mail: inlotec@r1n.net.mx POLAND SRI LANKA, THE MALDIVES lcal distributor before placing an COTE D'IVOIRE ISRAEL URL: hftp://rtn.net.mx Intemational Publishing Service Lake House Bookshop order. Center dEdition et de Diifusion Africaines GHANA Yoemot LLerature Ltd. Ul. Piekna 31(37 100, Sir Chtlampalam Gardiner Mawatha (CEDA) Epp Books Sernices RO. Box 56055 Mundi-Prensa Mexico S.A. do CV. on-677 Warzawa Colombo 2 ARGENTINA 04 B.P 541 PO. Box 44 3 Yohanan Hasandlar Street c/Rio Panuco, 141-Colonia Cnauhtemoc Tel: (48 2) 6284089 Tel: (941) 32105 OficinadelLibrno lntemacional AbidjanO4 TUC Tel Aviv 61560 06500 Maxico, D.F Fax: (482)621-7255 Fax: (94 1)432104 Av. Cordoba 1877 Tel: (225) 24 6510:24 6511 Accra Tol: (972 3) 5285-397 Tel: (52 5) 533-5658 E-mail: books%ips@ikp.atm.com.pt E-mail: LHL@-s.lanko-not 1120 Buenos Aires Fax: (225)250567 Fax: (9723) 5285-397 Fax: (525)5144799 URL: http:/vww. ipscg.waw.pVipsuexporV Tel: (541) 8154354 GREECE PORTUGAL SWEDEN Fax: (541) 815-4156 CYPRUS Papasonotiiu S.A. R.O.Y Ilemational NEPAL Livraia Potugal Wennergren-Williams AB E-mail: olilibmo@sallink.com Centerior Applied Research 35, Stoumara Str. PO Box 13056 Everest Media Intematbnal Senices (P) Ltd. Apartado 2681, Rua Do Cammo 70-74 P0 . Box 1305 Cyprus College 106 92 Athens Tel Aviv61130 GPO Box5443 1200 Lisbon S-171 25Solna AUSTRALIA, FIJI, PAPUA NEW GUINEA, 6, Diogenes Street, Engomi Tel:(301)364-1826 Tel: (9723)5461423 Kathmandu Tel: (1) 347-4982 Tel: (468)705-97-50 SOLOMON ISLANDS, VANUATU, AND PO. Eox 2006 Fax: (301) 36448254 Fax: (972 3) 5461442 Tel: (977 1) 472 152 Fax: (1) 347-0264 Fax: (46 9) 27-00-71 WESTERN SAMOA Nicosia E-mail: royilrevetvisbn.net.il Fax: (9771) 224 431 E-mail: mail@wwi.se D.A. Infomiation Services Tel: (357 2) 44-1730 HAITI ROMANIA 648 Whtehorse Road Fax: (357 2) 46-2051 Cuhure Diffusbin Palestinian Authority/Middle East NETHERLANDS Compani De Librarii Bucuresli S.A. SWITZERLAND Mtcham 3132 5, Rue Capois Index Inlormation Senicos De Lindeboom InOr-Publikalies Str. Lipscani no. 26, sector 3 Libmairie Payot Service Instoutionnel Victoria CZECH REPUBLIC C.P 257 POB. 19502 Jerusalem PO. Box 202,7480 AE Haaksbergen Bucharest C6tes-de-Montbenon 30 Tel: (61() 3 210 7777 National Informalion Center Poet-au-Prince Tel: (972 2) 6271219 Tel: (31 53) 574-0004 Tel: (401() 613 9645 1002 Lausanne Fax: (61) 3 9210 7706 pmdejna, Konviktska 5 Tel: (509) 23 9260 Fax: (972 2) 6271634 Fax: (31 53) 572-9296 Fax: (401() 312 4000 Tel (41 21) 341-3229 E-mail: serice@dadirect.com.au CS - 113 57 Prague 1 Fax: (509) 23 4858 E-mail: lindeboo/woddonline.nl Fax: (41 21) 341-3235 URL: hItpJ/,wvv .dodirct.com.au Tel: (42 2) 2422-9433 ITALY URL: http:/wwo .woddonline.nU-lindeboo RUSSIAN FEDERATION Fax: (422)2422-1484 HONG KONG, MACAO Licosa CommissoonariSansoni SPA lsdatelstvo ADECO Van Diemmen EdnionsTechniques AUSTRIA URL: hltp:lwww .nis.cz/ Asia 2000 Ltd. Via Duca Di Calabria, 1/1 NEW ZEALAND 9a, Kolpachniy Pereulek Ch. de Lacuez 41 GeoWld and Co. Sales & Circubltion Department Caselia Postale 552 EBSCO NZ Ltd. Moscow 101831 CH1807 Bhenay Weihburggasse 26 DENMARK Seabird House, unit 110142 50125 Firenze Privae Mail Bag 99914 Tel: (7095)9178749 Tel: (41 21)9432673 A-1011 Wen SamoundsLitteratur 22-28 Wyndham Street, Central Tel: (55) 645-415 New Market Fax: (7095)9179259 Fax: (41 21) 943 3605 Tel: (431) 512-47-314 Rosenoems Alle 11 Hong Kong Fax: (55) 641-257 Auckland Fax: (43 1)51247-31-29 DK-1970 Frederiksberg C Tel: (852) 2530-1409 E-mail: licosaltbthcc. Tel: (64 9) 524-8119 SINGAPORE, TAIWAN, THAILAND URL:hotp://www .gerold.co/at.online Tel: (45 31)351942 Fax: (852)2526-1107 URL: httpJ/www..hcbc.itAicosa Fax: (649)5248067 MYANMAR, BRUNEI Central Books Distribution Fax: (45 31) 357022 E-mail: salIesasia2000.com.hk Ashgate Publishing Asia Pacdic Pte. Ltd. 306 Silom Road BANGLADESH URL: httpJ/www.sl.cbs.ck URL: htp//www.asia2000.com.hk JAMAICA NIGERIA 41 Kaltang Pudding Road #04-03 Bangkok 10500 Micrno Industries Development Ian Randle Publisherns Ltd. Univesity Press Limited Goden Wheel Building Tel: (66 2) 235-5400 Assistance Society (MIDAS) ECUADOR HUNGARY 206 Old Hope Road, Kingston 6 Three Crowns Building Jericho Singapore 349316 Fax: (66 2) 237-4321 House 5, Road 16 Libn Mundi Euro Info Saece Tel: 876-927-2085 Private Mail Bag 5095 Tel: (65) 741-5166 Dhanmondi FVArea Librenta Intemacional Margiszgeti Europa Haz Fax: 876-9774243 lbadan Fax: (65) 742-9356 TRINIDAD & TOBAGO Dhaka 1069 PO. Box 1741-3029 H-1 138 Budapest E-mail: irpl@colis.com Tel: (23422)41-1350 E-mail: ashgate@asianconnect.com AND THE CARRIBBEAN Tel: (8802)326427 Juan Leon Mera 851 Tel: (36 1)3508024,3508025 Fax: (23422)41-2056 Systematic Studies Ltd. Fax: (8002)811188 Cuko Fax: (36 1) 3506032 JAPAN SLOVENIA St. Augustine Shopping Center Tel: (5932)521-606; (5932)544-185 E-mail: euroinfoGmail.matavhu Eastem Book Sernice NORWAY Gospodarski Vestnik Publishing Group Eastem Main Road, St. Augustine BELGIUM Fax: (5932)504-209 3-13 Hongo 3-chome, Bunkyo-ku NIC Into AVS Duna)ska cesta S Trinidad & Tobago, West Indies Jean Do Lannoy E-mail: libimuol @1ibrimundi.com.ec INDIA Tokyo 113 Book Depatment, Postboks 6512 EKerstad 1000 Ljubhlana Tel: (868)645-8486 Av. do Roi 2062 E-mail: librimu2@1ihrimundli.comr.oc, Allied Pobliahers Ltd. Tel: (81 3( 381840861 N64606 Dsol Tel: (386 81) 133 83 47; 132 12 30 Fax: (806()64584407 1060 Brussels 751 Mount Reed Fax: (813)38180864 Tel: (47 22) 97-4500 Fax: (38661)1338030 E-mail: tobe@trinklad.nel Tel: (322) 538-5169 EGYPT, ARAB REPUBLIC OF Madras -600002 E-mail: ordernsvt-ebs.co.jp Fax: (4722)974545 E-mail: repansekj@gvestnik.si Fax: (32 2) 5304841 Al Ahram Distbutibon Agency Tel: (91 44) 852-3930 URL: htpi/wwwb .bekkoame.orjp/-sv-ohbs UGANDA Al Galaa Street Fax: (91 44) 8524649 PAKISTAN SOUTH AFRICA, BOTSWANA Gustro Ltd. BRAZIL Cairo KENYA Mizo Book Agency Forosingle tites: PO Box 9997, Madhvani Building Publicacoes Tecnas Inlemacionais Ltda. Tel: (20 2) 578-6083 INDONESIA Afdca Book Service (E.A.) LId. 65, Shahrah-e-Quaid-e-Azam Oxford Univernsty Press Southem Africa Plot 16/4 Jinja Rd. Rua Pebxoto Gomide, 209 Fax: (20 2) 5784833 Pt. Indira Limhad Quaran House, Mfangano Street Lahore 54000 Vasco Boulevard, Goodwood Kampala 01409 San Paulo, SP Jalan Borobudur 20 PO. Box 45245 Tel: (92 42)7353601 PO. Box 12119, Nl City 7463 Tel: (256 41) 251 467 Tel: (5511)259-6644 The Middle EastObserver PO Box 181 Nairobi Fax: (92 42) 576 3714 Cape Town Fax: (25641)251 468 Fax: (5511) 2584990 41 Shert Street Jakarla 10320 Tel: (2542)223641 Tel: (27 21)5954400 E-mail: gus@swituganda.com E-mail: postmaster@petiuol.br Cailo Tel: (62 21) 390-4290 Fax: (254 2) 330 272 Oxfonrd Universfty Press Fax: (27 21) 595 4430 URL: hopM/www .uol.br Tel: (20 2) 393-9732 Faa: (62 21) 390-4289 5 Bangalore Toma E-mail: oxford@oup.co.za UNITED KINGDOM Fan: (20 2) 393-9732 KOREA, REPUBLIC OF Sharae Faisal Microinoo Ltd. CANADA IRAN Daebon Trading Co. Ltd. PO Box 13033 Forsubscption orderns: PO. Box3, Aeton, Hampshire GU34 2PG Renout Publishing Co. Ltd. FINLAND Ketab Sara Co. Publishers PO. Box 34, Youida, 706 Seoun Bldg Karachi-75350 Intemational Subsciption Service England 5369 Canotek Road Akalteeminen Kirjakauppa Khaled Eslamboli Ave., 6th Street 444 Youbdo-Dong, Yeongchengpo-Ku Tel: (92 21)446307 PO. Box 41095 Tel: (44 1420) 06848 Ottawa, Ontario KlJ 9,13 PO. Box 128 Delafrooz Alley No. 8 Seoul Fax: (92 21) 4547640 Craighall Fax: (44 1420) 65389 Tel: (613)745-2665 FIN-0010t Helsinki PO. Box 15745-733 Tel: (822)785-1631/4 E-mail: ouppak@TheOFice.net Johannesburg 2024 E-mail: wbank@aukminio.demon.co.uk Fax: (613) 745-7660 Tel: (3580)121 4410 Tehmn 15117 Fax: (82)7084-315 Tel: (2711)880-1448 URL: http://wwo.microinfo.co.uk E-mail: order .dept@renoufbooks.com Fax: (350 0)1214435 Tel: (98 21) 8717819; 8716104 Pak Book Corporatien Fax: (2711) 88046248 URL: http:Vowww.renoulbooks.com E-mail: akatilaus@sotckmann.fi Fax: (98 21)8712479 MALAYSIA Aziz Chambors 21, Queens Road E-mail: iss@is.co.za VENEZUELA URL: hep:/-llww .akateeminen.coml E-mail: ketab-saraneda.nelt.ir Univenity of Malaya Cooperative Lahore Tocni-Ciencia Libmos, S.A. CHINA Bookshop, Limaed Tel: (92 42) 636 3222; 636 0885 SPAIN Centro Culdad Comerncial Tamanco China Financial & Economic FRANCE Kowkab Publishers PO. Box 1127 Fax: (92 42) 636 2328 Mundi-Prensa Libros, S.A. Nvel C2, Caracas Publishing House World Bank Publications PO. Box 19575-511 Jabln Pantai Barnu E-mail: ptc@brain.net.pk Castello 37 Tel: (58 2) 959 5547; 5035; 0016 8, Da Fo Si Doxg Jie 66, avenue dIlna Tehran 59706 Kuala Lumpur 28001 Madrid Fax: (58 2) 959 5636 Beijing 75116 Pats Tel: (98 21) 258-3723 Tel:(603)756-5000 PERU Tel: (34 1)431-3399 Tel: (8610) 6333-8257 Tel: (331) 4049-30-56/57 Fax: (98 21) 258-3723 Fax: (60 3) 755-4424 Editorial Desarnollo SA Fax: (341) 575-3998 ZAMBIA Fax: (96 10) 6401-7365 Fax. (331) 40-69-3068 I-mai: umksop@tm.net.my Apartado 3024, Umar1 E-maO: librenai@mundiprensa.es Univeesity Bookshop, University of Zambia Tel: (51 14) 285380 URL: http:/www .mundiprensa.es/ Great East Road Campus China Book Impor Centre Fax: (51 14) 286628 PO. Box 32379 PO. Box 2825 Lusaka Beijing Tel: (2601) 252 576 Fan: (26061) 253 952 'ame RECENT WORLD BANK TECHNICAL PAPERS (continued) No. :360 Salman, The Legal Frameworkfor Water Users' Associations: A Comparative Study No. 361 Laporte and Ringold. Trends in Education Access and Financing during the Transition in Central and Eastern Europe. No. 362 Foley, Floor, Madon, Lawali, Montagne, and Tounao, The Niger Household Energy Project: Promoting Rural Fuelwood Markets and Village Management of Natural Woodlands No. 364 Josling, Agricultural Trade Policies in the Andean Group: Issues and Options No. 365 Pratt, Le Gall, and de Haan, Investing in Pastoralism: Sustainable Natural Resource Use in Arid Africa and the Middle East No. .366 Carvalho and White, Combining the Quantitative and Qualitative Approaches to Poverty Measurement and Analysis: The Practice and the Potential No. 367 Colletta and Reinhold, Review of Early Childhood Policy and Programs in Sub-Saharan Africa No. 368 Pohl, Anderson, Claessens, and Djankov, Privatization and Restructuring in Central and Eastern Europe: Evidence and Policy Options No. 369 Costa-Pierce, From Farmers to Fishers: Developing Reservoir Aquaculturefor People Displaced by Dams No. 370 Dejene, Shishira, Yanda, and Johnsen, Land Degradation in Tanzania: Perceptionfrom the Village No. 371 Essama-Nssah, Analyse d'une repartition du niveau de vie No. 372 Cleaver and Schreiber, Inverser la spriale: Les interactions entre la population, I'agriculture et l'environnement en Afrique subsaharienne No. 373 Onursal and Gautam, Vehicular Air Pollution: Experiences from Seven Latin American Urban Centers No. 374 Jones, Sector Investment Programs in Africa: Issues and Experiences No. 375 Francis, Milimo, Njobvo, and Tembo, Listening to Farmers: Participatory Assessment of Policy Reform in Zambia's Agriculture Sector No. 376 Tsunokawa and Hoban, Roads and the Environment: A Handbook No. 377 Walsh and Shah, Clean Fuelsfor Asia: Technical Optionsfor Moving toward Unleaded Gasoline and Low-Sulfur Diesel No. 378 Shah and Nagpal, eds., Urban Air Quality Management Strategy in Asia: Kathmandu Valley Report No.. 379 Shah and Nagpal, eds., Urban Air Quality Management Strategy in Asia: Jakarta Report No 380 Shah and Nagpal, eds., Urban Air Quality Management Strategy in Asia: Metro Manila Report No 381 Shah and Nagpal, eds., Urban Air Quality Management Strategy in Asia: Greater Mumbai Report No 382 Barker, Tenenbaum, and Woolf, Governance and Regulation of Power Pools and System Operators: An International Comparison No. 383 Goldman, Ergas, Ralph, and Felker, Technology Institutions and Policies: Their Role in Developing Technological Capability in Industry No. 384 Kojima and Okada, Catching Up to Leadership: The Role of Technology Support Institutions in Japan's Casting Sector No. 385 Rowat, Lubrano, and Porrata, Competition Policy and MERCOSLUR No. 386 Dinar and Subramanian, Water Pricing Experiences: An Interna.itional Perspective No. 387 Oskarsson, Berglund, Seling, Snellman, Stenback, and Fritz, A Planner's Guidefor Selecting Clean-Coal Technologies for Power Plants No. 388 Sanjayan, Shen, and Jansen, Experiences with Integrated-Conservation Development Projects in Asia No. 390 Foster, Lawrence, and Morris, Groundwater in Urban Development: Assessing Management Needs and Formulating Policy Strategies No. 392 Felker, Chaudhuri, Gytorgy, and Goldman, The Pharmaceutical Industry in India and Hungary: Policies, Insititutions, and Technological Development No. 393 Mohan, ed., Bibliography of Publications: Africa Region, 1990-97 No. 394 Hill and Shields, Incentivesfor Joint Forest Management in India: Analytical Methods and Case Studies No. 395 Saleth and Dinar, Satisfying Urban Thirst: Water Supply Augmentation and Pricing Policy in Hyderabad City, India No. 396 Kikeri, Privatization and Labor: What Happens to Workers When Governments Divest? No. 397 Lovei, Phasing Out Lead from Gasoline: Worldwide Experience and Policy Implications No. 399 Kerf, Gray, Irwin, Levesque, Taylor, and Klein, Concessionsfor Infrastructure: A Guide to Their Design and Award No. 401 Benson and Clay, The Impact of Drought on Sub-Saharan African Economies: A Preliminary Examination No. 403 Welch and Fremond, The Case-by-Case Approach to Privatization: Techniques and Examples THE WORLD BANK IsX1s II St,-cct, N \V \.,isl illgton, 1).C( 2() 1 ('SA\ Iclcplh,lnc: 2(02-477-1234 I;isimlilc: 202-477-0(31 I.cix \1('1 N 64145 \\RI A1)BA.\Nk \1I(. 248423 NNORIAI)I \N\k \\,,-1.1d\NId \V.h:http:\Xx\,-.h, s(,' I -,..lail hm,zksw( \\, -ldh ,nk. m,-, 1 jlElilllil 1 4 1 214| 9 780121 341926 ISBN 0-8213-4192-8