M a n a g in Managing Agricultural g A g ric u ltu Production Risk ra l P ro d u c Innovations in Developing Countries tio n R isk Agriculture and Rural Development Department 32727 THE WORLD BANK Agriculture & Rural Development Department World Bank 1818 H Street, N.W. Washington, D.C. 20433 http://www.worldbank.org/rural REPORT NO. 32727-GLB Managing Agricultural Production Risk Innovations in Developing Countries THE WORLD BANK AGRICULTURE AND RURAL DEVELOPMENT DEPARTMENT © 2005 The International Bank for Reconstruction and Development / The World Bank 1818 H Street, NW Washington, DC 20433 Telephone 202-473-1000 Internet www.worldbank.org/rural E-mail ard@worldbank.org All rights reserved. This volume is a product of the staff of the International Bank for Reconstruction and Development/The World Bank. The findings, interpretations, and conclusions expressed in this paper do not necessarily reflect the views of the Executive Directors of The World Bank or the governments they represent. The World Bank does not guarantee the accu- racy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. The material in this publication is copyrighted. Copying and/or transmitting portions or all of this work without permission may be a violation of applicable law. The International Bank for Reconstruction and Development/ The World Bank encourages dissemination of its work and will normally grant permission to reproduce portions of the work promptly. For permission to photocopy or reprint any part of this work, please send a request with complete information to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA, telephone 978-750-8400, fax 978-750-4470, http://www. copyright.com/. All other queries on rights and licenses, including subsidiary rights, should be addressed to the Office of the Publisher, The World Bank, 1818 H Street NW, Washington, DC 20433, USA, fax 202-522-2422, e-mail pubrights@worldbank.org. Contents ACRONYMS AND ABBREVIATIONS vii PREFACE AND ACKNOWLEDGMENTS ix EXECUTIVE SUMMARY xi 1 Introduction 1 2 Risk and Risk Management in Agriculture 5 Informal Mechanisms 6 Formal Mechanisms 8 3 Approaches to Agricultural Risk in Developed Countries 11 Crop Insurance Programs in Developed Countries 11 Why the Experience of Developed Countries is not a Good Model for Developing Countries 14 4 Innovation in Managing Production Risk: Index Insurance 15 Index Insurance Alternatives 15 Basic Characteristics of an Index 15 Relative Advantages and Disadvantages of Index Insurance 17 The Trade-off Between Basis Risk and Transaction Costs 17 Where Index Insurance Is Inappropriate 17 5 0 New Approaches to Agricultural Risk Management in Developing Countries 21 Role of Government 21 Policy Objectives 23 Constraints in Agricultural Risk Management 24 Risk Principles: Layering and the Role of Index Insurance 25 Addressing the Market Insurance Risk Layer 26 Market Failure Layer 29 iii iv Contents Policy Instruments 30 Index Insurance as a Source of Contingent Funding for Government Disaster Assistance and Safety Net Programs 32 6 From Theory to Practice: Pilot Projects for Agricultural Risk Transfer in Developing Countries 35 Nicaragua: A Seven-Year Incubation Period 36 Morocco 38 India: Private Sector Led Alternative Agricultural Risk Market Development 39 Ukraine 41 Ethiopia: Ethiopian Insurance Corporation and Donor Led Ex Ante Disaster Risk Management 43 Malawi and SADC: Weather Risk Transfer to Strengthen Livelihoods and Food Security 47 Peru: Government Led Systemic Approach to Agricultural Risk Management 48 Mongolia: World Bank Contingent Credit for Livestock Mortality Index Insurance 49 Global Strategy: The Global Index Insurance Facility (GIIF) 51 7 Potential Roles for Governments and the World Bank 53 Government Roles 53 World Bank Roles 54 REFERENCES 59 Appendix 1: Weather Risk Management for Agriculture 63 The Financial Impact of Weather 63 The Weather Market 64 Weather Risk and Agriculture 65 Structuring a Weather Risk Management Solution 67 Valuing Weather Risk 74 Weather Data 79 Further Reading 81 References 81 Appendix 2. Case Studies of Agricultural Weather Risk Management 83 Indexed-based Insurance for Farmers in Alberta, Canada: The AFSC Case Study 83 Alternative Insurance Through Weather Indices in Mexico: The Agroasemex Case Study 85 Weather Insurance for Farmers in the Developing World: Case Studies from India and Ukraine 90 Technology Application Case Studies: Grassland Index Insurance Using Satellite Imagery 107 References 108 NOTES 111 Contents v Tables 2.1 Risk Management Strategies in Agriculture 8 4.1 Advantages and Disadvantages of Index Insurance 18 5.1 Risk Transfer Strategies 28 6.1 Summary of Case Studies 36 6.2 Reasons for Buying Weather Index Insurance in India 43 A2.1 Options for CHU Contracts 85 A2.2 Total Liability Factored into the Agroasemex Business Plan for Autumn-Winter 2001­­2002 86 A2.3 Summary of the Methodology to Calculate the Eleven FCDD Indices 87 A2.4 Comparative Analysis Between the Observed Historical Severity Indices (indemnities/total liability) and the Estimated Severity Indices for the Crops and Risks Selected 88 A2.5 Specifications of Call Option Structures Considered by Agroasemex 89 A2.6 Estimated Commercial Premium for Weather Derivative Structures (in US$) 90 A2.7 Weather Insurance Contracts Offered to Groundnut and Castor Farmers 94 A2.8 Pilot Statistics, 2003 95 A2.9 Payout Structure Per Acre for Groundnut Weather Insurance Policy for Narayanpet Mandal, Mahahbubnagar District (2004) 96 A2.10 Payout Structure Per Acre for Castor and Groundnut Excess Rainfall Weather Insurance Policy for Narayanpet, Mahahbubnagar 97 A2.11 Relationship Between SHR and Winter Wheat Yields During the Vegetative Growth Phase of Plant Development 101 A2.12 Relationship Between SHR and Financial Losses Associated with Winter Wheat Yield Fluctuations 102 A2.13 Correlation Coefficients for the Interannual Variability of Cumulative Rainfall, Average Temperature, and the SHR Index Measured at Five UHC Weather Stations in Kherson 103 Boxes 2.1 Asset-Based Risk Management 7 5.1 Reinsurance 22 6.1 India Impact Assessment 42 7.1 Examples of Potential World Bank Investment Lending Projects to Facilitate Risk Management 57 Figures 2.1 Independent Versus Correlated Risk 9 3.1 Crop Insurance Premiums and Indemnities in the United States 12 4.1 Payout Structure for a Hypothetical Rainfall Contract 16 5.1 Framework for Governmental Agricultural Risk Management Policy Formulation 23 5.2 Average April to October Rainfall for Thirteen Malawi Weather Stations 26 5.3 Histogram of Simulated SADC Drought Events 29 5.4 Government-Sponsored DOC as Risk Transfer Product Between National and International Risk Markets 30 vi Contents 7.1 Potential Impacts of Natural Hazards 54 A1.1 Notional Value of All Weather Contracts in US$ 65 A1.2 Percentage of Total Weather Contracts by Location (excluding CME trades) 66 A1.3 Potential End User Market by Economic Sector 2003­2004 66 A1.4 Call Option Payout Structure and Wheat Grower's Losses 72 A1.5 Collar Payout Structure and Agrochemical Company's Deviation from Budgeted Revenue 73 A1.6 Schematic of Historical Revenues of a Business and the Impact of Weather Hedging 78 A2.1 Relationship Between the Daily Rate of Development of Corn Minimum and Maximum Temperatures 84 A2.2 Comparative Accumulated Distribution Probability Function Based on a "Probability of Exceedence Curve" for the Historical and Modeled Results (payouts in US$) 89 A2.3 Mahahbubnagar District Groundnut Yields Versus Groundnut Rainfall Index 93 A2.4 Payout Structure of Groundnut Weather Insurance Policy Held by Farmers with Small, Medium, and Large Land Holdings 94 A2.5 Payout Structure of Groundnut Weather Insurance Policy for Narayanpet Mandal, Mahahbubnagar District, 2004 97 A2.6 An Example of the Marketing Leaflet for Groundnut (DGN), Castor (DCN), and Excess Rainfall (EN) Protection in Narayanpet Mandal, Mahahbubnagar District, 2004 98 A2.7 Winter Wheat Yields for Kherson Oblast, 1971­2001 100 A2.8 Cumulative Rainfall and Average Temperature for Behtery Weather Station for April 15 to June 30, 1973­2002 104 A2.9 SHR Index for Behtery Weather Station, 1973­2002 105 A2.10 Sample Contract for Behtery Weather Station 106 Acronyms and Abbreviations ACP Africa-Caribbean-Pacific APF Agricultural Policy Framework of Canada APH actual production history ARD Agriculture and Rural Development Department of the World Bank Group BASIX Livelihood Promotion and Microfinance entity of Andhra Pradesh BIP base insurance product BSFL Bhartiya Samruddhi Finance Limited (part of BASIX) CAIS Canadian Agricultural Income Stabilization CAT catastrophe COFIDE Corporación Financiera de Desarollo S.A. (Development Finance Corporation located in Lima, Peru) CRDB Cooperative and Rural Development Bank Limited, a private commercial bank CRMG Commodity Risk Management Group (ARD, The World Bank) DECRG Development Economics Research Group of The World Bank DOC disaster option for CAT risk DPPC Disaster Prevention and Preparedness Commission (Ethiopia) DRP disaster response product EC/ACP European Commission/Africa-Caribbean-Pacific EIC Ethiopia Insurance Corporation ENESA Entidad Estatal de Seguros Agrarios, the National Agricultural Insurance Agency of Spain ENSO El Niño southern oscillation (sea surface temperatures) ESDVP Environmentally Sustainable Development Vice Presidency ESSD The World Bank Environmentally and Socially Sustainable Development Advisory Service FAO Food and Agriculture Organization of the United Nations FCIP Federal Crop Insurance Program FSE The Financial Sector Group of The World Bank GDP gross domestic product GIIF Global Index Insurance Facility (proposed by CRMG) vii viii Acronyms and Abbreviations GMO genetically modified organisms IBLI index-based livestock insurance ICICI A private general insurance Lombard company in India ICRISAT International Crops Research Institute for the Semi-Arid Tropics IFC International Finance Corporation of the World Bank Group IFFCO-Tokio A private general insurance company in India, a joint venture between Tokio-Marine and the Indian Fertilizer Association IFPRI International Food Policy Research Institute IMF International Monetary Fund INISER Instituto Nicaraguense de Seguros y Reaseguros Nicaraguan Institute for Insurance and Reinsurance ISMEA Istituto di Servizi per il Mercato Agricolo Alimentare (Italian Institute for Services to Agricultural Food Markets) KBS LAB Krishna Bhima Samruddhi Local Area Bank LIL learning and innovation loan MAMDA Mutuelle Agricole Marocaine d'Assurance MMPI Malawi Maize Production Index NASFAM National Smallholders Association NDVI normalized difference vegetation index NGO nongovernmental organization NMSA National Meteorological Services Agency OECD Organization for Economic Cooperation and Development OI Opportunity International PI production insurance RI reinsurance SADC Southern African Development Community SECO State Secretariat for Economic Affairs, Swiss Trade Commission SENAMHI Servicio Nacional de Meteorologia e Hidrologia del Peru (National Meteorology and Hydrology Service of Peru) SRA Standard Reinsurance Agreement (U.S. crop insurance) TCDAI Technical Committee for the Development of Agriculture Insurance (Peru) UNCTAD United Nations Conference on Trade and Development WFP World Food Program of the United Nations Preface and Acknowledgments This document was produced by Ulrich Hess, as task manager, and by Jerry Skees, Andrea Stoppa, Barry Barnett, and John Nash, using background papers written by Robert Townsend; Paul Siegel; and Jerry Skees, Barry Barnett, and Jason Hartell. (These papers can be viewed at the Commodity Risk Management Group (CRMG) web site, www.itf-commrisk.org.) Panos Varangis led the work for this study during its conceptual stage. The two appendixes are shortened versions of contributions by CRMG authors Joanna Syroka and Hector Ibarra to a forthcoming ISMEA (Istituto di Servizi per il Mercato Agricolo Alimentare) publication on innovations in agri- cultural risk management. Although motivated by the solid and growing literature on alter- native risk management techniques, this paper is ultimately driven by empirical results that would have been impossible to obtain with- out the development community's support and demand for action. At The World Bank, Karen Brooks and Richard Scobey, rural sec- tor managers in the Africa Region, supported the conceptual work and instilled a sense of realism and purpose into the ideas expressed here. Jock Anderson and Derek Byerlee in the Agriculture and Rural Development department continuously refreshed our ideas in the areas of agricultural risk management and food security risk man- agement. Kevin Cleaver and Sushma Ganguly, Sector Director and Sector Manager, respectively, in the Agriculture and Rural Development department, gave motivational advice and guidance. Ken Newcombe, ESDVP, encouraged this work and has become a champion of the Global Index Insurance Facility (GIIF). In his IFC days, Cesare Calari, FSE, was an early supporter of weather risk management concepts, and he continues to encourage this line of thinking in his various capacities. Rodney Lester, senior insurance expert in FSE, also contributed advice and support. Xavier Gine, DECRG, helped to shape our thinking on smallholder access to financial services. Our colleagues in the social development and social protection areas--notably Harold Alderman, Will Wiseman, and Elena Galliano--helped with the crossover to the social risk management realm, providing a better understanding of the needs of vulnerable populations and the relevance of insurance techniques for safety nets. Development partners have continuously prompted quality leaps forward through their particular expertise. Richard Wilcox of the UN World Food Program (WFP) pushes the weather insur- ix x Managing Agricultural Production Risk ance idea to new limits and has shaped our The demand for systematic techniques of agri- thinking at that level. Alexander Sarris, FAO, cultural risk management in developing countries and Lamon Rutten, UNCTAD, supported CRMG ultimately came from the people who deal with in the areas of commodity risk management. farmers and who partly make the farmers' risks The key concepts espoused in this paper have their own. The vision and inspiration of Nachiket been developed in the academic community as Mor, of ICICI Bank, India, and Vijay Mahajan, of well. Ronald Duncan and his group at the World BASIX, India, are the real motivators behind the Bank systematically explored index insurance astounding success of weather insurance tech- ideas in the early 1990s (Priovolos and Duncan niques. This paper and its proposals would be 1991). Also in the early 1990s, Peter Hazell, unthinkable without the ICICI Lombard and IFPRI, and Jerry Skees analyzed the shortcom- BASIX weather insurance pilots and their revela- ings of traditional crop insurance and suggested tion that farmers understand and appreciate the the weather index insurance alternative. transparency and timeliness of the product. This ESW insists on market-based insurance Ramesh and Vasumathi in Mahahbubnagar, techniques, the only sustainable way to transfer Ramana and Gunaranjan in Hyderabad and risk out of agriculture. At the same time, market Mumbai, Virat Divyakirti at ICICI Lombard, and gaps exist, and often markets fail the poor. CRMG Bindu Ananth, also at ICICI, were the architects of and its partners--by "crowding in" the private a simple innovation that promises to change India's sector--are building the bridges necessary to span rural landscape. Champions for pilot projects else- these gaps. None of this would have been possible where are Rachid Guessous, MAMDA, in Morocco; without the visionary thinking of leaders in the Ramon Serrano, INISER, in Nicaragua; and weather risk management markets. Ravi Nathan, Shadreck Mapfumo, OI, in Malawi. ACE Insurance of North America, in particular, has The authors wish to acknowledge the generous helped to globalize the market beyond OECD support of the Swiss State Secretariat for Economic countries thanks to creative partnership and risk- Affairs, SECO. SECO has supported CRMG's pilot sharing structures that include marketing partners programs in innovative agricultural risk manage- from developing countries. His vision continues to ment, and major lessons from these pilots inform inspire the market and our work. Crucial advisors this report. The European Commission and, in on the work and ideas of the CRMG as repre- particular, Henny Gerner are associated with the sented here are Brian Tobben and William Dick of work of CRMG and, by extension, with this ESW Partner RE; Juerg Trueb, of Swiss RE; and Rick through their constructive criticism of and support McConnell, formerly of the Agricultural Financial for the idea of the Global Index Insurance Facility Services Corporation, Alberta. Bruce Tozer, at (GIIF). Rabobank, and Roy Leighton, at Carlyon, have Finally, the authors express their sincere grat- advised and encouraged CRMG and the Inter- itude to World Bank reviewers Jock Anderson national Task Force for Commodity Risk Man- and Stephen Mink and to Celeste Sullivan and agement, throughout their existence, with Anne Goes, of GlobalAgRisk, Inc., for their edi- wisdom and passion. torial assistance. Executive Summary The creation of risk transfer markets for weather events in devel- oping and emerging economies is rapidly progressing. This docu- ment describes several sources of risk that create poverty traps for poor households and impede the development process, focusing on low-probability, high-consequence weather risk events as they relate to rural households. These types of risks are highly corre- lated and require special financing and access to global markets if they are to be pooled, rendered diversifiable, and improved in pric- ing. Thus, a significant contribution of this paper is the introduction of index insurance, highlighting its use at the micro-, meso-, and macrolevels for risk transfer. By using index insurance products, it is possible to organize systems that take advantage of global mar- kets to transfer the correlated risks associated with low-probability, high-consequence events out of developing countries. This docu- ment presents both a conceptual backdrop for understanding this system and a progress report on several World Bank efforts to assist countries in using their limited government resources to facilitate market-based agricultural risk transfer when faced with natural disasters. While global markets providing reinsurance for natural disasters are both large and growing, they are rarely interested in taking such risk from developing and emerging economies. In part, this is because developing countries have weak primary insurance mar- kets. Before agreeing to provide reinsurance, global reinsurers engage in due diligence investigations of primary insurers and of the risks the primary insurers wish to transfer. Compared to traditional insurance products, index insurance has far fewer problems with hidden information and hidden action. This reduces the reinsurers' due diligence and underwriting costs and makes accepting natural disaster risk from new insurance providers in developing countries more attractive. Nonetheless, natural disaster losses can be signifi- cant, and carefully crafted ways to finance such losses are critical preconditions for shifting the risk into global markets. Innovation in pooling these risks globally may also facilitate the transfer of natu- ral disaster risk from developing countries. One global innovation currently being prepared by the World Bank and the European Commission involves a Global Index Insurance Facility (GIIF). The GIIF will have three functions targeted at helping insurance providers in developing countries build capac- ity: (1) supporting the technical assistance and infrastructure needed xi xii Managing Agricultural Production Risk to develop index insurance based on quality data; responses are optimal. They fail to provide an effec- (2) aggregating and pooling risk from different tive safety net for the poor; they can be inequitable developing countries to improve pricing and risk and untimely; and they create a dependency that transfer into the global reinsurance and capital has dire consequences. markets; and (3) cofinancing certain insurance If the planning for and financing of extreme products on a bilateral basis from donor to develop- weather events were to occur ex ante, access to ing country. Importantly, the third function will be both formal and informal lending should improve. separate from the commercial activity represented As broader financial services become more acces- in the first and second functions. A global effort sible to the rural poor, newer technologies will to facilitate these three functions could represent be used, and improvements in productivity and a major breakthrough for those developing coun- incomes should follow. tries exposed to extreme natural disaster risk. Farmers around the world utilize various risk Another promising realm of innovation is the coping and risk management strategies, but many development of improved technology both to mea- of these strategies are inefficient. The economic sure weather and to link it to farming systems to development literature is full of cases illustrating forecast crop yields. Improved and less costly sys- how poor, risk-averse farmers often forego poten- tems for measuring weather events in developing tially higher incomes to reduce their risk exposure. countries will play a significant role in the potential Both individual households and the larger society success of many of the ideas presented here. Secure incur costs for smoothing consumption across and accurate measurement will influence both the income shocks. In many cases, following major pricing of index insurance and the demand from income shocks, the poor must resort to high inter- end users. Improvements in developing countries est rate loans. Many argue that the poor cannot first in measuring the vegetative cover using satel- afford to purchase ex ante insurance protection lite images and then in forecasting the value of that against extreme weather events, but the wide- vegetation in terms of crop yields or grazing value spread use of ex post loans suggests otherwise. could lead to the availability of enhanced types of The challenge remains of how to make insur- index insurance products. Additionally, more ance against extreme weather events both more sophisticated crop models linking weather, man- effective and more affordable. Two major consid- agement systems, and soil condition can be used to erations inhibit the development of risk transfer provide insurance products that protect against the markets for agricultural losses caused by extreme dominant random variable affecting production-- weather events: First, organizing ex ante financing the weather. for highly correlated losses can result in ex- Transferring risk out of developing countries is tremely large financial exposure; and, second, important for a number of reasons. Natural dis- asymmetric information problems, such as moral asters impede development, push households into hazard and adverse selection, lead to high trans- poverty, and drain fiscal resources. Many natural action costs. The latter also makes it nearly impos- disasters are directly tied to extreme weather events sible to provide traditional agricultural insurance that can have devastating impacts on agriculture. for small farmers, because the large fixed transac- Nearly three-fourths of the 1.3 billion people world- tion costs greatly increase the average cost, per wide living on less than US$1 per day depend on monetary unit, of insurance protection for small- agriculture for their livelihoods. In many countries holder agriculture. Unfortunately, there are few around the world, agricultural development clears successful examples to consider; the heavily sub- the way for overall economic development in the sidized crop insurance provided by governments broader economy, forging a strong link between in developed countries is both costly and ques- weather, the livelihoods of the poor, and develop- tionable in terms of net social welfare. ment. Yet, no effective ex ante solutions for deal- Researchers frequently find that economic deci- ing with weather risks in developing countries sion makers underestimate the likelihood and/or exist. Rather, developing countries, the World magnitude of low-probability, high-consequence Bank, and the donor community are currently loss events, leading to a reduced willingness to heavily exposed to natural disaster risk via ex pay for insurance to protect against these events. post responses such as financial bailouts, debt for- At the same time, because insurers have little giveness, and emergency response.1 None of these empirical information about the likelihood and/or Executive Summary xiii magnitude of extreme events, they tend to add widespread, correlated agricultural production large extra costs to premium rates for insurance losses. To the extent that institutions can be created products protecting against them. This diver- to aggregate and pool the low-probability, high- gence between what potential purchasers will pay consequence tail risk that results from writing and what insurers will accept results in agricul- insurance on these events, the divergence between tural insurance markets that clear less than insurers' willingness to accept and potential pur- socially optimal quantities of risk transfer. chasers' willingness to pay should decrease, caus- New conceptual models are being developed ing the market to clear at high quantities of risk to facilitate the transfer of extreme weather risk transfer. out of developing countries. This document This paper was written to inform a broad range reports on the progress of several ongoing efforts of decision makers about the progress being made by the Commodity Risk Management Group in risk transfer for natural disaster risk. While the (CRMG) at the World Bank that have been moti- focus here is on agriculture, many of the same con- vated by these models. All of these efforts are built cepts can clearly also be used for other sectors on the premise that index-based insurance prod- exposed to natural disaster risk. Two basic innova- ucts can effectively address the challenges of the tions dominate the conceptual framework: (1) use ex ante financing of highly correlated losses and of index-based insurance; and (2) layering risk to high transaction costs. Index insurance products facilitate risk transfer. In many cases, individuals pay indemnities based on an independent meas- will self-insure against the layer of risk com- ure highly correlated with realized losses. Unlike posed of high-probability, low-consequence traditional crop insurance, which attempts to losses. Some form of government intervention measure individual farm yields, index insurance may be required to achieve higher levels of risk makes use of variables largely exogenous to the transfer in the layer of risk composed of low-prob- individual policyholder, including area yield or ability, high-consequence losses. Between these weather events such as temperature or rainfall. two extremes lies a layer of risk that, with appro- This feature greatly reduces the need for priate risk transfer and pooling structures, can be deductibles and copayments, since it results in transferred using market mechanisms. very little exposure to asymmetric information Since catastrophe risks (CAT risks) are one of problems, such as moral hazard and adverse selec- the impediments to market development, a tion. By eliminating farm-level loss adjustment, framework has been developed for government index insurance products achieve lower transaction action in the management of agricultural risk that costs than are possible with traditional agricul- includes models for government intermediation tural insurance products. of catastrophic risk through government disaster Purchasers of index insurance products are options for CAT risk, or DOC. This framework pro- exposed to basis risk. Since index insurance poses that governments buy index-based cata- indemnities are triggered not by farm-level losses strophic risk coverage in international markets but rather by the value of an independent measure and offer them at rates lower than global market (the index), a policyholder can experience a loss rates to local insurers, who then pass the savings and yet receive no indemnity. Conversely, the on to end users in developing countries. This sys- policyholder may not experience a loss and yet tem would mitigate large-loss/infrequent risks nonetheless receive an indemnity. The effective- that are usually difficult and expensive to reinsure ness of index insurance as a risk management in traditional reinsurance markets and would ulti- tool depends on how positively correlated farm- mately allow local insurers to cover more people level losses are with the underlying index. against the extreme risks in an ex ante fashion. Importantly, since farmers have incentives to con- This paper includes several case studies illustrat- tinue to produce or to try to save their crops and ing the application of these concepts in countries livestock even in the face of bad weather events, around the world. While the specifics vary based on index insurance should provide for a more effi- each country's needs, all of the cases involve the use cient allocation of resources. of index insurance and/or the layering of risk to Since they are standardized and transparent, facilitate risk transfer. The final chapter of this doc- index insurance products can also function as re- ument describes potential future roles for the World insurance instruments that transfer the risk of Bank in the area of agricultural risk management. 1 Introduction This document presents innovations in agricultural risk management for natural disaster risk, with the focus on defining practical roles for governments of developing countries and the World Bank in devel- oping risk management strategies.2 Recent success stories demon- strate that the World Bank can play a role in assisting countries in taking actions that effectively use limited government resources to facilitate market-based agricultural risk transfer. This is important, as developing countries, the World Bank, and the donor community are currently heavily exposed to natural disaster risk without the benefit of ex ante structures to finance losses. Instead, at each big drought or other natural disaster, those affected must appeal for financial sup- port, leaving them vulnerable to the mercy of ad hoc responses from government, the international financial institutions, and donors. In most developing countries, livelihoods are not insured by inter- national insurance/reinsurance providers, capital markets, or even government budgets. In addition, natural disasters and price risk in agriculture also impede development of both formal and informal banking. Without access to credit, risk-averse poor farmers are locked in poverty, burdened with old technology, and faced with an ineffi- cient allocation of resources. Advances in risk transfer in developed countries are leading the way to solutions to many social problems. Shiller (2003) documents progress and charts a course for far more innovation as the democra- tization of finance and technology spur global risk pooling. Financial and reinsurance markets in developed countries are rapidly devising index-based instruments that allow for the transfer of systemic risks and even of livelihood risks. Innovations in risk transfer for natural disasters have been well documented (Doherty 1997; Skees 1999b). The challenge is to make these innovations relevant in developing countries and to facilitate knowledge and access. Is the absence of formal transfers of natural disaster risk inevitable in developing countries? Clearly not; formal global markets for off- setting natural disaster risks and weather risks are widely used in developed countries.3 This document demonstrates how these mar- kets can be used to insure natural disaster risk in developing coun- tries. Agricultural sectors in developing countries are much more exposed to the vagaries of weather than are those of richer countries, so this protection would be even more valuable to them. Is it a luxury to offer insurance to poor people who lack proper roads or even safe drinking water? Every government must set its 1 2 Managing Agricultural Production Risk own priorities. Careful consideration of the bene- BASIX (a microfinance entity in Andhra Pradesh) fits and costs of different interventions is critical. estimates that all of the 427 farmers who bought Still, the poor are forced to make production deci- weather insurance policies in 2003 have small- to sions using the objective of minimizing risk, rather medium-sized farms of between two and ten acres, than maximizing profits, and thus they must forego providing an average yearly income of 15,000 to more remunerative activities that could provide 30,000 Rupees, or between US$1 and US$2 per day. means of escape from their poverty. An effective and Currently, many farmers buying weather insur- timely insurance mechanism might allow people to ance in India are repeat customers. Clearly, these engage in higher risk, higher return activities with- farmers were not too poor to buy the product. Early out putting their livelihoods at risk. Spurring devel- survey results demonstrate that more than half of opment via improved financial markets is important those purchasing the insurance list managing risk for developing countries. as their primary reason. Some farmers might have Are there any effective precedents for agricul- chosen this new insurance option over the prospect tural insurance mechanisms in developing coun- of paying high interest to moneylenders when cash tries? While these innovations are just taking hold, is needed after a harvest failure. progress has been made with weather insurance Is India's insurance program sustainable? With for farmers in India, Ukraine, Nicaragua, Malawi, the pilot program now in its third year and other Ethiopia, and Mexico. Several other experiments insurance companies replicating and selling the are also documented in this work. Weather insured product, BASIX has mainstreamed the weather in- farmers in India say they either have a good crop-- surance product and automated delivery to an ex- in which case it does not matter if they do not recoup pected 8,000 clients for the 2005 season. Countries the insurance premium--or they have a monsoon in sub-Saharan Africa and Latin America are start- failure, in which case they receive an insurance pay- ing their own weather insurance projects at micro- out. This payout will at least cover the farmers' cash and macrolevels. Ethiopia is piloting a weather- outlay and perhaps provide them with enough extra insurance-based drought emergency response, for money to keep their children in school and to pre- example. Furthermore, weather insurance seems to serve assets they would otherwise be forced to liqui- be a good business. The Indian weather insurance date, often at greatly reduced prices. These farmers program has emerged without the support of gov- will be likely to invest a little more in the right seeds ernment subsidies. The Commodity Risk Manage- and fertilizer at the right time. Quantifying this ment Group (CRMG) of the World Bank has advised impact is difficult right now, but a large impact those who were ready to try these new approaches assessment will soon provide more information to agricultural risk management. on the effectiveness of this program. It is clear How can this process be operationalized in the already, however, that when offered the choice, World Bank and elsewhere? Task managers and many farmers in India will pay for fully priced practitioners may want to follow this work with weather insurance. Even farmers with access to potential projects, but how do they get started? the government-subsidized crop insurance prod- This document presents ideas on how to structure uct choose to buy the market-priced weather in- a solid framework of action. Among the important surance product. They say they like the objective public goods that governments and the World Bank nature of the weather index; they can check the might provide are, for example, weather stations weather station measurements themselves. They and risk financing for catastrophic protection. also like the timely payout. Indeed, on this count, Governments in drought-prone countries and the new rainfall index insurance, which pays on a donors and relief agencies should also be aware of timely basis, compares favorably to the national other kinds of projects that use risk management crop insurance product, which might pay only after markets to improve the response to weather-related as much as eighteen months. shocks. This document explores how current ad Is this insurance only suitable for large commer- hoc disaster relief mechanisms can be modified and cial farmers? One true advantage of weather insur- complemented by a more systematic response to ance is that it can be targeted to small farmers, as no recurrent droughts. monitoring is needed to verify farm-level losses. When assessing proper roles for government, The Indian experience clearly demonstrates that the first factors to consider are the economic bene- small farmers find value in weather insurance. fits that can be created by risk management tools, Introduction 3 the characteristics of the risks faced by farmers in search for new solutions. Chapter 4 explores alter- a specific area, and the challenges associated with nate solutions based on the concept of weather creating and maintaining risk management tools index insurance that covers farmers against weather such as insurance. In general, agricultural risk events leading to serious agricultural losses, high- management presents no "one-size-fits-all" policy lighting the advantages of such systems for devel- recommendation for the role of government. Most oping countries. Chapter 5 brings together two core governments consider at least four criteria when innovations: first, the use of index insurance to in- considering alternatives for addressing agricul- sure against detrimental weather events, a form tural risk management needs: (1) fiscal constraint; with significantly lower monitoring costs; and sec- (2) growth; (3) market-oriented risk-transfer; and ond, the layering of insurance products to segment (4) social goals of reducing poverty and vulnerabil- risk more efficiently, thus allowing for transfer of ity in rural areas. correlated risk. These innovations provide a rich Chapter 2 of this document begins with an framework for introducing new approaches to risk overview of risk in agriculture, focusing on how sharing and risk transfer in developing countries. decision makers currently cope with and manage Chapter 5 outlines an effective role for the World risk in developing countries and on the impedi- Bank and other donors in this important domain of ments to developing effective risk transfer markets. natural hazard risk management. Chapter 6 pro- High transaction costs, problems with correlated vides an overview of a number of ongoing agricul- risk, and the classic problems of moral hazard and tural risk pilot programs and case studies for in adverse selection clearly increase the cost of tradi- various countries. Finally, Chapter 7 makes rec- tional insurance. Chapter 3 reviews in detail the ex- ommendations for the role of the World Bank and periences of some developed countries with country governments in facilitating the develop- agricultural risk transfer. A clear message emerges ment of innovation in agricultural risk manage- about the costs to governments and the inefficien- ment. Following the core chapters, the report cies of these systems, supporting the need to search includes two detailed appendixes: the first explains for new solutions appropriate for developing coun- how to structure and price weather index insur- tries. The stark contrast between what is possible in ance; the second provides more background to risk a developed country versus what is possible in a transfer programs and experiences in Ukraine, developing country further motivates a continuing Mexico, Canada, and India. 2 Risk and Risk Management in Agriculture Agricultural risk is associated with negative outcomes stemming from imperfectly predictable biological, climatic, and price variables. These variables include natural adversities (for example, pests and diseases), climatic factors not within the control of agricultural pro- ducers, and adverse changes in both input and output prices. To set the stage for the discussion on how to deal with risk in agriculture, we classify the different sources of that risk.4 Agriculture is often characterized by high variability of production outcomes, that is, by production risk. Unlike most other entrepreneurs, agricultural producers cannot predict with certainty the amount of output their production process will yield, due to external factors such as weather, pests, and diseases. Agricultural producers can also be hindered by adverse events during harvesting or collecting that may result in production losses. Both input and output price volatility are important sources of market risk in agriculture. Prices of agricultural commodities are extremely volatile. Output price variability originates from both endogenous and exogenous market shocks. Segmented agricultural markets will be influenced mainly by local supply and demand con- ditions, while more globally integrated markets will be significantly affected by international production dynamics. In local markets, price risk is sometimes mitigated by the "natural hedge" effect, in which an increase (decrease) in annual production tends to decrease (increase) output price (though not necessarily farmers' revenues). In integrated markets, a reduction in prices is generally not correlated with local supply conditions, and therefore price shocks may affect producers in a more significant way. Another kind of market risk arises in the process of delivering production to the marketplace. The inability to deliver perishable products to the right market at the right time can impair producers' efforts. The lack of infrastructure and of well- developed markets makes this a significant source of risk in many developing countries. The ways businesses finance their activities is a major concern for many economic enterprises. In this respect, agriculture has its own peculiarities. Many agricultural production cycles stretch over long periods, and farmers must anticipate expenses they will only be able to recuperate after marketing their product. This leads to potential cash flow problems, which are often exacerbated by lack of access to credit and the high cost of borrowing. These problems can be classi- fied as financial risk. 5 6 Managing Agricultural Production Risk Institutional risk, that is, risk generated by un- the income gains might be larger than for less risky expected changes in regulations that affect produc- choices. This inability to accept and manage risk and ers' activities, constitutes another important source accumulate and retain wealth is sometimes referred of uncertainty for agricultural producers. Changes to as the "the poverty trap" (World Bank 2001). in regulations can have significant impact on the Once producers have decided to engage in farm- profitability of farming activities. This is particularly ing activities, the production strategy selected be- true for import/export regimes and for dedicated comes an important means of mitigating the risk of support schemes, but sanitary and phytosanitary crop failure. Traditional cropping systems in many regulations too can restrict producers' activities and places rely on crop and plot diversification. Crop impose costs on households. diversification and intercropping systems reduce Like most other entrepreneurs, agricultural pro- the risk of crop failure due to adverse weather ducers are responsible for all the consequences of events, crop pests, or insect attacks. Morduch (1995) their activities. Growing concern over the impact of presents evidence that households whose con- agriculture on the environment, however, includ- sumption levels are close to subsistence (and which ing the introduction of genetically modified organ- are therefore highly vulnerable to income shocks) isms (GMO), may cause an increase in producer devote a larger share of land to safer, traditional liability risk. Finally, agricultural households, along varieties such as rice and castor than to riskier, with other economic enterprises, are exposed to high-yielding varieties. Morduch also finds that personal risks to the well-being of people who work near-subsistence households diversify their plots on the farm and asset risks, including possible dam- spatially to reduce the impact of weather shocks age or theft of production equipment and assets. that vary by location. (See Box 2.1.) Apart from altering agricultural production In discussing how to design appropriate risk strategies, households also smooth income by diver- management policies, it is useful to understand sifying income sources, thus minimizing the effect strategies and mechanisms employed by producers of a negative shock to any one of them. According to deal with risk, including the distinction between to Walker and Ryan (1990), most rural households informal and formal risk management mechanisms in villages of semi-arid India surveyed by the Inter- and between ex ante and ex post strategies.5 As national Crops Research Institute for the Semi-Arid highlighted in the 2000/2001 World Development Tropics (ICRISAT) generate income from at least Report (World Bank, 2001), informal strategies two different sources; typically, crop income is ac- are identified as "arrangements that involve indi- companied by some livestock or dairy income. Off- viduals or households or such groups as commu- farm seasonal labor, trade, and sale of handicrafts nities or villages," while formal arrangements are are also common income sources. The importance "market-based activities and publicly provided to risk management of income source diversifica- mechanisms." The ex ante or ex post classification tion is emphasized by the Rosenzweig and Stark focuses on the point at which the reaction to risk (1989), who find that households with high farm takes place: ex ante responses take place before the profit volatility are more likely to have a household potential harming event; ex post responses take member engaged in steady wage employment. place after the event. Ex ante reactions can be further Buffer stock accumulation of crops or liquid as- divided into on-farm strategies and risk-sharing sets and the use of credit present obvious means by strategies (Anderson, 2001). Table 2.1 summarizes which households can smooth consumption. Lim these classifications. and Townsend (1998) show that currency and crop inventories function as buffers or precautionary INFORMAL MECHANISMS6 savings. Crop-sharing arrangements in renting land and Ex ante informal strategies are characterized by hiring labor can also provide an effective means diversification of income sources and choice of agri- of sharing risk among individuals, thus reducing culturalproductionstrategy.Onestrategyproducers producer risk exposure (Hazell 1992). Other risk can employ is simply to avoid risk. In many cases, sharing mechanisms, such as community-level extreme poverty makes people very risk averse; risk pooling, occur in specific communities or ex- producers facing these circumstances often avoid tended households where group members transfer activities that entail significant risk, even though resources among themselves to rebalance marginal Risk and Risk Management in Agriculture 7 Box 2.1 Asset-Based Risk Management Siegel (2005) broadens the risk discussion into an the potential to improve household livelihood options, asset-based risk management framework. This compre- yet their ultimate success will depend on the linkages hensive approach considers the dynamics of risks among assets, context, behavior, and outcomes. Thus, within a given context. The asset-based approach uses the real question to be asked is what optimal risk man- a "livelihood focus," recognizing that rural households agement instruments will allow households to maxi- hold a portfolio of assets that they allocate among a mize their objectives in terms of expected income and range of welfare generating activities and that the par- variability of income? ticular livelihood activities pursued reflect explicit (or The relationship between assets and productivity implicit) multidimensional objectives that include eco- explains the poverty cycle and the difficulty the poor nomic, social, cultural, and environmental outcomes have in improving their livelihoods. A household's (Chambers and Conway 1992; Carney et al. 1999). portfolio of assets influences their risk attitude and The asset-based approach helps clarify why and how their ability to respond to risk. Assets also determine households manage assets and risks to "select" certain the types of activities that can be undertaken. More livelihood strategies for achieving welfare outcomes productive activities are typically associated with given specific asset-context interface conditions. greater risk, so how assets are utilized will impact The asset-based risk management approach focuses productivity as a function of both expected income on the long-term implications of short-term decisions and variability of income. At the household level, about the allocation of assets. Coping strategies used agricultural risk management instruments reduce the by poor rural households can lead to the degradation variability of household incomes. The expectation is or decapitalization of assets, as when, for example, that by reducing risk and uncertainty, households will trees are cut down or children are removed from be able to accumulate assets and undertake more school, and these actions can contribute to a cycle of productive investments. poverty. Alternatively, livelihood strategies that lead In the design of risk management instruments, it is to improved asset portfolios, for example, invest- important to account for the unique context pre- ments in improved technology, training programs, sented in different situations. Risk management in- and empowerment through social and political net- struments must be tailored to specific constraints and works, can foster a virtuous cycle of sustainable objectives within the country, community, and growth. Asset accumulation and changes in liveli- household context. hood strategies are thus important for sustained In considering the potential applications of index improvements in household well-being. insurance in developing countries, it is important to Improved management of rural risk is critical to remember that index insurance is not necessarily ap- achieve rural growth and reduce poverty. It is critical, plicable or replicable for every situation. Nor should it however, to move beyond a narrow risk management be inferred that index insurance is a substitute for focus to a more holistic rural development approach other risk management strategies. Index insurance that focuses attention on building, enhancing, main- can, however, provide a starting point and, ideally, a taining, and protecting household assets. The develop- springboard for the development of a variety of risk ment of new rural risk management instruments offers management mechanisms. Note: A more detailed discussion of these issues can be found in "Looking at Rural Risk Management Using an Asset-Based Approach," a background paper for this report by Paul Siegel. In particular, the reader is directed to Figure 1, which depicts the relationships among assets, context, behavior, and outcomes. Source: Siegel 2005. utilities (World Bank 2001). These arrangements, Typical ex post informal income-smoothing however, while effective for counterbalancing the mechanisms include the sale of assets, such as land consequences of events affecting only some mem- or livestock (Rosenzweig and Wolpin 1993), or the bers of the community, do not work well in cases reallocation of labor resources to off-farm labor of covariate income shocks (Hazell 1992). activities. Gadgil, et al. (2002), argue that southern 8 Managing Agricultural Production Risk Table 2.1 Risk Management Strategies in Agriculture Formal Mechanisms Informal Mechanisms Market Based Publicly Provided S On-farm Avoiding exposure to risk Agricultural extension EI G Crop diversification and Pest management systems E T intercropping Infrastructures (roads, dams, A R Plot diversification irrigation systems) T S Diversification of income E T source N A Buffer stock accumulation X of crops or liquid assets E Adoption of advanced cropping techniques (fertilization, irrigation, resistant varieties) Sharing risk Crop sharing Contract marketing with others Informal risk pool and futures contracts Insurance S Coping with Sale of assets Credit Social assistance EI Reallocation of labor Social funds T G shocks S E T Mutual aid Cash transfer O A P R X T E S Source: Anderson 2001; Townsend 2005; World Bank 2001. Indian farmers who expect poor monsoon rains tions. Examples of such extra-regional risk sharing can quickly shift from 100 percent on-farm labor systems are found in the literature, including, credit activities to mainly off-farm activities. Fafchamps andtransfersbetweendistantrelatives(Rosenzweig, (1993), in his analysis of rain-fed agriculture among 1988; Miller and Paulson 2000); migration and mar- West African farmers, emphasizes the importance riages (Rosenzweig and Stark 1989); or ethnic net- of building labor flexibility into the production works (Deaton and Grimard 1992). Although these strategy. studies find some degree of risk sharing and thus As reported by Townsend (2005), in analyzing of insurance against weather, use of such systems the cost of risk on ex ante agricultural production is not so widespread as to cover all households, nor strategies, Rosenzweig and Binswanger (1993), do they come even close to providing a fully efficient Morduch (1995), and Kurosaki and Fafchamps insurance mechanism. Most households are there- (2002) all find considerable efficiency losses associ- fore still left with no insurance against correlated ated with risk mitigation, typically due to lack of risks, the main source of which is weather. specialization--in other words, farmers trade off income variability with profitability. FORMAL MECHANISMS Theneedtosmoothconsumptionnotonlyagainst idiosyncratic shocks but also against correlated Formal risk management mechanisms can be classi- shockscomesataseriouscostintermsofproduction fiedaspubliclyprovidedormarketbased(Table2.1). efficiency and reduced profits, thus lowering the Government action plays an important role in agri- overall level of household consumption. A major cultural risk management, both ex ante and ex post. consideration for innovation would be to shift cor- Ex ante education and services provided by agri- related risk from rural households (Skees 2003). cultural extension help familiarize producers with One obvious solution would be for rural households the consequences of risk and help them adopt to share risk with households or institutions from strategies to deal with it. Governments also reduce areas largely uncorrelated with the local risk condi- the impacts of risk by developing relevant infra- Risk and Risk Management in Agriculture 9 structure and by adopting social schemes and cash portant price discovery devices and market trend transfers for relief after shocks have occurred.7 indicators. As mentioned in the section on informal mech- For agricultural producers in developing coun- anisms, production and market risks are probably tries, access to futures and options contracts is prob- those with the largest impact on agricultural pro- ably the exception rather than the rule. Futures and ducers. Various market-based risk management options markets in developed countries represent solutions have been developed to address these important price discovery references for inter- sources of risk. national commodity markets, however, and indirect access to these exchange-traded instruments may be granted through the intermediation of collective Price Risk Management action by producer groups such as farmer cooper- One way producers have traditionally managed atives or national authorities.8 While an important price variability is by entering into preharvest agree- reality for some commodities, futures and options ments that set a specific price for future delivery. are not available for all agricultural products. These arrangements, known as forward contracts, allow producers to lock in a certain price, thus re- Production/Weather Risk Management ducing risk but also foregoing the possible benefits of positive price deviations. In specific markets, Insurance is another formal mechanism used in and for specific products, these arrangements have many countries to share production risks. Insurance evolved into futures contracts, traded on regulated does not as efficiently manage production risk, exchanges on the basis of specific trading rules and however, as derivative markets do price risks. Price for specific standardized products. This reduces risk is highly spatially correlated and, as illus- some of the risks associated with forward contract- trated in Figure 2.1, futures and options are ap- ing (for example, default). A further evolution in propriate instruments for dealing with spatially hedging opportunities for agricultural producers correlated risks. In contrast, insurance is an ap- has been the development of price options, a price propriate risk management solution for indepen- guarantee that allows producers to benefit from a dent risks. Agricultural production risks typically floor price while also allowing them to take advan- lack sufficient spatial correlation to be effectively tage of positive price changes. With price options, hedged using only exchange-traded futures or op- agents pay a premium to purchase a contract that tions instruments. At the same time, agricultural gives them the right (but not the obligation) to sell production risks are generally not perfectly spatially futures contracts at a specified price. Price options independent; therefore, insurance markets do not for commodities are regularly traded on exchanges, work at their best. Skees and Barnett (1999) refer to but they can also be traded in over-the-counter these risks as "in-between" risks. According to markets. Futures and options contacts can be ef- Ahsan, et al. (1982), "good or bad weather may have fective price risk management tools as well as im- similar effects on all farmers in adjoining areas," Figure 2.1 Independent Versus Correlated Risk Insurance Options and futures markets markets Perfectly Perfectly independent correlated (systemic) Auto, life, Crop Prices, fire yields interest rates Source: Miranda and Glauber 1997. 10 Managing Agricultural Production Risk and, consequently, "the law of large numbers, on that an insurance market will emerge diminishes. which premium and indemnity calculations are When considering an insurance purchase, the con- based, breaks down." In fact, positive spatial corre- sumer may have difficulty determining the value lation in losses limits the risk reduction obtainable of the contract or, more specifically, the probabil- by pooling risks from different geographical areas. ity and magnitude of loss relative to the premium This increases the variance in indemnities paid by (Kunreuther and Pauly 2001). Many decision mak- insurers. In general, the more the losses are posi- ers tend to underestimate their exposure to low- tively correlated, the less efficient traditional insur- frequency, high-consequence losses. Thus, they ance is as a risk-transfer mechanism. For many ideas are unwilling to pay the full costs of an insurance presented in this document, a precondition for suc- product that protects against these losses. Low- cess is a high degree of positive correlation of losses. frequency events, even when severe, are frequently The lack of statistical independence is not the discounted or ignored altogether by producers try- only problem with providing insurance in agricul- ing to determine the value of an insurance contract. ture. Another set of problems relates to asymmetric This happens because the evaluation of probability information, the situation in which the insured has assessments regarding future events is complex more knowledge about his or her own risk profile and often entails high search costs. Many people than does the insurer. Asymmetric information resort to various simplifying heuristics, but proba- causes two problems: adverse selection and moral bility estimates based on these heuristics may dif- hazard. In the case of adverse selection, farmers fer greatly from the true probability distribution have better knowledge than do the insurers about (Schade et al. 2002; Morgan and Henrion 1990). the probability distribution of losses. The farmers Evidence indicates that agricultural producers for- thus occupy the privileged situation of knowing get extreme low-yield events. The general finding whether or not the insurance premium accurately regarding subjective crop-yield distributions is that reflects the risk they face. Consequently, only farm- agricultural producers tend to overestimate the ers bearing greater risks will purchase the cover- mean yield and underestimate the variance (Buzby age, generating an imbalance between indemnities et al. 1994; Pease et al. 1993; Dismukes et al. 1989). paid and premiums collected. Moral hazard simi- On the other side, insurers will typically load larly affects the incentive structure of the relation- premium rates heavily for low-frequency, high- ship between insurer and insured. After entering consequence events where considerable ambiguity the contract, the farmer's incentive to take proper surrounds the actual likelihood of the event (Schade care of the crop diminishes, while the insurer has et al. 2002; Kunreuther et al. 1995). Ambiguity is limited effective means to monitor what may prove especially serious when considering highly skewed hazardous behavior by the farmer. Insurers may probability distributions with long tails, as is typical thus incur greater than anticipated losses. of crop yields. Uncertainty is further compounded Agricultural insurance is often characterized by when the historical data used to estimate probabil- high administrative costs, due, in part, to the risk ity distributions are incomplete or of poor quality, classification and monitoring systems that insurers a very common problem in developing countries. must put in place to forestall asymmetric informa- Small sample size creates large measurement error, tion problems. Other costs include acquiring the especially when the underlying probability distrib- dataneededtoestablishaccuratepremiumratesand ution is heavily skewed. Kunreuther et al. (1993) conducting claims adjustments. As a percentage of demonstrate via experimental economics that when the premium, the smaller the policy, typically, the risk estimates are ambiguous, loads on insurance larger the administrative costs. premiums can be 1.8 times higher than when insur- Spatially correlated risk, moral hazard, adverse ing events with well specified probability and loss selection, and high administrative costs are all im- estimates. portant reasons why agricultural insurance markets Together, these effects create a wedge between may fail. Cognitive failure among potential insur- the prices that farmers are willing to pay for cata- ance purchasers and ambiguity loading on the part strophic agricultural insurance and the prices that of insurance suppliers are other possible causes of insurers are willing to accept. Thus, functioning agricultural insurance market failure.9 private-sector markets may fail to materialize or, If consumers fail to recognize and plan for low- if they do materialize, they may cover only a small frequency, high-consequence events, the likelihood portionoftheoverallriskexposure(Pomareda1986). 3 Approaches to Agricultural Risk in Developed Countries To better understand agricultural risk management markets and government policies to facilitate access to risk management instru- ments, it is worthwhile to analyze critically the experiences of some developed countries. The experiences of the United States, Canada, and Spain are thus described for reference, but it is important to con- sider that these systems may not be replicable in or suitable for most developing countries. In addition, many developed countries have involved market support and income transfer programs that extend well beyond crop insurance. To the extent they are based on farm income, these programs involve levels of protection against severe crop failures. The European community has extensive policies focus- ing on income protection. CROP INSURANCE PROGRAMS IN DEVELOPED COUNTRIES This section presents overviews of agricultural risk management programs in three developed countries: the United States, Canada, and Spain. These countries have been able to implement substantial programs to reduce yield and revenue risk for agricultural produc- ers. While these programs offer a variety of risk management prod- ucts for farmers, the programs require levels of government support unfeasible for most countries. The United States In the United States, multiple peril yield and revenue insurance prod- ucts are offered through the Federal Crop Insurance Program (FCIP), a public/private partnership between the federal government and various private sector insurance companies.10 The program seeks to address both social welfare and economic efficiency objectives. With regard to social welfare, private companies selling federal crop in- surance policies may not refuse to sell to any eligible farmer, regard- less of past loss history. At the same time, the program aims to be actuarially sound. Policies are available for over one hundred commodities but in 2004 just four crops--corn, soybeans, wheat, and cotton--accounted for approximately 79 percent of the US$4 billion in total premiums. Excluding pasture, rangeland, and forage, approximately 72 percent of the national crop acreage is currently insured under the FCIP. 11 12 Managing Agricultural Production Risk About 73 percent of total premiums are for revenue as the standard reinsurance agreement (SRA). The insurance policies, while 25 percent are for yield in- SRA is quite complex, with both quota share rein- surance policies.11 surance and stop losses by state and insurance pool; MostFCIPpoliciestriggerindemnitiesatthefarm however, in essence, it allows the private insurance (or even subfarm) level.12 Yield insurance offers are companies to adversely select against the govern- based on a rolling four-to-ten-year average yield, ment. This is considered necessary since the compa- knownastheactualproductionhistory(APH)yield. nies do not establish premium rates or underwriting The federal government provides farmers with a guidelines but are required to sell policies to all base catastrophic yield insurance policy, free of any eligible farmers. premium costs.13 Farmers may then choose to pur- The federal costs associated with the U.S. pro- chase, at federally subsidized prices, additional gram have four components: insurance coverage beyond the catastrophic level. · Federalpremiumsubsidiesrangefrom100per- This additional coverage, often called "buy-up" cent of total premium for catastrophic (CAT) coverage, may be either yield or revenue insurance. policies to 38 percent of premium for buy-up Farm-level revenue insurance offers are based on policies at the highest coverage levels. Across the product of the APH yield and a price index that all FCIP products and coverage levels, the reflects national price movements for the particular commodity. average premium subsidy in 2004 was 59 per- For some crops and regions, defined along cent of total premiums. county barriers, area yield and/or area revenue · The federal government reimburses adminis- buy-up insurance policies are offered through FCIP. trative and operating expenses for private in- On a per acre insured basis, area-level insurance surance companies that sell and service FCIP products tend to be less expensive than farm-level policies. This reimbursement is approximately insurance products. Thus, in 2004, area yield and 22 percent of total premiums. area revenue policies accounted for 7.4 percent of · The SRA has an embedded federal subsidy total acreage insured but less than 3 percent of total with an expected value of about 14 percent premiums. of total premiums. The federal government also provides a rein- · The program, by law, can be considered ac- surance mechanism that allows insurance compa- tuarially sound at a loss ratio of 1.075. This nies to determine (within certain bounds) which implies an additional federal subsidy of policies they will retain and which they will cede 7.5 percent of total premiums. to the government. This arrangement is referred to On average, the federal government pays approx- imately 70 percent of the total cost for the FCIP. Farmer-paid premiums account for only about Figure 3.1 Crop Insurance Premiums and Indemnities 30 percent of the total cost. While the direct farmer in the United States subsidy varies by coverage level, the United States has consistently passed legislation increasing the subsidy level to farmers for crop and revenue insur- 3.5 ance products. The rate of subsidy is one component Crop Year 3.0 that has influenced the growth in overall premium. Premium subsidy Producer-paid Figure 3.1 clearly shows that the growth in premium sr 2.5 all subsidy is greater than the growth in farmer-paid o d 2.0 premiums. The rate of subsidy increased in 1995 n oilli and 2001. 1.5 b . S. U1.0 Canada14 0.5 In 2003, Canada revised its agricultural risk manage- 0.0 1991 1993 1995 1997 1999 2001 2003 ment programs. The "Business Risk Management" element of the new Agricultural Policy Framework Source: Babcock et al. 2004. (APF)iscomposedoftwomainschemes:Production Insurance and Income Stabilization. Approaches to Agricultural Risk in Developed Countries 13 The Production Insurance (PI) scheme offers gram generates a payment when a producer's producers a variety of multiple peril production or current year production margin falls below that production value loss products similar to many of producer's reference margin, which is based on an those sold in the United States. One major distinc- average of the program's previous five-year mar- tion, however, is that the Canadian program is mar- gins, less the highest and lowest. One important keted, delivered, and serviced entirely and jointly feature of CAIS is that producers must participate by federal and provincial government entities, in the program with their own resources. In partic- although it is the provincial authorities who are ular, a producer is required to open a CAIS account ultimately responsible for insurance provision. This at a participating financial institution and deposit allows provinces some leeway to tailor products to an amount based on the level of protection chosen fit their regions and to offer additional products. (coverage levels range from 70 percent to 100 per- Production insurance plans are offered for over cent of the "reference margin"). Once producers one hundred different crops, and provisions have file their income tax returns, the CAIS program ad- been made to include plans covering livestock losses ministration uses the tax information to calculate as well. Crop insurance plans are available based on the producer's program year production margin. either individual yields (or production value in the If the program year margin has declined below case of certain items, such as stone-fruits) or area the reference margin, some of the funds from the based yields. Unlike the U.S. program, Canadian producers' CAIS accounts will be available for producers are not allowed to separately insure dif- withdrawal. Governments match the producers' ferent parcels but rather must insure together all withdrawals in different proportions for different parcels of a given crop type. This means that low coverage levels. yields on one parcel may be offset by high yields on The total investment by federal and provincial another parcel when determining whether or not governments for the "business risk management" an overall production loss has occurred. Insurance programs is CAN$1.8 billion per year. In 2004, can also be purchased for loss of quality, unseeded approximately CAN$600 million was provided by acreage,replanting,spotloss,andemergency works. governments as insurance premium subsidies. The latter coverage is a loss mitigation benefit meant to encourage producers to take actions that reduce Spain the magnitude of crop damage caused by an in- sured peril. The Spanish agricultural insurance system is Cost sharing between the federal government structured around an established public/private and each province for the entire insurance program partnership. On the public side is the National is to be fixed at 60:40, respectively, by 2006. Federal Agricultural Insurance Agency (ENESA), which subsidies as a percentage of premium costs vary, coordinates the system and manages resources for however, from 60 percent for catastrophic loss subsidizing insurance premiums, and the Insurance policies to 20 percent for low deductible produc- Compensation Agency (Consorcio de Compensación tion coverage. Combined, the federal and provin- de Seguros) that, together with private reinsurers, cial governments cover approximately 66 percent provides reinsurance for the agricultural insurance of program costs, including administrative costs. market. Local governments are involved only to This is roughly equivalent to the percentage of total the extent that they are allowed to augment pre- program costs borne by the federal government mium subsidies offered at the national level. On in the U.S. program. Provincial authorities are the private side, insurance contracts are sold by responsible for the solvency of their insurance port- Agroseguro, a coinsurance pool of companies that folio. In Canada, the federal government competes aggregates all insurance companies active in agri- with private reinsurance firms in offering deficit culture. Farmers, insurers, and institutional rep- financing agreements to provincial authorities. resentatives are all part of a general commission Beginning in 2004, the Canadian Agricultural hosted by ENESA that functions as the managing Income Stabilization (CAIS) scheme replaced and board of the Spanish agricultural insurance system. integrated former income stabilization programs. Similar to programs in the United States and CAIS is based on the producer production margin, Canada, Spain's combined program offers insur- where a margin is "allowable farm income," includ- ance policies covering multiple perils. Policies are ing proceeds from production insurance minus available for crops, livestock, and aquaculture activ- "allowable (direct production) expenses." The pro- ities, with risks being pooled across the country by 14 Managing Agricultural Production Risk Agroseguro. Compared to the United States and facilitate similar income transfers, given the large Canada, however, farmers' associations are more segments of the population often engaged in farm- actively involved in implementation and develop- ing. Nonetheless, since a larger percentage of the ment of agricultural insurance. The government population in developing countries is typically in- has reserves to cover extreme losses, and, as a final volved in agricultural production or related in- resort, the government treasury covers losses that dustries, catastrophic agricultural losses will have occur beyond these reserves. a much greater impact on GDP than may occur in Total premiums for agriculture insurance poli- developed countries. cies purchased reached around US$550 million Policymakers should also carefully consider the ( 490 million) in 2003, of which approximately varying structural characteristics of agriculture in US$225 million ( 200 million) have been provided different countries. In general, farms in developing by the government (Burgaz 2004). The rationale be- countries are significantly smaller than are farms hind subsidizing agricultural insurance is that this in countries like the United States and Canada. For outlay serves as a disincentive for the government traditional crop insurance products, smaller farms to also provide free ad hoc disaster assistance. To typically imply higher administrative costs as a per- reinforce the point, Spanish producers are ineligi- centage of total premiums. A portion of these costs ble for disaster payments for perils for which in- are related to marketing and servicing (loss adjust- surance is offered. For noncovered perils, ad hoc ment) insurance policies. Another portion is related disaster payments are available, but only if the pro- to the lack of farm-level data and cost effective ducer had already purchased agricultural insur- mechanisms for controlling moral hazard. ance for covered perils. Developing countries also have far less access to global crop reinsurance markets than do devel- oped countries. Reinsurance contracts typically WHY THE EXPERIENCE OF involve high transaction costs related to due dili- DEVELOPED COUNTRIES IS gence. Reinsurers must understand every aspect of NOT A GOOD MODEL FOR the specific insurance products being reinsured (for DEVELOPING COUNTRIES example, underwriting, contract design, rate mak- ing, and adverse selection and moral hazard con- For various reasons, developing countries should trols). Some minimum volume of business, or the avoid adopting approaches to risk management prospect for strong future business, must be present similar those adopted in developed countries. to rationalize incurring these largely fixed transac- Clearly, developing countries have more limited tion costs. For a global reinsurer to be willing to fiscal resources than do developed countries. Even enter a market, the enabling environment must fos- more importantly, the opportunity cost of those ter confidence in contract enforcement and institu- limited fiscal resources may be significantly greater tional regulations. An enabling environment is, in than in a developed country. Thus, it is critical for a fact, a prerequisite for effective and efficient insur- developing country to consider carefully how much ance markets, and these components are largely risk management support is appropriate and how missing in developing countries. Setting rules assur- to leverage limited government dollars to spur in- ing that premiums will be collected and that indem- surance markets. In developed countries, govern- nities will be paid is not a trivial undertaking. The ment risk management programs are as much alternative risk management products discussed in about income transfers as they are about risk man- Chapter 5 are structured to overcome many of these agement. Developing countries cannot afford to problems. 4 Innovation in Managing Production Risk Index Insurance15 INDEX INSURANCE ALTERNATIVES16 Given the problems with some traditional crop insurance programs in developed countries, finding new solutions to help mitigate sev- eral aspects of the problems outlined above has become critical. Index insurance products offer some potential in this regard (Skees et al. 1999). These contingent claims contracts are less susceptible to some of the problems that plague multiple-peril farm-level crop insurance products. With index insurance products, payments are based on an independent measure highly correlated with farm-level yield or revenue outcomes. Unlike traditional crop insurance that attempts to measure individual farm yields or revenues, index insurance makes use of variables exogenous to the individual policyholder-- such as area-level yield or some objective weather event or measure such as temperature or rainfall--but have a strong correlation to farm-level losses. For most insurance products, a precondition for insurability is that the loss for each exposure unit be uncorrelated (Rejda 2001). For index insurance, a precondition is that risk be spatially correlated. When yield losses are spatially correlated, index insurance contracts can be an effective alternative to traditional farm-level crop insurance. Index products also facilitate risk transfer into financial markets where investors acquire index contracts as another investment in a diversified portfolio. In fact, index contracts may offer significant diversification benefits, since the returns generally should be un- correlated with returns from traditional debt and equity markets. BASIC CHARACTERISTICS OF AN INDEX The underlying index used for index insurance products must be cor- related with yield or revenue outcomes for farms across a large geo- graphic area. In addition, the index must satisfy a number of additional properties affecting the degree of confidence or trust that market par- ticipants have that the index is believable, reliable, and void of human manipulation; that is, the measurement risk for the index must be low (Ruck 1999). A suitable index required that the random variable mea- sured meet the following criteria: · observable and easily measured; · objective; · transparent; 15 16 Managing Agricultural Production Risk · independently verifiable; Consider a contract being written to protect · reportable in a timely manner (Turvey 2002; against deficient cumulative rainfall during a crop- Ramamurtie 1999); and ping season (for example, see Figure 4.1). The writer · stable and sustainable over time. of the contract may choose to make a fixed payment for every one millimeter of rainfall below the strike. Publicly available measures of weather variables If an individual purchases a contract where the generally satisfy these properties.17 strike is one hundred millimeters of rain and the For weather indexes, the units of measurement limit is fifty millimeters, the amount of payment for should convey meaningful information about the each tick would be a function of how much liabil- state of the weather variable during the contract ity is purchased. There are fifty ticks between the period, and they are often shaped by the needs one hundred millimeter strike and fifty millimeter and conventions of market participants. Indexes limit. Thus, if $50,000 of liability were purchased, are frequently cumulative measures of precipita- the payment for each one millimeter below one tion or temperature during a specified time. In hundred millimeters would be equal to $50,000/ some applications, average precipitation or tem- (100 - 50), or $1,000. perature measures are used instead of cumulative Once the tick and the payment for each tick are measures. known, the indemnity payments are easy to calcu- New innovations in technology, including the late. A realized rainfall of ninety millimeters, for ex- availability of low-cost weather monitoring sta- ample, results in ten payment ticks of $1,000 each, tions that can be placed in many locations and for an indemnity payment of $10,000. Figure 4.1 sophisticated satellite imagery, will expand the maps the payout structure for a hypothetical $50,000 number of areas in which weather variables can rainfall contract with a strike of one hundred mil- be measured as well as of the types of measurable limeters and a limit of fifty millimeters. variables. Measurement redundancy and auto- In developed countries, index contracts that pro- mated instrument calibration further increase the tect against unfavorable weather events are now credibility of an index. sufficiently well developed that some standardized The terminology used to describe features of contracts are traded in exchange markets. These indexinsurancecontractsresemblesthatusedforfu- exchange-traded contracts are used primarily by tures and options contracts rather than for other in- firms in the energy sector, although the range of surance contracts. Rather than referring to the point weather phenomena that might potentially be in- at which payments begin as a trigger, for example, sured using index contracts appears to be limited index contracts typically refer to it as a strike. They only by imagination and the ability to parameterize also pay in increments called ticks. the event. A few examples include excess or defi- cient precipitation during different times of the year, insufficient or damaging wind, tropical weather events such as typhoons, various measures of air Figure 4.1 Payout Structure for a Hypothetical Rainfall Contract temperature, measures of sea surface temperature, the El Niño southern oscillation (ENSO) tied to El NiñoandLaNiña,andevencelestialweather events $60,000 such as disruptive geomagnetic radiation from solar $50,000 t flare activity. Contracts are also designed for com- n e binations of weather events, such as snow and tem- m $40,000 y a p perature (Dischel 2001; Ruck 1999). The potential yti $30,000 for the use of index insurance products in agricul- n m ture is significant (Skees 2001). e $20,000 d nI A major challenge in designing an index insur- $10,000 ance product is minimizing basis risk. Basis risk $0 refers to the potential mismatch between index- 0 20 40 60 80 100 120 triggered payouts and actual losses. It occurs when Rainfall in mm an insured has a loss and does not receive an in- Source: Skees 2003. surance payment sufficient to cover the loss (minus any deductible) or when an insured has a loss and receives a payment that exceeds the amount of loss. Innovation in Managing Production Risk 17 Since index-insurance indemnities are triggered would never previously have considered. New risk by exogenous random variables, such as area yields management opportunities can develop if rele- or weather events, an index-insurance policyholder vant, reliable, and trustworthy indexes can be con- can experience a yield or revenue loss and not re- structed. A detailed technical overview of index ceive an indemnity. The policyholder may also ex- insurance is presented in Appendix 1. Key advan- perience no yield or revenue loss and still receive tages and challenges are summarized in Table 4.1. an indemnity. The effectiveness of index insurance as a risk management tool depends on how posi- THE TRADE-OFF BETWEEN BASIS tively correlated farm yield losses are with the underlying index. In general, the more homoge- RISK AND TRANSACTION COSTS neous the area, the lower the basis risk and the more Among the most significant issues for any insur- effective area-yield insurance will be as a farm-level ance product is the question of how much moni- risk management tool. Similarly, the more closely toring and administration is needed to keep moral a given weather index actually represents weather hazard and adverse selection to a minimum. To events on the farm, the more effective the index accomplish this goal, coinsurance and deductibles will be as a farm-level risk management tool.18 are used so that the insured shares the risk and any mistakes in offering too generous coverage are RELATIVE ADVANTAGES mitigated. Considerable information is needed to tailor insurance products and to minimize the basis AND DISADVANTAGES OF risk even for individual insurance contracts. In- INDEX INSURANCE creased information gathering and monitoring involve higher transaction costs, which convert Index insurance can sometimes offer superior risk directly into the higher premiums needed to cover protection compared to traditional farm-level, them. Index insurance significantly reduces these multiple- peril crop insurance. Deductibles, co- transaction costs and can be written with lower payments, or other partial payments for loss are deductibles and without introducing coinsurance. commonly used by farm-level, multiple-peril in- When farm yields are highly correlated with the surance providers to mitigate asymmetric informa- index being used to provide insurance, offering tion problems such as adverse selection and moral higher levels of protection can result in risk trans- hazard. Asymmetric information problems are fer superior even to individual multiple-peril crop much lower with index insurance because, first, a insurance (Barnett et al. 2005). producer has little more information than the in- The direct trade-off between basis risk and trans- surer regarding the index value, and second, indi- action costs has implications for achieving sustain- vidual producers are generally unable to influence able product designs and for outlining the role of the index value. This characteristic of index insur- governments and markets. Chapter 5 introduces ance means that there is less need for deductibles the idea of layering risk. These concepts also greatly and copayments. Similarly, unlike traditional insur- depend on understanding the trade-off between ance, few restrictions need be placed on the amount basis risk and transaction costs. At every level of of coverage an individual purchases. As long as risk transfer, someone must accept a certain degree the individual farmer cannot influence the realized of basis risk if the products are to be both sustain- value of the index, liability need not be restricted. able and affordable. In short, extremely high trans- An exception occurs when governments offer pre- action costs must be paid for. The social cost of mium subsidies as a percentage of total premiums. having products with some basis risk may be signif- In this case, the government may want to restrict icantly lower than the social cost associated with liability (and thus, premium) to limit the amount the high transaction cost entailed in attempting to of subsidy paid to a given policyholder. design products that have no basis risk. As more sophisticated systems (such as satellite imagery) are developed to measure events causing widespread losses, indexing major events should WHERE INDEX INSURANCE become straightforward and quite acceptable to IS INAPPROPRIATE international capital markets. Under these condi- tions, traditional reinsurers and primary providers Index insurance contracts will not work well for may begin offering insurance in countries they all agricultural producers. Many agricultural com- 18 Managing Agricultural Production Risk Table 4.1 Advantages and Disadvantages of Index Insurance Advantages Challenges Less moral hazard Basis risk The indemnity does not depend on the individual producer's Without sufficient correlation between the index and actual realized yield. losses, index insurance is not an effective risk management tool. This is mitigated by self-insurance of smaller basis risk Less adverse selection by the farmer; supplemental products underwritten by private The indemnity is based on widely available information, so insurers; blending index insurance and rural finance; and there are few informational asymmetries to be exploited. offering coverage only for extreme events. Lower administrative costs Precise actuarial modeling Underwriting and inspections of individual farms are not Insurers must understand the statistical properties of the required. underlying index. Standardized and transparent structure Education Contracts can be uniformly structured. Users must be able to assess whether index insurance will provide effective risk management. Availability and negotiability Standardized and transparent, the contracts may be traded Market size in secondary markets. The market is still in its infancy in developing countries and has some start-up costs. Reinsurance function Index insurance can be used to transfer the risk of wide- Weather cycles spread correlated agricultural production losses more easily. Actuarial soundness of the premium could be undermined by weather cycles that change the probability of the insured Versatility events, such as El Niño, for example. Index contracts can be easily bundled with other financial services, facilitating basis risk management. Microclimates These production conditions make rainfall or area-yield index based contracts difficult for frequent and localized events. Forecasts Asymmetric information about the likelihood of an event in the near future creates the potential for intertemporal adverse selection. Source: Authors. modities are grown in microclimates. Coffee grows Overfitting the data is another concern with on certain mountainsides in various continents and index insurance. If one has a limited amount of countries, for example, and fruits such as apples and crop yield data, fitting the statistical relationship cherries also commonly grow in areas with very between the index and that limited data can become large differences in weather patterns within only a problematic. Small sample sizes and fitting regres- few miles. In highly spatially heterogeneous pro- sions within the sample can lead to complex contract duction areas, basis risk will likely be so high as to designs that may or may not be effective hedging make index insurance problematic. Under these mechanisms for individual farmers. Standard pro- conditions, index insurance will work only if it is cedures that assume linear relationships between highly localized19 and/or can be written to protect the index and realized farm-level losses may be only against the most extreme loss events. Even in inappropriate. While scientists are tempted to fit these cases, it may be critical to tie index insurance complex relationships to crop patterns, interviews to lending, since loans are one method of mitigat- with farmers may reveal more about the types of ing basis risk. weather events of most concern. When designing Innovation in Managing Production Risk 19 a weather index contract, one may be tempted to extended the sales closing date and sold even more focus on the relationship between weather events rainfall insurance contracts. The company experi- and a single crop. When it fails to rain for an ex- enced very high losses and was unable to meet the tended period of time, however, many crops will full commitment of the contracts. Rainfall insurance be adversely affected. Likewise, when it rains for for agriculture in the United States suffered a sig- an extended period of time, resulting in significant nificant setback. The lesson learned is that when cloud cover during critical photosynthesis periods, writing insurance based on weather events, it is a number of crops may suffer. crucial to be diligent in following and understand- Finally, when designing index insurance con- ing weather forecasts and any relevant informa- tracts, significant care must be taken to assure that tion available to farmers. Farmers have a vested the insured has no better information about the like- interest in understanding the weather and climate. lihood and magnitude of loss than does the insurer. Insurance providers who venture into weather Farmers' weather forecasts are quite often highly ac- index insurance must know at least as much as curate. Potato farmers in Peru, using celestial obser- farmers do about conditional weather forecasts. If vations and other indicators in nature, are able to not, intertemporal adverse selection will render the forecast El Niño at least as well as many climate ex- indexinsuranceproductunsustainable.Theseissues perts (Orlove et al. 2002). In 1988, an insurer offered can be addressed; typically, the sales closing date drought insurance in the U.S. Midwest. As the sales must be established in advance of any potential fore- closing date neared, the company noted that farm- casting information that would change the proba- ers were significantly increasing their purchases of bility of a loss beyond the norm. But beyond simply these contracts. Rather than recognize that these setting a sales closing date, the insurance provider farmershadalreadymadeaconditionalforecastthat must have the discipline and the systems in place the summer was going to be very dry, the company to ensure that no policies are sold beyond that date. 5 New Approaches to Agricultural Risk Management in Developing Countries ROLE OF GOVERNMENT Should the lack of effective private-sector agricultural insurance mar- kets in developing countries be addressed through government inter- vention? High transactions costs preclude emergence of many markets, but this does not necessarily justify government intervention. In the case of high-frequency, low-consequence losses, govern- ment intervention is likely to distort incentives and create rent- seeking opportunities, possibly to an extent that actually reduces net social welfare. Farmers can employ other risk management mech- anisms to cover these losses. In fact, insurance products for high- frequency, low-consequence losses are seldom offered because the transaction costs associated with loss adjustment renders the insur- ance cost prohibitive for most potential purchasers. Governments may have no inherent advantage over markets in trying to facilitate the provision of individual farm-level yield or rev- enue insurance products. The private sector typically does not provide these insurance products in part because of information asymmetries that cause moral hazard and adverse selection problems (Miranda and Glauber 1997); it is difficult to see how a government provider would have any advantage in addressing this problem. In the case of low-frequency, high-consequence loss events, how- ever, government intervention may be justified. As explained in the section on production/weather risk management, research suggests that many decision makers tend to underestimate their exposure to low-frequency, high-consequence losses, a tendency reinforced when the decision maker believes the government will provide assistance in the event of a disaster. Thus, producers thinking in this way will be un- willing to pay the full costs of insurance products that protect against these losses. Those who do buy insurance against low-frequency, high- consequence losses often cancel the policy if they do not receive an indemnity for an extended period. Thus, it seems that to be success- ful agricultural insurance products must be constructed so that they make indemnity payments with reasonable frequency, for example, once every seven or ten years. On the supply side, insurers will typically load premium rates heavily for low-frequency, high-consequence loss events where con- siderable ambiguity surrounds the actual likelihood of the event. Together, these effects create a gap between the prices farmers will pay for catastrophic agricultural insurance and the prices insurers will accept. Thus functioning private sector markets fail to materialize, or, 21 22 Managing Agricultural Production Risk if they do materialize, they cover only a small por- sponsible for some losses might be an incen- tion of the overall risk exposure. This type of mar- tive for putting in place appropriate hazard ket failure is commonly cited as justification for management and mitigation measures. government interventions to facilitate provision of 3. A government's financial involvement in rein- products or services not otherwise provided (or surance may reduce political pressure to pro- provided in sufficient quantity) by private markets. vide distorting and often capricious ad hoc Subsidies for catastrophic reinsurance (see Box disaster relief. 5.1) are a type of government intervention that 4. Governments can potentially provide reinsur- can facilitate the provision of insurance for low- ance more economically than can commer- frequency, high-consequence loss events. Hardaker, cial reinsurers. A government's advantages, et al. (2004), provide the following arguments for including its deep credit capacity and unique such an approach: position as the country's largest entity, enable it to spread risks more broadly. 1. Governments already provide disaster re- lief; providing assistance through reinsur- If governments are to intervene in agricultural in- ance might be more efficient. surance markets, the social benefits of reducing the 2. The financial involvement of a government inefficiencies brought on by risk must outweigh the may address a moral hazard problem in its social cost of making agricultural insurance work. behavior: many catastrophes can either be This chapter presents a framework for government prevented or magnified by government poli- agricultural risk policy formulation that focuses cies or lack thereof. Government financial re- on policy objectives, constraints on government Box 5.1 Reinsurance Reinsurance is insurance for insurers. Just like insurance, Reinsurers seek to operate across boundaries in reinsurance is "fundamentally the promise to pay order to build globally diversified portfolios. More possible future claims against a premium today." than 250 reinsurers in 50 countries wrote annual re- Insurers often hold undiversifiable or extreme risk in insurance premiums of approximately US$176 billion their portfolios, and since they do not wish to retain in 2003.a Nonlife reinsurance premiums accounted all of it, they transfer some risk to reinsurance com- for US$146 billion, or about 14 percent, of the global panies, paying the reinsurers a premium to do so. nonlife primary insurance industry. Only US$25 billion Reinsurers also advise insurers on product development of these premiums are written outside North America and more complex risk-taking. and Western Europe.b The ten largest reinsurers write Reinsurance agreements can be proportional or about 54 percent of reinsurance premiums, and the nonproportional. With proportional agreements, two giants in the business, Munich RE and Swiss RE, insurers and reinsurers divide premiums and losses in write around US$49 billion of reinsurance a contractually defined proportion; with nonpropor- premiums.c tional agreements, the insurer usually pays all losses Securitization, an alternative to traditional reinsur- up to a defined amount and the reinsurer indemnifies ance, transfers catastrophic risks to capital markets in for losses above that limit. Quota-share and surplus the form of financial securities. Securitization has reinsurance are examples of proportional reinsurance been used for exposure to natural catastrophes, such agreements. Excess-of-loss and stop-loss agreements as earthquakes and hurricanes. are examples of nonproportional reinsurance. Notes: a. Standard & Poor's Global Reinsurance Highlights, 2004 Edition. b. Latin America: $US4.7 billion; Asia: $US13.8 billion; rest of the world: $US6.7 billion. For comparison, the World Bank disburses approximately $US0.5 billion per year in emergency assistance grants and loans to developing countries. c. This premium volume includes life and health reinsurance premiums. Source: Swiss Re 2004. New Approaches to Agricultural Risk Management in Developing Countries 23 action, risk principles, and potential policy instru- ing the volatility of their income and the likelihood ments (Figure 5.1). The framework is then used to that a risk event will wipe out hard-won asset gains. consider alternative models for government inter- A precondition for achieving sustainable growth vention in agricultural insurance markets. and poverty reduction is an ex ante system for dis- aster risk management. Disaster risk management covers severe and very infrequent events affecting POLICY OBJECTIVES mostly the poor, because the poor are more vul- Governments that seek to spur growth and eradi- nerable and tend to live in marginal and more risk- cate poverty almost inevitably mix economic poli- exposed areas. Susceptibility to and the experience cies meant to enhance efficiency and growth with of major natural disasters tend to trap people in social policies meant to address poverty and vul- poverty, due to the lack of efficient risk manage- nerability. Governments also often pursue equity ment at the household level.22 Government disaster or income redistribution objectives. Thus, govern- risk policies often entail some form of monetary ment policies related to agriculture and rural areas compensation for victims of natural disaster. The tend to pursue the following objectives: · Growth. Economic growth in rural areas--in particular higher agricul- tural yields and value-added process- Figure 5.1 Framework for Governmental Agricultural Risk Management Policy Formulation ing as well as development of off-farm activities--is perceived to be the best way out of poverty in the medium term. While better incentives for market players and an enabling in- Objectives frastructure are key drivers, better I Agricultural and rural economic growth management of agricultural produc- I Poverty reduction tion risk is also critical for growth, as it enhances access to credit and adop- Constraints tion of new technologies.20 · Reduction of poverty and vulnerability I Underdeveloped financial sector in rural areas. To achieve social and I Disaster risk equity goals, governments directly I Agricultural sector dominated by small farms intervene in a targeted manner, be- I Government fiscal limitations cause free markets do not necessarily alleviate poverty for those in society I Underdeveloped regulatory framework who cannot effectively participate in them. Safety nets provide one tool for Principles such government intervention.21 I Segment independent versus correlated risk Given limited resources in developing I Minimize rent seeking that creates market distortions countries and the existence of other sectors I Diversification of risk -- risk management -- risk layering requiring government attention, these objectives are typically pursued within an I Risk transfer cost optimization -- reduce transaction costs environment of binding fiscal constraints. The two objectives target different seg- Policy Instruments ments of the rural population and differ- ent risk profiles. Growth objectives focus I Mechanisms for transferring catastrophic risk layers on increasing profitability so that less poor I Limited government subsidies farmers can continue adopting production I Contingent funding for disaster relief and enhanced social safety nets technologies even when high-frequency, low-consequence loss events occur. Poverty Source: Authors. reduction policies seek to increase the aver- age income of poor farmers, thus decreas- 24 Managing Agricultural Production Risk challenge is to deliver timely and predictable aid in always robust across different projects, are required disaster situations. This requires ex ante planning to quantify risk management benefits. Still, it is rather than just ex post disaster responses. This also worthwhile to compare the net benefits of govern- implies efforts to forestall political demands for ment risk management programs with the net ben- ex post, ad hoc government disaster assistance. efits from other projects, if only to get a sense of the Indeed, a credible and reliable disaster risk man- orders of magnitude involved. agement system can put farmers and countries on a higher growth path by making people more com- Fiscal Constraints fortable with taking calculated and protected risks. Naturally growth and poverty-reduction objec- Government expenses for agricultural insurance tives overlap, but this makes it even more important programs can be quite high, a reality often masked to identify clear objectives and to design effective by how the actuarial performance is presented. and cost-efficient ways to achieve them. Mixing Governments typically report loss ratios, or cost objectives can lead to suboptimal outcomes. Many to premium ratios, as indemnities paid divided by government-facilitated crop insurance programs, total premiums collected. This method presents two for example, attempt to accomplish social welfare problems: first, due to government premium sub- and economic efficiency objectives simultaneously. sidies, farmers pay only a fraction of the total pre- mium; second, governments typically absorb most administrative and operating costs. When calculat- CONSTRAINTS IN AGRICULTURAL ing loss ratios for private sector insurance products, RISK MANAGEMENT administrative costs are included in the numerator. When considering only indemnity relative to pre- When making decisions about agricultural risk miums (without noting that significant portions of management programs, policymakers face a num- premiums are paid by the public sector), both the ber of constraints. They must consider whether the U.S. and Canadian crop insurance programs have, benefits of such programs outweigh the costs and in recent years, reported loss ratios around 1.0. if the benefits from putting resources into risk man- These loss ratios are then cited as evidence that the agement programs are greater than the benefits programs are actuarially sound. But when admin- of using these resources for other social needs. istrative and operating costs are added to the nu- Governments must construct risk management merator and government premium subsidies are programs that minimize distortions in resource subtracted from the denominator, so that the loss allocation and reduce opportunities for rent-seeking ratio is equivalent to the standard used for pri- behavior. They must take into consideration the vate sector insurance products, crop insurance status and development of financial and insurance loss ratios are about 3.6 for the United States and institutions within the country, any regulatory con- 2.9 for Canada.23 Hazell (1992) estimates similar straints on the operations of those institutions, and ratios for a number of government-based crop in- the infrastructure for enforcing contracts. Finally, surance programs. His estimates for programs in policymakers must consider the dichotomy, present the Philippines, Japan, and Brazil, for example, in many countries, between smallholder farms and show loss ratios (as defined in the private sector) large farms producing for export markets. exceeding 4.0. Policymakers often suggest agricultural insur- Cost-Benefit Analyses of Agricultural Risk ance programs as alternatives to free ex post dis- Management Projects aster assistance. In principle, insurance programs have many advantages over ex post disaster as- Traditional economic analyses of projects (or other sistance. Disaster assistance programs, it is often sector interventions) weigh social benefits against argued, for example, can generate perverse in- social costs, usually in monetary terms. In theory, centives that increase the magnitude of losses in this procedure should make it possible to compare subsequent disaster events (Barnett 1999; Rossi et al. the net benefits from these projects with the net 1982). But, in practice, agricultural insurance pro- benefit of a government risk management program. grams have often evolved into alternate vehicles for Conducting such a comparison is not a trivial exer- transferring wealth from the public sector to agri- cise, however, because numerous assumptions, not cultural producers.Furthermore,notmuchevidence New Approaches to Agricultural Risk Management in Developing Countries 25 indicates that agricultural insurance programs have New products will be required if agricultural been successful in forestalling free ex post govern- insurance is to take root in countries with under- ment disaster assistance. In the United States, for developed traditional insurance sectors. Insurance example, more and more costly crop insurance pro- productsbasedonanindexrecognizedandaccepted grams have coexisted with disaster payments for byinternationalreinsurers,forexample,canprovide well over twenty years (Glauber 2004). opportunities to bypass in-country insurance capac- ity constraints. If the reinsurer accepts the index data and settlement procedures, the insurer's capi- Operational Constraints: Minimize tal becomes somewhat less relevant than for tradi- Distortions/Rent-Seeking Opportunities tional lines of insurance; this is because the reinsurer Governments should only invest public resources isnotreallyacceptingtheinsurer'sunderwritingrisk in developing agricultural insurance if the social but only the risk inherent in the index. Experience costs of the inefficiencies resulting from the lack of with reinsurance for weather index contracts re- such insurance products outweigh the social costs veals that reinsurers may even be willing to take 100 percent of the risk. For operational and regu- of government intervention. Social costs include latory reasons, however, international reinsurers not only the opportunity costs of public resources prefer to deal with professionally-run companies used to create and maintain the agricultural insur- to source the risk. ance products but also any resource allocation dis- tortions that result when farmers and rural decision makers respond to the incentives created by the Structure of Agricultural Sectors insurance products. This can include rent-seeking Agricultural dominated by smallholders imposes and regressive effects that benefit mostly large com- clear constraints on the large scale roll-out of sophis- mercial farmers. ticated crop insurance programs or, indeed, of any agricultural risk management scheme. Farmers Contract Enforcement with one hectare of land or less will never offer an attractive marketing target for insurance compa- Contract enforcement is critical to achieving effec- nies. The challenge is to identify suitable aggre- tive and sustainable risk management programs. gators of risk, such as microfinance institutions, It is very difficult to develop insurance contracts if banks or cooperatives, or even local authorities who the legal and regulatory environment does not exist can enroll farmers in group insurance programs. for contract enforcement. Purchasers will lose trust Agricultural sectors need to be segmented, with in the program if indemnity payments are not made distribution channels tailor-made to specific needs on a timely basis or if they are frequently tied up in and local customs.25 lengthy legal procedures.24 Likewise, insurers will lose trust in the program if they are forced to pay Regulatory Constraints indemnities for losses that the contract was not in- tended to cover. Agricultural risk transfer involves financial con- tracts that are regulated according to prudential principles. Insurance companies must organize the Level of Financial Sector Development financing to pay for the possibility of the worst case scenario. This constrains the type and sophistica- Complex agricultural insurance programs are un- tion of contracts, which may also be constrained by likely to be sustainable unless they are accompanied limitations in the regulator's ability to understand by adequate insurance capital and expertise. In and supervise new products. developing countries, insurance sectors are often underdeveloped and concentrated in very few lines of business, for example, automobile, property, and RISK PRINCIPLES casualty insurance. Insurance companies in devel- Layering and the Role of Index Insurance oping countries also tend to be based in urban areas and to shy away from doing business in rural areas, Segmenting risk into different "layers" is a key risk where the insurance market is characterized by management principle. Consider, for example, high transaction costs and small policies. Figure 5.2, which shows the probability distribution 26 Managing Agricultural Production Risk Figure 5.2 Average April to October Rainfall for Thirteen Malawi Weather Stations X 505 X 1776 1.0% 99.0% Limit Strike 0 500 700 1000 1500 2000 2500 April-October rainfall (mm) Source: Authors. for average April to October rainfall at thirteen ADDRESSING THE MARKET weather stations in Malawi.26 Suppose that farmers INSURANCE RISK LAYER start incurring production losses whenever rainfall is less than one thousand millimeters. The domain Referring again to Figure 5.2, suppose that an insur- of losses might be segregated into three risk layers, ance provider writes a rainfall index insurance con- with different entities holding each layer: tract with a strike of seven hundred millimeters and a limit of five hundred millimeters. Limits are com- · For rainfall in excess of seven hundred mil- monly used by weather index insurance writers to limeters, farmers would retain the loss risk, avoid open-ended exposure to catastrophic weather either individually or with financial service events. The insured would select the amount of providers: the risk retention layer. insurance (the liability) and the payment per tick · For rainfall between five hundred and seven would be calculated using this formula. hundred millimeters, the risk would be trans- ferred to an insurance company via a weather Liability index insurance product: the market insur- Payment Per Tick = Limit - Strike ance layer. · For rainfall levels below five hundred milli- Assume that a farmer has a crop with an expected meters, the risk in this example would not be value of $15,000. At only five hundred millimeters insured due to cognitive failure and ambigu- of rainfall, the farmer is estimated to lose two-thirds ity loading: the market failure layer.27 of the value of the crop. Thus, the farmer purchases Farmers would absorb losses in the risk retention $10,000 of liability, with a payment for each tick layer using self-insurance strategies such as those (each millimeter of rainfall) of fifty ($10,000 divided described in Chapter 2. Strategies for effectively by (700 - 500)). If the realized value for the rainfall transferring the other risk layers are described index is six hundred millimeters, for example, the below. indemnity will be $5,000 ((700 - 600) × $50).28 New Approaches to Agricultural Risk Management in Developing Countries 27 The limit of five hundred millimeters caps the would be reinsured. If there are opportunities to di- insurance provider's loss exposure on the index versify risks within the pool, however, this strategy insurance product. Without the limit, the contract could lead to lower reinsurance premiums relative would be extremely expensive, since it would pro- to either of the other two strategies, since the risk of tect against losses in the extreme lower tail of the the overall pool (rather than each individual con- probability distribution. Buyers would exhibit cog- tract) would be reinsured. The first strategy does nitive failure regarding the probability of events not involve the government in the transfer of risk. with less than five hundred millimeters of rainfall, The other two strategies may involve government, while insurance providers would load the premium either in facilitating risk transfer (for the second for ambiguity regarding these same events. Thus, strategy) or in pooling risk and facilitating risk even if insurance was available to protect against transfer (for the third strategy). rainfall events of less than five hundred millimeters, few transactions would be likely, since the premium Pooling of Risk would exceed most buyer's willingness to pay. The third risk transfer strategy identified above in- Spatial Correlation of Risk volves pooling risks within the country or region. Risk pooling is based on the statistical law of large Weather events that cause agricultural losses are numbers, which states that the more uncorrelated often highly spatially correlated. In the presence of risks are added to a portfolio, the lower the vari- such spatial correlation, index insurance products, ance in the outcomes of the overall portfolio. For an such as the rainfall index insurance described above, insurer, this results in lower capital needs and, can be effective risk transfer mechanisms. Once the therefore, lower capital costs. risk is transferred from the farmer to a local insur- Index-based insurance contracts can be pooled ance provider, however, spatial correlation makes it and transferred in a number of ways. In one method, very difficult for the local insurance provider to gen- the reinsurance contract can be based on a basket erate much risk reduction through pooling. Unless index that is a weighted average of the indexes con- some mechanism exists for transferring the spatially tained in the pool. A risk management program correlated loss risk out of the region or country, local being considered for Malawi would have private in- insurance providers will be reluctant to offer insur- surers sell rainfall-based index insurance contracts ance products, even if those products protect only for various weather stations around the country. against losses in the market insurance layer. The government would purchase reinsurance pro- tection and sell it to the insurers. For reinsurance coverage, the government could use the Malawi Risk Transfer Strategies Maize Production Index (MMPI), a weighted aver- At least three strategies exist for transferring risk age of weather station indexes with each station's from index insurance contracts: (1) direct transfer contribution weighted by the corresponding ex- of contracts into reinsurance markets; (2) packaged pected maize production from that location. The transfer of independent contracts; and (3) pooling more highly spatially correlated the risks on the of risk and subsequent transfer of the pool tail risk. underlying indexes, the better the basket index (See Table 5.1.) Under the first two strategies, no will perform as a reinsurance mechanism (that is, basis risk occurs, insofar as every single contract is the lower the reinsurance basis risk). But, of course, reinsured against payouts that exceed a defined the more highly spatially correlated the risks on the level. Since no pooling occurs prior to the risk trans- underlying indexes, the less advantage there is to fer, however, direct and packaged risk transfer pooling within the country as opposed to simply strategies will likely have higher reinsurance pre- transferring the underlying weather station indexes mium rates than will the transfer of pooled risks, to the reinsurance market using either of the first even if the reinsurer offers portfolio-adjusted pric- two strategies identified above. ing. Under the third strategy of pooling risk prior A pool of index insurance risks can also be trans- to transfer, insurers could be exposed to some basis ferred using traditional stop-loss reinsurance. In this risk, insofar as a pool of indexes does not perfectly case, in exchange for a reinsurance premium, the reflect the payout likelihood of each individual con- reinsurer would simply cover all losses in excess of tract, and only the excess risk of the overall pool a predefined percentage (for example, 110 percent) 28 Managing Agricultural Production Risk Table 5.1 Risk Transfer Strategies Strategy Advantages/Disadvantages Role of Government Direct risk transfer Contracts are transferred directly from No basis risk. Pooling occurs at rein- Government is not involved in facilitat- insurers to reinsurers. surer level. If spatial diversification op- ing risk transfer. portunities exist, reinsurance premium rates will likely be higher than if risks were pooled at insurer level (even if the reinsurer offers portfolio adjusted rein- surance premiums). Reinsurer will need to perform extensive due diligence on index but little due diligence on insurer. Packaged risk transfer Contracts are bundled among compa- Same as above, only may pay lower Either government or an association of nies and transferred to one (syndicate) reinsurance premium rates because insurers can facilitate the bundling and of reinsurers. bundling reduces transactions costs for transfer of contracts to the reinsurance the reinsurer. market. Pooling and transfer Contracts are pooled within the country Some basis risk. If spatial diversifica- Either government or an association of and/or region with only the tail risk of tion opportunities exist, reinsurance insurers can facilitate the risk pooling the pool transferred to reinsurers. premium rates will be lower than with and transfer of pool tail risk to the rein- other strategies. In the case of pool surance market. reinsurance based on traditional stop- loss coveragea transactions costs may be higher, since the reinsurer will need to perform due diligence not only on the index but also on the pool. In case of reinsurance based on index insur- ance, pool due diligence is avoided, but basis risk would be higher.b Notes: a. For the agricultural insurance pool proposed by the Mongolian project of the World Bank, see the case study in Chapter 6. b. See the Agroasemex case in Appendix 2. Source: Authors. of the total premium dollars in the pool. With this This concept can be extended to the pooling of type of reinsurance (and unlike reinsurance based multicountry risks within a region. Weather risk on a basket index), the pool would not be exposed can be retained and managed internally if the areas to basis risk. The transactions costs for the reinsurer under management are significantly diverse in their will be much higher compared to the basket index weather risk characteristics. This immediately sug- based reinsurance, however, since the reinsurer gests that the weather sensitivity of neighboring will need to conduct due diligence on not only the countries must be taken into account when consid- underlying indexes but also the underwriting of ering a country's weather-risk profile and its need the pool. All other things being equal, higher trans- for outside reinsurance. Consider the example of actions costs will lead reinsurers to charge higher the region of the Southern African Development reinsurance premiums. Despite this, if spatial di- Community (SADC; Figure 5.3). Analysis shows versification opportunities are sufficiently high, that, on average, two countries in the region suffer pooling may reduce risk exposure to such an extent a drought each year. The distribution of drought that reinsurance premium costs are reduced. events in SADC is extremely long-tailed, however, New Approaches to Agricultural Risk Management in Developing Countries 29 Figure 5.3 Histogram of Simulated SADC Drought Events X $80m X $350m X $1.1b 69.0% 95.5% 100% 4000 Fund Reinsurance Securitization 3500 )st n e 3000 v e t h g u 2500 or d 7 7 6 2000 5 f o t u o( 1500 y c n e u q 1000 er F 500 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1 1.05 1.1 More SADC drought-risk exposure ($ billion) Source: Hess and Syroka 2005. with the possibility of widespread drought events countries by transferring risk as part of a regional that could potentially devastate the region. strategy rather than by transferring the risk one A SADC pool of rainfall-based index insurance country at a time. The SADC pooling approach contracts could be constructed, with each member above, for example, would reduce insurance costs country being charged an actuarially fair assess- by 22 percent for one of the countries, Malawi, due ment of the risk transferred to the pool. Suppose to risk-pooling effects (Hess and Syroka 2005). the financial impact to the pool of four SADC coun- Managing a pool requires a high degree of under- tries experiencing simultaneous droughts is about writing and actuarial sophistication, however. US$80 million. The pool may wish to transfer the Reinsurers will conduct due diligence and will be risk of losses beyond US$80 million to the inter- very reluctant to write traditional excess of loss national reinsurance market. This could be done reinsurance unless they are convinced that the pool in layers with, for example, one layer of US$80 to is being managed appropriately. 350 million being transferred using reinsurance mechanisms.29 Losses in excess of US$350 million, MARKET FAILURE LAYER as might occur with simultaneous droughts in ten SADC countries, occur with a frequency of about At the catastrophic loss layer represented by market 1 percent. Instruments such as catastrophe (CAT) failure, private decision makers will likely not pur- bonds might be used to transfer this extreme layer. chase adequate insurance due to cognitive failure, CAT bonds allow the transference of very large ambiguity loading of premiums rates, and perhaps, exposures into financial markets and often have expectations of government or donor disaster re- tenures of up to three years. lief. Some form of government intervention may be More efficient means of transferring risk imply required to facilitate adequate transfer of the risk in that costs could be greatly reduced for the member this layer. 30 Managing Agricultural Production Risk POLICY INSTRUMENTS sign Disaster Option for CAT risk (DOC) index rein- surance contracts for catastrophic risks. Returning Risk layering provides an extremely helpful con- to the example in Figure 5.2, a DOC could insure ceptual framework for thinking about government against rainfall less than five hundred millimeters intervention in risk transfer markets. The discus- with a payment per tick of say, $50. Primary insur- sion of the market insurance layer described situa- ers could then offer coverage beyond the earlier tions in which government packaging or pooling of imposed limit of five hundred millimeters and risk could potentially reduce the transaction costs transfer the catastrophic tail risk to the government associated with risk transfer and thus the premi- using the DOC. Even if primary insurers are selling ums paid by end users. This section explores other traditional crop insurance, they could use a DOC possible government interventions, including gov- to transfer part of the catastrophic tail risk in their ernment facilitation of risk transfer in the market portfolio of crop insurance policies.31 DOCs could failure layer, the role of government subsidies in be offered for a variety of strikes and settlement risk transfer markets, and potential uses of index weather stations, as long as the coverage is for cata- insurance instruments to finance government dis- strophic risk layers and can be offset in international aster relief and safety net policies. weather risk markets. The government could even offer other DOC indexes (for example, excess rain- fall or wind speed) to reinsure other lines of insur- Government Disaster Option for CAT Risk: ance, such as property and casualty (see Figure 5.4). A Policy for the Market Failure Layer30 The government would reinsure DOCs in inter- Cognitive failure and ambiguity loading occur pri- national reinsurance or capital markets using any marily with events in the extreme tail of the loss of the three risk transfer strategies described ear- distribution, the area previously termed the market lier.32 Since DOCs would address only extreme failure layer. For this reason, and as a substitute for catastrophic loss events, reinsurance premium rates ad hoc disaster relief payments, governments may would likely contain an ambiguity load. Premiums decide to cofinance risk transfer mechanisms for could be subsidized to offset part of this ambiguity these events. A government, for example, could de- load so that DOC purchasers would pay something Figure 5.4 Government-Sponsored DOC as Risk Transfer Product between National and International Risk Markets DOC Market Risk (market failure insurance retention layer) layer layer layer 0 500 700 1000 1500 2000 2500 April-October rainfall (mm) Source: Authors. New Approaches to Agricultural Risk Management in Developing Countries 31 closer to a pure premium rate.33 DOCs could be risk into international markets. This reduces tailor-made to individual insurers' needs; for exam- the potential for perverse incentives that could ple, DOCs could be based on individual weather encourage excessive risk taking. stations or written as regional weighted average baskets of weather stations. Strikes should be set Subsidies34 so that the DOC covers only infrequent events (for example, events with an expected frequency of Governments frequently subsidize agricultural once every thirty years or less). This is the domain insurance products. These subsidies take a variety of the probability distribution over which poten- of forms. The government may cofinance insur- tial insurance purchasers tend to experience cog- ance purchasing with direct premium subsidies, nitive failure and insurance providers engage in reimburse primary insurers for administrative or ambiguity loading. Primary insurers and ultimately product development costs, or provide reinsur- insured parties would pay a premium for this cata- ance at below market premium rates. Regardless strophic protection, but it would be significantly less of the form, government subsidies are generally than what the market would charge. designed to increase insurance purchasing by low- Those who reinsure DOC contracts will insist on ering the premiums charged to agricultural insur- verifying the credibility of the underlying indexes. ance purchasers. The premium required to transfer the risk to inter- Such subsidies are extremely controversial. They national markets would provide a baseline for set- tend to benefit operators of larger farms more than ting DOC premium rates. those of smaller farms. A wide range of stakeholders The risk-layering approach proposed here would can and will engage in rent seeking once subsidies institutionalize the social role of government in sub- are introduced. Subsidies are costly to maintain and sidizing extreme risk events at the local level. Pre- are subject to close scrutiny regarding social costs mium rates could be subsidized to offset ambiguity versus social benefits. Many times, subsidies are loading. Furthermore, by organizing DOC contracts provided based on the rationalization that agricul- at the local level, victims of isolated severe events tural insurance markets are missing or incomplete, that fail to capture national policymakers' attention without careful consideration of the core reasons could still receive some structured assistance. why such market limitations exist. This document The following list summarizes the major advan- has carefully considered why agricultural insurance tages of offering index-based DOCs: is missing or incomplete in many settings: adverse selection and moral hazard, high transaction costs, · DOC contract provisions established ex ante cognitive failure and ambiguity loading, and expo- allow for better planning than do ad hoc dis- sure to highly correlated loss events. Any govern- aster payments. ment subsidies should be carefully targeted to · DOCs provide a structure that provides more address one of these specific sources of market fail- spatial and temporal equity in government ure. Even then, however, the costs of addressing disaster assistance. that market failure may simply be too high to justify · DOCs facilitate commercial insurance product use of limited government resources to that end. development by providing a means by which The rents resulting from even the most carefully catastrophic risk layers can be effectively trans- targeted subsidies can still be captured by politi- ferred into international markets. cally powerful elites. Government insurance subsi- · DOCs can be subsidized to address the mar- dies may crowd out demand for private sector risk ket failure associated with ambiguity loading transfer instruments. The World Bank supports the and cognitive failure. development of financial institutions that operate · GovernmentscanestimatetheirownDOCsub- profitably on a commercial basis by offering prod- sidy cost exposure based on actuarial estimates ucts and services that meet the needs of a wide range of the risk inherent in the index. Reinsurance of clients, including the poor. Thus, any World Bank coverage adds a market check on the credibil- efforts to facilitate the provision of risk transfer ity of the index and the adequacy of DOC pre- instruments should be based on careful consider- mium rates. ation of whether subsidies or grants can be pro- · While DOCs may be partially subsidized, end vided without distorting or inhibiting the growth users still pay part of the cost to transfer the of private sector financial markets. 32 Managing Agricultural Production Risk Some types of subsidies are likely to be less as much or more than they do risk management distorting than others. Subsidies and grants for objectives. Such subsidies are typically provided on supporting financial intermediaries and financial a percentage basis. This clearly benefits higher risk infrastructure, such as technical assistance and data areas relatively more than lower risk areas. Even systems needed to develop effective index insurance attempts to subsidize to levels that represent a pure products, generally create little distortion. Beyond premium or expected loss basis may favor higher distortions in the markets, legitimate reasons exist risk areas relatively more than lower risk areas, since for supporting infrastructure to improve market ac- in a commercial market, premium rates for higher cess among the rural poor. Finally, some public sup- risk areas would likely contain higher catastrophic port for product development may be justifiable loads. Thus, any attempt to introduce premium because of the free rider problem. Innovative insur- subsidies will likely be distorting. ance products are costly to develop, yet it is difficult In principle, if subsidies are targeted to the mar- to recoup these costs in a competitive market. Any ket failure layer, as described above, market dis- firm can simply copy and compete with the new tortions should be minimal. Given the ambiguity product without the expense of recovering product loading and cognitive failure that occur in this layer, development costs. Unfamiliarity with index insur- carefully targeted subsidies (such as cofinancing ance products can heighten these problems in many of DOCs) may even be welfare enhancing. For the developing countries. market insurance layer, however, subsidies should, Examples of subsidies for financial intermedi- in general, be avoided. Any subsidies in the market aries and infrastructure include: insurance layer should be targeted to reducing · Providing technical assistance to financial uncertainty loads in premium rates. Commercial intermediaries to improve systems that en- insurers will tend to load premium rates based on hance efficiency, such as management infor- the quantity and quality of data used to generate mation systems; pure premium rates. The better (worse) the data · Developing and introducing demand-driven used to generate the pure premium rates, the lower products on a pilot basis; (higher) the premium load. These loads could be · Helping to develop or improve service deliv- offset with cofinancing from donors. Here again, ery mechanisms that enable greater outreach however, donors should be very clear about the into rural areas; level of these subsidies and the intent behind them. · Covering a portion of the cost of establishing new branches in areas lacking financial inter- INDEX INSURANCE AS A SOURCE mediaries to serve the poor; OF CONTINGENT FUNDING · Creating capacity within regulatory and super- visory bodies; FOR GOVERNMENT DISASTER · Supporting the creation of industry assoc- ASSISTANCE AND SAFETY iations; NET PROGRAMS · Developing training institutes and insurance information agencies; In addition to rural economic growth, governments · Supporting data for weather stations or other also want to manage disaster assistance efforts more data to be used to develop effective indexes; effectively and to combat poverty by pursuing social and and equity objectives. Rather than listing the multi- · Providing technical assistance to develop new tude of social policy responses to these objectives, products in an emerging market in develop- this document focuses on the link between funding ing countries. for social policy tools and risk. Specifically, index in- surance is proposed as a source of contingent fund- ing for government disaster assistance and safety Premium Subsidies net programs. While it is common for developed countries to cofinance premiums for farmers with direct pre- Ex Ante Disaster Risk Management mium subsidies, these types of subsidies are par- ticularly problematic. Generally, direct premium Disaster financing has generally focused on pro- subsidies reflect income enhancement objectives viding resources for ex post relief operations to New Approaches to Agricultural Risk Management in Developing Countries 33 cope with shocks rather than on making dedicated increases, necessitating the scale-up of the safety resources available ex ante. This has often meant net. Because the emergency response capabilities providing in-kind emergency resources rather than of existing safety nets are currently limited, how- cash resources. Additional transient needs are met ever, they could be complemented with index- through emergency relief operations that often based disaster insurance. duplicate ongoing interventions: that is, through The scaled-up safety net is limited by two factors: public works and assistance to the vulnerable. More- · Design. Safety nets often focus on addressing over, due to delays in declaring emergencies and chronic poverty rather than transient poverty. mobilizing and then distributing resources, relief Although efforts have been made to scale up often takes significant time to arrive and, indeed, safety nets in time of drought, for example, this can arrive too late. has proved difficult due to delays in mobiliz- Index insurance could be used to provide con- ing financing and organizing activities. tingent ex ante funding for emergency relief opera- · Capacity. Existing safety net operations have in- tions. The relief could be distributed through normal creasingly focused on implementation through emergency channels but would benefit from ex ante local government structures. This is a positive fundingandtimelierprovisionofassistance.Current development, as it will lead to enhanced local funding for emergency activities in food-insecure capacity in the long run, but capacity at the countriesisbasedonaprotractedappeals-basedsys- local level is limited, and scaling up rapidly temthatdeliversfoodaidwellaftercropfailuresand and effectively in times of need requires sub- weathershocks.Bythistime,thepeopleaffectedmay stantial existing capacity. have already had to sell productive assets and/or migrate. Additionally, the support that does come Safety nets could be enhanced using index insur- is not consistent; delivered as a result of appeals to ance. A rainfall index, for example, could be used individual donors subject to their own approval to automatically trigger payments to districts in processes and budget cycles, deliveries are unpre- which the drought-affected population is concen- dictable. The use of index insurance as a means of trated, with the sums insured based on this popu- contingent funding for emergency assistance may lation's likely size. Targeting to the household level mitigate some of the shortcomings of the current would then be used to determine which individu- system. Index insurance provides timely and pre- als in the district should receive payments. Cash fi- dictable payouts during emergencies; by funding nancing would be distributed to districts early (that early relief they preserve livelihoods and to some is, immediately after the weather shock and before extent preempt emergencies (Skees et al. 2005; Goes harvest) to scale up existing safety nets as rainfall and Skees 2003). measures indicate where production shortfalls will occur. This plan distributes cash during the critical coping period, several months earlier than under Safety Nets current emergency arrangements, and before the Safety nets respond to the needs of the poorest and hungry period has set in.35 This mechanism would most vulnerable by providing livelihood support not replace emergency operations but would in- and contributing to immediate food security, often stead provide timely contingent funding to scale up through community-driven public works schemes existing safety net structures. Providing assistance and transfers to vulnerable labor-poor individuals. in the early stages of a disaster event may preempt In times of adverse climatic shocks, the number the need for more extensive, long-term emergency of households in need of assistance dramatically responses. 6 From Theory to Practice Pilot Projects for Agricultural Risk Transfer in Developing Countries36 The previous chapter presented the conceptual foundations for de- veloping risk transfers. This more pragmatic chapter offers concrete examples of the progress made in using index insurance for agricul- tural risk transfer in several developing countries. Index insurance is not a new concept. Chakravati in India was writing about this type of insurance as early as 1920. Sweden and Quebec, Canada, had area- yield insurance programs beginning in the 1950s and 1970s, respec- tively. The United States introduced the Group Risk Plan in 1992 (Skees et al. 1997). The concept of index insurance based on area rain- fall follows many earlier efforts with area-yield insurance. The World Bank and other donors were involved in crop insurance projects in the 1970s and 1980s. These efforts were soon abandoned, however, as many of the problems with introducing multiple-peril crop insurance in developing countries became insurmountable con- straints. Hazell (1992) emphasized the problems with traditional crop insurance and recommended using rainfall insurance. Hazell and Skees (1998) participated in the World Bank's first efforts to return to crop insurance work, undertaken in Nicaragua. Skees and Miranda (1998) followed the work in Nicaragua, and this lead to the develop- ment of the Skees, Hazell, and Miranda (1999) document. In 1999, a team of World Bank professionals and outside consultants obtained a Development Market Place award to work in Morocco, Nicaragua, Ethiopia, and Tunisia. Many of the efforts describe in this chapter fol- low the conceptual development of that project. As with any innovation, the adoption of this new insurance prod- uct went through various life cycle stages. Often an idea is largely ignored for decades before being slowly adopted. After the idea has been tested, the replication phase begins. The overall effortsdescribed in this document are just entering the replication phase. Initial efforts to introduce the concepts in Nicaragua and Morocco have been slow to develop into projects. Nonetheless, these efforts and the experience of performing feasibility studies in these countries proved invaluable in the overall adoption process. Table 6.1 lists the chapter's country case studies in the order in which they are presented. Nicaragua and Morocco are covered first, as they were the first two countries to undertake the work, followed by IndiaandUkraine,bothcountriesinwhichweatherindexinsurance has been used. Ethiopia, Malawi, and the SADC appear next, pre- sented together because of the common elements in their experiences. Thenextlistedcountries,PeruandMongolia,eachdemonstrateunique aspects. Finally, the current progress of the Global Index Insurance 35 36 Managing Agricultural Production Risk Table 6.1 Summary of Case Studies Objectives Initial Work Better Social and by the Disaster Poverty Conceptual Significance of Country World Bank Status Growth Risk Mgmt Reduction Risk-Transfer Model Nicaragua 1998 Pilot in 2005 Direct link to loans and reduction of in- terest rates when farmers purchase index insurance Morocco 2000 No project More efficient and effective drought risk management for cereal producers India 2003 Three years Large scaling-up and mainstreaming of of sales weather insurance for smallholders Ukraine 2002 First sales in Regulatory approval under traditional 2005 insurance legislation and piloting of weather index insurance (first weather insurance contracts sold in April 2005) Ethiopia, 2003 Pilot in 2006 World Bank addressing rural risk in com- Micro prehensive manner; weather insurance for smallholders Ethiopia, 2003 Pilot 2006 WFP/WB jointly developed ex ante Macro weather insurance based financing of early response to weather failure leading to negative coping strategies Malawi 2004 Pilot 2005 Weather insurance for groundnut farmers SADC 2004 Feasibility Introduction of scaled-up safety nets; stage improved food security risk management comprehensively Peru 2004 Pilot planned Systematic approach to dealing with for 2006 agricultural risk by government Mongolia 2001 Pilot planned World Bank pilot project mainstreaming for 2006 designed to learn if herders will pay a commercial rate for mortality index insur- ance; prepaid indemnity pool coupled with a structure to completely protect the financial exposure Global Index Concept note Reinsurance intermediation for micro- Insurance and macrolevel insurance for insurers, Facility governments, and banks Source: Authors. Facility is described; this effort, much broader in nificant economic activity. In 2003, agriculture ac- scope than the individual country efforts, could sig- counted for nearly 18 percent of the US$4.1 billion nificantly facilitate risk transfer for all of the preced- GDP of Nicaragua, and thirty percent of popula- ing programs as well as any future activity. tion is involved in agricultural activities. The major commoditiesproducedincludecoffee,meat,shrimp, NICARAGUA corn, sugar, and beans. Since the 1990s, however, agriculture has had little or, often, negative growth. A Seven-Year Incubation Period With its agricultural production hindered by expo- sure to drought and flood risks, Nicaragua has re- Country Context and Risk Profile mained a net food importer of cereals and grains. The contribution of agriculture to the Nicaraguan Nicaragua has provided the World Bank's first GDP has been in decline, but it still remains a sig- experience in recent history of serious consideration From Theory to Practice 37 of rainfall insurance. Hazell and Skees provided pilot test area, the key person should investi- the first feasibility study in the spring of 1998. gate new regions with the potential to stand Subsequently, Skees and Miranda (1998) examined on their own, with private support, in which the issue in more detail and made specific recom- to inaugurate additional pilot programs; fos- mendations about rainfall insurance in the major tering similar activity in other regions will help cereal production area of northwest Nicaragua, entice the international reinsurance commu- which suffers major risks to cereal production from nity. Additional responsibilities for the key insufficient or excess rainfall, concluding that the person would be facilitating an active educa- risk of both could be hedged using rainfall index tion program and managing and deploying insurance contracts sold to individual farmers. funds for advertising and promotion. Nonetheless, Skees and Miranda also pointed to large hurdles blocking such an introduction in a Discussion of these concepts was progressing developing country and offered four key recom- in Nicaragua's public and private sectors when mendations for the development and sustainability Hurricane Mitch arrived with its devastation in of such an insurance scheme: October 1998. After this event, the World Bank's technical assistance efforts in Nicaragua shifted to · Analytical work and development of human cap- developing an aggregate weather index that would ital. Extensive data analysis and modeling provide disaster financing to the government dur- would be necessary to design and price the ing severe weather events. This work developed to insurance contracts. Training Nicaraguans in the point at which a specific set of weather stations these methods would be equally important were indexed into a single aggregate index to pro- in developing capacity within the country tect against catastrophic risk; the index was even for future efforts. priced in the global reinsurance markets. After the · Pilot development for demonstration, education, contract was priced, however, the government re- and evaluation. In its first year, the pilot should jected the idea, maintaining that they did not need start small and target primarily learning and to purchase insurance because they could depend demonstration. Education, marketing, and on the global community for assistance when major sales would be primary goals. Only three sta- catastrophes occurred. Subsequent to this decision, tions should be used in the first year: Leon; no further activity on index insurance has been pur- San Antonio; and Chinandega. The market sued in Nicaragua. Nevertheless, the Nicaraguan thus delineated would be contiguous and experience provides a number of significant lessons: would cover no more area than eight hundred square kilometers. To obtain the most effec- · It takes time to develop innovation. The literature tive risk management, only producers within on innovation emphasizes that it takes time, ten kilometers of the stations should purchase sometimes as much as a generation, for new the rainfall contracts. ideas to gain acceptance. The Nicaraguan ex- · Infrastructure development and pilot expansion. perience perfectly illustrates this observation. During year one of the pilot, investments in The original weather insurance idea was pre- additional secure weather stations should be sented in Nicaragua seven years ago, but new made to increase the density of stations within products deriving from those ideas are only the original eight hundred square kilometer now being introduced. One reason Nicaragua market area. By year two, sales and exposure may be proceeding now is because in the should increase to about US$10 million. meantime other countries have ventured into · In-country project management and support. It is this domain. essential to have a key person in Nicaragua to · The expectation that countries will purchase cata- manage and support the pilot project. This strophic protection presents an inherent moral person should know all aspects of the project hazard. The excellent work completed follow- and take an active role in every dimension of ing Hurricane Mitch to develop a mechanism the project. Central goals for this individual for the government of Nicaragua to indem- would be monitoring the activity and provid- nify catastrophic losses from extreme weather ing international reinsurers with the con- events met with a cool reception. The govern- fidence necessary to participate. Beyond the ment was likely correct in its conclusion that 38 Managing Agricultural Production Risk this type of protection was not needed, since becomes a useful tool for facilitating investments in the global community has been very respon- the agricultural sector. sive with free aid after major catastrophes. · Linking index insurance to banking in Nicaragua MOROCCO is an excellent addition to ongoing work else- where around the globe. Early indications are Country Context and Risk Profile that Nicaragua's banks have agreed to reduce In Morocco, 47 percent of the total population and interest rates for production loans for farmers most of the poor live in rural areas. Agriculture who purchase the new weather index insur- plays a crucial role in rural livelihoods. On average, anceproducts.Nicaraguamaybethefirstcoun- agriculture accounts for about 17 percent of the try to forge an explicit tie between interest rates GDP, but this percentage fluctuates, mainly due to and the amount of index insurance purchased. climatic--especially rainfall--variations. Moroccan This is an important development that should agriculture is characterized by a dichotomy between be evaluated and more fully understood. the traditional and commercial sectors. The tradi- tional sector consists of small farms in rain-fed areas Proposed Agricultural Risk involved predominantly in cereal, legume, and live- Management Structure stock production; the commercial sector operates mainly in irrigated areas. Farm surveys indicate that In November 2004, CRMG responded to INISER's about 70 percent of farms are small in size (under interest in developing a local weather index insur- 5 hectares) and account for 23 percent of total land ance market for agriculture. CRMG provided tech- under cultivation. Farms less than 20 hectares (ha) nical assistance to analyze potential markets for a in size represent 96 percent of farms in operation. pilot project in 2005 and decided to concentrate on Average farm size in Morocco is 5.7 ha. Almost developing a pilot project to secure lending for the 90 percent of Moroccan agriculture is nonirrigated, groundnut sector. Banks have expressed interest and the dependence of most crops on adequate in internalizing some part of the risk reduction by rainfall translates into wide variations in yields lowering interest rates and providing financing for and production. Drought caused cereal produc- farmers to pay premiums as incentives for a pro- tion, for example, to fall from 9.5 million tons in active financial risk management approach. 1994 to 1.6 million tons in 1995. Armed with prototype contracts, INISER/CRMG has launched consultations with end users, finan- Current Response cial intermediaries, and the insurance regulator. Final contracts have been designed and priced by In 1995, the Moroccan government activated the reinsurers, although they still await approval from Programme Secheresse (Drought Program), a state- the regulator. The pilot project is expected to begin sponsored insurance program managed by the local operations in the spring-summer of 2006. mutual agricultural insurance company (MAMDA) The government of Nicaragua had adopted a to address the drought problem by implementing "wait-and-see" strategy, based on several previous a yield insurance scheme. The program, revised in failures to launch either traditional or weather 1999, is structured on the coverage of three revenue index insurance for agriculture. It was not until the levels: 1,000, 2,000, and 3,000 Moroccan Dirhams most recent proposal was being developed and the (MAD) per hectare (ha). Payments are triggered by government could clearly see the interest and par- a ministerial declaration certifying the occurrence ticipation of the international financial markets that of drought. For the first revenue threshold, the pay- it opened the door for serious policy dialogue on out is based on an area-yield base mechanism, the issue. In particular, the government has offered while for the 2,000 and 3,000 MAD/ha levels, spe- to support INISER in the implementation phase cific farm yield assessments are required. The pro- with economic resources as well as guidance for gram proved to be popular, but it also encountered scaling up the current pilot project. This has opened typical yield insurance problems, such as high costs the door to work with several productive sectors, for supporting insurance premiums and severe including small farmers, in a comprehensive con- management problems related to individual farm text of economic development in which insurance yield assessment (Hess et al. 2003). From Theory to Practice 39 Proposed Agricultural Risk was that rainfall precipitation in the selected areas Management Structure showed a downward trend, and the reinsurance company involved in the deal made the cost of the Given the limitations of the Drought Program, the insurance prohibitive for producers. The experience Moroccan government agreed to participate in a developed through Morocco's feasibility study World Bank research project aimed at exploring the and planned implementation project, however, feasibility of weather-based insurance as an alter- generated expertise that led to the realization of native to traditional yield insurance. The investi- other WB-facilitated deals (for example, in India) gations led the team to conclude that a drought and of other independent programs (for example, insurance program based on rainfall indexes could in Colombia). have potentially significant benefits over the current scheme, minimizing moral hazard and adverse selection risk and promoting a more rapid, stream- INDIA lined pay-out process, in addition to increasing the Private Sector Led Alternative Agricultural potential interest of international reinsurers and Risk Market Development capital markets in investing in the program. Based on analysis of rainfall and cereal-yield data across Country Context and Risk Profile the country, the study determined that an index- A 1991 household survey addressing rural access based rainfall insurance product could be feasible to finance in India revealed that barely one-sixth in Morocco. Following the feasibility study, an of rural households had loans from formal rural international team sponsored by the IFC and the finance institutions and that only 35 to 37 percent Italian Technical Assistance Trust Fund assisted of the actual credit needs of the rural poor were MAMDA in structuring the insurance coverage to being met through these formal channels (Hess be launched as a pilot program in some cereal grow- 2003). A survey based on the Economic Census of ing regions. 1998 (Hess 2003) shows that Indian formal financial intermediaries reportedly met only 2.5 percent of Products the credit needs of the unorganized sector through commercial lending programs.37 The product proposed was a rainfall index insur- ance contract that would indemnify cereal pro- Current Response ducers when the rainfall index in a given area fell below a specified threshold. Farmers, then as now, responded to the lack of for- The indexes, developed by local agronomists mal financial services by turning to moneylenders; together with farmers' representatives, added im- reducing farming inputs; overcapitalizing and in- portant insights into the relationship of rainfall to ternalizing risk; and/or by overdiversifying their yield. They were not just cumulative measures of activities, leading to suboptimal asset allocation. rainfall but included specific weights for different Smallholders cannot risk investing in fixed capital plant growth phases and a "capping" procedure to or concentrating on the most profitable activities take into account the loss of water in excess of stor- and crops, because they cannot leverage the start- age capacity and hence unavailable to contribute up capital and they face catastrophic risks, such as to plant growth. This process allowed the indexes drought, that could wipe out their livelihoods at developed to reach correlation values of over 90 per- any time. The challenge for banks is to innovate cent (Stoppa and Hess 2003), and they were greatly low-cost ways to reach farmers and help them better appreciated by the potential end users. manage risk. Constraints Proposed Agricultural Risk Management Structure Despite the wide consensus gained by the pro- posed rainfall index contracts among government An initial study explored the feasibility of weather officials, insurers, and producers, the implemen- insuranceforIndianfarmerstodetermineifitwould tation of the planned pilot programs in Morocco be possible to extend the reach of financial ser- did not take place. The main reason for this failure vices to the rural sector by reducing exposure to 40 Managing Agricultural Production Risk weather risk (Hess, 2003). The study identified sev- farmers. In total, over 400 farmers bought insurance eral potential project partners. In response to this through BASIX in 2004, and a further 320 groundnut study, CRMG, in collaboration with the Hyderabad- farmers, members of a the Velugu self-help group based microfinance institution BASIX and the organization in the Anantapur district, bought in- Mumbai-based insurance company ICICI Lombard, surance directly from ICICI Lombard. Several a subsidiary of ICICI Bank, initiated a project to farmers were repeat customers from the 2003 pilot. launch the first weather insurance initiative ever In contrast to 2003, ICICI Lombard did not seek undertaken in India: a small weather insurance reinsurance for the BASIX farmer/weather insur- pilot program for groundnut and castor farmers in ance portfolio in 2004. the Andhra Pradesh district of Mahahbubnagar. In 2004, a number of other transactions also took The insurance contracts were designed by ICICI place within the Indian private sector in response Lombard, with technical support from CRMG and to the 2003 pilot program initiated by CRMG. In in consultation with BASIX, to protect farmers from 2004, BASIX themselves bought a crop lending drought during the groundnut growing season. portfolio insurance policy based on weather in- The products were marketed and sold in the four dexes. For the first time, BASIX used this protection villages selected by the extension officers of Krishna to cover their own risk and passed neither the cost Bhima Samruddhi Local Area Bank (KBS LAB)38 nor the benefits to their farmers. The protection using workshops and meetings with the BASIX allowed BASIX to keep lending to drought-prone borrowers. In total, 230 farmers (154 groundnut areas by mitigating default risk through the insur- and 76 castor farmers) bought the insurance for ance policy claims in extreme drought years. BASIX khariff, themonsoonseasonfromJunetoSeptember, bought a policy, insured by ICICI Lombard with 2003. Most purchasers fell into the small farmer structuring support from CRMG and reinsured into category, with less than 2.5 acres of landholding. the international weather market, covering three The entire portfolio of weather insurance contracts business locations. sold by BASIX was insured by ICICI Lombard, During 2004, not only did BASIX expand their with reinsurance from one of the leading inter- weather insurance program, a number of other in- national reinsurance companies. stitutions, including the originator ICICI Lombard, ICICI Lombard was also involved in another began expanding the market for weather insur- project in khariff 2003 in Aligarh, Uttar Pradesh, ance in India. In 2004, IFFCO-Tokio, a joint venture where 1,500 soya farmers bought protection against insurance company, launched weather insurance excessive rainfall. ICICI Lombard filed all the nec- contracts similar to the 2003 contracts, selling essary forms and terms of insurance with the Indian over 3000 policies to farmers throughout India. insurance regulator, registering their products In conjunction with ICICI Lombard, the govern- before the programs were launched. ment of Rajasthan launched a weather insurance A second pilot program was launched in khariff program for orange farmers, insuring 783 orange 2004 and introduced significant changes to the 2003 farmers from insufficient rainfall in khariff 2004; design following farmer feedback from the pilot they also covered 1036 coriander farmers in rabi program, with technical assistance from CRMG. (the October to March growing season) 2004. The The program was extended to four new weather National Agricultural Insurance Company (NAIC), stationlocationsintwoadditionaldistrictsinAndhra responsible for the government-sponsored area- Pradesh: Khammam and Anantapur. The weather yield indexed crop insurance scheme, also launched insurance contracts were offered to both BASIX a pilot weather insurance scheme for twenty dis- borrowers and nonborrowers and marketed and tricts throughout the country in 2004, reaching sold through KBS LAB in the Khammam and nearly 13,000 farmers; the scheme was even men- Mahahbubnagar districts and through Bhartiya tioned in the Indian government's budget for the Samruddhi Finance Ltd. (BSFL)39 in the Anantapur financial year 2004­2005. It is estimated that nearly district using village meetings, farmer workshops, 20,000 farmers bought weather insurance through- and feedback sessions during the month leading up out India in 2004. to the groundnut and castor growing season. New In2005,BASIX/ICICILombardfurtherimproved contracts were also offered for cotton farmers in the itsweatherinsuranceproductandautomatedunder- Khammam district and an excess rainfall product writing and claims settlements. In 2005, BASIX sold for harvest was offered to all castor and groundnut area-specific weather insurance products in all of From Theory to Practice 41 its fifty branches, finally selling 7,685 policies to sugar beets, wheat, and barley) in all twenty-five 6,703 customers in thirty-six locations in six Indian oblasts in the 1970 to 2001 period show a substan- states. In addition, ICICI Lombard scaled up its tial geographic spread of the agricultural values agricultural weather insurance sales and expanded concentrated in central and southern Ukraine. The into other sectors, while NAIC and IFCCO-Tokio correlation of crop yields between eastern Ukraine stepped up their efforts to sell weather insurance and the southern region near Odessa is nearly zero, products and to develop better products for farm- facilitating risk pooling and in-country retention of ers. New insurance providers such as HDFC Chubb a large share of natural risks. also entered the market. It is estimated that during 2005, 250,000 farmer bought weather insurance Current Response throughout the country. In partnership with ICICI Lombard, over seventy new automated weather In this market, the types of insurance policies cur- stations were installed by private company Delhi- rently offered are input cost insurance, generally based National Collateral Management Services linked to agricultural credit collateral requirements Limited, on which weather insurance contracts were and limited to very low insured sums, and harvest written for the 2005 monsoon season. The company insurance, covering hail, storm, excessive precipi- plans to scale-up their installations throughout the tation, frost, and fire risk. Drought insurance is of- country with more insurance-provider partners fered by only a few companies and is not generally in 2006. covered. Two crop insurance pools, one composed Monitoring will be an important element of the of five companies and the other of sixteen, were new pilot programs. Ultimately, it will be necessary founded in 2003 as part of attempts to provide more to learn not only if farmers are buying these prod- secure crop insurance to Ukrainian farmers. The in- ucts but how the purchases are changing their be- surance companies agreed to pool their agricul- havior and the lending behavior of local financial tural risks to improve their risk-bearing capacity institutions. Box 6.1 describes the initial steps being and to obtain access to international reinsurance taken to monitor the Indian weather insurance markets. Nevertheless, crop insurance policy sales products. An early result of monitoring efforts-- were very limited (around eighty for both pools). learning why farmers purchase the insurance--is Market participants cited the following reasons for reported in Table 6.2. the low uptake: inability to pay for the policy, un- clear loss adjustment and underwriting procedures, mistrust of insurance companies, and insufficient UKRAINE information available to farmers. Moreover, by pro- Country Context and Risk Profile viding ad hoc disaster assistance to farmers in 2003 and 2004, the government of Ukraine (GoU) low- Rural financial institutions in Ukraine increasingly ered incentives for farmers to pay for commercial use future harvests as collateral, since farm equip- insurance premiums. According to recent market ment is generally antiquated and of limited value. information, by the end of 2004, the biggest agricul- These lenders also tend to require harvest insur- tural insurance pool had shrunk to six companies. ance to hedge against crop losses.40 The major banks active in agricultural lending, such as Aval (with Policy Objectives a total of 4600 loans and 30 percent market share), do not lend on the basis of uninsured collateral, so The GoU has experimented with compulsory crop to obtain credit a farmer must have a proper insur- insurance and is now establishing a crop insurance ance policy written by a preapproved insurer. To subsidization scheme. The regulator has approved provide for the lending insurance needs of farmers, weather index insurance as an insurance product, most banks set up their own insurance companies. and a few weather insurance policies were sold to Most farmers do not yet understand the particular farmers in the first pilot sales season of 2005. nature of weather index insurance, but they are A feasibility study by CRMG presents a risk man- familiar with weather risk and would like to have agement framework and considers several options protection against multiple natural perils. for government intervention in the sector. An in- Crop risk is diverse throughout Ukraine. Crop- vestment phase would consist of the acquisition and yield data for five major crops (maize, sunflowers, installation of automated weather stations, includ- 42 Managing Agricultural Production Risk Box 6.1 India Impact Assessment CRMG and DECRG designed a baseline surveya that the crop to which the rainfall insurance was linked was implemented by the International Crop Research and the premium and payouts, but not the trigger Institute (ICRISAT). The survey was conducted to study levels. In fact, insurance trigger levels are expressed the introduction of the rainfall insurance products in millimeters of cumulative rainfall, but most farmers designed by ICICI Lombard and marketed through do not understand the concept of a millimeter. Most BASIX. The main objectives were to assess, first, the farmers determine when to sow by analyzing the take-up rate, that is, the factors influencing the deci- moisture in the ground, and, indeed, only 10 percent sion to purchase the insurance product, and, second, were able to make an estimate in millimeters of the the impact of the insurance product in the treated minimum accumulated rainfall required to sow. villages as compared to the control villages. A sample Nonetheless, take-up was high. Buyers said they was drawn from Hindupur, Anantapur district, and purchased the insurance for security reasons (exposure Narayanpet, Mahahbubnagar district, of 1,052 farm- to rain, large cultivation of castor or groundnut, etc.) ing households, including 267 buyers, 186 nonbuyers and because they were advised to do so by others. Yet who attended a marketing meeting, and 299 nonatten- initially, many buyers thought of the insurance policy dees in the treated villages. In addition, 300 farming as a gamble. They put money at risk in the hope of households were interviewed in control villages. making a profit if the accumulated rainfall was below Anantapur and Mahahbubnagar are characterized a certain threshold. To support this claim, we find that by low and uncertain rainfall, low levels of irrigation, risk-loving people are more likely to buy the policy as and shallow and infertile soils. Anantapur has virtually well as those that believe that the monsoon rains will a groundnut monoculture, while Mahahbubnagar has start later, for whom the gamble has favorable odds. castor bean, groundnut, sorghum, pigeon pea, maize, In addition, buyers are generally more educated, farm cotton, paddy, and finger millet crops. Crop failure is more land (total and irrigated), have more savings at very frequent in these districts, mostly triggered by the time of purchase, and are more likely to trust the droughts. Indeed, 80 percent of farmers considered insurance product and BASIX, as compared to non- drought their main risk. In a drought year, farmers can buyers. At the time of the survey, most farmers in lose about 25 percent of income. Drought affects most treatment villages reported that they would like to pur- villagers at the same time, rendering informal insur- chase the insurance for the next khariff (main mon- ance networks useless. Instead, in bad years, farmers soon) in June 2005. In addition, 14 percent of poorer sell livestock or their few assets and migrate to urban farmers said they would like to open savings accounts areas or other states. In addition, they borrow from in November to save for the premium. Again, when formal and informal rural financial institutions. The asked why they would like to buy the insurance in union and state governments offer employment gener- 2005 (see Table 6.2), 60 percent cited security reasons, ation schemes, watershed development programs, and but a full 30 percent cited the experience of a payout other welfare schemes to stem migration and assuage in 2004. the misery of the people. This willingness to purchase the policy as a result The rainfall insurance product was explained by of a payout is particularly telling in the context of the BASIX and ICICI in village meetings. Most people who introduction of a new product. Farmers may be uncer- heard about the meeting decided to attend; of those, tain that BASIX will honor its promise and thus may de- 35 percent attended because they trusted BASIX and cide to wait and see and not change behavior. Indeed, another 35 percent because friends and neighbors the preliminary analysis conducted suggest that while attended. Only 27 percent of the buyers purchased the there are no differences in input usage or area devoted insurance during the marketing meeting, because the to cash crops for farmers that do not trust Basix or the product was new and meeting attendees lacked the product, it does seem that trust in BASIX allows buyers requisite funds. Meeting participants well understood to use the insurance policy as a hedging instrument. Note: a. Financed by Swiss Trade Commission, SECO. Source: This information is based on preliminary findings by economist Xavier Giné (DECRG, World Bank), working in collaboration with Don Larson (DECRG, World Bank), Robert Townsend, professor at the University of Chicago, and James Vickery, an economist at the Federal Reserve Bank of New York. From Theory to Practice 43 Table 6.2 Reasons for Buying Weather Index Insurance in India Khariff 2004 Khariff 2005 Reasons for Buying Insurancea Freq. % Freq. % Security/risk reduction 144 54.8 181 53.2 Could not afford to lose harvest income 25 9.5 11 3.2 Low premium 19 7.2 1 0.3 Advice from progressive farmers 18 6.8 0 n/a Other trusted farmers bought insurance 7 6.5 5 1.5 Advice from village officials 10 3.8 1 .3 High payout 10 3.8 10 .9 Concentration on castor crop 7 2.7 4 1.2 Product was well explained 5 1.9 0 n/a Concentration on groundnut crop 4 1.5 0 n/a Luck 4 1.5 5 1.5 Paid out for previous year 0 n/a 107 31.5 Advice from BUA members 0 n/a 11 3.2 TOTAL 263 100 340 100 Note: a. The categories listed were created from open-ended survey responses to the question, "Why did you buy the insurance product for the last khariff?" The same categories may not apply for both years. Source: ICRISAT survey, courtesy Xavier Gine. ing analysis of the density of the network required an efficient and well-regulated risk pool can over- to cover Ukraine's weather exposure and design come this market failure. Risk layers representing of an adequate maintenance program to ensure the relatively frequent (but mild) adverse events would quality of observations across time. be insured by the GoU risk fund. Intermediate risk In addition, the GoU could consider a backstop layers (for example, events occurring once in twenty facility for weather risk insurance retention. years to once in one hundred years) could be trans- Ukrainian insurance companies would need inter- ferredtotheGoUBackstopFacility.Thecatastrophic national reinsurance for insuring against systemic risk layer (the once in one hundred year event) risks. A risk pool "facility" in Ukraine would allow could be transferred to international reinsurance for the underwriting of agricultural reinsurance markets. based on preestablished guidelines to retain as much risk inside the country as possible. This pool ETHIOPIA would then reinsure itself through a GoU fund. Ethiopian Insurance Corporation and Donor Extreme or catastrophic risk would be reinsured Led Ex Ante Disaster Risk Management on the international reinsurance market based on transparent and competitive premium ratemaking Country Context and Risk Profile principles; that is, once the pool and the GoU fund are depleted, international reinsurers would pay Ethiopia is one of the poorest and least developed the remaining claims. Aggregation and layering countries in the world, ranking 169 of the 175 coun- of risk would help interest reinsurers in reinsuring tries in the Human Development Index. More than risk in Ukraine, causing them to price risk compet- 85 percent of the population make their living in itively. Individual insurance companies sometimes the agricultural sector, which accounts for 39 per- face insurmountable difficulties even accessing cent of Ethiopia's GDP (2002/2003) and 78 per- international reinsurance markets, let alone obtain- cent of foreign earnings. In Ethiopia, agriculture ing competitive prices. The combination of intro- is predominantly rain-fed, and more than 95 per- ducing a transparent index insurance product and cent of its output comes from subsistence and 44 Managing Agricultural Production Risk smallholder farmers. The staple diet for the majority dictable, and untimely to provide an effective in- of Ethiopians consists of coarse grains, including surance function. maize,teff(acerealgrain),andsorghum.Production In 2003, in part to address this issue, the gov- of coarse grains is valued at around US$380 million ernment of Ethiopia (GoE), donors, United Nations and cereals at US$585 million. agencies, and nongovernmental organizations At the household level, adverse weather patterns, (NGOs), launched the New Coalition for Food primarily lack of rain, are detrimental to yields and Security with the goal of achieving food security outputs and result in significant income losses and for the part of the Ethiopian population categorized negative impacts on farmers' livelihoods. Ethiopia as "chronically food-insecure" and to improve sig- faces highly variable rainfall and suffers from both nificantly the food security for the additional ten national and regional droughts that can have ex- million people vulnerable to becoming so in the next treme impacts on farmers who utilize traditional fiveyears.Toachievethesegoals,startinginJanuary agricultural practices with little irrigation and who 2005, the organizations began working through rely on the country's thirty-five million head of live- the government to introduce a productive safety stock. This rainfall variability, in addition to limit- net for five to six million people. The safety net is ing the ability and motivation of farmers to invest not meant to serve as an emergency activity but in agricultural technology and yield-increasing to change the vulnerability and risk profile of the assets, reduces overall production, which can de- chronically food-insecure. Responses to chronic crease both household consumption and income. and to emergency food shortages began to be ad- At the national level, average grain production in dressed through different channels: the former, the country is 8.9 million metric tons (MT) and is essentially a development activity, fell to the pro- subject to recurrent drought. The Ethiopian min- ductive safety net program coordinated by the Food istry of agriculture has indicated that the level of Security Coordination Bureau, and the latter, a re- production is too low to feed the whole population sponse mechanism to unpredictable humanitarian even in good rainfall years. needs, to the Disaster Prevention and Preparedness Commission (DPPC). Accordingly, those house- holds not covered by the safety net program but still Current Response considered in need of government relief assistance With 10 percent of the population of seventy-two will fall under the emergency program through million requiring food aid assistance each year, early warning and annual needs assessments. food insecurity is a chronic issue. Emergency re- sponses have been frequent if not constant, ac- Proposed Agricultural Risk counting for an annual average of 870,000 MT of Management Structures food aid between 1994 and 2003. In 2003, a record thirteen million Ethiopians required emergency as- To address the current situation in Ethiopia, sistance as a result of drought and the consequent two agricultural risk management structures are failed harvest in 2002. These emergency responses currently being considered, one at the farmer or have saved millions of lives in the short term, but microlevel and the other at the government or destitution has worsened, assets have eroded, and macrolevel. vulnerability has increased. The uninsured loss of income and assets caused by natural disasters, pri- Microlevel Weather Insurance marily droughts, in developing countries such as The state-owned Ethiopia Insurance Corporation Ethiopia threatens the lives and livelihoods of vul- (EIC) plans to launch a small pilot weather insur- nerable populations. Insurance is a critical require- ance program for wheat and pepper farmers in ment for development, as uninsured losses lock southern Ethiopia in the wereda (district) of Alaba, entire populations in vicious cycles of deepening SNNPR. The EIC has previously experimented destitution. It is estimated that in sub-Saharan with agricultural insurance for farmers, but it met Africa approximately 120 million people are at risk with little success. The EIC is keen to explore new from natural disasters and that, for these popula- potential products to address the risks of larger, tions, humanitarian aid provides the only insur- commercial farmers. A pilot program, for which ance protecting their lives and livelihoods. But it receives technical support from CRMG, is due to humanitarian aid is often too unreliable, unpre- start in April 2006. Part of the work includes the From Theory to Practice 45 demand assessment and participatory design of The poverty reduction strategy is character- contracts with Alaba farmers ized by strong country ownership and focuses on a broad-based participatory process. In particular, Macrolevel Ex Ante Funding of Emergency the GoE favors a gradual shift from food assistance, Relief Operations assistance in kind, toward financial assistance that The World Bank and the United Nations World could be used to purchase food from the domestic Food Programme (WFP) have launched a pilot to market. The New Coalition for Food Security at- investigate the feasibility of index-based weather tests to the government's ambitious poverty reduc- insurance as a reliable, timely, and cost-effective tion strategy: the main features of the safety net are way of funding emergency operations in Ethiopia. multiannual funding, transition toward cash-based The intention is to address the extreme emergency programming, scaled-up public/community works, drought situations that put pressure on donor linkages with broader food security programs, har- budgets and GoE strategic grain and cash reserves. monized budgeting, and monitoring and evalua- The pilot is designed to serve vulnerable popula- tion. The Food Security Coordination Bureau has tions who are neither food-insecure nor included in been created, under the Ministry of Agriculture and the country's new safety net program but who are Rural Development, to coordinate all food security "at risk" to income and asset losses and consump- programming, including the safety net. Targeting tion shocks resulting from the more severe natural the nonchronically hungry but food-insecure or disasters. It is estimated that at least a further 35 per- vulnerable populations, an index-based weather cent of the population, above those considered insurance approach for Ethiopia aiming to provide chronically food-insecure and covered by the safety contingency cash funding for responses to severe net, is at risk from hunger in the event of an extreme and catastrophic drought clearly aligns with the drought such as that in 1984. A traditional food aid government's strategy and complements the safety response to a catastrophic drought in today's prices net initiative. would be estimated to cost about US$1.6 billion The objective of the macrolevel pilot project is to for all beneficiaries, chronic and nonchronic.41 In contribute to an ex-ante risk-management system preparing for a future drought, rather than rely to protect the livelihoods of Ethiopians vulnerable on traditional funding approaches based on pro- to severe and catastrophic weather risks. The pilot tracted appeals to international donors, the insur- will use a weather derivative to demonstrate the ance approach focuses on transferring the risk to the feasibility of establishing contingency funding for reinsurance and capital markets. Such a mechanism an effective aid response in the event of contractu- will ensure predictable and timely availability of ally specified severe and catastrophic shortfalls in funds with which the DPPC can launch emergency precipitation. WFP will put in place a small hedge relief operations and appropriate interventions in for Ethiopia's 2006 agricultural season from March the event of a well-defined rainfall deficit at harvest to October 2006, demonstrating the possibility of time. Some of the benefits of this type of insurance- indexing and transferring the weather risks of based emergency funding include objective pay- least-developed countries and facilitating price dis- outs, timely delivery, and funding in cash. In the covery for Ethiopian drought risk in international case of Ethiopia, the insurance approach would financial markets. In effect, in the pilot stage of the allow intervention four to six months earlier than initiative,theWFPwillbethecounterpartytoacom- does the traditional appeals-based system. mercial transaction with the international risk mar- ket. Donors will pay for the premium associated with this risk transfer. Ideally, however, the ulti- Policy Objectives mate aim of the initiative would be for the GoE to Both proposed agricultural risk management struc- take responsibility for the risk management pro- tures are in line with the GoE current poverty reduc- gram as part of its overall long-term poverty reduc- tion strategy, which focuses on (1) agricultural-led, tion strategy. rural-based growth, recognizing the importance of improving the environment for exports, private Constraints sector growth, and rural finance; and, linked to this, (2) food security. Clearly the microlevel weather in- Two major constraints might, in the short term, limit suranceinitiativesarecomplementarytothegovern- the proposed risk management frameworks. The ment's primary focus on agricultural development. first involves the weather-observing network and 46 Managing Agricultural Production Risk the weather data available in Ethiopia. The National operations in Ethiopia. It is therefore an appropri- Meteorological Services Agency (NMSA) is respon- ate proxy for representing economic loss due to sible for a network of over five hundred weather sta- drought and also a simple, objective basis for index tions and rain gauges throughout Ethiopia. Not all insurance. The appropriate index must be based on of these weather stations, however, offer reporting a weighted average, or "basket," of as many stations capabilities or historical data of a quality sufficient as possible to capture the macrolevel nature of the to transfer risk to the international markets or even risk the GoE faces. The government may be able to to perform an actuarial analysis of the weather cope with small, localized droughts by transport- risks involved. Furthermore, given the large size ing food supplies from other regions of the country and challenging topography of the country, the and by sourcing government budget reserves. spatial distribution of the network is inadequate to Retaining such risks will most probably be a more protect the entire country from weather risk. These cost-effective solution than would seeking insur- issues will hamper both micro- and macrolevel ef- ance, and Ethiopia should be able to take advan- forts. On the microlevel, initially, only farmers who tage of any natural diversification of the country live near good weather stations will benefit from to reduce its insurance costs. In situations where the availability of weather insurance. Furthermore, drought severely affects a single region or affects the EIC may find it difficult to secure reinsurance several regions or the entire nation, however, the for this risk until the quality and security of the government may find this reallocation of resources NMSA network improves. On the macrolevel scale, unmanageable, making it appropriate to utilize the the weather protection can only be designed using basket-based insurance product to fund the ex- weather stations that adhere to the strict quality re- pected emergency relief operations in a predictable quirements of the international weather market. and timely manner. The basket approach also re- This will naturally limit the scope of the project in duces the risk of reliance on one weather station its first years. and the associated issues of moral hazard and basis The second constraint, more relevant for the risk. On this note, including more stations in the bas- macrolevel weather-risk transfer, involves fiscal ket not only provides better national coverage and, issues: namely, the ability of the government of hence, enhances the representation of the index, it Ethiopia eventually to take over the ex ante fund- also increases the placement potential of the struc- ing of the emergency relief operations program and ture in the international reinsurance markets. In to take responsibility for the premium payments 2006, the index to be piloted is based on a basket necessary to establish and maintain this funding of 26 weather stations distributed throughout the mechanism. agricultural producing areas of the country. In the pilot stage of the program, the WFP will Products and Risk Transfer Structure be the counterparty to any commercial transaction with the international risk market and donors will Both micro- and macrolevel proposals focus on pay for the premium associated with this risk trans- index-based weather risk management solutions. fer. In the event of an extreme and catastrophic At the microlevel, the EIC will market and sell drought, however, any payment triggered by the weatherinsurancecontractstokebeles(smallgroups insurance would be made available to the GoE of farmers) and/or farming cooperatives to protect DPPC. This would allow the early provision of re- theirfarmermembersfromthefinancialcostsassoci- sources to the GoE and thus to the beneficiaries to ated with crop failure as a result of adverse weather. ensure appropriate consumption smoothing and to The products will be similar in concept to the prod- avoid distressed sales of assets, a vital outcome if ucts offered to farmers in India (see Appendix 2), but the intervention is to play an effective and protec- it will be sold at the group rather than individual tive role. With the availability of cash, the interven- level in line with farmer preferences identified dur- tion can also be used to fund activities other than ing discussions and focus groups in Alaba. The EIC food aid that have already been established in other will then seek international reinsurance for their parts of the country, such as cash-transfers, food- portfolio of weather risk. for-work, or cash-for-work schemes. Ultimately, the At the macrolevel, lack of rainfall is the domi- long-term objective of these insurance plans would nant, immediate cause triggering emergency relief be for the GoE to go directly to the market and take From Theory to Practice 47 responsibility for the program rather than having it At the intermediary level, banks can package continue to operate through the intermediary WFP. loans and weather insurance into a single product, a weather-indexed crop production loan. Farmers would enter into higher interest rate loan agree- MALAWI AND SADC ments that include weather insurance premiums Weather Risk Transfer to Strengthen that the bank would then pay to the insurer. In case Livelihoods and Food Security42 of a severe drought impacting crop yields, the borrower would pay only a fraction of the usual Country Context and Risk Profile loan due and would thus be less likely to default, Malawi is dominated by smallholder agriculture, strengthening the bank's portfolio and risk profile. with farmers cultivating mostly maize, the staple Historical simulations in Malawi of such products food. Maize is very weather sensitive and requires from maize demonstrated that the years of reduced a series of inputs. The economy and farm liveli- loan payments coincided with the drought years in hoods are affected by rainfall risk (and resulting which farmers suffered from much lower yields, food insecurity), soil depletion, lack of credit, and mainly 1992 and 1994. Recently, CRMG partnered limited access to inputs. Malawi suffers serious with Opportunity International (OI) to develop capacity constraints because it is ravaged by poverty weather insurance products to secure credit for and AIDS. Very few people have the energy and groundnut farmers. Nearly 1000 policies were skills to build financial service programs. sold in October 2005 for the 2005/2006 groundnut growing season. At the macrolevel, a specific nationwide maize Current Response production index for the entire country could form Malawi once had a paternalistic state culture. The the basis of an index-based insurance policy or op- role of the state in agricultural marketing (mainly erate as an objective trigger to a contingent credit tobacco but also maize) is still strong. Prices are not line for the government in the event of food emer- free, and smallholder incentives are distorted due gencies that put pressure on government budgets. to food aid and sales of subsidized maize by the state Applying the Lilongwe maize farmer index ap- marketing board. The state and donors respond to proach to the macrolevel situation, a Malawi Maize recurrent drought-induced food crises by ad hoc Production Index (MMPI) can be defined as the disaster relief programs. weighted average of farmer maize indexes mea- sured at weather stations located throughout the country, with each station's contribution weighted Proposed Agricultural Risk by the corresponding average or expected maize Management Structures production in that location. Given the objective na- At the micro- or farm-level, weather-based index ture of the MMPI and the quality of weather data insurance allows for more stable income streams from the Malawi Meteorological Office, such a and could thus protect peoples' livelihoods and im- structure could be placed in the weather risk rein- prove their access to finance. An insurance product surance market. Analysis shows that Malawi could can be based on a crop production index constructed need up to US$70 million per year to financially from weather data recorded at the airport weather compensate the government in case of an extreme station in Lilongwe (Malawi's capital). Analysis food emergency. Given the size of this figure, such and simulations conducted for the Lilongwe area a transaction would be treated on a stand-alone indicate that the match between potential insurance basis, with an estimated premium of approximately payouts and farm-yield losses would be adequate. three times the expected loss for the reinsurer. In All that is needed is for demand to be aggregated at this case, the expected loss--given forty years of product distribution channels such as the National historical rainfall data and assuming the govern- Smallholders Association (NASFAM). Rural finan- ment retains the cost associated with deviations cial institutions could finance the insurance premi- in maize production up to 25 percent away from ums and lower interest rates to borrowers, since the normal--would be US$2.32 million, implying a financial institutions stand to benefit from reduced premium of US$6.96 million or an insurance rate default risk. of 10 percent for such a product. 48 Managing Agricultural Production Risk The weather index/drought risk management PERU approach suggested for Malawi could be extended Government Led Systemic Approach to to a regional level to include all members of SADC Agricultural Risk Management at some future point. Weather risk can be retained and managed internally if the areas under manage- Country Context and Risk Profile ment are significantly diverse in their weather risk PeruiscurrentlynegotiatingaFreeTradeAgreement characteristics. This immediately suggests that the with the United States. Agriculture, because of its weather sensitivity of neighboring countries, the lack of competitiveness, is one of the most vulner- SADC members, must be taken into account when able sectors when an economy is opened. In this considering Malawi's weather risk profile and its context, the Peru's Ministry of Agriculture (MA) need for outside insurance. Analysis of the SADC is preparing a multidimensional strategy involv- region shows that, on average, two countries suffer ing extension services to farmers and innovative drought each year. The distribution of drought financial schemes, with the private sector partici- events in SADC is extremely long-tailed, however, pating to facilitate access to better technology and with the possibility of widespread drought events new markets. Because of farmers' lack of bankable that could potentially devastate the region. This collateral, the MA intends to facilitate the emer- indicates that the most efficient way to layer and gence of a sustainable private agriculture insur- thus manage the risk is as follows: ance market. · SADC Fund: The size of the SADC fund could be set at US$80 million, the average financial Current Response impact of four average droughts in the region, with each member contributing its share ac- Two major efforts in the last decade have attempted cording to an actuarially fair assessment of to introduce agriculture insurance in Peru, but the the expected claim of each country. results were disastrous. Lack of technical knowl- · Reinsuranceand/orcontingentcreditlines: SADC- edge and exposure to catastrophic events like wide events incurring a financial loss of, say, El Niño generated big losses in the industry. From US$80 million to $350 million could be trans- the consumers' perspective, these schemes were ferred to the weather-risk reinsurance/profes- not transparent and lack of education translated sional investor market. Alternatively, in such into dissatisfaction about the scope and use of these situations, the SADC members could have financial instruments. Currently, crop insurance or access to a World Bank contingent credit line. similar instruments are not available to farmers. · Securitization: The final and extreme layer of risk, such as drought in ten countries, occur- Proposed Agricultural Risk ring 1 percent of the time, could be securi- Management Structure tized and issued as a CAT bond (investors lose the principal if the event occurs in ex- The government of Peru (GoP) created a special change for a higher coupon) in the capital commission in 2003 to draft a strategic plan for markets. The advantage of capital markets the implementation of an agriculture insurance for this risk transfer is the immense financial scheme. The treasury ministry, agriculture depart- capacity of these markets and also the longer ment, insuranceregulator,privateanddevelopment tenure of CAT bonds: up to three years and bank representatives, farm unions, and insurance possibly longer. representatives participated in the discussions and recommendationsforthestrategicworkplan.Aspe- A more efficient means of transferring risk implies cific body designed for that purpose is the Technical that costs could be greatly reduced for the member Committee for the Development of Agriculture countries by transferring risk as part of a regional Insurance (TCDAI), which was created by ministe- strategy rather than by transferring that risk one rial resolution in September 2004 and is housed in country at a time. The SADC fund approach out- the agriculture ministry. The TCDAI is currently lined above, for example, would reduce insurance working on several technical studies related to the costs by 22 percent for Malawi due to risk pooling design and implementation of agriculture insur- effects. ance in Peru. From Theory to Practice 49 Policy Objectives 4. Regulatory review: The purpose of this activ- ity is to develop a strategic work plan with The main objectives of the GoP are (1) to maintain the insurance regulator to prepare the neces- the prudent fiscal, monetary, and exchange rate sary technical documentation for the index policies essential to attract investment and promote insurance product to be approved under the continued growth; and (2) to complement growth guidelines of property insurance. with direct interventions that address inequality and poverty, focusing on excluded groups: indige- The TCDAI has defined the following crops and nous people, Afro-Peruvians, and at-risk groups areas of interest for the feasibility study: such as youths and single mothers (Peru, 2004­06). Rice--San Martín Mango--Piura Constraints Yellow maize--Lima Potato--Huanuco In addition to fiscal constraints, Peru's agricultural Coffee--Cuzco sector is divided into two: a group of powerful Cotton (Tangis)--Ica export-oriented, high-value agricultural produc- Cotton (Pima)--Piura ers concentrated in twelve valleys along the coast Asparagus--Lima and a group of smallholder agricultural producers occupying the sierra (highlands) and selva (jungle) Risk-Transfer Structure areas. The GoP seeks to enhance risk-taking capacity in the country generally by facilitating special risk Products transfer arrangements with insurance companies The technical committee, assisted by CRMG, pro- in Peru, particularly those wishing to launch agri- posed a four-part work plan: cultural insurance. Specifically, the GoP wishes to set up a US$50 million fund, managed by the lead- 1. Design of prototype index contracts: The feasibil- ing second-tier bank (COFIDE), to take agricultural ity of these contracts is tested for several crops risk. In addition, the technical committee plans to in the three main agricultural areas of Peru develop for insurers index-based products directly (coastal, sierra, and selva). The contract de- transferable into international risk markets. sign requires weather data from the Peruvian weather service (SENAMHI), acquisition of which is a priority for the work plan. MONGOLIA 2. Demand assessment: This activity will aim at World Bank Contingent Credit for Livestock gauging the demand for weather insurance by Mortality Index Insurance43 type of producer and will include participatory Country Context and Risk Profile design sessions addressing questions such as whattypesofcontractstodevelopandforwhat The economy of the Mongolian countryside is periods. This activity will include training based on herding: agriculture contributes nearly potential end users (farmers) regarding index one-third of the national GDP, and herding ac- insurance basics (for example, types of in- counts for over 80 percent of agriculture. Animals demnities, how indemnities and premiums provide sustenance, income, and wealth, protect- are calculated, and how contracts are settled). ing nearly half the residents of Mongolia. Shocks 3. Delivery model design: Based on a mapping of to the well-being of animals have devastating im- rural financial intermediation in Peru, this plications for the rural poor and for the overall activity will evaluate segmented delivery Mongolian economy. Major shocks are common models to be used for real distribution chan- as Mongolia has a harsh climate, and animals are nels to farmers with small- and medium- herded with limited shelter. From 2000 to 2002, sized farms with viable production potential. elevenmillionanimalsperishedduetoharshwinters Prototype contracts by institution and client (dzuds). The government of Mongolia has struggled segment will be used in working with poten- with the obvious question of how to address this tial intermediaries. problem. 50 Managing Agricultural Production Risk The Mongolian government requested specific pool. Insurers must replace the reinsurance cost assistance in coping with extreme livestock losses. and the exposure above 100 percent for the pre- Given the nature of highly correlated death rates paid indemnity pool. for animals in Mongolia, an index-based livestock In the syndicated pooling arrangement, partici- insurance (IBLI) product was proposed and in pants share underwriting gains and losses based on May 2005, and the World Bank approved a loan the share of herder premium they bring into the to Mongolia to finance the Index-Based Livestock pool. Each insurer also pays reinsurance costs con- Insurance Project. This project will support a three- sistent with the book of business they bring into the season pilot program in three Mongolian states and pool. This gives the reinsurance pool the benefits of includes a contingent debt facility to serve as a the pooling arrangement and provides the oppor- mechanism for protecting against extreme losses tunity to build reserves for the overall activity. The during the pilot. The major objective of the pilot reinsurance pool pays for the first layer of losses program is to determine the viability of IBLI in beyond the 105 percent stop-loss. Once the re- Mongolia, including testing herders' willingness to insurance pool is exhausted, the government of pay for an IBLI product. The index would pay in- Mongolia can call upon the contingent debt to pay demnities based on adult mortality rates by species for any remaining losses. and by soum (province). By law, Mongolia per- A major advantage of having a prepaid indem- forms a census of animals each year. Elaborate sys- nity pool is that all other lines of the insurance busi- tems are in place to assure the quality of the data. ness are protected from the extreme losses that can The proposed pilot involves three distinct layers of occur from writing a highly correlated agricultural risk: (1) self-retention by the herder; (2) a base in- risk policy. In the long-term vision, the syndicate surance product (BIP) for mortality rates in a cer- will be well positioned to find risk-sharing partners tain range; and (3) a disaster response product in the global community quickly, as the pooling (DRP) for livestock losses beyond the layer covered arrangement is both risky and profitable. Reinsurers by the insurer. might be willing to provide capital and enter quota- An index-based insurance program was recom- share arrangements on that risk. To the extent that mended because of significant concerns about the the risks within the pool are standardized, using the moral hazard, adverse selection, and extreme mon- same measures and procedures, one can also envi- itoring costs associated with any individual live- sionthismechanismasameanstosecuritizetherisk. stock insurance program in the vast open spaces of Finally, the design also offers the opportunity to Mongolia. Weather index insurance was consid- transition the system to the market once it is learned ered; however, it was determined that the weather whether herders find the BIP an acceptable product events contributing to livestock deaths were too and demonstrate a willingness to pay. complex to develop this alternative. The project The first challenge to the risk transfer structure is will support continued research to strengthen the the uncertainty of the livestock mortality index- mortality index by incorporating other indexes, for based on an annual government census of all ani- example, the Normalized Difference Vegetation mals in the country. Several systems are in place to Index (NDVI), as a means of establishing a more monitor potential problems during the pilot, for ex- secure index for paying losses. ample, the movement of animals across soum bor- While it is believed that the index-insurance ders. From the perspective of the reinsurer, even the product can be effectively underwritten, signif- governmentcouldhavetheincentivetotamperwith icant financial exposure for a nascent insurance the data if this data determines the level of reinsur- market with extremely limited access to global ance claims. The project seeks to establish systems to risk-shifting markets remains among the largest verifylossesusingthird-partyaudits.Asecondchal- challenges. Given concerns about financing ex- lenge is the sustainability of the proposed pooling treme losses, the pilot design involves a syndicate mechanism that determines reinsurance premiums pooling arrangement for companies. Pooling risk for each participating insurer using advanced mod- among the insurance companies offers some op- eling procedures. Human capital within the country portunity to reduce the exposure for any individ- must be developed to perform these duties. Pooling ual insurer. In the short term, the government of mechanisms generally tend to fail because of collec- Mongolia will offer a 105 percent stop-loss on the tive action problems and high transaction cost. The pooled risk of the insurance companies. Herder challenge in Mongolia will be to move the pooling premiums go directly into a prepaid indemnity mechanism to a private sector entity by the comple- From Theory to Practice 51 tion of the pilot; otherwise, if left to the government Facility (GIIF) that would intermediate weather, to maintain, the system will likely be unsustainable. disaster, and price risk (all index-based) among developing country-based primary insurers, gov- ernments, banks, and organized markets. CRMG GLOBAL STRATEGY is in intense dialogue with market makers as to The Global Index Insurance Facility (GIIF) the risk-taking capabilities of the GIIF, with a focus toward "crowding-in" rather than "crowding-out" Background the private sector. The facility would consist of a Theeconomicgrowthprospectsofdevelopingcoun- 100m capital investment in a risk-taking entity tries are negatively impacted by external shocks, that would underwrite global weather, disaster, which create both short- and long-term physical and price risks in developing and, in particular, and financial distress. The lack of coherent and the African-Caribbean-Pacific (ACP) countries. The timely response to shocks, coupled with indirect main objective of the facility would be to achieve impacts on growth and investment, compound the returns on equity and build a diverse portfolio of cost of direct physical damage. Uninsured enter- risk from developing countries not previously prises do not develop their full earnings potential transferred to the capital and insurance markets, because they engage in low-risk/low-return ac- thereby leveraging private risk transfer. The main tivities to minimize downside risks. Generally, developmentobjectivewouldbetoalleviatepoverty too much capital goes into nonremunerated self- by facilitating effective disaster insurance and risk insurance. OECD countries, on the other hand, reduction, allowing countries and enterprises to tend to be better equipped to manage shocks since profitably invest resources rather than waste them they have larger diversified economies that can with inefficient self-insurance. The GIIF would fur- withstand such events and because private assets ther facilitate risk transfer by absorbing transaction are insured. Demand for risk management instru- costs for developing country clients through cofi- ments is often frustrated by market gaps and entry nancing of premiums, funded separately by EC/ barriers. International reinsurers, for example, ACP funds, and through reinvestment of dividends require substantial minimum risk amounts: "The by public sponsors. greatest challenge is not to find capacity, but to find a large enough portfolio to make it worth under- Types of Risks Underwritten by the GIIF writing" (Tobben 2005). The GIIF seeks to close the gap between the The GIIF would provide cover for disaster, weather, developing country's demand for insurance against and price risks by underwriting index-based insur- severe shocks at public and private levels and ance contracts. Index insurance also allows very the index insurance markets. The World Bank timely automatic settlements, which is crucial for Commodity Risk Management Group (CRMG) effective disaster response. Price risk management already addresses the knowledge gap through contracts will be based on liquid exchange-traded technical assistance and the demonstration effects instruments, set at market prices. All indexes must of pilot transactions, but credit and market gaps be objective, transparent, published, and sustain- will limit its ability to scale up. GIIF would lower able; price indexes must be liquid. The GIIF would the entry barrier for international risk transfer by regularly publish insurable indexes. pooling smaller transactions, thereby helping to scale up risk transfer from developing countries. Exit Strategy The GIIF seeks to catalyze a commercial market Present for index-based insurance products in develop- The European Commission allocated a total of ing countries by "crowding in" the private sector. 25 million for a commodity risk management Following GIIF's start-up phase, it is expected that facility and submitted the concept to the Council the market for developing country risk will be suf- Working Group of Member States as part of the ficiently developed and competitive to offer risk "conditional billion" package, the final tranche management products to end-user countries and of the Ninth EDF/2003 to 2007. CRMG is putting clients at a reasonable cost. This period could vary together a proposal for a Global Index Insurance from seven to ten years. 7 Potential Roles for Governments and the World Bank Agricultural producers and other rural residents are often exposed to a variety of biological, geological, and climatic factors that can nega- tively affect household income and/or wealth, as well as tremendous variability in output and/or input prices. Given this environment, risk-averse individuals often make investment decisions that reduce risk exposure but also reduce the potential for income gains and wealth accumulation. Thus, risk contributes to the "poverty trap" experienced by rural people in many developing countries. For a variety of reasons (discussed in Chapter 2), markets for trans- ferring these risks are typically either very limited or nonexistent. This "market failure" has stimulated a number of policy responses. Many developed countries have highly subsidized, farm-level agri- cultural insurance programs. Critics argue that, in addition to being very expensive, these programs stimulate rent-seeking activity, are highly inefficient, and may actually increase risk exposure by encour- aging agricultural production in high-risk environments (Chapter 3). Given fiscal constraints in most developing countries, highly subsi- dized, farm-level agricultural insurance programs are not a realistic policy option. Index-based insurance products have been proposed as an alterna- tive risk-transfer mechanism for rural areas in developing countries. While not a panacea for all risk problems, index-based insurance products may prove to be valuable instruments for transferring the financial impacts of low-frequency, high-consequence systemic risks out of rural areas (Chapter 4). For a variety of reasons, however, gov- ernment intervention may be required to generate socially optimal quantities of risk transfer. Governments must carefully consider the extent and nature of any intervention in markets for index-based insurance products (Chapter 5). These efforts can be facilitated by World Bank policy advice, lending instruments, and monitoring and evaluation systems (see World Bank 2004; 2005b). This chapter sets out policy and operational implications for governments and subse- quently for the World Bank operational agenda. GOVERNMENT ROLES Risks in rural areas must be managed at the macro-, meso-, and micro- levels.Governmentsneedto(1)understandthecountry'sruralriskpro- file; (2) quantify the impact of this risk on the economy and revenues; (3) design a rural risk management framework; and (4) implement risk reduction and risk transfer.44 53 54 Managing Agricultural Production Risk Identify the Risk Profile for Private and physical and indirect losses for different types of Public Assets and Business Flows assets and economic activities? As represented in Figure 7.1, a variety of indirect business flow losses A natural risk assessment identifies the types of often compound the direct physical losses caused risks that affect major private and public assets and by natural hazards. economic activities in rural areas.45 This assessment distinguishes between micro- and macrolevel risk and considers both geographical and seasonal vari- Design a Rural Risk Management Framework ations. Identification of risks at the microlevel is Government intervention in risk transfer markets typically based on household surveys as well as specific risk surveys. The objective is to understand must be based on a careful analysis of market the types of risks that affect households and the na- shortcomings and a clear statement of how gov- ture of those risks. At the macrolevel, the assess- ernment involvement will address those short- ment would consider the aggregate economic effect comings (Chapter 5). A well-designed rural risk of household risk with a particular focus on gov- management framework clearly delineates public ernment budget exposure. and private roles in the ex ante world of risk re- duction and risk financing and also in the ex post worldofemergencyresponse.Thisframeworktakes Quantify Risk Impacts at All Levels country-specific objectives and constraints into ac- Once the major risks have been identified, govern- count instead of replicating developed country his- ments need to quantify the potential impact of torical models (Chapter 3). The objective is to learn those risks. What is the magnitude of potential from these historical examples and then to apply that understanding to country-specific efforts that incorporate new and innovative risk transfer in- struments (Chapter 4). To plan appropriately, pri- Figure 7.1 Potential Impacts of Natural Hazards vate decision makers need to know where and how government would intervene at different risk levels. Where a credible and reliable insurance cover is in Natural Hazard Risk in the Rural Sector place, for example, agricultural enterprises might intensify production. Implement a Risk Management Strategy Direct Losses Indirect Losses To be successful, a well-conceived risk management strategy must be supported by a credible govern- Fixed Capital Flow Losses ment commitment that is sufficiently funded over the long term. While appropriate government roles Public Assets Loss of Tax Revenue and will vary to reflect country-specific circumstances, · Government Buildings Tax Base one strategy might be government intermediation · Public Infrastructure Business Interruption of index-based risk management products made Private Assets Reallocation of Investments · Industrial Infrastructure available in international capital and reinsurance · Residential Infrastructure markets and government creation of infrastruc- Inventories ture to support the development and implemen- · Staple Foods/Other Crops tation of new private risk management products. · Inputs WORLD BANK ROLES Short Term Consequences Long Term Consequences Humanitarian Crisis Scarce data The World Bank can engage in a number of activi- (Attracts International Attention) (Usually not recorded) ties that, in coordination with governments, may Source: Authors. lead to increased risk-transfer opportunities for agricultural producers and other rural residents in developing countries. In general, these activities Potential Roles for Governments and the World Bank 55 include educational efforts, incorporating risk man- World Bank should routinely consider how to facil- agement into holistic rural development strategies, itate the development of risk management instru- investment lending operations designed to encour- ments and should be prepared to support this age the development of risk transfer markets, ex process through policy advice and, in some cases, ante coordination of donor responses to natural dis- lending operations. Often, this may require reform- asters, and monitoring and evaluation of the per- ing collateral,macroeconomic,orregulatorypolicies. formance of index insurance instruments. Risk management instruments using international markets, for example, cannot operate properly while exchange controls are in place. Often, local regula- Building Global Knowledge of the Index tions affecting insurance or financial markets also Approach to Agricultural Risk Management must be revised. The World Bank is uniquely placed to reach govern- Because government or World Bank involve- ments and decision makers on all continents. The ment in any risk management program may require World Bank, in general, and ARD (the Agriculture trade-offs with other means of enhancing rural de- and Rural Development department), in particular, velopment and reducing vulnerability (for example, can facilitate technology transfer across continents. irrigation, infrastructure, and so on), the program This economic and sector work of ARD will be dis- should be embedded in an overall rural develop- seminated outside the World Bank: in fiscal year ment strategy so any trade-offs can be carefully 2006. CRMG is planning Global Distance Learning weighed. This will also allow formation of linkages events that will have a component on agricultural with other rural development objectives (for exam- risk management concepts and also two workshops ple, rural finance). The overall rural development in two different regions, possibly in connection with strategy should take a holistic approach to risk weather insurance pilot project launches. Inside management, recognizing that diversification of in- the World Bank, information sharing will take come sources (remittances, off-farm employment, place mainly through "brown bag" lunches and and others) is often an important means of reduc- workshops. ing rural vulnerability. In addition to formal risk management markets, the strategy should consider what reforms are needed to encourage income Incorporating Risk Management Strategies diversification and to allow farmers a full range into Rural Development Strategy of choices in a functioning marketplace. This may Formulation and Development Policy include, for example, market liberalization and Lending Programs privatization; investments in transportation, com- While the World Bank and the IMF have a long his- munication, and market infrastructure; legal rights tory of assisting governments in dismantling un- guaranteeing market access (especially for women sustainable mechanisms for managing price risk, and ethnic minorities); provision of market infor- this often took place in the absence of alternative mation; and measures to better integrate rural and risk management tools or a clear risk management nonrural labor markets (see Siegel 2005; Lanjouw agenda for deregulated markets. This gap has con- and Feder 2001; Lloyd-Ellis 1999; and Mead and tributed to a breakdown in marketing arrangements Liedholm 1998). Attention should also be dedicated and credit channels, so that these efforts have some- to safety nets designed to minimize the need to liq- times not produced the projected results (Kherallah uidate productive assets in times of emergency and et al. 2002). While the task will be neither quick nor to be scaled up quickly and efficiently at need (see easy, the importance of addressing issues of collat- Jorgensen and Van Domelen 1999; Jutting 1999; eral policies and institutional development as inte- and Morduch 1999). gral to reform is now widely understood. While the index-based risk management tools Creating Investment Lending Operations discussed here are not a cure-all, they can help credit that Encourage Risk Management institutions, producer organizations, and (in some cases) producers to manage production risk directly; At the macrolevel, a number of World Bank instru- by doing so they can help reconnect farmers to out- ments (and those of other donors) exist or are being put and credit markets. In assisting policymakers explored that can cushion the fiscal and balance of in the design of a country's reform programs, the payments adjustments required when countries 56 Managing Agricultural Production Risk face shocks from natural disasters or international The target of the project, as currently conceptu- price movements of major commodity exports or alized, would be individual farmers, but a project imports. These include automatic mechanisms to like this could be targeted at the mesolevel as well. adjust debt service--or even to augment financing-- Pooling risk at the subregional level (a complex in response to exogenous shocks. (For a full discus- climate system) can reduce financing requirements sion, see World Bank 2004; 2005b.) by taking advantage of scale. The subregion as a At the mesolevel, risk management tools can be whole is more attractive to international insur- used to improve the functioning of government so- ance markets (due to risk-spreading) than would cial safety net programs, either at central or decen- be individual countries. Other direct benefits in- tralized levels. Index-based insurance instruments, clude the faster spread of ideas and the more ef- for example, could be used to provide ex ante fective development of capacity made possible by contingent funding that would allow safety net cross-country collaboration and the presence of programs to expand when they are most needed, preexisting regional institutions ready to support without the delays and uncertainties caused by re- project implementation. liance on budgeting processes or on external aid. Likewise, use of index-based insurance by individ- ualfarmers,associations,processors,orruralfinance Donor Coordination institutions would reduce their degree of uncer- Like farmers, governments may suffer from a form tainty and facilitate primary producers' access to of moral hazard. Donor response to catastrophes credit and input markets. can reduce the interest of the developing country In addition to policy advice, the primary World government in using markets to shift natural disas- Bank tool now being used to support development ter risk, as was the case in Nicaragua following the of risk management markets, investment lending overwhelming donor response following Hurricane projects may also be useful in some cases. Examples Mitch. Donor responses, however, cannot be pre- can be found in World Bank-facilitated price risk dicted with certainty and often are not timely. management efforts. In Turkey, for example, a com- Furthermore, the international community may modity market development learning and innova- tion loan (LIL) had the objective of first supporting overlook localized disasters, which may devastate a the development of physical commodity markets, community despite having limited impact beyond which in the long term could evolve into a domes- it. A better solution would be to take advantage of tic platform for trading futures contracts. The pro- these donations in a more structured and ex ante ject financed the upgrading of testing laboratories, fashion. Donors could, for example, contribute to an warehouse facilities, and regional market infra- insurance pool for the country or region. The World structure, and it provided technical assistance to Bank--particularly the teams in countries espe- enhance and harmonize grades and standards for cially prone to disasters--can play a leading role in some commodities, upgrade the warehouse receipts this through the consultative group process. system, and improve the operations of the commod- A special case of aid in response to disaster is ity market regulatory authority. While there Turkey food aid following a serious drought. Here, the still has no domestic futures trading, progress has need for an improved approach is particularly been made toward the more limited objectives of acute, as in-kind assistance often has counterpro- establishing better linkages between producers ductive effects in undermining development of and buyers and of encouraging forward contract- local production and marketing channels. Also, ing for spot delivery, providing another means of aid given ex post in response to droughts is often reducing price risk. In addition, the project has fa- late in arriving, forcing starving victims to liqui- cilitated more efficient price discovery: the prices date productive assets, thus perpetuating a cycle for cotton and wheat determined on two exchanges of poverty. Use of an index-based instrument to participating in the project now serve as the offi- fund emergency food aid holds the promise of a cial record of domestic market prices for those two much more rapid response, since payment would commodities. be triggered by weather events far in advance of Another project being explored focuses on the the actual food shortages, and of far less disruption establishment of a regional system of weather in- of local markets, since food aid agency payouts surance in southern Africa (see Box 7.1). would be made in cash that would be used to pro- Potential Roles for Governments and the World Bank 57 Box 7.1 Examples of Potential World Bank Investment Lending Projects to Facilitate Risk Management Global level technical assistance from international experts (including CRMG), and premium cost-sharing funds. Global Index Insurance Facility: The facility would These premium support funds would compensate consist of a capital investment in a risk-taking entity for the extra premium costs that international and that would underwrite global weather, disaster, and national insurers add in the infancy stages of the price risks in developing countries. The main devel- product and as a result of data uncertainty. These opment objective would be to absorb costs for initial premium support funds would be phased out as transactions for developing country clients through volumes increased and as the extra costs for pre- cofinancing of premiums, funded both separately and miums declined. through reinvestment of dividends by public spon- Knowledge transfer: Travel costs, expertise, design sors. The main commercial objective of the facility of methodologies and tools to quantify risk exposure, would be to generate a modest return to its share- underwriting guidelines, manuals, operational system holders through active management of a diverse port- development, and study tours. folio of developing country risk not previously Financial backing of risk-taking entities: Government transferred to the capital and insurance markets. The mediation of catastrophic risk between international facility would perform several commercial functions risk insurance markets and insurers or other risk takers providing benefits to developing countries. in the country; governments could either set up sepa- National level rate risk-taking vehicles or enter into contingent credit agreements with the World Bank to lower annual Infrastructure: Fallback stations, new weather stations, maintenance of weather stations, communications premium costs. equipment for weather services, contract with data Regional level vetting services (such as the U.K. Met Office), set-up of weather databases (online), and the cleaning and Financial contribution to a regional index insurance enhancing of weather data. fund: Pooling systemic risk at the regional level, signif- Regulatory assessment: Review of legislation, icantly lowering premium costs and warranting set-up drafting of new regulations, general policy frame- of a regional risk fund that would insure its members work review, and country-specific policy framework according to sound actuarial rates before it lays off risk review (including recommendations on subsidy in international markets. levels, national weather risk funds, basis risk matching Climate prediction and forecasting technologies: funds, and so on). Can be cost effectively rolled out only at a regional International market/pilot transactions: Travel level that achieves economies of scale and enforces to international reinsurance market contacts, collaboration. Source: Authors. cure food locally, to the extent possible, or to pay the farm level, and their development impacts need beneficiaries directly. The World Bank is collabo- to be evaluated. CRMG has launched a first base- rating with the World Food Program and other line study with DECRG (Research Department of donors to pilot such an approach. the World Bank). Generally, utility at the farm level can be gauged by the level of take-up of unsubsi- dized and unbundled products and, particularly, Monitoring and Evaluation of Transactions the level of repeat buying. Panel studies will reveal The work on index insurance in developing coun- the actual impact of these products. Indicators tries is still in an early stage, and its development are the level of inputs used and the diversification impact is not yet proven. A number of assumptions of farm activities, particularly the share of cash about the value of these instruments, their utility at crops in the overall portfolio. Another important 58 Managing Agricultural Production Risk linkage will be to gauge whether index insurance circumstances and not under others. Mistakes will products improve access to credit or improve the be made. Learning from those mistakes will require terms of credit for small farmers in developing careful evaluation and subsequent adjustments. At countries. Both the Indian and the Mongolian pilot this stage, the key value added from index insur- project have very explicit monitoring and evalua- ance products appears to be the opportunity for tion components that will attempt to gauge these structured ex ante financing of catastrophic risk tied activities. to highly correlated losses resulting from weather As with any innovation, index insurance prod- risk in agriculture. Such risk cannot be pooled at the ucts for agricultural production risk will go through local level, and the special structures introduced in some significant changes in the next few years. It is this ESW give hope that they can be shifted into likely that we will learn that they work under some global markets. References Ahsan, S. M., A. A. G. Ali, and N. J. Kurian. 1982. "Towards Preferences of Cranberry Growers." Northeastern Journal of a Theory of Agricultural Insurance." American Journal of Agricultural and Resource Economics 18 (October): 109­17. Agricultural Economics 3: 520­29. Doherty, N. A. 1997. "Financial Innovation in the Management Anderson, J. R. 2001. "Risk Management in Rural Development: of Catastrophic Risk." Paper presented at ASTIN/AFIR A Review." Rural Development Strategy Background Conference, Australia, August. Paper 7, Rural Development Department, The World Bank, Fafchamps, M. 1993. "Sequential Labor Decisions under Washington, D.C. Uncertainty: An Estimable Household Model of West African Babcock, B., C. Hart, and A. Stoppa. 2004. "La gestione del Farmers." Econometrica 61 (September): 1173­97. rischio nell'agricoltura Usa." In La Gestione Del Rischio in Gadgil, S., P. R. Seshagiri Rao, K. Narahari Rao. 2002. "Use Agricoltura: Strumenti E Politiche, ed. F. Filippis. Rome, of Climate Information for Farm-Level Decision Making: Italy: Coldiretti. Rainfed Groundnut in Southern India." Agricultural Systems Barnett, B. J. 1999. "U.S. Government Natural Disaster 74: 431­57. Assistance: Historical Analysis and a Proposal for the Future." Glauber, J. W. 2004. "Crop Insurance Reconsidered." American Disasters: The Journal of Disaster Studies, Policy & Management Journal of Agricultural Economics 86 (December): 1179­95. 23: 139­55. Goes, A., and J. R. Skees. 2003. "Financing Natural Disaster Risk Barnett, Barry J., J. Roy Black, Yingyao Hu, and Jerry. R. Skees. Using Charity Contributions and Ex Ante Index Insurance." 2005. "Is Area-Yield Insurance Competitive with Farm-Yield Paper presented at the annual meetings of the American Insurance?" Journal of Agricultural and Resource Economics Agricultural Economics Association, July 27­30, Montreal, 30 (2): 285­301. Canada. Burgaz, F. J. 2004. "Il sistema delle assicurazioni agricole com- Hardaker, J. B., R. B. M. Huirne, J. R. Anderson, and G. Lien. binate in Spagna." In La Gestione Del Rischio in Agricoltura: 2004. Coping with Risk in Agriculture, 2nd ed. Wallingford, Strumenti E Politiche, ed. F. Filippis. Rome, Italy: Coldiretti. Oxon, U.K.: CABI Publishing. Buzby, J. C., P. L. Kenkel, J. R. Skees, J. W. Pease, and F. J. Harwood, J., R. Heifner, K. Coble, J. Perry, and A. Somwaru. Benson. 1994. "A Comparison of Subjective and Historical 1999. "Managing Risk in Farming: Concepts, Research, and Yield Distributions with Implications for Multiple-Peril Crop Analysis." Economic Research Service (ERS), U.S. Department Insurance." Agricultural Finance Review 54: 15­23. of Agriculture. Byerlee, D. 2005. Agricultural Growth for the Poor: An Agenda for Hazell, P. B. R. 1992. "The Appropriate Role of Agricultural Development. Directions in Development Series. Washington, Insurance in Developing Countries." Journal of International DC: World Bank. Development 4: 567­81. Carney, D., M. Drinkwater, T. Rusinow, K. Neefjes, S. Wanamali, Hazell, P. B. R., and J. R. Skees. 1998. "Potential for Rainfall and N. Singh. 1999. "Livelihood Approaches Compared: A Insurance in Nicaragua." Report to the World Bank, March. Brief Comparison of the Livelihoods Approaches of DFID, Hess, U. 2003. "Innovative Financial Services for Rural India: CARE, Oxfam, and UNDP." Department for International Monsoon-Indexed Lending and Insurance for Smallholders." Development (DFID), London. Agriculture and Rural Development Working Paper 9, The Chambers, R., and G. Conway. 1992. "Sustainable Rural World Bank, Washington, D.C. www.itf-commrisk.org. Livelihoods: Practical Concepts for the 21st Century." IDS Hess, U., and J. Syroka. 2005. "Weather-Based Insurance in Discussion Paper 296, Institute for Development Studies, Southern Africa: The Case of Malawi." Agriculture and Rural Brighton, U.K. Development (ARD) Discussion Paper 13, The World Bank, Deaton, A., and F. Grimard. 1992. "Demand Analysis and Tax Washington, D.C. Reform in Pakistan." The World Bank Living Standards and Hess, U., A. Stoppa, K. Richter. 2003. "Weather Risk Management Measurement Survey Working Paper, No. 85, The World for Agriculture and Agribusiness in Developing Countries." Bank, Washington, D.C. In Climate Risk and the Weather Market: Financial Risk Dercon, S., ed. 2005. "Insurance against Poverty." UNU-WIDER Management with Weather Hedges, ed, R. Dischel. London: Risk StudiesinDevelopmentEconomics.Oxford:OxfordUniversity Books. Press. Jorgensen, S. L., and J. Van Domelen. 1999. "Helping the Poor ------. 2002. "Income Risk, Coping Strategies, and Safety Nets." Manage Risk Better: The Role of Social Funds." Paper pre- The World Bank Economic Research Observer 17 (Fall). sented at the IADB Conference on Social Protection and Dischel, R. 2001. "Double Trouble: Hedging." Weather Risk: Poverty, February. A Special Report. Risk Magazine and Energy and Power Risk Jutting, J. 1999. "Strengthening Social Security Systems in Rural Management August: 24­26. Areas of Developing Countries." ZEF Discussion Paper on Dismukes, R., G. Allen, and B. J. Morzuch. 1989. "Participation Development Policy. Center for Development Research, in Multiple-Peril Crop Insurance: Risk Assessments and Risk Bonn, Germany. 59 60 Managing Agricultural Production Risk Kherallah, M., C. Delgado, E. Gabre-Madhin, N. Minot, and for Agricultural Development: Issues and Experience, eds. P. B. R. M. Johnson. 2002. Reforming Agricultural Markets in Africa. Hazell,C.Pomareda,andA.Valdes.Baltimore:JohnsHopkins Baltimore: Johns Hopkins University, Press for the Inter- University Press. national Food Policy Research Institute. Priovolos, T., and R. C. Duncan, eds. 1991. Commodity Risk Kunreuther, H., R. M. Hogarth, and J. Meszaros. 1993. "Insurer Management and Finance. Washington, D.C.: Published for the Ambiguity and Market Failure." Journal of Risk and Uncertainty World Bank by Oxford University Press. 7: 71­87. Ramamurtie,S.1999."WeatherDerivativesandHedgingWeather Kunreuther, H., and M. Pauly. 2001. "Ignoring Disaster, Don't Risks."InInsuranceandWeatherDerivatives:FromExoticOptions Sweat Big Stuff." Wharton Risk Center Working Paper 01-16- to Exotic Underlyings, ed. H. Geman. London: Risk Books. HK, Wharton Business School, University of Pennsylvania. Rejda, G. E. 2001. Principles of Risk Management and Insurance, Kunreuther, H., J. Meszaros, R. M. Hogarth, and M. Sprance. 7th ed. Boston: Addison Wesley Longman. 1995. "Ambiguity and Underwriter Decision Processes." Rosenzweig, M. 1988. "Risk, Implicit Contracts and the Family Journal of Economic Behavior and Organization 26: 337­52. in Rural Areas of Low-Income Countries." Economic Journal Kurosaki, T., and M. Fafchamps. 2002. "Insurance Market 98 (December): 1148­70. Efficiency and Crop Choices in Pakistan." Journal of Rosenzweig, M., and H. P. Binswanger. 1993. "Wealth, Weather Development Economics 67 (April): 419­53. Risk and the Composition and Profitability of Agricultural Lanjouw, P., and G. Feder. 2001. "Rural Non-Farm Activities Investments." Economic Journal 103 (January): 56­78. and Rural Development: From Experience toward Strategy." Rosenzweig, M., and O. Stark. 1989. "Consumption Smoothing, Rural Strategy Discussion Paper 4, The World Bank, Migration, and Marriage: Evidence from Rural India." Journal Washington, D.C. of Political Economy 97: 905­26. Lim, Y., and R. Townsend. 1998. "General Equilibrium Models Rosenzweig,M.,andK.Wolpin.1993."CreditMarketConstraints, of Financial Systems: Theory and Measurement in Village Consumption Smoothing and the Accumulation of Durable Economies." Review of Economic Dynamics 1: 59­118. Production Assets in Low-Income Countries: Investments Little, P. D., A. P. Castro, M. P. Stone, W. Negatu, and T. Mogues. in Bullocks in India." Journal of Political Economy 2: 223­44. 2004. "Moving in Place, Drought and Poverty Dynamics in Rossi, P., J. Wright, and E. Weber-Burdin. 1982. Natural Hazards South Wollo, Ethiopia." BASIS CRSP Policy Conference: and Public Choice: The State and Local Politics of Hazard Mitigation. Combating Persistent Poverty in Africa, November 15­16, New York: Academic Press. Four Points Sheraton Hotel, Washington, D.C. Ruck, T. 1999. "Hedging Precipitation Risk." In Insurance and Lloyd-Ellis, H. 1999. "Endogenous Technological Change and Weather Derivatives: From Exotic Options to Exotic Underlyings, Wage Inequality." American Economic Review 89: 47­77. ed. H. Geman. London: Risk Books. Mahul, O., and J. R. Skees. 2005. "Managing Agricultural Schade, C., H. Kunreuther, and K. Kaas. 2002. "Low-Probability Catastrophic Risk at the Country Level: The Case of Livestock Insurance Decisions: The Role of Concern." Wharton Risk Mortality Risk in Mongolia." Working Paper, The World Center Working Paper No. 02-10-HK, Wharton Business Bank, Washington, D.C. School, University of Pennsylvania. Marshall, J. F. 1989. Futures and Options Contracting: Theory and Shiller, R. J. 2003. The New Financial Order: Risk in the 21st Practice. Cincinnati, OH: South-Western Publishing Company. Century. Princeton: Princeton University Press. Mead, D. C., and C. Liedholm. 1998. "The Dynamics of Micro Siegel, P. B. 2005. "Looking at Rural Risk Management Using an and Small Enterprises in Developing Countries." World Asset-Based Approach." Paper prepared for the Commodity Development 26: 61­74. Risk Management Group, Agricultural and Rural Develop- Miller, D., and A. Paulson. 2000. "Informal Insurance and Moral ment Department, ESW, The World Bank, Washington, D.C. Hazard: Gambling and Remittances in Thailand." Contributed Skees, J. R. 2003. "Risk Management Challenges in Rural paper No. 1463, Econometric Society World Congress. Financial Markets: Blending Risk Management Innovations http://econpapers.repec.org/paper/ecmwc2000/1463.htm. with Rural Finance." Thematic papers presented at the Miranda, M. J., and J. W. Glauber. 1997. "Systemic Risk, USAID Conference: Paving the Way Forward for Rural Reinsurance, and the Failure of Crop Insurance Markets." Finance: An International Conference on Best Practices, June AmericanJournalofAgriculturalEconomics79(February):206­15. 2­4, Washington, D.C. Morduch,J.1999."TheMicrofinancePromise."JournalofEconomic ------. 2001. "The Bad Harvest: More Crop Insurance Reform: Literature 37: 1569­1614. A Good Idea Gone Awry." Regulation: The CATO Review of ------.1995."IncomeSmoothingandConsumptionSmoothing." Business and Government 24: 16­21. Journal of Economic Perspectives 3: 103­14. ------. 1999b. "Opportunities for Improved Efficiency in Risk Morgan, M., and M. Henrion. 1990. Uncertainty: A Guide to Sharing Using Capital Markets." American Journal of Dealing with Uncertainty in Quantitative Risk and Policy Agricultural Economics 81: 1228­33. Analysis. Cambridge: Cambridge University Press. ------. 1999a. "Agricultural Risk Management or Income Orlove, B., J. Chiang, and M. Cane. 2002. "Ethnoclimatology in Enhancement?" Regulation: The CATO Review of Business and the Andes: A Cross-Disciplinary Study Uncovers a Scientific Government 22: 35­43. Basis for the Scheme Andean Potato Farmers Traditionally Skees, J. R., and B. J. Barnett. 1999. "Conceptual and Practical Use to Predict the Coming Rains."American Scientist 5: 428­35. Considerations for Sharing Catastrophic/Systemic Risks." Pease, J., E. Wade, J. R. Skees, and C. M. Shrestha. 1993. Review of Agricultural Economics 21: 424­41. "Comparisons between Subjective and Statistical Forecasts of Skees, J. R., B. J. Barnett, and J. Hartell. 2005. "Innovations in Crop Yields." Review of Agricultural Economics 15: 339­50. Government Responses to Catastrophic Risk Sharing for Peru, Government of. Hoja de Ruta (Road Map) 2004­06. Agriculture in Developing Countries." Paper prepared for Pikor, G., and A. Wile. 2004. "L'esperienza canadese nella ges- the Commodity Risk Management Group, Agricultural and tione del rischio in agricoltura." In La Gestione Del Rischio in Rural Development Department, ESW, The World Bank, Agricoltura: Strumenti E Politiche, ed. F. Filippis. Rome, Italy: Washington, D.C. Coldiretti. Skees, J. R., J. R. Black, and B. J. Barnett. 1997. "Designing and Pomareda, C. 1986. "An Evaluation of the Impact of Credit Rating an Area-Yield Crop Insurance Contract." American Insurance on Bank Performance in Panama?" InCrop Insurance Journal of Agricultural Economics 79: 430­38. References 61 Skees, J. R., and U. Hess. 2003. "Evaluating India's Crop Failure World Bank CRMG and Inter-American Development Bank Policy: Focus on the Indian Crop Insurance Program." Paper workshop on Innovative Agricultural Insurance for Central delivered to the South Asia Region of the World Bank, America, May 9­11, Antigua. November. Townsend, R. 2005. "Weather Insurance in Semi-Arid India." Skees, J. R., P. B. R. Hazell, and M. Miranda. 1999. "New Paper prepared for the Commodity Risk Management Approaches to Public/Private Crop-Yield Insurance." EPTD Group, Agricultural and Rural Development Department, Discussion Paper No. 55, International Food Policy Research ESW, The World Bank, Washington, D.C. Institute, Washington, D.C. www.cgiar.org/ifrpi. Turvey, C. G. 2002. "Insuring Heat Related Risks in Agri- Skees, J. R., and M. J. Miranda. 1998. "Rainfall Risk Contracts for culture with Degree-Day Weather Derivatives." Paper pre- Nicaragua." Report to the World Bank, Washington, D.C. sented at the AAEA Annual Conference, July 28­31, Long Skees, J. R., P. Varangis, D. Larson, and P. Siegel. 2005. "Can Beach, CA. Financial Markets Be Tapped to Help Poor People Cope with Walker, T. S., and J. G. Ryan. 1990. Village and Household Weather Risks?" In Insurance against Poverty, ed. S. Dercon. Economies in India's Semi-Arid Tropics. Baltimore: The Johns Oxford: UNU-WIDER Studies in Development Economics, Hopkins University Press. Oxford University Press. World Bank. 2005b. "Managing the Debt Risk of Exogenous Stoppa, A., and U. Hess. 2003. "Design and Use of Weather Shocks in Low-Income Countries." The World Bank, Derivatives in Agricultural Policies: The Case of Rainfall- Washington, D.C. Index Insurance in Morocco." Paper prepared for the ------. 2005a. "Rural Finance Approach." ESW report prepared International Conference on Agricultural Policy Reform and by the Agricultural and Rural Development Department, The the WTO: Where Are We Heading? June 23­26, Capri, Italy. World Bank, Washington, D.C. Swiss Re. 2004. "Understanding Reinsurance: How Reinsurers ------. 2004. "Exogenous Shocks in Low-Income Countries: Create Value and Manage Risk." Technical Publishing Series, Policy Issues and the Role of the World Bank." Report pre- Swiss Reinsurance Company, Zurich, Switzerland. www. pared for technical briefing of the Board, ARD, PREM, and swissre.com. FRM, No. 9, March 9, The World Bank. Tobben, B. (Partner Re). 2005. "Weather Risk Markets for ------. 2001. World Development Report 2000/2001: Attacking Agriculture in Central America." Paper presented at the Poverty. Washington, D.C.: The World Bank. Appendix 1 Weather Risk Management for Agriculture46 The emerging weather risk market offers new risk zens in the event of weather-related disasters and management tools and opportunities for agricul- must increase spending for rehabilitation and re- ture. The aim of this appendix is to illustrate how construction of infrastructure and assets as a result an end user in the agricultural industry could use of damage incurred. Moreover, the economy of a a market-based solution to mitigate the financial country is also at risk to weather through business impact of weather on its business operations. The interruption, supply shocks, diversion of domestic appendix draws information from the wealth of investment from productive activities to mitigation literature written on the subject of weather risk of the disasters' impacts and, for some countries, a management to provide the reader with a step-by- reduction in foreign investment in the aftermath of step guide to how weather risk management in- an extreme weather-related event. While often such struments could be developed for and used in the effects are reversible and short-term, the impact on agricultural sector. After discussing the financial the economy of a poor country can be significant impact of weather on agriculture, this Appendix ex- andlonglasting.Between1997and2001,theaverage plores the key steps required to structure a weather damagepernaturaldisasterinlow-incomecountries risk management solution, from identifying the was 5.8 percent of GDP (IMF, 2003). Evidence from risk to execution. Also discussed are the pricing of sixteenCaribbeancountriesshows,forexample,that weather risk management instruments, with a brief one percentage point of GDP in direct damage from overview of how the weather market approaches natural disasters can reduce GDP growth by half a and values weather risk and the implications for the percentage point in the same year (Auffret 2003). end user. Finally, the Appendix treats the prerequi- Furthermore, the humanitarian cost of weather- sites for weather risk management instruments: the related disasters is also greater in the developing weather data used to construct weather indexes and world: approximately 80 percent of all fatalities settle contracts and the data cleaning and analy- due to weather disasters between 1980 and 2003 oc- sis necessary when pricing and structuring a poten- curred in the "uninsured world," comprising pre- tial transaction. Selected references suggest further dominantly low-income countries (Loster 2004). reading on weather risk management. Even noncatastrophic weather events have a fi- nancial impact. The U.S. Department of Commerce estimates that nearly one-third of the U.S. econ- THE FINANCIAL IMPACT omy, or US$1 trillion (U.S. Congress 1999) is mod- ulated by the weather, and that up to 70 percent of OF WEATHER all U.S. companies are weather sensitive. Weather Weather risk impacts individuals, corporations, risk can impact a business through its overall prof- andgovernmentswithvaryingdegreesoffrequency, itability or simply through the success or failure of severity, and cost. Around the world, people face the an initiative as a consequence of the weather. Like vagaries of the weather on a daily basis. The media governments, businesses can face both demand- continually reports catastrophic weather events-- and supply-driven weather risks. Energy compa- floods, hurricanes, and droughts--that impact indi- nies, for example, can be exposed to demand-driven viduals' property, health, and lives. Consequently, weather risk. In the event of a warmer than average governmentsarealsofinanciallyexposedtoweather winter, for instance, gas companies, in particular risk. They are called upon to provide direct finan- those dealing with domestic customers, face a po- cial, nutritional, and housing support to their citi- tential drop in gas sales as customers use less gas 63 64 Managing Agricultural Production Risk than expected to heat their homes. Therefore, even from such ancillary risks that create unpredictable if the company has adhered to prudent price risk earnings streams. Just as interest rate and currency management practices by protecting their sales risks are currently managed through market-based margin from fluctuations in the gas supply price, solutions, weather risks that increase business un- weather-driven demand fluctuations can lead to certainty can now be neutralized, allowing a com- a drop in sales volume below expected levels that pany to focus on its core business and to protect significantly affects budgeted revenues. A supply- earnings per share forecasts and growth. side example of weather risk can be found in the construction industry. Because building materi- THE WEATHER MARKET als have specific weather requirements, cold and wet weather conditions can impact construction In 1997, a formal weather risk market was born in progress; concrete, for example, cannot be poured the United States through the first open market de- in wet or below-freezing conditions. Contractors rivative transaction indexed to weather. Motivated must assume this supply-driven weather risk, by the deregulation of the energy industry, which which can significantly delay a construction project led to the break-up of regulated monopolies in elec- and result in hefty penalties if the project is not tricity and gas supply, the nascent weather market completed on schedule. responded to energy companies' need to increase Weather has traditionally been the scapegoat in operational efficiency, competitiveness, and share- business for poor financial performance (Clemmons holdervalue.In1996,theKansas-basedenergycom- 2002). Annual reports, financial statements, and pany, Aquila, entered into a transaction with New press releases frequently contain declarations such York-basedConsolidatedEdisonthatcombinedtem- as, "[c]ooling degree days were 21 percent below perature and energy indicators, protecting the lat- last year's quarter and 16 percent below normal. ter against a cool August that would reduce power The effects of milder weather compared with last sales. The first publicized transaction in 1997, how- year had a negative impact on [earnings before in- ever, was between energy companies Koch Energy terestandtaxes]ofabout$35millionforthequarter" and Enron. Additional deals soon followed, with (Duke Energy 2003); "4 cents per share [decline] other energy market participants wanting protec- for lower gas deliveries due to warmer weather tion against risks, primarily temperature, associated in the fourth quarter of 2003" (Energy East 2004); with volumetric fluctuations in energy. and "Europe's performance continued to be im- In 2001, the Weather Risk Management Asso- pacted by unfavorable summer weather with vol- ciation (WRMA)--the industry body--commis- ume down 12 percent in the third quarter and sioned PricewaterhouseCoopers (PWC) to conduct year-to-date volume down 6.5 percent" (Coca-Cola a survey of weather risk contracts executed among 2004). Given such examples, it is not surprising that WRMA members and survey respondents from the financial community has begun to seek prac- October 1997 to March 2001 and since then on an tical solutions to controlling the financial impact annual basis. Since 1997, the survey has shown that of weather. Centrica Plc, for example, one of the over US$20 billion has been transacted through largest domestic gas suppliers in Great Britain, is the weather risk market47: the market has grown to one of a number of utilities that has chosen to man- around US$4.6 billion outstanding risk for the year age its weather risk in order to "protect the com- April 2003 to March 2004 (PWC 2003; 2004; see pany against variability in earnings of its gas retail Figure A1.1), although some believe this to be an business due to abnormal winter temperatures in underestimate.48 Active trading occurs in U.S. the UK" (Ulrich 2002), and it has been doing so since European, and Japanese cities (Figure A1.2); most 1998. London-based Corney and Barrow Wine Bars notable among the few transactions occurring out- Limited deploys several weather hedges to provide side these three main trading hubs are agricultural financial protection against cool summers resulting transactions in Mexico, India, and South Africa. The in poor customer patronage: "After the exceptional market has also evolved to include nonenergy appli- summer of 2003 Corney and Barrow was keen to cations. Survey respondents, when asked to list re- secure protection against the possibility of the re- quests received from potential end users of weather verse experience [in 2004]" (XL Trading 2004). With risk management products, identified end users in the emergence of a market for weather risk man- the retail, agriculture, transport, and leisure and agement products, a business can now be protected entertainment industries (Figure A1.3), although Appendix 1. Weather Risk Management for Agriculture 65 energy still contributes approximately 56 percent Figure A1.1 Notional Value of All Weather Contracts of the potential weather risk management end user in US$ market. As a result of this expansion, the market has also broadened its product offering to include transactions on nontemperature indexes49 such as 5,000 rainfall, wind, and snow. CME contracts Today, the key market participants include 4,500 Non-CME contracts (re)insurers, investment banks, and energy compa- nies. (Re)insurers and investment banks provide 4,000 weather risk management products to end user customers--such as Corney and Barrow Wine Bars ) 3,500 n Limited and Centrica Plc--and form the primary oilli market; all three participate in a secondary mar- m 3,000 S ket in which players transfer weather risk among U $( themselves through over-the-counter (OTC) finan- e 2,500 ul a cialtransactionsandexchange-basedderivativecon- v l a tracts on the Chicago Mercantile Exchange (CME)50 n 2,000 oit to diversify and hedge their portfolios. o N Weather risk management is also being intro- 1,500 duced to the developing world through the work of 1,000 organizations such as the World Bank Commodity Risk Management Group (CRMG) and the United 500 Nations World Food Program (WFP). The World Bank was involved in the first index-based weather 0 risk management program--in India in June 2003-- 1998/ 1999/ 2000/ 2001/ 2002/ 2003/ and it is currently working on several projects 1999 2000 2001 2002 2003 2004 around the world. The small pilot program was Weather trading year launched by Hyderabad-based microfinance insti- Source: Author's figures, using PricewaterhouseCoopers industry tution BASIX and the Indian insurance company survey data from 2003 and 2004. ICICI Lombard, in conjunction with CRMG, when 230 groundnut farmers in Andhra Pradesh bought weather insurance to protect against low monsoon rainfall (Hess 2003). Currently the WFP, in con- Indeed 13 percent (PWC 2004) of the end user re- junction with the World Bank, is investigating the questsintheweathermarketarenowfocusedonthe feasibility of weather-based insurance as a reliable, agricultural sector (Figure A1.3). Weather affects timely, and cost-effective way of funding emer- manyaspectsoftheagriculturalsupplyanddemand gency operations in countries such as Ethiopia (The chain. From the supply side, weather risk manage- Economist 2004). Work is also underway to see if ment can help control both production or yield risk developingcountrygovernmentsinsouthernAfrica and quality risk. canbenefitfromweatherriskmanagementproducts Technology plays a key role in production risk and strategies (Hess and Syroka 2005). The global in farming. The introduction of new crop varieties weather-risk market is particularly interested in and production techniques offers the potential for these types of transactions, as they provide much improved efficiency; however, agriculture is also sought after diversification to their books through often affected by many uncontrollable events re- new locations and risks. lated to weather--including excessive or insuffi- cient rainfall, hail, extreme temperatures, insects, WEATHER RISK AND and diseases--that can severely impact yields and AGRICULTURE production levels. Countless examples can be given on the impact of cold temperatures on deciduous One of the most obvious applications of weather fruit (Guaranteed Weather 2005b), deficit rainfall risk management products, weather insurance or on wheat (Stoppa and Hess 2003), excess rainfall weather derivatives is in agriculture and farming. on potato yields (Meuwissen et al. 2000), and even 66 Managing Agricultural Production Risk Figure A1.2 Percentage of Total Weather Contracts by Location (excluding CME trades) 100 )st 90 c art n 80 o c E M 70 C . c x E( 60 st c art 50 n o c l 40 at ot f o 30 e g at n 20 e cr e P 10 0 1998/1999 1999/2000 2000/2001 2001/2002 2002/2003 2003/2004 Weather trading year NA West NA South NA Mid-West NA East Asia Europe Other Source: Author's figures, using Pricewaterhousecoopers industry survey data from 2003 and 2004. temperature stress on cattle and thus dairy produc- Figure A1.3 Potential End User Market by Economic tion (Guaranteed Weather 2005a). In 2003, 59 per- Sector 2003­2004 cent of Ukraine's winter grain crop was destroyed due to winterkill temperatures (USDA 2003) and 40 to 50 percent of northeastern England's oil rape- seed crop was lost due to excessive rain at harvest in August 2004 (BBC 2004). The costs associated Agriculture 13% with drops in expected or budgeted production due to such uncontrollable factors can have a signifi- Retail cant impact on a producer's revenues and contrac- 9% tual obligations. A producer may seek protection against adverse weather conditions affecting crop Construction Energy 7% yield. Weather can also impact the quality, if not the 56% absolute production levels, of a crop (Guaranteed Weather 2005c). Transportation 4% On the demand side, weather also affects related Other agricultural products through the use of pesticides, 11% fertilizers, and herbicides. Agricultural chemical producers, for example, can use weather risk man- agement instruments to hedge against the costs Source: Author's figures, using PricewaterhouseCoopers industry associated with fluctuations in the demand for survey data from 2004. chemicals by farm operators. The cotton boll weevil, for example, which costs cotton producers in the Appendix 1. Weather Risk Management for Agriculture 67 United States US$300 million per year,51 is a weather gions at risk from weather and the weather stations sensitive pest; its numbers vary from year to year that reflect that risk; identifying the time period largely due to the severity of the winter. In ex- during which risk is prevalent; and identifying tremely cold winters, weevil numbers drop sig- the weather index providing the best proxy for the nificantly, directly affecting the net earnings of an weather exposure. This last step is the most critical agrochemical company. Chemical producers could in designing an index-based weather risk manage- hedge their earnings volatility caused by fluctua- ment strategy. Rather than measuring the actual tions in pesticide sales by purchasing a weather risk impact on crop yields--or related fluctuations in management instrument specifically indexed to the demand, supply, or profitability--the index acts as phenology of the pests their products target. a proxy for the loss experienced due to weather and Index-based weather insurance is a relatively is constructed from actual observations of weather new product, and the use of weather risk manage- at one or more specific weather stations. ment products in the agricultural sector is still in its Location and Duration infancy, with very few publicized transactions in the United States and Europe. A number of agri- All weather contracts are based on the actual ob- cultural transactions have occurred outside of the servations of weather variables at one or more spe- main weather market trading hubs, however, most cific weather stations. Transactions can be based on notably in Canada (Ontario--maize; Alberta-- observations from a single station or a basket of forage), Argentina (Sancor--dairy), South Africa several stations or on a weighted combination of (Gensec Bank--apple cooperative freeze cover), readings from multiple stations. (More information and India (ICICI Lombard--groundnut, cotton, on the weather station and data requirements for coriander, and orange). Given weather is one of weather risk management instruments appears the biggest risks faced by farmers, weather-indexed below.) If an individual farmer is interested in pur- risk management products have been suggested as chasing weather protection for his particular crop, a potential alternative to the traditional crop insur- the index-based weather contract must be written anceprogramsforsmallholderfarmersintheemerg- on the weather station nearest the farmer's land to ing markets. provide the best possible coverage for the farmer client. A larger grower, with several production re- gions, may be more interested in purchasing a STRUCTURING A WEATHER RISK weather contract based on several weather stations MANAGEMENT SOLUTION to reflect the weather conditions in all areas cov- ered by the business. The grower's risk manage- Developing a successful weather risk management ment strategy can be either to purchase a weather and transfer program for agriculture involves four contract on each of the identified weather stations essential steps: or to purchase a single contract on a weighted aver- · Identifying significant exposure of an agricul- age of several stations, with the weightings chosen tural grower/producer to weather; to reflect the importance of the different stations to · Quantifying the impact of adverse weather on the overall weather exposure of the business. The their revenues; approach chosen depends on the risk preferences · Structuring a contract that pays out when and risk retention appetite of the grower, although adverse weather occurs; and weighting is generally the cheaper and more effi- · Executingthecontractinoptimalformtotrans- cient approach. Retaining localized risks will most fertherisktotheinternationalweathermarket. probably be a more cost-effective solution than would transferring them to a third-party, while still Each of the steps is outlined in the following four providing protection in situations where adverse subsections, and they are fully explored in the case weather affects several regions and involves the studies in the next appendix. overall production portfolio of a producer. The lat- ter approach will also reduce the risk of reliance on one weather station and hence the associated issue Identifying the Risk of basis risk,52 covered below. Identifying weather risk for an agricultural grower All contracts have a defined start and end date or producer involves three steps: identifying the re- to limit the period over which the underlying index 68 Managing Agricultural Production Risk is calculated. This calculation period describes the ity risk by purchasing a contract that pays if a spec- effective dates of the risk protection period during ified undesirable weather event occurs or a spec- which relevant weather parameters are measured ified desirable weather fails to occur. The index at the specified weather stations. For agricultural possibilities are limitless and flexible to match the end users, the duration of the weather contracts will exposure of the agricultural grower or producer, as be determined by the specific requirements of their long as the underlying data are of sufficient qual- business. Contract duration is flexibility to address ity. A few examples of weather indexes for specific individual end-user business exposures; contracts agricultural exposures appear below. Although the can be weekly, monthly, seasonal, and even multi- examples are based on temperature and precipita- annual. Final settlement of the weather contracts tion, the principles apply to all weather parameters typically occurs up to forty days after the end of recorded by ground-based meteorological weather the calculation period, once the collected weather stations. More examples are given in the case studies data have been cross-checked and quality con- in Appendix 2. trolled by the relevant data-collecting body, usu- ally the National Meteorological Service.53 Example 1: Growing Degree Days Growing Degree Days (GDDs) is a common index Underlying Indexes used in the agricultural sector, similar to HDDs and A weather index can be constructed using any com- CDDs in the energy sector. GDDs are a measure- bination of measurable weather variables and any ment of the growth and development of plants number of weather stations that best represent the (both crops and weeds) and insects during a grow- risk of the agricultural end user. Common variables ing season. Organisms that cannot internally regu- include temperature and rainfall, although transac- late their own temperature are dependent on the tions on snowfall, wind, sunshine hours, river flow, temperature of the environment to which they are relative humidity, and storm/hurricane location exposed. Development of an organism does not and strength are also possible and are becoming occur unless the temperature is above a minimum more frequent. Unlike energy indexes, in which the threshold value, known as the base temperature, relationshipbetweenenergydemandandweatheris and a certain amount of heat is required for devel- more transparent and is linked primarily to temper- opment to move from one stage to the next. The ature, weather indexes for agriculture demonstrate base temperature varies for different organisms and more complex, albeit still quantifiable, relationships is determined through research and scientific con- between crop yields or pesticide use. siderations. A GDD is calculated by the following The normal process for designing an index-based equation: weather insurance contract for an agricultural grower, for example, involves identifying a mea- Daily GDD = max(0,Taverage - L); surable weather index strongly correlated to crop yield rather than measuring the yield itself. After Taverage = (Tmax - Tm ) 2 (1) in gathering the weather data, an index can be de- signed by (1) looking at how the weather variables where L is the baseline temperature and Taverage is the have or have not influenced yield over time; (2) dis- daily mean temperature, defined as the average of cussing key weather factors with experts, such as the daily maximum (Tmax) and minimum (Tmin) tem- agrometeorologists and farmers; and/or (3) refer- peratures. If this average is greater than the thresh- ring to crop growth models using weather vari- old temperature L, the GDD accumulated for that ables as inputs for yield estimates or phenology day is the threshold temperature minus the daily models illustrating how weather variations relate average temperature. If the daily average tempera- to pest development. A good index must account ture is less than the base temperature, then the GDD for the susceptibility of crops to weather factors for that day is zero. Adding the GDD values of con- during different stages of development, the biolog- secutive days gives the accumulated GDDs over ical and physiological characteristics of the crop, a specific period. Accumulated GDDs are a good and the properties of the soil. If a sufficient degree proxy for establishing the development stages of of correlation is established between the weather a crop, weed, or insect and can give an indication index and crop yield or quality, a farmer or an agri- as to the development and maturity of a crop or the cultural producer can insure his production or qual- proper scheduling of pesticide or herbicide appli- Appendix 1. Weather Risk Management for Agriculture 69 cations. Measuring the amount of heat accumu- of Kherson is often unstable, hence complete winter lated over time provides a physiological time scale wheat crop failure due to winterkill is a potential that is biologically more accurate than are calendar risk in the southern steppe zone of Ukraine; the crop days (Neild and Newman 2005), and specific or- usually dies in years with no snow cover or when ganisms, pest or plant, need different accumulated the stable snow cover appears late in winter, as it GDDs to reach different stages of development. By did in 2003. A winterkill index, based on days when comparing accumulated GDD totals with those of the daily minimum temperature is less than -16°C, previous years, it can be seen if a normal amount of could therefore be used by a farmer to obtain pro- heat energy has been made available to a crop. In tection against such crop failure risk. A farmer could general, assuming adequate moisture supplies are enter into a contract with the recovery of the full available, the total GDDs received by the end of the value of the crop, as expected under normal weather growing season are often related to crop yield, and conditions, if the recorded daily minimum air tem- therefore GDDs can be a good index for crop pro- perature is less than -16°C for four or more con- duction. The cumulative temperature index can be secutive days at any time during the winter period used to establish a relationship between GDDs and from November to March. production and thus ultimately with a producer's revenues. Example 3: Deficit Rainfall and Drought Meteorological drought is usually defined in terms Example 2: Event-based Indexes of deviation of precipitation from normal levels Crop damage can also be the result of specific or and duration of a region's dry periods. Agricultural critical temperature events that can be detrimen- drought refers to situations in which soil moisture tal to yield or quality. Freezing conditions, for in- content no longer meets crop growing needs in an stance, were reported to have caused more than area due to insufficient rainfall. Crops, particu- US$600 million in damage to the U.S. citrus crop in larly rain-fed crops, often have a minimum overall a single week of December 1998, with US$300 mil- threshold of cumulative rainfall necessary for suc- lion occurring in Tulare County, California, alone cessful and healthy plant development. Dry beans, (Guaranteed Weather 2005b). Critical temperatures for example, can consume up to 368 mm of water causing crop damage may vary depending on the during the growing season, depending on plant va- length of time that temperatures remain below riety, soils, climate, and weather conditions (Efetha freezing as well as on the variety, health, and devel- 2002). For dry-land corn farming, 450 to 500 mm or opment stage of a plant. Preventative and proactive more of rainfall during the growing season is re- measures can often be taken to protect crops from quired for high yields (Neild and Newman 2005). such events, but these may have limited impact or These water requirements must be met by natural become more difficult for crops that are farmed in rainfall, stored soil moisture from precipitation large areas, such as cereals and grains. prior to the growing season, or supplemental irri- Winter wheat yields at harvest, for example, de- gation. Therefore, a deficit of rainfall below these pend to a great extent on how well the plants sur- levels, in the absence of irrigation, can cause plant vive the winter hibernation period. In the territory moisture stress that affects development and re- of Kherson, in Ukraine, winter wheat crops have duces yields. A simple cumulative rainfall index been known to die when air and therefore soil tem- can be developed to suit a grower's specific insur- peratures fell below a critical level for one day or ance requirements with regard to such decreases in longer. These winterkill events cause damage and rainfall and yield. Looking at historical yield data, death of the plants' tillering node: "[with little or no for example, can establish an empirical relationship snow, plants begin to die when] the daily minimum between seasonal cumulative rainfall and yield. The air temperature drops below -16 deg C; [a crop can distribution of rainfall during the growing season be completely lost if this happens for] four days or at specific stages of a plant's development is in a row or in the minimum temperature drops often more important than total rainfall, however, below -21 deg C" (Adamenko 2004). Snow cover and customized indexes must be developed to cap- considerably improves conditions of winter wheat ture this risk (Stoppa and Hess 2003). Such indexes hibernation, as the difference between air and soil may also include other weather parameters, such temperature increases from 0.5 to 1.1°C per cen- as temperature and relative humidity. Crop growth timeter of snow cover. Snow cover on the territory models or historical yield data can be used to infer 70 Managing Agricultural Production Risk the empirical relationship between rainfall amounts vest. Producers with fixed-price delivery contracts and yield/quality for specific soil and crop types. or those using price risk management instruments toprotectthemselvesfrommarketfluctuationsinthe price of their crop at harvest time know the financial Quantifying the Risk value of each kilogram or metric ton they produce Once the index has been identified, it must be cali- and hence can quantify the financial cost of a short- brated to capture the financial impact of the speci- fall in production. If a grain producer, for example, fied weather exposure as measured by the index. knows he will receive $X per metric ton of crop, the Two approaches are possible at this stage: identify- following relationship must hold for his change in ing the financial exposure per unit of the defined revenue: index, and/or establishing the limit, the total finan- cial protection, required per risk period, that is, the Revenue = X × (Actual Yield - Expected Yield) × H maximum payout necessary in a worst-case sce- = X × V = ± X × H × a(I) × I (3) nario. The approach chosen depends on the nature of the underlying index and weather event. If the A good weather hedge must offset the negative weather exposure is event driven, for example, Revenue fluctuation in the event of a drop in yield such as a Category 5 hurricane hitting a particular from budgeted levels if a producer is to protect his location or a cold winterkill event destroying an earnings. In order to perfectly replicate his position, entire wheat crop, the latter approach is more ap- the farmer could enter into a weather contract with propriate. If the weather exposure is of a cumula- the following incremental payout P per unit index: tive nature, such as drought or Growing Degree Days,theformerapproachshouldbechosen.Taking P = X × H × a(I) × I (4) intoconsiderationthemaximumprotectionrequired per risk period can also inform the financial expo- sure per unit index. Therefore, his overall position would be: Unit Exposure Revenue + AP = -X × Y × H After developing weather indexes to capture the + X × H × a(I) × I = 0 (5) impact of adverse weather conditions on a specific crop's yield, it is straightforward to calculate the Producers may have contractual obligations to de- financial impact of these events for producers. In liver a predefined amount of their farmed product designing the index, expert scientific agrometeo- to a buyer at harvest time, with associated penalties rological assessments, either in conjunction with if these obligations are not met. In such a situation, crop model output or with verification using his- it would be straightforward to quantify and struc- torical yields, have been employed to construct an ture a hedging product to protect producers from underlying index that best proxies the weather sen- these contractual costs in the event of weather- sitivity of the crop in question. Having identified related shortfalls in production. the index, the crop yield, Y, or volume, V, variabil- ity per unit of the defined index, I, can be defined, The Limit as follows: Most weather contracts have a limit, which corre- sponds to the maximum financial payout or recov- Y = V H = a(I)I (2) ery from the contract in a worst-case scenario, such as a complete crop failure. The maximum payout where, a(I) is some function of I that relates the can be set by either considering the value-at-risk for index to the yield Y, and H is the planting area of the the producer in the event of a total crop failure or crop. In order to calibrate an appropriate weather by looking at historical index, production, and contract, the variation in crop yield must now be sales data to find the worse-case scenario histori- converted into a financial equivalent that mirrors cally in order to establish a limit. Alternatively, a the producer's exposure. This can be done, for ex- producer may simply want to insure his produc- ample, by considering a producer's production and tion and input costs in order to recover these out- input costs per hectare planted or by considering his lays if the crop fails. If a producer's production expected revenue from the sale of the crop at har- costs are $Z per hectare farmed, $Z will therefore Appendix 1. Weather Risk Management for Agriculture 71 correspond to the maximum payout, the limit of · The term, T: the risk protection period of the the weather contract, for each hectare the producer contract, including the start and end date of wishes to insure. The unit exposure P will therefore the contract; be as follows: · A strike, K: also known as an attachment level, the level at which the weather protec- P = (-Y Expected Yield) × Z tion begins; = (a(I)I Expected Yield) × Z, for Y < 0 · The payout rate, X: the financial compensation (6) per unit index deviation above (call) or below (put) the strike at maturity, defined as the unit Structuring the Product exposure in the previous section; and · The limit, M: the maximum payout per risk Structure Type protection period. Once the index has been identified and calibrated, the next step is to structure a contract that pays The payout, Pcall, of a call option can be defined when the specified adverse weather occurs, thus using the following equation: performing a hedging or risk-smoothing function for the agricultural grower or producer. Derivative Pcall = min(max(0, I - K) × X, M) (7) and insurance products form the mainstay of the weather risk management market. While the two The payout, Pput, of a put option can be defined as instruments feature different regulatory, account- follows: ing, tax, and legal issues, the risk transfer charac- teristics and benefits are often the same. One of the Pput = min[max(0,K - I) × X, M] (8) drivers of market growth has been the flexibility between both instruments and the possibility of tai- The type of option purchased depends on the risk loring risk management solutions to a client's needs profile of the buyer. Assume, for example, a winter (CorballyandDang2002).Ariskmanagementprod- wheat grower loses 4 percent of his expected yield uct can be either of the following: every day that the maximum daily temperature · A traditional insurance-style product, that is, rises above 30°C in the months of May and June, risk transfer that results in downside protec- incurring a cost per day per hectare of 16. The tion in exchange for a premium; for example, grower has 10,000 hectares of wheat under cultiva- a call or put option structure. Or, tion and is prepared to accept yield losses due to · Arisk-exchangederivative-basedproduct,that heat stress of up to 480,000, but he wants protec- is, a product based on giving away upside in tion for any losses in excess of that amount. In this goodyearsorseasonstofinancedownsidepro- case, the grower may consider purchasing a call op- tection; for example, a collar or swap structure. tion, either in derivative or insurance form, with the following specifications: Call and Put Options Reference Weather A call option gives the buyer of the option the right, Station (RWS): Growerstown, ID No. 12345 but not the obligation, to buy the underlying index Index: Daily Tmax > 30 C, measured at a predefined level at the maturity, or end date, of at RWS the contract.54 In exchange for this right, the buyer Calculation Period: 1 May 2005 to 30 June 2005 pays a premium to the seller. Similarly, a put op- (inclusive) tion gives the buyer the right, but not the obliga- Call Strike: 3 events tion, to sell the underlying index at a predefined Payout Rate: 160,000 per event above the strike level at contract maturity; in exchange for this right, Limit: 1,600,000 the buyer of the option pays a premium to the seller. Every option contract and, in general, most weather To secure such protection the grower must pay a contracts are defined by a set of standard specifica- premium, but he is allowed to recover 160,000 for tions including: each day in May and June that the daily maximum · The reference index, I, and weather station(s): temperature exceeds 30°C in excess of the strike complete specification of the index and data level. Figure A1.4 illustrates the impact of such a used to construct it; hedging strategy on the revenues of the grower: by 72 Managing Agricultural Production Risk if the weather is beneficial for the business. In Figure A1.4 Call Option Payout Structure and Wheat essence, the business can forego a portion of profit Grower's Losses to offset the cost of reduced revenues by selling a put option and then buying a call option from the provider, or vice versa. A collar, therefore, com- 2,000,000 bines both a call and put option, but it does not in- t 1,500,000 u volve an exchange of premium from the end user o Call option y a payout structure to the provider. A collar is a means by which two p )s 1,000,000 t c or parties can exchange risk; hence, collars may often art u E( 500,000 n be structured with asymmetric call and put options o )ts c o r c 0 to make the risk exchange of equal value to both e ht m a ui parties. This approach may not be applicable to all e ­500,000 w m Net losses of grower weather risk management problems in agriculture. d er with weather hedge n p ­1,000,000 a g Furthermore, businesses may be averse to giving s ni ess d up profits in a good year. A very simple example of ol ul ­1,500,000 c a possible application can be found by considering r x e E(­2,000,000 w a local agrochemical company whose sales of a par- or Grower losses due G ­2,500,000 ticular pesticide vary depending on the number of to heat events pest growing degree days (PGDDs) recorded in ­3,000,000 0 2 4 6 8 10 12 14 their sales region during the winter. When the Number of heat events recorded PGDDs are high, pest attack incidents in- (daily Tmax > 30° C) in May-June crease, and pesticide sales increase accordingly. Source: Authors. When PGDDs are low, demand for pesticides drops and sales are low. The company has quantified this risk and finds that, on average, it loses or gains $12,000 per PGDD from budgeted revenues if the accumulated PGDDs are below or above the 1700 purchasing the call option, his downside exposure PGDDs expected in the region's normal winter. The is now limited to 480,000, unless the number of company may be interested in a collar agreement heat events exceeds an unprecedented 13 during because, not only is it costless to enter into, it also the calculation period. Modifications can obviously reduces the company's weather related revenue be made to this simplified example to better repli- volatility. In this case, the company may consider cate the exposure of the grower; a more sophisti- purchasingacollarwiththefollowingspecifications: cated product may be based, for instance, on an Reference Weather index that considers only consecutive days of ex- Station (RWS): Growerstown, ID No. 12345 cessive temperature, includes relative humidity, or Index: Cumulative PGDDs measured establishes a nonlinear payout rate that increases at RWS compensation as the number of heat events during Calculation Period: 1 November 2005 to 31 March the calculation period increases. Alternatively, the 2006 (inclusive) Call Strike: 1800 PGDDs grower many want to purchase a digital call option, Put Strike: 1600 PGDDs an all-or-nothing structure that will pay the grower Payout Rate: $12,000 per PGDD above/ a lump sum, rather than incremental payouts, if the below strikes heat stress reaches a critical level at which most of Limit: $2,400,000 the crop will be lost. Similarly, an end user buying a put option would protect himself from events The historical distribution of November to March when the index drops below the strike level. PGDDs in Growerstown is found to be symmetric around the 1700 PGDD average with a standard Collars and Swaps deviation of 100 PGDDs; hence the call and put op- A business may be averse to paying an upfront pre- tions have strikes equidistant of the average to cre- mium for risk protection. An alternative is a con- ate a zero-cost collar. Figure A1.5 illustrates the tract in which the business receives downside impact of such a hedging strategy on the revenues protection in return for sacrificing upside revenue of the company: the collar reduces a potential two Appendix 1. Weather Risk Management for Agriculture 73 standard deviation fluctuation in revenues for the Figure A1.5 Collar Payout Structure and Agrochemical company from +/- $2,400,000 to +/- $1,200,000. Company's Deviation from A swap is a contract in which a buyer makes a Budgeted Revenue payment to the seller when a weather index rises aboveapredefinedstrikelevelandentitlesthebuyer to receive a payment from the seller when the index 5,000,000 ) falls below the same level. Essentially, a swap is a $( ) Deviation from d put and a call option with the same strike, payment m 4,000,000 n a ui budgeted revenue rate, and limit, which, like a collar, is costless to e m u er 3,000,000 n Collar payout p enter. In the example above, rather than using a col- e v t structure er c lar contract, the local agrochemical company could d art 2,000,000 et n o "sell" a swap contract to a provider with a strike of e c g d g 1,000,000 1700PGDDsandapayoutrateof$12,000perPGDD. u ni b d This would ensure that the business achieves no s' ul y c 0 n x a more or less than its budgeted revenue. Swaps are E( p t m u ­1,000,000 derivative OTC contracts that are commonly traded o o c y a in the secondary derivative weather risk market; m p t ­2,000,000 orf c they are rarely used outside the energy industry, n art oit n ­3,000,000 however, as they do not always offer the best corre- o ai c Deviation from budgeted v r lation to the underlying risk. Swaps are only avail- e e D ht ­4,000,000 revenue with weather hedge a able in derivative form (Raspe 2002). e w ­5,000,000 1350 1450 1550 1650 1750 1850 1950 2050 Exotic Structures Cumulative PGDDs from November-March In theory, a weather risk management solution can Source: Authors. take any form or combination of options, swaps, triggers, and indexes. Possible exotic combinations include knock-in or knock-out options, which grant the buyer a standard call or put option if a partic- tially provide total revenue insurance for agricul- ular knock-in or knock-out threshold is breached, tural producers whose revenues depend on both either on the same or a different index (for example, the price at which they sell their produce and the a heat stress call option for wheat that is only trig- volume they produce. Such contracts exist and are gered if precipitation during the same calculation traded in the OTC energy derivatives markets. period drops below a critical level); compound op- tions, known as "an option on an option," that grant Risk Retention and Premium the buyer the right to purchase an underlying op- It is clear that an important aspect to consider when tion at some future date (for example, a multiyear structuring an index-based solution is the retention structure that gives the buyer an option to buy an of risk by the party seeking protection. This means option on the weather conditions for the next grow- defining the index trigger level at which the weather ing season at the end of the current season); and protection begins. The strike determines the insured structures with a variable start date depending on party's level of risk retention and is the key to pric- the timing of a pre-specified event (such a structure ing and success in transferring the risk. A strike may be appropriate for crops with variable plant- very close to the mean of the index indicates a low ing dates that can be associated with cumulative level of risk retention by the end user and a high rainfall or growing degree day totals). probability that the contract will pay out. This im- Reference indexes may also include nonweather plicitly means a large premium, as well as the pos- variables. Temperature contingent commodity call sibility of inspiring little interest in the weather options, for example, may give a purchaser the right market if the location or nature of the risk is outside but not the obligation to buy an underlying com- the main liquid trading hubs or variables. A strike modity at a prespecified price and volume only if farther away from the mean reduces the probabil- certain temperature, that is, growing conditions, ity of a payout and hence the premium of the con- have been met. Such exotic structures could poten- tract, as the entity is retaining the more frequent, 74 Managing Agricultural Production Risk near-the-mean risk internally and transferring less secondary OTC and exchange-traded deriva- to the market. The level of risk retention will depend tives market. on the risk appetite and business imperatives of the · Specialized hybrid companies or funds. These end user and the sensitivity to the premium associ- include organizations such as Coriolis Capital ated with entering into a contract. To reduce the (formerly Société Générale) and Guaranteed premium payment, for instance, the wheat grower Weather Trading Ltd., which were established in the call option example above could increase the specifically to trade and invest in weather risk. strike for heat stress events. By retaining more risk, Such hybrid entities deal in weather deriva- all things being equal, the producer would reduce tives and reinsurance and offer weather risk the premium of the contract. Alternatively, the solution products to customers. grower could reduce the payment rate to partially, Theenergycompaniesresponsibleforthebirthofthe instead of fully, hedge his exposure. Premium pay- marketplace--Enron, Aquila, Southern Company, ment terms must be defined before entering a and Entergy Koch (now Merrill Lynch)--are no weather contract, and an overview of how such longer active in the weather market. Although the contracts are priced by weather market providers market is still predominantly driven by energy re- appears in the following section. lated weather risk, with energy companies and several banks hedging their energy portfolios with weather derivatives, the major source of secondary Execution market liquidity is now driven by the three pre- The Market Providers dominant types of counter-party outlined above, through the hedging of end-user deals or the taking The main providers of risk capacity, product struc- of speculative positions. turing, and/or pricing for end-user customers in the current weather risk market can be categorized into Regulatory Issues three main groups: Depending on the jurisdiction, weather risk man- · Insurance and reinsurance companies that agement products can be classified as financial (de- view noncatastrophic weather insurance as a rivative),insurance,orgamingcontracts.Depending natural extension of their traditional business on their classification, these contracts are subject to and given analysis capabilities. Examples specific tax and accounting treatments, which can include ACE, AXA, Munich Re, Partner Re, render one form more optimal than another for an Swiss Re, Tokio Marine and Fire Insurance, end user's purposes and business. Interested parties and XL Capital. Most of these entities also offer are strongly advised to contact their local financial derivative products and, although some may services authority, insurance regulator, or a profes- choose to retain the risk by dealing in a large sional specializing in insurance law to find out how amount of diversified end-user business, sev- weather contracts are treated in their jurisdiction eral are among the most active portfolio man- and the legal and financial implications associated agers in the secondary market, using financial with each (Raspe 2002). derivatives contracts to manage their weather risk portfolios, including both high- and low- VALUING WEATHER RISK frequency risk. Pricing Overview · Banks that structure weather risk solutions to fit the needs of their clients. Examples in- The premium of an index-based weather contract is clude ABN AMRO, Calyon, Deutsche Bank, determined actuarially by conducting a rigorous Goldman Sachs, Merrill Lynch, and Rabobank. analysis of the historical weather to reveal the sta- Banks have a large potential client base for tistical properties and distribution of the defined weather derivative products and may find weather index and, therefore, the payouts of the in- many marketing and cross-selling opportuni- surance or derivative contract. Such an analysis in- tiesinmanydifferentsectorsofbusiness.Banks cludes (1) cleaning and quality control of the data, generally do not have as much risk capacity as that is, using statistical methods to in-fill missing do the (re)insurers; they often pass the posi- data and/or to account for significant changes, if tions of their end-user customers to other mar- any, as a result of instrumentation or station loca- ket providers or actively hedge positions in the tion changes; (2) checking the cleaned data for sig- Appendix 1. Weather Risk Management for Agriculture 75 nificant trends and detrending to current levels if These three parameters quantify the expected appropriate (this is particularly pertinent for tem- (a) and variable or risky (b, c) payouts of the con- perature data, which, in general, exhibit a strong tract and must be determined from the historical warming trend in the Northern Hemisphere); and weather data, either by using the historical index (3) performing a statistical analysis on the cleaned values from the available cleaned and detrended and detrended data and/or a Monte Carlo simula- dataset or by using the data to calibrate a Monte tion, using a model calibrated by the data, to deter- Carlo simulation model to generate thousands of mine the distribution of the defined weather index possible realizations of I in order to fill out the and the subsequent payouts of the contract. By de- distribution of payouts and to determine better termining the frequency and severity of weather estimates of E(P), (P), and VaRX(P). A complete events specified by the index, an appropriate pre- description of the various methods for determining mium can be calculated. these payout statistics are beyond the scope of this It should be noted that the premium charged by appendix, but an overview of possible approaches providers in the weather market may depend on appears in the following subsection. It is clear, several factors, not all as objective as the underlying however, that E(P), (P), and VaR99(P) will vary statistical analysis of the weather data. Institutions with the strike, payout rate, and limit. charge different risk margins, or discounts, over Having established values for the expected and the expected value or fair price to potential buyers; variable payout parameters, the price of a contract thesechoicesaredrivenbytheriskappetite,business is then determined by the risk preferences of the imperatives, and operational costs of the provider (re)insurance company or financial institution pro- (Henderson et al. 2002). An overview of pricing is viding the risk protection: that is, by how they mea- given in this section, and the implications of the sure the cost of risk with respect to return for the premium charged for the end user will also be dis- purposes of pricing, risk management, and capital cussed. The data issues associated with points 1 and allocation (Henderson et al. 2002). As a result, this 2 above will be covered later in this Appendix. aspect of the risk pricing process is the most sub- Expected Loss and Risk Margin jective, as it is largely driven by the institutional constraints and risk appetite of the provider. It To illustrate the pricing process, an index-based is clear, however, that the provider will charge weather contract is structured as a call option (see E(P) plus an additional risk margin for taking the above). The payout, P, of the contract is determined weather risk from the end user, that is, by the following equation: P = min[max(0,I - K) × X, M] Premium = E(P) + Risk Margin (10) (9) There are many methods for measuring risk and where K is the strike, I is the index measured during hence for determining a risk taker's risk margin. the calculation period, X is the payout rate per unit Two examples of simple methods that have been index, and M is the limit of the contract. To calculate suggested (Henderson et al. 2002) for the weather the premium for the contract, one must determine market are the Sharpe Ratio and the Return on VaR; the following parameters: both measure expected excess return in terms of · The expected loss of the contract, E(P), that is, some measure of risk and hence determine the "cost theaverageorexpectedpayoutofthestructure of risk" for the contract seller. each year; · The standard deviation of the payouts of the Sharp Ratio, = [Premium - E(P)] (P) contract, (P), that is, a measure of the vari- ability of the contract payouts; and Premium = E(P) + (P) (11) · The xth-percentile of the payouts, that is, a measure of the value-at-risk (VaR) of the con- Return on Var(99%), = [Premium - E(P)] tract for the seller, VaRX(P). The 99 percent VaR, for example, represents the economic [VaR (P) - E(P)] 99 loss for the provider that is expected to be Premium = E(P) + exceeded, with 1 percent probability, at the end of the calculation period of the contract. [VaR99 (P) - E(P)] (12) 76 Managing Agricultural Production Risk The Sharpe Ratio uses standard deviation as the pected (E(P)) and variable (VaR99(P), (P)) payouts underlying measure of risk; therefore represents of the contract must be determined. This section the "cost of standard deviation" as determined by briefly outlines three possible approaches, repre- the seller's risk preferences. One of the benefits of senting varying degrees of difficulty and effort, relating risk to the standard deviation of payouts is commonly used by weather market participants. In that it constitutes an easy parameter for estimating; general, providers may use several or all of these however, it is a symmetric measure of risk captur- methods to crosscheck results and compute a con- ing the mean width of the payout distribution, and, tract price. for traditional risk exchange products, the payout distribution is often not symmetric but has a long Historical Burn Analysis tail. The Return on VaR method uses VaR(99%) as Historical Burn Analysis (HBA) is the simplest the underlying measure of risk and therefore rep- method of weather contract pricing. It involves resents the "cost of VaR." Value-at-Risk (VaR) is a taking historical values of the index, which may term that has become widely used by insurers, cor- be based on raw, cleaned, and possibly detrended porate treasurers, and financial institutions to sum- weather data, and applying the contract in question marize the total risk of portfolios. The advantage of to them. Assuming the data used to calculate the VaR99 is that it is computed from the loss side of the historical indexes are of good quality for the risk payout distribution, where loss is defined with re- analysis, HBA can give a useful and intuitive first spect to the expected payout E(P), and therefore indication of the mean and range of possible pay- captures the potential financial loss to the seller. outs of a weather contract from which parameters Using the Return on VaR method is more appropri- such as E(P) and (P) can be calculated. The method ate for pricing structures that protect against low- is simple and can easily be done in a spreadsheet. frequency/high-severity risk, which have highly The disadvantage of HBA is that it gives a limited asymmetric payout distributions. VaR99 is a harder view of possible index outcomes: it may not capture parameter to estimate, however, particularly for the possible extremes, and it may be overly influ- strike levels set far away from the mean, and it is enced by individual years in the historical dataset. usually established through Monte Carlo simula- Estimates of parameters such as VaR99(P) therefore tion. The worst-case recorded historically can often become very difficult, although the largest historical be used as a crosscheck for VaR. In both methods outlined above, and quantify the risk loading value is always a good reality check when consider- ing the possible variability of payouts. Additionally, appropriate for the risk preferences of the provider. the confidence level that can be attached to averages It is also worth noting that weather market par- ticipants can often enter into financial derivatives and standard deviation calculated from historical contracts to manage their weather risk portfolios data is limited by the number of years of data avail- and actively hedge positions in the secondary OTC able. The standard error in the average decreases and exchange-traded derivatives market. This is as the number of years included in the average in- particularly true if the end-user risk is in a location creases, however; although weather stations often included in or positively correlated to locations have thirty to forty years of historical data, the rep- commonly traded in the market. Moreover, even if resentative nature of older data for today's weather a market provider chooses to retain the risk inter- and climate should also be questioned (see below). nally, a new potential contract may look attractive in comparison to the overall portfolio of the risk Historical Distribution Analysis taker; that is, it may be a contract that, like hedg- Much can be gained from understanding the statis- ing, will reduce the relative and VaR99 parame- tical properties of the underlying index. If index ters and the overall risk position of the portfolio values are calculated using historical meteorologi- and hence reduce or increase the premium while cal data, then looking at the distribution of these maintaining the same cost of risk . A reasonable index values and ascertaining the probability dis- estimate for and , given prices in the weather tribution function of the index will provide a better market, are = 15­30% and = 5­10%. estimate of the parameters necessary to specify that function and, therefore, the expected and vari- Approaches to Pricing Weather Risk able payouts of the contract. Historical Distribution In order to price a weather contract, given the over- Analysis (HDA) involves determining the probabil- view above, the parameters that quantify the ex- ity distribution that best fits the historical (possibly Appendix 1. Weather Risk Management for Agriculture 77 detrended)indexdata.Theprocessisverymuchone rated in the pricing process though the E(P) and of trial and error, and various standard tests and possibly (P) terms by their dependence on E(I) goodness-of-fit statistics, each with strengths and and (I). The weather market actively follows fore- weaknesses, can be used to pick the best distribution cast information and will modify its estimates of from a potential selection; these include Quantile- E(I) and (I) based on historical information if nec- Quantile plots, calculation of moments, and statisti- essary (Jewson 2004b). Complex daily simulation cal tests such as chi-squared, Kolmgorov-Smirnov, methods can also be used. Building models that cor- Anderson-Darling, root-mean squared error, and rectly capture the physical relationships between maximum likelihood methods. By determining the many meteorological variables at many sites at a distribution and therefore the parameters necessary daily resolution poses significant scientific, mathe- to define it, such as the mean and standard devia- matical, and programming challenges (Brody et al. tion, the E(P) and (P) VaR99(P) can be calculated 2002), however, and should be required only for either by simulation from the distribution (see path-dependent contracts or nonlinear structures below) or analytically, depending on the type of that depend on several variables or critical daily distribution chosen and the underlying complexity values. of the contract to be priced. Closed form solutions can be derived for call and put options using dif- End-User Perspective ferent underlying distributions (Jewson et al. 2005), such as the Normal distribution, kernel density, and On receiving a price quotation for a weather risk Gamma distribution. Although the HDA method is management solution from a market provider, an more accurate than HBA for computing expected agricultural grower or producer must decide if, and variable payouts (Jewson 2004a), and often sim- given the price, such a solution is the best strategy pler due to the availability of analytical formulas, it for the business to manage its weather risk. Some assumes the underlying distribution is a correct of the advantages and disadvantages for end users representation of the data. Fitting and putting too of using a market-based risk management tool are much emphasis on a distribution that does not cap- highlighted below. A grower can take many tech- ture the higher moments of variability, for exam- nical and practical measures to make crops more ple, can lead to an underestimate of variability and, resilient to the vagaries of the weather; examples therefore, the premium. include better irrigation systems, new strains of seed, or new farming technologies. Likewise, an Monte Carlo Simulation agricultural product sales company, for example, Once a distribution is identified to represent an may choose to diversify into other products to re- index, constraints associated with the length of the duce their overall exposure to a particular weather historical data record are no longer valid, and thou- event. Although such strategies will not be covered sands of realizations of the index can be simulated, in this appendix, end users should consider the to estimate the contract statistics to any arbitrary relative cost and efficiency of choosing such ap- degree of statistical accuracy, using the distribu- proaches over an insurance or derivative weather tion to make Monte-Carlo simulations. The Index based-solution. Ideally, the end user should focus Simulation (IS) method is commonly used for pric- on the most cost-efficient and effective means for ing weathercontracts.Indexvaluescanbesimulated dealing with weather risk by determining the opti- statistically by drawing samples from the chosen mal interaction of risk retention, risk transfer, and distribution to generate large numbers (years) of potential operational strategies to create a compre- artificial index values. The weather contract struc- hensive risk management solution. ture is applied to each of these values to create a range of payout outcomes that can be used to cal- Revenue Volatility and Value-at-risk culate the price of the contract. The IS method is From an agricultural end user's perspective, the particularly good for cumulative contracts, such cost of E(P) is essentially already embedded in the as GDDs, or for contracts that depend on several business: it is the average annual cost (loss) of weather variables where the correlation between weather inherent in running the business in ques- these variables can be included in the simulation tion, be it farming a crop in a particular region or process.AnadditionaladvantageoftheISandHDA selling a specific agrochemical product. In other methods is that weather forecasts can be incorpo- words, without protection, the grower or producer 78 Managing Agricultural Production Risk can expect to lose this amount on average each tection, particularly in a worst-case scenario, for year. Therefore, the premium the grower or pro- his business. This can, to a certain extent, be quanti- ducer ultimately pays for a weather risk manage- fied with historical information. The relevant ques- ment product is only the risk margin charged by the tion the end user should consider is whether the provider over the expected loss. This is illustrated payout from a risk management contract based on by the schematic below (Figure A1.6). By purchas- a weather index effectively reduces the end user's ing a tailored weather hedge, an end user receives value-at-risk (VaR); in other words, the end user a reduction of revenue volatility due to weather, should determine whether the contract reduces but at a cost: the risk margin. Reducing the volatil- the potential for economic loss with a given prob- ity at an appropriate cost, however, increases the ability within a given time horizon (Hull 2000). A return per unit risk, or the quality of earnings of grower or producer's VaR is an effective measure the end user. of the overall vulnerability of the business to exter- Obviously, the end user must also consider the nal shocks, be it price movements or fluctuations in efficacy of the weather hedge and decide whether supply and demand for his product. Weather pro- the risk management contract offers adequate pro- tection that limits a business's potential downside Figure A1.6 Schematic of Historical Revenues of a Business and the Impact of Weather Hedging Unhedged expected revenue The risk margin Assuming a contract is priced by actuarial without weather protection methods, if the annual premiumwas equal to the expected loss of the contract, then, e u on average, the payout of the contract n Hedged expected revenue would equal the premium over time and e v with weather protection the unhedged and hedge expected e revenue would be the same. R Unhedged Value-at-Risk without Weather Protection Hedged Value-at-Risk with Weather Protection Time Source: Authors. Appendix 1. Weather Risk Management for Agriculture 79 revenue exposure reduces the end user's overall the same rainfall patterns or totals during the calcu- VaR. Minimizing VaR also has the associated cost-- lation period as do the locations an end user wishes the risk margin--but it raises the question as to toprotect.Forthisreason,weathermarketproviders whether a business could withstand extreme sys- do not offer contracts based on hail; hail is a highly tematic shocks and their ramifications without pro- localizedmeteorologicalphenomenon,andalthough tection, limiting losses in catastrophic years. it can be indexed to an observing weather station, The birth of the weather market has created an such indexing may not be an effective risk man- opportunity for businesses to attain protection on agement strategy for an end user. Although histor- their income statements from the impact of noncat- ically an index and losses may correlate strongly-- astrophic weather variations. Previously, traditional showing that an index could be used as an under- insurance products dealt primarily with losses lying trigger to indemnify losses in an insurance affecting the balance sheet by protecting physical contract (see above)--a good correlation is not a assets from damage due to catastrophic weather. guarantee that the underlying contract payout will A business that protects its revenues and, as a re- match the actual loss experienced. Basis risk, there- sult, has a less volatile revenue stream may benefit fore, which can often be minimized by effective or by receiving, for example, a lower cost of debt or intuitive structuring and by using local stations an increased access to credit and, for public com- (Hess and Syroka 2005), is always an issue when panies, potentially improved stock valuations or dealing with an index-based risk management solu- stronger credit ratings (Malinow 2002). Eliminating tion. A potential basis risk outcome can be quanti- the uncertainty associated with noncatastrophic fied by using historical data; however, the key point weather-related risk allows an operation to con- to consider, as outlined above, is the efficacy of the centrate on its core business and focus on control- hedge and the effective reduction in the insured lable targets and growth. These benefits associated party's overall operational VaR (Hess 2003). with reducing revenue volatility and VaR, in rela- tion to the effective cost of hedging, are considera- WEATHER DATA tions for the end user. Just like the weather market Data Requirements providers, end users must also decide how they value risk in relation to return in the context of their In order to implement a successful weather risk business. It must define how much risk it is willing management program, the data used to construct to hold and the budgeted cost at which it is willing the underlying weather indexes must adhere to to do so. strict quality requirements, including reliable and trustworthy on-going daily collection and report- Basis Risk ing procedures; daily quality control and cleaning; A major concern with index-based weather risk an independent source of data for verification, for managementproductsisbasisrisk:thepotentialmis- example, GTS (Global Telecommunication System) match between contract payouts and the actual loss weather stations; and a long, clean, and internally experienced. On considering weather-index insur- consistent historical record permitting proper actu- ance as a product for growers, Skees (2003) writes, arial analysis of the weather risks involved (at least "[t]he effectiveness of index insurance as a risk man- thirty years of daily data are ideal). agement tool depends on how positively correlated The premium associated with weather risk man- farm yield losses are with the underlying area yield agement strategies is based on a sound actuarial or weather index." As with the regulatory concerns analysis of the underlying risk. The commercial risk regarding the definition of insurance (described taker will charge a premium reflecting the given above), this statement relates to the question of probability and severity of specific weather events; whether insurance based on a weather index can hence the quality of historical and on-going weather substituteforatraditionalcropinsurancepolicyand data is paramount. Nearly all weather contracts indemnify the grower for his losses. are written on data collected from official National Basis risk is a concern with all weather variables, Weather Service (NWS) weather stations; ideally, but it is particularly important for rainfall, which these will be automated stations reporting daily to exhibits a high degree of spatial and temporal vari- the World Meteorological Organization (WMO) ability. The weather station on which a weather GTS providing data in the internationally recog- contract is based, for example, may not experience nized standard format that then undergo standard 80 Managing Agricultural Production Risk WMO-established quality control procedures. End current values. The methods of cleaning and ad- users without access to weather data satisfying the justing data often involve statistical procedures be- above criteria, or living in areas in which the spatial yond the scope of this appendix. An awareness of coverage of a NWS weather station network may the possible need for cleaning and adjusting data is not be sufficient to fully represent their weather risk recommended, however, and the approaches used profiles, may not able to benefit from weather risk are briefly outlined below. Cleaned and adjusted management solutions. datasets can also be purchased from private ven- All contracts traded in the active secondary OTC dors with proprietary data estimation models, such derivative market are based on climatic weather as RMS and AIR. data collected and published by the NWS of the country in question. Historical climate data, and Detrending Data daily up-dates, can be purchased from each NWS, a list of which can be found on the WMO website Meteorological data often contain trends that arise (www.meteo.org/wmo). In the United States, for due to natural climate variability, urban heating ef- example, the primary source of weather data is the fects, or the impact of global warming. Regardless National Climatic Data Center. In Great Britain, of the cause, in some circumstances it may be use- weather data can be purchased from Weather- ful to be able to remove such trends from the data. xchange (www.weatherxchange.com), a joint ven- Such a procedure is known as detrending. The aim ture with the U.K. Met Office set up to support the of detrending data for pricing weather risk is to ob- European weather derivatives market. Weather- tain better estimates or forecasts of E(I), (I), and xchange provides quality-controlled historical cli- VaRX(I) based on historical data. Warming trends, mate and SYNOP datasets across Great Britain and for instance, can significantly impact the defining has distribution rights to data from several NWS parameters of the underlying data. By failing to ac- organizations across Europe, including those of count for such trends, E(I) may be significantly un- Germany, Italy, France, Netherlands, Austria, and derestimated and (I) overestimated, which can Spain. Data can also be purchased from private lead to mispricing of contracts that settle based on data vendors, such as Risk Management Solutions/ future data. Many different mathematical methods EarthSat (www.rms.com, www.earthsat.com) and exist for detrending data, each based on a different Applied Insurance Research (AIR; www.air.com). set of assumptions. Private vendors often offer additional value-added In essence, the aim of detrending is to statistically services such as data cleaning and adjusting (see model the underlying process by decomposing a below). dataset into a deterministic trend and a stochastic noise term around the trend: Cleaning and Adjusted Data D(t) = Y(t) + (t),(t) ~ N(0,2 ) (13) Despite the NWS quality control procedures, data from some meteorological observing stations may where,D(t)istheprocessrepresentedbythedataset, still have missing and erroneous values. Stations Y(t) is the deterministic and therefore predictable may also have undergone instrumentation and/or component, (t) is a normally distributed noise com- station location changes, which can introduce sys- ponent with a mean of zero, and standard devia- tematic changes to a historical dataset. A station tion and t is unit time. Determining how much of moved from a rural to an urban location, for ex- the historical data variability is attributed to Y(t) ample, may now be in a location several degrees gives an indication of how well a particular model warmer than before, creating an artificial jump in represents the underlying data. The method and ap- the station's historical temperature record. Records proach chosen for detrending data can be highly of station or instrumentation changes are usually subjective, and the decision to detrend or not should kept by the NWS for each weather station. For data be informed by some underlying criteria (Jewson to be usable for pricing weather risk management and Penzer 2004). Choosing a detrending method products, the raw data must be cleaned to correct that is better than another at predicting future data for errors and missing values and checked and per- values--or even not detrending at all--is prefer- haps adjusted for nonclimatic inhomogeneities that able to using a method that increases uncertainty could make the historical data unrepresentative of in predicting future values. The performance of Appendix 1. Weather Risk Management for Agriculture 81 different methods can be compared by consider- dations for pricing weather risk. Another good ing characteristics of the distribution of errors in source of articles and information on weather deriv- the predictions they make. By using the historical atives and the market is the Artemis website at data to back-test various detrending methods and http://www.artemis.bm and the industry body approaches, estimates of the uncertainty around website,theWeatherRiskManagementAssociation, the trend can be found and can inform the error http://www.wrma.org. The Guaranteed Weather associated with a particular method for estimating website at http://www.guaranteedweather.com/ future values. casestudies.phpincludesawealthofcasestudiesand Identifying trends and their cause is itself a sub- weather risk management examples. Information on jective process, however, and care should always be weather risk management in the developing world takentocheckthesensitivityofdetrendingresultsto can be found at http://www.itf-commrisk.org. the underlying method used. Crosschecking several detrending methods and approaches and visually REFERENCES sense-checking the data are always recommended. The weather market often uses the ten-year average Adamenko, T. 2004. "Agroclimatic Conditions and Assessment of an index as a quick first-guess estimate for E(I). of Weather Risks for Growing Winter Wheat in Kherson Oblast." The World Bank Commodity Risk Management The simplest and most commonly used method Group (CRMG) and International Finance Corporation for detrending data is polynomial detrending. The PartnershipEnterpriseProjects(IFC-PEP).UnpublishedReport aim of this method is to fit a polynomial function from the Ukrainian Hydrometeorological Centre, Kiev, July. of time to a meteorological dataset, usually a first- Auffret, P. 2003. "High Consumption Volatility: The Impact of Natural Disasters." World Bank Working Paper 2962, The order polynomial trend--a linear trend--that fits a World Bank, Washington, D.C. straight line through a set of data points (Weisstein Banks, E., ed. 2002. Weather Risk Management: Markets, Products 2002). Examples of other detrending techniques and Applications. New York: Palgrave Macmillan. BBCNews.2004."FarmersAbandonRottingHarvest."21August. include the moving average (Henderson et al. 2002), http://news.bbc.co.uk/1/hi/uk/3585638.stm. LOESS (Cleveland 1979), and low-pass filter (Von Brody, D. C., J. Syroka, and M. Zervos. 2002. "Dynamical Pricing Storch and Zwiers 1999) methods. of Weather Derivatives." Quantitative Finance 2 (June). Clemmons,L.2002."IntroductiontoWeatherRiskManagement." In Weather Risk Management: Markets, Products and Application, FURTHER READING ed. E. Banks. New York: Palgrave Macmillan. Cleveland, W. S. 1979. "Robust Locally Weighted Regression and Smoothing Scatterplots." Journal of the American Statistical Information for this appendix was taken from the Association 74 (1979): 829­36. wealth of sources available on the subject of weather Coca-Cola Enterprises Incorporated. 2004. "Coca-Cola Enter- risk management. These works are strongly recom- prises, Inc. Reports Third-Quarter 2004 Results." News mendedtointerestedreadersforfurtherinformation Release, 28 October. http://ir.cokecce.com/releaseDetail.cfm? ReleaseID=146706. and discussion. An excellent in-depth introduction Corbally, M., and P. Dang. 2002. "Risk Products." InWeather Risk to the weather market can be found in Banks (2002), Management: Markets, Products and Applications, ed. E. Banks. Weather Risk Management: Markets, Products, and New York: Palgrave Macmillan. Dischel, R., ed. 2002. Climate Risk and the Weather Market. Applications. Information on general weather risk London: Risk Books. and weather risk management issues can be found Duke Energy. 2003. "Duke Energy Reports Third Quarter in Dischel (2002), Climate Risk and the Weather 2003 Results." News Release, 30 October. http://www. Market, and Geman (1999), Insurance and Weather duke-energy.com/news/releases/2003/Oct/2003103001f.asp. The Economist. 2004. "Hedging against the Horsemen." Derivatives. The Social Science Research Network 11 December. at http://www.ssrn.org offers a large quantity of Efetha, A. 2002. "Irrigation Management of Dry Beans." papers and articles on many aspects of weather Agriculture, Food and Rural Development, Government of Alberta, June. http://www1.agric.gov.ab.ca/$department/ derivatives, including analytical pricing methods, deptdocs.nsf/all/irr4426?opendocument. simulation models, detrending methods, and the Energy East Corporation. 2004. "Energy East Corporation use of forecasts. The papers written by Dr. Stephen Announces 2003 Financial Results." News Release, 30 January 2004.http://www.corporate-ir.net/ireye/ir_site.zhtml?ticker= Jewson are particularly recommended and can be EAS&script=410&layout=-6&item_id=490266. found on http://www.stephenjewson.com. The Geman, H., ed. 1999. Insurance and Weather Derivatives: From recent publication, Weather Derivative Valuation: The Exotic Options to Exotic Underlyings. London: Risk Books. Meteorological, Statistical, Financial and Mathematical GuaranteedWeather.2005c."BreweryBarleyRiskManagement." Agriculture Weather Risk Resources Case Study. Acquisition Foundations (Jewson et al. 2005), contains an excel- date: April 2005c. http://www.guaranteedweather.com/ lent summary of this work and the theoretical foun- page.php?content_id=26. 82 Managing Agricultural Production Risk ------. "Freeze Risk to Citrus Crops." 2005b. Agriculture of Economics and Risk Management in Agriculture, March Weather Risk Resources Case Study. Acquisition date: 23­25, Gulf Shores, Alabama. http://www.guaranteed- April 2005b.http://www.guaranteedweather.com/page.php? weather.com/page.php?content_id=29. content_id=27. Neild, R. E., and J. E. Newman. 2005. "Growing Season ------. "The Effects of Temperature Stress on Dairy Production." CharacteristicsandRequirementsoftheCornBelt."Agronomy 2005a. Agriculture Weather Risk Resources Case Study. Extension, National Corn Handbook, Report NCH-40, Purdue Acquisition date: April 2005a. http://www.guaranteed- University. Acquisition date: April 2005. http://www.ces. weather.com/page.php?content_id=25. purdue.edu/extmedia/NCH/NCH-40.html. Henderson, R., Y. Li, and N. Sinha. 2002. "Finance and PricewaterhouseCoopers (PWC). 2004. Annual Weather Risk Accounting Treatment." In Weather Risk Management: Markets, Management Association (WRMA) membership survey Products and Applications, ed. E. Banks. New York: Palgrave on weather data, National Economic Consulting Group, Macmillan. PricewaterhouseCoopers, Washington, D.C.http://www. Hess, U. 2003. "Innovative Financial Services for Rural India." wrma.org/about.html. Agriculture and Rural Development (ARD) Working Paper ------. 2003. Annual Weather Risk Management Association 9, The World Bank, Washington, D.C. (WRMA) membership survey on weather data, National Hess, U., and J. Syroka. 2005. "Weather-Based Insurance in Economic Consulting Group, PricewaterhouseCoopers, Southern Africa: The Case of Malawi." Agriculture and Rural Washington, D.C. http://www.wrma.org/about.html. Development (ARD) Discussion Paper 13, The World Bank, Raspe, A. 2002. "Legal and Regulatory Issues." In Weather Risk Washington, D.C. Management: Markets, Products and Applications, ed. E. Banks. Hull, J. 2000. Options, Futures and Other Derivatives, 4th ed. Upper New York: Palgrave Macmillan. Saddle River: Prentice-Hall International. Skees, J. R. 2003. "Risk Management Challenges in Rural International Monetary Fund (IMF). 2003. "Fund Assistance for Financial Markets: Blending Risk Management Innovations Countries Facing Exogenous Shocks." Prepared by the Policy with Rural Insurance." Paper presented at Paving the Way Development and Review Department in consultation with Forward for Rural Finance: An International Conference on Area Finance and Fiscal Affairs, 8 August. Best Practices, June 2­4, Washington, D.C. Jewson,S.2004a."ComparingthePotentialAccuracyofBurnand Stoppa, A., and U. Hess. 2003. "Design and Use of Weather Index Modelling for Weather Option Valuation." Working Derivatives in Agricultural Policies: The Case of Rainfall Paper, Social Science Research Network Electronic Paper Index Insurance in Morocco." International Conference on Collection, 10 January. http://ssrn.com/abstract=486342. AgriculturalPolicyReformandtheWorldTradeOrganization: ------. 2004b. "Making Use of the Information in Ensemble Where Are We Heading, June 23­26, Capri, Italy. www.eco- Weather Forecasts: Comparing the End to End and Full stat.unical.it/2003agtradeconf/Contributedpapers/Stoppaan Statistical Modelling Approaches." Online e-print archives, dHess.PDF. Cornell University Library, 20 September. http://arxiv.org/ Ulrich, J., 2002. "Managing Director Centrica Energy Risk PS_cache/physics/pdf/0409/0409097.pdf. Management Group." Press Release, November. http:// Jewson, S., A. Brix, and C. Ziehmann. 2005. Weather Deriva- www.artemis.bm/html/press_releases/extpress58.htm. tive Valuation: The Meteorological, Statistical, Financial and U.S. Congress. 1999. William Daley, Commerce Secretary, re- Mathematical Foundations. Cambridge: Cambridge University marks to Congress in 1998, quoted in M. Golden and E. Silliere, Press. "Weather derivatives are becoming a popular hedge." Wall Jewson, S., and J. Penzer. 2004. "Weather Derivative Pricing Street Journal, 2 February 1999. and a Preliminary Investigation into a Decision Rule for U.S. Department of Agriculture (USDA). 2003. "Ukraine: Detrending." Working Paper, Social Science Research Extensive Damage to Winter Wheat." Online report, Network Electronic Paper Collection, 11 November. http:// Production Estimates and Crop Assessment Division ssrn.com/abstract=618590. (PECAD), Foreign Agricultural Service, 23 May. http:// Loster, T., Munich Re. 2004. "Risking Cost of Natural Disasters www.fas.usda.gov/pecad2/highlights/2003/05/Ukraine_ and Their Impacts on Insurance." Paper presented at the Trip_Report/. ProVention Consortium International Conference, October, Von Storch, H., and F. W. Zwiers. 1999. Statistical Analysis in Zurich, Switzerland. Climate Research. Cambridge: Cambridge University Press. Malinow, M. 2002. "Market Participants: End Users." In Weather Weisstein, E. W. 2002. "Least Squares Fitting." From Risk Management: Markets, Products and Applications, ed. MathWorld--A Wolfram Web Resource, 16 September. E. Banks. New York: Palgrave Macmillan. http://mathworld.wolfram.com/LeastSquaresFitting.html. Meuwissen, M. P. M., M. A. P. M. van Asseldonk, and R. B. M. XL Trading Partners. 2004. "Corney and Barrow Hedges Huirne. 2000. "The Feasibility of a Derivative for the Potato Weather Exposure with XL Trading Partners Ltd." Press re- Processing Industry in the Netherlands." Report based on a lease,26May.http://www.artemis.bm/html/press_releases/ presented paper at the Meeting of the Southern Association extpress124.htm. Appendix 2 Case Studies of Agricultural Weather Risk Management This appendix presents four case studies--from nualCHUbelowThresholdCornHeatUnit(TCHU) Canada, Mexico, India, and Ukraine--showing the levelatthespecifiedweatherstation.TheCHUindex successful application for agricultural end users of falls into the Growing Degree Day category, dis- weather risk management insurance and derivative cussed briefly in Appendix 1, and represents the en- products. The first section of this appendix focuses ergyavailableforthedevelopmentofcorn.Giventhe on the Agriculture Financial Services Corporation smallwindowforagriculturalproductioninCanada, (AFSC), the Canadian financial crown corporation the availability of sufficient solar energy is vital for of Alberta that has been offering Growing Degree the development of this crop. The CHU is estimated Day products to maize farmers in the province since from daily maximum and minimum temperature, 2000. The second section covers Agroasemex, the beginning on May 15 each year. The Celsius-based Mexican agricultural reinsurance company that has formula used to calculate daily CHUs is defined as been using weather derivatives to manage agricul- follows (Brown and Bootsma, 1993): tural portfolio risk since 2001. The third section pre- sents two case studies from the recent work of the CHU = 0.5 × Ymin + 0.5 × Ymax (1) World Bank Commodity Risk Management Group in developing agricultural weather risk markets in Ymin = 9 5 × [Tmin - 4.4] (2) India and Ukraine. The Technology Application Case Studies described at the end of this appendix Ymax = 3.33 × [Tmax - 10.0] - 0.084 × [Tmax - 10]2 (3) briefly outlines the principles of the AFSC program to insure grassland for pasture on an index basis using satellite imagery and the grassland insurance where Tmin and Tmax are the daily minimum and program in Spain. maximum temperatures, respectively. The daily CHU values are calculated from these temperatures. The daytime relationship involving INDEXED-BASED INSURANCE FOR Tmax, uses 10°C as the base temperature (if Tmax FARMERS IN ALBERTA, CANADA is less than 10, its value is set at 10) and 30°C as the optimum temperature, as warm-season crops do The AFSC Case Study not develop when daytime temperatures fall below 10°C and develop at a maximum rate at around Corn Heat Unit Insurance 30°C. The nighttime relationship involving Tmin The Corn Heat Unit insurance program is aweather uses 4.4°C as the base temperature below which index-based insurance product offered by the AFSC daily crop development stops. (If Tmin is less than toprotectfarmersagainstthefinancialimpactofneg- 4.4, its value is set at 4.4.) The CHU value is calcu- ative variations in yield for irrigated grain and silage lated by taking into account the functional relation- corn. The contract is designed to insure against lack ship between daytime and nighttime temperatures of Corn Heat Units (CHU) over the growing sea- and the daily rate of crop development, as shown in son. It has been offered on a pilot basis since 2000 Figure A2.1. The nighttime relationship is a straight and was planned to last until 2005. The program is line (Equation 2), while the daytime relationship ap- scheduled for a thorough evaluation to assess its pears as a curve that records greater CHUs at 30°C impact over the next year. The index has been de- than at higher or lower temperatures (Equation 3). signed to indemnify the policyholder against an an- The accumulation of CHU stops on the first day on 83 84 Managing Agricultural Production Risk buying the insurance policy, farmers must elect the Figure A2.1 Relationship Between the Daily Rate of dollar coverage per acre, select the weather station Development of Corn Minimum and Maximum Temperatures for settlement purposes, and indicate if they prefer a hail endorsement to the contract or the variable price benefit. 35 The farmer must insure all the seeded acres of Nighttime minimum eligible corn and must insure a minimum of five temperature relationship 30 acres for each crop: grain and silage crops are con- sidered separate for the purposes of referring to a t n e 25 specific insurance contract. Only producers grow- m p Daytime maximum ol ing grain or silage corn on irrigated land in AFSC e temperature relationship 20 v e designated areas are eligible to buy a CHU insur- d p ance contract. The farmer must complete seeding or 15 c f by May 31 and must declare the final number of o et seeded acres and a legal description for the location a 10 R of each crop no later than June 1. The insurable crop 5 shall be grown within the risk area boundaries as determined solely by AFSC. Furthermore, the AFSC 0 is responsible for controlling the use of these con- 0 5 10 15 20 25 30 35 40 45 tracts to ensure that they are used only for insur- Minimum and maximum temperature (degrees Celsius) ance purposes. For control and product evaluation Source: Brown and Bootsma 1993. purposes, the farmer is required to present a har- vested production report, stating the production of allinsuredcrops,nolaterthanfifteendaysaftercom- pletionof the harvest but no later than December 15 of each calendar year. which a minimum temperature of minus two de- The premium payable under the CHU contract grees Celsius or less is recorded, after 700 CHU have is due upon receipt of the contract by the farmer. A been accumulated. This means the accumulation table of premium rates and payment rates for grain continues until the first killing frost hits the crop. and silage corn is made available to the farmer and An early frost setback is also built into the AFSC indicates the base premium rate and the percentage calculation.55 payment triggered, depending on the heat unit The weather data for settlement of the contracts level recorded at the station chosen.56 The formula are provided by the federal and provincial weather to calculate the indemnity for each insurable crop stations and compiled by the Irrigation Branch of is given by the following equation: the Alberta Government. Contract end users can select a weather station for the settlement from the Indemnity = Dollar Coverage per Acre federal and provincial stations available, choosing × Payment Rate × Number of Insured Acres the station that best represents the temperatures on their farms. Weather stations used for CHU insur- ance are divided into three groups based on simi- If a farmer chose to insure one hundred acres at lar historical heat accumulations. Weather stations $225 per acre, for example, and the accumulated within each group have similar threshold options, CHU payment rate was 30 percent of the expected premium rates, and loss payment functions. level, a claim of $6,750 dollars would result. The Coverage is available in $25 Canadian Dollar maximum indemnity payable is 100 percent of the (CD) increments with a minimum of CD$100 per Dollar Coverage per Acre (including the additional acre for both grain and silage corn and a limit of dollar coverage if the Variable Price Benefit is acti- CD$225 and CD$300, respectively. Farmers can vated) multiplied by the number of insured acres. buy the insurance product until April 30 of the year Producers can choose between two CHU in- to be covered for that year's growing season. When surance deductibles or threshold options (High Appendix 2. Case Studies of Agricultural Weather Risk Management 85 and Low "Trigger"); see Table A2.1. Payments Table A2.1 Options for CHU Contracts begin sooner under the high threshold option, so this choice has a higher cost than the low thresh- Deductible or Trigger (Annual CHU) old option. Claims are based on accumulated CHUs calcu- Station Long-Term Low High lated using the temperature data recorded at the Grouping Normal Option* Option** selected weather station. CHUs accumulated before A 2,505 2,260 2,380 the killing frost are compared to the threshold B 2,387 2,160 2,280 chosen by the producer at the weather station. If C 2,332 2,100 2,220 the annual CHUs are less than the chosen thresh- old, the insurance program starts to make payments *Approximately 90 percent of long-term CHU normal. according to a predetermined table. The further the ** Approximately 95 percent of long-term CHU normal. annual CHUs are below the threshold, the greater Source: AFSC. the insurance payment. The main peril for producers is lack of heat dur- ing the growing season, but this insurance plan also includes a provision for late spring frost. A cultural insurance. Agroasemex relies heavily on late spring frost can set back corn plant growth and the traditional reinsurance market to protect its affect production. To trigger this provision, a tem- agricultural portfolio from inordinate losses. As a perature of less than zero degrees Celsius must be result of a 70 percent increase in the retrocession recorded on or after June 1 and prior to the record- rates of 2001, Agroasemex's search for new alter- ing of 700 CHUs at the weather station. If both natives led it to analyze the comparative efficiency these conditions are met, 50 CHUs will be deducted of the weather derivatives market. The purpose of from the accumulated total CHUs at the end of the this case study is to present the background, design, year for the first day and an additional 15 CHUs will and guiding principles behind the weather deriva- be deducted for every other day between June 1 and tive structure ultimately created for use as a hedge the day the frost in question occurred. fortheAgroasemexagriculturalportfolio.Itisworth It is important to point out that the CHU con- noting that the institution's weather derivative tract with the hail endorsement is designed to pro- transaction in 2001 was the first of its kind in the tect corn against two major perils: lack of heat and developing world. This simplified case study will hail. The grain and silage corn farmers are also eli- outline the approach and thought processes behind gible for traditional crop insurance contracts based the structuring of the Agroasemex weather risk on individual records; nevertheless, the premiums transfer program. are lower for the CHU contract because of AFSC's reduced transaction costs. It should also be noted Designing a Weather Risk Transfer Solution that the premiums paid by the farmers for the CHU for the Agroasemex Agricultural Portfolio contract are subsidized by approximately 55 per- cent, so the farmer pays only 45 percent of the cost Selection of Risks of the contract. The subsidy is 40 percent for the hail There are two agricultural production cycles in endorsement. The federal and provincial govern- Mexico: spring-summer and autumn-winter. The ments coshare the financial burden of the program, formerisprimarilyarain-fedproductioncycle,while and they subsidize all AFSC's administration costs. the latter is generally irrigated. The Agroasemex weather risk transfer program was specifically de- ALTERNATIVE INSURANCE signed for the autumn-winter cycle of 2001 to 2002. The main weather risks for agriculture during this THROUGH WEATHER INDICES cycle were potentially large negative deviations in IN MEXICO temperature and excess rainfall. For some areas, The Agroasemex Case Study where irrigation was not used, lack of rainfall was also an important risk. The percentages of crops dis- Agroasemex is a Mexican government-owned re- tributed in five states were included in the weather insurance company operating exclusively in agri- risk transfer program. 86 Managing Agricultural Production Risk The crops and weather risks were selected given Oncetheseverityindexwascalculatedforeachcrop, their relative importance in the portfolio, the con- thenextstepwastofindamathematicalrelationship sistency of the numerical analysis between negative between the SI and the weather index most rele- deviations in the agricultural portfolio and the pro- vant to the crop. Agroasemex performed linear tection provided by the proposed weather deriva- least square regressions for each crop severity index tive structure, and the availability of consistent and to establish the SI­weather-index relationship: high-quality historical weather data. Based on the original risk profile and business plan report for the yt = m0 + m1xt + t (5) autumn-winter cycle of 2001­2002, the total liability for the crops and risks selected for the weather risk where transfer program are shown in Table A2.2. The total expected traditional reinsurance pre- yt = Indemnities ; (6) mium for the entire Agroasemex portfolio was esti- Total Liabilityt mated to be US$1,917,422. The subset in Table A2.1 and represents approximately 10 percent of the risk in the entire portfolio for 2001­2002. xt = FCDDt, Transforming Weather Indices into the Expected Indemnities of the Agroasemex where FCDD (Factores Climaticos Dañinos Diarios)-- Agricultural Portfolio damage degree days or periods--that represent the index that captures the critical weather risk of each The following method was used to establish the crop in the portfolio outlined in Table A2.3 (see relationship between weather indices and the ex- below); t is a normally distributed noise term; and pected indemnities of the Agroasemex agricultural the estimators for the linear gradient and intercept, portfolio. First, a severity index was created for each m1 and m0, were calculated using a least squares crop in the portfolio in order to understand, at the regression method. portfolio level, how important this crop risk would The gradient estimator for m1, in particular, is be when a given weather phenomenon, as captured very important, as it establishes the relationship by an index, occurred. A very simple severity index between the individual severity indices and the (SI) is defined as follows:57 relevant weather indices. Once all the linear re- gressions for each crop are performed and all SI = Indemnities Total Liabilityti (4) the linear estimators are calculated, the expected indemnities (in monetary terms) for each severity t = 19991 92,1992 93 ...1999 2000; index, given a certain weather index (FCDD) and Autumn-Winter Cycles total liability, can be calculated as follows: i = Crop (Indemnities) = (Total Liability) × FCDD × m (7) t t t 1 Table A2.2 Total Liability Factored into the Agroasemex FCDDs: The Weather Indices Business Plan for Autumn-Winter 2001­2002 The FCDD terms for each crop in the preceding sec- (the basis of the design of the weather tion represents the weather index or indices that derivative contract) best capture the weather risk for that crop. If we are analyzing the exposure of beans to low tem- Total Liability State Crop (US$ Million) peratures, for example, the FCDD index could be defined as the number of days that the daily mini- Nayarit Tobacco 22.4000 mum temperature drops below a specified daily Sinaloa Beans 0.1917 threshold during the growing season. To construct Sinaloa Chickpeas 0.4600 the appropriate weather indices for the Agroasemex Tamaulipas Sorghum 1.8200 portfolio, the relevant weather historical informa- Sinaloa--Sonora Maize 2.0190 tion was collected: five Mexican weather stations Source: Authors. on the Pacific Ocean coast were chosen to represent thewesternareaofthecountry(Sonora,Sinaloa,and Nayarit), while two U.S. airport stations (McAllen Appendix 2. Case Studies of Agricultural Weather Risk Management 87 and Brownsville) were used to represent the north- drops below 12°C, the expected tobacco yields will eastern area (Tamaulipas). be below average. Hence 12°C is the minimum It is important to note that even though each temperature threshold level for tobacco crop dam- severity index, as defined above, is a seasonal ag- age: DDD-12 represents Damage Degree Days with gregate, the types of risks relevant for an agricul- a 12°C threshold. The DDD-12 index is defined as tural portfolio of crops can occur over very short follows: periods of time; for example, crop damage due to frost can occur in just one day. Therefore the selec- DDD-12 = max(0 ,12-Tmin ) (8) tion of the individual weather indices for each crop was based on two criteria: first, and primarily, on where the DDD-12 summation is over each day the agronomical surveys and experience of the tech- in the growing period of tobacco: November 1 to nical personnel of Agroasemex, and second, on the March 31 of the following year. Daily minimum strength of the mathematical relationship obtained temperature, Tmin, is measured at a single weather when comparing the available data on indemnities station, Capomal, in Santiago Ixcuintla, Nayarit. for the crop in question, with the weather index The data are aggregated at a seasonal level. The (Equation 4)--this was done both on a daily basis DDD-12 estimation is consistent with the El Niño, (data on indemnities were available in daily reso- as the worst year recorded of cold temperatures lution) and on a seasonal basis. affecting the tobacco-producing area. To understand how each individual FCDD was In total, eleven independent FCDDs were de- estimated, consider the example for the weather signed to represent the exposure of the crops and index chosen for tobacco in Nayarit: DDD-12. Low risks selected. The FCDD calculation methodologies temperature is the greatest risk for tobacco crops in usingdailyweatherdataarepresentedinTableA2.3 Nayarit; when the daily minimum temperature for all crops in the portfolio. Table A2.3 Summary of the Methodology to Calculate the Eleven FCDD Indices Weather FCDD Calculation Methodology State Crop FCDD Station (in mm and deg Celsius) Calc. Period Nayarit Tobacco DDD-12 Capomal DDD-12 = Sum Daily [max (0, 12 - Tmin)] Dec 1­Mar 31 EMNF 1 Capomal EMNF = Sum Daily [Rainfall Station 1] + Nov 1­Feb 28 2 La Concha Sum Daily [Rainfall Station 2] EMMA 1 Capomal EMNF = Sum Daily [Rainfall Station 1] + Mar 1­Apr 30 2 La Concha Sum Daily [Rainfall Station 2] Sinaloa Beans DDD-5 Sanalona DDD-5 = Sum Daily [max (0, 5 - Tmin)] Oct 1­Apr 30 DDD-3 Sanalona DDD-3 = Sum Daily [max (0, 3 - Tmin)] Dec 1­Dec 31 EMF 1 Sanalona EMF = Sum Daily [Rainfall Station 1] + Nov 1­Mar 31 2 El Fuerte Sum Daily [Rainfall Station 2] + Sum Daily 3 Jaina [Rainfall Station 3] MAX-5 1 Sanalona MAX-5 = max (MP - 200, 0); Nov 1­Mar 31 2 El Fuerte MP = max (Sum 5-day D3) - max rainfall 3 Jaina for a consecutive period of 5 days, where D3 = Daily Rainfall Station 1 + Daily Rainfall Station 2 + Daily Rainfall Station 3 Chickpeas EMG Sanalona EMG = Sum [max (Daily Rainfall - 55, 0)] Nov 1­Apr 15 Tamau-lipas Sorghum MAXPS 1 Brownsville PS = Sum [max (250 - CMP1, 0)] + 2 Oct 1­May 31 2 McAllen Sum [max (250 - CMP2, 0)]; CMP1 = Monthly Cum. Rainfall Station 1 CMP2 = Monthly Cum. Rainfall Station 2 Sinaloa Maize DDD-5 Sanalona DDD-5 = max [D5 - 22, 0]; Oct 1­Apr 30 Sonora D5 = Sum Daily [max (0, 5 - Tmin)] DDD-3 Sanalona DDD-3 = Sum Daily [max(0, 3 - Tmin)] Dec 1­Dec 31 Source: Authors. 88 Managing Agricultural Production Risk The mathematical relationship between each and using the FCDD-indemnity relationships FCDD index and the indemnities for the corre- established in Table A2.3. sponding crop in the Agroasemex portfolio were The values of the severity index for each crop established using equations 4 through 6, defining a were calculated using both the historical and the means of converting FCDD indices into expected modeled data for comparison. The results showed indemnities in monetary terms. By combining this that the combined weather index established for information,thebasketofalltheexpectedindemnity the Agroasemex portfolio had an acceptable pre- indices was used to replicate the overall weather dictive power, mainly because it captured the large exposure of the agricultural portfolio. This "com- historical deviations in the portfolio (Table A2.4). bined index"--essentially the sum of all the ex- The results demonstrate that the combined pected crop indemnity indices--was used as an weather index model explains about 93 percent of underlying proxy and therefore hedge for the the variability demonstrated by the empirical data. weather exposure of a portfolio. A derivative struc- ture based on this combined index, such as a call Valuation of the Weather Derivative option, is therefore conceptually the same as a stop- Structure and the Agroasemex Transaction lossreinsurancestrategyfortheportfolio,asweather is the greatest risk to Agroasemex. MonteCarlosimulation,asdescribedinAppendix1, was used to generate an estimate of the distribution of the possible results of the combined weather Historical Back-Testing index and therefore the maximum liability of the The strength of the approach outlined above--to Agroasemex portfolio (see Figure A2.2).58 The green establish a basket of indices that best captures the line in Figure A2.2 is constructed using only histor- weather exposure of the Agroasemex agricultural ical information, while the darker, smoother line is portfolio--was back-tested by using annual histor- established from the stochastic Monte Carlo simula- ical indemnity and total liability information from tion analysis of the underlying weather variables. It the Agroasemex direct insurance operations from is clear that the historical payout of the Agroasemex 1990 to 2001. The historical portfolio indemnity portfolio has never exceeded US$1.65 million, while records were compared to the estimated indemni- the simulation analysis generates more extreme ties, given the total liability observed for that year results, exceeding the US$2.5 million level. Table A2.4 Comparative Analysis Between the Observed Historical Severity Indices (indemnities/total liability) and the Estimated Severity Indices for the Crops and Risks Selected Tobacco Beans Chickpeas Sorghum Maize Total Obs. Est. Obs. Est. Obs. Est. Obs. Est. Obs. Est. Obs. Est. 0.052 0.060 0.000 0.086 0 0 0.038 0.057 0.017 0.019 0.020 0.018 0.003 0 0.085 0.093 0 0 0.021 0.023 0.004 0.009 0.027 0.043 0 0.015 0.133 0.122 0 0 0.016 0.021 0.007 0.002 0.109 0.113 0.043 0.042 0.233 0.178 0 0 0.037 0.033 0.009 0.006 0.059 0.047 0 0 0.131 0.158 0 0 0.019 0.018 0.067 0.068 0.164 0.178 0.117 0.126 0.000 0.004 0.017 0.017 0.067 0.069 0.052 0.046 0.403 0.407 0.117 0.104 0.010 0.071 0.142 0.142 0.109 0.103 0.008 0.006 0.167 0.140 0 0 0.084 0.061 0.011 0.011 0.033 0.027 0.007 0.006 0.099 0.115 0 0 0.064 0.175 0.003 0.003 0.010 0.014 r = 0.985 r = 0.968 R = 0.988 r = 0.702 r = 0.999 r = 0.970 r2 = 0.971 r2 = 0.936 R2 = 0.976 R2 = 0.492 r2 = 0.999 R2 = 0.939 Note: Figures are in decimals. Source: Authors. Appendix 2. Case Studies of Agricultural Weather Risk Management 89 Figure A2.2 Comparative Accumulated Distribution Probability Function Based on a "Probability of Exceedence Curve" for the Historical and Modeled Results (payouts in US$) 1 Stochastic Stochastic 0.9 probability of probability of e 0.8 attachment exhaustion c n e 0.7 Structure A: 32.2% 1.4% d e Structure B: 26.4% 1.4% e c 0.6 x Structure C: 21.8% 1.4% e f Structure D: 17.7% 1.4% 0.5 o ytili 0.4 b a b 0.3 or P 0.2 0.1 0 0 500,000 1,000,000 1,500,000 2,000,000 2,500,000 3,000,000 Result ($) Source: Authors. The original analysis performed by Agroasemex ate loading of 30 percent was considered by focusedonfourpossiblecalloptionderivativestruc- Agroasemex. tures, which varied in the strike price and limit of · Loading Based on the Uncertainty due to Gaps payout that could be used as an alternative to a tra- in the Historical Weather Data: When missing ditional stop-loss reinsurance contract to manage data exceed 1 percent of data points, market the portfolio risk (Table A2.5). players usually design a sensitivity analysis The historical results and the stochastic analysis to estimate the impact of using alternative for the actuarial fair value of risk for each call op- in-fillingmethods(seeAppendix1)andcharge tion structure (average and standard deviation) for the uncertainty that arises as a result of are summarized in Table A2.5. In addition to the such gaps in the historical record. No estab- actuarial fair value of risk, the market the pre- lished method exists for calculating this uncer- mium charged for risk management solutions tainty loading in the market, which generally combined the expected or fair value of the risk-- depends on the risk appetite of the individual the pure risk premium--with an additional risk weather risk taker. margin. Considering market standards at the time,59 · Loading for Administrative Expenses: A mar- the following risk loadings above the expected gin of 15 percent was added. value were considered: The weather stations used for the project in Mexico · Loading Based on Standard Deviation:60 Mar- were carefully selected. Nevertheless, missing ket standards 20 to 40 percent. An intermedi- data ranged from 2.70 percent to 9.20 percent. The Table A2.5 Specifications of Call Option Structures Considered by Agroasemex Structure A B C D Strike Price (US$) 1,000,000 1,100,000 1,200,000 1,300,000 Payout Limit (US$) 1,200,000 1,100,000 1,000,000 900,000 Source: Authors. 90 Managing Agricultural Production Risk weather data gaps were in-filled by Risk Manage- Developments Since 2001 ment Solutions (RMS) on a monthly basis, based on After devising the initial weather derivative transac- data collected from neighboring weather stations. tion presented above, Agroasemex devoted its insti- In order to quantify the sensitivity and robustness tutional efforts and experience to developing a local of the in-filling method, instead of filling gaps with weatherriskmarket.Theseactivitiesincludedathor- data inferred from the most correlated stations, the ough review of the weather data; further improve- gaps were also in-filled with the most extreme ob- ments to the weather observation infrastructure, in servations from a sample of stations that had accept- conjunction with the Mexican National Weather able correlations to the station with the missing data Service; and training and education for potential end points, both for temperature and rainfall. The uncer- users within Mexico. The greatest interest generated tainty loading due to missing data was estimated to by the 2001 transaction was from the Mexican gov- be 50 percent of the resulting change in the average ernment regarding their catastrophic weather expo- payout, as a result of this sensitivity analysis, plus sure:since2001,Agroasemexhassoldweatherindex 50 percent of the change in the standard deviation insurance to three Mexican states to cover the states' observed. The results were aggregated to complete catastrophic exposure related to agriculture. In turn, the analysis; Table A2.6 shows the estimated com- Agroasemex has bought protection for this risk, on a mercial premium (expected value plus risk margin) quota share basis, in the international weather deriv- calculatedforthefourweatherderivativestructures. atives market. The three transactions together have Despite the risk loading, Agroasemex eventu- an approximate notional value of US$15 million, ally bought structure D from the market. The main with several other states in the coverage pipeline. motivations for this choice were the following: There are unofficial reports that the international · The transaction included the donation of three market has also closed several transactions with the automated weather stations, worth approxi- private industry in Mexico as a result of this first mately US$36,000, as fallback stations. Taking weather derivative transaction. this cost into account, the ratio of the com- mercial price of the derivative to the pure risk WEATHER INSURANCE FOR premium was the lowest for structure D: 1.57 FARMERS IN THE vs. 1.62 for the nearest structure. DEVELOPING WORLD · To establish credibility and brand recognition for future weather transactions. Case Studies from India and Ukraine · To set a market reference for the risk margin, so that future, larger deals could be negotiated The Commodity Risk Management Group (CRMG) under more narrow risk margins. at the World Bank started working on pilot weather Table A2.6 Estimated Commercial Premium for Weather Derivative Structures (in US$) Analysis and Statistics Structure A Structure B Structure C Structure D Last Ten-Year HBA Pure Risk Premium 181,447 151,447 121,447 91,447 Standard Deviation Loading 83,372 69,669 55,987 42,347 15% Margin 46,733 39,020 31,312 23,611 Full Price 311,552 232,229 186,622 141,157 Simulation Analysis Pure Risk Premium 133,460 104,291 80,252 60,528 Standard Deviation Loading 80,241 70,226 60,638 51,634 Data Uncertainty Loading 31,750 27,584 23,693 20,136 15% Margin 43,315 30,797 24,863 19,793 Full Price 288,766 232,898 189,447 152,091 Source: Authors. Appendix 2. Case Studies of Agricultural Weather Risk Management 91 risk management projects in 2003. The CRMG was BASIX: Weather Insurance for Groundnut involved in its first index-based weather risk man- and Castor Farmers agement contract in India in June 2003. Since then, Established in 1996, BASIX has since emerged as the number of projects has grown. CRMG is cur- one of India's leading microfinance institutions. It rently working on pilot projects for smallholders in has systematically addressed the issues of risk mit- IndiaaswellasprojectsinPeru,Nicaragua,Ethiopia, igation and cost reduction with the twin aims of at- Thailand, Kenya, Malawi, and Ukraine. Providers tracting investment from the mainstream capital in the global weather risk market are extremely in- markets while maintaining and expanding its lend- terested in such new transactions both to diversify ing in rural areas, including lending for agriculture their weather portfolios through new locations and in drought-prone regions (Hubka forthcoming). risks and to offer opportunities for business growth BASIX is the umbrella name used to denote a group and expansion. of companies focused on the provision of micro- TwocasestudieswillillustratesomeoftheCRMG credit and investment services as well as on im- work in this new area. The first case study examines proving the livelihoods of its clients and borrowers. the developing weather market in India, particu- To date, BASIX has approximately 150,000 borrow- larly the recent work of the Mumbai-based insur- ers and 8,600 savers in 7,800 villages in ten Indian ance company ICICI Lombard General Insurance states, disbursing US$37 million in loans since 1996; Company Ltd. and the Hyderabad-based micro- currently 49 percent of loans are for nonfarm activi- finance institution BASIX in making weather in- ties (Hubka forthcoming). Its goal is to affect one surance available to smallholder farmers in Andhra millionlivelihoodsby2010:500,000directlythrough Pradesh. This case study provides an example of the financial services and another 500,000 through role of insurance in access to finance for farmers indirect means. BASIX thinks of itself not as a exposed to weather risk. The second case study microfinance institution but as "a new generation focuses on the 2005 weather insurance pilot pro- livelihood promotion institution,"62 implying that gram for winter wheat farmers in the southern credit alone is not the solution to the problems of oblast of Kherson in Ukraine. rural areas. BASIX manages its risk at two levels: first, it Weather Insurance for Agriculture in India manages its own, institutional-level risk through customer selection and lending practices and part- In 1991, a household survey in India addressing rural access to finance revealed that barely one-sixth nerships withotherinstitutions;andsecond,ithelps of rural households had loans from formal rural its borrowing customers to reduce their risk (Hubka finance institutions and that only 35 to 37 percent forthcoming). By helping customers to mitigate and of the actual credit needs of the rural poor were manage their own risk, and hence the risk of default- being met through these formal channels (Hess ing on their loans, BASIX in turn protects the quality 2003). These findings implied that over a half of all of its own portfolio. In 2003, in order to further ex- rural household debt was to informal sources, such tend the risk management offerings it provides its as moneylenders charging annual interest rates clients, BASIX joined forces with ICICI Lombard, ranging from 40 to 120 percent. A survey based on and with technical support from CRMG, they de- the Economic Census of 1998 (Hess 2003) showed signed, developed, and piloted a weather insur- that India's formal financial intermediaries report- ance product for farmers with small and medium edly met only 2.5 percent of the credit needs of the holdings in Andhra Pradesh. unorganized sector through commercial lending BASIX recognized that, in many areas, farmers' programs.61 yields depend critically on rainfall and that its loan In this context, the CRMG, in collaboration default rates were highly correlated to drought. with the Hyderabad-based microfinance institu- Furthermore, BASIX found that the losses sustained tion BASIX and the Indian insurance company by individual farmers from below average rainfall ICICI Lombard, a subsidiary of ICICI Bank, initi- wereonaccountofseveralfactors,notdirect impacts ated a project to explore the feasibility of weather on yields alone (KBS LAB 2004). In addition to insurance for Indian farmers and to determine if, weather-related yield loss affecting an individual by reducing exposure to weather risk, it would be farmer's ability to meet credit repayments--with possible to extend the reach of financial services to credit default disrupting the next season's loan dis- the rural sector. bursal and hence the farmer's agricultural cycle-- 92 Managing Agricultural Production Risk the systematic nature of drought leads to area-wide new insurance pilot to farmers who had accessed production drops, resulting in local price inflation finance, BASIX would form a base from which they and harder credit terms for the next growing season could begin to understand the interaction between for all producers. such a product, credit repayment, and, ultimately, The government-sponsored area-yield indexed their crop-loan portfolio default rates. crop insurance scheme offered by the National Agricultural Insurance Company (NAIC) is com- The Weather Insurance Contract Design pulsory for all crop-loan borrowers using Indian Groundnut is the primary rain-fed crop grown banks and the only crop insurance option available in the Mahahbubnagar district during the June to to BASIX customers. BASIX, as have others (Hess September monsoon, or khariff, season, followed and Skees 2003), found, however, a number of in- by castor. While most of the cultivation of ground- efficiencies in the federal program in relation to nut and castor is during the khariff, crops are also drought. In particular, they noted that the NAIC cultivated in the winter, or rabi, growing season, in program only led to recovery in extreme situations, pockets of irrigated land. The economics of culti- that is, following district drought declarations by vating groundnut and castor per acre during the the state government, which were often the result khariff and rabi seasons were established through of political maneuvering rather than objective cri- interactions with the BUA members in feedback teria. Furthermore, in the NAIC program, recovery sessions and workshops organized by KBS LAB was based on minimum crop prices and in general and ICICI Lombard, with additional information occurred two to three years after the failed harvest. and crosschecking from the local agricultural uni- By comparison, index-based weather insurance versity in Hyderabad. Total input costs for ground- offered the potential of a transparent, objective, and nut were estimated at Rs 6,500 (khariff) and timely settlement processes for economic losses as- Rs 6,000 (rabi), and for castor at Rs 3,000 (khariff) sociated with noncatastrophic weather risk, with and Rs 3,100 (rabi). recovery based on fair market price estimates. With The aim of the 2003 pilot program was to design the requirements of farmers in rain-sensitive re- weather insurance contracts to insure farmers' input gions in mind, BASIX considered these to be com- and production costs. The initial weather insur- pelling reasons to launch a pilot weather insurance ance contracts designed for the castor and ground- program. nut farmers were based on a weighted rainfall index of rainfall collected and recorded at the First Pilot Program: 2003 Indian Meteorological Department (IMD) official The initial pilot launched by BASIX and ICICI district weather station in the district capital town, Lombard was based in the Mahahbubnagar dis- Mahahbubnagar. High-yield rainfall correlations trict of Andhra Pradesh, with an objective of sell- were measured for khariff crops in the area; never- ing weather insurance policies to two hundred theless agronomic information was used to enhance groundnut and castor farmers through Krishna and strengthen the yield-rainfall relationship for Bhima Samruddhi Local Area Bank (KBS LAB), a the contract structures. In the case of groundnut, for BASIX subsidiary licensed by the Reserve Bank of example, the most critical periods--when ground- India providing microcredit and savings services nut is most vulnerable to low rainfall and therefore in three districts.63 The farmers selected for the water stress--are the emergence periods immedi- initial pilot were members of a Bore Well Users' ately after sowing and the flowering and pod- Association (BUA)64 in four BUA villages in the filling phase two to three months after emergence Mahahbubnagar district: Kodur, Pamireddypally, (Narahari Rao et al. 2000). On the basis of farmer in- Utkoor, and Ippalapaddy. In 1999, for example, terviews, agrometeorological studies (Gadgil et al. the BUA in Pamireddypally received an agricul- 2002),localyieldinformation,andmodelssuchasthe tural loan from BASIX. With a 100 percent repay- United Nations Food and Agriculture Organization ment rate, and therefore a good BASIX credit history (FAO) water satisfaction index (UNFAO 2005), a andstanding,theywereplanningtoborrowafurther groundnut-specific rainfall index was developed. amount for the financial year 2003­2004. Based on The index was defined as a weighted sum of cu- this strong customer relationship, BASIX launched mulative rainfall during the period from May 11 the weather insurance pilot in Pamireddypally and to October 17, the average calendar dates for the the other three villages. In particular, by linking the groundnut growing season. Individual weights Appendix 2. Case Studies of Agricultural Weather Risk Management 93 were assigned to consecutive ten-day periods of weather station were determined to be 653mm the growing season, so the index gave more weight and 439mm, respectively. These reference-weighted to the critical periods during the crop's evolution index values represent the expected growing con- when groundnut is most vulnerable to rainfall ditions that produce satisfactory yields for farmers variability. Furthermore, a ten-day cap on rainfall of these crops in the region. The weather insurance of two hundred millimeters was introduced to the contracts were designed so that payouts started at index because excessive rain does not contribute to 95 percent of this reference level. Farmers partic- plant growth. The individual weights were deter- ipating in the program received a payment if the mined by groundnut water requirements, as ad- index fell below the predetermined threshold, vised by local agrometeorologists, that maximized indicating that the insured should be granted an correlation between district groundnut yields and indemnity to cover lost production and input the rainfall index (Figure A2.3) but defined homoge- costs as a result of lower than expected yields. The nous rainfall periods, making the contract under- initial pilot limited how much insurance a farmer standable and more marketable to the farmers and could purchase by offering three different fixed less susceptible to basis risk (see Appendix 1). More contracts depending on the size holding of the information on the index construction can be found farmer wanting to buy the insurance (Table A2.7). in Hess (2003). The payout schedule as a function of index for The average or reference weighted index value small, medium, and large farmers is given in for groundnut and castor at the Mahahbubnagar Figure A2.4. Figure A2.3 Mahahbubnagar District Groundnut Yields Versus Groundnut Rainfall Index 1000 1200 Correlation: 45% 1971­2001, 69% 1997­2001 Mahabubnagar district groundnut yields 900 Rainfall index measured at Mahabubnagar weather station 1000 800 700 ) ) 800 m er at m( * c e 600 x e h/ d g k( ni ll dl 500 600 af ei ni y t ar u t n u d n 400 n d u n or u 400 or G G 300 200 200 100 0 0 1971 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 Year Source: District yield data are from the government of Andrha Pradesh, Bureau of Statistics and Economics in Hyderabad. Rainfall data from 1994­1996 are missing. 94 Managing Agricultural Production Risk Table A2.7 Weather Insurance Contracts Offered to Groundnut and Castor Farmers Category Premium (Rs) Farmer Eligibility Sum Insured (Rs) Groundnut Small 450 < 2.5 acres land holding 14,000 Medium 600 2.5­5 acres land holding 20,000 Large 900 > 5 acre land holding 30,000 Castor Small 255 < 2.5 acre land holding 8,000 Medium 395 > 2.5 acre land holding 18,000 Source: Authors. The Marketing and Sales Campaign holdings. Of the 154 groundnut farmers, 102 were The products were marketed and sold by KBS LAB women who belonged to Velugu (light) self-help extension officers to the four villages through work- groups. Velugu works with four hundred thou- shops and meetings with the BUA members. The sand poor women organized into self-help groups sales period ended on April 30, 2003. In total, 230 in Andhra Pradesh. Funded by the World Bank, farmers bought the insurance: 154 groundnut farm- Velugu is implemented by the Society for Elim- ers and 76 castor farmers, most having small land ination of Rural Poverty (SERP), an autonomous society setupbythegovernmentofAndhraPradesh to fulfill its poverty alleviation objectives. The women were keen to purchase protection against Figure A2.4 Payout Structure of Groundnut Weather the vagaries of the monsoon, as all their households Insurance Policy Held by Farmers with and most of their fellow villagers grew groundnut. Small, Medium, and Large Land Holdings These fellow villagers were the primary customers of the women in the self-help groups; therefore these women felt the impacts of a poor monsoon 30000 season additionally through drops in sales and pur- ).s chases of their services and hence wanted to protect Large farmers R( themselves also. ) 25000 m The entire portfolio of weather insurance con- ui m tracts soldbyBASIXwasinsuredbyICICILombard, er p 20000 with reinsurance through one of the leading inter- g ni Medium farmers d national reinsurance companies. ICICI Lombard ul c filed all the necessary forms and terms of insur- x 15000 e( y ance with the Indian insurance regulator, regis- cil o tering their products before the program was p 10000 f o launched. t u o At the end of the contract term, the final values y a P 5000 of the weighted indices at Mahahbubnagar weather Small farmers station were calculated by multiplying the cumula- tive rainfall totals in each ten-day period from 0 0 10 20 30 40 50 60 70 80 90 100 May 11 to October 17, 2003, by the specific weight Percentage of reference groundnut rainfall index assigned to that period. The weighted rainfall in- Source: Authors. dices for groundnut and castor were calculated to be 516mm and 490mm, respectively, for khariff 2003, triggering a payout for groundnut farmers and no Appendix 2. Case Studies of Agricultural Weather Risk Management 95 payout for castor farmers. Groundnut farmers with Table A2.8 Pilot Statistics, 2003 small,medium,andlargeholdingsrecovered Rs 320, Rs 400, and Rs 480, respectively, within two weeks Statistic Groundnut Castor Total of the end date of the contract, after the rainfall data were collected and crosschecked by the IMD (see Total number of 154 76 230 Table A2.8). farmers insured Aggregate value 2,250,000 858,000 3,108,000 Farmer Feedback of insurance (Rs) Aggregate premium 71,700 22,880 94,580 The overall farmer feedback from the first pilot paid (Rs) was positive; the farmers welcomed the new prod- Aggregate amount 50,417 0 50,417 uct and appreciated the objective nature of the of claims (Rs) Net Incurred Claim 70.3 0 53.3 weather insurance contracts and the timely pay- to Net Premium ment of claims. In particular, groundnut farmers Earned (%) received a timely recovery from the policies they purchased, even though the Mahahbubnagar dis- Source: KBS LAB. trict was not declared a drought area by the gov- ernment of Andhra Pradesh in 2003 and, as a result, no payments were made from the govern- ment's crop insurance program. The following the crop was made at the beginning of the growing positive aspects of the pilot, as reported by KBS season; they believed more emphasis should have LAB from feedback sessions with the BUA mem- been given to this phase. Other shortfalls, as re- bers in Pamireddypally in January 2004, included ported by KBS LAB after feedback sessions with the following: the BUA members in January 2004, included the · Farmers had the opportunity to reflect on following: rainfall shortages and the economic losses · RainfalldatawerecollectedatMahahbubnagar associated with them and to learn about the weather station, but the farmers felt the sta- concept and process of rainfall insurance; tion did not represent the rainfall of their vil- · Farmers were happy that they could buy rain- lage well. fall insurance to protect themselves from the · Claim calculation criteria were not clearly most critical risk to their farming operations; communicated to the farmers during the sales · The product was introduced through KBS and marketing campaign; in particular, the LAB, a credible source of services and facili- farmers were more comfortable with indexing ties for the farmers; and claims in millimeters rather than in percentile · Claims were paid in a timely manner. points, and the farmers did not understand Some shortfalls were perceived in the product de- the nonlinear payout function of the insur- sign, however; in particular, the farmers expected ance contract and were expecting a linear re- that more weight would be given to the initial sow- lationship between the rainfall index and the ing period of groundnut. Moisture stress at sowing claim amount. In 2003, for example, a 22 per- was associated with the greatest financial risk for cent shortfall occurred in the rainfall index; farmers, as the farmers invest most of their pro- hence the farmers expected Rs 2,800 as the duction costs at sowing time. If the plants do not claim amount: 22 percent of the Rs 14,000 sum germinate and survive the establishment period, insured for small-hold farmers. the entire crop will be lost along with the invest- · Farmers felt that the product should offer ment costs, and the farmer will have to resow, phase-wise payouts for each growing phase, incurring further input and production expendi- subject to the maximum limits, so that it would tures. In 2003, for example, the groundnut farmers beclearhowtheweightsandthereforepayouts expected a greater payout than the amount recov- related to each growing stage. The farmers ered, as the rains during sowing were delayed also requested that in the future the insurance and not optimal. The farmers felt the index did company send a progress report on the rain- not properly reflect that most of the investment in fall for each of the crop phases in order to 96 Managing Agricultural Production Risk facilitate a better understanding within the and, as a result, in general all rain gauges were ten farming community. kilometers away from the faming villages involved · Farmers noted that excess rainfall at harvest in the scheme. This limited the basis risk to farmers, couldresultinseverecroplossesandrequested because the gauges were closer to their actual farms, that protection against the risk of excess rain- but made it more difficult and indeed impossible to fall be offered under the weather insurance find international reinsurance for the final portfolio product. of weather insurance contracts sold by BASIX and insured by ICICI Lombard. In 2004, therefore, ICICI Second Pilot Program: 2004 Lombard chose to keep the risk itself without inter- The second pilot program in khariff 2004 intro- national reinsurance support. duced significant changes to the 2003 design. The The biggest difference in 2004, however, was the program was extended to four new weather refer- design of the weather insurance contracts. In light of ence station locations in two additional districts in the farmer feedback from khariff 2003, the drought Andhra Pradesh: Khammam and Anantapur. The protection products for 2004 were structured by di- weather insurance contracts were offered to both viding the groundnut and castor growing seasons BASIX borrowers and nonborrowers and were mar- into three phases each, corresponding to the plants' keted and sold through KBS LAB in Khammam three critical growing periods: (1) establishment and and Mahahbubnagar districts and through Bhartiya vegetative growth, (2) flowering and pod formation, Samruddhi Finance Ltd. (BSFL)65 in Anantapur dis- and (3) pod filling and maturity. With a departure trict at village meetings, farmer workshops, and from the weighted index design, the new contracts feedback sessions in the month leading up to the specified a cumulative rainfall trigger for each of groundnut and castor growing season. A portion of the three phases, with an individual payout rate the weather insurance contracts were written on and limit for each phase. The groundnut drought local rain gauges monitored by the government of insurance policy offered to farmers in Narayanpet Andhra Pradesh, rather than on the district IMD sta- mandal in Mahahbubnagar district, for example, tions. Because 60 percent of agriculture in Andhra appears in Table A2.9. Pradesh is rain-fed, the government of Andrea Trigger levels and payout rates were determined Pradesh maintains a network of 1,108 rain gauges in consultation with local agrometeorologists and throughout the state. This monitoring is done at the farmers and with reference to local yield data as smallest administrative unit in the state, known as in 2003. Premiums and threshold levels vary by a mandal, which is a grouping of approximately fif- weather station, depending on the risk profile of teen villages. In Andhra Pradesh there are forty to each individual location. This simplified design was fifty mandals in each district, and each mandal has introduced to give clarity to the recovery process one rain gauge: 232 of the rain gauges are owned by by clearly associating each critical growth phase the IMD, and all conform to World Meteorological with an individual deficit rainfall protection struc- Organization specifications. Records begin in 1956, ture. If the rainfall deficit reached the lower limit and historical data can be purchased from the in each phase, the total payout limit for that phase Government Bureau of Statistics and Economics in would be triggered to indemnify farmers for the Hyderabad.Thesecondpilotusedtheseraingauges, severe corresponding crop losses associated with Table A2.9 Payout Structure Per Acre for Groundnut Weather Insurance Policy for Narayanpet Mandal, Mahahbubnagar District (2004) Phase Dates Strike (mm) Limit (mm) Payout Rate (Rs) Limit (Rs) Establishment and Vegetative Growth June 10­July 14 75 20 15 3,000 Flowering and Pod Formation July 15­August 28 110 40 10 2,000 Pod Filling and Maturity August 29­October 2 75 10 5 1,000 Source: Authors. Appendix 2. Case Studies of Agricultural Weather Risk Management 97 the lack of rainfall. Figure A2.5 shows the contract Figure A2.5 Payout Structure of Groundnut Weather payout structure. In a further departure from the Insurance Policy for Narayanpet Mandal, 2003 pilot, the contracts were designed to be sold Mahahbubnagar District, 2004 per acre. A farmer could buy as many acres of protection as he wished, provided he actually cultivated that 3500 many acres of the crop to be insured. The premium 3000 associated with the product in Table A2.9 is Rs 250 ).s Phase 1 per acre insured, for a sum insured of Rs 6,000 per R( 2500 acre. New contracts were also offered for cotton es a h 2000 farmers in Khammam district, and an excess rain- p Phase 2 r e fall product for harvest was offered to all castor and p 1500 t u groundnut farmers with the structure shown in o y 1000 a Table A2.10. P 500 In total, over 400 farmers bought insurance Phase 3 through BASIX in 2004, and a further 320 ground- 0 0 20 40 60 80 100 120 140 nut farmers, members of a Velugu self-help group Cumulative phase rainfall (mm) organization in Anantapur district, bought insur- ance directly from ICICI Lombard. Several farmers Source: Authors. were repeat customers from the 2003 pilot. In con- trast to 2003, ICICI Lombard did not seek re- insurance for the BASIX farmer weather insurance portfolio in 2004. As in 2003, all contracts were set- tled promptly, within thirty days of the end of the emphasis on access to credit and credit repayment calculation period. An example of the marketing by farmers. leaflet developed by KBS LAB and ICICI Lombard The Future for BASIX Weather Insurance detailing the weather insurance contracts for cas- tor, groundnut, and excess rainfall for Narayanpet In 2004, a number of other transactions also took mandal is shown in Figure A2.6. For example, in place within the Indian private sector in response khariff 2004, the rainfall in Narayanpet mandal to the 2003 pilot. In 2004, BASIX bought a crop-loan was not good for groundnut farmers. The rainfall recorded at the local mandal rain gauge measured 12mm for Phase 1 and 84.2mm for Phase 2; rain- Table A2.10 Payout Structure Per Acre for Castor fall during Phase 3 was above average, at 112mm. and Groundnut Excess Rainfall Weather Farmers who bought this policy received a payout Insurance Policy for Narayanpet, of Rs 3,258 per acre insured on September 22, 2004. Mahahbubnagar In autumn 2004, CRMG commissioned a base- line survey to be conducted for the World Bank by Dates September 1­October 10 Rainy Day Index Daily rainfall greater than the International Crops Research Institute for the or equal to 10 mm Semi-Arid Tropics (ICRISAT) in Hyderabad to as- Premium Rs200 per acre insured certain the overall farmer feedback for the first two Limit Rs6,000 per acre insured years of weather insurance. The survey, involving Excess Rainfall Payout Structure one thousand farmers, some of whom have been involved in both pilot programs, will be used as Number of Consecutive Rainy Days Claim Amount (Rs) base from which the impact, efficiency, and accept- ability of the weather insurance concept can be 4 1,500 measured. The results provide strong guidelines 5 1,500 and direction for future weather insurance pro- 6 3,000 grams in India, particularly regarding the issues of 7 6,000 scalability and sustainability. The results also in- Source: Authors. dicate how these new products function in the overall rural finance framework, with particular 98 Managing Agricultural Production Risk Figure A2.6 An Example of the Marketing Leaflet for Groundnut (DGN), Castor (DCN), and Excess Rainfall (EN) Protection in Narayanpet Mandal, Mahahbubnagar District, 2004 Source: ICICI Lombard/BASIX. portfolio insurance policy based on weather indices. dynamic contract start date determined by a rain- For the first time, BASIX used this protection to fall trigger and minimum and maximum limits to coveritsownriskandpassedneitherthecostnorthe the rainfall counted (for example, rainfall below benefits to its client farmers. The protection allowed two millimeters per day is not counted). In addi- BASIX to keep lending to drought-prone areas by tion, BASIX simplified and largely automated the mitigating default risk through the insurance policy underwriting process, which is why BASIX could claims in extreme drought years. BASIX bought a roll out weather insurance to every branch. Intense policy to cover three business locations, which was training sessions with loan officers, who became insured by ICICI Lombard and then reinsured into literally one-stop-shop full customer service agents, the international weather market. allowed BASIX to service a large array of rainfall In 2005, BASIX scaled-up the weather insurance insurance products. At the same time, the policies program for farmers, extending the projects to all became more general "monsoon failure" policies, of their branches in seven Indian states for khariff meaning they were area-specific rather than crop- 2005, with a sales target of ten thousand policies. specificproducts,targetinggenerallivelihoodlosses BASIX sold 7,685 policies to 6,703 customers in of farmers that have diversified agricultural port- thirty-six locations in six Indian states during the folios at risk to weather, rather than losses associ- 2005 monsoon season. The new policies featured a ated with yield variations of a specific crop. For the Appendix 2. Case Studies of Agricultural Weather Risk Management 99 first time BAISX also worked with another insur- 2004; this was scaled up to include more crops ance provider, NAIC, as well as ICICI Lombard, to and farmers in 2005. The NAIC, responsible for sell weather insurance policies in some locations. In the government-sponsored area-yield indexed crop 2005, over seventy new automated weather stations insurance scheme, also launched a pilot weather were installed throughout India, by private com- insurance scheme for twenty districts throughout pany Delhi-based National Collateral Management the country in 2004, reaching nearly 13,000 farm- ServicesLimited(NCMSL)inpartnershipwithICICI ers; the scheme was even mentioned in the gov- Lombard, on which weather insurance contracts ernment of India budget for the financial year 2004 were written, including many BASIX contracts. By to 2005. In 2005, NAIC sold weather insurance to establishing stations closer to the farmers, BASIX approximately 125,000 farmers throughout India. In had more reliable automatic stations as settlement the same year, ICICI Lombard scaled up its agri- bases for their contracts and more accurate products cultural weather insurance sales, reaching ap- for their farmers. NCMSL plans to scale-up their in- proximately 100,000 farmers, and expanded into stallations throughout the country with more insur- other economic sectors. New insurance providers ance provider partners in 2006, which will benefit such as HDFC Chubb also entered the market in end users like BASIX in subsequent seasons. 2005. In total it is estimated that during kharrif 2005 BASIX is also interested in making the insurance 250,000 farmers bought weather insurance through- available to landless laborers and self-help group out the country. Given this strong level of interest women in its operating regions, whose livelihoods and the potential size of the end user market, agri- also suffer from the vagaries of the monsoon. In culture weather risk management in India is set 2004, three hundred women bought a weather in- to grow (Divyakirti 2004). surance policy from ICICI Lombard directly, travel- ing by train to Hyderabad. Weather Insurance for Agriculture BASIX's ultimate goal is to offer weather-indexed in Ukraine66 loans to their borrowers. BASIX can package a loan and a weather insurance contract (Hess 2003), based Ukraine is one of the biggest grain and oilseed pro- on the drought indices described above, for exam- ducers in the world and the agricultural sector is of ple, into one product, such as a weather-indexed great importance for the national economy: agri- groundnutproductionloan.Thefarmerwouldenter culture accounts for 14 percent of the country's into a loan agreement with a higher interest rate that GDP.67 For their production, Ukrainian farmers face accounts for the weather insurance premium that multiple perils, such as drought, excess rain, and BASIX would pay to the insurer. In return, in the frost, which make their incomes unpredictable and eventofadroughtasdefinedbytheindex,thefarmer limit their access to credit. will not repay all the dues. In the event of a moder- Empirical evidence demonstrates that the largest atedrought,insteadofpayingtheloanprincipaland risk to crop production in the Kherson oblast interest, the farmer would repay the principle only; (province) is weather, namely drought in spring in the event of a severe drought, he would only need and summer and low temperatures in winter. to repay part of the principle. Traditional multiple-peril products offered by local During 2004 and 2005, not only did BASIX ex- insurance companies somewhat addressed winter pand their weather insurance program, a number risks, but drought coverage was excluded from the of other institutions, including the originator, ICICI insurance products available to farmers. In addi- Lombard, began expanding the market for weather tion, the insurance companies did not have the pro- insurance in India. IFCCO-Tokio, a joint venture in- fessional staff with agricultural expertise nor the surance company, launched weather insurance con- infrastructure necessary to offer comprehensive tracts similar to the 2003 contracts in 2004, selling agricultural insurance products. Consequently the over three thousand policies to farmers throughout farmers did not trust the insurance companies and India in 2004 and over sixteen thousand in 2005. In the policies offered. High administrative costs and conjunction with ICICI Lombard, the government asymmetry of information further compounded of Rajasthan launched a weather insurance pro- theseproblems,renderingtheagriculturalinsurance gram for farmers for the 2004 growing seasons, in- system in the country ineffective. suring 783 orange farmers from insufficient rainfall In 2001, the CRMG introduced the concept of in khariff 2004 and 1036 coriander farmers in rabi index-based weather insurance to Ukraine in col- 100 Managing Agricultural Production Risk laboration with IFC-PEP. The concept of weather The Kherson Oblast insurance appeared particularly feasible in Ukraine A cursory glance at winter wheat yield data for the because of a widespread system of 187 weather sta- Kherson oblast shows a significant interannual vari- tions, eight in Kherson, and the excellent quality of ability in yield in the region (Figure A2.7), which re- data. After extensive consultations with the farm- flects the agroclimatic risk inherent to the oblast. ers, local authorities, and agricultural scientists, IFC- Formal interviews with winter wheat farmers in the PEP decided to investigate the feasibility of weather region indicated the greatest perceived risks were insuranceinthesouthernoblastofKherson.In order related to weather. toreachtheacceptablevolumeofcontractsales,IFC- PEP decided that the weather pilot project should Designing the Index concentrate on regional farmers' most important Historical yield data for Kherson are unreliable (not crops susceptible to weather risk. Potential crops reported accurately) for the purposes of index con- included winter wheat, spring barley, sunflower, struction, as the data does not faithfully represent and corn. Of these, winter wheat has the biggest the actual production in the rayons (subregions) of planted area and considerable value at risk: 1.5 to the oblast. In order to design an effective weather 2 million tons is produced in the oblast annually risk management instrument, key weather factors with an approximate crop value of US$200 million, had to be discussed with experts, such as agrome- and, in addition, most of this crop is cultivated teorologists and farmers, and crop models using without irrigation. Furthermore, financial institu- weather variables as inputs for yield estimates had tions in the oblast had recently started to accept to be developed. To this end, a report (Adamenko standing crops of grain as security for agricultural 2004)wascommissionedbytheCRMGandICF-PEP loans, despite concerns over lack of sufficient insur- from the Ukrainian Hydrometeorological Center ance protection. (UHC) in Kiev to assess the agroclimatic conditions With this basis in 2004, the CRMG together with IFC-PEP Agribusiness Development Project agreed and weather risks for growing winter wheat in the to run a small pilot project for the Kherson oblast in Kherson oblast. In the absence of reliable yield data, spring 2005. expert assessment and the results from the report based on the UHC oblast-specific crop model were used as the basis for constructing an appropriate weather index for winter wheat in Kherson. Figure A2.7 Winter Wheat Yields for Kherson Oblast, 1971­2001 Identified Weather Risks According to the UHC report (Adamenko 2004) the 45 most significant weather risks for growing winter wheat in the Kherson oblast are (1) winterkill dur- ) er 40 at ing the crop's hibernation period from December to c e March, and (2) moisture stress during the vegeta- h r e 35 tive growth period from mid-April to June. p sl at Winter wheat yields at harvest depend to a great ni 30 u extentonhowwelltheplantssurvivethewinterand q( dl the hibernation period. In the territory of Kherson, ei 25 y theprimarycauseofwinterwheatwintercropdeath t a e is one day or more of air temperature and, therefore, h 20 w r soil temperature below the critical level. These win- et ni terkill events cause damage and death of the plants' 15 W tillering node. Snow cover considerably improves conditions for winter wheat hibernation, as the dif- 10 1971 1976 1981 1986 1991 1996 2001 ference between air and soil temperature increase Harvest year by 0.5 to 1.1°C for each centimeter of snow cover. Source: Hess et al. 2005. The crop usually dies in years without snow cover or when the stable snow cover appears late in win- ter, as it did in 2003. Appendix 2. Case Studies of Agricultural Weather Risk Management 101 Low moisture is the other main limiting factor in Kherson, April 15 to June 30, the SHR is defined for high winter wheat yields in the Kherson oblast. as follows: In fact, lack of moisture in the soil and air during the vegetative growth period is the main cause of SHR = Daily Rainfall 15April-June lowwinterwheatyields.Inparticular,allfiverayons of the oblast are subject to frequent droughts; the (0.1 × Average Daily Temperature ) 15 April-June probability of a severe and medium drought (de- fined subsequently) during the vegetative period in It holds for periods when daily average tempera- the region is 15 to 20 percent and 40 to 50 percent, tures are consistently above +10°C. This period, on respectively. The first critical period in which winter average, begins on April 15 in the Kherson oblast. wheat yield formation is highly susceptible to mois- The SHR does not always serve as a reliable crite- ture stress is the phase from leaf-tube formation to rion of agricultural drought because it does not ac- earing. Due to the climatic conditions of the region, count for soil moisture, but because soil dryness, this period lasts from April 15 to May 25. The water unlike rainfall and average temperature, is gener- requirements for winter wheat during this stage, ally notanobservedvariable,theSHRistheonlyob- when compared to the climatic conditions for this jective indicator that can be used to capture drought period for the oblast, are estimated by the UHC to risk during the vegetative period. Conditions for be 80 percent of the optimum. During the most re- obtaining the best harvest are when the SHR is cent years, in 50 percent of cases the moisture con- between 1.0 and 1.4. When the SHR is greater ditions during this period were close to optimum than or equal to 1.6, plant yields will be depressed (1998, 1999, 2001), while in the other 50 percent of by excessive moisture. When the SHR is less than cases they were insufficient (2000, 2002, and 2003). or equal to 0.6, plants are depressed by drought The second critical period for winter wheat is the conditions. In general, the isoline SHR = 0.5 coin- phase from earing to milk ripeness, which is the cides with regions of semidesert climate condi- kernel formation stage; this lasts, on average, from tions. ResultsfromtheUHCcropmodel(Adamenko May 22 to June 14, but it can extend later into June. 2004) that suggest the impact on yields of SHR dur- Lack of moisture during this period directly de- ing the vegetative growth stage between April 15 creases the number of kernels in a wheat ear and and June 30 are defined in Table A2.11. leads to excessive drying of the kernels. The water The SHR can therefore be used as an index to requirements for winter wheat during this stage, monitor the impact of air drought on winter wheat when compared to the climatic conditions for this crop yields. period for the oblast, are estimated by the UHC to be 90 percent of the optimum. Quantifying the Impact of Weather There are two possible levels for weather insurance The Selyaninov Hydrothermal Ratio Index (SHRI)68 protection that can identify the appropriate limit for The previous findings indicate the need to include drought risk in a meaningful insurance product. Table A2.11 Relationship Between SHR and Winter An example of a product that has been suggested Wheat Yields During the Vegetative for Kherson oblast is outlined in this section. Agri- Growth Phase of Plant Development cultural drought can take two forms: air drought and soil drought. Air drought describes conditions SHR Description Yield Loss (%) in which precipitation is low and high air tempera- ture persists against a background of low relative 1.6 Excessive humidity 30+ air humidity. This leads to unfavorable conditions 1.3­1.6 Damp -- 1.2­1.0 Sufficient humidity -- for plant vegetation and drastically reduces crop 0.9­0.7 Dry -- yields. Soil drought describes the excessive dryness < 0.7 Drought conditions -- of soil, resulting in a scarce supply of moisture avail- 0.5­0.6 Medium drought 20 able for crop growth and development. Air drought, 0.4­0.5 Severe drought 20­50 characterized by a long rainless period, high air tem- < 0.4 Extreme drought 50+ perature, and low air humidity, is often described Source: Hess et al. 2005. using the Selyaninov Hydrothermal Ratio (SHR). For the vegetative growth period for winter wheat 102 Managing Agricultural Production Risk a weather insurance contract: Production costs and Calculating the limit and payout rate for a con- expected revenue. The former, in general, is more tract to protect farmer revenue is a little more diffi- appropriate for catastrophic weather risks early in cult, as harvest-time commodity prices are not the growing season, such as winterkill, when the known in advance when the insurance is purchased. farmer has an opportunity to resow another crop Furthermore, commodity prices also often vary in for summer harvest if the winter wheat crop is response to extreme production shocks, and it is completely destroyed. The latter is, in general, often difficult to quantify the production (weather) more appropriate for weather risks later in the price correlation. Estimates for the harvest-time growing season, when there is no opportunity for price can be made, however; for example, the pre- resowing, yet conditions, such as an April to June vious year's harvest-price or the five-year average drought, can cause yield to vary significantly from of the September price from the local commodities the expected levels. The choice of a factor, how- exchange could be used as a best estimate, or the ever, depends on the preferences of the farmer. government minimum support price could be used Informal interviews with farmers in the oblast in- as a lower boundary for the selling price. dicate that farmers are less concerned with win- terkill risk than with drought risk, even though it Structuring a Weather Insurance Contract can potentially cause complete damage, because of the potential to resow. The Sum Insured Winter wheat farmers spend a maximum of In order to ensure that the insurance product has (Ukrainian Hryvna) UAH 1000 per hectare on pro- some relationship with the true risk exposure of the duction and inputs costs during the crop's entire farmer, the limit of the insurance contract is nego- growing season. The limit of a mid-April to June tiable with the farmer; however, it cannot exceed a drought insurance contract to cover production maximum estimated by the potential insured loss and input costs should therefore be set at UAH to the farmer, as outlined in above. In the design of 1000 per hectare insured. In the event of total crop the contract, an upper limit on the risk volume per failure as a result of a very extreme drought, for client will be set at the total area of the crop planted example, say a SHR < 0.15 event, the farmer would multiplied by the expected selling price, determined be indemnified for UAH 1000 per hectare insured as mentioned above by the previous year's selling to compensate for the loss of the investment. The price according to records, the five-year average, or payout rate of the insurance contract can be deter- the government's minimum support price. mined from the information in the UHC report and is summarized in Table A2.12. Contract Specifications As outlined in Appendix 1, in addition to defining the index, the buyer/seller information (names, Table A2.12 Relationship Between SHR and Financial crop, and hectarage insured), limit and tick-size, an Losses Associated with Winter Wheat index-based weather insurance contract must also Yield Fluctuations include the location (weather station of reference), the calculation period, the strike or deductible, and SHR Payout per Hectare the premium. In the case of Ukraine, to provide the best possible coverage for the farmer client, index- 0.6­0.51 UAH 200 (20% loss) 0.5­0.46 UAH 300 (30% loss) based insurance contracts must be written on the 0.45­0.41 UAH 400 (40% loss) UHC weather station nearest to the farmer's land. 0.4­0.36 UAH 500 (50% loss) Indeed, the extent of the UHC weather observing 0.35­0.31 UAH 600 (60% loss) network may be a limiting factor for the applicabil- 0.3­0.26 UAH 700 (70% loss) ity of this type of insurance in regions that do not 0.25­0.21 UAH 800 (80% loss) 0.2­0.16 UAH 900 (90% loss) have a UHC station. The correlation coefficients < 0.15 UAH 1000 (100% loss) for the interannual variation in cumulative rain- fall, cumulative average temperature, and SHR for Source: Hess et al. 2005. April15toJune30from1973to2002forfiveweather stations in the oblast are given in Table A2.13. Appendix 2. Case Studies of Agricultural Weather Risk Management 103 Table A2.13 Correlation Coefficients for the Interannual Variability of Cumulative Rainfall, Average Temperature, and the SHR Index Measured at Five UHC Weather Stations in Kherson Oblast Station Name Behtery Genichesk Kherson N Kahowka N Sirogozy Station Location April 15­June 30 Cumulative Rainfall Correlation Coefficients (1973­2002) Behtery 1 4615 N 3218 E Genichesk 0.72 1 4610 N 3449 E Kherson 0.74 0.59 1 4638 N 3234 E N Kahowka 0.70 0.41 0.65 1 4649 N 3329 E N Sirogozy 0.35 0.54 0.39 0.50 1 4651 N 3424 E April 15­June 30 Cumulative Temperature Correlation Coefficients (1973­2002) Behtery 1 4615 N 3218 E Genichesk 0.93 1 4610 N 3449 E Kherson 0.98 0.93 1 4638 N 3234 E N Kahowka 0.98 0.95 0.99 1 4649 N 3329 E N Sirogozy 0.95 0.95 0.98 0.98 1 4651 N 3424 E April 15­June 30 SHR Correlation Coefficients (1973­2002) Behtery 1 4615 N 3218 E Genichesk 0.72 1 4610 N 3449 E Kherson 0.74 0.59 1 4638 N 3234 E N Kahowka 0.74 0.44 0.68 1 4649 N 3329 E N Sirogozy 0.38 0.58 0.42 0.50 1 4651 N 3424 E Source: Hess et al. 2005. A very loose rule-of-thumb is that farmers liv- leaf-tubing to kernel formation growth period of ing within a thirty kilometer radius of the weather winter wheat. Final settlement of the weather in- stations may purchase weather insurance indexed surance contracts typically would occur up to to that station. Temperature exhibits less spatial forty-five days after the end of the calculation pe- variability than does rainfall. The benefit of the riod, once the collected weather data have been SHR index is that, by combining cumulative rain- cross-checked and quality controlled by the UHC. fall with temperature, the spatial variability of the The strike would be set at a predefined SHR level index, in comparison to indexes of cumulative appropriate to the weather station under consid- rainfall alone, is slightly reduced. In this example, eration. A pricing example for winter wheat the calculation period for the SHR drought insur- drought risk is given below for Behtery weather ance contract is April 15 to June 30 to cover the station. 104 Managing Agricultural Production Risk Example: Pricing Drought Risk as Measured The payout of a SHR index insurance contract at by the SHR Index Behtery is determined by the following equation: In Behtery, droughts of varying intensity happen Payout = min(max(0,K - SHR) × X, M) quite frequently. Although irrigation is partially used by farmers in this area, farmers have ex- where K is the strike, SHR is the SHR index mea- pressed interest in products that protect against ex- sured during the calculation period, X is the payout treme drought. Figure A2.8 shows the cumulative rate, determined by the structure of the contract, average temperature and cumulative daily rainfall and M is the limit of the contract. A reasonable es- measured at the Behtery station from 15 April to 30 timate for the risk loading factors , , given prices June 1973 to 2002. The temperature data exhibit in the weather market, are = 25% and = 5%. By strong trends, hence the data must be detrended to simply taking the thirty years of payouts in Figure make the historical data consistent with recent A2.9, the payout statistics for a weather insurance warmer conditions that may make severe drought contract with a strike level of SHR = 0.4 can be cal- events more frequent in Behtery now than thirty culated as follows: E(SHR) = UAH 70, (SHR) = years ago. The weather data from the UHC are of UAH 220 and VaR97(SHR) UAH 800. A first-order high quality and do not need to be cleaned or qual- estimate of an appropriate premium to charge a ity controlled prior to analysis. The data are de- farmer for an insurance contract with a strike level trended by fitting and removing a best-fit least of SHR = 0.4 at Behtery Weather Station, therefore, is mean square linear trend to the cumulative average between UAH 110 and 125 per hectare for a sum in- temperature totals for April 15 to June 30 (see sured of UAH 1000.69 (See Figure A2.10 for the terms Appendix 1). Figure A2.9 shows the corresponding of an example of a prototype contract for Behtery.) SHR index: medium droughts (SHR < 0.6) have oc- curred nine times in the past thirty years and severe The 2005 Pilot in Kherson droughts (SHR < 0.4) twice. The driest conditions According to Ukrainian legislation, in order to be occurred in 1996, with SHR = 0.21. able to introduce a new product, such as index- Figure A2.8 Cumulative Rainfall and Average Temperature for Behtery Weather Station for April 15 to June 30, 1973­2002 200 1600 ) 180 C Detrended cumulative temperature 1400 g e 160 d( 1200 er ) Raw cumulative temperature ut m 140 ar m( e p ll 1000 120 m af et ni e ar g 100 800 e ar vit e v al a 80 u 600 yli m a u d 60 C Cumulative rainfall e 400 vit 40 al u m 200 u 20 C 0 0 1970 1975 1980 1985 1990 1995 2000 2005 Year Source: Authors. Appendix 2. Case Studies of Agricultural Weather Risk Management 105 Figure A2.9 SHR Index for Behtery Weather Station, 1973­2002 900 1.60 er 800 Payout in UAH per hectare 1.40 ut c urts 700 1.20 e c n 600 ar us 1.00 x ni e 500 d R H 0.80 nI S R el 400 H S p SHR Index m 0.60 a 300 x e f o 0.40 t 200 u o y a 0.20 P 100 0 0.00 197374 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 00 012002 Harvest year Source: Hess et al. 2005. based weather insurance, to the market, the par- tential of basis risk. The regulator further stated ticipating company (or companies) must design that the insured area must not be greater than the and register the rules of insurance with the state seeded area and, for the purpose of this product, regulatory body. Although the law on insurance-- a farmer's report declaring the seeded area should the leading document regulating the insurance be sufficient proof of the maximum possible area industry--does not specifically reference "index" for insurance. insurance, other legislative documents introduce The weather insurance contract designs and mar- index-based products in relation to agricultural ap- keting materials for the proposed pilot program in plications; for example, relating to agricultural in- Kherson were finalized following receipt of State surance and state finance support of the agricultural Regulator approval of the rules of weather index sector. As a result, there was no direct legislative insurance for agricultural applications. Using feed- barrier prohibiting the use of index-based products back and workshop sessions, IFC-PEP worked with in Ukraine. In April 2005, the regulator agreed to the insurance partner in Kherson oblast to target register rules of insurance that permit the develop- groups--including farmers, agribusinesses, and ment of different types of index-based insurance financial institutions--who could benefit from the products for agribusiness applications. new insurance products. Only two weather insur- The insurance company partner, Kiev-based ance contracts protecting against drought were sold Credo Classic, working with IFC-PEP and CRMG, during the brief marketing period, primarily due to submitted the necessary package of documents the timing of the pilot and late regulatory approval. to the regulator in Kiev. This included drafting The protection period for the first pilot finished in and registering the rules of insurance for index- July 2005. The results of the small first pilot have based weather insurance products with the regu- been communicated to the public to raise awareness lating body. The rules of insurance were accepted about index insurance and the pilot experience: the at the beginning of April 2005, clearing the way concept and methodologies developed have been for the first weather insurance pilot in Ukraine. made publicly available. Presently, the insurance The regulator confirmed that, given the nature of company leading the pilot in Kherson is already the product, the insurer is not required to carry out providing consultations to other markets players field checks and loss adjustments, despite the po- in Ukraine on designing index-based products in- 106 Managing Agricultural Production Risk Figure A2.10 Sample Contract for Behtery Weather Station Buyer Farmer Z 1 Wheat Street, Behtery, Kherson, UA Seller ABC Insurance Company Hectares of Winter Wheat 100 Hectares Insured Calculation Period April 15, 2005 to June 30, 2005 (inclusive) Location Behtery Behtery Weather Station Index, SHR SHR = Index 1 / ( Index 2 * Scaling Factor) Where: Index 1 = Cumulative Capped Daily Rainfall measured during the Calculation Period at Location. Measuring Unit: mm Index 2 = Cumulative Daily Average Temperature measured during the Calculation Period at Location. Measuring Unit: Degrees Celsius Scaling Factor = 0.1 Capped Daily Rainfall Capped Daily Rainfall = min (50, Daily Rainfall Total) Measuring Unit: mm Strike, K 0.4 Maximum Payout, M UAH 1000 per Hectare Insured Settlement Calculation 1. If the Index SHR is greater than the Strike K no payment is made. 2. If the Index SHR is less than or equal to the Strike K the Buyer receives a payout X per hectare insured from the Seller according to the following Settlement Calculation: If 0.36 < max (K ­ SHR, 0) < 0.41, X = UAH 500 If 0.31 < max (K ­ SHR, 0) < 0.36, X = UAH 600 If 0.26 < max (K ­ SHR, 0) < 0.31, X = UAH 700 If 0.21 < max (K ­ SHR, 0) < 0.26, X = UAH 800 If 0.16 < max (K ­ SHR, 0) < 0.21, X = UAH 900 If max (K ­ SHR, 0) < 0.16, X = UAH 1000 Maximum Settlement The maximum payment that can be made from the Seller to the Buyer is UAH 100,000. Premium The Buyer will pay the Seller a premium of UAH 12,000 for the weather protection outlined above. Settlement Data Ukrainian Hydrometeorological Centre, Kiev Settlement Date Within 45 days of the end of the Calculation Period. Source: Hess et al. 2005. Appendix 2. Case Studies of Agricultural Weather Risk Management 107 house and drafting the insurance rules for these new smaller the amount of red light is, in turn, reflected products.Therearealsoplanstoscaleupweatherin- by the plant and recorded by the satellite, therefore surance activities to cover more crops and regions the larger the NDVI value. in 2006. Another important input for the use of NDVI as index insurance is the design of an appropriate mask. A mask is simply a set of geo-referenced TECHNOLOGY APPLICATION information identifying specific land features that CASE STUDIES can be laid over the satellite imagery information. Grassland Index Insurance Using The overlaying of this information allows some of Satellite Imagery the satellite imagery to be extracted from the infor- mation file prior to making production assessments. In recent times, the availability of new technology, such as satellite imagery, has sparked the introduc- Grassland Insurance in Alberta tion of new initiatives to insure grasslands. The most (AFSC operated)70 common technical justifications for the adoption of satellite imagery (SI), as the principle of area-yield In 2001, Alberta launched a pilot project using insurance, are the following: (1) SI can measure pas- satellite imagery to define a historical "benchmark" ture health and growth and represents a multiple- production and assess annual pasture production. peril insurance approach; (2) SI can economically The pilot was limited to a geographical area of the reduce the size of the area on which pasture growth province where pasture is the predominant land andpotentialinsurancepaymentsarebased,thereby cover. An NDVI, scaled appropriately to reflect reducingbasisriskascomparedtootherapproaches native pasture production, was calculated for each (that is, the cage clipping alternative); and (3) SI can township in the pilot area. Insured farmers received assess pasture conditions throughout the growing payments according to a predetermined payment season and thereby lends itself to "intra-seasonal schedule when the annual township NDVI fell coverage options." This section will discuss the use below the historical benchmark NDVI for the of satellite imagery in creating useful indices to in- township. The program was expanded slightly in sure grassland following a parametric and objective 2002 to the portion of the province in which the procedure and will describe relevant experiences in square kilometer resolution (pixel image) of the CanadaandSpain,thetwocountriesthathavemade NOAA satellite system was considered practical the most effective use of this kind of parametric for pasture.71 insurance. The mask used for the project selects only infor- mation known to be at least 85 percent native or im- Use of the Normalized Difference Vegetation proved pasture at a quarter section level (160 acres). Index (NDVI) for Insurance Purposes In the pilot area, where satellite imagery insur- ance operated, a significant percentage of land, 80 One of the satellite networks with more information to 90 percent, is native pasture. Areas of crop irriga- available for these purposes comes from the NOAA tion and some bush land also need to be extracted, satellite. The NOAA satellite has blue, green, red, ortheysignificantlyinfluencetheprogramoutcome. infrared, and thermal sensors and takes one image If a quarter section of land has irrigation, it is re- per day for every square kilometer of the earth's sur- moved from the program dataset. face. The NDVI is a type of vegetative index based The process for calculating a township NDVI on the relationship between red light and near- included the use of daily images to estimate the infrared light. Healthy vegetation absorbs the red NDVI for each square kilometer section and scaled light from the sun and uses it for photosynthesis to identify variations in pasture observations to gen- while reflecting near-infrared light from the sun. erate a pasture vegetative index (PVI). All weekly The formula used to calculate the NDVI is given by: "pixel image" PVI values within a township are averaged to get the weekly township PVI value. NDVI = (NIR - "Red") (NIR + "Red") While ample data existed to calculate the PVI, little accurate "in-field" pasture information was avail- where NIR is near-infrared light andRed is red light. able to judge whether the PVI actually correlated The more red light is absorbed by the plants, the to pasture growth. In the past, however, AFSC had 108 Managing Agricultural Production Risk operated a cage clipping system that allowed it to Grassland Insurance in Spain obtain production estimates. The availability of in- The parametric insurance scheme in Spain was formation allowed pursuit of a statistical procedure engineered mainly to cover farmers from droughts to assess the efficiency of the index indicator to re- affecting the pasture areas. The index utilized is flect the variations in volume of grassland, basi- also the NDVI (estimated from NOAA images). cally by comparing historical PVI values to pasture The product has been offered since 2001 for all the production trends over time, and to confirm any farms performing extensive livestock production, correlation with farmers. The pasture production data were available for specifically cattle, sheep, horses, and goats, and is correlation comparisons from 1991 to 1999 from the designed to cover the farmers experiencing more cage clippings at designated and consistent sites. than thirty dry days (defined as based on the aver- In addition, AFSC personnel compared satellite im- age historical information on pasture). agery to trends in precipitation measured at select In contrast to the previous case study, the insur- Environment Canada weather stations. Correlation able index is based only on pure imagery, that is, no results, however, were not good (approximately verification with actual yields was performed. The r = 0.65). Through a series of client meetings, AFSC index is therefore constructed using a historical evo- asked farmers to identify their two best and two lution of the pixels to create a curve, and the indem- worst pasture production years in the last fifteen- nity is defined when the actual observations in a year period. Since a PVI value could be calculated particular year are located below the average curve, for each township from 1987 to 2000, farmers could based on eighteen years of data. see whether the extreme PVI values compared to Also in contrast to the weekly NDVI values, this their recollections of historical pasture production scheme is based on a ten-day period NDVI index. trends. Production shortfalls due to drought and A Maximum Value Composite Index (MVCI) is es- cool early season temperatures appeared to be iden- timated for each ten-day period to eliminate the ef- tified in the historical PVI values. Geographical dif- fect of clouds. The reference curves built from the ferences among township PVI values corresponded MVCI are smoothed using different algorithms and to the anecdotal production perceptions of farmers are defined as beginning on the first ten-day period surveyed. of October and finalized on the last ten-day period To augment the information acquired by satellite of September of the next calendar year. Whenever imagery, AFSC developed research plots through- information is not available for a particular period, out the pilot pasture area to measure rainfall and a linear interpolation method is used to fill the miss- the growth of pasture under cages and to note ing gaps. changing pasture conditions over the growing sea- The mask in this scheme is based on the Corine son throughout the pilot area (thirty in total). The Land Cover (CLC-90), which is used to discrimi- correlations were improved substantially through nate between areas with and without grassland this process. production. The deductible is calculated from the Pasture insurance is sold in the spring of each ten-day period and is defined as the historic aver- year, but farmers must make their purchasing de- age MVCI for each area, minus 1.25 standard devi- cisions by the end of February. Farmers must in- ations from the average MVCI. The second item of sure all the acres of pasture within the same the deductible is related to the amount of ten-day category--native, improved, or bush pasture--but periods below the individual deductible for each a lower than normal PVI value in one township is time window. The time deductible is three periods not offset by a higher than normal PVI in another. below the reference threshold for every ten-day pe- Coverage and premium are expressed in dollars riod, which is equivalent to thirty days with dry and derived by multiplying the pounds of pasture vegetative indicators. production expected in each forage risk area, as determined by AFSC, by 80 percent of one of the four price options available to the farmer. The pre- REFERENCES mium rate for the 2003 native pasture insurance Adamenko,T.2004."AgroclimaticConditionsandAssessmentof program was 21 percent (60 percent is subsidized Weather Risks for Growing Winter Wheat in Kherson Oblast." by the government). The World Bank Commodity Risk Management Group Appendix 2. Case Studies of Agricultural Weather Risk Management 109 (CRMG) and International Finance Corporation Partnership Hess, U., and J. R. Skees. 2003. "Evaluating India's Crop Failure Enterprise Projects (IFC-PEP), unpublished report from the Policy: Focus on the Indian Crop Insurance Program." Paper Ukrainian Hydrometeorological Centre, Kiev, July. delivered to the South Asia Region of the World Bank, Agriculture Financial Services Corporation (AFSC). 2005. November. "Canada-AlbertaInsuranceProgramsfor2005AnnualCrops." Hess, U., J. R. Skees, H. Ibarra, J. Syroka, and R. Shynkarenko. Promotional and informational brochure published by AFSC. 2005."Ukraine,InitialFeasibilityStudyofDevelopingWeather www.afsc.ca. Index Insurance, Crop Disaster Assistance in Ukraine." World Brown, D. M., and A. Bootsma. 1993. "Crop Heat Units for Corn Bank Working Paper. and Other Warm Season Crops in Ontario." Agriculture and Hubka, A. Forthcoming. "BASIX Case Study." Innovations in Rural Finance, Commodity Risk Management Group, The Rural Division, Ministry of Food and Agriculture Factsheet Agdex #: 111/31, Government of Ontario, Canada, October. World Bank. KBS LAB. 2004. "Weather (Rainfall) Insurance." Visual pre- http://www.gov.on.ca/OMAFRA/english/crops/facts/ sentation prepared by corporate managers of KBS Bank, 93-119.htm. 25 January. Divyakirti, V. 2004. "Saving for a Rainy Day." Environmental Narahari Rao, K., S. Gadgil, P. R. Seshagiri Rao, and K. Savithri. Finance, October. 2000. "Tailoring Strategies to Rainfall Variability: The Choice Gadgil, S., P. R. Seshagiri Rao, and K. Narahari Rao. 2002. "Use of the Sowing Window." Current Science 78: 1216­30. of Climate Information for Farm-Level Decision Making: United Nations Food and Agriculture Organization (FAO). 2005. Rainfed Groundnut in Southern India." Agricultural Systems "Crop Water Management: Crop Water Information." Online 74: 431­57. information, The Land and Water Development Division Hess, U. 2003. "Innovative Financial Services for Rural India: (AGL) Water Resources, Development and Management Monsoon-Indexed Lending and Insurance for Smallholders." Service (AGLW), Water Management and Irrigation Systems Agriculture & Rural Development Working Paper 9, The Group, acquisition date: May 2005. http://www.fao.org/ag/ World Bank. agl/aglw/cropwater/cwinform.stm. Notes EXECUTIVE SUMMARY CHAPTER 4 1. The ex ante or ex post classification focuses on when the 15. This section is based on the background paper by Skees et al. reaction to risk takes place: prior to the occurrence of the 2005. Appendix 2 offers additional technical details. potential harmful event (ex ante) or after the event has 16. This paper does not address the responses to the price risk occurred (ex post). management needs of developing countries, as CRMG is preparing a separate analysis and evaluation (possibly in an ESW) of its ongoing transaction support and capacity- CHAPTER 1 building work in this area. 17. By contrast, area-yield indexes in developing countries 2. While the focus of this document is on natural disaster risks, often are not measured in a reliable and timely manner. the World Bank is also heavily involved in assisting the 18. Basis risk also exists with traditional farm-level, multiple- transfer of commodity price risk for certain commodities. peril crop yield insurance. Typically, a very small sample CRMG will produce a separate document on lessons learned size is used to develop estimates of the central tendency in the price risk management area in 2006. in farm-level yields (for example, four to ten years in the 3. Given the combination of price risk and weather risk man- United States). Given simple statistics about the error of agement transfer, farmers with storage can reduce risk and small sample estimates, it can easily be demonstrated that improve income by storing commodities and bargaining for these procedures sometimes generate large mistakes when higher prices. estimating expected farm-level yield. This makes it possible for farmers to receive insurance payments when yield losses CHAPTER 2 have not occurred and to fail to receive payments when payable losses have occurred. Thus, basis risk occurs not only in index insurance but also in farm-level yield insur- 4. For similar classifications, see Hardaker et al. 2004; and Harwood et al. 1999. ance. Another type of basis risk results from the estimate of 5. For other classifications, see Hazell 1992; World Bank realized yield. Even with careful farm-level loss adjustment 2001; Anderson 2001; Dercon 2002; Townsend 2005; Siegel procedures, it is impossible to avoid errors in estimating the 2005. true realized yield. These errors can also result in under- and 6. This section is based on Townsend 2005. overpayments. Longer series of data are generally available 7. See Dercon 2002. See also World Bank 2001 for a discussion for area-level yields or weather events than for farm-level of the role of safety nets in risk management in developing yields. Because of this, the square-root of n rule suggests countries. there will be less measurement error for index insurance 8. Examples are the Tanzanian coffee and cotton hedging products than for farm-yield insurance products when esti- activities of a major cooperative and CRDB Bank Ltd., the mating the central tendency. If the standard deviation of the leading private agricultural bank in the New York coffee random variable used for the index is lower than the stan- and cotton futures markets. dard deviation of farm-level yields (as would be the case if 9. See the Skees, Barnett, and Hartell (2005) background paper the index is based on area-level yields), the index insurance for more discussion of "cognitive failure" and "ambiguous will have even less measurement error relative to a farm- loading." level insurance product. 19. Temperature, for example, can be measured with field lodged temperature gauges that automatically transmit CHAPTER 3 data to a central server. 10. For more detailed reviews of the U.S. program, see Glauber 2004; Skees 1999a; and Skees 2001. CHAPTER 5 11. The remaining 2 percent of the premiums pays for a variety of other insurance products. 20. Byerlee (2005) distinguishes between growth strategies for 12. Under certain conditions, policyholders can choose to divide irrigated high potential systems and areas with limited mar- farms into separately insured smaller units. ket access in marginal dry lands. Strategies for these two 13. The catastrophic policy only covers yield losses in excess very different types of agricultural systems put different of 50 percent of the APH yield at a rate of indemnity only emphases on agricultural policy options of intensification, 60 percent of the expected market price. diversification, increasing farm size, enhancing off-farm ac- 14. Information in this section is based on Pikor and Wile 2004. tivities, or encouraging exit from agricultural activities. 111 112 Managing Agricultural Production Risk 21. Dercon (2005) also cites the importance of macroeconomic CHAPTER 6 stability and better functioning asset markets because they increase the usefulness of self-insurance. In addition, "Better 36. More details on several of these case studies as well as ad- access to alternative economic activities and increased ditional country examples are presented in Appendix 2. income-earning opportunities could strengthen income- 37. The unorganized sector corresponds to India's informal or based strategies. Public safety nets could be a useful alterna- submerged economy, small-scale nonregistered businesses, tive, although initiatives to develop such programs should for example, particularly in rural areas. take into account their effect on existing risk-coping strate- 38. This is a BASIX subsidiary and a Reserve Bank of India li- gies. Strengthening self-insurance through group-based censed bank providing microcredit and savings services in savings, for example, is an alternative that remains insuffi- three districts. ciently explored" (161). 39. BSFL is another BASIX subsidiary company. Launched in 22. Little, et al. (2004), describe how disastrous droughts in 1998, BSFL is the "flagship" company of the group and is a Ethiopia were the key external factor that "pushed vulnera- Reserve Bank of India registered nonbank financial com- ble households into poverty out of which many had not re- pany engaged in microcredit and retailing insurance and the covered by 2003," three years after the major drought event. provision of technical assistance. Moreover, "the occurrence of periodic droughts tends to 40. A 2002 IFC survey of agricultural enterprise participants wipe out asset gains that poor households attain" (15­17). in Ukraine reveals that the failure of farmers to repay credit 23. These estimates are from Skees, et al. (2005). U.S. Summary was often due to low sale prices, limited product demand, of Business data were used for the U.S. estimate, and data lack of market information, and high interest rates. Only from Pikor and Wile (2004) were used for the Canadian 12 percent of respondents cited bad harvests as the reason estimate. for farmers' inability to repay their debts. In the years before 24. Timely payment of claims was one of the key reasons for the survey was taken, farmers experienced marketing prob- the success of the Indian weather insurance pilot programs. lems for grains and good harvests. Crop failures due to frost See Appendix 2 for the Indian experience with weather and drought in the 2002 to 2003 season may have signifi- insurance. cantly altered farmers' perceptions. 25. CRMG, for example, conducts participatory sessions with 41. The estimate of $1.6 billion was determined by assuming farmers to identify contract and delivery model designs. WFP costs for the 1999­2000 drought, in which the WFP was In Ethiopia, smallholders designate kebeles (local elected responsible for 45 percent of the total food aid deliveries ap- leaders of around six hundred farmers) to collect insur- pealed for by the DPPC. Using that cost estimate to deter- ance premiums for group insurance. In one Malawian vil- mine 100 percent of the cost of the drought in 1999­2000, lage, residents wanted local leaders to contract weather then multiplying this cost by the magnitude of the 1984 insurance that covered the smaller farmers under the pro- drought (assumed to be the worst case scenario), the total grams of the smallholder farmers' association. In India, cost today of a 1984 drought was estimated to be $1.6 billion. microfinance institutions function as trusted intermedi- 42. See Hess and Syroka (2005) for more details on Malawi and aries for small farmers. In some places, cooperatives have the SADC region. gained the trust necessary to deliver insurance products to 43. Skees provided some of the background for this section; see farmers. also Mahul and Skees (2005). 26. This probability distribution was developed using proce- dures that smooth historical data. In reality, few observations have been made below the five hundred millimeter level. CHAPTER 7 27. To be clear, the threshold where cognitive behavior begins is unknown. In this example, five hundred is used for il- 44. See Appendix 1 for a four-step design of a risk management lustration purposes only. If the value were known with plan at the microlevel. certainty, it would also be relatively easy to develop an an- 45. For information on this topic, see the World Bank Hazard alytical solution for the optimal subsidy level. Risk Management Unit Web site. 28. A more detailed discussion of index insurance is found in Appendix 1. 29. International donors could also reinsure this layer through APPENDIX 1 a contingent credit. 30. This section draws on an idea formulated in Skees and Hess 46. This appendix abridges a chapter in a forthcoming Istituto (2003) proposing a "standing disaster insurance program." di Servizi per il Mercato Agricolo Alimentare (ISMEA) pub- 31. See the Agroasemex case study in Appendix 2 for details on lication on innovations in agricultural risk management. reinsuring an agricultural insurance portfolio with a weather 47. The last PWC Survey was published in June 2004; this fig- index contract. ure therefore includes transactions up to March 2004.Anew 32. DOC contracts would most likely be reinsured using direct PWC survey is expected in June 2005. or packaged transfers of the underlying indexes. Pooling 48. In the publication Energy Risk, survey respondents estimated prior to transfer is likely to offer only minimal benefits, since that the market was worth around 45 percent more in 2004. in-country spatial diversification opportunities are gener- The WRMA survey relies on figures from nineteen compa- ally limited for catastrophic layers. nies, all members of the Washington-based organization. 33. See Appendix 1 for details on pricing methodologies. Some large weather trading operations, such as Deutsche 34. This section borrows from World Bank (2005a). Bank and Calyon, are not WRMA members, however, mak- 35. The weather shock insurance safety net concept has been ing the true size of the market difficult to determine. launched by a Malawian government official, Patrick 49. Most energy-related weather transactions are based on tem- Kabambe, and is more broadly based on the work on covari- perature indexes, such as Heating Degree Days (HDDs) and ateshockinsuranceinAfricabytheWorldBankCRMG-Social Cooling Degree Days (CDDs), designed to correspond to Protection unit (Harold Alderman and Will Wiseman) and fluctuations in demand for gas (heating) and power (cool- the CRMG-Southern Africa rural sector unit (Rick Scobey). ing, that is, air conditioning). Notes 113 50. In 1999, the Chicago Mercantile Exchange (CME) began list- Therefore, the following pricing exercise does not include ing and trading standard weather futures and options con- any "detrending" procedures such as those described in tracts on temperature indexes. They now list twenty-two Appendix 1. locations in the United States, Europe, and Japan. 59. This information was provided by RMS, who worked with 51. National Cotton Council of America, http://www. Agroasemex on the initial project. cotton.org/. 60. The Sharpe Ratio method is presented in Appendix 1. 52. Basis risk is a potential mismatch between insured party's 61. The unorganized sector in India corresponds to the informal actual loss and the weather contract payment. or submerged economy, such as small-scale nonregistered 53. More information on the weather station and data require- businesses, found particularly in the rural areas. ments and providers appears below. 62. www.basixindia.com. 54. To be precise, this definition describes a European Option, 63. BASIX Annual Report 2003­04. an option that can only be exercised at the end of its life, that 64. The BUA is a project of the Andhra Pradesh Government; it is, at maturity. In general, this is the most appropriate type subsidizes 85 percent of the cost of community bore wells of option on an underlying weather index. Other types of dug for irrigation of lands belonging to multiple village options includeAmerican Options, an option that can be ex- households. The remaining 15 percent of the bore well cost ercised at any time during its life; Bermudan Options, an op- is met by the individual BUAmembers, in proportion to the tion that can be exercised on specific dates during its life; land they irrigate. and Asian Options, an option with a payout function that 65. BSFL is another BASIX subsidiary company. Launched in depends on the average value of the underlying index dur- 1998, BSFL is the "flagship" company of the group and ing a specified period. is registered with the Reserve Bank of India as a nonbank financial company engaged in microcredit and retailing insurance and the provision of technical assistance. Source: APPENDIX 2 www.basixindia.com. 66. This section is from Hess et al. 2005. 55. Specific information on this is not available for public dis- 67. As of 2003. The source of this information is the World closure. Development Indicators database, August 2004. 56. The actual premium and payment rates are not available for 68. Information on SHR is from Adamenko 2004. public disclosure and are omitted from this paper. Since the 69. See Appendix 1 for details regarding the pricing of weather lack of heat units affects the end use of grain corn more that insurance contracts. it does silage corn, the table of premium and payment rates 70. The information for this section is from AFSC 2005. differs for the two types of crop. 71. The NOAA satellite system was used because historical 57. Besides working as a severity index, this mathematical rela- satellite images were readily available. To be effective, how- tionship is a percentage relationship, allowing the compari- ever, any nonpastureland had to be excluded from the satel- son of figures from different years without concern for the lite images. With the square kilometer resolution of the scale of the measurement or inflation rates. It also helps elim- satellite image, pastureland outside the pilot area is situated inate variations in the total sum insured on a yearly basis. in smaller land parcels and within other crop and forested 58. The weather information for the Mexican transaction was land. Moving beyond the pilot area, with this resolution, reviewed directly by Risk Management Solutions (RMS; would dictate the exclusion of many pixels that do not meet www.rms.com) which determined that no significant trends, the minimum pasture content criteria. Without a minimum particularly in the temperature data, occurred in the infor- number of pixel images, the sample size for a township pro- mation used to construct the weather derivative structure. duction estimate is not credible.