68607 v2 China: Innovations in Agricultural Insurance Technical Annexes June 2007 THE WORLD BANK ii Table of Contents Annex 1: International Experience with Agricultural Insurance ................................ 1 1.1. An Introduction to Agricultural Insurance Programs .............................................. 1 1.2. Examples of Government Support to Agricultural Insurance.................................. 3 1.3. Reinsurance.............................................................................................................. 5 1.4. Six Case Studies....................................................................................................... 6 1.5. References.............................................................................................................. 24 Annex 2: Agricultural Risk Assessment ....................................................................... 25 2.1. Approach................................................................................................................ 25 2.2. Heilongjiang........................................................................................................... 28 2.3. Xinjiang.................................................................................................................. 42 2.4. Shanghai................................................................................................................. 54 2.5. Hainan .................................................................................................................... 58 2.6. Methodology .......................................................................................................... 62 Annex 3: Macro Risk Policy Framework ..................................................................... 69 3.1. Data Set Review..................................................................................................... 69 3.2. Comparison of Data Sets ....................................................................................... 72 3.3. Developing a Macro Portfolio Insurance Model for China ................................... 85 3.4. Expanding the Macro Risk Model ......................................................................... 88 3.5. Limitations and Conclusions.................................................................................. 92 Annex 4: Legal and Regulatory Framework.............................................................. 102 4.1. Insurance Law...................................................................................................... 102 4.2. Regulatory Issues ................................................................................................. 106 4.3. Legal and Regulatory Issues—Index-Based Insurance ....................................... 106 Annex 5: Index-Based Product Development............................................................. 117 5.1. Index Insurance: Background and Product Description ...................................... 117 5.2. Index Insurance Options for China ...................................................................... 134 5.3. Heilongjiang......................................................................................................... 139 5.6. Xinjiang Uygur Autonomous Region .................................................................. 153 5.5. Hainan .................................................................................................................. 163 5.6. Shanghai............................................................................................................... 172 Annex 6: Reinsurance Capacity Issues ....................................................................... 182 6.1. Chinese Reinsurance Market ............................................................................... 182 6.2. Chinese Agricultural Reinsurance Market........................................................... 183 6.3. Individual Company Agricultural Reinsurance Programs................................... 185 6.4. Global Agricultural Insurance and Reinsurance Markets.................................... 194 6.5. Agricultural Reinsurance Capacity Issues ........................................................... 197 Tables Table A1.1: Financial Performance of Crop Insurance in Seven Countries....................... 3 Table A1.2: Types of Government Support to Crop Insurance, Selected Countries.......... 3 Table A1.3: Legal Changes Increasing Premium Subsidy (Percentages)........................... 8 iii Table A1.4: Alberta Province Agricultural Insurance Subsidies...................................... 14 Table A2.1. Heilongjiang Yield Losses Due to Drought and Flood................................. 31 Table A2.2. Soybean Yield Losses in Selected Years ...................................................... 31 Table A2.3. Soybean Yield Losses in Selected Years ...................................................... 32 Table A2.4: Nenjiang County and Qixingpao Farm Average Annual Loss ..................... 32 Table A2.5: Jixian County and Farm 291 Average Annual Loss ..................................... 34 Table A2.6: Shuangcheng City and Hongguang Farm Average Annual Loss ................. 35 Table A2.7: Nenjiang County Loss Ratio Exposure......................................................... 37 Table A2.8: Qixingpao Farm Loss Ratio Exposure.......................................................... 38 Table A2.9: Nenjiang County and Qixingpao Farm Combined Loss Ratio Exposure..... 38 Table A2.10: Jixian County Loss Ratio Exposure............................................................ 39 Table A2.11: Farm 291 Loss Ratio Exposure................................................................... 39 Table A2.12: Jixian County and Farm 291 Combined Loss Ratio Exposure................... 39 Table A2.13: Shuangcheng City Loss Ratio Exposure..................................................... 40 Table A2.14: Hongguang Farm Loss Ratio Exposure ...................................................... 40 Table A2.15: Shuangcheng City and Hongguang Farm Combined Loss Ratio Exposure41 Table A2.16: Heilongjiang Combined Loss Ratio Exposure............................................ 41 Table A2.17: Xinjiang Percentage of Planted Hectares; Yield Losses Caused by Drought and Flood; Cotton and Maize Yield Losses................................................................ 45 Table A2.18. Cotton Yield Losses in Selected Years....................................................... 46 Table A2.19: Manasi County and Farm 8 Average Annual Loss..................................... 47 Table A2.20: Tuokexun County and Farm 4 Average Annual Loss ................................ 48 Table A2.21: Pishan County Average Annual Loss ......................................................... 49 Table A2.22: Manasi County Loss Ratio Exposure.......................................................... 51 Table A2.23: Farm 8 Loss Ratio Exposure....................................................................... 51 Table A2.24: Manasi County and Farm 8 Combined Loss Ratio Exposure..................... 51 Table A2.25: Tuokexun County Loss Ratio Exposure ..................................................... 52 Table A2.26: Farm 4 Loss Ratio Exposure....................................................................... 52 Table A2.27: Tuokexun County and Farm 4 Combined Loss Ratio Exposure ................ 53 Table A2.28: Pishan County Loss Ratio Exposure........................................................... 53 Table A2.29: Xinjiang Combined Loss Ratio Exposure .................................................. 54 Table A2.30: Shanghai Average Annual Loss.................................................................. 57 Table A2.31: Shanghai Loss Ratio Exposure ................................................................... 58 Table A2.32: Hainan Average Annual Loss ..................................................................... 60 Table A2.33: Hainan Loss Ratio Exposure....................................................................... 62 Table A3.1: Estimates of Crop Value by Province—All Crops and Seven Key Crops ... 74 Table A3.2: Market Share of Each Crop by Province ...................................................... 75 Table A3.3: Province Level Coefficients of Variation for Major Crops .......................... 77 Table A3.4: Share of Cause of Loss Ranked by Province................................................ 81 Table A3.5: Loss Exceedance for Loss Ratios for Heilongjiang, Shanxi, and Heilongjiang, and Shanxi Pooled Together ................................................................ 87 Table A3.6: Sorting Provinces by Value of (Actuarially Fair) Stop Loss ........................ 90 iv Table A3.7.(a–z): Detailed Examination of Losses with Province Crop Data and Province Cause-of-Loss Data..................................................................................................... 93 Table A3.7.a: Anhui.......................................................................................................... 93 Table A3.7.b: Fujian ......................................................................................................... 93 Table A3.7.c: Gansu ......................................................................................................... 93 Table A3.7.d: Guangdong................................................................................................. 94 Table A3.7.e: Guangxi...................................................................................................... 94 Table A3.7.f: Guizhou ...................................................................................................... 94 Table A3.7.g: Hebei.......................................................................................................... 95 Table A3.7.h: Heilongjiang............................................................................................... 95 Table A3.7.i: Henan.......................................................................................................... 95 Table A3.7.j: Hubei........................................................................................................... 96 Table A3.7.k: Hunan......................................................................................................... 96 Table A3.7.l: Inner Mongolia ........................................................................................... 96 Table A3.7.m: Jiangsu ...................................................................................................... 97 Table A3.7.n: Jiangxi........................................................................................................ 97 Table A3.7.o: Jilin............................................................................................................. 98 Table A3.7.p: Liaoning ..................................................................................................... 98 Table A3.7.q: Ningxia....................................................................................................... 98 Table A3.7.r: Qinghai ....................................................................................................... 99 Table A3.7.s: Shaanxi ....................................................................................................... 99 Table A3.7.t: Shandong .................................................................................................... 99 Table A3.7.u: Shanxi ...................................................................................................... 100 Table A3.7.v: Sichuan..................................................................................................... 100 Table A3.7.w: Tianjin ..................................................................................................... 100 Table A3.7.x: Tibet ......................................................................................................... 101 Table A3.7.y: Xinjiang ................................................................................................... 101 Table A3.7.z: Yunnan ..................................................................................................... 101 Table A5.1: Term-Sheet Features for a Weather-Index Contract (Rainfall) .................. 129 Table A5.2: Illustrative Climate Characteristics, Heilongjiang (Location: Harbin)....... 140 Table A5.3. Heilongjiang—Main Crop Types and Weather Hazards ............................ 144 Table A5.4: Technical Features of a Prototype Corn or Soybean Rainfall-Deficit Index Product ...................................................................................................................... 147 Table A5.5: Rainfall Payout Example at Nenjiang, Heilongjiang.................................. 149 Table A5.6: Illustrative climate characteristics, Xinjiang. (Location: Urumqi) ............. 154 Table A5.7: Xinjiang—Main Crop Types and Weather Hazards................................... 157 Table A5.8: Example Term Sheet for GDD Index ......................................................... 161 Table A5.9: Illustrative Climate Characteristics, Hainan. (Location: Haikou) .............. 165 Table A5.10: Hainan—Main Crop Types and Weather Hazards ................................... 169 Table A5.11: Illustrative Climate Characteristics, Shanghai.......................................... 174 Table A5.12: Shanghai—Main Crop Types and Weather Hazards................................ 177 Table A6.1: Agricultural Insurance Premiums—Top 10 Territories (2005) .................. 194 v Table A6.2: Agricultural Reinsurance Premiums—Top 10 Territories (2005).............. 195 Table A6.3: Leading Agricultural Reinsurers’ Estimated Global Share (2005)............. 196 Boxes Box A1.1 Types of Federal Crop Insurance Program Policies........................................... 9 Box A1.2: Institutional Design of the Fondos .................................................................. 20 Box A5.1: Main Types of Crop Insurance...................................................................... 118 Box A5.2: Basis Risk...................................................................................................... 123 Box A5.3: Summary of Advantages and Challenges of Index Insurance....................... 124 Box A5.4: Comparison of Indicative Expected Cost Levels Involved in Underwriting and Administration Functions of Traditional and Index Insurance ................................. 126 Box A5.5: Identifying Potential Indexes for Crop Exposures ........................................ 127 Box A5.6: Example Phases in Prototype Maize Drought Index (Thailand)................... 130 Box A5.7: Example of “Varsha Bimaâ€? Index Insurance (India)—Coverage Options ... 130 Box A5.8: Pricing Index Products—An Overview......................................................... 132 Box A5.9: Meteorological Data Requirements for Underwriting .................................. 134 Box A5.10: Growing Degree Days—Definition ............................................................ 160 Box A5.11: Terminology of Typhoon Classification Used in China ............................. 166 Box A6.1: Sunlight Insurance Company Crop Stop-Loss Treaty (2005 and 2006) ....... 187 Box A6.2: Anxin, 2005 Crop Stop-Loss Reinsurance Treaty ........................................ 190 Box A6.3: PICC Coinsurance Pool, Zhejiang Province (2006 Reinsurance Structure) . 192 Box A6.4: Hainan Province—Proposed Pools and Reinsurance Structure, 2007 .......... 193 Figures Figure A1.1: Crop Insurance Premium and Subsidies in the United States ..................... 12 Figure A1.2: Mexican Insurance System after Public Policy Change.............................. 19 Figure A2.1: Heilonjiang—Nenjiang County Yield-Loss Exposure................................ 33 Figure A2.2: Heilongjiang—Qixingpao Farm Yield-Loss Exposure ............................... 33 Figure A2.3: Heilongjiang—Jixan County Yield-Loss Exposure .................................... 34 Figure A2.4: Heilongjiang—Farm 291 Yield-Loss Exposure.......................................... 35 Figure A2.5: Heilongjiang—Shuangcheng City Yield-Loss Exposure............................ 36 Figure A2.6: Heilongjiang—Hongguang Farm Yield-Loss Exposure ............................. 36 Figure A2.7: Xinjiang—Manasi County Yield-Loss Exposure........................................ 47 Figure A2.8: Xinjiang—XPCC No. 8 Yield-Loss Exposure............................................ 47 Figure A2.9: Xinjiang: Tuokexun County Yield-Loss Exposure ..................................... 48 Figure A2.10: Xinjiang—XPCC No. 4 Yield-Loss Exposure.......................................... 49 Figure A2.11: Xinjiang—Pishan County Yield-Loss Exposure....................................... 50 Figure A2.12: Shanghai—Yield-Loss Exposure .............................................................. 57 Figure A2.13: Hainan—Yield-Loss Exposure.................................................................. 61 Figure A3.1: National Crop Share by Province ................................................................ 76 vi Figure A3.2: Comparison of Annual Loss from Province Crop Data and Cause-of-Loss Data for Anhui Province ............................................................................................. 79 Figure A3.3: National Annual Loss from Province Crop Data and Cause-of-Loss Data. 79 Figure A3.4: Loss Exceedance Curves for Anhui Province ............................................. 80 Figure A3.5: Average Annual Loss from Drought ........................................................... 82 Figure A3.6: Average Annual Loss from Floods.............................................................. 83 Figure A3.7: Average Annual Loss from Hail.................................................................. 83 Figure A3.8: Average Annual Loss from Freeze and Frost.............................................. 84 Figure A3.9: Average Annual Loss from All Cause-of-Loss Data .................................. 85 Figure A3.10: Loss Exceedance Curves for Loss Ratios for Heilongjiang, Shanxi, and Pooled Business from These Two Provinces.............................................................. 87 Figure A3.11: Model for Joint Sharing of Catastrophic Financing .................................. 89 Figure A3.12 Risk Grouping for Different Stop-Loss Values.......................................... 91 Figure A3.13: Differences in Probability Distribution of Central Government Payments under Different Rules for Stop Loss ........................................................................... 92 Figure A5.1: Payout Structure for a Hypothetical Rainfall Contract.............................. 119 Figure A5.2: Annual Precipitation, Heilongjiang ........................................................... 140 Figure A5.3: Annual Precipitation, Xinjiang.................................................................. 154 Figure A5.4: Annual Precipitation, Hainan .................................................................... 164 Figure A5.5: Annual Precipitation, Shanghai ................................................................. 174 vii Abbreviations A&O Administrative and Operating AFSC Agriculture Financial Services Corporation (Canada) AGR adjusted gross revenue AGROSEGURO Agrupación Española de Entidades Aseguradores de los Seguros Agrarios Combinados (Spain) AIC Agricultural Insurance Company of India AICI Agricultural Insurance Corporation of India APH actual production history CAIS Canadian Agricultural Income Stabilization CAT Catastrophe Coverage (United States) CCIS Comprehensive Crop Insurance Scheme (India) CCS Consorcio de Compensación to Seguros (Spain) CDD cooling degree day CIRC China Insurance Regulatory Commission CRC Crop Revenue Coverage (United States) CUPIC China United Property Insurance Company ENESA National Agricultural Insurance Agency or La Entidad Estatal de Seguros Agrarios (Spain) FAO Food and Agriculture Organization FB Finance Bureau (China) FCIC Federal Crop Insurance Corporation FCIP Federal Crop Insurance Program (United States) FMD foot-and-mouth disease GDD growing degree day GNPI gross net premium income GRIP group risk income protection GRP Group Risk Plan (United States) HDD heating degree day HRG Heilongjiang Reclamation Group (China) IAIS International Association of Insurance Supervisors IMD Indian Meteorological Department IP Income Protection (United States) ISP Portuguese Insurance Institute or Instituto de Seguros de Portugal ISDA International Swaps and Derivatives Association viii Abbreviations (continued) MPCI multiple-peril crop insurance NAIC National Agricultural Insurance Company of South Korea NAIC National Association of Insurance Commissioners (United States) NAIS National Agricultural Insurance Scheme (India) NASS National Agricultural Statistics Service (United States) OLS ordinary least squares OTC over-the-counter P&C property and casualty PI production insurance PICC People Insurance Company of China PML probable maximum loss R&D research and development RA revenue assurance RMA Risk Management Agency (United States) SAIC Sunlight Agricultural Insurance Company (China) SIPAC Integrated System of Protection against Climatic Events or Sistema Integrado de Protecção Contra as Aleatoridades Climáticas (Portugal) SRA standard reinsurance agreement (United States) TSI total sum insured USDA United States Department of Agriculture VaR value at risk WRMA Weather Risk Management Association WMO World Meteorological Organization WTO World Trade Organization XPCC Xinjiang Production and Construction Corps (China) All dollar amounts are U.S. dollars unless otherwise indicated. ix x Annex 1: International Experience with Agricultural Insurance Beginning with a general introduction to agricultural insurance programs in other countries (Section 1.1), this annex discusses the different ways that governments of selected countries have provided support to their crop insurance programs (Section 1.2) and governmental provision of reinsurance protection in several of these countries (Section 1.3), and presents case studies (Section 1.4) detailing how the governments of six countries—the United States, Canada, Spain, Portugal, Mexico, and India—have been involved in crop insurance programs. Because these countries exhibit a wide range of government support to and intervention in agricultural insurance, reviewing their programs can provide some basis for understanding many of the recommendations that are made in this report. 1.1. An Introduction to Agricultural Insurance Programs Crop insurance is available in 50 countries around the world, with 84 percent of the insurance market in North America and Europe (primarily the United States, Canada, and Spain). Among the factors motivating the desire to introduce both crop insurance and livestock insurance is the increased frequency of natural disasters during recent decades. However, given the high cost to government that characterizes most agricultural insurance programs, such programs are not feasible models for all countries. This annex includes a brief introduction to the types of government support and interventions in agricultural insurance programs. To illustrate the wide range in the level of government support and intervention, it presents six case studies: the United States, Canada, Spain, Portugal, Mexico, and India. The most common form of government support to agricultural insurance involves up- front premium subsidies tied to what are intended to be commercial premiums. In addition, governments in some countries provide financial subsidies toward the costs of crop insurance administration and loss assessment, and/or for the costs of research and development (R&D) into new products and programs. Other forms of government intervention in agricultural insurance include the provision of catastrophe reinsurance protection and, occasionally, the establishment of state agricultural insurance companies. 1 Specific crop insurance products are structured in a variety of ways, but most take one of two main forms: named-peril crop insurance and multiple-peril crop insurance (MPCI). Named-peril programs insure against losses resulting from a specific risk. Hail insurance, for example, has been offered by the private sector for over 100 years in the United States and Europe. Farmer cooperatives in France and Germany had crop/hail insurance dating 1 Prior to the 1980s, governments in many developing countries established state multi-peril crop insurance programs (examples include Latin America, the former Soviet Union, and South Asia). In nearly all cases these national programs have been discontinued due to poor financial results. Since the mid-1980s the trend has been toward private sector-led agricultural insurance programs, with or without support from governments. 1 back to the 1820s. Its success in the private sector is due to the nature of the risk: Hail is an infrequent event that causes localized damage and rarely impacts a widespread area. Thus, if insurer spread their risk over a large area, the losses are manageable with the premium income from a single season. Hail is typically an independent risk akin to house fires or automobile accidents, insurable under traditional insurance approaches. In contrast, MPCI protects against crop losses resulting from any of a list of risks, including adverse weather, pest infestation, and disease. MPCI creates a greater insurance challenge because of the multitude of risks that are covered. Indemnities are paid on the basis of individual losses as determined by insurance adjusters. MPCI programs may offer the most accurate compensation for farmers but often require a high level of government support. The cost to administer these programs is extremely high, because each claim potentially requires a farm visit to determine the level of loss. To reduce the cost to farmers and encourage greater participation, governments often subsidize premiums. For an insurance product to be actuarially sound over the long run, the costs (indemnities plus administrative costs) must be less than premium revenue. People often ignore administrative costs in figuring loss ratios, considering only the ratio of indemnities to premium payments. Administrative costs include operating expenses and the cost of capital. Once the administrative costs are taken into account, as in the equation below, a true indicator of actuarial performance is revealed. I = Indemnities A = Administrative cost P = Premium (I + A) / P < 1 For the U.S. crop insurance program, for example, the government finances product development and covers the administrative and operational costs incurred by the private companies that sell the insurance. This is common practice for multiple-peril insurance programs in other parts of the world, including Canada, Spain, Japan, and Brazil, among others. Although the figures in Table A1.1 are out of date, they illustrate that these programs are supported with intended or unintended premium subsidies and some high levels of administrative costs. For example, the experience of Japan from 1985–89 shows that indemnities are slightly less than the premium, but administrative costs are 3.57 times more than the premium, due to the level of monitoring needed to control adverse selection and moral hazard. The result is a dilemma: Either excessively high costs are incurred in controlling adverse selection and moral hazard problems, or the consequence is bad actuarial performance that many times results in unintended subsidies as governments incur the cost of indemnities that exceed premiums. These numbers seem indicative of experience in establishing a government-supported MPCI program. To emphasize the point, the U.S. I/P ratio is now much closer to 1, but the (A+I)/P ratio is now approaching 4 due to increases in subsidies. 2 Table A1.1: Financial Performance of Crop Insurance in Seven Countries Country Time Period I/P A/P (A+I)/P Brazil 1975–81 4.29 0.28 4.57 Costa Rica 1970–89 2.26 0.54 2.80 India 1985–89 5.11 n/a n/a Japan 1947–77 1.48 1.17 2.60 Japan 1985–89 0.99 3.57 4.56 Mexico 1980–89 3.18 0.47 3.65 Philippines 1981–89 3.94 1.80 5.74 United States 1980–89 1.87 0.55 2.42 Source: Hazell 1992. Worldwide, there are few examples of MPCI programs that do not have significant government subsidies. Chapter 3 provides a review of the various costs associated with these types of insurance products, showing why many governments ultimately choose to subsidize these programs. 1.2. Examples of Government Support to Agricultural Insurance Cross-national comparisons of crop insurance are helpful in understanding the various ways governments support agricultural insurance around the world. Table A1.2 provides a cross-national comparison of government support to crop insurance for selected countries. Table A1.2: Types of Government Support to Crop Insurance, Selected Countries Country Forms of Government Financial Support Year Agricultural Public- Premium Subsidies on Financial Public Public- Insurance sector Subsidies Administrative Support to Sector Crop Private Pool Multiple- Costs of Crop Research & Reinsurance Program (Coinsurers) Peril Insurance Development Incepted Crop and Insurer Training United 1930s NO NO √ √ √ √ States Canada 1970s NO √ √ √ √ √ Spain 1980 √ NO √ NO NO √ Portugal 1979 NO NO √ NO NO √ Italy 1970s NO NO √ NO NO NO Mexico 1990 NO NO √ NO √ √ 3 Chile 2000 √ NO √ NO √ NO India 1985 NO √ √ √ NO √ South 2001 √ NO √ √ NO √ Korea France 2005 NO NO √ NO NO NO Sources: Authors. Agricultural insurance pools currently operate in Spain, Chile, and South Korea. Portugal operated a pool program between 1979 and 1990 until the pool was disbanded. In all the other countries listed, crop insurance is provided by individual insurance companies. Today only one country, Canada, has a public (state) agricultural insurer. In all the other countries listed in Table A1.2, the agricultural insurers are either private commercial companies or, occasionally, cooperative or mutual insurers (for example, the national cooperative insurer, NAIC, in South Korea; and the fondos, small mutual insurers, in Mexico). Table A1.2 shows that the most common form of state support in these selected countries is subsidies on premiums paid by growers. These subsidy levels vary from about 30 percent in Mexico to as high as 85 percent in Portugal, for some groups of farmers. In addition, the United States provides the catastrophe (CAT) program (50 percent insured yield coverage) essentially free of charge to growers (apart from a small policy- processing fee). The complexity of the form and level of premium subsidies for agricultural insurance varies from country to country. Several countries have opted for a single flat-rate premium subsidy that applies to all crop and livestock programs and to all types of farmers, of either large or small holdings, in favored or in marginal areas. Countries with single premium subsidy rates include: • Turkey: 50 percent premium subsidy applicable to all programs; • Chile: 50 percent premium subsidy applicable to all programs; • South Korea: 50 percent premium subsidy applicable to all crops; • France: 25 percent premium subsidy level applicable to all crops (pilot). Other countries, including Spain and Portugal, have opted for differential-premium subsidies based on such criteria as the following: • Type of crop (more susceptible crops with higher rates carrying higher subsidies); • Favored regions where government is promoting agriculture; • Type of farmer (in Spain, for example, young farmers can obtain an additional premium subsidy); • Type of insurance contract (individual versus collective or group contract, which carries an additional subsidy; • Time period—annual versus multiyear agricultural contract (higher subsidies to promote multiyear contracts). 4 Canada and the United States provide high levels of government intervention in crop insurance, with financial support in the form of crop premium subsidies and financial assistance with the administrative costs of insurance. For example, the U.S. government subsidizes the costs of sales distribution through agents and the costs of loss adjustment, provides reinsurance protection (often at very favorable terms), and financially supports product R&D, and training of growers, crop insurance field staff, and so on. The oldest of the programs listed is the U.S. Federal Crop Insurance Corporation (FCIC) Multiple-Peril Crop Insurance Program, which traces its origins to the dust bowl of the mid-1930s and policies of the Franklin D. Roosevelt administration to stabilize farm incomes and livelihoods. The newest of these schemes is the pilot subsidized MPCI program of the government of France, begun in 2005. In Italy, private sector insurance (crop, hail, and some freeze coverage) is available to farmers. Premium subsidies are available to all growers registered with regional producer associations. Annual crop insurance premiums amount to about $350 million. In Chile, the government provides50 percent premium subsidies from the Internal Economic Ministry budget, administered by the Ministry of Agriculture. A new national crop insurance initiative, launched in September 2000, provided named-peril yield- shortfall coverage for losses below 67 percent of average yield. Currently, coverage is extended to 18 crops, including cereals and horticultural crops. In the future, the program will include tree fruits and vines. The program is insured by a pool of four private commercial Chilean insurers, led by Mapfre Chile. The program is reinsured by leading international agricultural insurers (proportional and nonproportional reinsurance). In South Korea, since 2005 the national cooperative insurer, National Agricultural Insurance Company (NAIC), has been charged by the government to administer the pool program on behalf of a group of local companies that act as coinsurers. The NAIC pool is supported by three private insurance companies and one Korean reinsurer. 1.3. Reinsurance Another important form of government intervention in agricultural insurance is the provision of catastrophe reinsurance protection, found in six of the ten countries listed in the Table A1.2—the United States, Canada, Spain, Portugal, Mexico, and South Korea (since 2005). Spain has a national reinsurer Consorcio de Compensación to Seguros (CCS), which acts both as a direct coinsurer under the pool with 10.5 percent share and a stop-loss reinsurer on the viable lines, and as an excess-of-loss reinsurer on the more volatile experimental lines. Reinsurance programs in selected countries are described below: • Portugal has a government stop-loss reinsurance treaty with different retentions (priorities) according to each frost risk zone, varying from a very low priority of 65 5 percent in the most frost prone zone to a high priority of 110 percent loss ratio in lower frost prone zones. The government’s liability above the retention levels is open-ended. Insurance companies are required to coinsure 15 percent of the excess losses. Government stop-loss reinsurance can be contracted voluntarily by any of the 12 Portuguese private insurers. In 2005, the gross net premium income (GNPI) was $35 million. • Spain has a stop-loss treaty for viable lines in three layers, up to 160 percent loss ratio. Excess-of-loss reinsurance is provided for experimental lines. In 2005, the GNPI was $750 million. For the viable lines, individual pool members are free to purchase stop-loss reinsurance on their retentions from international reinsurers. • South Korea has private international stop-loss reinsurance treaties for losses in excess of a 110 percent loss ratio up to a 170 percent loss ratio. The government provides unlimited stop-loss reinsurance for losses in excess of the 170 percent loss ratio on a base premium of $50 million in 2005. • Chile has a reinsurance pool that currently places a 60 percent Quota Share Reinsurance Treaty with leading European reinsurers. 1.4. Six Case Studies This section presents overviews of agricultural risk-management programs in six countries—the United States, Canada, Spain, Portugal, Mexico, and India—involved in implementing substantial programs to reduce yield risk and revenue risk for agricultural producers. Although they offer a variety of risk-management products for farmers, these programs also require significant government support that would not be feasible for most countries. Crop Insurance Experience in the United States and Canada The United States and Canada have long-established MPCI programs. Over the past 20– 30 years, both countries have made significant changes in their crop insurance programs. In both countries these programs are subsidized by the federal (United States) or federal and provincial (Canada) governments. Both of these crop insurance programs merit critical review, because they are often promoted to lower-income countries as an effective and efficient government response to perceived market failure of the agricultural insurance sector. Many elements of the crop insurance programs of both countries are similar and will be reviewed in greater detail below, but at least two major differences between the U.S. and Canadian crop insurance programs merit noting. These differences involve the delivery system and the extent of product uniformity across the country. In the United States, crop insurance is sold and serviced (including claims adjustment) by private sector insurance companies. In Canada, crop insurance is sold and serviced by provincial government entities. In the United States, the Risk Management Agency (RMA) of the United States Department of Agriculture (USDA) is responsible for approving any new insurance 6 products and maintaining existing crop insurance products (including setting premium rates). As a federal agency, the RMA ensures a large degree of product uniformity across different geographic regions of the United States. In Canada, provinces have some degree of autonomy to tailor insurance products to regional needs. The provincial government entities that sell and service crop insurance have a relationship with the Canadian federal government whereby the federal government provides some subsidy and reinsurance capacity under the broader umbrella of the Canadian agricultural safety net policy. Both the United States and Canada have expanded their insurance offerings beyond farm- level yield insurance. Both now offer farm-level revenue insurance products, which account for approximately three-fourths of the U.S. agricultural insurance premiums. The U.S. revenue insurance products are crop-specific. 2 Canada has moved aggressively to offer multicrop, whole-farm revenue insurance. Both the United States and Canada also have agricultural insurance products that trigger indemnities based on area-level (rather than farm-level) yield or revenue shortfalls. Common elements of U.S. and Canadian crop insurance The U.S. and Canadian crop insurance programs share several common elements. Each of the programs: • Attempts to address both social welfare and economic efficiency objectives embedded in the same program; • Offers core, farm-level, multiple-peril insurance products; • Employs government premium subsidies that are a percentage of total premium; • Uses government funds and/or government agencies to absorb the administrative and operating costs of the program; • Requires significant government expenses; and • Involves a major role for government in pooling and holding some of the most catastrophic risk exposure. Each of these common features is analyzed in more detail in the discussion below. Any country rethinking its existing approaches to agricultural insurance programs, or attempting to design new insurance products, should critically assess the U.S. and Canadian experience with respect to each of these common features. The United States 3 Multiple-peril yield and revenue insurance products are offered through the Federal Crop Insurance Program (FCIP), 4 a public-private partnership between the federal government and private sector insurance companies. The FCIP, regulated and financed by the federal government, is administered wholly by private insurance companies through a 2 See Glauber (2004) for a review of U.S. revenue insurance products and their growth. 3 See Glauber (2004) and Skees (2001) for more detailed reviews of the U.S. program. 4 The FCIP is administered through the USDA Risk Management Agency by the Federal Crop Insurance Corporation. 7 nationwide network of more than 26,000 local insurance agents. The federal government is responsible for setting the terms, conditions, and rates on all crop coverage, which are then marketed by the private companies on their own paper. The program seeks to address both social welfare and economic efficiency. Regarding social welfare, private companies that sell federal crop insurance policies may not refuse to sell insurance to any eligible farmer, regardless of past loss history. At the same time, the program aims to be actuarially sound. In 2006, a total of 1.1 million crop policies were sold, with premium totaling $4.6 million, of which $2.7 million was subsidy. The total amount insured was nearly $50 billion. Forms of government support The following are examples of government support for MPCI: • Insurance legislation: The 1980 Federal Crop Insurance Act and subsequent amendments in 1990 and 1994, and the 2000 Agricultural Risk Protection Act. • Premium subsidies: Table A1.3 shows the premium subsidy changes over time by the various coverage levels. In 2006, nearly 59 percent of the total premium value was a subsidy from the government. Table A1.3: Legal Changes Increasing Premium Subsidy (Percentages) Coverage 1994 Level 1980 Act Reform Act 2000 Act 55 30 46.1 64 65 30 41.7 59 75 16.9 23.5 55 85 -- 13 38 Sources: Authors, from USDA data. • Subsidies paid to private insurers to cover administrative and operating (A&O) expenses: The A&O subsidy covers all delivery and loss adjustment costs and varies depending on the policy type. The average A&O expense reimbursement is about 22 percent of the unsubsidized premium. Traditionally a very high proportion of these payments have gone toward agent commissions (brokerage). • Reinsurance: Government provides reinsurance below market rates, and private insurers may cede nearly all of the risk on some policies. The average cost of these reinsurance subsidies is roughly 14 percent of the unsubsidized premiums. • Financing crop insurance R&D: Government finances the bulk of R&D in developing new programs for new crops and also performs actuarial reviews and rate adjustments. 8 The FCIP dates back to 1938, but direct premium subsidies were not introduced until the 1980 legislation, before which participation rates were about 10 percent nation wide, although some higher risk areas had participation rates that exceeded 30 percent. In 2006, about 80 percent of the eligible plantings were insured with some form of crop or revenue insurance. Summary of major insurance policies Box A1.1 describes the different types of insurance policies that are available to farmers. Policies are available for more than 100 commodities, but in 2004 four crops—corn, soybeans, wheat, and cotton—accounted for approximately 79 percent of the $4 billion in total premiums. Excluding pasture, rangeland, and forage, approximately 72 percent of the national crop acreage is currently insured under the FCIP. About 73 percent of total premiums are for revenue insurance policies, and 25 percent are for yield insurance policies. 5 Most policies trigger indemnities at the farm (or even subfarm) level. 6 Yield insurance offers are based on a rolling 4–10-year average yield known as the actual production history (APH) yield. The federal government provides farmers with a base catastrophic yield insurance policy, free of any premium costs. Farmers may then choose to purchase, at federally subsidized prices, additional insurance coverage beyond the catastrophic level. This additional coverage, often called “buy-upâ€? coverage, may be either yield or revenue insurance. Farm-level revenue insurance offers are based on the product of the APH yield and a price index that reflects national price movements for the particular commodity. Area-yield and area-revenue buy-up insurance policies are offered through the FCIP for some crops and regions. On a per-acre insured basis, area-level insurance products tend to be less expensive than farm-level insurance products. In 2006, area-yield and area- revenue policies accounted for about 10 percent of total premiums. Box A1.1 Types of Federal Crop Insurance Program Policies Farmers may select from various types of policies. MPCI policies are available for most insured crops. Some of the policies listed below are being tested in pilot programs and are only available in selected states and counties. Yield-based (APH) insurance coverage Actual production history (APH) policies insure producers against yield losses due to natural causes such as drought, excessive moisture, hail, wind, frost, insects, and disease. The farmer selects the amount of average yield to insure—from 50 to 75 percent (in some areas up to 85 percent)—and also selects the percent of the predicted price he or she wants to insure (55–100 percent of the crop price established annually by RMA). If the harvest is less than the yield insured, the farmer is paid an indemnity based on the 5 The remaining 2 percent of premium is for a variety of other insurance products. 6 Under certain conditions, policyholders can choose to divide farms into smaller units that are insured separately. 9 difference. Indemnities are calculated by multiplying this difference by the insured percentage of the established price selected when crop insurance was purchased. Group risk plan (GRP) policies use an area-yield index as the basis for determining a loss. Payments are not based on the individual farmer's loss records. When the county yield for the insured crop, as determined by the National Agricultural Statistics Service (NASS), falls below the trigger level chosen by the farmer, an indemnity is paid. Yield levels are available for up to 90 percent of the expected county yield. GRP protection involves less paperwork and costs less than the farm-level coverage described above. However, individual crop losses may not be covered if the county yield does not suffer a similar level of loss. This type of insurance is most often selected by farmers whose crop losses typically follow the county pattern. The dollar plan provides protection against declining value due to damage that causes a yield shortfall. The amount of insurance is based on the cost of growing a crop in a specific area. A loss occurs when the annual value of the crop is less than the amount of insurance. The maximum dollar amount of insurance is stated on the actuarial document. The insured may select a percentage of the maximum dollar amount equal to CAT, or additional coverage levels. The dollar plan is available for several crops, including fresh market tomatoes, strawberries, and cherries (on a pilot program basis in limited areas only). Revenue insurance plans Note: All revenue-based options determine revenue differently; the provisions of each policy define revenue. Group risk income protection (GRIP) makes indemnity payments only when the average county revenue for the insured crop falls below the revenue chosen by the farmer. Adjusted gross revenue (AGR) insures the revenue of the entire farm rather than an individual crop by guaranteeing a percentage of average gross farm revenue, including a small amount of livestock revenue. The plan uses information from a producer's Schedule F tax forms and the current year's expected farm revenue to calculate the policy revenue guarantee. Crop revenue coverage (CRC) provides revenue protection based on price and yield expectations by paying for losses below the guarantee at the higher of an early-season price or the harvest price. Income protection (IP) protects producers against reductions in gross income when either a crop's price or yield declines from early-season expectations. The policy provisions determine coverage. Revenue assurance (RA) provides dollar-denominated coverage. The producer selects a dollar amount of target revenue from a range defined by 65–75 percent of expected revenue. The policy provisions determine coverage. Policy endorsements Catastrophic coverage (CAT) pays 55 percent of the established price of the commodity on crop losses in excess of 50 percent. The premium on CAT coverage is paid by the federal government; however, producers must pay a $100 administrative fee for each 10 crop insured in each county. Limited-resource farmers may have this fee waived. CAT coverage is not available on all types of policies. Source: USDA, RMA website http://www.rma.usda.gov/policies/ The federal government also provides a reinsurance mechanism that allows insurance companies to determine (within certain bounds) which policies they will retain and which they will cede to the government. This arrangement is referred to as the standard reinsurance agreement (SRA). The SRA is quite complex, with both quota-share reinsurance and stop losses by state and insurance pool. It essentially allows private insurance companies to adversely select against the government. This practice is considered necessary because the companies are required to sell policies to all eligible farmers and do not establish premium rates or underwriting guidelines. Costs There are three main components of federal costs associated with the U.S. program: 1. Federal premium subsidies range from 100 percent of total premium for CAT policies to 38 percent of premium for buy-up policies at the highest coverage levels. Across all FCIP products and coverage levels, the average premium subsidy in 2006 was 59 percent of total premiums. 2. The federal government reimburses administrative and operating expenses for the private insurance companies that sell and service FCIP policies. This reimbursement is approximately 22 percent of total premiums. 3. The SRA has an embedded federal subsidy with an expected value of about 14 percent of total premiums. On average, the federal government pays approximately 70 percent of the total cost for the FCIP. Farmer-paid premiums account for only about 30 percent of the total cost. Although the direct-farmer subsidy varies by coverage level, Table A1.3 shows that the United States has consistently passed legislation to increase the subsidy level to farmers for crop and revenue insurance products. The rate of subsidy is one component that has influenced the growth in overall premium. Figure A1.1 shows the overall growth of premium and premium subsidy. 11 Figure A1.1: Crop Insurance Premium and Subsidies in the United States 8.0 U . B nD llas 7.0 6.0 .S illio o r 5.0 Total Premium 4.0 3.0 2.0 1.0 Subsidy 0.0 1991 1993 1995 1997 1999 2001 2003 2005 Crop Year Source: Authors’ graph from USDA data. Canada 7 Public sector MPCI programs operate in 10 provinces as a partnership between the provincial and federal governments of Canada. Forms of government support The following are examples of how the government of Canada supports agricultural insurance: • Insurance legislation: The Farm Income Protection Act of 1991. • Subsidies on crop premiums: Provincial and federal governments subsidize crop premiums paid by growers. Subsides range from 50 percent to a maximum of 80 percent. • Subsidies on administrative costs of provincial insurance programs: Financed 50- 50 between provincial governments and the federal government—producers do not contribute toward the administrative costs of MPCI provision. • Stop-loss reinsurance: Federal government reinsures five provinces with stop-loss reinsurance (the province of British Colombia purchases private international stop-loss reinsurance). • Farm income disaster program: Shared by provincial and federal governments. In 2003, Canada revised its agricultural risk-management programs. The “business risk managementâ€? element of the new Agricultural Policy Framework is composed of two main schemes: production insurance (PI) and income stabilization. 7 Information in this section is based on Pikor and Wile, 2004. 12 The PI scheme offers producers a variety of multiple-peril production or production value loss products that are similar to many of those sold in the United States. A major distinction is that the Canadian program is marketed, delivered, and serviced entirely and jointly by federal and provincial government entities, although the provincial authorities are ultimately responsible for insurance provision. This arrangement allows provinces some leeway in tailoring products to their regions and offering additional products. PI plans are offered for more than 100 different crops, and provisions have also been made to include plans for livestock losses. Crop insurance plans are available, based on individual yields (or production value in the case of certain items, such as stone-fruits) or area-based yields. Unlike the U.S. program, Canadian producers are not allowed to insure different parcels separately, but rather must insure together all parcels of a given crop type. This means that low yields on one parcel may be offset by high yields on another parcel when determining whether or not an overall production loss has occurred. Insurance can also be purchased for loss of quality, unseeded acreage, replanting, spot loss, and emergency works. The latter coverage is a loss-mitigation benefit meant to encourage producers to take actions that reduce the magnitude of crop damage caused by an insured peril. By 2006, cost sharing between the federal government and each province for the entire insurance program was to be fixed at 60-40, respectively. However, federal subsidies as a percentage of premium costs vary from 60 percent for catastrophic loss policies to 20 percent for low-deductible production coverage. Combined, federal and provincial governments cover approximately 66 percent of program costs, including administrative costs. This is roughly equivalent to the percentage of total program costs borne by the federal government in the U.S. program. Provincial authorities are responsible for the solvency of their insurance portfolio. In Canada, the federal government competes with private reinsurance firms by offering deficit financing agreements to provincial authorities. Beginning in 2004, the Canadian Agricultural Income Stabilization (CAIS) scheme replaced and integrated former income stabilization programs. CAIS is based on the producer-production margin, where a margin is “allowable farm income,â€? that includes proceeds from production insurance, minus “allowable (direct production) expenses.â€? The program generates a payment when a producer’s current-year production margin falls below that producer’s “reference margin,â€? which is based on an average of the program’s previous five-year margins, less the highest and lowest. One important feature of CAIS is that producers must participate in the program with their own resources. In particular, a producer is required to open a CAIS account at a participating financial institution and deposit an amount based on the level of protection chosen (coverage levels range from 70 percent to 100 percent of the reference margin). Once producers file their income tax returns, the CAIS program administration uses the tax information to calculate the producer’s program-year production margin. If the program-year margin has declined below the reference margin, some of the funds from producer’s CAIS account will be available for withdrawal. Governments match the producers’ withdrawals in different proportions for different coverage levels. 13 The total investment by federal and provincial governments for the “business risk managementâ€? programs is CAN$1.8 billion per year. In 2004, governments provided roughly CAN$600 million as insurance premium subsidies. Table A1.4 provides an example of how different types of crop insurance are funded. It shows agricultural insurance subsidies administered by Agriculture Financial Services Corporation (AFSC). 8 In the province of Alberta, 80 percent of premium subsides are covered jointly by the federal government and the Alberta provincial government for coverage at 50 percent or below. Table A1.4: Alberta Province Agricultural Insurance Subsidies Premium Financing Indemnities Administration Costs (Subsidies) Multiple-Peril Crop Insurance Coverage up to Canada 40% Crop Insurance Fund Alberta 50% 50%a of risk Alberta 40% and Reinsurance Canada 50% Producers 20% Funds Coverage greater than 50% up to 80% Producers 50% and hail Alberta 30% endorsement Canada 20% Farm Income Disaster N/A Alberta 100%b Producer fees cover about 25%; Alberta covers balance Crop Hail Insurance Producers 100% Hail Insurance Fund Producers 100% a. Coverage does not include hail endorsement b. In accordance with the cost sharing between the province of Alberta and the government of Canada, the government of Canada provides companion program funding to Alberta to be used for agricultural programs. The province of Alberta, in turn, may provide some of these companion program funds to the AFSC to fund the indemnities of the Farm Income Disaster Program. Sources: Authors, from AFSC data. Spain The agricultural insurance system in Spain is structured around an established public- private partnership known as AGROSEGURO (Agrupación Española de Entidades Aseguradores de los Seguros Agrarios Combinados), formed in 1980 to provide farmers with insurance for crop, livestock, aquaculture, and, most recently, forestry. 8 The AFSC is the Canadian financial crown corporation of Alberta that has been offering growing degree day (GDD) products to maize farmers in the province since 2000. 14 Forms of government support The following are examples of how the government of Spain supports agricultural insurance: • Insurance legislation; • Subsidies on agriculture insurance premiums paid by farmers and herders; • Coinsurance and reinsurance through the Insurance Compensation Consortium or CCS. Key parties involved include the following: • Administrator: ENESA (The National Agricultural Insurance Agency or La Entidad Estatal de Seguros Agrarios) coordinates the system and manages resources for subsidizing insurance premiums. • Coinsurers: In 2005, there were 36 private and mutual Spanish and international insurance companies and the state catastrophe reinsurer, CCS. CCS had 12.5 percent market share, and 23 companies had a market share of less than 1 percent each. The coinsurance pool clearly illustrates the principle of a large number of companies pooling risks, with each company bearing a small share of risk. • Managing underwriter: AGROSEGURO, owned by the 36 shareholders/coinsurers, has been appointed by the coinsurers to underwrite, adjust, and settle claims on their collective behalf. • International commercial reinsurers: Providers of (1) stop-loss reinsurance to pool reinsurers on their viable line retentions and (2) multiyear catastrophe stop-loss insurance to CCS. The public sector entities are ENESA, which coordinates the system and manages resources for subsidizing insurance premiums, and CCS, which, together with private reinsurers, provides reinsurance for the agricultural insurance market. Local governments are involved only to the extent that they are allowed to augment premium subsidies offered at the national level. On the private side, insurance contracts are sold by AGROSEGURO. Farmers, insurers, and institutional representatives are all part of a general commission hosted by ENESA that functions as the managing board of the Spanish agricultural insurance system. Similar to the United States and Canada, insurance policies offer multiple-peril coverage in a combined program. However, only a very small fraction of AGROSEGURO’s overall liability is MPCI, and the company underwrites hail and named peril policies, as well as a large livestock portfolio, both of which are entirely separate from the MPCI drought and flood policies. Policies are available for crops, livestock, and aquaculture activities, with these risks being pooled across the country by AGROSEGURO. Unlike the United States and Canada, farmer associations are more actively involved in implementation and development of agricultural insurance. Government has reserves to cover extreme losses, and as a final resort, the government treasury is used to cover losses that may occur beyond these reserves. 15 Total premiums for agriculture insurance policies purchased reached about $550 million (€490 million) in 2003, of which approximately $225 million (€200 million) was provided by the government (Burgaz, 2004). Up-front premium subsidies (from both the central government and the autonomous regions) were equal to 56 percent of the total premium volume on average in 2004. The 2006 state budget to support costs of agricultural insurance premiums was $315 million (€240 million). The rationale for subsidizing agricultural insurance is that it will serve as a disincentive for the government to also provide free ad hoc disaster assistance. To reinforce the point, Spanish producers are not eligible for disaster payments for perils for which insurance is offered. For noncovered perils, ad hoc disaster payments are available, but only if the producer has already purchased agricultural insurance for covered perils. AGROSEGURO In 2005, AGROSEGURO underwrote almost 400,000 crop and livestock policies with a net premium volume of €480 million. The same year, a total of 1,808,000 crop claims and 610,000 individual animal claims were adjusted and settled, with a claim cost of €609 million and an overall loss ratio of 127 percent. The year 2005 was exceptionally bad for spring frost and drought losses. In 2004, AGROSEGURO employed a full-time professional staff of 249, comprising management, technicians, administrators, and clerical support staff. AGROSEGURO maintains its own network of 375 trained and specialized crop loss adjusters located throughout Spain. In addition, the company retains a network of 139 livestock loss adjusters. Viable lines include the less volatile and lower-risk crop programs, which are insured by the pool of 45 commercial insurers. Experimental lines are insured directly by CCS and include the more volatile crops and peril combinations, including the systemic perils of drought and flood, which can lead to catastrophe losses. Drought is only offered as an experimental coverage for selected crops and programs (the integral winter cereal program, for example, and yield-shortfall policies for wine grapes); flood has been included as an insured peril only since 1999. To minimize antiselection, flood is a compulsory peril on all crop insurance lines; drought and flood are offered only on experimental lines and are reinsured by CCS. In 2005, AGROSEGURO underwrote about 200 viable and experimental crops, livestock, and marine aquaculture lines, plus forestry insurance covering a wide range of crop types, including cereals, oilseeds, horticultural crops, leaf and fibers, tree fruits and vines, and livestock types. The company offers a comprehensive range of single-peril hail, named-peril, and multiple-peril crop insurance policies. In every year from 2001 to 2005, excluding 2002, the experimental loss ratio has been higher than the viable loss ratio. 16 AGROSEGURO reinsurance AGROSEGURO reinsurance structure in 2005, offered via CCS, provides the following stop-loss reinsurance protection: • Layer 1: 50 percent of all losses up to 90 percent of GNPI; • Layer 2: 95 percent of 40 percent in excess of 90 percent of GNPI; • Layer 3: 90 percent of 30 percent in excess of 130 percent of GNPI; • Layer 4: 100 percent of losses in excess of 160 percent of GNPI (unlimited stop loss). CCS allows AGROSEGURO coinsurers to purchase stop-loss reinsurance on their 5 percent retention on Layer 2 and on their 10 percent retention on Layer 3. For experimental lines, CCS provides excess of loss reinsurance for any loss in excess of $792,000 (€600,000). The loss ratio for viable and experimental claims as a percentage of premiums was 224 percent in 2005. However, the CCS multiyear stop-loss retrocession treaty (with international reinsurers that reinsure for losses of 40 percent in excess of 145 percent of GNPI) was free of claims in 2005 (the actual loss ratio of 125 percent was well below the priority). In addition, this stop-loss retrocession treaty was claims-free for the full five-year contract from 2001 to 2005. Portugal In Portugal, the Integrated System of Protection against Climatic Events (Sistema Integrado de Protecção Contra as Aleatoridades Climáticas—SIPAC) is a program underwritten exclusively through private sector insurers. Approximately 15 commercial insurance companies are registered and approved by the Portuguese Insurance Institute, or Instituto de Seguros de Portugal (ISP), to underwrite the Uniform Crop Insurance Policy. Forms of government support Government support to agricultural insurance in Portugal includes the following: • Insurance legislation • ISP-approved uniform crop policy, underwritten by all crop insurers • Basic coverage hail plus fire • Complementary coverage hail plus fire, tornado, excess rain, frost and snow • Setting of reference rates for calculation of premium subsidies • Subsidies on crop insurance premiums paid by growers • Stop-loss reinsurance protection 17 For each crop in each region, ISP quotes reference rates , which are used to calculate premium subsidies. In 1999, the minimum premium subsidy levels available to all growers were 40–45 percent for the basic coverage (hail and fire). Growers are eligible for additional subsides of 10–20 percent if they elect to purchase the more expensive multiple-peril policy (complementary coverage). Crops with higher reference rates carry additional subsidies (20 percent in the case of crops with reference rates greater than 8.4 percent). Until 1999, the maximum subsidy level was 85 percent. In 2000, the maximum premium subsidy level was reduced to 65 percent. The government of Portugal has provided financial support through the reinsurance program (Fondo de Calamidades) since 1996. The implementing agency is the Instituto de Financiamento e Apoio ao Desenvolvimento da Agricultura e das Pescas. The government’s stop-loss program is optional. It is a nonproportional stop-loss reinsurance program with limits of liability for the direct insurer that are established for each climatic risk zone. Mexico Between 1990 and 2001, Mexico transitioned to a combination of private and public sector crop and livestock insurance, moving to private sector crop and livestock insurance in 2001. Since 2001, however, Agroasemex, a government-owned reinsurance company operating exclusively in agricultural insurance, has been reinstated as a public sector agricultural reinsurer. There are currently four commercial companies that underwrite crop and livestock policies. Peril and multiple-peril crop insurance products are available for a wide range of crops throughout Mexico. Annual crop insurance premiums average around $60 million. Forms of government support Government support to agricultural insurance in Mexico includes the following: • Insurance legislation • Creation of state-owned reinsurer, Agroasemex • Technical assistance and ongoing training and education by Agroasemex to the fondos • Subsidy Agroasemex Agroasemex relies heavily on the traditional reinsurance market to protect its agricultural portfolio from inordinate losses. While searching for new alternatives after a 70 percent increase in the retrocession rates of 2001, Agroasemex investigated the comparative efficiency of the weather derivatives market. This case study presents the background and guiding principles to designing the weather derivative structure that was ultimately used as a hedge for the Agroasemex agricultural portfolio, and to structuring the Agroasemex weather risk transfer program. It is noteworthy that the institution’s weather derivative transaction in 2001 was the first of its kind in the developing world. 18 From 1990 to 2001, the state-owned Agroasemex offered direct insurance to farmers. In 2001, the government decided to take the next step toward consolidating a private agricultural insurance market. Agroasemex was transformed into a reinsurance company specializing in the agricultural market, offering reinsurance to both the fondos 9 and private companies. The government decided that Agroasemex would also become a development agency dedicated to research on alternative insurance and reinsurance schemes that would promote the development of the agricultural insurance market through participation of both private companies and fondos, and Agroasemex reoriented its support to the fondos system by providing technical assistance and ongoing training for their registration, formation, operation, and reinsurance coverage. Figure A1.2 depicts the Mexican agriculture insurance system after the public policy change in 2001. Figure A1.2: Mexican Insurance System after Public Policy Change FEDERAL GOVERNMENT AGROASEMEX INTERNATIONAL (Reinsurer) REINSURANCE PRIVATE FONDOS COMPANIES FARMERS Source: Ibarra and Mahul 2004. In 2000, the agricultural insurance market had been almost equally distributed between Agroasemex, the fondos, and the private companies. In 2001, when Agroasemex became a reinsurer, the private companies were the first to benefit, but in 2002 the fondos became the dominant player in the agricultural insurance market with 862,765 hectares (53.7 percent of the market) insured. The average insured area per fondo is 3,300 hectares, and the total number of farmers associated as members of the fondos system is between 70,000 and 90,000. In 2003, more than 240 fondos provided agriculture insurance services to their members, accounting for 50 percent of the total insured agricultural area in Mexico. According to Mexican laws, fondos are nonprofit organizations constituted by the farmers as civil associations without the need to provide any capital endowment, except their willingness to associate among themselves. From a risk-financing perspective, fondos pool crop-yield risks from farmers with similar risk profiles. The concept of insurance through mutual- type organizations was developed in Mexico based on a sound insurance market approach (including proper underwriting of risks based on technical principles, constitution and 9 Fondos are self-insurance funds that have been operating in Mexico since 1988. 19 investment of adequate financial reserves, and loss adjustment procedures based on technical guidelines and rates developed according to sound actuarial methodologies). In addition, there are advantages to the mutual-type organizational principles and structure of incentives that can keep transaction costs under control. The fondos are managed by the farmers on a democratic basis, and the Treasury Ministry supervises their activities (see Box A1.2). In the case of Mexico, the state-owned reinsurance company, Agrosemex, provides the fondos with technical assistance and unlimited reinsurance capacity. The fondos experience in Mexico is a valuable example of how mutual-type organizations can provide agricultural insurance services on a profitable basis in developing countries. The fondos system illustrates how market-based agricultural insurance, with limited and targeted subsidies, can enable asset-poor farmers involved in profitable business activities to maintain the viability of their business in the long run despite their exposure to weather-related production shocks, while asset-poor farmers involved in activities with a lower return must rely on heavily subsidized social insurance (Gurenko and Mahul 2004). Box A1.2: Institutional Design of the Fondos Approval of new fondos. The Treasury Ministry is responsible for approving the application of a new fondo. To be eligible, the interested farmers present a feasibility study conducted by the licensed insurance-reinsurance institutions interested in offering reinsurance to the new fondo. The feasibility study includes all technical and financial information necessary to demonstrate the feasibility of the proposed fondo. Therefore the feasibility of the new fondo is evaluated on ex ante and technical criteria. Formal group agreement to form a fondo. The objectives and outreach of the organization are defined, and the proposed bylaws are approved and formally registered, in a session of a General Assembly of farmers wanting to form the fondo. All fondos formally register their bylaws with a public notary. Any group of farmers has the right to name a formal representative to represent the voting interests of that group to the General Assembly. Having a representative provides a transparent information mechanism that helps members of the fondo understand the types of services offered and gives members a voice when key decisions for the fondo are evaluated in the Assembly. Reports of financial situation at an individual fondo level. The Treasury Ministry is the authority that defines the type and content of the reports regarding the organization, operation, accounting, investments, reserves, and any other aspect related to the operation of the fondo. Each fondo submits a balance sheet and income statement to the Treasury Ministry on an annual basis, so the Ministry can evaluate the financial situation of each fondo. These financial statements are audited by a licensed external accounting firm and approved by the General Assembly Area of influence. A fondo operates only in the region where its members (farmers) conduct their crop production. The area of influence is included in the request for registry and is approved by the Treasury Ministry. Limiting the geographical area helps maintain a competitive advantage when underwriting risk and following up the loss reports, and maintains control over loss adjustment procedures. The area of influence can be 20 broadened only with permission of the Treasury Ministry and the internal approval of the General Assembly of the fondo. Decision making for the fondos. The final authority for approval of important issues regarding the fondos is the General Assembly, ensuring that important decisions affecting the fondos are made by the majority of farmers in the fondo and not by small interest groups. New memberships approval. All applications for new members are approved by the General Assembly of the fondo. The request is presented with a technical and economic feasibility analysis by crop, emphasizing the real or estimated loss ratio for at least the previous five years. Members of the fondo can formally assess the risk exposure of the proposed new members. This selection process for new members is an important mechanism that controls adverse selection. Source: Ibarra and Mahul 2004. India India is the second-most populous country in the world, with more than a billion people. Seventy percent of the population lives in rural areas, and 60 percent of the surface area is agricultural land. With 120 million farmers in India, agriculture accounts for 20 percent of the country’s GDP. Although commercial agriculture is a relatively small part of the economy, two-thirds of the population depends on agriculture for their livelihood. The livestock sector is very large, with 220 million cattle, 124 million goats, 94 million buffalos, and 58 million sheep, in addition to camels and poultry. Livestock insurance is available, but only a very small proportion of livestock is insured. India has no rival when it comes to the number of agricultural insurance policies sold. In 2005, 18 million Indian farmers were insured under the National Agricultural Insurance Scheme (NAIS). Like many lower-income countries, India’s agricultural sector is dominated by small farms, with an average farm size of 1.5 hectares. As early as 1920, the Indian scholar J. S. Chakravati recognized that this condition required a unique approach in structuring a crop insurance program. He proposed an area-based approach— basing losses on an index of area yields and foregoing the nearly impossible task of measuring farm-level yields on such small farms—as the most suitable structure for Indian agriculture. . The challenge is how to provide risk-management solutions for India’s millions of smallholder farmers while operating a sustainable system. A 1947 crop insurance study commissioned by the Indian government favored homogenous area-based insurance over an individual approach, but the concept was not adopted by the states. The first crop insurance program, introduced in 1972, paid based on individual farm-level losses. The scheme covered only 3,000 farmers and had a loss ratio of claims to premium that exceeded 8. An area-based approach, introduced to Indian farmers in 1979 under the Pilot Crop Insurance Scheme, paid indemnities for yield shortfalls based on the average area yield as measured by crop-cutting samples. The insurance was available on a voluntary basis to 21 farmers obtaining agricultural loans. Linking insurance to credit has been a consistent feature of subsequent schemes. In 1985, the area-index MPCI yield-shortfall program was expanded into 16 states and 2 union territories under the name of the Comprehensive Crop Insurance Scheme (CCIS) and implemented by the General Insurance Corporation of India. The Agricultural Insurance Company of India (AIC) was created in 2002 by government decree to improve the performance of the NAIS, which has been in place since 1999. The program has expanded to nearly every state and territory in India during recent years. The government crop insurance is compulsory for borrowers of crop credit and voluntary for nonborrowers. Only about 20 percent of the farmers who purchased crop insurance are nonborrowers. Until now, the program was not operated on an actuarial basis, which explains in part the poor actuarial performance—indemnities exceeded premiums by 4 to 1 over the period 1985–2002. Furthermore, one state received more than half of the indemnities and had a loss ratio greater than 10 to 1 for the same period. Forms of government support Government support to agricultural insurance in India includes the following: • Insurance legislation • Insurance schemes • Agriculture Insurance Company of India • Subsidy Throughout the history of the CCIS and NAIS, the program has been heavily subsidized by both the government of India and the state governments on a 50-50 basis. The subsidies take several forms including: • Maintaining average premium rates between 1.5 percent and 3.5 percent, or well below actuarial rates, to make crop insurance as widely affordable to India’s farmers as possible; • Premium subsidies, which in 2006 are provided only to small and marginal farmers with less than 2 hectares; • Excess of loss reinsurance for claims excess of 100 percent loss ratio (food crops and oil seeds) and 150 percent loss ratio (commercial crops and horticultural crops); • Subsidies on AIC’s administrative and operating expenses. India’s crop insurance programs have undergone many changes in an effort to develop programs that serve the needs of farmers and are financially sustainable over the long run. The AIC is working to expand its market and to move toward an actuarial rating system. It is also working to address many of the shortcomings of the government program. One significant shortcoming has been the timeliness of payments. Given that it can take from six months to more than one year to estimate area yields with crop-cutting data, there are times when farmers do not receive payment before they must plant another crop. The government is moving toward blended products to address this problem. In this 22 case, the blended product involves a more timely payment that can be made using a rainfall-index insurance—and a second payment, if warranted, using the area-yield estimates. This type of double-trigger policy offers a model that merits further investigation for agricultural insurance programs for lower-income countries, which also offer the opportunity to combine a government program and a private sector product. Rainfall-index insurance products Partially in response to some of the shortcomings of the government program, ICICI Lombard of Mumbai, the largest private general insurance company in India, introduced new rainfall insurance products in 2003, with technical assistance supported by the World Bank. ICICI Lombard recognized that either deficit or excess rainfall accounted for the vast majority of crop losses in India. The timing and intensity of annual monsoons has a great influence on agricultural production. A major advantage of the new rainfall insurance products is that payments can be determined much more quickly given that the only information needed is how much it has rained. In 2003, the microfinance institution BASIX partnered with the insurance company ICICI Lombard to introduce index-based rainfall insurance on a pilot basis in the Mahahbubnagar District of Andhra Pradesh. By the third insurance season, the program had expanded to 7,685 policies issued to 6,703 customers in 36 locations in 6 states during the 2005 monsoon season. The expansion of the rainfall insurance program can be attributed largely to the effort of BASIX to modify and improve the product according to client feedback. BASIX has invested considerably in improving the accessibility of the product by simplifying the delivery system, training agents, and incorporating new technology. Other insurance companies, including the government company Agricultural Insurance Corporation of India (AICI), are now developing their own weather insurance products. The public agriculture insurance company AIC has introduced a rainfall insurance product since 2004 and is pilot testing index-based rainfall and yield insurance for coffee. In 2005, AICI sold about 120,000 weather-based insurance policies for a total value of about $5 million, and its market share was 50 percent. Nevertheless, with only three years of sales, it is still too early to tell how well the products will perform in the long term. Conclusions The six case studies discussed above provide examples of a wide array of government support and intervention in crop insurance programs. Many developed countries, such as the United States, Canada, Spain, and Portugal, have instituted agricultural programs which are highly subsidized by the public sector. The costs of premium subsidies, insurance administration, and loss assessment create large financial obligations for these countries. Such prohibitively high costs have encouraged more innovative approaches in countries with more limited financial resources, such as Mexico and India. The partnership between the state-owned reinsurance company Agroasemex and the fondos demonstrates how market-based agricultural insurance, with limited and targeted government support, can be successful alternatives in managing weather risk. The rainfall insurance products introduced by in India are another innovative way market-based 23 insurance has been successful in reducing the costs associated with agricultural insurance. Such market-based approaches, tailored to the specific country, have potential to manage the costs of weather risk in a more sustainable way. 1.5. References Burgaz, F. J. “Il Sistema Delle Assicurazioni Agricole Combinate in Spagna.â€? La Gestione Del Rischio in Agricoltura: Strumenti E Politiche. Rome, Italy: Edizioni Tellus, 2004. Chakravati, J. S. Agricultural Insurance: A Practical Scheme Suited to Indian Conditions. Bangalore, India: Government Press, 1920. Glauber, J. W. “Crop Insurance Reconsidered.â€? American Journal of Agricultural Economics 86 (2004): 1179–1195. Gurenko, E. and O. Mahul (2004). Enabling productive but asset-poor farmers to succeed: a risk financing framework.. World Bank Policy Research Working Paper 3211. Hazell, P. B. R. “The Appropriate Role of Agricultural Insurance in Developing Countries.â€? Journal of International Development 4 (1992): 567–581. Ibarra, H. and O. Mahul. “Self-Insurance Funds as Agriculture Insurance Providers: The Case of FONDOS in Mexicoâ€? Working Paper from the Contractual Savings and Insurance Practice, Financial Sector Operations and Policy Department, World Bank, November 2, 2004. Pikor, G., and A. Wile. “L'esperienza Canadese Nella Gestione Del Rischio in Agricoltura.â€? La Gestione Del Rischio in Agricoltura: Strumenti E Politiche. Rome, Italy: Edizioni Tellus, 2004. Skees, Jerry “The Bad Harvest: Crop Insurance Has Become a Good Idea Gone Awry.â€? Regulation: The CATO Review of Business and Government 24 (2001): 16–21. Skees, Jerry “Agricultural Risk Management or Income Enhancement?â€? Regulation: The CATO Review of Business and Government 22 (1999): 35–43. 24 Annex 2: Agricultural Risk Assessment This annex presents a risk assessment for each of the four provinces or municipalities that were visited and that are the focus of this report—Heilongjiang, Xinjiang, Shanghai, and Hainan—and emphasizes differences in risk exposure between different crops and geographic regions. Further, to illustrate some important concepts regarding risk pooling, a hypothetical insurance company is assumed for each province or municipality. The hypothetical insurance company shows how overall risk exposure decreases when risks are spread across multiple crops and geographic regions. 2.1. Approach A risk assessment begins with an estimate of the probability that losses of various magnitudes will occur. This probability estimate is also the basis for developing insurance premium rates. In this annex, these estimates are presented in tables that show the average annual loss and in figures that contain loss exceedance curves. A loss exceedance curve shows the probability that losses will exceed various magnitudes. These measures, however, are not adequate for determining the amount of contingent capital that might be required to pay indemnities in the event of extreme losses. To know how much contingent capital (reinsurance, reserves, and so on) might be required, the insurer needs some estimate of how large the indemnities might be, relative to premiums collected. This part of the risk assessment is based on loss ratios (indemnities divided by premiums) for a hypothetical insurance company that diversifies its portfolio of crop insurance policies across different crops and geographic regions. For each of the four provinces or municipalities, pilot counties or state farms were targeted for risk assessment. Between 15 and 20 years of yield data were obtained and analyzed for these counties and farms. In addition, supplementary information was obtained through discussions with agricultural experts in each of the provinces or municipalities. No risk data on livestock mortality were available, so the risk assessment focuses exclusively on crop-yield risk. It is important to note that the available yield data are generally at the township level for counties and the team level for state farms. 10 Due to aggregation bias, a risk assessment based on aggregate data will systematically underestimate risk at lower levels of aggregation (such as village or household). Thus, the risk assessment presented here should be interpreted as a minimum estimate of the actual risk exposure at lower levels of aggregation. It is also important to note that the loss ratios for the hypothetical insurance companies are calculated based on an assumption that multiple-peril crop insurance is purchased by each township or team in the pilot counties or farms. The insurance is assumed to have no 10 In the case of Shanghai, the available yield data were at the county or district level. 25 deductible and actuarially fair premium rates (that is, over the time period analyzed, premium rates for each county or farm were set so that total premiums collected were exactly equal to total indemnities paid). Despite these assumptions, it is important to note that it is not recommended that insurers in China: (1) adopt multiple-peril crop insurance; (2) sell crop insurance policies with no deductible; or, (3) set crop insurance premium rates at actuarially fair levels. These assumptions were adopted for this analysis for several reasons: First, township- or team-level data were not available on the magnitude of yield losses caused by specific perils. Thus, the assessment of loss ratio exposure must be based on an assumption of multiple-peril insurance. Second, for crop insurance sold to farm households, 20–30 percent deductibles are typical. However, given that the available yield data are at the township or team level rather than the farm household level, the calculations assume no deductible, which should generate loss ratios that are roughly equivalent to what one would expect using typical deductibles at the farm household level. Third, crop insurance premium rates are typically loaded for things like operating expenses, reserve building, and return on equity. However, applying a premium load will reduce overall loss ratios. Thus, actuarially fair premium rates are assumed to generate conservative estimates of possible loss ratio exposure. Details on the statistical procedures used throughout this risk assessment are described in the last section of the annex. Due to data limitations, and the assumptions required to compensate for those data limitations, insurance companies or other decision makers should not make decisions based on specific quantitative estimates provided in this annex. The purpose of this annex is solely to demonstrate some basic concepts regarding: (1) risk assessment measures and procedures; (2) relative differences in risk across crops and regions; and (3) risk reduction through pooling across crops and regions. Agricultural Data Collection System Provincial statistical bureaus collect data on agricultural variables, including hectares planted and tons produced for major crops. These data are available at the township, county, and provincial levels. State farms collect similar data at the farm and team levels. The provincial statistical bureaus are affiliated with the National Bureau of Statistics, which prescribes the methods used for data collection. Data on hectares planted are collected through survey samples. Data on tons produced are estimated based on yields measured at a number of sampling points. Livestock data are collected annually via a census but do not include information on livestock mortality. In 1997, the U.S. National Agricultural Statistics Service began working with the National Bureau of Statistics to upgrade sampling procedures used in China and to implement multilayer consistency checks. The latter helps to ensure that the summation of production data collected at lower levels of aggregation (such as townships) is consistent with external evidence measured at higher levels of aggregation (for example, data on commodities being processed or exported). If external consistency checks indicate that aggregate production has been underestimated or overestimated, the National Bureau of Statistics will investigate the discrepancy. If need be, the estimates at 26 lower levels of aggregation will be scaled until the summation is consistent with the external evidence. These changes, introduced in 1997, were fully implemented by 2003. County- and township-level data are available from the National Bureau of Statistics or provincial statistical bureaus. These data do not include information for state farms. Provincial- and national-level data do include information for state farms and are available on the Web site of the National Bureau of Statistics. Data on hectares planted and tons produced were used to conduct a risk assessment for the four provinces and municipalities that are the focus of this report. Various mechanisms were used to test for both spatial and temporal consistency in the data on hectares planted and tons produced. No problems were noted with the provincial level data. For the township- and team-level data, only infrequent problems were noted on spatial consistency. These problems appeared to be random outliers caused by recording or keypunching errors. When the nature of the problem was obvious (for example, a decimal out of place) the outlier was corrected. When it was not possible to discern the nature of the problem, the outlier was treated as a missing observation. There were, however, frequent and pervasive problems with the temporal consistency of the township- and team-level data. The supporting information indicated that planted area was measured in hectares and the amount produced was measured in metric tons. However, this was often not the case. In fact, a common problem was that the units used to measure planted area and amount produced were not consistent throughout the time series. The measure of planted area sometimes switched between mu and hectares, and in some cases the measure of amount produced switched between thousands of jin and metric tons. Similarly, the scale of measure would sometimes change through the time series. For example, planted area would be measured in hectares for some years and in thousands of hectares for other years. The combined inconsistencies in the unit of measure and scale of measure, and the fact that these inconsistencies were different across counties and farms, made it extremely difficult to determine the exact nature of the data problems. Each of the pilot counties and farms had unique data problems that had to be investigated and resolved. After the data were cleaned, provincial- and township- or team-level annual yields were calculated by dividing the tons produced by the hectares planted for each year. To account for changes in technology, production practices, and other widespread structural changes, the provincial- and township- or team-level yields were trend-adjusted using the estimated provincial-level yield trends. (See Section 2.6, below). This change effectively converts historical yields into current equivalent values. Weather Data Collection System It was not possible to make a detailed study of the available quantity and quality of weather data. However, China is a member of the World Meteorological Organization (WMO), and the China meteorological service, through its bureaus in each province, conform to WMO standards. There is a good network of meteorological stations, and daily weather data are generally available for these stations, though not necessarily in electronic format for the full period of recording. Additional sources of weather data 27 include various hydrological organizations and state farms. More specific information on weather data availability in each of the four provinces or municipalities is provided below. In the provinces visited were many organizations—meteorological bureaus, provincial academies, universities—with expertise in agro-meteorology. In addition, there is a good level of technical know-how at the level of state farms and reclamation bureaus. 2.2. Heilongjiang Agricultural land in Heilongjiang is controlled either by state farms or collectives of independent farmers. Land that is controlled collectively is administered by farmer organizations (for example, village committees). The Heilongjiang Reclamation Group (HRG) administers most of the state farms in the province. State farms under the HRG are administered through a two-tiered system that includes the farm and individual production teams. HRG has 103 farms and approximately 2 million hectares of farmland (20 percent of the total farmland in Heilongjiang). Heilongjiang is located in the northeastern part of China in what is classified as a frigid temperature zone. The major crops produced in the province are soybeans, maize, and rice. Heilongjiang has almost 4 million hectares planted in soybeans and more than 2 million hectares in maize. Over the past 26 years, rice production has increased significantly while wheat production has decreased. Currently, there are approximately 1.6 million hectares of rice, up from about 0.2 million hectares in 1980. In 2004, rice and soybeans each accounted for about 29 percent of the total value of crop production in the province. Maize accounted for about 14 percent of the total crop value. The remaining 28 percent was from a variety of crops, including barley, flax, melon, millet, rapeseed, sorghum, sunflower seed, sugar beet, tobacco, tubers, various vegetable crops, and wheat. There are currently only about 0.25 million hectares of wheat, down from about 2 million hectares in 1980. The current provincial-level expected yield for soybeans is approximately 1.5 tons per hectare. Expected soybean yields trended upward until the late 1990s before being pulled down by yield shortfalls in 2000, 2001, and 2003. Agricultural experts in Heilongjiang indicated that in recent years soybean production had moved from the more drought- prone southern part of the province to the less drought-prone north. Still, some soybeans are produced in the south as a rotational crop with maize. The current provincial-level expected yield for maize is approximately 4 tons per hectare. Expected maize yields trended upward until the mid 1990s but, due to drought, have trended downward ever since. Typically, maize is planted in non-irrigated fields in the southern part of the province. The current provincial-level expected yield for rice is approximately 6.8 tons per hectare. Expected rice yields have trended steadily upward since 1980 when the expected yield was only 3 tons per hectare. Rice is generally produced in irrigated fields in the southern part of the province. Maize is planted in Heilongjiang between April 20 and the end of June, depending on latitude and weather conditions. Soybeans are planted during the same period. In regions that produce both maize and soybeans, soybeans are generally planted 5–10 days after the 28 maize has been planted. Rice is typically transplanted from the nursery to the field in early May. All three crops are harvested in about October. Meteorological Stations There are 83 meteorological stations in Heilongjiang, operated by the provincial weather administration, 32 of which participate in international data exchange. Some automated stations have been in place since 2001. HRG has 90 additional stations and the hydrological services operate stations and rain gauges. Prior to 2001, HRG operated weather stations independently, but since that time they have been integrated into the provincial meteorological weather administration. A typical distance between stations was advised as 50 kilometers. Data collections started as early as 1952 in some locations. Monthly data were available for 32 internationally reporting stations covering the period 1971 to 2000. Prior to 1991, data in the HRG area is not in an electronic format. Generally, daily data are not recorded over a long period in electronic format, except for those stations that report internationally as part of the WMO network, but daily records are still available at the stations. Data for Pilot Counties and Farms The pilot areas selected for analysis in Heilongjiang were Nenjiang County and Qixingpao Farm (HRG) in the northern part of the province, Jixian County and Farm 291 (HRG) in the central part of the province, and Shuangcheng City and Hongguang Farm (HRG) in the southern part of the province. For each of the six pilot counties or farms annual data were provided on hectares planted and tons produced for major crops. In the northern part of the province, township-level data were available for Nenjiang County for the period 1986–2005, and team-level data were available for Qixingpao Farm (HRG) for the period 1984–2004. In the central part of the province, township-level data were available for Jixian County for the period 1985–2005, and team-level data were available for Farm 291 (HRG) for the period 1984–2005. In the southern part of the province, township-level data were available for Shuangcheng City for the period 1985–2004, and team-level data were available for Hongguang Farm (HRG) for the period 1986–2004. In addition, provincial-level data on planted hectares and tons produced were available for major crops for the period 1979–2004. Monthly rainfall and temperature data were available for the weather stations nearest each pilot county or farm. If a specific weather peril was a significant cause of loss, and if it seemed possible that the available historical weather data could measure the magnitude of that weather peril, then the weather data were examined to investigate potential relationships between yield shortfalls and extreme weather events. Data were also provided on hectares covered by, and affected by, natural disasters. Hectares covered by natural disaster are those that experience a yield loss of more than 10 percent. Hectares affected by natural disaster are those that experience a yield loss of more than 30 percent. These data were available at the provincial level and in some cases at the township and team levels. However, frequently the township- or team-level data on hectares covered by, and affected by, natural disasters were unusable due to inconsistencies in the data provided and large numbers of missing observations. Where 29 usable data were available, these data allowed for consistency checks against the yield data. However, for risk assessment their value is limited since they are not crop-specific nor do they indicate the exact amount of yield loss. Primary Perils Drought is the primary peril for non-irrigated crops produced in Heilongjiang. Expected rainfall varies across the province between 370 millimeters and 600 millimeters and falls principally in the summer months. Agricultural experts reported that up to 70 percent of all crop yield losses in Heilongjiang were due to drought. Drought is particularly problematic for corn and soybeans produced in the central and southern parts of the province. However, since 2001 Sunshine insurance company reports that only 36 percent of their indemnity payments were for losses caused by drought while 28 percent were for losses caused by flooding (including waterlogging). This may reflect the fact that Sunshine sells exclusively to HRG farms, many of which are on low-lying reclaimed land. Other perils are hail, disease, late frost in spring, early frost in autumn, and cold summer temperatures. The provincial data on hectares covered by, and affected by, natural disasters confirm that drought is a major peril. Between 1982 and 2004 almost 9 percent of the hectares planted in Heilongjiang had drought-related yield losses in excess of 30 percent. Another 10 percent of the hectares planted had drought-related yield losses of 10–30 percent. During the same period, almost 6.5 percent of the hectares planted had flood-related (including waterlogging) yield losses in excess of 30 percent and another 6.5 percent had flood-related yield losses of 10–30 percent. Table A2.1 provides insights into the spatial characteristics of these two perils. For selected years, the table shows the percentage of planted hectares (for all crops) with yield losses greater than 10 percent due to drought and flooding. The table also shows the provincial-level yield loss for soybeans and maize in the same years. The table demonstrates that yield losses caused by drought are much more spatially correlated than those caused by flooding. Yield losses in 1982, 1989, 2000, 2001, and 2003, were caused primarily by drought. In each of those years, provincial-level yields for soybeans and/or corn were significantly below expected values. In contrast, flooding was the primary cause of yield losses in 1981, 1985, 1988, 1994, and 1998. Flooding tends to be localized. Further, the excess rainfall that causes flooding or waterlogging in low-lying or poorly drained areas may actually increase yields in areas that are less flood-prone or have better drainage. Thus, despite extensive flooding in 1988, 1994, and 1998, there was little or no provincial-level yield loss for either soybeans or maize. In 1985 flooding (and also drought) caused provincial-level yield losses for maize but not soybeans. In 1981 flooding caused losses in soybeans but not maize. 30 Table A2.1. Heilongjiang Yield Losses Due to Drought and Flood Percentage of Planted Hectares (with Yield Losses of at least 10%) Caused Provincial-Level Soybean Provincial-Level Maize by Yield Loss Relative to Yield Loss Relative to Year Drought Flooding Expected Value Expected Value 2003 46% 13% -20% -4% 2001 37% 1% -18% -12% 2000 36% 0% -14% -3% 1998 5% 26% -3% No Loss 1994 15% 27% No Loss No Loss 1989 33% 10% -20% -23% 1988 7% 40% No Loss No Loss 1985 18% 28% No Loss -14% 1982 41% 1% -8% -13% 1981 5% 34% -12% No Loss Source: Authors’ calculation. At a provincial level, for the period 1979–2003 the largest soybean yield losses occurred during the drought years of 1989, 2000, 2001, and 2003. For these years, Table A2.2 shows the township- and team-level yield losses for the six pilot counties and farms. Although yield losses in 1989 were pervasive, the losses in 2000 seem to have been concentrated primarily in the southern part of the province. In 2001 and 2003, yield losses occurred in the north, central, and southern regions, but not for all pilot counties or farms. Table A2.2. Soybean Yield Losses in Selected Years North Central South Nenjiang Jixian Shuangcheng Qixingpao County County Hongguang City Weighted Farm Weighted Farm 291 Weighted Farm Average Weighted Average Weighted Average Weighted Township- Average Township- Average Township- Average Level Team-Level Level Yield Team-Level Level Yield Team-Level Year Yield Loss Loss Yield Loss Loss Yield Loss Yield Loss 2003 0% -41% -13% 0% -10% -1% 2001 -8% -1% -10% -22% -21% -3% 2000 -5% -2% -2% 0% -21% -37% 1989 -23% -30% -11% -41% -23% -41% Source: Authors’ calculation. 31 At a provincial level, for the period 1979–2003 the largest maize yield losses occurred in 1982, 1985, 1989, and 2001. Major droughts occurred in 1982, 1989, and 2001. Losses in 1985 were the result of both drought and flooding. For these years, Table A2.3 shows the township- and team-level yield losses for the four pilot counties or farms where maize is produced. As with soybeans, the 1989 yield losses were pervasive. The 1985 yield losses seem to have been concentrated in the central part of the province. In 2001, Farm 291 (HRG) experienced an 8 percent yield loss, and Shuangcheng City experienced a 4 percent loss, but there was no yield loss for either Jixian County or Hongguang Farm (HRG). No yield data is available for the pilot counties and farms in 1982. Table A2.3. Soybean Yield Losses in Selected Years Central South Farm 291 Hongguang Farm Jixian County Shuangcheng City Weighted Average Weighted Average Weighted Average Weighted Average Team-Level Team-Level Township-Level Township-Level Year Yield Loss Yield Loss Yield Loss Yield Loss 2001 -8% No Loss No Loss -4% 1989 -50% -24% -16% -39% 1985 -14% -15% NA -1% Source: Authors’ calculation. Risk Assessment of Pilot Counties and Farms Risk assessment was conducted for each of the six pilot counties and farms. Yield-loss exposure is presented using both tables and graphs. The tables present the average annual loss, and the graphs contain loss exceedance curves that indicate the probability of losses in excess of some magnitude. The procedures used to generate these measures are described in the last section of the annex. North Nenjiang County and Qixingpao Farm (HRG) are the pilot county and farm located in the northern part of Heilongjiang province. For both Nenjiang County and Qixingpao Farm, soybeans are the dominant crop, although both also produce wheat. Table A2.4 presents measures of average annual loss. Figures A2.1 and A2.2 present Nenjiang County and Qixingpao Farm yield-loss exceedance curves for wheat and soybeans. The results clearly show that, in this area, wheat is riskier than soybeans. Table A2.4: Nenjiang County and Qixingpao Farm Average Annual Loss Wheat Soybean Nenjiang County (township level) 14% 12% Qixingao (team level) 14% 12% Source: Authors’ calculation. 32 Figure A2.1: Heilonjiang—Nenjiang County Yield-Loss Exposure Heilongjiang: Nenjiang County Yield Loss Exposure (Township-Level Data) 100% 90% 80% 70% Probability 60% 50% 40% 30% 20% 10% 0% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Loss Relative to Expected Value Wheat Soybean Source: Authors’ calculation. Figure A2.2: Heilongjiang—Qixingpao Farm Yield-Loss Exposure Heilongjiang: Qixingpao Farm Yield Loss Exposure (Team-Level Data) 100% 90% 80% 70% Probability 60% 50% 40% 30% 20% 10% 0% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Loss Relative to Expected Value Wheat Soybean Source: Authors’ calculation. 33 Central Jixian County and Farm 291 (HRG) are the pilot county and farm located in the central part of Heilongjiang province. Rice, maize, and soybeans are the primary crops produced in Jixian County and Farm 291. Table A2.5 presents measures of average annual loss. Figures A2.3 and A2.4 present Jixian County and Farm 291 yield-loss exceedance curves for rice, maize, and soybeans. In Jixian County, maize is riskier than both soybeans and rice. For Farm 291, maize is riskier than rice, which, in turn, is riskier than soybeans. Table A2.5: Jixian County and Farm 291 Average Annual Loss Rice Maize Soybean Jixian County (township level) 14% 21% 16% Farm 291 (team level) 25% 30% 14% Source: Authors’ calculation. Figure A2.3: Heilongjiang—Jixan County Yield-Loss Exposure Heilongjiang: Jixan County Yield Loss Exposure (Township-Level Data) 100% 90% 80% 70% Probability 60% 50% 40% 30% 20% 10% 0% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Loss Relative to Expected Value Rice Maize Soybean Source: Authors’ calculation. 34 Figure A2.4: Heilongjiang—Farm 291 Yield-Loss Exposure Heilongjiang: Farm 291 Yield Loss Exposure (Team-Level Data) 100% 90% 80% 70% Probability 60% 50% 40% 30% 20% 10% 0% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Loss Relative to Expected Value Rice Maize Soybean Source: Authors’ calculation. South Shuangcheng City and Hongguang Farm (HRG) are the pilot county and farm located in the southern part of Heilongjiang province. Maize is the primary crop produced in Shuangcheng City, although small amounts of rice and soybeans are produced as well. In contrast to other producers in the southern part of the province, Hongguang Farm produces primarily soybeans, although maize is produced as well. Table A2.6 presents measures of average annual loss. Figure A2.5 presents Shuangcheng City yield-loss exceedance curves for rice, maize, and soybeans. Figure A2.6 presents Hongguang Farm yield-loss exceedance curves for maize and soybeans. For Shuangcheng City, both rice and soybeans are riskier than maize. The results for Hongguang Farm are interesting in that the average annual loss for soybeans is higher than for maize, but, figure A2.6 clearly shows that the probability of large losses is much higher with maize than with soybeans. Table A2.6: Shuangcheng City and Hongguang Farm Average Annual Loss Rice Maize Soybean Shuangcheng City (township level) 12% 8% 13% Hongguang Farm (team level) NA 13% 15% Source: Authors’ calculation. 35 Figure A2.5: Heilongjiang—Shuangcheng City Yield-Loss Exposure Heilongjiang: Shuangcheng City Yield Loss Exposure (Township-Level Data) 100% 90% 80% 70% Probability 60% 50% 40% 30% 20% 10% 0% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Loss Relative to Expected Value Rice Maize Soybean Source: Authors’ calculation. Figure A2.6: Heilongjiang—Hongguang Farm Yield-Loss Exposure Heilongjiang: Hongguang Farm Yield Loss Exposure (Team-Level Data) 100% 90% 80% 70% Probability 60% 50% 40% 30% 20% 10% 0% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Loss Relative to Expected Value Maize Soybean Source: Authors’ calculation. 36 Loss Ratio Risk Exposure When establishing premium rates, it is helpful for an agricultural insurer to have estimates of the average annual-loss and yield-loss exceedance curves. But these measures are not adequate for determining the amount of contingent capital that might be required to pay indemnities in the event of extreme losses. To know how much contingent capital (reinsurance, reserves, and so on) might be required, the insurer needs some estimate of how large indemnities might be relative to premiums collected. Next, consideration is given to the potential loss ratios for a hypothetical insurance company that sells multiple-peril crop insurance in Heilongjiang. An insurance loss ratio is calculated as indemnities paid, divided by premiums collected. The last section of the annex contains details on how the loss ratios presented here were calculated. The loss ratio exposure for a hypothetical insurance company is presented as the probable maximum loss (PML) for various return periods. Loss ratio PMLs are valuable for insurance companies to demonstrate the insurer’s exposure to extreme loss events. To maintain financial solvency over the long term, insurers should have sufficient access to contingent capital to pay indemnities from at least a 100-year return loss event. North Loss ratio PMLs for Nenjiang County and Qixingpao Farm are presented in tables A2.7 and A2.8. Results are presented for soybeans, wheat, and a portfolio that contains the expected value of soybean and wheat production in the county or farm. In principle, the portfolio will diversify the risk so that the loss ratio PMLs for the portfolio will be lower than those for the dominant crop in the portfolio. However, as described in the last section of this annex, the risk reduction obtained by building a portfolio of insurance policies for multiple crops will depend upon: (1) the extent to which the portfolio is dominated by a single crop (or region), and (2) the correlation between yield risk exposures for the various crops (or regions). In Nenjiang County, the correlation between soybean and wheat yields is only 0.28. However, soybeans account for more than 90 percent of the portfolio, so the portfolio does not greatly reduce loss ratio PMLs relative to just insuring soybeans alone. For Qixingpao Farm, soybeans account for 85 percent of the portfolio, but the correlation between wheat and soybean yields is -0.10. For this reason, the portfolio has noticeably lower loss ratio PMLs relative to just insuring soybeans alone. Table A2.7: Nenjiang County Loss Ratio Exposure Wheat Soybean Portfolio Probable Maximum Loss (10-year return) 3.7 3.4 3.1 Probable Maximum Loss (20-year return) 4.2 3.9 3.8 Probable Maximum Loss (100-year return) 5.2 5.0 4.9 Source: Authors’ calculation. 37 Table A2.8: Qixingpao Farm Loss Ratio Exposure Wheat Soybean Portfolio Probable Maximum Loss (10-year return) 3.2 2.9 2.5 Probable Maximum Loss (20-year return) 4.4 3.7 3.1 Probable Maximum Loss (100-year return) 5.7 4.7 3.9 Source: Authors’ calculation. Table A2.9 presents loss ratio PMLs for a combination of Nenjiang County and Qixingpao Farm. Nenjiang County accounts for 90 percent of the combined expected value of soybean production and 84 percent of the combined expected value of wheat production. However, the loss ratio PMLs for the combination of Nenjiang County and Qixingpao Farm are noticeably lower than those for Nenjiang County alone, particularly for soybeans. This is because the correlation between Nenjiang County and Qixingpao Farm soybean yield losses is only 0.19. The correlation between Nenjiang County and Qixingpao Farm wheat yield losses is 0.66. Table A2.9: Nenjiang County and Qixingpao Farm Combined Loss Ratio Exposure Wheat Soybean Portfolio Probable Maximum Loss (10-year return) 3.4 3.0 2.9 Probable Maximum Loss (20-year return) 4.0 3.5 3.4 Probable Maximum Loss (100-year return) 5.0 4.4 4.4 Source: Authors’ calculation. These findings suggest that to cover a 1-in-100-year event, an insurance company insuring soybeans in either Nenjiang County or Qixingpao Farm (but not both) would need sufficient access to contingent capital to pay indemnities of up to 5 times premiums collected. If the insurance company were insuring both Nenjiang County and Qixingpao Farm, the amount of contingent capital needed would drop to about 4.5 times premiums collected. Central Loss ratio PMLs for Jixian County and Farm 291 are presented in tables A2.10 and A2.11. Results are presented for rice, maize, soybeans, and a portfolio that contains the expected value of each crop in the county or farm. In Jixian County the correlation between rice and maize yield losses is 0.75, the correlation between maize and soybean yield losses is 0.72, and the correlation between rice and soybean yield losses is 0.48. Soybeans account for 39 percent of the portfolio, and maize and rice account for 36 percent and 25 percent, respectively. For Farm 291, rice accounts for 83 percent of the portfolio, and maize and soybeans account for 10 percent and 7 percent, respectively. The correlation between rice and maize yield losses is 0.89, the correlation between maize and soybean yield losses is 0.33, and the correlation between rice and soybean yield 38 losses is 0.28. Since rice and maize together account for 93 percent of the portfolio and yield losses for the two crops are so highly correlated, the portfolio loss ratio PMLs for Farm 291 are similar to those for rice and maize separately. Table A2.10: Jixian County Loss Ratio Exposure Rice Maize Soybean Portfolio Probable Maximum Loss (10-year return) 3.7 3.2 3.3 3.0 Probable Maximum Loss (20-year return) 4.1 3.7 3.8 3.4 Probable Maximum Loss (100-year return) 5.0 4.5 4.6 4.2 Source: Authors’ calculation. Table A2.11: Farm 291 Loss Ratio Exposure Rice Maize Soybean Portfolio Probable Maximum Loss (10-year return) 2.6 2.5 2.7 2.5 Probable Maximum Loss (20-year return) 2.9 2.8 3.1 2.8 Probable Maximum Loss (100-year return) 3.5 3.4 3.9 3.4 Source: Authors’ calculation. Table A2.12 presents loss ratio PMLs for a combination of Jixian County and Farm 291. Jixian County accounts for 85 percent of the combined expected value of soybean production, but the correlation between Jixian County and Farm 291 soybean yield losses is 0.49; so the combined soybean loss ratio PMLs are somewhat less than those for Jixian County alone. Jixian County accounts for 76 percent of the combined expected value of maize production, and the correlation between Jixian County and Farm 291 maize yield losses is 0.87. Even so, the combined maize loss ratio PMLs are less than those for Jixian county alone. Farm 291 accounts for 78 percent of the combined expected value of rice production. The correlation between Jixian County and Farm 291 rice yield losses is 0.63; so the combined rice loss ratio PMLs are identical to those for Farm 291 alone. Table A2.12: Jixian County and Farm 291 Combined Loss Ratio Exposure Rice Maize Soybean Portfolio Probable Maximum Loss (10-year return) 2.6 3.0 3.0 2.6 Probable Maximum Loss (20-year return) 2.9 3.3 3.4 3.0 Probable Maximum Loss (100-year return) 3.5 4.1 4.1 3.6 Source: Authors’ calculation. Note that according to the risk assessment, average annual losses are much higher for the central region than for the northern region. However, the loss ratio PMLs are generally higher for the northern region than for the central region. Both regions can experience extreme losses, but the loss ratio PML measures the magnitude of extreme losses relative 39 to an actuarially fair premium rate. The premium rate will be lower (higher) when average annual losses are lower (higher). Since the northern region has much lower average annual losses, it also has lower actuarially fair premium rates, so extreme events generate larger loss ratios. This demonstrates the point made earlier that when considering the amount of contingent capital required, the insurer cannot simply use measures of annual average loss. Instead, the insurer must consider the magnitude of extreme events relative to average annual loss (for example, loss ratio PMLs). South Loss ratio PMLs for Shuangcheng City and Hongguang Farm are presented in tables A2.13 and A2.14. For Shuangcheng City, results are presented for rice, maize, soybeans, and a portfolio that contains the expected value of each crop in the city. For Hongguang Farm, results are presented for maize, soybeans, and a portfolio of maize and soybeans. In Shuangcheng City the correlation between soybean and maize yield losses is 0.71, the correlation between rice and maize yield losses is 0.68, and the correlation between rice and soybean yield losses is 0.34. Maize accounts for 83 percent of the portfolio, and rice and soybeans account for 12 percent and 5 percent, respectively. The loss ratio exposure for the portfolio is very similar to that for maize, because maize dominates the portfolio. Further, the correlations between maize and the other two crops are very high, thus very little diversification is accomplished by including the other crops into a portfolio. For Hongguang Farm, soybeans account for 73 percent of the portfolio and maize accounts for 27 percent. Yet even though soybeans dominate the portfolio, and even though the maize loss ratio PMLs are higher than those for soybeans, the portfolio of soybeans and maize actually has lower loss ratio PMLs than those for soybeans alone. This occurs because the correlation between soybean and maize yield losses is unusually low at -0.17. Table A2.13: Shuangcheng City Loss Ratio Exposure Rice Maize Soybean Portfolio Probable Maximum Loss (10-year return) 3.5 3.0 3.1 2.7 Probable Maximum Loss (20-year return) 6.0 4.5 3.6 4.6 Probable Maximum Loss (100-year return) 7.6 5.8 4.3 5.9 Source: Authors’ calculation. Table A2.14: Hongguang Farm Loss Ratio Exposure Maize Soybean Portfolio Probable Maximum Loss (10-year return) 3.4 1.8 1.6 Probable Maximum Loss (20-year return) 3.9 2.0 1.8 Probable Maximum Loss (100-year return) 4.8 2.4 2.1 Source: Authors’ calculation. Table A2.15 presents loss ratio PMLs for a combination of Shuangcheng City and Hongguang Farm. Shuangcheng City accounts for 100 percent of the combined expected 40 value of rice production and 99 percent of the combined expected value of maize production. Thus, the combined rice and maize loss ratio PMLs are the same as for Shuangcheng City alone. Shuangcheng City accounts for 69 percent of the combined expected value of soybean production. The correlation between Shuangcheng City and Hongguang Farm soybean yield losses is only 0.14; so the combined soybean loss ratio PMLs are significantly less than for Shuangcheng City alone. Table A2.15: Shuangcheng City and Hongguang Farm Combined Loss Ratio Exposure Rice Maize Soybean Portfolio Probable Maximum Loss (10-year return) 3.5 3.0 2.5 2.7 Probable Maximum Loss (20-year return) 6.0 4.5 2.9 4.5 Probable Maximum Loss (100-year return) 7.6 5.8 3.5 5.7 Source: Authors’ calculation. It should be noted that findings regarding relatively low soybean yield loss exposure and low loss ratio PMLs are contrary to the information made available for this review. This is largely because of the results for Hongguang Farm. The low soybean yield loss exposure for Hongguang Farm—combined with the low correlations between soybean yield losses on Hongguang Farm and maize yield losses on Hongguang Farm and soybean yield losses for Hongguang Farm and Shuangcheng City—suggests that soybean production on Hongguang Farm may be irrigated. Regardless, the soybean results are likely not representative of other areas in the south of Heilongjiang Province. Province level Table A2.16 presents loss ratio PMLs for a portfolio consisting of the six pilot counties and farms. In the six pilot counties/farms (and considering only these three crops) soybeans account for 45 percent of the expected value of production, maize is 39 percent, and rice is 16 percent. In total for the province (and again considering only these three crops), rice is approximately 45 percent of the expected value of production, soybeans are 37 percent, and maize is 18 percent. Thus, relative to the province as a whole, the six pilot counties and farms have more soybeans and maize but less rice. Even so, table A2.16 clearly demonstrates how diversifying across crops reduces loss ratio PMLs. Table A2.16: Heilongjiang Combined Loss Ratio Exposure Soybean Maize Rice Portfolio Probable Maximum Loss (10-year return) 2.7 2.6 2.4 2.2 Probable Maximum Loss (20-year return) 3.1 3.7 2.8 2.6 Probable Maximum Loss (100-year return) 3.7 4.8 3.5 3.4 Source: Authors’ calculation. 41 Weather Trend Analysis and Extreme Events Monthly averages of daily rainfall and daily average temperature were provided for a number of weather stations in Heilongjiang. Three weather stations were selected for analysis because of their proximity to the three pilot regions. Specifically, for Nenjiang County and Qixingpao Farm (HRG), the pilot county and farm located in the northern part of Heilongjiang province, data for the Nenjiang weather station were provided for the period 1971–2000. For Jixian County and Farm 291 (HRG), the pilot county and farm located in the central part of Heilongjiang province, data for the Jiamsui weather station were provided for the period 1971–2000. For Shuangcheng City and Hongguang Farm (HRG), the pilot county and farm located in the southern part of Heilongjiang province, data for the Shuangcheng weather station were provided for the period 1985–2004. To consider possible relationships with maize and soybean yields, three weather variables were selected for analysis: cumulative rainfall from June through August, July daily average temperature, and August daily average temperature. It was not possible to document strong relationships between these weather variables and maize and soybean yield losses. However, this should not be interpreted as a definitive finding, because the analysis was very limited. Only monthly weather data were available, whereas more detailed analyses of the relationship between weather variables and yield losses would be possible with daily weather data. Only a limited number of weather variables were considered. The analysis was restricted to the limited number of years for which data were available for both yields and weather—in some cases, as few as 16 years. Most researchers would agree that this time series is not long enough to draw definitive conclusions about the relationship between weather variables and yield losses. Finally, the analysis focused only on the six pilot counties and farms and associated weather stations. A more thorough analysis would look at many more geographic locations. 2.3. Xinjiang Agricultural land in Xinjiang is either owned collectively by farmers or owned by the state. Land that is owned collectively is administered by farmer organizations such as village committees. Most of the state farms in Xinjiang are under the administration of the Xinjiang Production and Construction Corps (XPCC). The 175 state farms under the XPCC are administered through a two-tiered system that includes the farm and individual production teams. Xinjiang has an arid climate. Some areas in the northern part of the province receive up to 400 millimeters of rainfall per year, but most regions receive only 100–200 millimeters. The southern part of the province receives less than 100 millimeters of rainfall per year. Most rainfall occurs in the spring. Much of the crop production is irrigated, snowmelt from the mountains being the primary source of irrigation water. The length of the growing season depends on latitude and altitude but is generally limited to 120–180 frost-free days. Upland cotton is produced in the northern part of the province, and extra long staple cotton is produced in the south. A total of approximately 1.1 million hectares of cotton are produced in the province, up from only 161,000 hectares in 1979. In addition, the 42 province produces about 700,000 hectares of wheat (down from 1.3 million hectares in 1979) and 500,000 hectares of maize (down from 600,000 hectares in 1979). Oil-bearing crops, such as rapeseed and peanuts, are also produced in Xinjiang, along with rice, soybeans, sugar beets, tobacco, and various fruits, vegetables, and tree nuts. The current provincial-level expected yield for cotton is approximately 1.6 tons per hectare. Expected cotton yields have consistently trended upward since 1979, when the expected yield was less than 0.5 tons per hectare. The current provincial-level expected yield for wheat is a little more than 5 tons per hectare. Expected wheat yields have consistently trended upward since 1979, when the expected yield was less than 2 tons per hectare. The current provincial-level expected yield for maize is approximately 7 tons per hectare. Expected maize yields have also consistently trended upward since 1979, when the expected yield was a little more than 2 tons per hectare. Planting dates vary with latitude and weather conditions, but, in general, cotton is planted in mid- to late March in the south and late March to early April in the north. Cotton is harvested in September and October. Wheat is planted in late August to early September and harvested in mid-April to mid-May. Maize is planted in late April to mid-May and harvested in late July to mid-August. Some maize is double-cropped with wheat. Meteorological Stations The Xinjiang Weather Administration operates 106 meteorological stations of which 19 participate in international exchange. In addition to stations operated by the Weather Administration, there are also weather stations operated by other agencies. For instance, XPCC has some 40 stations. In addition, because of the importance of snow to the irrigation network, the provincial hydrology bureau operates its own weather stations, which focus on snow monitoring, as well as 132 river stations measuring water volume and quality. Remote sensing is also used to monitor snowfall. Data for Pilot Counties and Farms Five pilot counties and farms, located in three distinct regions, were selected for analysis. Manasi County and Farm 8 (XPCC) are located along the Manasi watershed in northern Xinjiang. Tuokexun County and Farm 4 (XPCC) are also located in northern Xinjiang. They are east of Manasi County and Farm 8 (XPCC) on the Turpan oasis basin that is surrounded by mountains. Pishan County is in extreme southern Xinjiang. For each of the five pilot counties or farms annual data were provided on hectares planted and tons produced for major crops. For the three pilot counties, the data were provided at the township level. For the two XPCC farms, the data were provided at the team level. For Manasi County, township-level data were available for the period 1986–2004 (excepting 1991) for wheat and maize, and for the period 1986–2005 for cotton. For Tuokexun County, township-level data were available for the period 1986–2004 for “grainâ€? (an aggregate category that can include both food and feed grains) and cotton. For Pishan County, township-level data were available for the period 1988–2005 for wheat and cotton. For both XPCC farms, team-level data were available for the period 43 1986–2005 for wheat, maize, and cotton. In addition, provincial-level data on planted hectares and tons produced were available for major crops for the period 1979–2004. Monthly rainfall and temperature data were available for the weather stations nearest each pilot county or farm. If a specific weather peril was a significant cause of loss, and if it seemed possible that the available historical weather data could possibly measure the magnitude of that weather peril, then the weather data were examined to investigate potential relationships between yield shortfalls and extreme weather events. Data were also provided on hectares covered by, and affected by, natural disasters. (As mentioned earlier, hectares covered by natural disaster are those that experience a yield loss of more than 10 Percent, and hectares affected by natural disaster are those that experience a yield loss of more than 30 percent.) These data were available at the provincial level and in some cases at the township and team levels. However, frequently the township- or team-level data on hectares covered by, and affected by, natural disasters were unusable due to inconsistencies in the data provided and large numbers of missing observations. Where usable data were available, these data allowed for consistency checks against the yield data. However, for risk assessment their value is limited, because they are not crop-specific nor do they indicate the exact amount of yield loss. Primary Perils One of the primary crop perils in Xinjiang is dry, hot winds that desiccate crops. These winds sometimes carry sand that also damages crops. Other perils include inadequate irrigation water, cold summer temperatures, hail, disease, early frost in autumn, late frost in spring, and occasional flooding. Statistical analysis of the data on hectares covered by, and affected by, natural disasters indicated that nearly half of the yield loss is due to drought and about one quarter is from hail. Flooding and frost or freeze account for about 12 percent each of the reported causes of loss. However, it is important to note that wind damage is not included in the cause-of-loss data. In the southern part of Xinjiang, the major peril for cotton is sandstorms during April and May. Sandstorms also occur in the north, though much less often than in the south. Sandstorms destroy cotton seedlings, so farmers incur replanting costs. The crop insurance companies operating in Xinjiang report that approximately 50 percent of their cotton losses are due to wind and sandstorms. Unusually cold temperatures during the growing season or early frost in autumn can also cause yield losses in cotton. These perils are related to the risk of sandstorm, because replanting increases the probability of yield losses due to cold temperatures or damage due to early frost. Cotton is also affected by hail and increasingly, due to monoculture cotton production, pests and disease. With cotton it is also important to remember that yield variability is not the only type of production risk exposure. Cold temperatures during the growing season or early autumn frosts can cause significant quality losses. The primary perils for wheat are unusually high temperatures during the milking stage and insufficient snowfall during the winter to insulate the crop from extremely cold temperatures. Grapes and other fruits are affected by late spring or early autumn frosts. 44 Most crops are affected by the risk of insufficient irrigation water. The amount of available irrigation water depends, in part, on the amount of snow that falls in the mountains during the proceeding winter. But more important than annual winter snow fall is the amount of snow pack, which results in important lag effects, because the amount of snow fall in a given year may affect the amount of available irrigation water for several years into the future. Temperature also has a major effect on the amount of available irrigation water. If spring temperatures in the mountains are unusually cold, there will be less snowmelt to feed the streams. If spring temperatures are unusually warm, flooding can occur, especially in the north. During the field visit it was reported that significant flooding occurs in the north in one out of every ten years. Table A2.17 presents the percentage of planted hectares (for all crops) with yield losses greater than 10 percent due to either drought or flood (including waterlogging). These percentages were calculated using the provincial-level data on hectares covered by, and hectares affected by, various natural disasters. The table also shows the provincial-level yield loss for cotton and maize in the same years. The percentage of planted hectares with yield losses caused by either drought or flood is quite low relative to other provinces in China. The 1989 drought caused cotton yield losses and, to a much lesser extent, maize yield losses. The 2001 drought seems to have affected only cotton, but the 1995 drought affected only maize. Wheat is not included in table A2.17 because significant provincial-level yield losses have not occurred in any of the years shown. Table A2.17: Xinjiang Percentage of Planted Hectares; Yield Losses Caused by Drought and Flood; Cotton and Maize Yield Losses Percentage of Planted Hectares Provincial-Level Provincial-Level with Yield Losses of at least 10% Cotton Yield Loss Maize Yield Loss Caused by Relative to Expected Relative to Expected Year Drought Flooding Value Value 2003 11% 2% -1% -2% 2002 7% 7% No Loss -3% 2001 12% 1% -11% No Loss 2000 11% 1% No Loss No Loss 1995 12% 2% No Loss -6% 1994 16% 3% No Loss No Loss 1991 25% 0% No Loss No Loss 1989 18% 0% -13% -2% 1986 14% 0% No Loss No Loss 1978 28% 0% NA NA Source: Authors’ calculation. 45 At a provincial level, for the period 1979–2003 the largest cotton yield losses occurred in 1988, 1989, and 2001. As indicated above, 1989 and 2001 were drought years (cotton was also affected by an April windstorm in 2001), but the cause of loss in 1988 was evidently something other than drought or flood. For these years, table A2.18 shows the provincial-level cotton yield loss along with township- and team-level yield losses for the five pilot counties and farms. The 1989 drought affected cotton yields in all five pilot counties and farms, but Farm 8 (XPCC), Manasi County, and Farm 4 (XPCC) suffered the largest losses. The 2001 drought and the 1988 loss event primarily affected Farm 8 (XPCC) and Manasi County. Table A2.18. Cotton Yield Losses in Selected Years Tuokexun Pishan Farm 8 Manasi County Farm 4 County County (XPCC) Weighted (XPCC) Weighted Weighted Weighted Average Weighted Average Average Average Township- Average Team- Township- Township- Team-Level Level Yield Level Yield Level Yield Level Yield Year Yield Loss Loss Loss Loss Loss 2001 -31% -29% No Loss -10% -5% 1989 -23% -51% -27% -4% -7% 1988 -27% -26% -7% No Loss -1% Source: Authors’ calculation. Risk Assessment of Pilot Counties and Farms Risk assessment was conducted for each of the five pilot counties or farms. Yield-loss exposure is presented using tables and graphs. The tables present the average annual loss. The graphs contain loss exceedance curves that indicate the probability of losses in excess of some magnitude. The procedures used to generate these measures are described in the last section of the annex. Manasi Watershed Manasi County and Farm 8 (XPCC) are the pilot county and farms located along the Manasi watershed in the northern part of Xinjiang province. For both Manasi County and Farm 8, cotton is the dominant crop, although both also produce wheat and maize. Table A2.19 presents measures of average annual loss. Figures A2.7 and A2.8 present Manasi County and Farm 8 yield-loss exceedance curves for wheat, cotton, and maize. The results clearly show that, in this area, cotton is riskier than wheat or maize. 46 Table A2.19: Manasi County and Farm 8 Average Annual Loss Wheat Cotton Maize Manasi County (township level) 4% 14% 10% Farm 8 (team level) 7% 12% 6% Source: Authors’ calculation. Figure A2.7: Xinjiang—Manasi County Yield-Loss Exposure Xinjiang: Manasi County Yield Loss Exposure (Township-Level Data) 100% 90% 80% 70% Probability 60% 50% 40% 30% 20% 10% 0% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Loss Relative to Expected Value Wheat Cotton Maize Source: Authors’ calculation. Figure A2.8: Xinjiang—XPCC No. 8 Yield-Loss Exposure Xinjiang: XPCC No. 8 Yield Loss Exposure (Team-Level Data) 100% 90% 80% 70% Probability 60% 50% 40% 30% 20% 10% 0% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Loss Relative to Expected Value Wheat Cotton Maize Source: Authors’ calculation. 47 Turpan Basin Tuokexun County and Farm 4 (XPCC) are the pilot county and farm located along the Turpan Basin in the northeastern part of Xinjiang province. For both Tuokexun County and Farm 4, grain (primarily wheat and maize) is the dominant crop, although both also produce cotton. Table A2.20 presents measures of average annual loss. Figures A2.9 and A2.10 present Tuokexun County and Farm 4 yield-loss exceedance curves. The data for Tuokexun County do not separate wheat and maize. Instead they are combined into an aggregate category called “grain.â€? Even though the cotton average annual loss is higher than that for grain, figure A2.9 shows that for very low probability loss events the magnitude of losses for grain is higher than for cotton. Similarly, for Farm 4 the average annual loss for cotton is higher than for maize, but for very low probability loss events the magnitude of maize losses is higher than for cotton. Table A2.20: Tuokexun County and Farm 4 Average Annual Loss Grain Wheat Cotton Maize Tuokexun County (township level) 5% NA 7% NA Farm 4 (team level) NA 10% 14% 10% Source: Authors’ calculation. Figure A2.9: Xinjiang: Tuokexun County Yield-Loss Exposure Xinjiang: Tuokexun County Yield Loss Exposure (Township-Level Data) 100% 90% 80% 70% Probability 60% 50% 40% 30% 20% 10% 0% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Loss Relative to Expected Value Grain Cotton Source: Authors’ calculation. 48 Figure A2.10: Xinjiang—XPCC No. 4 Yield-Loss Exposure Xinjiang: XPCC No. 4 Yield Loss Exposure (Team-Level Data) 100% 90% 80% 70% Probability 60% 50% 40% 30% 20% 10% 0% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Loss Relative to Expected Value Wheat Cotton Maize Source: Authors’ calculation. South Pishan County is the pilot county from the southern part of Xinjiang. Wheat is the dominant crop in Pishan County, but cotton is produced as well. Table A2.21 presents measures of average annual loss. Figure A2.11 shows yield-loss exceedance curves. The results for Pishan County are similar to those for the other regions of Xinjiang in that cotton is riskier than wheat. Table A2.21: Pishan County Average Annual Loss Wheat Cotton Pishan County (township level) 10% 12% Source: Authors’ calculation. 49 Figure A2.11: Xinjiang—Pishan County Yield-Loss Exposure Xinjiang: Pishan County Yield Loss Exposure (Township-Level Data) 100% 90% 80% 70% Probability 60% 50% 40% 30% 20% 10% 0% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Loss Relative to Expected Value Wheat Cotton Source: Authors’ calculation. Loss Ratio Risk Exposure When establishing premium rates, it is helpful for an agricultural insurer to have estimates of the average annual loss and yield-loss exceedance curves, but these measures are not adequate for determining the amount of contingent capital that might be required to pay indemnities in the event of extreme losses. To know how much contingent capital (reinsurance, reserves, and so forth) might be required, the insurer needs some estimate of how large indemnities might be, relative to premiums collected. Next, consideration is give to the potential loss ratios for a hypothetical insurance company that sells multiple-peril crop insurance in Xinjiang. An insurance loss ratio is calculated as indemnities paid divided by premiums collected. The last section of the annex contains details on how the loss ratios presented here were calculated. The loss ratio exposure for a hypothetical insurance company is presented as the probable maximum loss for various return periods. Loss ratio PMLs are valuable for insurance companies, because they demonstrate the insurer’s exposure to extreme loss events. To maintain financial solvency over the long term, insurers should have sufficient access to contingent capital to pay indemnities from at least a 100-year return loss event. Manasi Watershed Loss ratio PMLs for Manasi County and Farm 8 are presented in tables A2.22 and A2.23. Results are presented for wheat, cotton, maize, and a portfolio that contains the expected value of production for each of the three crops in the county or farm. In Manasi County, cotton accounts for 82 percent of the portfolio. For Farm 8, cotton accounts for 90 50 percent of the portfolio. Yet, the portfolio has noticeably lower PMLs relative to insuring cotton alone, because the correlations between yield losses for the three crops are very low. In Manasi County, the correlation between wheat and cotton yield losses is 0.24. For wheat and maize the correlation is 0.05, and for maize and cotton the correlation is -0.25. For Farm 8, the correlation between wheat and maize yield losses is 0.22. For maize and cotton the correlation is -0.03, and for wheat and cotton the correlation is -0.14. Table A2.22: Manasi County Loss Ratio Exposure Wheat Cotton Maize Portfolio Probable Maximum Loss (10-year return) 1.8 3.1 3.1 2.6 Probable Maximum Loss (20-year return) 2.1 4.3 3.8 3.6 Probable Maximum Loss (100-year return) 2.6 5.7 5.0 4.7 Source: Authors’ calculation. Table A2.23: Farm 8 Loss Ratio Exposure Wheat Cotton Maize Portfolio Probable Maximum Loss (10-year return) 2.7 3.3 2.3 3.1 Probable Maximum Loss (20-year return) 3.4 3.7 2.5 3.5 Probable Maximum Loss (100-year return) 4.4 4.5 3.0 4.3 Source: Authors’ calculation. Table A2.24 presents loss ratio PMLs for a combination of Farm 8 and Manasi County. Farm 8 accounts for 80 percent of the combined expected value of cotton production, 77 percent of the combined expected value of wheat production, and 58 percent of the combined expected value of maize production. Yet, the loss ratio PMLs for the combination of Farm 8 and Manasi County are noticeably lower than those for Farm 8 alone. The correlation between Manasi County and Farm 8 yield losses is 0.69, 0.46, and 0.03 for cotton, maize, and wheat, respectively. Table A2.24: Manasi County and Farm 8 Combined Loss Ratio Exposure Wheat Cotton Maize Portfolio Probable Maximum Loss (10-year return) 2.3 3.0 2.2 2.8 Probable Maximum Loss (20-year return) 3.2 3.4 3.0 3.2 Probable Maximum Loss (100-year return) 4.2 4.0 3.9 3.8 Source: Authors’ calculation. These findings suggest that to cover a 1-in-100-year event, an insurance company that was insuring cotton for Manasi County or Farm 8 (but not both) would need sufficient access to contingent capital to pay indemnities of up to 5 or 6 times premiums collected. 51 If the insurance company were insuring both Manasi County and Farm 8, the amount of contingent capital needed would drop to 4 times premiums collected. Turpan Basin Loss ratio PMLs for Tuokexun County and Farm 4 are presented in tables A2.25 and A2.26. For Tuokexun County, results are presented for grain and cotton. Since grain is an aggregate category that contains both wheat and maize, it was not possible to construct a portfolio for Tuokexun County. For Farm 4, results are presented for wheat, cotton, maize, and a portfolio that contains the expected value of production for each of the three crops. Farm 4 loss ratio PMLs are very similar for wheat and cotton. Maize has much larger loss ratio PMLs for 1-in-20-year and 1-9n-100-year events. However, the portfolio of all three crops has much lower loss ratio PMLs than any of the crops individually. There are two reasons for this. First, the portfolio value is spread across all three crops. For Farm 4, cotton accounts for 40 percent of the portfolio value. Maize and wheat account for 36 percent and 24 percent of the portfolio value, respectively. Second, the yield-loss correlations for the three crops are quite low. The correlation between maize and cotton yield losses is 0.33. For maize and wheat the correlation is 0.08 and for wheat and cotton the correlation is -0.24. Table A2.25: Tuokexun County Loss Ratio Exposure Grain Cotton Probable Maximum Loss (10-year return) 3.4 2.6 Probable Maximum Loss (20-year return) 4.1 3.2 Probable Maximum Loss (100-year return) 5.1 4.1 Source: Authors’ calculation. Table A2.26: Farm 4 Loss Ratio Exposure Wheat Cotton Maize Portfolio Probable Maximum Loss (10-year return) 3.6 3.6 3.6 2.5 Probable Maximum Loss (20-year return) 4.4 4.2 7.0 3.6 Probable Maximum Loss (100-year return) 5.6 5.3 8.9 4.7 Source: Authors’ calculation. Table A2.27 presents cotton loss ratio PMLs for a combination of Tuokexun County and Farm 4. The combined cotton loss ratio PMLs are much lower than those for either Tuokexun County or Farm 4 alone. Tuokexun County accounts for 61 percent of the combined expected value of cotton production. More importantly, the correlation between Tuokexun County and Farm 4 cotton yield losses is very low at -0.24. 52 Table A2.27: Tuokexun County and Farm 4 Combined Loss Ratio Exposure Cotton Probable Maximum Loss (10-year return) 2.3 Probable Maximum Loss (20-year return) 2.5 Probable Maximum Loss (100-year return) 3.1 Source: Authors’ calculation. These findings suggest that to cover a 1-in-100-year event, an insurance company that was insuring cotton for Tuokexun County or Farm 4 (but not both) would need sufficient access to contingent capital to pay indemnities of up to 5.5 times premiums collected. If the insurance company were insuring both Tuokexun County and Farm 4, the amount of contingent capital needed would drop to a little more than 3 times premiums collected. South Loss ratio PMLs for Pishan County are presented in table A2.28, which shows results for wheat, cotton, and a portfolio that contains the expected value of production for each of the two crops. Wheat accounts for 67 percent of the portfolio value. The portfolio PMLs are significantly lower than those for wheat alone, because the correlation between wheat and cotton yield losses is only 0.38. Table A2.28: Pishan County Loss Ratio Exposure Wheat Cotton Portfolio Probable Maximum Loss (10-year return) 3.7 3.0 2.9 Probable Maximum Loss (20-year return) 4.1 3.3 3.4 Probable Maximum Loss (100-year return) 5.0 4.1 4.2 Source: Authors’ calculation. Province level Table A2.29 presents loss ratio PMLs for portfolios based on the five pilot counties and farms. The cotton PMLs are based on all five pilot counties and farms. The wheat PMLs are based on Manasi and Pishan Counties and XPCC Farms 4 and 8. The maize PMLs are based on Manasi County and XPCC Farms 4 and 8. In the five pilot counties and farms (and considering only these three crops) cotton accounts for 84 percent of the expected value of production in the portfolio, wheat is 9 percent, and maize is 7 percent. In total for the province (and again considering only these three crops), cotton is approximately 64 percent of the expected value of production, wheat is 23 percent, and maize is 13 percent. Thus, relative to the province as a whole, the five pilot counties and farms have more cotton but less wheat and maize. Even so, the insurance portfolio that includes some wheat and maize reduces the loss ratio PMLs relative to insuring cotton alone. 53 Table A2.29: Xinjiang Combined Loss Ratio Exposure Wheat Cotton Maize Portfolio Probable Maximum Loss (10-year return) 2.0 3.0 2.2 2.7 Probable Maximum Loss (20-year return) 2.3 3.3 2.8 3.0 Probable Maximum Loss (100-year return) 2.7 4.0 3.7 3.6 Source: Authors’ calculation. Weather Trend Analysis and Extreme Events Monthly averages of daily rainfall and daily average temperature were provided for a number of weather stations in Xinjiang. Three weather stations were selected for analysis because of their proximity to the three pilot regions. Specifically, for Manasi County and Farm 8, the pilot county and farm located in the Manasi watershed, data were provided for the Urumqi weather station. For Tuokexun County and Farm 4, the pilot county and farm located in the Turpan Basin, data were provided for the Tulufan weather station. For Pishan County, data were provided for the Hetian weather station. All weather data were for the years 1986–2005. To consider possible relationships with cotton yields, three weather variables were selected for analysis: cumulative rainfall from June through August, July daily average temperature, and August daily average temperature. It was not possible to document strong relationships between these weather variables and maize and soybean yield losses. However, this should not be interpreted as a definitive finding, because the analysis was very limited. For example, only monthly weather data were available, whereas more detailed analyses of the relationship between weather variables and yield losses would be possible with daily weather data. Only a limited number of weather variables were considered. The analysis was restricted to the limited number of years for which data were available for both yields and weather–—in some cases, as few as 19 years. Most researchers would agree that this time series is not long enough to draw definitive conclusions about the relationship between weather variables and yield losses. Finally, the analysis focused only on the five pilot counties and farms and associated weather stations. A more thorough analysis would look at many more geographic locations. 2.4. Shanghai Shanghai Municipality is located at the mouth of the Yangtze River in southeastern China. It contains three counties and 17 urban districts. Agriculture accounts for only about 2 percent of the municipality’s GDP, but the municipal authorities encourage agricultural production in an effort to increase self-sufficiency. Shanghai lies in a coastal zone with temperate climate conditions. Freezing temperatures rarely occur, and annual average rainfall is approximately 1,100 millimeters. The topography is flat, and production is on low lying areas of land. There is a heavy reliance 54 on drainage of these areas, with a high investment in drainage canals and flood-control infrastructure. Crops produced in Shanghai include rice, wheat, rapeseed, melons, and a variety of vegetables and tree fruit. Vegetables are produced both in fields and in greenhouses. Shanghai produces approximately 140,000 hectares of vegetables and 30,000 hectares of fruit orchards (primarily tangerines, peaches, pears, grapes, and apples), up from 111,000 hectares of vegetables and 12,000 hectares of fruit orchards as recently as 1999. Shanghai produces approximately 110,000 hectares of rice, down from 325,000 hectares in 1979. Approximately 30,000 hectares of rapeseed are produced in Shanghai, down from 95,000 hectares as recently as 1992. Shanghai also produces about 20,000 hectares of wheat, down from 100,000 hectares in 1998. High-value crops such as vegetables and fruits dominate crop production in Shanghai. In 2004, rice accounted for about 15 percent of the total value of crop production in Shanghai. Wheat and rapeseed each accounted for only about 1percent. The current provincial-level expected yield for rice is approximately 8.0 tons per hectare. Expected rice yields have trended steadily upward from about 4.5 tons per hectare in 1980. The expected rapeseed yield is about 1.8 tons per hectare, down from about 2.1 tons per hectare in 1980. The expected wheat yield is approximately 3.8 tons per hectare, which is about the same as in 1980. Meteorological Stations Shanghai has 190 weather stations, 70 of which are automated (although most measure only rainfall), 10 are national stations, and only 1 station reports internationally. Many of the automated stations have only recently been installed. Data for Pilot Counties and Farms Three pilot counties or districts were selected for analysis: Jiading District, located approximately 20 kilometers northwest of downtown Shanghai; Nanhui District, approximately 25 kilometers southeast of downtown Shanghai; and Chongming County, an island with an area of more than 1,200 square kilometers, situated in the mouth of the Yangtze River north of downtown Shanghai. For each of the three pilot counties or districts, annual data were obtained on hectares planted and tons produced for grain (primarily rice but also including some wheat), rapeseed, melons, and vegetables. These data were available for the period 1985–2004 and are aggregated to the county or district level. All three counties or districts produce all of these crops. However, Chongming County produces much more rapeseed and grain than either Jiading or Nanhui Districts. Nanhi District produces more melons than the other counties or districts. All three produce vegetables. Monthly rainfall and temperature data were available for Shanghai from 1986–2004. However, since neither rainfall (apart from rainfall associated with typhoon) nor temperature is a significant crop peril in Shanghai, these data were not evaluated for the risk assessment. Data were also provided on typhoon and tropical storm occurrences along the southeast coast of China from 1986–2005. Each typhoon or tropical storm was 55 categorized according to maximum wind speed. Also included are the date and the location where the typhoon or tropical storm made landfall. Although these data were instructive, much longer time series of data would be required to conduct risk assessment for low-frequency events such as typhoons. In addition, to assess the impact of typhoons on crop production, estimates of crop losses associated with historical typhoon events would be required. Municipality-level data were also provided on hectares covered by, and affected by, drought, flood, hail, and frost between 1979 and 2004. (As mentioned earlier, hectares covered by each of these natural disasters are those that experience a yield loss of more than 10 percent, and hectares affected are those that experience a yield loss of more than 30 percent. Statistical analyses of these data suggest that nearly half of lost hectares are from flooding (including waterlogging) and about one-quarter are from frost or freeze events. Hail and drought account for about 12 percent each of all causes of loss. In 1999, 31 percent of planted hectares experienced at least a 10 percent yield loss due to frost or freeze. In the same year, 15 percent of planted hectares experienced at least a 10 percent yield loss due to flooding. In 1994, 21 percent of planted hectares experienced at least a 10 percent yield loss due to drought. No other year in the time series had more than 15 percent of planted hectares with at least a 10 percent yield loss due to one of these natural disasters. It is important to note that typhoon is not among the perils for which these data are collected. Also, for risk assessment purposes these data are of limited value since they are not crop-specific nor do they indicate the exact amount of yield loss. Primary Perils Typhoon is the primary peril affecting crop production in Shanghai. For example, Typhoon Maisha caused major damage to crops and aquaculture on August 5, 2005. Typhoons of Maisha’s magnitude occur in Shanghai about every 10 years. Typhoons generally occur between June and September. High rainfall from typhoons causes flooding and waterlogging. Typhoons also cause wind damage to orchard fruit and greenhouses. Other perils that cause crop losses include hail, drought, prolonged periods of extremely high or low temperatures, and frost or freeze. For rice, an important peril is excessive rain during harvest (late November). Rice yield losses can also be caused by high winds or high temperatures during flowering (late September). Risk Assessment of Pilot Counties and Farms Using the available data for the three pilot counties and districts, a risk assessment was conducted for Shanghai. Data were provided for vegetables and melons. However, these data are aggregated over many different crop species that are produced at different times during the year, so the data are not appropriate for purposes of assessing risk. Thus, the risk assessment focuses on grain (rice and wheat) and rapeseed. Yield-loss exposure is presented using both tables and graphs. The tables present the average annual loss. The graphs contain loss exceedance curves that indicate the probability of losses in excess of some magnitude. It should be noted that with only limited years of yield data it is particularly difficult to model yield risk exposure due to catastrophic events such as typhoons. Infrequent but catastrophic events may be either underrepresented or 56 overrepresented in the available yield data. The procedures used to generate these measures are described in the last section of the annex. Table A2.30: Shanghai Average Annual Loss Grain Rapeseed Shanghai (county and district levels) 3% 13% Source: Authors’ calculation. Figure A2.12: Shanghai—Yield-Loss Exposure Shanghai: Yield Loss Exposure (County/District Level Data) 100% 90% 80% 70% Probability 60% 50% 40% 30% 20% 10% 0% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Loss Relative to Expected Value Grain Rapeseed Source: Authors’ calculation. In comparing the yield-loss exposure for Shanghai to that for other provinces, it is important to remember that the Shanghai data are at the county and district levels rather than the township and team levels, as in other provinces. The more aggregate nature of the Shanghai data will greatly bias the yield-loss exposure downward in comparison to the other provinces. More importantly, the risk assessment greatly underestimates the risk that would exist at a village or household level. Loss Ratio Risk Exposure When establishing premium rates, it is helpful for an agricultural insurer to have estimates of the average annual loss and yield-loss exceedance curves. But these measures are not adequate for determining the amount of contingent capital that might be required to pay indemnities in the event of extreme losses. To know how much contingent capital (reinsurance, reserves, and so on) might be required, the insurer needs some estimate of how large the indemnities might be, relative to premiums collected. 57 Next, consideration is given to the potential loss ratios for a hypothetical insurance company that sells multiple-peril crop insurance in Shanghai. An insurance loss ratio is calculated as indemnities paid divided by premiums collected. The last section of the annex contains details on how the loss ratios presented here were calculated. The loss ratio exposure for a hypothetical insurance company is presented as the probable maximum loss for various return periods. Loss ratio PMLs are valuable for insurance companies, because they demonstrate the insurer’s exposure to extreme loss events. To maintain financial solvency over the long term, insurers should have sufficient access to contingent capital to pay indemnities from at least a 100-year return loss event. Loss ratio PMLs for Shanghai are presented in table A2.31. Due to the aggregation inherent in the vegetable and melon data, the analysis focuses only on grain and rapeseed. Further, since grain is also an aggregate category that includes both rice and wheat, it is not possible to consider a portfolio that includes both grain and rapeseed. Table A2.31: Shanghai Loss Ratio Exposure Grain Rapeseed Probable Maximum Loss (10-year return) 3.5 2.9 Probable Maximum Loss (20-year return) 6.9 4.4 Probable Maximum Loss (100-year return) 8.8 5.7 Source: Authors’ calculation. These findings suggest that to cover a 1-in-100-year event, an insurance company that was insuring grain or rapeseed (but not both) would need sufficient access to contingent capital to pay indemnities of 6–9 times premiums collected. 2.5. Hainan Hainan Province is an island in the South China Sea off the coast of Guangdong Province. Much of the island has a tropical climate. Colder temperatures and higher rainfall occur in higher elevations—primarily in the central and southeastern parts of the island. Major crop commodities produced in Hainan include rubber, banana, rice, coconut palm, oil palm, betel palm, pepper, sisal hemp, lemon grass, cashew, cocoa, wheat, sweet potato, cassava, taro, maize, Chinese sorghum, millet, beans, sugarcane, peanut, sesame, tea, pineapple, litchi, longan, plantain, citrus, mango, watermelon, parambola, and jackfruit. In addition, more than 120 kinds of vegetables are grown. Most rubber production is on state farms in a higher elevation belt that runs across the central part of the island from northwest to southeast. Currently, there are about 470,000 hectares of rubber trees under cultivation in Hainan. A typical rubber yield is about 0.7 tons per hectare. Rubber is harvested from March to October. Two-thirds of the banana production is in the southwestern part of the province. Most banana production is on large-scale commercial farms that sell in the international market. Only about 30 percent of banana production is by small-scale farmers, and most of that is in Ledong County. In 58 total, Hainan has 35,000 hectares of bananas with an average yield of about 30 tons per hectare, though large commercial operations often have yields in excess of 40 tons per hectare. The harvest period depends on where the bananas are grown. In the main southwestern production region, bananas are harvested from February through June. Hainan also has about 335,000 hectares of rice, with an average yield of about 4.3 tons per hectare. Meteorological Stations Hainan has five weather stations that report internationally. Two are located offshore. The three that are on the island are at Haikou, Dongfeng, and Sanya. Data for Pilot Counties and Farms One county and one farm were selected for analysis. Ledong County is a major banana producing county located in southwestern Hainan. Annual township-level data on tons produced and hectares planted of bananas was available for Ledong County for 1990– 2005. Nongken Farm is a major rubber producer in Hainan. Annual team-level data on tons produced and hectares planted of rubber trees was available for Nongken Farm for 1985–2005. Monthly rainfall and temperature data were available for Hainan for 1986–2005. However, since neither rainfall (apart from rainfall associated with typhoon) nor temperature is a significant crop peril in Hainan, these data were not evaluated for the risk assessment. Data were also provided on typhoon and tropical storm occurrences in the vicinity of Hainan for 1986–2005. For each typhoon or tropical storm, data was provided on the dates of influence, location of landfall on Hainan, air pressure, and wind speed. Although these data were instructive, much longer time series of data would be required to conduct risk assessment for low-frequency events such as typhoons. In addition, to assess the impact of typhoons on crop production, estimates of crop losses associated with historical typhoon events would be required. Provincial-level data were also provided on hectares covered by, and affected by, drought, flood, hail, frost, and typhoon for 1987–2004. (As noted earlier, hectares covered by each of these natural disasters are those that experience a yield loss of more than 10 Percent, and hectares affected are those that experience a yield loss of more than 30 percent.) Statistical analyses of these data indicate that about one-third of lost hectares are from drought and about one-third are from flooding (including waterlogging). Typhoon data are only available for a limited number of years. However, between 2000 and 2003, 22 percent of hectares planted had at least 10 percent loss due to typhoon. In 1989, 45 percent of planted hectares experienced at least a 10 percent yield loss due to flooding. In 1988, 20 percent of planted hectares experienced at least a 10 percent yield loss due to flooding, and another 20 percent of planted hectares experienced at least a 10 percent yield loss due to drought. In 1991, drought and hail each caused about 17 percent of planted hectares to experience at least a 10 percent yield loss. This is the only year in the time series with significant hail losses. In 1999, 14 percent of planted hectares experienced at least a 10 percent yield loss due to frost. This is the only year in the time series with significant frost losses. For risk assessment purposes these data are of limited 59 value, because they are not crop-specific nor do they indicate the exact amount of yield loss. Primary Perils Typhoon is the primary peril affecting crop production in Hainan. Typhoons generally occur between May and November, with a peak in August and September. Crop losses are caused by wind damage, flooding, and waterlogging. On average there are 3.1 typhoon landings on the island per year, and if tropical depression is included, the figures increase to 7.7 per year. Major typhoons occurred in 1966, 1973, 1989 and 2005. In 2005, Typhoon Dawai hit Hainan, causing more damage than any typhoon in the past 30 years. Agricultural losses were estimated at RMB 8 billion. More than 10 percent of the rubber trees in Hainan were broken by the typhoon. One county lost 20 percent of its rubber trees. Some farms lost up to 80 percent of their trees. Typhoon risk is not uniform throughout the province. The east of the island (the main direction of arrival of typhoons) has a higher frequency. Banana production has been moved to the west of the island, largely as a result of typhoon exposure. Drought can also cause yield losses for rubber and other crops in Hainan. Commercial banana production is all irrigated. Risk Assessment of Pilot Counties and Farms The risk assessment focuses on banana production in Ledong County (located in southwestern Hainan) and rubber production for Nongken Farm. Rubber is produced primarily in the higher elevations in the central and southeastern parts of Hainan. It is important to note that these data only reflect yield loss for bananas and rubber. They do not reflect the value of damage to the rubber or banana trees. It is also important to note that the risk assessment is based on yield data through 2005. Typhoon Dawai hit Hainan in September 2005 and caused tremendous damage to both banana and rubber trees. However, its impact on yields in subsequent years is not reflected in these data. Yield-loss exposure is presented using both tables and graphs. The tables present the average annual loss, and the graphs contain loss exceedance curves that indicate the probability of losses in excess of some magnitude. It should be noted that with only limited years of yield data it is particularly difficult to model yield risk exposure due to catastrophic events such as typhoons. Infrequent but catastrophic events may be either underrepresented or overrepresented in the available yield data. The procedures used to generate these measures are described in the last section of the annex. Table A2.32: Hainan Average Annual Loss Ledong Nongken Banana Rubber Hainan (township-level bananas and team-level rubber) 10% 9% Source: Authors’ calculation. 60 Figure A2.13: Hainan—Yield-Loss Exposure Hainan: Yield Loss Exposure (Township/Team Level Data) 100% 90% 80% 70% Probability 60% 50% 40% 30% 20% 10% 0% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Loss Relative to Expected Value Ledong Banana Nongken Rubber Source: Authors’ calculation. Loss Ratio Risk Exposure When establishing premium rates, it is helpful for an agricultural insurer to have estimates of the average annual loss and yield-loss exceedance curves. But these measures are not adequate for determining the amount of contingent capital that might be required to pay indemnities in the event of extreme losses. To know how much contingent capital (reinsurance, reserves, and so on) might be required, the insurer needs some estimate of how large the indemnities might be, relative to premiums collected. Next, consideration is given to the potential loss ratios for a hypothetical insurance company that sells multiple-peril crop insurance in Hainan. An insurance loss ratio is calculated as indemnities paid divided by premiums collected. The last section of the annex contains details on how the loss ratios presented here were calculated. The loss ratio exposure for a hypothetical insurance company is presented as the probable maximum loss for various return periods. Loss ratio PMLs are valuable for insurance companies, because they demonstrate the insurer’s exposure to extreme loss events. To maintain financial solvency over the long term, insurers should have sufficient access to contingent capital to pay indemnities from at least a 100-year return loss event. Loss ratio PMLs for Hainan are presented in table A2.33. The loss ratio PMLs for bananas are larger than those for rubber. The portfolio is weighted to reflect the relative value of production for these two crops in Hainan (74 percent rubber and 26 percent banana). Despite the fact that rubber dominates the portfolio and banana and rubber losses are highly correlated at 0.78, the portfolio generates significant reduction in loss ratio exposure, especially for 1-in-100-year loss events. 61 Table A2.33: Hainan Loss Ratio Exposure Ledong Nongken Portfolio Banana Rubber Probable Maximum Loss (10-year return) 2.8 1.8 2.0 Probable Maximum Loss (20-year return) 3.5 2.9 2.2 Probable Maximum Loss (100-year return) 4.5 3.9 2.6 Source: Authors’ calculation. These findings suggest that to cover a 1-in-100-year event, an insurance company insuring bananas or rubber (but not both) would need sufficient access to contingent capital to pay indemnities of between 4 and 4.5 times premiums collected. For the portfolio, this drops to less than 3 times premiums collected. However, again caution is warranted in using these findings, because it is very difficult to model loss ratio exposure for crops that are susceptible to typhoon. The available time series of yield data may not be representative of the actual probability and magnitude of typhoon losses. 2.6. Methodology The following subscripts are used throughout this discussion of risk assessment methods: c = a specific crop. t = a specific year. i = a specific township or team. j = a specific county, district, or state farm. k = a specific province or municipality. Yield-Loss Exposure Province- and municipality-level annual yields were calculated as follows: tonsckt yield ckt = . hectaresckt Township- and team-level annual yields were calculated as follows: tonscit yield cit = . 11 hectarescit All data on tons and hectares are from the National Bureau of Statistics, provincial statistical bureaus, or state farms. 11 For Shanghai, township-level yield data were not available; so county- and district-level yield data were used instead. 62 Historical yield data often reflect yield trends associated with changes in agricultural technologies. To conduct risk assessment it is necessary to convert historical yield data to current technology levels. Although it is possible to detrend historical yield data at the township and team levels, these data are more variable than yield data at higher levels of aggregation. Thus, there would be more statistical error in the estimates of yield trends. With aggregate (for example, province or municipality) yield data, pooling reduces the effects of local idiosyncratic factors on yields, so trend estimates are more likely to reflect systemic technology changes. A common practice is to estimate yield trends at higher levels of aggregation and then apply the results when detrending yield data at lower levels of aggregation. For this analysis, LOESS statistical procedures were used to estimate yield trends from the province- and municipality-level yield data. LOESS uses nonparametric methods for estimating local regression surfaces and is more flexible than traditional ordinary least squares (OLS) models. (For an explanation of LOESS, see http://support.sas.com/rnd/app/papers/loesssugi.pdf.) The LOESS trend procedure generated predicted yields for each crop, province or municipality, and year. Those predicted yields are designated as: predicted yield ckt . The following ratio was then calculated: predicted yield ck 2004 ratiockt = . predicted yield ckt Notice that for 2004, the ratio is equal to 1. If the yield trend is uniformly positive, the ratio will be greater than 1 for years prior to 2004. Township- and team-level yields were then detrended to 2004 technology as: detrended yield cit = yield cit × ratiockt ∀i ∈ k . For some townships and teams, 2005 yield data were available. The ratio was set equal to 1 to calculate detrended yields for 2005. For each crop and township or team, expected yields were calculated as the average over the time series of the detrended yields, where m is the first year of available yield data and n is the last year: n ∑ detrended yield cit expected yield ci = t =m n−m 63 For each crop and township or team, the annual percentage yield loss relative to the expected yield was calculated as: ⎛ expected yield ci − detrended yield cit ⎞ percentage yield losscit = max⎜ 0, ⎜ ⎟. ⎟ âŽ? expected yield ci ⎠ The annual weighted average percentage loss for each crop and county or farm was then calculated as: percentage yield losscit × weight cit percentage yield losscjt = ∀i∈ j ∑ weightcit i …where: weight cit = expected yield ci × planted hectarescit . In other words, the weight was the tons of production expected from each township or team given, the planted hectares. Estimates of percentage yield losscjt can be calculated only for the limited years of available yield data. Kernel smoothing was used to derive a continuous nonparametric density function from the discrete estimates. Loss exceedance curves are based on the cumulative distribution of the kernel-smoothed density function. Expected annual average losses are the 50th percentile of the cumulative distribution. Loss Ratio Exposure The loss ratio risk exposure for hypothetical insurance companies was based on an assumption that multiple-peril crop insurance was purchased by each township or team in the pilot counties and farms. The insurance is assumed to have no deductible and actuarially fair premium rates (that is, over the time period analyzed, premium rates for each county or farm were set so that total premiums collected were exactly equal to total indemnities paid). Despite these assumptions, it is important to note that it is not recommending that insurers in China: (1) adopt multiple-peril crop insurance; (2) sell crop insurance policies with no deductible; or, (3) set crop insurance premium rates at actuarially fair levels. There are various reasons why these assumptions were adopted for this analysis. First, township- or team-level data were not available on the magnitude of yield losses caused by specific perils. Thus, the assessment of loss ratio exposure must be based on an assumption of multiple-peril insurance. Second, for crop insurance sold to farm households, 20–30 percent deductibles are typical. However, given that the available yield data are at the township and team levels rather than at the farm household level, the calculations assume no deductible. This should generate loss ratios that are roughly equivalent to what one would expect using typical deductibles at the farm household level. Third, crop insurance premium rates are typically loaded for items such 64 as operating expenses, reserve building, and return on equity. However, applying a premium load will reduce overall loss ratios. Thus, actuarially fair premium rates are assumed to generate conservative estimates of possible loss ratio exposure. To estimate loss ratio exposure, a county- or farm-level actuarially fair premium rate (assuming no deductible) was calculated as: n ∑ percentage yield loss cjt premium ratecj = t =m n−m …where n and m are as defined earlier. In other words, the actuarially fair premium rate is the simple average of the empirical estimates of percentage yield losscjt . A county- or farm-level annual loss ratio is then calculated as: percentage yield losscjt loss ratiocjt = . premium ratecj Again, estimates of loss ratiocjt can be calculated only for the limited years of available yield data. Kernel smoothing was used to derive a continuous nonparametric density function from the discrete estimates. Loss ratio PMLs are based on the cumulative distribution of the kernel-smoothed density function. Specifically, loss ratio PMLs for 10- , 20-, and 100-year returns correspond to the 10, 5, and 1 percentile of the cumulative distribution. Portfolios If the loss ratios for insuring different crops and/or regions are less than perfectly correlated, then a diversified portfolio of insurance policies from these different crops and regions can reduce loss ratio variability relative to that for a single crop or region. Presented first is a conceptual discussion of how a diversified portfolio can reduce loss ratio variability. Next, there is a description of how portfolios were constructed for the analyses of loss risk exposure presented earlier. Conceptual discussion of portfolios Define xa as a weight that reflects the percentage of the overall portfolio value that is due to the ath crop-region combination (for example, insurance on soybean production in Nenjiang County, Heilongjiang). Also define the rate of return for the insurance portfolio as 1 minus the loss ratio. Thus, if the loss ratio is 0.70, the rate of return is 30 percent. If the loss ratio is 1.30, the rate of return is -30 percent. If the portfolio consists of only two crop-region combinations (for example, insurance on Nenjiang County soybeans and insurance on Nenjiang County wheat, or insurance on Nenjiang County soybeans and insurance on Jixian County soybeans) then: 65 σ portfolio = xa σ a + xb σ b2 + 2 xa xbσ ab 2 2 2 …where σ portfolio is the standard deviation of the rate of return for the portfolio, σ a is the 2 variance of the rate of return for the ath crop-region combination, and σ ab is the covariance of the rates of return for ath and bth crop-region combinations. More generally, if there are q crop-region combinations in the portfolio of insurance policies, then: q q σ portfolio = ∑∑ x x σ a b a b ab . …where σ aa = σ a is the variance of the rate of return for the ath crop-region 2 combination. The covariance is related to the correlation coefficient as: σ ab = σ aσ b Ï? ab …where Ï? ab is the correlation coefficient between the ath and bth crop-region combinations. This decomposition demonstrates the factors that affect the variability in the rate of return (or loss ratio) for the portfolio. They are: 1) the relative weights of the various elements of the portfolio, x1 , x2 . . . xq ; 2) the standard deviation of the rate of return for each of the various elements of the portfolio, σ 1 , σ 2 , . . . σ q ; and, 3) the correlations between the rates of return for each of the various elements of the portfolio, Ï? ab ∀ a = 1, 2, . . . q; b = 1, 2, . . . q; a ≠ b . The decomposition also demonstrates the following generalizations about the amount of risk reduction (or reduction in loss ratio PMLs) that can be obtained by pooling various crop-region combinations of insurance policies into a portfolio: 1) the more (less) that one crop-region combination dominates the portfolio, the less (more) risk reduction is obtained by creating the portfolio; 2) the higher (lower) the correlation between rates of return for each of the various crop-region combinations in the portfolio, the less (more) risk reduction is obtained by creating the portfolio. 66 Constructing portfolios for the analyses of loss risk exposure To construct portfolios for the analyses of loss ratio exposure, it was necessary to aggregate across different crops. This implies that the estimates of percentage yield losscjt must be converted into monetary units. To calculate annual indemnities for the portfolio, the following steps were required: First, the county- or farm-level annual expected yield was calculated as a weighted average of the township- or team-level annual expected yields: n ∑ expected yield × weightci cit expected yield cjt = t =m ∀i ∈ j ∑ weight i cit …where weight cit is as defined previously. Note that expected yield ci does not vary through the time series but weight cit does vary due to the variability in planted hectarescit . This is why expected yield cjt varies through the time series. An overall county- or farm-level expected yield was then calculated as: n ∑ expected yield cjt expected yield cj = t =m . n−m Second, expected hectarescj was calculated as the average of planted hectarescj for the last three years of the time series. Third, for each crop-region combination, the annual indemnity was then calculated as : indemnitycjt = percentage yield losscjt × expected yield cj × expected hectarescj . Fourth, the total annual indemnities for the portfolio are: portfolio indemnityt = ∑∑ indemnitycjt . c j To calculate premiums for the portfolio the following steps were required: First, the actuarially fair premium rate for each crop-region combination was calculated as: n ∑ indemnity cjt premium ratecj = t =m n−m …and the annual premium was calculated as: 67 premiumcj = premium ratecj × expected yield cj × expected hectarescj . Note that the premium does not vary over time. The total annual premiums for the portfolio (which, again, do not vary over time) are: portfolio premium = ∑∑ premiumcj . c j The annual loss ratio for the portfolio can then be calculated as: portfolio indemnityt portfolio loss ratiot = portfolio premium Again, estimates of portfolio loss ratiot can be calculated only for the limited years of available yield data. Kernel smoothing was used to derive a continuous nonparametric density function from the discrete estimates. Loss ratio PMLs are based on the cumulative distribution of the kernel-smoothed density function. Specifically, loss ratio PMLs for 10-, 20-, and 100-year returns correspond to the 10, 5, and 1 percentile of the cumulative distribution. 68 Annex 3: Macro Risk Policy Framework Following the work presented in Annex 2 on risk assessment, this annex examines the same issues for all provinces in China. Two major data sets are used for this macro analysis: (1) production data for seven major crops by province, and (2) cause-of-loss data by hectares and by province for all crops comingled. Both data sets are available for 1980–2004. Additionally, 2004 data for total crop value for all crops are available by province. Together, these three data sets provide the foundation for developing a macro risk framework to gain insights into the provincial and national cost and risk of crop insurance in China. This annex includes the following sections: (3.1) review of both data sets, and description of the procedures used to convert these data into loss values; (3.2) comparison of the data sets to identify trends in the data and the relative risk for major crops and provinces; (3.3) development of a macro portfolio model using the cause-of-loss data; (3.4) expansion of a risk-financing section to illustrate insurance principles under the strong assumption that the macro portfolio model developed in Section 3.3 would be reflective of a mature crop insurance program in China. 3.1. Data Set Review This section presents a review of both data sets and a description of the procedures used to convert these data into loss values. Data and Methods Using Provincial Crop Data China has great variance in risk across crops and geographic areas. The first data set is from the National Bureau of Statistics on tons produced and hectares planted for seven major crops in 30 provinces. A price matrix for these seven crops was also created by each province. Hectares planted and production data are used to develop the yield per hectare over the time series: tons cpt yield cpt = hectares cpt …where: c = crop (Corn, Wheat, Peanuts, Rice, Rapeseed, Soybeans, or Cotton) p = province t = year (1978–2004) When data series for a province are incomplete, data from a neighboring province are used to develop a “proxyâ€? value. Thus, although there are missing crops for many provinces (because those crops are not produced at a significant level), every crop array that has data will have a completed array of data via this proxy replacement method. 69 Historical yield data often reveal yield trends associated with changes in agricultural technologies. To conduct risk assessment, it is necessary to adjust historical yield data to current technology levels. For this analysis, LOESS statistical procedures are used to estimate yield trends from the provincial yield data. LOESS procedures use nonparametric methods for estimating local regression surfaces and are more flexible than using traditional ordinary least squares (OLS) models. (For an explanation of LOESS, see http://support.sas.com/rnd/app/papers/loesssugi.pdf.) The LOESS trend procedure generated predicted yields for each crop, province, and year. Those predicted yields are designated as predicted yield cpt . As a final step to the trend process, each graph was visually inspected. In cases where the LOESS procedure tended to overfit the data, linear spline regressions were fitted to develop the trend yield. This was also done with visual data inspection. After the predicted yields were developed, the following ratio was then calculated to normalize all yields to the 2004 base: predicted yield cp 2004 ratio cpt = predicted yield cpt For 2004, the ratio is equal to 1. If the yield trend is uniformly positive, the ratio will be greater than 1 for all years prior to 2004. This ratio allows for detrending the province yields to 2004 technology, using the following simple equation: detrended yieldcpt = yield cpt × ratiocpt These trend-adjusted yield data allow for evaluation of yield loss relative to the expected yield and can be used to compare yield risk for different crops in different regions. A limitation of these data is that they are aggregate data and cannot be used to estimate the yield risks that are present for household farm units. Nonetheless, these data do provide a basis for understanding trends in losses, and how crop losses may be correlated within the same province or across provinces. Losses are developed when the trend-adjusted yield is below the predicted yield of the series of data for that crop. Thus, a series of normalized losses can be developed as a percentage of the predicted yield for 2004. If detrended yield cpt > predicted yield cp2004, then percentage yield loss cpt = 0. ( predicted yield cp 2004 − detrended yield cpt ) Or else, percentage yield losscpt = predicted yield cp 2004 The average annual loss is the simple average of the loss cpt. : 70 n ∑ percentage yield loss cpt average annual loss cp = t =m n−m Average annual loss can also be used as the basis for developing premium rates. For this analytical work, the simple average loss can be used as a proxy premium rate. When this is done, premium values will be equal to indemnity payouts over the long run. No agricultural insurer would price insurance in this fashion, because extra “loadsâ€? would need to be added to the price to cover the many cost items discussed in Chapter 3. Nonetheless, these data provide the basis for the first part of this macro risk analysis; these average annual losses are used to understand more about the potential correlation among crop losses and to provide some focus on how average annual losses could be changing over time. Even though these values are aggregate data, they can be used for these purposes. However, to use them most effectively, it is also important to normalize the values in a portfolio problem by using the price of the commodities and the share of the planting for recent years. Developing a portfolio analysis with normalized values will follow the risk assessment section. Data and Methods of Using the Cause-of-Loss Data A highly useful data set was discovered in the annual statistics for the government. These data contain crop losses at the provincial level and are organized by cause of loss. They include losses due to: (1) drought, (2) floods, (3) freeze or frost, (4) hail, and (5) typhoons. The data are by province and are sorted into three categories: • Total sown hectares; • Number of hectares with 10–30 percent damage; and • Number of hectares with more than 30 percent damage. The first task is to use the three data points to estimate an equivalency of hectares totally lost. With this estimate, the percentage of yield loss can be estimated for each year by taking the ratio of equivalent hectares loss divided by total hectares sown. Special procedures were developed to make an estimate of the equivalent hectares loss. It is important to first estimate the median values for the percentage of loss of 10–30 percent and for the range of loss above 30 percent. For example, if it is known that the median was 20 percent damage and there were 10,000 damaged hectares in that range (the 10–30 percent range), there would be an equivalent hectares loss of 2,000 hectares that are “totally lostâ€? (0.20 times 10,000). Likewise if it is known that the median loss was 35 percent in the category of losses above 30 percent, and there were 10,000 hectares in that grouping, the equivalent hectares loss would be 3,500 hectares (0.35 times 10,000). If the total sown hectares were 100,000, then the percentage loss would equal 5,500 divided by 100,000, or 5.5 percent. 71 To formalize the relationships, consider data that are in the category of 10–30 percent loss as Bin 1 and those with more than 30 percent loss as Bin 2. As the percent of the total hectares sown increases in each of these bins, the median value for the bin should also be greater. In principle the relationship is as follows: Median of Bin 1 = f (percent of sown hectares in Bin 1, percent of sown hectares in Bin 2) Median of Bin 2 = f (percent of sown hectares in Bin 2) Two multiple regression equations were developed to operationalize these relationships. Finally, an upper limit on the median values was imposed for each bin. The upper limit used for Bin 1 was 22 percent, and the upper limit for Bin 2 was 50 percent. The sensitivity of these somewhat arbitrary limits was tested and found to be acceptable. These procedures are repeated for each cause of loss. Once there is an estimate of the median for the two ranges, the equivalent hectares loss above 10 percent is calculated as: equivalent hectares loss = (median estimate1 * hectares1 + median estimate2 * hectares2 ) …where 1= values for range between 10 and 30; 2= values for the range of 30 and above. Finally, the key variables for percentage loss by each cause of loss can be calculated for all provinces and all years. Given that this value is for all crop hectares, it also represents aggregate losses for all crops. equivalent hectares losslpt percentage losslpt = hectares sownlpt …where l = cause of loss (drought, flood, freeze or frost, hail, typhoon). To obtain the estimate of multiple perils, each of the five estimates of cause of loss are added together. percentage loss pt = ∑1 percentage losslpt . 5 3.2. Comparison of Data Sets This section discusses comparison of data sets to identify trends in the data and the relative risk for major crops and provinces. Given estimates of the historic percentage loss from the two data sets, it is possible to examine a number of important relationships to obtain insights into risk relationships. For this comparison to be most effective, the data must be normalized with prices and crop value, which allows the relative values of the crops to be factored into the analysis—a requirement when considering any insurance portfolio problem. However, before moving to the portfolio problem, it is useful to provide basic statistics. The first comparison simply examines the estimates of the crop value, by crop, to the total crop value in 2004. 72 Table A3.1 provides the overall profile of crop production by value of all crops in China. Total crop value is RMB 1,780 billion. Of this total and from the data available, value estimates for the seven key crops is RMB 689 billion, or about 40 percent of all crop value in China. Given that some of the key crops have missing data for certain provinces, this is a low estimate of the percentage of total value of these seven crops. Still, the data appear to be relatively complete. That less than 50 percent of the total crop value comprises these seven dominant crops illustrates the great diversity of crops in China. As further evidence of that diversity, Table A3.2 provides an indication of the share of each of the seven crops relative to other crops for each province. For the crop data that are present, rice appears to represent the largest market share at about 15 percent of all crop value. Corn and wheat each appears to account for about 7 percent of all crop value. Peanuts, cotton, and soybeans each represents about 3 percent of all crop value. Rapeseed represents less than 2 percent of all crop value. Figure A3.1 provides a geographic view of the national crop share data that appear in Table A3.1. This figure reveals that the primary crop production is in the eastern-central region of China. 73 Table A3.1: Estimates of Crop Value by Province—All Crops and Seven Key Crops (2004) Seven Crop National Crop Province All Crops Seven Crops Share in Share Province (%) Values in Million RMB (%) Shandong 10.6 189,170 63,609 34 Henan 9.0 160,290 74,027 46 Jiangsu 7.0 124,240 49,138 40 Hebei 6.4 113,570 40,330 36 Sichuan 5.5 98,770 43,176 44 Guangdong 5.4 96,000 23,423 24 Hubei 5.2 92,160 37,359 41 Hunan 4.9 87,400 41,845 48 Anhui 4.7 84,200 44,835 53 Guangxi 3.5 62,310 22,790 37 Heilongjiang 3.5 62,020 39,121 63 Liaoning 3.4 61,130 18,442 30 Zhejiang 3.3 59,260 16,125 27 Fujian 3.0 52,580 10,297 20 Yunnan 2.9 51,690 20,174 39 Xinjiang 2.9 51,500 23,225 45 Jiangxi 2.8 49,110 26,076 53 Jilin 2.7 48,620 25,172 52 Shaanxi 2.3 41,370 12,832 31 Inner Mongolia 2.3 41,150 13,461 33 Gansu 1.9 33,140 8,410 25 Guizhou 1.8 31,770 14,661 46 Shanxi 1.6 29,050 9,930 34 Hainan 1.0 17,090 n.a. n.a. Shanghai 0.6 10,930 1,882 17 Tianjin 0.5 9,530 2,524 26 Beijing 0.5 9,270 953 10 Ningxia 0.4 7,130 3,301 46 Qinghai 0.2 3,420 1,083 32 Tibet 0.1 2,660 552 21 n.a. = not applicable. Sources: Authors. 74 Table A3.2: Market Share of Each Crop by Province Other Province Corn Wheat Peanut Rice Rapeseed Soybean Cotton Crops (%) Anhui 4 13 4 21 4 4 3 47 Beijing 4 3 1 0 0 1 1 90 Fujian 0 0 2 16 0 1 0 80 Gansu 6 12 0 0 2 2 3 75 Guangdong 1 0 4 20 0 1 0 76 Guangxi 3 0 4 28 0 2 0 63 Guizhou 12 3 1 21 6 3 0 54 Hebei 10 14 5 1 0 1 4 64 Heilongjiang 11 1 0 27 0 23 0 37 Henan 7 21 7 4 1 2 5 54 Hubei 2 3 3 23 5 2 3 59 Hunan 2 0 1 38 3 2 2 52 Inner 19 4 0 2 2 5 0 67 Mongolia Jiangsu 2 7 3 20 3 2 3 60 Jiangxi 0 0 3 46 2 1 1 47 Jilin 29 0 2 13 0 7 0 48 Liaoning 14 0 3 11 0 2 0 70 Ningxia 15 17 0 12 0 2 0 54 Qinghai 0 14 0 0 18 0 0 68 Shaanxi 9 13 1 3 2 2 1 69 Shandong 8 12 8 1 0 1 5 66 Shanghai 0 1 0 14 1 1 0 83 Shanxi 17 10 0 0 0 3 3 66 Sichuan 6 6 3 22 4 3 0 56 Tianjin 6 6 0 2 0 1 11 74 Tibet 1 13 0 0 7 0 0 79 Xinjiang 6 10 0 2 1 1 27 55 Yunnan 10 3 0 21 1 3 0 61 Zhejiang 0 0 0 23 1 2 0 73 Sources: Authors. 75 Figure A3.1: National Crop Share by Province Sources: Authors. Table A3.2 provides a further breakdown of the market share of each of the seven crops relative to the total crop value in each province. These data provide still more details about the variation of crops that are grown across China. The trend-adjusted yield data can now be used to examine a number of risk issues. The most straightforward comparison involves examining the coefficient of variation (CV): standard deviationcp coefficent of variation cp = mean detrended yield cp Although these values provide a first indication of the relative risk of different crops for different provinces, it is critical to recognize that these are aggregate data and the values are likely several magnitudes lower than CV values that would emerge from farm-level data. Farm data can easily have CV values that are in the range of 30–50 percent. Nonetheless, a number of these data exhibit quite high CV values for province-level data. (See Table A3.3.) Any time province-level CVs exceed 20 percent, it is a good indicator of high risk for that crop. 76 Table A3.3: Province Level Coefficients of Variation for Major Crops Province Rice Wheat Corn Soybean Rapeseed Peanut Cotton (%) Anhui 7 16 13 17 19 12 16 Beijing n.a. 3 7 13 n.a. 8 12 Fujian 2 14 9 7 7 7 Gansu 12 9 10 17 8 n.a. 13 Guangdong 3 15 3 6 21 17 n.a. Guangxi 5 7 8 7 10 7 n.a. Guizhou 9 8 7 17 9 11 n.a. Hebei 10 7 5 13 29 7 26 Heilongjiang 7 17 9 12 26 26 n.a. Henan 12 6 13 16 24 10 17 Hubei 4 9 7 11 7 5 16 Hunan 2 7 8 12 5 2 17 Inner Mongolia 7 9 8 20 21 19 n.a. Jiangsu 4 8 7 14 10 8 11 Jiangxi 3 7 12 11 8 8 18 Jilin 15 19 14 14 n.a. 22 n.a. Liaoning 11 18 17 18 n.a. 22 19 Ningxia 8 9 5 30 n.a. n.a. n.a. Qinghai n.a. 7 n.a. n.a. 8 n.a. n.a. Shaanxi 10 9 12 21 14 13 26 Shandong 9 5 9 12 33 12 16 Shanghai 3 10 11 20 16 Shanxi 15 11 12 17 32 14 13 Sichuan 10 4 9 17 5 10 19 Tianjin 14 6 14 20 n.a. 14 16 Tibet n.a. 4 18 n.a. 11 n.a. n.a. Xinjiang 10 2 3 12 7 n.a. 7 Yunnan 4 12 4 27 13 7 n.a. Zhejiang 4 7 7 10 9 5 15 n.a. = not applicable. Sources: Authors. 77 Examining Losses Relative to Expected Value The crop yield data are used to examine the losses relative to the expected value of the crop. To focus this issue, only the top two crops, by value, in each province are used to develop these estimates. The losses relative to expected value (percentage yield loss as presented above) can be combined for the two crops by developing a weighted average based on the relative value contribution of each of the two crops. For example, if a province has 30 percent of the total crop value in rice and 20 percent of the total crop value in corn, focusing on only these top two crops will provide a weight of 60 percent for rice and 40 percent for corn (30/50 and 20/50). The two-crop average annual loss is: Two-crop percentage loss pt = (percentage yield losspt1 * w1 +percentage yield loss pt2 * w2) …where w1 is the weighted percent of the crop contribution from the No. 1 crop and w2 is the weighted percent of the crop contribution from the No. 2 crop. The time series from 1983 to 2004 provides an opportunity to examine the losses of the province more closely. Results are presented in Table A3.7 (at the end of this annex). The results from the crop province data are also complemented by results from the cause-of- loss data where the percentage loss is: equivalent hectares losslpt percentage losslpt = hectares sownlpt In every case the annual average loss from the cause-of-loss data is greater than the average annual loss from either of the top two crops or the top two crops combined. Again, this is because the crop loss data suffer from a very strong aggregation bias. Only results from the first province, Anhui, are discussed in detail to give the reader a guide for how to evaluate the details in Table A3.7. Anhui represents about 5 percent of crop production in China. Figure A3.2 provides a mapping of the estimate of annual loss from the combined two-crop data (rice and wheat) and the cause-of-loss data. These values are 83 percent correlated for this province. This level of correlation provides some indication that the data sources are consistent, although these data are derived from very different processes. Both data sets appear to capture the extreme events in this province. These correlations are reported for all provinces as “two crops and cause of lossâ€? in Table A3.7 (at the end of this annex). The correlation is not as strong in many provinces as it is for Anhui Province. A national annual loss estimate can also be developed by focusing only on the two major crops in each province and the cause-of-loss data. These values are developed by summing up the value across all losses from the cause-of-loss data and the data for the two major crops for each province by year. The data in Figure A3.3 presents a plot of the national annual loss estimates and a trend line for both data sources. The average annual loss from the cause-of-loss data is 8 percent, versus 2.3 percent for the province crop data. This is a good indicator of the aggregation bias that is introduced with the province 78 crop data. These data are correlated at 75 percent. More striking is that the trend on losses is nearly identical from these two very different data sources and methods for estimating national annual losses. Figure A3.2: Comparison of Annual Loss from Province Crop Data and Cause-of- Loss Data for Anhui Province 30 Two crop 25 Cause of Loss 20 15 Percent 10 5 0 1983 1986 1989 1992 1995 1998 2001 2004 Year Sources: Authors. Figure A3.3: National Annual Loss from Province Crop Data and Cause-of-Loss Data 12 Cause of Loss Data 10 8 Percent 6 Crop Data 4 2 0 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 Year Sources: Authors. 79 Kernel smoothing is used to derive a continuous nonparametric density function from the discrete estimates that are represented by the data, such as the estimates presented in Figure A3.2. Loss exceedance curves are based on the cumulative distribution of the kernel-smoothed density function. The probable maximum losses for 1-in-10-year, 1-in- 20-year, and 1-in-100-year loss ratios are estimated and reported throughout Table A3.7 (at the end of this annex). The data from the kernel smoother is plotted for Anhui Province in Figure A3.4. This figure shows the loss exceedance curve for rice, wheat, the two combined crops, and from the cause-of-loss data for all crops. Although the wheat crop is obviously more risky than the rice, combining the two provides a two-crop distribution that lies between the two crops, as expected. Of more interest is that even though there is a significantly greater aggregation bias with the province data than with the cause-of-loss data, the extreme values for the crop losses exceed those of the cause- of-loss data. Although the cause-of-loss data are more reflective of the underlying risk, they also represent a portfolio of all crops. Thus, the extreme values from the cause-of- loss data are many times less than the extreme values from the aggregate single-crop data. This is also evident in Table A3.7, because the 1-in-100-year return value is 30 percent for all crops, based on the cause-of-loss data, 32 percent for the two crops, and 49 percent for wheat. These patterns repeat themselves in many provinces, as is demonstrated throughout Table A3.7. Figure A3.4: Loss Exceedance Curves for Anhui Province 100 90 80 Rice 70 Wheat Two Crops Odds of the Event (%) 60 Cause of Loss 50 40 30 20 10 0 38.4 48.6 35.8 46.1 33.3 43.5 30.7 40.9 25.6 28.1 23.0 20.5 17.9 15.4 10.2 12.8 0.0 2.6 7.7 5.1 Annual Loss as a Percent of Crop (%) Sources: Authors. 80 Average Annual Loss for Various Causes of Loss The cause-of-loss data can be examined more closely to explain where various causes of loss dominate. Again, the average annual loss is estimate by the four main causes of loss: (1) drought, (2) flood, (3) hail, and (4) freeze and frost. Table A3.4 presents the share of cause of loss by event and by province. Figures A3.5–A3.8 provide maps of the average annual loss for the various cause-of-loss events. To give some national perspective, from 1982–2004, 24 percent of all hectares sown experienced losses of 10 percent or more. Nearly 17 percent of all hectares were impacted at this level by drought, 8 percent by floods, 3 percent by hail, and nearly 2 percent by freeze. Typhoon data were less reliable and are not reported. When taking the weighted average of the cause-of-loss and the crop-value matrix, drought accounts for 52 percent of all losses, flood for 28 percent, hail for 10 percent, and frost or freeze for about 6 percent of all losses. The residual (4 percent) is for the typhoon data. These data are not reliable, because there were many missing values. It is likely that typhoon losses are greater than 4 percent. Table A3.4: Share of Cause of Loss Ranked by Province with Largest Market Share of Crop Value Market Share Province Percent of Average Annual Loss by Cause of Loss (%) Drought Floods Hail Freeze (%) Shandong 10.4 71 15 11 3 Henan 8.8 63 24 9 3 Jiangsu 6.9 36 40 12 10 Hebei 6.3 72 10 16 2 Sichuan 5.4 52 32 11 4 Guangdong 5.3 33 42 10 9 Hubei 5.1 43 43 7 6 Hunan 4.8 45 44 6 4 Anhui 4.6 41 45 6 7 Guangxi 3.4 53 32 7 5 Heilongjiang 3.4 53 34 6 6 Liaoning 3.4 68 22 7 2 Zhejiang 3.3 26 47 11 6 Fujian 2.9 32 44 7 10 Yunnan 2.9 50 24 12 13 Xinjiang 2.8 49 12 25 13 Jiangxi 2.7 36 49 7 6 Jilin 2.7 59 28 9 4 Shaanxi 2.3 71 17 7 4 Inner Mongolia 2.3 74 12 9 5 81 Gansu 1.8 72 10 11 7 Guizhou 1.8 51 25 17 6 Shanxi 1.6 80 7 8 4 Hainan 0.9 37 37 5 7 Shanghai 0.6 12 47 12 25 Tianjin 0.5 72 10 17 0 Beijing 0.5 69 8 21 1 Ningxia 0.4 71 8 11 10 Qinghai 0.2 63 8 23 5 Tibet 0.1 58 20 9 11 Sources: Authors. Figure A3.5: Average Annual Loss from Drought Sources: Authors. 82 Figure A3.6: Average Annual Loss from Floods Sources: Authors. Figure A3.7: Average Annual Loss from Hail Sources: Authors. 83 Figure A3.8: Average Annual Loss from Freeze and Frost Sources: Authors. Figure A3.9 presents the average annual loss from all cause-of-loss data. This is the sum of the individual average annual percentage loss values. These rates are also used to develop the base premium rates that are used in the macro model presented below. The actual values appear in Table A3.7 under the “all cropâ€? column that corresponds to the “average annual lossâ€? row. The PML values for the distribution of annual losses also appear in this table. 84 Figure A3.9: Average Annual Loss from All Cause-of-Loss Data Sources: Authors. 3.3. Developing a Macro Portfolio Insurance Model for China Although it is useful to examine the average annual loss, as has been done above, it also is important to develop a national macro portfolio model and to simulate an insurance program. The focus now is on using only the cause-of-loss data, since it is more representative of the underlying risk within each province. The average annual loss values are used as a proxy for premium rates that are expected to only offset the indemnities that are paid over the time period 1980–2004 for the cause-of-loss data. Of course, no insurance company would operate by setting premiums so that only the indemnity payment could be made over time, but doing so here provides a good base for making comparisons and for developing a portfolio model. This “break-evenâ€? premium rate is simply the average of the annual percentage loss values from the cause-of-loss data. To convert everything to a value basis, annual indemnities are simply the product of the percentage annual loss and the most recent crop value for the province (2004): Indemnity pt = percentage annual loss pt × crop value p 2004 The total annual premiums for the portfolio (which, again, do not vary over time) are simply the sum of the premium for any combination of provinces. 85 The annual loss ratio for the portfolio can then be calculated as the sum of all indemnities, by year, for all provinces in the portfolio, divided by the sum of all premiums for all provinces in the portfolio: portfolio loss ratiot = ∑ p portfolio indemnityt ∑ p portfolio premium …where p is any combination of provinces. Again, estimates of portfolio loss ratiot can be calculated only for the limited years of available yield data (1980–2004). Kernel smoothing was used to derive a continuous nonparametric density function from the discrete estimates. Loss ratio PMLs are based on the cumulative distribution of the kernel-smoothed density function. Specifically, loss ratio PMLs for 10-, 20-, and 100-year returns correspond to the 10, 5, and 1 percentile of the cumulative distribution. The loss ratio history can be estimated for a single province or for any combination of provinces. It is critical to note that the province portfolio is implicitly a perfectly spread portfolio for all crops in the province. This means that the risk represented by these portfolios will again be low relative to a portfolio of business that would be generated by an actual portfolio of business in a province. A true portfolio would likely be limited by the crops and would not be perfectly spread over the province. The province data is not broken out by perils. Thus, the risk assessment is for all perils, as would be represented by a multiple-peril crop insurance program. However, this should not be misinterpreted as an endorsement of MPCI. As was clearly laid out in the recommendations presented in Chapter 4, other insurance products should take priority over MPCI. To illustrate the value of pooling across provinces, two provinces are examined in Figure A3.10. Loss exceedance curves—for Heilongjiang, Shanxi, and the pooled business from these two provinces—clearly demonstrate the value of pooling. Even further benefits are demonstrated if all provinces could be pooled together. For example, the 1-in-10-year loss ratio is estimated to be 150 percent for Heilongjiang, 176 percent for Shanxi, and 138 percent when these two provinces are pooled into one book of business. The correlation between the annual average loss values for Heilongjiang and Shanxi is a negative 26 percent. This explains the significant portfolio effect of combining these two provinces into a single insurance portfolio. A national pool would perform only slightly better than the two-province pool with a 1-in-10 value of 130 percent; a 1-in-20 value of 137 percent; and a 1-in-100 value of 147 percent. 86 Table A3.5: Loss Exceedance for Loss Ratios for Heilongjiang, Shanxi, and Heilongjiang, and Shanxi Pooled Together Heilongjiang Shanxi Pooled (%) a Average Annual Loss Ratio 100 100 100 Probable Maximum Loss (10-Year 150 176 138 Return) Probable Maximum Loss (20-Year 170 196 148 Return) Probable Maximum Loss (100-Year 229 226 165 Return) a. The average annual loss ratio is 100% by definition, because premiums have been set at their actuarially fair value. Sources: Authors. Figure A3.10: Loss Exceedance Curves for Loss Ratios for Heilongjiang, Shanxi, and Pooled Business from These Two Provinces 100 90 80 Heilongjiang Shanxi 70 Pooled 60 Odds of the Event (%) 50 40 30 20 10 0 24 48 72 96 120 144 168 192 216 0 Loss Ratio (%) Sources: Authors. 87 3.4. Expanding the Macro Risk Model It is now possible to develop a macro model using the province annual-loss data by year from 1980-2004 and the total crop value data by province. To simplify the analysis, the expected annual-loss data for the province are used as the premium rates. Once again, these values are far too low for any sustainable insurance program, since they include only one of the cost components presented in Chapter 3 of the main report. Cost of insurance = Expected Annual Loss + Expense Loads + Cost of Capital Nonetheless, the values should provide a reasonable basis for considering the underlying risk of an expanded MPCI program in China. As a point of reference, the average premium rate for all of the provinces in the macro model is about 7.8 percent of value insured. Given the 2004 crop value of RMB 1,780 billion, if there were a uniform 10 percent participation across China, the premium total would be RMB 13.9 billion. A Monte Carlo model is developed with the 25 years of correlated annual-loss values. To provide some level of catastrophic accounting, the most extreme value in the 25-year series is increased by 10 percent for each province. These 25 years are used with Simetar’s multivariant generation model to develop 1,000 random draws that maintained the embedded correlations among the province loss values. The random draws are developed from the empirical marginal distributions represented by the actual data, with the extreme value increased by 10 percent as described above. Once again, it must be emphasized that the province data represents a nearly perfectly spread book of business, and the PML values for actual limited books of business by specific insurance companies are likely to be larger, since they are unlikely to be as well-diversified. Thus, the specific levels that might be considered for government stop-loss reinsurance should be different than those that emerge from this macro model. This model allows for consideration and review of a number of the issues that are highlighted in Chapter 4 of the main report. In particular, it is now possible to focus on a national MPCI program for China and the potential implications of following some of the recommendations that were presented in Chapter 4. The following key recommendations are based on the Monte Carlo model: • The central government could offer free stop-loss reinsurance at a 50 percent proportion above certain extreme levels. The stop-loss levels would be different, based on the relative risk of the province, and the reinsurance would be offered free. This would be the central government’s subsidy contribution to agricultural insurance. • The central government offers a fully priced stop loss for the other 50 percent proportional value, or leaves it to the provincial or local government to facilitate reinsurance protection for the remaining 50 percent. In other words, the decision to provide more subsidies by paying for the other 50 percent of the stop-loss layer offered by the central government would be left to the individual provincial governments. 88 • Global reinsurers would continue to fill in the gaps with reinsurance at the lower levels of stop loss. This type of structure can be formalized with a specific case, just as was presented in Chapter 4. It this case the stop loss would be set at 150 percent for all provinces. This can serve as a base case for comparing different structures. Figure A3.11: Model for Joint Sharing of Catastrophic Financing* 150% loss ratio and Central Government pays for Central Government sells above 50% of catastrophic losses reinsurance on the other 50% of catastrophic losses to either the insurance companies or the provincial government, if they wish to add subsidies 110–150% loss ratio Reinsurance and/or other alternative risk transfer solutions pay losses 75–110% loss ratio Reserves and/or credit pay losses 0–75% loss ratio Premiums pay losses *Between the central government and either provincial insurance company or provincial government. Sources: Authors. In this example, the stop loss for the central government would begin at 150 percent of the losses above gross premium income. For 50 percent of these losses, the central government would provide the reinsurance free of charge, as its subsidy contribution. For the other 50 percent of these losses, the central government would sell the stop-loss reinsurance either to the agricultural insurance company or to the provincial government, which would then offer that portion of the stop loss free of charge to the insurance company. This could be the subsidy level for the provincial insurance company. To the extent that the stop loss is set at less than amounts that are relatively high, these types of subsidies would be far less distorting than a flat subsidy that is a percentage of premium. The stop-loss subsidy would give the insurance company greater incentives to pay more attention to the product designs and underwriting risks, given that it would depend on a global reinsurer to provide reinsurance for the layer of risk below the central-government stop-loss threshold. The Monte Carlo macro model can be used to investigate the potential fiscal implications of a structure whereby the central government provides a free stop loss for every province—for 50 percent of the excess losses above 150 percent—for each of the insurance companies represented by these perfectly balanced portfolios. With these data, the government would be transferring an average equivalent of 2 percent of total premium for such a stop loss. The provincial governments or the provincial crop insurance companies would need to pay an average of 2 percent to cover the other 50 percent. These values are average rates. However, nearly one-half of the provinces would be paying less than 2 percent, and some 37 percent of them would pay 2–5 percent, with the remaining 17 percent paying in excess of 5 percent. Given this model and with full 89 excess-of-loss protection above 150 percent of premium, the average value of a reinsurance layer that pays for losses in excess of 110 percent of premium, with limits at 150 percent, would be about 8.4 percent. As presented in Chapter 4, the recommendation for the central government is to segment the provinces into at least three groupings, and then to provide different stop-loss values based upon the risk segmentation. Table A3.6 shows the average value associated with providing different stop-loss values by province. It also shows the different sorting that was done to determine what stop loss to provide to each province. Finally, these same groupings are mapped in Figure A3.12. There is roughly an equal dispersion of provinces into these three categories. Those in the low category would be provided a free stop loss from the central government at 150 percent; the middle category receives a free stop loss at 175 percent; and the high category receives a free stop loss at 200 percent. For these perfectly distributed portfolios, this segmented stop-loss program represents an average transfer from the central government of about 1 percent per province, as is highlighted by the bold values appearing in Table A3.6. The province insurance company or the provincial government would have to purchase a matching stop loss to cover the other 50 percent of losses in excess of each of these values. Table A3.6: Sorting Provinces by Value of (Actuarially Fair) Stop Loss Sorting Stop Loss at Stop Loss at Stop Loss at Extreme Province 150% 175% 200% Losses Guangxi 1% 0% 0% Low Hebei 0% 0% 0% Low Heilongjiang 1% 0% 0% Low Hubei 1% 0% 0% Low Hunan 0% 0% 0% Low Shaanxi 1% 0% 0% Low Shandong 1% 0% 0% Low Sichuan 1% 0% 0% Low Yunnan 1% 0% 0% Low Zhejiang 1% 0% 0% Low Fujian 2% 1% 0% Mid Gansu 2% 0% 0% Mid Guangdong 2% 1% 0% Mid Guizhou 2% 1% 0% Mid Henan 2% 1% 0% Mid Inner_Mongo 2% 1% 0% Mid Shanxi 2% 1% 0% Mid Xinjiang 2% 1% 0% Mid Anhui 5% 3% 2% High Beijing 3% 2% 1% High Hainan 3% 2% 1% High Jiangsu 4% 2% 1% High Jiangxi 4% 3% 2% High Jilin 5% 3% 1% High Liaoning 5% 2% 1% High Ningxia 3% 3% 2% High Qinghai 7% 5% 3% High Tianjin 2% 2% 1% High Tibet 11% 9% 7% High Sources: Authors. 90 Figure A3.12 Risk Grouping for Different Stop-Loss Values Sources: Authors. The implications of setting different stop-loss values are clear—the average premium transfer is 2 percent for the central government with a flat 150 percent stop loss, and 1 percent with the differential stop loss. The distributions around these two values are presented in Figure A3.13. For the flat stop-loss model at 150 percent, there is a 10 percent chance that the payouts will be in excess of 5.6 percent of premium value. That frequency drops to about 2 percent for the model with differential stop loss. By offering these stop losses, the central government effectively pools the risk across China. 91 Figure A3.13: Differences in Probability Distribution of Central Government Payments under Different Rules for Stop Loss 9% Distribution of Central Government Payments Percent of Distribution with differential stop loss (150; 175; and 200) 6% Distribution of Central Government Payments with only a stop loss of 150 3% 0% 0% 1% 2% 3% 4% 5% 6% 7% 8% 9% 10% Percent of Premium from Central Government Stop Loss Sources: Authors. 3.5. Limitations and Conclusions Results from this annex demonstrate a number of important principles about spreading risk, and general implications of some of the recommendations that are presented in Chapter 4. Still, the macro model developed here has numerous limitations, and significant caution is needed in reaching conclusions about which levels of stop loss the central government should use. To emphasize this point, a macromodel was developed using the average annual loss data from the two-crop models that emerged from the province crop data and that are presented in Table A3.7.(a–z). For this macromodel, the central government would need to set the average stop loss at 250 percent to achieve the same average transfer as the macro model that set all province stop losses at 150 percent. To obtain more precise answers regarding specific thresholds that should be used for the government in setting stop-loss contracts, it will be critical for professionals in China to obtain the best data possible to perform a modeling exercise similar to that which is presented here. 92 Table A3.7.(a–z): Detailed Examination of Losses with Province Crop Data and Province Cause-of-Loss Data Table A3.7.a: Anhui Anhui Correlation —Two Crops and Cause of Loss 83% Rice Wheat Two Crops All Crops (%) Share of Value of All Crops 21 13 34 100 Average Annual Loss 2 5 3 8 Probable Maximum Loss (10-Year Return) 14 19 17 16 Probable Maximum Loss (20-Year Return) 17 26 26 22 Probable Maximum Loss (100-Year Return) 20 49 32 30 Sources: Authors. Table A3.7.b: Fujian Fujian Correlation — Two Crops and Cause of Loss 32% Rice Peanuts Two Crops All Crops (%) Share of Value of All Crops 16 2 18 100 Average Annual Loss 0.6 2.2 0.8 5.6 Probable Maximum Loss (10-Year Return) 14.1 13.2 7.5 10.1 Probable Maximum Loss (20-Year Return) 15.9 15.7 15.3 11.4 Probable Maximum Loss (100-Year Return) 17.2 17.5 16.6 13.2 Sources: Authors. Table A3.7.c: Gansu Gansu Correlation — Two Crops and Cause of Loss 62% Wheat Corn Two Crops All Crops (%) Share of Value of All Crops 12 6 18 100 Average Annual Loss 2.9 4.2 3.4 10.5 Probable Maximum Loss (10-Year Return) 13.9 15.1 10.4 18.4 Probable Maximum Loss (20-Year Return) 16.9 17.9 15.6 19.9 93 Probable Maximum Loss (100-Year Return) 20.6 21.6 21.9 22.4 Sources: Authors. Table A3.7.d: Guangdong Guangdong Correlation — Two Crops and Cause of Loss 28% Rice Peanuts Two Crops All Crops (%) Share of Value of All Crops 20 4 23 100 Average Annual Loss 1.5 6.3 2.2 5.7 Probable Maximum Loss (10-Year Return) 13.9 20.3 15.4 9.5 Probable Maximum Loss (20-Year Return) 15.9 26.7 26.7 10.5 Probable Maximum Loss (100-Year Return) 17.9 46.2 30.1 12.5 Sources: Authors. Table A3.7.e: Guangxi Guangxi Correlation — Two Crops and Cause of Loss 56% Rice Peanuts Two Crops All Crops (%) Share of Value of All Crops 28 4 32 100 Average Annual Loss 1.7 2.3 1.8 6.2 Probable Maximum Loss (10-Year Return) 7.9 10.1 7.1 9.1 Probable Maximum Loss (20-Year Return) 12.7 12.6 8.9 10.0 Probable Maximum Loss (100-Year Return) 14.7 14.5 14.2 11.5 Sources: Authors. Table A3.7.f: Guizhou Guizhou Correlation — Two Crops and Cause of Loss 41% Rice Corn Two Crops All Crops (%) Share of Value of All Crops 21 12 33 100 Average Annual Loss 3.0 2.4 2.8 6.7 Probable Maximum Loss (10-Year Return) 13.7 15.2 9.3 10.7 94 Probable Maximum Loss (20-Year Return) 16.6 17.5 14.9 12.3 Probable Maximum Loss (100-Year Return) 20.1 20.3 18.7 14.7 Sources: Authors. Table A3.7.g: Hebei Hebei Correlation — Two Crops and Cause of Loss 34% Wheat Corn Two Crops All Crops (%) Share of Value of All Crops 14 10 24 100 Average Annual Loss 1.8 2.0 1.8 8.5 Probable Maximum Loss (10-Year Return) 3.9 5.7 5.9 7.9 Probable Maximum Loss (20-Year Return) 5.7 6.9 7.0 8.0 Probable Maximum Loss (100-Year Return) 7.7 7.9 7.9 8.1 Sources: Authors. Table A3.7.h: Heilongjiang Heilongjiang Correlation — Two Crops and Cause of Loss 52% Rice Soybeans Two Crops All Crops (%) Share of Value of All Crops 27 23 50 100 Average Annual Loss 2.2 4.6 3.3 9.3 Probable Maximum Loss (10-Year Return) 13.8 13.8 9.0 13.8 Probable Maximum Loss (20-Year Return) 16.2 16.4 14.4 15.0 Probable Maximum Loss (100-Year Return) 18.6 19.0 18.0 18.6 Sources: Authors. Table A3.7.i: Henan Henan Correlation — Two Crops and Cause of Loss 47% Wheat Peanuts Two Crops All Crops (%) Share of Value of All Crops 21 7 28 100 Average Annual Loss 1.8 3.7 2.3 7.0 95 Probable Maximum Loss (10-Year Return) 14.2 15.2 8.3 11.5 Probable Maximum Loss (20-Year Return) 16.0 18.2 15.0 12.8 Probable Maximum Loss (100-Year Return) 18.0 21.9 17.5 14.7 Sources: Authors. Table A3.7.j: Hubei Hubei Correlation — Two Crops and Cause of Loss 52% Rice Rapeseed Two Crops All Crops (%) Share of Value of All Crops 23 5 28 100 Average Annual Loss 1.0 2.7 1.3 8.5 Probable Maximum Loss (10-Year Return) 14.3 14.9 7.9 13.4 Probable Maximum Loss (20-Year Return) 16.0 17.5 15.1 14.5 Probable Maximum Loss (100-Year Return) 17.7 20.4 17.1 16.4 Sources: Authors. Table A3.7.k: Hunan Hunan Correlation — Two Crops and Cause of Loss 42% Rice Rapeseed Two Crops All Crops (%) Share of Value of All Crops 38 3 41 100 Average Annual Loss 0.6 1.9 0.7 7.8 Probable Maximum Loss (10-Year Return) 3.1 7.4 6.8 9.8 Probable Maximum Loss (20-Year Return) 7.4 8.9 7.4 10.3 Probable Maximum Loss (100-Year Return) 8.1 10.3 8.0 10.5 Sources: Authors. Table A3.7.l: Inner Mongolia Inner Mongolia Correlation — Two Crops and Cause of Loss 70% Corn Soybeans Two Crops All Crops (%) 96 Share of Value of All Crops 19 5 24 100 Average Annual Loss 3.0 6.9 3.9 14.1 Probable Maximum Loss (10-Year Return) 14.2 20.6 15.8 22.2 Probable Maximum Loss (20-Year Return) 16.3 31.3 26.5 23.8 Probable Maximum Loss (100-Year Return) 19.5 51.1 30.7 26.5 Sources: Authors. Table A3.7.m: Jiangsu Jiangsu Correlation — Two Crops and Cause of Loss 47% Rice Wheat Two Crops All Crops (%) Share of Value of All Crops 20 7 28 100 Average Annual Loss 1.3 2.9 1.7 6.9 Probable Maximum Loss (10-Year Return) 14.3 14.5 8.0 12.8 Probable Maximum Loss (20-Year Return) 15.9 17.2 15.1 14.9 Probable Maximum Loss (100-Year Return) 17.6 19.7 17.2 19.3 Sources: Authors. Table A3.7.n: Jiangxi Jiangxi Correlation — Two Crops and Cause of Loss 68% Rice Peanuts Two Crops All Crops (%) Share of Value of All Crops 46 3 48 100 Average Annual Loss 0.9 1.5 0.9 6.0 Probable Maximum Loss (10-Year Return) 3.4 6.1 6.5 7.5 Probable Maximum Loss (20-Year Return) 7.0 7.3 7.3 8.0 Probable Maximum Loss (100-Year Return) 8.2 8.4 8.2 8.5 Sources: Authors. 97 Table A3.7.o: Jilin Correlation — Two Crops and Cause of Jilin 81% Loss Corn Rice Two Crops All Crops (%) Share of Value of All Crops 29 13 43 100 Average Annual Loss 4.4 5.7 4.8 13.8 Probable Maximum Loss (10-Year Return) 13.9 15.4 12.3 22.3 Probable Maximum Loss (20-Year Return) 17.3 18.6 17.3 24.2 Probable Maximum Loss (100-Year Return) 21.8 22.8 25.0 26.0 Sources: Authors. Table A3.7.p: Liaoning Liaoning Correlation — Two Crops and Cause of Loss 88% Corn Rice Two Crops All Crops (%) Share of Value of All Crops 14 11 25 100 Average Annual Loss 6.8 4.2 5.6 12.9 Probable Maximum Loss (10-Year Return) 27.1 20.8 21.5 24.8 Probable Maximum Loss (20-Year Return) 29.8 24.8 25.5 28.1 Probable Maximum Loss (100-Year Return) 32.4 29.4 30.8 32.1 Sources: Authors. Table A3.7.q: Ningxia Ningxia Correlation — Two Crops and Cause of Loss 34% Wheat Corn Two Crops All Crops (%) Share of Value of All Crops 17 15 32 100 Average Annual Loss 3.6 1.6 2.7 8.2 Probable Maximum Loss (10-Year Return) 11.9 12.1 14.3 14.3 Probable Maximum Loss (20-Year Return) 17.0 15.8 18.0 16.0 Probable Maximum Loss (100-Year Return) 19.8 19.6 19.8 18.8 Sources: Authors. 98 Table A3.7.r: Qinghai Qinghai Correlation — Two Crops and Cause of Loss 72% Rapeseed Wheat Two Crops All Crops (%) Share of Value of All Crops 18 14 32 100 Average Annual Loss 2.9 3.0 2.9 11.5 Probable Maximum Loss (10-Year Return) 23.1 19.5 19.3 20.6 Probable Maximum Loss (20-Year Return) 25.3 23.1 22.5 23.6 Probable Maximum Loss (100-Year Return) 26.9 26.4 26.4 26.6 Sources: Authors. Table A3.7.s: Shaanxi Shaanxi Correlation — Two Crops and Cause of Loss 77% Wheat Corn Two Crops All Crops (%) Share of Value of All Crops 13 9 22 100 Average Annual Loss 3.4 3.6 3.5 11.8 Probable Maximum Loss (10-Year Return) 20.5 18.0 18.0 18.2 Probable Maximum Loss (20-Year Return) 23.2 21.5 20.7 19.7 Probable Maximum Loss (100-Year Return) 24.7 24.2 24.0 22.0 Sources: Authors. Table A3.7.t: Shandong Shandong Correlation — Two Crops and Cause of Loss 61% Wheat Corn Two Crops All Crops (%) Share of Value of All Crops 12 8 20 100 Average Annual Loss 2.1 3.1 2.5 7.8 Probable Maximum Loss (10-Year Return) 17.6 16.6 17.3 12.4 Probable Maximum Loss (20-Year Return) 20.4 20.2 20.2 13.6 Probable Maximum Loss (100-Year Return) 23.2 23.0 22.5 15.2 Sources: Authors. 99 Table A3.7.u: Shanxi Shanxi Correlation — Two Crops and Cause of Loss 76% Corn Wheat Two Crops All Crops (%) Share of Value of All Crops 17 10 27 100 Average Annual Loss 4.7 4.7 4.7 14.8 Probable Maximum Loss (10-Year Return) 18.7 16.8 17.5 21.0 Probable Maximum Loss (20-Year Return) 21.5 20.3 20.3 22.2 Probable Maximum Loss (100-Year Return) 23.4 23.0 23.0 23.4 Sources: Authors. Table A3.7.v: Sichuan Sichuan Correlation — Two Crops and Cause of Loss 48% Rice Corn Two Crops All Crops (%) Share of Value of All Crops 22 6 28 100 Average Annual Loss 0.6 3.4 1.2 6.0 Probable Maximum Loss (10-Year Return) 5.6 10.5 11.7 9.0 Probable Maximum Loss (20-Year Return) 10.2 13.5 13.0 10.0 Probable Maximum Loss (100-Year Return) 11.0 16.0 14.8 12.0 Sources: Authors. Table A3.7.w: Tianjin Tianjin Correlation — Two Crops and Cause of Loss 17% Cotton Corn Two Crops All Crops (%) Share of Value of All Crops 11 6 17 100 Average Annual Loss 5.8 5.5 5.7 8.7 Probable Maximum Loss (10-Year Return) 30.7 21.5 22.0 16.2 Probable Maximum Loss (20-Year Return) 35.1 25.9 26.4 19.1 Probable Maximum Loss (100-Year Return) 41.4 31.7 32.2 22.5 Sources: Authors. 100 Table A3.7.x: Tibet Tibet Correlation — Two Crops and Cause of Loss 47% Wheat Rapeseed Two Crops All Crops (%) Share of Value of All Crops 13 7 20 100 Average Annual Loss 1.4 4.3 2.4 6.1 Probable Maximum Loss (10-Year Return) 12.3 16.0 16.9 16.2 Probable Maximum Loss (20-Year Return) 18.9 19.4 19.6 18.9 Probable Maximum Loss (100-Year Return) 21.4 21.9 21.9 21.6 Sources: Authors. Table A3.7.y: Xinjiang Xinjiang Correlation — Two Crops and Cause of Loss 1% Cotton Wheat Two Crops All Crops (%) Share of Value of All Crops 27 10 36 100 Average Annual Loss 2.5 0.7 2.1 3.9 Probable Maximum Loss (10-Year Return) 7.2 6.0 11.5 6.8 Probable Maximum Loss (20-Year Return) 10.0 10.8 12.5 7.5 Probable Maximum Loss (100-Year Return) 12.6 11.8 13.4 8.9 Sources: Authors. Table A3.7.z: Yunnan Yunnan Correlation — Two Crops and Cause of Loss 20% Rice Corn Two Crops All Crops (%) Share of Value of All Crops 21 10 31 100 Average Annual Loss 1.5 1.2 1.4 5.5 Probable Maximum Loss (10-Year Return) 5.5 5.7 10.0 7.9 Probable Maximum Loss (20-Year Return) 9.1 6.6 10.6 8.6 Probable Maximum Loss (100-Year Return) 10.7 10.8 10.9 9.7 Sources: Authors. 101 Annex 4: Legal and Regulatory Framework The purpose of legislation is to give legal effect to a desired policy objective. Given that the government of China is in the early stages of formulating policy, and this review has presented a number of different policy options for consideration, it would be premature to provide detailed recommendations for legislative reform. Lessons can be drawn from international experience, however, where laws deal with specific aspects of agricultural insurance, such as setting up national programs (for example, the U.S. group risk plan) or establishing agricultural insurance pools (as in Turkey, for example). Annex 1 (above)—in particular the case studies in Section 1.4 of that annex—provides evidence of the wide range of agricultural insurance risk-management programs in use throughout the world. Since most have government support and intervention as an underlying objective, it would be a mistake for the government to look to other jurisdictions to provide a legislative model that can be adapted for China, particularly if the government wishes to see a comprehensive agricultural insurance law enacted. Once policy has been agreed to and formulated, then it would make sense to review the legislation of some other jurisdictions to see how they have dealt with specific issues. 4.1. Insurance Law The Insurance Law of the People’s Republic of China is ambiguous concerning agricultural insurance, which is excluded under Article 149 of the Law. Need for Separate Agricultural Insurance Law (Article 149 Issue) Unfortunately, there apparently is a widely held assumption, based on Article 149, that there is no law providing for agricultural insurance in China, and therefore no legal and regulatory framework governing agricultural insurance. This assumption results in uncertainty in the marketplace and difficulties for insurers, some of whom have been successfully challenged in the courts. It is not clear, however, whether this assumption of lack of legal provision for agricultural insurance in China is justified. First, Article 149 conflicts with Article 3, which provides that the Insurance Law shall apply to all insurance activities within the territory of the People’s Republic of China. Second, it is clear that even agricultural insurance companies are currently subject to approval and supervision by the China Insurance Regulatory Commission (CIRC). The only legal basis for this requirement is the current Insurance Law. Therefore, the original purpose of Article 149 may have been to enable the enactment of an agricultural insurance law that would apply to agricultural insurance in addition to the current Insurance Law, which seems a more reasonable interpretation, since much of the Insurance Law is appropriate for agricultural insurance as well as other forms of insurance. This approach should be recommended to the government. Nevertheless, given the confusion in the insurance market, it is important that this matter is resolved as soon as possible. 102 High-Level Review of Insurance Law Given the ongoing revision of the Insurance Law, a high-level review of the Law has been undertaken to determine whether there appear to be any issues that could have serious impact on the development of agricultural insurance. (It should be noted that a full legal analysis of the Insurance Law would be a major task that is beyond the scope of this project. It also would require access to other documents and materials that were unavailable for this review, including all the regulations that have been promulgated under the Law. A full review would require a detailed study of the interface of this Law with other legislation, which also is well beyond the scope of this project. Finally, there was no review of Article 3 of the Law, which relates to life insurance.) Under the circumstances, no attempt has been made to identify all the relevant issues, or even all the critical issues. The review is expected to provide the government with some areas to consider in the current revision exercise. It is believed that the current revision of the Insurance Law is extremely timely. The Law was enacted in 1995—the year after the International Association of Insurance Supervisors (IAIS) was established—and has been subject to only one relatively minor set of amendments since. It is not surprising that the Insurance Law is now rather out of date and does not comply with a number of international standards promulgated by the IAIS since 1995 for the regulation and supervision of insurance business. A significant number of regulations have been made under the Law, which go some way toward implementing international standards. However, it is not clear to us that the Insurance Law enables all of these regulations to be made. To the extent that the Insurance Law will apply to insurance companies offering agricultural insurance, and to the extent that it will govern agricultural insurance products and contracts, it is an essential foundation stone for the development of agricultural insurance in China. It is therefore vital that the Law is comprehensive, up to date, and compliant with international standards, as far as they are applicable to China. The government is encouraged to undertake a comprehensive “root and branchâ€? review of the Insurance Law. In this review, it is recommended that the government bear in mind the recommendations that made in Chapter 4 of the report, concerning the advisability of framework legislation. Some detailed and specific provisions in the Insurance Law would serve much better if placed in regulations—for example, the specification of time periods, detailed provisions relating to the contents of an insurance contract, specific capitalization provisions, and specific amounts relating to other financial resourcing provisions. A number of general legal and regulatory issues relating to index-based insurance are considered later in this annex. However, a few provisions in the Insurance Law could have implications for the introduction of index-based insurance: • As noted in the main report, Article 2 of the Law provides, regarding general insurance business, that insurance is a contract of indemnity. This provision may be considered an impediment to index-based insurance, given that payment is based on an index, not on actual loss, under an index-based contract. 103 • Article 11 of the Law, in common with the laws of many jurisdictions, requires an insured applicant to have an insurable interest in the subject matter insured. In the event that there is no insurable interest, the insurance contract is not binding. Unless this article is amended, great care will have to be taken to ensure that index-based contracts are issued only to insured farmers that have a demonstrable insurable interest in the subject matter of the insurance contract. Insurable interest is defined as an “interestâ€? over the subject matter of the insurance, but the Insurance Law does not provide any definition of “interestâ€? or any guidance as to what constitutes an interest. However, this should not be a problem in the majority of cases, provided that the policy is taken by the farmer or household that owns the crops. • Article 21 requires the applicant, insured, or beneficiary “at the time of being aware of the occurrence of an insurance accident … to notify the insurer in a timely manner.â€? The term “insurance accidentâ€? is taken to mean the event that triggers a claim. In the case of an index-based insurance contract, a claim will be triggered by the value of an underlying index, which generally will be known to the insurance company before the insured. Given the nature of an index-based contract, it is considered inappropriate for the insured to be required to notify of a claim. Instead, the onus should be on the insurance company to advise the insured that the value of the index has triggered a claim. It does not appear to be possible for the insurance contract to override this article. The recommendation, therefore, is that consideration be given to amending the article with respect to index- based contracts. Articles 24, 25, and 26 also would need similar amendments. • Article 22 is also considered inappropriate for an index-based insurance contract. In effect, this article places the onus for providing proof of loss on the insured. Although this requirement is appropriate for an insurance contract that indemnifies the insured against a specific loss or damage, it is inappropriate for an index-based contract. Evidence that the value of the index has triggered a claim will be more readily available to the insurance company than to the insured. Again, since it does not appear to be possible for the insurance contract to override this article, it is recommended that consideration be given to amending the article with respect to index-based contracts. • Article 23 imposes a strict 10-day time limit on an insurance company for the payment of a claim. This provision may be appropriate for contracts of indemnity, which are settled on a claim by claim basis, but not necessarily for index-based contracts. This is a good example of a detailed provision which is really inappropriate for primary legislation. • Article 38 appears to give the insured a right to terminate an insurance contract after it has commenced, in which case the premium, less that for the period to termination, is to be returned to the insured. The language is not clear, but if this interpretation is correct, it would destroy the basis for an index-based contract by permitting the insured to speculate on the probability that the index value has been reached—after the contract has been commenced—and recoup the unused part of his premium. Under an agricultural weather- index insurance contract, it is important that the selling season terminates before the farmer has sufficient knowledge to make an informed judgment on whether the index value is likely to be reached. This precaution is essential to avoid adverse selection. This article, if it gives an entitlement to terminate the insurance, will also encourage adverse 104 selection against the insurer. For example, if the underlying index is rainfall, and the insured event is flood, farmers will usually have a very good idea, prior to the conclusion of the period of coverage, whether there is likely to be excessive rainfall. They would be able to use this knowledge to terminate the contract, thus depriving the insurance company of the profits that it needs to build up reserves for future years. It is recommended that this article be reviewed. • Article 41 provides that an insured must minimize his losses if the insured event occurs, and that necessary and reasonable expenses incurred in minimizing the loss must be paid by the insurer. Given that under an index-based contract payment is against the index and not on assessed losses, this article is wholly inappropriate for an index-based contract. It is important that consideration should be given to disapplying this article with respect to index-based contracts. Article 42 (partial losses) also should be reviewed, because it may cause problems with respect to an index-based contract. • Article 48 provides that “necessary and reasonable expensesâ€? paid by the insurer and insured for the purposes of investigating and determining the nature and cause of an insurance accident shall be borne by the insurer. This is considered inappropriate for an index-based contract. The government might consider revising several other articles of the Insurance Law: • It is unclear whether the term “insuranceâ€? includes reinsurance, but Article 28 suggests that it does, in most jurisdictions. If so, the wording of Articles 5 and 6 should be reconsidered in light of the practice of allowing insurance companies to use foreign reinsurance companies where there is insufficient Chinese reinsurance capacity. • Article 36 is difficult to follow in the English translation. It appears to state that if, within the period of the insurance contract, the risk to the property insured increases, the insured must notify the insurer, and the insurer has the right to increase the premium or to terminate the contract. The article goes on to say that if the insured does not comply with this obligation, the insurer is not liable to indemnify the insured under the contract. If this is the correct interpretation, this article appears inappropriate for an agricultural insurance contract, where the risk may well increase during the season. Although no conclusion has been reached concerning this article, the article should be reviewed. • Article 37 (return of portion of premium) is also difficult to follow in translation. As a requirement, it would not work for agricultural insurance. However, it is noted that the article does permit the insurance contract to override its provisions. It is important that all agricultural insurance contracts provide such an override. • Article 106 requires the supervisory authority (now CIRC) to determine “basic insurance clauses and premium rates for major coverage of commercial insurance.â€? It is understood that this requirement is no longer being followed in practice, but it is recommended that the article be reviewed and amended as appropriate. The Government may wish to consider a number of other changes or additions to the Insurance Law to accommodate agricultural insurance. In particular, the reserving 105 provisions in Article 93 for general insurance companies (which are very basic and should be reviewed) may not be appropriate for agricultural insurance business, where the season may not equate to a full year. There are no provisions enabling exemptions to be made with respect to microinsurance. 4.2. Regulatory Issues A brief overview meeting with CIRC helped clarify some of the key regulatory issues. It was not sufficient, however, to support a comprehensive review of the regulatory issues concerning agricultural insurance. Such a review would have required a considerable time commitment as well as translation of substantial numbers of regulations, insurance company returns, and other documents. However, analysis of the insurance market and the insurance companies, and in particular the agricultural risk assessment undertaken for this review, shows that there are significant problems with respect to the mismatch between the agricultural risks carried by insurance companies and their capitalization and financial resources. This situation is compounded by factors (as discussed in the report) such as the concentration of risk (with respect to crops and region), the types of policies issued (MPCI and index), and the lack of access to reinsurance with respect to some of the risks carried. This is clearly a key issue for the prudential regulation of insurance companies, and it must be addressed. For insurance companies that are not specialists in agricultural insurance, significant agricultural losses could have severe impact on the ability to pay claims in other areas. A number of other issues that have regulatory implications have also been addressed. For example, in paragraphs 2.77 to 2.79 of the report it is noted that a number of companies are not adjusting their rates on an actuarial basis, and that in general rates are inadequate to cover the long-term average claims, let alone to generate adequate levels of return on equity. Although this most likely will eventually lead to a withdrawal from the market, in the event of an extreme event, it could lead to a failure of the insurer. 4.3. Legal and Regulatory Issues—Index-Based Insurance An index-based insurance product is a form of index-based risk-transfer product, which in most cases may be written as an insurance contract or as a capital markets product, such as a derivative. It is necessary for CIRC and the capital markets regulator in China to work together to ensure that index-based risk-transfer products are properly classified. Legal Differences between Derivatives and Insurance Products There are a number of differences between a derivative and an index-based insurance product. In its response to the white paper, “Weather Financial Instruments,â€? which was to be published by the U.S. National Association of Insurance Commissioners (NAIC), 12 12 The white paper was never actually published. 106 the Weather Risk Management Association (WRMA) argued that there is a spectrum available of commercial risk-transfer products, which WRMA describes as contingent commercial contracts. At one end of the spectrum is the traditional insurance product and at the other end are various types of capital market products, including derivative contracts. The most common capital market product used to transfer agricultural risk is the weather derivative. Although insurance products and derivatives may be considered to occupy different positions along the spectrum of contingent commercial contracts, and although they may be considered to have similar commercial and economic features, from a legal and regulatory perspective, they are entirely different products. Insurance The Insurance Law defines insurance in terms similar to those of many other jurisdictions. In essence, an insurance contract has the following features: • One party (the insured) pays a sum of money (the premium) to the other party (the insurer). • In return for the premium, the insurer agrees to accept the risk of an uncertain event occurring at a future time (this is the effect, but not the wording, contained in Article 2 of the Insurance Law). • The insured must have an insurable interest in the subject matter of the insurance. • The insurer agrees to indemnify the insured for loss or damage that the insured sustains upon occurrence of the insured event. • The insurance contract has a specified period (the term). The rationale for the requirements of insurable interest and indemnification or compensation for loss are discussed further later in this annex. Derivatives A derivative is commonly understood to be a financial contract or instrument that “derivesâ€? its value from some other underlying asset, rate, or index (Braddock, 1997). Derivatives may be exchange-traded or traded privately—the “over-the-counterâ€? (OTC). When used as part of a risk-management strategy, derivatives will typically take the form of forward contracts and options. The parties to an OTC derivative are relatively free to agree on the terms of the contract, although the contract will often be based on a standard model. It is not known whether standard contracts such as the ISDA Master Agreement 13 are used for derivative contracts in China. Weather derivatives are financial instruments by which the purchaser seeks to hedge the risk to yields or revenue streams from weather events. The value of a weather derivative 13 The ISDA Master Agreement is a model agreement published by the International Swaps and Derivatives Association. 107 derives from the underlying weather index, which could take a variety of different forms (including temperature, rainfall, snowfall, wind speed, or sea surface temperature, or a non-weather index, such as area yields or livestock loss). Principal Differences between an Insurance Contract and a Derivative The terms of a derivative contract include the payment of a sum of money from one party to another, the transfer of risk, and the stipulation of a definite contract term, therefore sharing a number of the elements of an insurance contract set out above. However, there are two important legal differences: • The requirement (in the case of an insurance contract) that the insured person has an “insurable interest;â€? in the property insured; and • The principle that an insurance contract is designed to indemnify for loss. There is a third difference between insurance and derivatives: Derivatives are usually tradable, either on a recognized exchange or privately, but insurance contracts are not. As insurance is linked to an insurable interest, an insurance contract usually can be assigned only if the insured property is transferred, although there is no reason an insured person cannot assign the proceeds of his insurance contract. Insurable interest As already indicated, the Insurance Law requires the insured to have an insurable interest in the subject matter of the insurance. The concept of insurable interest with respect to general insurance is complex, and a full consideration is beyond the scope of this project. Furthermore, the legal meaning of insurable interest differs from one jurisdiction to another. As noted earlier, the Insurance Law defines insurable interest in relation to an interest, but it does not go on to define what an “interestâ€? is. Consider an example from another jurisdiction: An English case (decided in 1806 but still representative of English law) determined, in effect, that to have an insurable interest in property, a person must have an existing right to an interest in the property or a right under contract. A person will also have an insurable interest in property if that person is under any legal liability with respect to that property. The definition of insurable interest is wider than this in some jurisdictions and may include, for example, a contingent interest in property. (This is further discussed below, under “Legal Issues.â€?) In contrast, there is no requirement for a party to a derivative contract to have any interest in the subject matter underlying the derivative, although that party may have such an interest. Loss The Insurance Law, in common with other jurisdictions, provides that the insured under an insurance contract may not claim under the contract unless the insured has sustained a 108 loss. It is important to note that the issue of whether a loss has been sustained is a different issue from that of the amount that may be claimed under an insurance contract in respect to the loss. Insurance contracts often provide for an insured person to be indemnified in respect to loss, but it is not necessarily a requirement. The relationship between the actual loss and the amount that may be claimed is obviously of critical importance in the case of index insurance, where an index is used as a proxy for loss. This issue will therefore be covered more fully under “Legal Issues,â€? below. With respect to a derivative contract, there is no requirement that a party receiving payment under the contract has suffered a loss. The “Gaming Riskâ€? Insurance contracts and derivatives are both types of contract that impose contingent payment obligations on one party. Gaming contracts (bets and wagers) are also contracts that impose contingent payment obligations. In some jurisdictions, index-based risk- transfer contracts (whether insurance or derivative) have been classified by the law or the regulator as a gaming contract, which could result in the contract being declared void by a court at some future time. (This point is discussed later in this annex.) Historically, the insurable interest requirement and the requirement that an insurance contract should indemnify or compensate for loss or damage developed to ensure that insurance contracts were not used as a form of gaming. Without these requirements, an insurance contract is, in effect, a wager on the occurrence of the insured event. Such wagers introduced extreme moral hazard and, in the case of life insurance, were even an inducement to murder! It was for this reason that the United Kingdom first enacted legislation some 260 years ago to introduce the requirement for an insurable interest with respect to marine insurance and, a few years later, with respect to life insurance. Even earlier, the courts in the United Kingdom had established a requirement for an insurable interest with regard to buildings insurance. Public policy considerations have led many jurisdictions to enact legislation that makes gaming contracts void and unenforceable, unless entered into within a defined regulatory framework. Such legislation usually contains certain exemptions that include insurance contracts. In many, but not all, jurisdictions there are also specific exemptions for financial contracts such as derivatives. However, these statutes are not always fully effective, a matter covered in more detail later in this section of the annex. Not considered herein are gaming laws in China. It is possible that this is not an issue, but consideration should be given to whether China’s gaming laws could have an impact on index-based insurance contracts. Regulatory Differences between Derivatives and Insurance Products Given the differences described above, insurance and derivative products are subject to entirely different regulatory regimes. 109 As in most, if not all jurisdictions, China’s insurance industry is highly regulated. Insurance regulation usually has two principal objectives: • Reducing systemic risk (which means, in a regulatory context, the risk that a collapse of an insurer will threaten the stability of the financial markets generally); and • Protecting policyholders. Given that the insured policyholder must rely on, and trust, the insurer to meet his claim should the insured event take place at some time in the future, it is particularly important to minimize the risk of the insurer’s not being in a position to meet the claim. Basic principles for insurance supervision promulgated by the IAIS are contained in its “Insurance Core Principles and Methodology,â€? published in October 2003. However, a number of other principles, standards, and guidance papers have also been produced on numerous topics, including: • Conduct of insurance business; • Capital adequacy and solvency; • Supervision of reinsurers; • Licensing; • Derivatives; and • Insurance regulation and supervision in emerging market economies. An examination of the IAIS principles, standards, and guidance papers is beyond the scope of this project, but they will be well known to CIRC, and none covers index-based insurance specifically. However, they must be kept in mind when designing an index insurance product, because such products will be subject to the same standards of supervision and regulation as any other general or short-term insurance product. Derivatives When considering the introduction of an agricultural risk-transfer product, it may be appropriate to consider whether the product should be an insurance product or a derivative. Although the legal framework for the regulation of capital markets (if any) has not been considered, it is most unlikely that it will offer the same level of protection as it does with respect to insurance. The reason for this is that regulations for capital markets are usually intended for sophisticated and knowledgeable market participants, such as banks and other financial institutions, hedge funds, and large institutions. Insofar as the market participants (such as banks, insurance companies, and funds) are themselves regulated, their use of derivatives will be subject to prudential supervision by the responsible regulator. However, where the market players are not subject to any form of prudential supervision, the sale and purchase of derivatives is often unregulated, unless covered by investor protection legislation. 110 Index risk-transfer contracts at the farm level It is unlikely that the average small farmer would have a detailed knowledge of complex financial products, including derivatives. It is most unlikely, therefore, that the farmer would fall into any category of “sophisticated and knowledgeable market participant.â€? Small farmers may be able to match or beat an insurance company when assessing the risks they face, but they will not have the knowledge of the financial markets necessary to use derivatives. If an agricultural risk-transfer product falls outside the regulatory framework, the farmer would not have the benefit of any regulatory protection. Even where weather derivatives do fall within a regulatory regime, the regime will not include the strict financial resourcing requirements applicable to insurance companies, nor is their sale likely to be subject to market conduct rules equivalent to those that international standards require to apply to the sale of insurance. It is advisable that index-based contracts not be offered to farmers as a risk-transfer mechanism unless they are insurance contracts. To permit otherwise would result in farmers not having the benefit of the regulatory regime established by the Insurance Law and the various regulations made under it. Capital market products, such as weather derivatives, are inappropriate risk-transfer products at the small-farm level. Of course, this does not suggest that capital market products, including weather derivatives, may not be used in appropriate circumstances as a risk-management tool by insurers and reinsurers doing business within the agricultural sector—or by lenders, for example, to hedge weather-related portfolio default. Legal Issues As discussed above, it is essential to bear in mind the differences between index insurance and derivatives when designing an index insurance product. If the product is not well designed, there is a risk that it could be drafted as a capital markets product. China is not alone in having an Insurance Law that does not contemplate index insurance. 14 Already provided herein is a list of the Law’s articles that may hinder the development of an index-based insurance market. If the argument made above with respect to the unsuitability of weather derivatives at the small-farm level is accepted, it is important to keep in mind the essential elements of an insurance contract when designing an index insurance product. One of the principle advantages of an index insurance contract is that, as payment is made against an index, the costs entailed in assessing individual losses are eliminated. This should reduce the cost of the product, making it a more affordable risk-transfer 14 Although it is interesting to note that Mongolia’s new insurance law (brought into force July 2005) specifically provides that an insurance contract can be index-based. 111 mechanism for small farmers. However, this very advantage introduces a risk that the product may not be classified as insurance on the grounds that: • There is no requirement on the insured to prove actual loss; and • The insured may receive compensation that is more, or less, than his actual loss. Because index insurance is such a new concept, it is not clear how great this risk actually is. Contracts of property insurance are usually contracts of indemnity. Under such contracts, the insurer promises to indemnify the insured against any loss caused by the occurrence of the insured event, which clearly requires the insurer to compensate the insured only for the actual loss that the insured has sustained. It has to be accepted that it is difficult to argue that an index contract is an indemnity contract. As already noted, the Insurance Law, in translation, only refers to indemnity insurance with respect to non-life insurance. It has been long established under English law, and under the law of some other common law countries, that parties are free to opt out of this contract. English law recognizes an alternative type of insurance contract known as the “valued policy.â€? Article 39 of the Insurance Law also appears to recognize the concept of “valued policies.â€? It is considered that a strong argument can be made that an index contract should be regarded as having some characteristics of a valued policy. In a valued insurance contract, the parties to the contract agree in advance on the value to be placed on the insured property, and therefore on the sum payable in the event of its loss. Under English law, the insurer and the insured are bound by the value they placed on the property (in the absence of fraud or special circumstances that invalidate the agreement), and the agreed value must be paid by the insurer to the insured, even if it is greater than his actual loss. The Insurance Law does not currently include this detail, but the government may wish to consider it in the revision exercise. Marine policies are, and traditionally have been, written as valued policies. It is understood that this practice developed with respect to marine insurance to save the expenses involved in settling the amount of the actual loss—a similar rationale to that of index insurance. However, valued insurance contracts are also used with respect to non-marine insurance, for example: • Where the subject matter of the insurance is difficult to value, such as a work of art; and • Where the value of the insured property fluctuates frequently. For this review, no information is available regarding the status of valued policies generally in civil law jurisdictions, such as China. However, an index contract differs from the traditional valued insurance contract in two ways. First, it is usual for the value to be an agreed sum. Second, the agreed value is paid only in the case of a total loss. In the event that there is only a partial loss, the insurer is liable to pay only that proportion of the agreed value that is equal to the proportion of the loss, which is specifically provided for in Article 39 of the Insurance Law. 112 As already stated, under a valued policy, the parties agree on the value of the insured property—the payment to be made in the event of a total loss. Under an index contract, the index serves not just to trigger the insurer’s liability, but also as the parties’ pre- estimate of the insured’s individual loss, which will vary according to the value of the underlying index. This extends the accepted definition of a valued policy. However, there seems to be no reason in principle that the parties—if they can agree to the value of property on the basis of a total loss—cannot also agree on the value of partial losses calculated in accordance with an agreed index. If the parties agree on a value that is manifestly excessive, however, the contract resembles a gaming contract more than an insurance contract. Therefore, in the case of an agricultural index insurance contract, it is suggested that the insured farmer not be able to insure a value greater than the value of his crop, animals, and so on, unless there is clear justification, such as likely economic loss. Further, the values insured in index insurance programs are frequently the costs of growing the crop, which is a lesser figure than the expected value of the revenue. It is also important that the agreed value is a genuine pre- estimate of the agreed loss, which will depend in part upon how good a proxy the underlying index is. It should be noted that there is a difference between an insurance contract that provides for the maximum sum insured (the ceiling on the amount of the actual loss that may be claimed), which is still an indemnity contract, and a valued policy. For example, when drafting a valued contract, care must be taken to ensure that that the contract is not written as a contract with a maximum insured sum. There are a number of reported English cases where a contract has been held by the court to be an indemnity contract, not a valued contract, because the agreed amount was held to the maximum sum insured, not to the agreed value of the property insured. Given the wording of Article 39 of the Insurance Law, this appears to be the case in China too. The laws of many civil law jurisdictions provide that payment made under an insurance contract not exceed the amount of the loss sustained by the insured. Such a provision could be construed as requiring all insurance contracts to be indemnity contracts. Given the greater potential for basis risk under an index-based contract, this could preclude index-based insurance. Although the Insurance Law does not contain an express provision to this effect, concern is already noted that Article 2 (which defines non-life insurance as an indemnity contract) could have this effect. All insurance products have some basis risk. For example, different loss assessors will place different values on the same loss. The market value of insured products can fluctuate, sometimes significantly. A typical “new-for-oldâ€? property insurance is a good example of a traditional insurance product that contains significant embedded basis risk. If the underlying index can be demonstrated to be a good proxy for the loss, it can be argued that the basis risk embedded in the index-based insurance product is no more or less than other typical insurance products. 113 Insurable Interest An index-based insurance contract should be drafted so as to clearly include the element of insurable interest. In the case of an indemnity policy, insurable interest is implied, because if the insured does not have an insurable interest, there will be no loss to indemnify. However, the circumstances are different in the case of an index-based insurance contract, which arguably should be regarded as analogous to a valued policy. However, in both cases—an index-based insurance contract and a traditional valued insurance contract—the insured person must have an insurable interest and must be able to prove that interest, although it is a legal requirement of an insurance contract. In the context of agricultural insurance, however, whether index-based or not, there may be complications in establishing insurable interest. For example, is a herder entitled to take out an insurance contract for animals that the herder looks after on behalf of somebody else? Is a farmer or grower entitled to take out an insurance contract for crops grown on land that does not belong to that farmer? Consideration should be given to whether these situations may be problems in China. It is suggested that in all agricultural index insurance contracts, the issue of insurable interest may be covered by requiring the insured to declare his interest in the agricultural product or animals to be insured. Of course, if the farmer makes a false declaration and is subsequently found not to have an insurable interest, this will still result in the contract not being an insurance contract, but at least the issue will have been brought to the attention of the farmer, thus minimizing the legal risk. The real risk is more practical. It is that farmers will not understand the concept of insurable interest. This could result in an insurer voiding insurance contracts on the basis that there is no insurable interest, which would damage the credibility of the product. It is therefore important that any project to introduce index-based insurance includes provision for educating policyholders and potential policyholders on the consequences of entering into an index-based contract without having an insurable interest. This requirement should also be clearly stated in the policy documentation. Regulatory Issues Although by no means a new product in the international marketplace, index-based insurance is a relatively new insurance product in China. CIRC will have to consider carefully what criteria is should apply to an application by an insurance company for the introduction of an index-based product. It is suggested that CIRC may wish to consider the following issues, among others: • Is the proposed index-based insurance product properly an insurance product under the Insurance Law? CIRC will need to be satisfied, for example, that there is an insurable risk and that monies paid under the contract represent compensation for the insured’s loss. 114 • Is CIRC reasonably confident that the product has been properly rated and that it will be sufficiently profitable to enable the insurance company to build up adequate reserves? The setting of premiums for traditional products requires considerable expertise and good data, but as is suggested in the main report, the rating of index-based products usually requires excellent historical data and the application of sound actuarial principles. • CIRC will probably want to be reasonably satisfied that sufficient research has been undertaken into historical data to justify the use of the selected index: • The data used to calculate the index is reliable and accurate; • The data is verifiable; • The body or person responsible for the data is trusted and independent; and • The insurers’ underwriters have sufficient knowledge and experience to rate the product. • Given that index-based insurance is a relatively new product, CIRC should probably be concerned as to whether the insurance companies offering the product have sufficient management and technical skills to deploy it. • CIRC will need to understand the risk profile of the index-based product. It should be concerned that, even if historical weather patterns have been fully analyzed, other factors such as changing weather patterns, unusual weather events, and microclimates may make the risk profile difficult to assess. • CIRC should probably understand the risk-management strategy of the insurance company proposing to offer index-based insurance. If catastrophic risk is to be insured, what arrangements will be made for managing the catastrophic risk? In particular, will the insurance company offering the product be able to obtain reinsurance coverage in the international marketplace (in the absence of any program whereby the national or provincial government provides a stop loss). • Given that the product is so different from traditional insurance, there is a danger that farmers purchasing index-based insurance will not fully understand it. CIRC should therefore ensure effective and adequate supervision of the insurance company’s market conduct, because misrepresenting the product could seriously damage market confidence and would be likely to result in poor acceptance of the product in future years. • CIRC should be concerned that, if an insurer should fail, for whatever reason, failure by the insurer to pay claims under the index-based contract could seriously damage market confidence in the index-based product. • CIRC will also need to be satisfied that the insurance company has sufficient resources to offer the product, and to pay potential claims. If not, the insurance company’s resources could be depleted to the detriment of its ability to meet claims made under their other lines of business. 115 • CIRC will also need to consider whether the current regulations governing capitalization, solvency, and reserving are adequate for index-based insurance. • CIRC should determine the degree of basis risk. If the basis risk is too high, it may result in poor acceptance of index-based insurance by farmers in the medium to long term. The basis risk will depend upon the suitability of the selected index or indexes. 116 Annex 5: Index-Based Product Development This annex explores the feasibility of index-based insurance products for China. The early sections provide a description of index insurance, key advantages and challenges of index insurance, and a description of international experiences in index insurance. The last four sections analyze the potential role of index insurance products in each of the four provinces visited. 5.1. Index Insurance: Background and Product Description The concept of index insurance, as applied to agriculture, evolved during the 1990s, because models for agricultural insurance in developed countries could not readily be transferred to many developing countries. Experience shows that yield-based (multiple- peril) crop insurance is subject to moral hazard, adverse selection, and high administrative cost, even in countries with large farm size and high technology, such as the United States. 15 Yield-based crop insurance not only covers hazards beyond the control of farmers (such as weather events), but it also covers risks of poor farm management practices, which may be partially or fully under the farmer’s control. Further, covariate risks, such as drought, give rise to major aggregate financial exposures that differ from hail insurance, where damage is localized and can be managed within a balanced portfolio of insured farms. Index insurance offered the possibility of overcoming many of the shortcomings of MPCI, and of addressing highly covariate risks, with simplified loss assessment, lower delivery costs, and easier access to small farmers. The difficult experience with MPCI has to be balanced against a successful experience with some specific named-peril crop insurance programs, notably hail insurance, which continue to function successfully in many countries—in the private insurance sector and typically without government support. Decisions as to the most appropriate crop insurance product are complex, and depend on crop and climatic characteristics, farm size and rural organization, linkages between insurance and rural services, and the role of public and private sectors. Internationally, there is very limited experience in index insurance, compared to a long history of traditional insurance products for crops and livestock; so, index insurance is still considered to be developmental, but fast emerging. Despite substantial research into index insurance, there are relatively few concrete examples of operational programs that have multiplied beyond the pilot stage. Features of Index Insurance In traditional crop insurance, individual farm yields or damage are measured in the field, or at harvest time, to calculate the indemnity to be paid to that farmer. (See Box A5.1.) In 15 Hazell, P. B. R. “The Appropriate Role of Agricultural Insurance in Developing Countries.â€? Journal of International Development 4 (1992): 567-581. 117 contrast, index insurance relies on the measurement of an objective and independent parameter, which is highly correlated with the actual loss incurred by a farmer. Measurements such as rainfall or temperature are used as a proxy for such yield loss. Under parametric index insurance, payouts are based solely on the measurement of a particular parameter (for example, of rainfall at a named meteorological station) according to an agreed payout scale (established in the insurance policy) related to the rainfall actually recorded at a specific meteorological station. Under aggregate index insurance, payouts are based on an index developed from the aggregated statistics of farm production or yield in specified districts (for example, area yield statistics for crops, or mortality index for livestock). Box A5.1: Main Types of Crop Insurance Traditional Damage-based indemnity insurance (named-peril crop insurance) is calculated by measuring the percentage damage in the field, soon after the damage occurs. The percentage damage measured in the field, less a deductible expressed as a percentage, is applied to the preagreed sum insured. The sum insured may be based on production costs, or on the expected revenue. Where damage cannot be measured accurately immediately after the loss, the assessment may be deferred until later in the crop season. Damage- based indemnity insurance is best known for hail, but is also used for frost and excessive rainfall. It is not feasible to insure drought using direct percentage-damage field assessment, although yield measurements can be converted to a percentage loss. Yield-based crop insurance (MPCI) establishes an insured yield (tonnes per hectare) as a percentage of the historical average yield of the insured farmer. The insured yield is typically 50–70 percent of the average yield of the farmer. If the final actual yield is less than the insured yield, the claim is calculated as the difference between actual yield and insured yield, multiplied by a preagreed value of sum insured per unit of yield. MPCI typically covers all possible causes of yield reduction, because causes of loss are difficult to identify and differentiate. MPCI is able to cover drought risk, but as a product it has many problematic features. Index Weather index insurance makes the payout to the farmer based on measurements of a weather parameter made at a nominated weather station. The amount of the payout is calculated from a preagreed scale according to the occurrence of the weather parameter, measured over defined period(s) of time. Area yield index insurance makes the payout to the farmer based on average yield of a district or region. The insured yield is established as a percentage of the average yield of the district or region. If the district or regional actual yield falls below the district or regional insured yield, all insured farmers in that district or region are paid on a scale according to the shortfall of actual yield below the insured yield. There is no measurement of individual farmer yields. This type of index is highly dependent on the availability and integrity of historical yield data at a district or regional level. Sources: Authors. 118 Weather-index insurance includes the following features: • The index is based on a weather parameter (for example, rainfall) or more than one parameter (such as rainfall and temperature); • The index is typically designed for a specific crop type; • The period of start and end of coverage is specified, targeting a specific period when the crop is vulnerable to the specified weather hazard; • The period of measurement may be a whole crop cycle, or it may be divided into several separate phases during the crop cycle; • The measurements are made at a nominated weather station, typically an official weather station of the national weather service; • The index starts to pay out when the weather parameter exceeds or falls below a specified level (the “thresholdâ€?, or “strikeâ€?) 16; • The index pays out an increasing amount (for example, dollars per millimeter of rain, or per degree of temperature) as the deviation in the weather parameter increases (the “incrementâ€? or “tickâ€?); • The index pays out a maximum amount at a specified level of the weather parameter (the “limitâ€? or “exitâ€?). An example of a payout structure for rainfall deficit is shown in Figure A5.1. In this instance the index payout threshold is 100 millimeters of rainfall, falling in a specified period, and the payout limit is reached when rainfall falls to 50 millimeters of rainfall. Figure A5.1: Payout Structure for a Hypothetical Rainfall Contract Sources: Authors. 16 Index insurance has its history in the derivative markets, where terminology differs from that commonly used in the insurance market. This annex adopts the insurance market terminology of “threshold,â€? “increment,â€? and “limit.â€? 119 The properties of an index required from the random variable are that it must be observable and easily measured, objective, transparent, independently verifiable, reported in a timely manner, and consistent over time. The most common example of parametric index insurance is weather-index insurance. New indexes under development include those developed using satellite imagery, or remote sensing index, principally to establish vegetation growth as an indicative measure of crop yield or pasture production. This type of index is largely experimental at present. Area-yield-index insurance includes the following features: • The index is based on the average yield of a particular crop type, measured over a specified unit area of insurance: • The unit area of insurance is determined by the availability of officially recorded historical yield data at a regional level (for example, the unit area could be a district or region, as opposed to a province). The intent is that the yield outcome of all farmers located in the unit area be sufficiently homogenous for individual farmer yields to be closely correlated with the area yield. The unit area must be sufficiently large that no individual farmer can influence the area yield, and therefore the payout. • An index is created with a threshold yield below the average yield (for example, 80 percent of average yield). A payout scale, in increments of yield shortfall, is calculated, subject to a limit, in a similar structure as for weather-index insurance. When actual average regional yields fall below the threshold yield, payouts are made according to the agreed scale of actual regional-yield shortfall below the threshold yields. • An important feature of yield-index insurance is that there is no field loss assessment at the individual farmer level. However, assessment of actual regional yield is reliant on an accurate and consistent sampling of actual yields. All farmers are paid according the regional-yield result, which has the positive feature (compared to MPCI insurance) that farmers still have the incentive to maximize their own individual production. • Area-yield insurance is fully dependent on high-quality historical yield data. Because the payout is based on regional yield, which is influenced by multiple variables, it can also be influenced by factors other than weather. For example, at times of low product price, fewer inputs may be applied by farmers, with consequent regional-yield reduction. As with weather-index insurance, area-yield index insurance is only feasible for crops and regions that are subject to highly correlated catastrophe risks, such as drought. Payouts will not reflect localized perils such as hail, which may affect individual farms. A statistical difficulty is that accurate historical yield data often is not available at the regional or district level. Consistent data may be available only at an aggregated level, for example provincial level. Larger aggregation of statistics often masks local-yield variation, even from covariate risks such as drought. Area-yield insurance has been tried in the United States (GRP), Canada, Brazil, and India. Where individual farmer-yield 120 insurances are also offered, area-yield insurance typically has a low uptake and is not favored by farmers. Micro, Meso, and Macro Applications for Index Insurance Index insurance has applications for three possible levels of aggregation of risk: • Microlevel index insurance—An individual farmer is the policy holder. This is the most common type of index insurance and, where possible, can be a retail product. For weather index, the measurements typically are made at a single weather station closest to, and most representative of, the weather at the farmer’s location. • Mesolevel index insurance—The policy holder is an “aggregatorâ€? of risk—for example, a bank holding a portfolio of loans in a region, or a processing and packing company that is financially dependent on the throughput of a product. Measurements made at several weather stations may be needed to build an index representing the risk at meso level. • Macrolevel index insurance—An index is set to measure the large-scale impact of a peril over an entire region (for example, an island or province), and the policy holder has an interest in such a widespread event. Macrolevel indexes may use the results of multiple weather stations (as with mesolevel indexes) but can also use other indexation of major weather events, notably regional categorization of the strengths of typhoons. The distinction between these applications is important, since there are some types of weather events, notably typhoon, where the impact of the hazard is very complex and cannot be predicted at a local level. In these instances, a microlevel index may not be feasible on account of basis risk, but a macrolevel index may still offer an opportunity for development of index insurance, allowing the transfer of risk at macrolevel. Advantages of Index Insurance The main advantages associated with index insurance, compared to yield-based multiple- peril crop insurance, can be summarized as follows: Reduced adverse selection and moral hazard Under traditional MPCI, an individual farmer is more likely to buy insurance if he is a higher risk (adverse selection), and the farmer may be able to influence the claim (moral hazard). With index insurance, the farmer has no ability or incentive to influence the claim, since payout is based on an independent weather parameter. Elimination of field loss assessment Loss assessment is typically the greatest challenge to any traditional crop insurance program, due to the need to mobilize large numbers of skilled or semi-skilled assessors, who need to have some agronomic knowledge. The ability to make payouts without field assessment can open up opportunities that would otherwise be uninsurable. This issue 121 will be considered in more detail in relation to organizational structure and product design for traditional products in China. Easier distribution of product and enrollment of farmers Yield-based crop insurance programs require considerable work in establishing insured yields, and in recording details of farmers. Although extension and farmer education remains essential for index products, the actual enrollment process is less complex, and the product can be sold by less-skilled personnel. Further, the range of insured parties with an interest in crop production can be wider than farmers alone. Products could be relevant also to traders and those whose income is dependent on production of the crops concerned. Lower administrative costs Because of the lower distribution and, particularly, lower loss-assessment costs, the overall administrative cost margin, and therefore overall premium cost, should be lower than for a traditional crop insurance product. Challenges of Index Insurance Although in many cases index insurance is an effective alternative to traditional insurance products, it is important to understand the challenges associated with it. Basis risk Basis risk, the most critical feature of index insurance, significantly limits the applicability of index instruments. Basis risk is the difference between the payout, as measured by the index, and the actual loss incurred by the farmer. Because no field loss assessment is made under index insurance, the payout may be either higher or lower than the actual crop loss of suffered by the farmer. In agriculture, many variables ( such as local soils, management practices, localized occurrence of the weather peril, topography, and so on) influence the farmer’s final crop yield. Local microclimates, management practices, or crop variety features therefore increase the likelihood of basis risk. Examples of poorly correlated risks are hail and localized frost (which may also be affected by topography). Index insurance, therefore, is best suited to weather hazards that are highly correlated (covariate) and cause homogenous damage within a reasonable distance of a weather station. The extent of basis risk is influenced by several issues: • Covariance of the hazard to be insured by index: Hazards with high covariance (where spatial impact of the hazard is similar over a wide area) are rainfall deficit (leading to drought), high temperature conditions, and cool advective winds. Hazards with low covariance are hail and frost. 122 • Severity of the occurrence of the hazard: Under catastrophe situations (for example, severe drought), the impact of the peril may become more covariate and widespread, and severe damage more homogenous. • Extent of local microclimates: Local microclimates are often found where there is extensive topography (hilly areas), or in areas close to the sea. • Type of crop and its vulnerability to the hazard being insured: Some crops are very sensitive to the precise level of a weather parameter. For example, insuring fruit against spring frost gives rise to significant basis risk, because the amount of damage varies markedly due to the precise timing of the frost in relation to sensitive flowering period, the precise temperature and time a temperature threshold is passed, varietal differences, and cultural differences such as the type of training and trellising of the plants. Defining the parameters of a weather index to capture such damage is extremely difficult. • Increasing the network of weather stations: In view of the above constraints, it may be possible to reduce basis risk by increasing the density of weather stations used for the index program. Low-cost automatic weather stations are being added, to increase the density of the station network and ensure that a station is available at a reasonable distance from all insured farmers. Box A5.2: Basis Risk A major challenge in designing an index insurance product is minimizing basis risk. Basis risk refers to the potential mismatch between index-triggered payouts and actual losses. It occurs when an insured has a loss and does not receive an insurance 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. Since index insurance indemnities are triggered by exogenous random variables, such as area yields or weather events, an index insurance policyholder can experience a yield or revenue loss and not receive an indemnity. The policyholder may also experience no yield or revenue loss and still receive an indemnity. The effectiveness of index insurance as a risk-management tool depends on how positively correlated farm yield losses are with the underlying index. In general, the more homogeneous the area, the lower the basis risk and the more effective area-yield insurance will be as a farm-level risk- management tool. Similarly, the more closely a given weather index actually represents weather events on the farm, the more effective the index will be as a farm-level risk- management tool. Source: World Bank 2005. 17 Data availability Another challenge of index insurance is the need for accurate and complete data sets. This requirement applies to the historical record for underwriting and pricing purposes, 17 World Bank, Managing Agricultural Production Risk: Innovations in Developing Countries, World Bank 2005. (See http://siteresources.worldbank.org/INTARD/Resources/Managing_Ag_Risk_FINAL.pdf.) 123 and for recording of the parameter during the period of insurance. For weather-index insurance, meteorological data sets are required. For area-yield index, regional- or district-yield data is required. Integrity of weather stations Weather stations used for index insurance must have sufficient security that they cannot be tampered with. Preferably they should also have automatic, as opposed to manual, recording of data. Increasing the density of stations also provides better backup stations, which can be used to cross-check, as well as serving as backup stations in cases of station failure or tampering. Farmer extension and education Index insurance is a new concept for farmers—a concept they need to understand, especially since it has the capacity to reach small farmers who are outside the scope of traditional insurance. Experience in India suggests that small-size, low-income farmers can understand weather-index insurance and weather insurance payouts, if appropriate extension service is adopted, and training is provided to staff responsible for selling such insurance. Financial management of catastrophe events Index insurance is particularly suited to catastrophic perils, such as drought, for which financial management by insurers is difficult. For catastrophe perils such as drought, spatial spread of risk is not possible, and financial risks have to be spread temporally. This means that substantial reserves must be established, or that effective access to reinsurance is facilitated. Index insurance can open up new reinsurance markets that may offer increased capacity availability, as compared to traditional agricultural reinsurance, where the market is very restricted. Box A5.3: Summary of Advantages and Challenges of Index Insurance Advantages Challenges Less moral hazard Basis risk The indemnity does not depend on the Without sufficient correlation between the individual producer’s realized yield. index and actual losses, index insurance is not an effective risk-management tool. This Less adverse selection is mitigated by self-insurance of smaller The indemnity is based on widely available basis risk by the farmer, supplemental information, so there are few informational products underwritten by private insurers, asymmetries to be exploited. blending index insurance and rural finance, and offering coverage only for extreme Lower administrative costs events. Does not require underwriting and inspections of individual farms. Precise actuarial modeling Insurers must understand the statistical 124 Standardized and transparent structure properties of the underlying index. Uniform structure of contracts. Education Availability and negotiability Required by users to assess whether-index Standardized and transparent, could be insurance will provide effective risk traded in secondary markets. management. Reinsurance function Market size Index insurance can be used to more easily The market is still in its infancy in transfer the risk of widespread correlated developing countries and has some start-up agricultural production losses. costs. Versatility Weather cycle Can be easily bundled with other financial Actuarial soundness of the premium could services, facilitating basis risk be undermined by weather cycles (such as management. El Niño events) that change the probability of the insured events. Microclimates Make rainfall or area-yield index-based contracts difficult for more frequent and localized events. Forecasts Asymmetric information about the likelihood of an event in the near future will create the potential for intertemporal adverse selection. Source: World Bank 2005. Other Considerations The advantages and challenges for index insurances have to be considered in relation to other underwriting and cost considerations experienced in traditional agricultural insurance products. (See Box A5.4). These criteria are important in determining whether traditional or index products are likely to be more feasible. 125 Box A5.4: Comparison of Indicative Expected Cost Levels Involved in Underwriting and Administration Functions of Traditional and Index Insurance Function Traditional Index Comment Establishing insured Key function: Insurers Not required: Use an Individual farmer yield yield must establish farm or index as an agreed basis setting not feasible in district level yield. for payout. small-scale farming. Cost: High Cost: Low Underwriting Needs assessment of Not required, but insurers Product must be adapted individual risk or need to screen clients to to local weather situation localized district risk. check for insurable to minimize basis risk. interest. Cost: High Cost: Low Policy Sales Sales process requires Sales process also Education and extension high skills since it requires good product remains important for any involves underwriting knowledge. No major crop or index product. decisions. underwriting decisions in sales process. Cost: High Cost: Medium Paperwork and Generally complex. Simplified certificates or A key to cost reduction is information technology coupons. effective IT in head office (IT) and districts Cost: High Cost: Medium Field inspection Check for crop Not required. The insurer should emergence. monitor crop growing conditions in all cases. Cost: High Cost: Low Loss adjustment Needs inspection of crop Not required: Payment This category is one of damage and claim according to measured the most important adjustment. index. differences between traditional and index products. Cost: High Cost: Low Claims payment Settlement of claim. Settlement of claim. Once a claim is finalized, similar payment costs are Cost: Low Cost: Low incurred. Source: Manuamorn 2005 18. Designing an Index Insurance Product—General Principles and Steps Following are summaries of several steps involved in designing an index product. A more detailed explanation, with examples, is provided in a publication by the World Bank (World Bank 2005). 18 Manuamorn, O. Scaling up Micro-Insurance: The Case of Weather Insurance for Smallholders in India, World Bank 2005. 126 Step 1: Identifying significant exposure to weather This stage involves analysis of the causes of crop loss or damage—and identification of the key weather perils involved—in a particular region. The intention is to identify whether criteria likely to lead to a suitable index approach are met. (See Box A5.5.) Technical questions to be answered include: • Which weather events are causing crop losses? • Is there a good correlation of crop yields and weather event(s)? • Is the weather event widespread, or are there many microclimates? • In which phases of the crop cycle do the weather events occur? • Are the weather events “sudden onsetâ€? events, or are they progressive during the crop season? • What is the availability of weather stations in the region? • In what area is the crop located, and what is the economic importance of the crop in the region? • What is the profit margin of the crop, and is demand for insurance likely? • What are the organizational linkages for marketing an index product? • Is a more traditional named-peril or multi-peril crop insurance product also feasible, and what are the advantages and disadvantages of index versus traditional product for this crop and region? Box A5.5: Identifying Potential Indexes for Crop Exposures A weather index can be constructed using any combination of measurable weather variables and any number of weather stations that best represent the risk of the agricultural end user. Common variables include temperature and rainfall, although transactions on snowfall, wind, sunshine hours, river flow, relative humidity, and storm or hurricane location and strength are also possible and are becoming more frequent. Unlike energy indexes, in which the relationship between energy demand and weather is more transparent and is linked primarily to temperature, weather indexes for agriculture demonstrate more complex, albeit still quantifiable, relationships between crop yields or pesticide use. The normal process for designing an index-based weather insurance contract for an agricultural grower, for example, involves identifying a measurable weather index strongly correlated to crop yield rather than measuring the yield itself. After gathering the weather data, an index can be designed by (1) looking at how the weather variables have or have not influenced yield over time; (2) discussing key weather factors with experts, such as agrometeorologists and farmers; and/or (3) referring to crop growth models using weather variables as inputs for yield estimates or phenology models illustrating how weather variations relate to pest development. A good index must account for the susceptibility of crops to weather factors during different stages of development, the biological and physiological characteristics of the crop, and the properties of the soil. If a sufficient degree of correlation is established between the weather index and crop yield or quality, a farmer or an agricultural producer can insure his production or quality risk by purchasing a contract that pays if a specified undesirable weather event occurs or a specified desirable weather fails to occur. The index possibilities are limitless and 127 flexible to match the exposure of the agricultural grower or producer, as long as the underlying data are of sufficient quality. Source: World Bank 2005. Step 2: Quantifying the impact of adverse weather on farmer revenues This stage investigates more fully the relationship between the weather peril and its financial impact on crop production. Questions to be addressed include: • What is the extent of correlation between the selected peril and crop yield or damage? • Is there a crop-yield data series that can demonstrate the history of crop-yield loss due to the proposed weather hazard? • Are there crop-damage data relating to the weather hazard? • What level of basis risk exists? The investigation of the relationship between weather and crop yield is normally carried out in consultation with farmers, agrometeorologists and local agricultural experts. The objective is to characterize the requirements for optimum crop growth during each plant growth phase (germination, vegetative stage, reproductive phase, and so on), and to establish the relationship between the weather parameter under consideration, and crop loss. In the case of rainfall-deficit index (drought index), existing standard agro- meteorological models can be used to provide valuable input for establishing the appropriate threshold, tick size, and limits needed in each phase of the crop growth cycle. These models take into account that factors other than rainfall (such as soil water capacity, residual moisture, and temperature) have an impact on final crop production. The start-of-season dates, crop growth period (days), and end-of-season dates are established in the model. The simpler models use rainfall as the sole weather parameter. In climates where the start-of-season date is dependent on occurrence of a given amount of rainfall to permit crop germination, a variable start date of the season (dependent on the given rainfall being reached within a number of days, for example seven days) can be determined, and may be incorporated into the index design. The objective is to design a prototype index, for which a term sheet is typically produced. Table A5.1 provides an example of a term sheet for a rainfall contract, with comments in italics. 128 Table A5.1: Term-Sheet Features for a Weather-Index Contract (Rainfall) Product feature Description Crop type Specific crop type or types for which the index is designed. Index Description of the index type (for example, rainfall index) Start of contract [date] The start of the contract is normally a specified calendar date End of contract [date] Specific calendar date, or number of days after start of the contract Phase 1 2 3 Duration (days) [number] [number] [number] Threshold (mm) The amount of the meteorological [number] [number] [number] measurement (e.g., rainfall, temperature) at which payout starts (also known as a trigger) Limit (mm) The amount of the meteorological [number] [number] [number] measurement at which maximum payout is made Increment (RMB/mm) The amount of RMB to be [number] [number] [number] paid out for each increment of the meteorological measurement above or below the threshold. Sum insured per phase (RMB) [number] [number] [number] Premium (RMB) [number] The premium payable Reference weather station [Name of reference weather station] Nominated weather station, and supervisory authority. A secondary station is normally specified, as backup. Maximum policy limit [number] The maximum RMB amount which can be paid under the contract (RMB) over all phases Additional features of an Rainfall less than a given amount is not counted in the aggregate rainfall index which may be (This feature recognizes that minimal amounts of rainfall cannot be used by relevant plants.) Rainfall more than a given amount in 24 hours is not counted in the aggregate rainfall (This feature recognizes that large rainfall in a single day will run off and not available to plants.) Dynamic starting date (The start date for an index can be designed to incept only after a given cumulative amount of rainfall has fallen–being the amount normally required for planting.) Sources: Authors. Although the typical growth cycle of an annual crop may be in the region of 140 days, in which there are therefore 14 decads (10-day periods), the construction of the index needs to be simplified into a few phases. In India, indexes are simplified into three phases for rainfall insurance. Phase 1 is for crop establishment (typically 30 days); Phase 2 is for vegetative growth and flowering (typically 50 days); and Phase 3 is yield formation and ripening (typically 60 days). Each of these phases carries its own threshold, increment (tick size), and limit. Box A5.6 shows a prototype drought index for maize index in Thailand. Note that in this case an increasing sum insured is used in each phase, to reflect the production costs incurred up to the end of each growth phase. 129 Box A5.6: Example Phases in Prototype Maize Drought Index (Thailand) Phase Phase 1 Phase 2 Phase 3 Seedling Emergence Vegetative Physiological Maturity Growth stage to Knee-High Days 30 21 30 Trigger (mm.) 35 50 60 Limit (mm.) 15 20 30 Tick size (Baht/mm./rai) 42 21 21 Sum Insured (Baht/rai) 1,200 1,600 1,700 Source: World Bank 2005. An important feature of index design is its flexibility, in providing options for the periods of coverage, thresholds, limits, and increments. In the case of the Indian Agricultural Insurance Company, three options are available: a seasonal rainfall insurance; a phased rainfall-distribution index; and an index targeting only crop emergence. (See Box A5.7.) This is a generic product, targeting any crop grown in the specific season. Box A5.7: Example of “Varsha Bimaâ€? Index Insurance (India)—Coverage Options Option I: Seasonal Rainfall Insurance Coverage is against negative deviation of 20 percent and beyond in actual rainfall (in millimeters) from normal rainfall (in millimeters) for the entire season. Actual rainfall is the monthly cumulative rainfall from June to November (with June to September or October for short- and medium-duration crops). The payout structure is designed so that the yield is correlated to various ranges of adverse deviation in rainfall. The sum insured per hectare is the maximum payout corresponding to the maximum potential loss. The claim payout shall be on a graded scale (in slabs), corresponding to different degrees of adverse deviation in actual rainfall. Option II: Rainfall Distribution Index Coverage is against adverse deviation of 20 percent and beyond in actual rainfall index from normal rainfall index for the entire season. The index is constructed to maximize the correlation for weekly rainfall within the season. The indexes vary from Indian Meteorological Department (IMD) station to IMD station and crop to crop. The sum insured per hectare is the maximum payout corresponding to the maximum potential loss. The claim payout shall be on a graded scale (in slabs), corresponding to different degrees of adverse deviation in actual rainfall index. 130 Option III: Sowing Failure Coverage is against adverse deviation in actual rainfall (in millimeters) from normal rainfall (in millimeters) beyond 40 percent between June 15 and August 15. The sum insured per hectare is the maximum input cost incurred by the cultivator till the end of the sowing period, and is prespecified. The claim payout shall be on a graded scale, corresponding to different degrees of rainfall deviation. The maximum payout of 100 percent of sum insured is on the basis of actual rainfall data within a month from end of indemnity period. Sources: Authors. This index contract is designed to assist farmers to manage their exposure to drought in maize production, and at the same time to facilitate their access to credit for the purchase of hybrid seed varieties and fertilizers. In this case, a single policy was purchased through the lending bank on behalf of a group of borrowing farmers. Step 3: Structuring and pricing the contract Structuring: Although weather-index insurances originated in the derivatives market, the underwriting of index insurance contracts by insurance companies has led to contracts with structures similar to those of traditional insurance products. For example, an index product can fit into the normal administrative structure (underwriting, marketing, claims, and reinsurance) of an insurance company for traditional products. while offering the opportunity for much simpler procedures in all departments. Pricing: A summary of methodology for pricing index products is shown in Box A5.8. Because index products for agriculture are relatively new, the World Bank has provided considerable technical support for those new products from the outset. Compared to pricing of conventional agricultural insurance, which is subject to multiple variables, pricing of weather-index products is relatively simple. The only variable used for pricing is the data series of the weather parameter concerned. However, pricing for insurance or reinsurance purposes does need to consider the quality, data gaps, time length, and trends of the weather data. In particular, climate change requires that the more recent trends (and increased volatility) are considered. In the case of area-yield index insurance, more attention is required to biases in data quality, collection techniques, and trends. The final commercial premium which an insurer needs to charge is based on the pure risk premium, plus a risk margin, plus a share of overhead costs. The extent of the required risk margin is largely determined by the expected volatility of the outcome of the portfolio of risks to be insured. In summary, the volatility of the portfolio determines the amount of capital that the insurer, or reinsurer, needs to allocate to the business. An estimate of value at risk (VaR) is made based on analysis of the probability of outcomes, for example the expected worst payout in 99 years out of 100 (VaR 99 percent). Further details of methods and approaches to pricing index risks are beyond the scope of this report. 131 An issue for China, and recommendation of this study, is to establish technical assistance services to support commercial insurers in each province, with an objective of developing skills for index design and pricing within provincial insurance companies. Although reinsurers routinely deal with such pricing, these techniques often are not available to direct insurers who are starting index insurance programs. Box A5.8: Pricing Index Products—An Overview The premium of an index-based weather contract is determined actuarially by conducting a rigorous analysis of the historical weather to reveal the statistical properties and distribution of the defined weather index and, therefore, the payouts of the insurance or derivative contract. Such an analysis includes (1) cleaning and quality control of the data, that is, using statistical methods to in-fill missing data and/or to account for significant changes, if any, as a result of instrumentation or station location changes; (2) checking the cleaned data for significant trends and detrending to current levels if appropriate (this is particularly pertinent for temperature data, which, in general, exhibit a strong warming trend in the Northern Hemisphere); and (3) performing a statistical analysis on the cleaned and detrended data and/or a Monte Carlo simulation, using a model calibrated by the data, to determine the distribution of the defined weather index and the subsequent payouts of the contract. By determining the frequency and severity of weather events specified by the index, an appropriate premium can be calculated. It should be noted that the premium charged by providers in the weather market may depend on several factors, not all as objective as the underlying statistical analysis of the weather data. Institutions charge different risk margins, or discounts, over the expected value or fair price to potential buyers; these choices are driven by the risk appetite, business imperatives, and operational costs of the provider (Henderson et al. 2002). Source: World Bank 2005. Step 4: Implementing the contract—legal and regulatory This step brings together the tasks for launching the product. Although the activities are similar to a conventional insurance product (and may include market testing, drafting the insurance policy, publicity, distribution, premium collection, and education), an additional step for index insurance is to confirm legal and regulatory compliance. Index insurance has specific characteristics that distinguish it from conventional insurance. These characteristics relate to two regulatory topics: Insurable interest: Under insurance law in most countries, there must be a demonstrable “insurable interestâ€? in the insured subject matter. Whereas a traditional crop insurance contract states the location, crop type, area, basis of calculation of the sum insured, and obligations of the farmer and insurer, an index insurance contract can operate independently of demonstration of an insurable interest. Insurance regulators have normally required that index policies be sold only to clients who can demonstrate, as a minimum, the intention to grow the specified number of hectares of that crop type. 132 Indemnity: Under indemnity insurance, and most insurance laws, the amount of the loss must be measured, and the insured must be compensated according to the terms and conditions of the insurance policy. In index insurance, there is no measurement of loss of the crop, and the index policy makes a payout based on measurement of the weather event. Although this may require some changes to insurance regulations commonly in force, most insurance regulators have accepted the concept and practice of index insurance payouts and, where required, agreed to the amendment of regulations. (See Annex 4 for further detail). Reinsurance: There are reinsurers who specialize in index reinsurance. Many are the same companies who provide traditional reinsurance, but they underwrite index insurance in different internal departments. Thus, existing relationships between insurance and reinsurance companies can sometimes be maintained and developed. The same considerations apply to structuring reinsurance treaties for index insurance as exist for conventional crop insurance. Flood Index Insurance Flood index: The descriptions of index insurance in this chapter have related to weather- index insurance and area-index insurance. At present, the World Bank is undertaking research into whether the concept of weather index can be extended to flood risk. This research was started because of the importance of flood as a risk to agriculture in Asia. The results of this research will not be available until mid-2007, when publication of a technical paper is proposed. The basic premise is to make use of new technology to obtain objective and independent spatial measures of flood occurrence. These technologies are: • Remote sensing, which allows the use of satellite images to interpret, at a high resolution, the spatial extent and duration of flooding; • Flood modeling, which allows the mapping of flood plains, using historical data from rainfall and river flow to predict the frequency of recurrence of flood at any point in a river valley, leading to establishment of premium rating zones; • Agrometeorological modeling, which allows the modeling of the expected impact of yield reduction from different durations of flood inundation according to the growth phase of rice (or other crops); • Geographical information systems, which allow an insurer’s portfolio of clients to be geo-referenced and assigned to zones of expected equal risk. Preliminary findings of this study are promising, suggesting that these technologies can be harnessed to flood insurance, but only for inundation flood and not for flash flood. Further, there remain considerable underwriting and technology issues still to be addressed. These considerations will dictate how the final flood product may be formulated, if feasible. Since this type of index is at a developmental stage, flood insurance is not considered further as an option in this report, although it may become a possibility at some time in the future. 133 5.2. Index Insurance Options for China Before going forward with index insurance in China, certain technical, organizational, and financial factors need to be considered. Technical Considerations The following technical issues have an impact on the feasibility of index insurance: Weather stations Weather-index insurance can be operated only where there are weather data of sufficient quality and length. Payouts are based on measurements made at stations, and certain criteria are needed as a guideline for station integrity. Box A5.9. shows the criteria required by a lead reinsurer in the weather business. Box A5.9: Meteorological Data Requirements for Underwriting There are market makers who are keenly interested in offering rainfall-index insurance in developing countries. For example, Partner Reinsurance Company, New Solutions Department, presented the following list of items that are needed to get them interested in offering such contracts: - Historic weather data (prefer 30-plus years of data, especially to cover extreme risk) - Limited missing values and out-of-range values (prefer less than 1 percent missing) - Data integrity - Availability of a nearby station for a “buddy checkâ€? - Consistency of observation techniques: manual versus automated - Limited changes of instrumentation, orientation, and configuration - Reliable settlement mechanism - Integrity of recording procedure - Little potential for measurement tampering Source: Annual meeting of the International Task Force on Commodity Risk Management, jointly sponsored by the Food and Agriculture Organization (FAO) and the World Bank at the FAO, Rome, May 5–6, 2004. China is a member of the World Meteorological Organization, and the China meteorological service, through its bureaus in each province, conforms to WMO standards. (For more information, see Sections 5.5–5.8, below.) A detailed study of the available data quality and quantity was not practical. However, there is a generally good network of meteorological stations, and it is understood that there were no major national discontinuities in the data set that would be available for the implementation of weather-index insurance. Apart from the data collected by the China Meteorological Bureau, there are additional meteorological data sources from other organizations, such as those for hydrological purposes and for the reclamation groups. As noted above, basis risk is a significant shortcoming of index insurance. In India, the difficulties of basis risk are addressed 134 through the introduction of an additional network of automatic weather stations, to increase the network of recording stations. Although no historical data clearly exist at such stations, a high density of stations allows statistical interpolation of the expected meteorological parameters expected at such stations, and also provides additional cross- validation of actual results. Higher densities of stations allow better detection of any tampering with specific stations. Such stations may carry a capital cost of some $10,000, or considerably less for a station solely recording rainfall. A guideline for rainfall-deficit index insurance, in areas not subject to major microclimates, requires that the distance between stations be no more than 20 kilometers, but compliance with this requirement is highly dependent on the local microclimatic and topographical conditions. The extent of such local microclimates must be assessed in each situation. Increasing the density of the station network, to reduce basis risk and to facilitate index insurance, would be feasible. Meteorological data series For development of index insurance, data on a daily record is required. Daily records generally remain available at stations, even if not currently held electronically at all stations for the full period of recording. Further, at least 20 years of data were generally held for the main network of stations, and more in some cases. The data held within the bureau appeared to conform to international standards that were likely to be acceptable for the development of index insurance. Agrometeorological research In the provinces visited, there were many organizations with expertise in agrometeorology. This expertise was within meteorological bureaus, or in provincial academies or universities. In addition, there is a good level of technical know-how at the level of state farms and reclamation bureaus. Thus, expertise exists in organizations that could support insurers in the design of indexes. Agricultural production data The availability of reliable time-series yield data, on a local level, would be a prerequisite to the development of area-yield index insurance, and would also serve to validate any weather-index products at the design phase. It is understood that problems exist in the quality of time-series yield data at a local level. Microlevel index products The above features demonstrate that the right meteorological data and specialist organizations generally exist to support weather-index insurance development, although the availability and quality of localized-yield data series is doubtful. The latter may limit the scope of area-yield index insurance. However, the ranges of crops and weather-hazard combinations for the application of index insurance are not universal. The high proportion of irrigated crops in China limits the scope of rainfall-index insurance for drought risks. 135 Macrolevel index products This type of index product is a new concept and only experimentally available in a few countries. As with microlevel index, however, the existence of weather data should enhance the potential for macrolevel risk transfer within China in the future. The primary client for macrolevel products would be insurance companies. Macrolevel weather-index products would be offered to insurance companies by reinsurers, but application of this concept would be secondary to conventional reinsurance, as insurers develop their business and their exposures can be transferred using conventional reinsurance, without the basis risk inherent in a macrolevel index reinsurance product. The recommendation is that conventional reinsurance at a provincial level should be pursued as a priority, at least at the present stage of development. Organizational Considerations The following organizational issues need to be taken into consideration when assessing the feasibility of index insurance: Underwriting Index insurance products are underwritten by insurance companies in the same manner as other insurance products. A conclusion is that a mix of conventional and index insurance products is likely to be necessary to meet the risk-management needs in each province; therefore an insurer will be the primary party responsible for the underwriting and administration of insurance business in China. Risk transfer arrangements for index products (through reinsurance) follows similar organizational considerations as for traditional products, even if the types of risk insured are often covariate, giving rise to high-capacity requirements. In scaling up agricultural insurance activity in China, both traditional and index products can be accommodated within the same insurance companies. Underwriting of index policies does not require individual-farmer, or localized, risk assessment during the actual sales process. Thus, the procedures for acceptance of the risk can often be simplified in comparison to a conventional insurance product. These product simplification procedures have implications for the educational levels, and staff numbers, of persons involved in underwriting, distribution, and claims functions within insurance companies. Notably, MPCI requires staff with significant skills in agronomy and yield assessment. Distribution Staff educational requirements (level of agronomic training) for distribution are similar to those for underwriting. Nevertheless, index products are a new concept for farmers and may require at least as much explanation to potential clients as do traditional products. Distribution channels also should be selected appropriately. Distribution opportunities for index products could be wider than for traditional products. 136 The past distribution structure for the Peoples Insurance Company of China (PICC) was through either (1) provincial, district, and village organizational channels, where individual villages would collectively decide whether to participate in an insurance program and the policy would be issued at the village level; or (2) to state farms. In contrast, the special situation of the Heilongjiang Reclamation Group and Xinjiang Production Corps allowed Sunlight Insurance Company and China United, respectively, to access their own marketing channels to farmers. The need to have cadres of skilled and trained agronomists to support traditional insurance products is a serious impediment to operation of complex crop insurance. The simplicity of the index product, and lack of specialist loss-assessment requirements, opens up almost any marketing channel. Some authors have suggested that weather insurance should be opened up to allow marketing of weather insurance policies to any member of the public with some interest in the outcome of the crop (for example rural traders and processors). Such a market is developing in India. In China, the simplified distribution opportunity of index products could open up new distribution channels that are not possible for more sophisticated traditional products. In particular, an index product could be distributed by linkages to microfinance institutions, banks, or local (village) administrative authorities. Given that HRG and XPCC are unique organizations in China, less sophisticated distribution channels would be important to scaling up agricultural insurance in China. Loss assessment It is with loss assessment that the main differences between traditional and index insurance become apparent, with index insurance providing significantly better opportunities. At its most complex, MPCI requires that each insured unit reporting a loss have an actual assessment of the yield loss. Such assessment requires that persons knowledgeable in crop production and crop science are involved to identify the insured, and uninsured, causes of loss. Such work is highly skilled. At its intermediate level, a damage-based crop insurance policy with a percentage-damage assessment procedure can be undertaken by less-skilled personnel. In Hainan, for example the field assessment of wind damage to rubber trees is simplified to a number of easily measured criteria: the height at which a stem is broken; the angle at which a tree is leaning; the number of trees on the property. Such measurement can be performed by persons with limited or no agronomic education. The constraint on traditional product expansion as a result of loss- assessment staffing is, therefore, reduced for index products. Financial Considerations Financial issues that have an impact on index insurance include the following: Sums insured Annexes profiling each insurer, as part of this study, describe the levels of sums insured for the current traditional crop insurance products, relative to the production costs of growing crops. In China the sums insured (and therefore insurance compensation) was often considered inadequate to cover even the production costs incurred. Because agricultural insurance is risky—and because premiums need to be set at levels adequate to cover claims, operational expenses, and a risk margin—there is always a balance 137 between affordable premium and optimum sum insured. In China, low sums insured have been established both for affordability reasons and to limit catastrophe exposure of insurance companies. It is recommended that the target should be to provide sufficient sum insure at least to cover production costs, including some allowance for the farmer’s labor. The premiums associated with a sum insured equivalent to the expected revenue of crop sales would often be unaffordable. Hence, an accepted guideline is to provide a sum insured equivalent to the production costs. It also makes evaluation easier, since production costs are relatively stable, as compared to expected revenues. The first stage in developing insurance products, whether index or conventional, is likely to be to cover production costs. Further refinements in index product design will increase the sum insured to reflect the accumulated production costs incurred according to the crop growth phase when the weather event occurs. (See Box A5.6, above.) The flexibility of the index product allows different sums insured to be assigned to different growth phases, as required, for particular crop types. Similar increases in sums insured are also used in China in some existing products. Index products are flexible in the setting of thresholds, increments, and limits. This allows some flexibility the selection of affordable premium rates, which itself allows the selection of sum insured, since the premium paid by the farmer is the product of the sum insured and the premium rate. Pricing The pricing of weather-index products is derived from analysis of the meteorological data (as described in Step 3, above). Due to the restriction to a single weather hazard (as opposed to multiple hazards of traditional coverage), combined with lower administration and loss adjustment costs, index insurance can be offered at a lower premium rate than can a traditional contract; but the covered hazards are restricted, and direct comparisons are difficult. For example, on traditional products, claims costs may be higher where antiselection and moral hazard are difficult to control, leading to a need for insurers to charge higher premium rates. A valuable approach with index insurance is to determine the level of protection that can be provided for a given premium (for example, by adjusting thresholds, limits, and increments). A series of options can be derived that reflect affordability and that proxy actual loss as closely as possible. Adjusting the thresholds to cover only extreme weather events allows more catastrophe-level products to be offered. However, regions or crops that are prone to frequent damaging weather events may be unsuitable for growing such crops on an economic basis, and still cannot be insured at economical premium levels with index products, as with conventional products. Next Steps in Implementation In the event that provincial authorities decide to proceed with index insurance, the following phases are suggested: 138 Phase 1: Prototype index product identification and design: • Identification of key hazards, seasons, and crop types for pilots • Identification of target clients and distribution channels • Data collection and analysis (including meteorological, agro-meteorological, crop area and production, and so on) • Product design, policy drafting • Pricing Phase 2: Pilot project planning • Formation of stakeholder group, organizational roles and responsibilities • Development of pilot project plan • Design operational processes Phase 3: Testing the product concept with farmers Phase 4: Pilot project implementation (1–3 years) Phase 5: Monitoring and evaluation of the pilot Phase 6: Expansion into commercial-scale operation 5.3. Heilongjiang Located in the northern frontier of China, Heilongjiang Province covers 469,000 square kilometers. It is situated between the temperate and the frigid zones, with continental monsoon climate. Overview of Agriculture in the Province Heilongjiang has 9.8 million hectares of cultivated land. According to the China National Bureau of Statistics, Heilongjiang is primarily a rainfed production area, notably for soybeans (3.9 million hectares), corn (2.2 million hectares) and wheat (0.3 million hectares), but with 2.28 million hectares under irrigation, mainly for rice (1.6 million hectares). It is an important livestock-producing province for dairy cattle, hogs, and poultry, mainly raised under intensive conditions. Climate and Production Systems Rainfall in the province—typically 370–600 millimeters (Figure A5.2.)—falls principally in the summer months. There is marked seasonality, with very cold winters. Illustrative climate characteristics in Harbin are shown in Table A5.2. Crops are spring sown, except for wheat which may be overwintered. 139 Figure A5.2: Annual Precipitation, Heilongjiang Source: Oregon State University, Spatial Climate Analysis Service. Table A5.2: Illustrative Climate Characteristics, Heilongjiang (Location: Harbin) Latitude: 45.9° Longitude: 126.6° Elevation: 155m Prc. Tmp. Tmp. Tmp. Grnd Rel. Wind Month Prc. Prc. cv Wet mean max. min. Frost hum. Sun (2m) ETo ETo mm/m mm/d % days °C °C °C days % % m/s mm/m Mm/d Jan 3.5 0.1 75.2 6.0 -18.9 -13.1 -24.8 31.0 72.4 60.7 2.5 9.3 0.3 Feb 5.2 0.2 84.8 5.8 -15.3 -8.8 -21.8 28.2 68.6 65.1 2.5 11.2 0.4 Mar 9.9 0.3 74.4 5.4 -4.5 1.7 -10.7 29.0 56.9 65.4 3.2 40.3 1.3 Apr 22.0 0.7 78.4 6.4 6.5 13.2 -0.2 18.0 49.3 58.1 3.8 96.0 3.2 May 39.2 1.3 57.1 10.5 14.4 21.3 7.5 3.5 50.2 56.8 3.6 145.7 4.7 Jun 81.4 2.7 48.7 12.8 19.9 25.9 13.9 0.0 64.2 55.4 2.8 144.0 4.8 Jul 161.3 5.2 38.8 15.4 22.8 27.7 18.0 0.0 77.2 50.0 2.5 133.3 4.3 Aug 119.5 3.9 55.9 13.6 21.0 26.1 16.0 0.0 78.3 53.7 2.3 117.8 3.8 Sep 62.1 2.1 52.2 10.8 14.4 20.5 8.4 2.1 70.1 60.9 2.5 90.0 3.0 Oct 24.3 0.8 65.3 7.4 5.5 11.5 -0.5 19.1 62.5 62.1 3.0 62.0 2.0 Nov 8.4 0.3 69.4 6.0 -5.7 -0.5 -11.0 28.2 66.8 59.9 3.1 27.0 0.9 140 Dec 5.0 0.2 103.3 6.6 -15.3 -9.9 -20.7 31.0 71.7 55.4 2.7 12.4 0.4 Total 541.8 889.0 Key: Prc.= precipitation; cv= coefficient of variation; Tmp= temperature; grnd= ground; rel. hum = relative humidity; ET= evapotranspiration. Source: FAO Climate Information Tool (www.fao.org). Weather stations in Heilongjiang Within Heilongjiang are 83 stations operated by the provincial weather administration, 32 of which participate in international data exchange. Some automated stations have been in place since 2001. HRG has 90 additional stations, and the hydrological services operate stations and rain gauges. Prior to 2001, HRG operated weather stations independently, but since that time they have started integrating into the provincial meteorological weather administration. A typical recommended distance between stations is 50 kilometers. Data collections started as early as 1952 in some locations. The project was provided with monthly data for 32 internationally reporting stations covering the period 1971–2000. Prior to 1991 data in the HRG area is not in an electronic format. Generally, daily data are not recorded over a long time period in electronic format, except those stations that report internationally as part of the WMO network, but daily records are still available at the stations. Those stations involved in a pilot area may enter daily data into electronic format. Cloud seeding for rainfall enhancement Weather management has been undertaken for two purposes: hail prevention and rain enhancement, with hail prevention being the main activity. For data analysis for a drought-index insurance program, it would be necessary to determine whether rainfall enhancement had caused a bias to the natural rainfall record. No data were obtained on the extent of this activity, or on the impact on the historical rainfall records. Depending on the extent of the past activity, choosing locations that were not in a command area for any rainfall enhancement activities would avoid the additional task of such interpretation. Identification of crops and weather hazards Key exposures to crops: A matrix of main cultivated crops and key perils is shown in Table A5.3. Key risk exposures in the province were explained as follows: Drought: Up to 70 percent of all losses are due to drought, the main cause of crop loss. Principle crops affected are corn, soybean, and wheat. There are regional differences, and drought is a particular problem in the south of the province, where the wheat area is greatly reduced, and soybean production has been moved to the north. There also has been increased incidence of drought. Spring is characterized by dry and cold winds, and low rainfall at spring and early summer is critical to crop establishment. Rainfall in 141 summer was not reliable. (Typhoons do not pass over the province, but the impact of distant coastal typhoon was often beneficial in bringing rains and alleviating drought.) Flood and waterlogging: Major production areas in the province are from land reclaimed by the HRG, and these areas are often low-lying, with shallow elevations. Irrigation has brought areas of river valleys and basins into production. Periodic floods occur in these areas as a result of higher river flow and prolonged or excessive rainfall during July through September. Up to 42 million mu (2.8 million hectares) were potentially prone to flooding. Flood and waterlogging is therefore a complex, and localized, problem, also linked to soil alkalinity 19, and to drainage management. Hail: HRG, through Sunlight Insurance Mutual, operates hail-prevention systems (rocket-launched cloud seeding) in some hail-prone areas, principally for tobacco and corn. Operational costs are met from a deduction from insurance premiums, and capital costs are met by the provincial authorities. Although the hail prevention is considered highly cost effective, hail remained a localize peril faced by farmers, especially outside of the prevention command area, and is dependent particularly on topography. Low summer temperatures: Rice is grown at the northernmost limit of its climatic range. Cooler summers can give rise to insufficient growing-degree days to reach satisfactory maturity, with consequent loss of yield and quality. Autumn frosts: Early autumn frosts were sometimes a problem, especially if poor spring sowing conditions, or resowing, gave rise to late planting for several crop types. Winterkill: Autumn-seeded wheat is dependent on snow cover to protect the seedlings from extreme cold conditions. Reportedly, this was not a commonly recurring problem and that the area under wheat had decreased significantly. Disease: No specific disease issues were discussed, but disease of plants is understood to be a significant cause of loss in the province. Key exposures to livestock: Livestock production systems are intensive, with limited feedstuffs dependency on natural pasture. Hence, exposure of livestock to weather hazards, such as drought on pasture lands, was considered to be limited. (Main exposures faced by livestock producers were disease and variable feed prices, often influenced by production levels.) 19 FAO Land and Water Development Division, 1999, (referring to China as a whole) notes: “In 1996, the area subject to waterlogging was 24.58 million hectares, of which 20.28 million hectares were controlled through drainage. In 1995, the power drained area was 4.2 million hectares. Saline-alkaline cultivated areas cover 7.73 million hectares, of which 5.51 million hectares have been improved or reclaimed. The total cultivated area protected from floods is 32.69 million hectares. In northern China in particular, waterlogging, salinization and alkalization have been the main constraints on agricultural production. In this region, there are about 6.70 million hectares of low-yielding farmland prone to waterlogging (out of a total of 18.09 million hectares of farmland), and 33,500 hectares of saline-alkaline soils.â€? (http://www.fao.org/ag/agl/aglw/aquastat/countries/china/index.stm) 142 Data on natural hazard damages: Annex 2, above, considers the risk assessment in Heilongjiang Province. Data on the area covered and area affected by natural hazards indicate, in decreasing order of magnitude the hazards of drought, flood, hail, and frost. Data on insured losses: Annex 6, below, includes analysis of the causes of insured loss in Sunlight Insurance Company’s portfolio. Since 2001, 36 percent of losses are a result of drought, 28 percent from flood or waterlogging, 10 percent from frost, 9 percent from hail, 13 percent from pest and disease, and 4 percent from other perils. These results are consistent with the other sources of information obtained on risk exposures. 143 Table A5.3. Heilongjiang—Main Crop Types and Weather Hazards Total Total Area of Province MAIN CROP Oil-bearing Fiber Sugar Orchard selected Cultivated TYPES Rice Wheat Corn Soybean Tubers crops Cotton crops Beet Tobacco Vegetables crops crops area Provincial area ('000 1587.8 255.0 2179.5 3913.6 346.4 411.1 0.0 98.5 75.5 31.5 291.6 39.5 9230.0 9888 ha) 20 Peril Drought Prolonged high temperatures Typhoon Flood Excess rain Waterlogging Prolonged low temperature Spring frost Autumn frost Winterkill 20 Source: China Statistical Bureau, 2004 data. 144 Hail Wind Pest Disease Sea water storm surge Key: = perils of significant importance in the province = specific crop perils noted This table includes only crop and peril combinations discussed for this review. It is not intended to be exhaustive. The crop-peril matrix can serve as an approach for developing a comprehensive database on weather risk exposures. Sources: Authors. 145 Organizational Considerations for Index Products in Heilonjiang Sunlight Agricultural Insurance Company was formed two years ago as successor to the former well-established agricultural mutual insurance co-operatives of the HRG. It has a long history of insurance experience. HRG local officials are highly integrated into the sales and the loss-assessment processes of the company. Sunlight currently does not operate outside the HRG areas, which account for approximately 20 percent of the province’s cultivated area, but wishes to extend its operational base outside HRG. Currently insured products and perils Sunlight currently offers comprehensive crop insurance coverage (MPCI), including drought and uncontrollable pest and disease, to their clients in HRG. MPCI products have been offered to HRG clients for a long and continuous period. Sunlight has certain structural advantages in operating MPCI, a product with inherent difficulties. First, due to the structure of HRG, all farmers are insured, and this reduces or eliminates the problems of antiselection (only the least favorable farmers insuring). Second, farmers within HRG broadly follow collective farm production decisions and farm inputs recommended by HRG for optimum production, which further reduces the exposure to individual farmer- yield outcomes and moral hazard. Third, loss adjustment is able to make use of local cooperative resources. Although any yield-based product remains technically complex, particularly in relation to setting of insured yields and in loss assessment, the organized structure of HRG may assist in mitigating these operational insurance difficulties. Sunlight benefits from the individual farmer historical crop-yield information, gathered over a long period. The above benefits do not exist outside HRG areas. Sunlight advised that it would be very difficult to expand its current MPCI product outside of the HRG command area. Further, the main exposure is drought, which provides an opportunity for index products outside of the HRG area. Drought-index insurance for corn and soybean: It is considered that the most interesting opportunity for an index product in the province is drought (rainfall deficit) index insurance for corn and soybean. This potential product is of interest because it fits several criteria: • Corn and soybean are two main economic crops grown in the province; • Corn and soybean are grown under rainfed conditions, and drought was the key exposure of the province; • Corn and soybean are grown over wide areas with relatively limited variations in terrain or microclimate—conditions that would reduce basis risk, depending on the exact density of weather stations employed in an insurance program; • Drought is inherently difficult to insure, from an operational and financial point of view; • Drought (rainfall deficit) index insurance is the most tested of index products in other countries, and corn is a crop type for which indexes have been developed. A 146 soybean drought-index product has not yet been developed elsewhere, but is technically feasible. Table A5.4: Technical Features of a Prototype Corn or Soybean Rainfall-Deficit Index Product Parameters to be developed during the product-design phase: Product feature Description Crop type Corn and soybean (two separate products) Index Aggregate rainfall during each phase (mm) Policy duration (days) [number] a Phase 1 2 3 Duration (days) [number] [number] [number] Trigger (mm) [number] [number] [number] Limit (mm) [number] [number] [number] Increment (RMB/mm) [number] [number] [number] Sum insured per phaseb (RMB) [number] [number] [number] Maximum policy limit (RMB) [number] Reference weather stationc [Name of reference weather station] Maximum policy limit (RMB) [number] Premium (RMB) [number] Additional relevant features of an index Rainfall less than a given amount is not counted in the aggregate rainfall. Rainfall more than a given amount is not counted in the aggregate rainfall. Dynamic starting date (day 1 of the index dependent on a given cumulative amount of rainfall—as deemed necessary for planting) Notes: a. More than 3, or less than 3, phases can be considered, e.g., Emergence–Vegetative; Flowering/Reproductive; Maturing phases. b. Sum insured per phase is frequently based on the cumulative production costs incurred up to the end of that phase. c. The reference weather station is the meteorological station where observations are recorded and on which claims payouts will be calculated. Sources: Authors. Correlation Between Rainfall and Yield A preliminary analysis of correlation of rainfall data in the growing season (June, July, and August) in the yields of soybeans and corn (in the study areas for which data were collected in northern and southern Heilongjiang) indicates that there is not a clear correlation. This suggests the need for a more detailed analysis of these local statistics of 147 yields and causes of loss. However, it is clear that in a rainfed area such as Heilongjiang there will be a relationship between rainfall and yield. Monthly rainfall data were available from several stations in Heilongjiang. During the development of a product, it would be necessary to consider all data relating to the crops in specific areas—crop planting dates and crop calendar, crop input requirements, meteorological data, yield data, varieties, and so on—to design an appropriate index contract. Parameters within the contract would be selected so as to reduce basis risk. Table A5.5 provides an example of how rainfall data for a selected weather station can be analyzed statistically to explore the payouts associated with different thresholds and limits in an index product. Since only monthly data were available, a hypothetical example—setting three phases (June, July and August)—is shown, and is intended to be illustrative only. Thresholds were arbitrarily set at 40 percent of the average monthly rainfall, and the limit at 5 percent of the average monthly rainfall. The number of years in which there would have been some payout under these thresholds is 6 years out of 30 years of data. In designing a real index product, the phases and their length and start and finish dates would be determined based on crop type, planting dates, crop growth phases, and so on, as described in this report. It should also be noted that, in estimating the long- term payouts expected, it is customary to determine the most appropriate expected probability density function (distribution) that reflects the rainfall pattern, and to model payouts based on that function. This allows a comparison of actual payouts made in the past period for which actual rainfall is available, and the modeled payout, which will normally be higher, since extreme events are underrecorded in short-term time series. It should also be noted that a risk margin and expenses margin need to be added to derive a commercial premium rate. The example shown in Table A5.5 is intended solely to illustrate the way in which it is possible to work with different options of threshold and limit. Similarly, different options of insured value (reflecting increased production costs during the crop season) are frequently used in different phases, with the intention of reimbursing actual costs incurred up to that phase, and having the effect of minimizing premium costs. Similarly, different options of increment (tick size) can be used between the threshold and the limit. 148 Table A5.5: Rainfall Payout Example at Nenjiang, Heilongjiang Examples: June July August Threshold (mm) : a 340 580 420 Limit (mm)b: 40 70 50 c Year Actual rainfall (mm): Percentage paid out : June July August June July August 1971 883 843 1473 0% 0% 0% 1972 813 1636 873 0% 0% 0% 1973 248 1395 361 31% 0% 16% 1974 463 870 1711 0% 0% 0% 1975 1007 1447 385 0% 0% 9% 1976 743 1721 458 0% 0% 0% 1977 832 1457 556 0% 0% 0% 1978 1556 1323 180 0% 0% 65% 1979 1003 816 992 0% 0% 0% 1980 1083 1616 1150 0% 0% 0% 1981 864 3891 1258 0% 0% 0% 1982 183 443 1193 52% 27% 0% 1983 482 1797 686 0% 0% 0% 1984 464 1841 2403 0% 0% 0% 1985 953 427 1616 0% 30% 0% 1986 761 1572 1056 0% 0% 0% 1987 413 1719 1454 0% 0% 0% 1988 578 2572 630 0% 0% 0% 1989 1050 1799 945 0% 0% 0% 1990 1364 858 1387 0% 0% 0% 1991 813 1740 1171 0% 0% 0% 1992 428 1125 1740 0% 0% 0% 1993 1215 1803 1329 0% 0% 0% 1994 880 1924 332 0% 0% 24% 1995 1429 971 819 0% 0% 0% 1996 1021 1453 646 0% 0% 0% 1997 614 960 1428 0% 0% 0% 1998 2179 751 761 0% 0% 0% 1999 489 1489 1386 0% 0% 0% 2000 523 812 1189 0% 0% 0% Average: 844 1436 1052 149 Empirical loss costd: 2.77% 1.90% 3.80% Modeled loss coste: 6.72% 5.03% 7.31% Notes: a. Threshold set at 40% of average monthly rainfall each month b. Limit set at 5% of average monthly rainfall each month c. Percentage payout structure shown in Figure A5.1. (Note that different sums insured can be set per phase, to reflect increasing input costs in growing the crop.) d. Empirical loss cost is the average actual payout for that month. e. Modeled loss cost is the average modeled payout rate for that month. (Modeling a distribution allows for extreme events not shown in the actual data set, and estimates the long term payout expected. ) Sources: Authors. Other Microlevel Index Insurance Product Opportunities in Heilongjiang Additional opportunities for microlevel index products in Heilongjiang include the following: Extension to other rainfed crop types A similar approach to the product described for corn and soybean rainfall deficit could be extended to other field crop types, such as wheat, barley, or oilseed crops. Design features would be similar, but triggers, limits, and increments would be adapted to the crop types concerned. Similarly, sums insured should reflect the production-cost economics for each crop type. Generic index for rainfed crops For rainfed crops with similar agronomic features, it might be possible to develop a generic index. Although such an index presents no technical problems regarding operation, time did not permit determining whether there are specific crop types that have similar water requirements, which would mean limited basis risk between the payout from a generic product and the crop-yield reduction expected. Such a product could be foreseen where the farmer would have some choices for the sum insured purchased, to reflect the level of intensity of his or her production system. Short time-window products A feature of the above corn and soybean rainfall product is that the policy is operative for the full period of the crop cycle (planting to harvest). Weather-index insurance is very flexible, and products can be developed to target a specific weather hazard known to cause yield reduction during a particular period during the crop cycle. Specific 150 opportunities were not explored, but possible options to be investigated further could include: • Prolonged cold temperature conditions affecting rice during period of growth and maturing—reduced growing degree days (GDD). (Note: This hazard is already covered in HRG areas under the MPCI policy.); • Early autumn frost affecting any crop types with special quantity or quality vulnerabilities (such as fruits and vegetables); • Special exposures to unexpected excess rainfall, giving rise to inability to harvest, or loss of the crop quantity or quality by excessive rainfall at time of maturity. Although it was not explored whether such exposures exist, they may be potentially indexable. Provision of coverage in specific windows, and for a single hazard, may be perceived by the farmer as providing only narrow coverage regarding hazards and timeframe. However, if that hazard is of a severity that can cause infrequent but severe loss, a weather index can still be an effective way of covering the specific hazard, if a weather station is located nearby. Unlikely Exposures for Index Insurance in Heilongjiang Some exposures, such as the following, are very unlikely to have an application in Heilongjiang. Exposures with high basis risk Features that make the index approach very problematic are circumstances where the vulnerability of the crop is not predictable; or where the hazard is very complex, or is localized. Examples of hazards that are not candidates for indexation are: • Hail in any crop type; • Flood, waterlogging, and salination of soils. Any development of area-yield index would be dependent on assessment of the quality of district-yield data. Conclusions for Index Insurance in Heilongjiang Some index products have a place in agricultural insurance in Heilongjiang, and others are problematic, as the following summaries explain. Corn and soybean index product There is a potential for a corn and soybean drought-index product, which could be researched for implementation in the province. The product could be distributed through insurance companies such as Sunlight Insurance, and the nonretained portion of risk could be reinsured internationally. 151 Other index products There are other smaller potential opportunities for index products, addressing specific weather hazards—for example, insufficient GDD (due to prolonged cold temperatures) for cotton production; early autumn frosts; or excessive and prolonged rain at harvest time for vulnerable crop types, such as fruits or vegetables. The drought-index coverage could also be extended to other rainfed crops. Existing insurance products Many hazards cannot be addressed by a weather-index approach (for example, hail, pest and disease, and flood risk). Consistent with observations in other provinces, there is a very important ongoing role for traditional insurance products, and particularly for named-peril crop insurance products where loss assessment can be based on percentage- damage loss assessment (for example, hail). There are potential benefits of promoting exchange of best international practices with provinces within China, through (1) organizing central training and technical support activities in China; and (2) promoting networking with specialist agricultural insurance organizations internationally. A conclusion is that, although index insurance is likely to have an important place in the future mix of products in China, traditional crop insurance also has equal or greater importance. China’s available manpower to carry out simplified percentage-damage loss assessment is an important consideration in facilitating traditional named-peril crop insurance. Implications of offering a corn and soybean drought-index product in parallel with MPCI product Sunlight Insurance offers an existing MPCI policy type for corn and soybean, within the HRG area, and would need to consider whether drought-index insurance could be marketed as an alternative type of policy, or should be considered as a replacement. It is recommended that no such decision should be made until there is an evaluation, through pilot testing, of the index insurance product. A potentially more attractive opportunity for Sunlight would be to adopt the index policy to spearhead expansion in other areas of the province, outside HRG. These areas do not have the benefits of HRG infrastructure, which facilitates the operation of MPCI. In particular, it could be extremely difficult and costly to attempt implementation of a comprehensive loss-assessment system for drought outside of HRG areas, without the benefits of local HRG staffing. A major advantage of the index product is that there is no requirement for a loss-assessment network. Macrolevel opportunities for index reinsurance There is potential for transferring major drought risk through an index reinsurance product linked to many weather stations. This type of reinsurance could be an alternative to traditional reinsurance in the future, if required (for example, if reinsurance from traditional markets became restricted), to provide reinsurance capacity for the traditional MPCI portfolio of Sunlight Insurance. However, such a product would leave basis risk 152 with Sunlight Insurance, since the payout would only be a proxy, measured on rainfall, to the actual outcome of the MPCI portfolio, which would be measured by field loss assessment. Since Sunlight Insurance enjoys favorable reinsurance relationships for its MPCI portfolio, it seems unlikely that there would be benefits in pursuing this option at present. Organization of product development and implementation Meetings occurred with several organizations with relevant expertise for development of index products, which could participate with Sunlight Insurance. These organizations are Heilongjiang Provincial Meteorology Bureau, Heilongjiang Academy of Agricultural Sciences, Northeastern Agricultural University; and Heilonjiang Reclamation Group Meteorological Bureau. 5.6. Xinjiang Uygur Autonomous Region Xinjiang Uygur Autonomous Region (Xinjiang) represents about one-sixth of the total area of China, with 1.66 million square kilometers. Xinjiang is bordered by Mongolia, Russia, Kazakhstan, Kyrgyzstan, Tajikistan, Afghanistan, Pakistan, and India. Overview of Agriculture in the Autonomous Region Of Xinjiang’s 68 million hectares (42 percent) appropriate for farming and animal husbandry, 48 million hectares are natural grasslands used for grazing, 666,700 hectares are man-made pastures, 9 million hectares are reclaimed land, 4 million hectares are cultivated farmland, and 4.8 million hectares are available for forestry, with 1.5 million hectares in current production. According to China National Bureau of Statistics (2004), Xinjiang’s most dominant crop is cotton (1.137 million hectares, 30 percent of China’s production). Other important crops are wheat (686,000 hectares), corn (518,000 hectares), oil- bearing crops (215,000 hectares), and vegetables (186,000 hectares). There are 396,000 hectares planted in orchards, including grapes. High-technology agriculture is practiced in irrigated areas. Xinjiang has 46 million head of livestock. Climate and Production Systems Xinjiang has a typical continental climate, with low rainfall. Major differences exist in its climate, ranging from desert (the Gobi desert) to mountains with 8,611-meter elevation. Cultivated crops are grown in irrigation zones managed from rivers fed from snowmelt. These zones are located at the base of the central Shan mountains, which run west-to-east through the region. Extensive pastures in the north and west support nomadic grazing systems. The climate and production systems are therefore highly varied. Figure A5.3 shows the region’s distribution of annual rainfall, which is typically about 200 millimeters in the areas of irrigated agriculture. Table A5.6 shows illustrative climate characteristics in Urumqi. Of note is the restricted length of growing season (120–180 frost-free days, depending on district and altitude), with consequent risk of crop timings and maturity. 153 Figure A5.3: Annual Precipitation, Xinjiang Source: Oregon State University, Spatial Climate Analysis Service. Table A5.6: Illustrative climate characteristics, Xinjiang. (Location: Urumqi) Latitude: 43.7825° Longitude: 87.586388888889° Elevation: 1 220m Prc. Tmp. Tmp. Tmp. Grnd Rel. Wind Month Prc. Prc. cv Wet mean max. min. Frost hum. Sun (2m) ETo ETo mm/m mm/d % days °C °C °C days % % m/s mm/m mm/d Jan 6.5 0.2 99.4 7.6 -13.7 -8.4 -19.0 31.0 75.1 53.8 1.1 9.3 0.3 Feb 7.4 0.3 101.2 6.0 -11.6 -6.1 -17.2 28.3 74.1 58.9 1.3 11.2 0.4 Mar 16.2 0.5 103.6 6.8 -1.7 3.4 -6.8 26.7 67.4 59.8 1.8 37.2 1.2 Apr 27.1 0.9 73.7 6.1 9.3 15.5 3.1 9.7 46.8 64.7 2.5 99.0 3.3 May 29.6 1.0 77.1 5.5 16.2 22.7 9.8 2.1 41.4 65.3 2.7 158.1 5.1 Jun 35.7 1.2 71.2 7.0 20.7 27.1 14.4 0.7 41.2 65.5 2.5 177.0 5.9 Jul 28.1 0.9 75.8 7.9 23.1 29.5 16.7 0.3 40.0 68.6 2.4 195.3 6.3 Aug 22.0 0.7 90.4 6.1 21.8 28.5 15.2 0.4 38.6 72.0 2.3 179.8 5.8 Sep 23.0 0.8 77.1 5.2 15.9 22.4 9.5 2.1 42.4 73.2 2.2 123.0 4.1 Oct 20.4 0.7 82.2 4.5 6.9 12.7 1.2 13.9 54.9 70.9 1.8 65.1 2.1 Nov 14.2 0.5 80.7 6.1 -3.1 1.7 -7.9 27.2 70.3 57.4 1.4 24.0 0.8 Dec 9.4 0.3 86.7 7.7 -11.1 -6.2 -16.0 30.9 75.9 47.2 1.0 9.3 0.3 Total 239.6 1 088.3 154 Key: Prc.= precipitation; cv= coefficient of variation; Tmp= temperature; grnd= ground; rel. hum = relative humidity; ET= evapotranspiration. Source: FAO Climate Information Tool (www.fao.org). Weather stations in Xinjiang The Xinjiang Weather Administration operates 106 weather stations, of which 19 participate in the international exchange and 54 participate in the national exchange. In addition to stations operated by the Weather Administration, there are also weather stations operated by other agencies. For instance, Xinjiang Production and Construction Corp has some 40 stations. In addition, and because of the importance of snow to the irrigation network, the hydrology bureau of the region operates its own weather stations, focused on snow monitoring, as well as 132 river stations measuring water volume and quality. Remote sensing is used to monitor snowfall. Identification of crops and weather hazards Key exposures to crops: A matrix of main cultivated crops and key perils exposures is shown in Table A5.7. It is understood that key risk exposures in the province are as follows: • Water shortage: Meeting the crop water requirements for irrigated crops in Xinjiang is complex. Available water resources are the annual snowfall (leading to river flow that feeds irrigation networks) and some wells from which water is pumped. Annual rainfall supplements water requirements but is not a very significant contributor, given the low rainfall and relative evapotranspiration rates. River flow is also affected by temperatures. High temperatures give rise to higher rates of snow melt and available irrigation water. Unexpectedly cold conditions can reduce available water. Further, the balance between retained water, irrigation water, and underground water varies by region, including location within the irrigation network and water allocation. The result is that there can be localized areas where there are periods of water stress during the growing cycle. Further, there are periods of specific sensitivity to water stress (for example, the milk phase in wheat), and sufficient water during the end of April and early May is particularly important. Natural rainfall also affects forage available for livestock in herding areas. • Flood and rainstorms: During the spring there is the potential for flooding from overflow of the rivers that feed the irrigation network, when fast flowing (flash flood) rivers descend the foothills of the mountain. Summer rainstorms, often on mountain slopes, can cause summer flooding. Flooding can also affect livestock grazing areas (such as Elie). Flood risk is strongly influenced by river engineering measures. Cropping is not permitted inside flood lines, where flood risk is considered high. • Dry heat and strong winds: Periods of dry heat and winds cause desiccation of crops. Hot, dry winds are a well known phenomenon in the north of China, and 155 the nation is making major efforts to limit desertification. These winds do not have marked impact on cotton, but they do affect wheat and fruit. Sand accumulation is a specific problem that affects only those production areas close to deserts and basins. Wind is most serious in the east of the region, sandstorm in the southern part. Hot dry winds can also be a problem at grape harvest periods. It is noted that the largest cause of claims (60 percent) arising to PICC for cotton insurance is from wind damage. • Prolonged cold temperatures and strong winds (“cold currentâ€?): One of the most devastating hazards, especially for cotton, is periods of prolonged low temperatures during the growing season. Insufficient growing degree days (GDD) for satisfactory growth of cotton results in poor yield, poor quality (downgrading), and late development. Recent years with cold temperatures were 1997, 2001, and 2003. A major problem of cold current is likely to occur approximately once every 10 years. • Spring frost: In annual crops, especially cotton, late spring frost occurring after normal planting time can necessitate replanting—a frequent experience. For fruit, there can be damage from spring frost at the time of flowering (for example, end of April and early May). • Autumn frost: Due to the short growing season and continental climate, there is potential for damage to mature crops by frost prior to harvest. Vulnerable crops are cotton and corn. • Excessive rainfall: A specific risk for grapes at the point of harvest is excessive rain, which causes splitting. Generally, low rainfall conditions favor production of cotton, where sunlight and heat enhance cotton quality, and untimely rainfall (especially when followed by overcast conditions) causes downgrade. Similarly, tomatoes (an important crop in Xinjiang) suffer from excess rainfall in countries where machinery is used for harvest. Hand labor for harvesting may reduce the risk associated with autumn rainfall for this crop. Excess rainfall was no a major risk in Xinjiang. • Hail: Hail is a significant problem in Xinjiang. Although it is infrequent, it can be very severe. Individual hailstorms, a result of major summer storms, can cover a large area and be intense. On a visit to Manasi district, it was advised that 60,000 mu of crops (out of 400,000 mu in the district) were totally lost in a single storm. Some rocket-based hail protection is practiced in the region. • Pest and disease: Cotton pests (bollworm, aphids, red spider mite), common to all cotton-growing countries, were cited as a key hazard. Integrated pest management is practiced. Although pest incidence is linked to weather conditions, the relationship is complex and is not considered further here. Key exposures to livestock: Extensive livestock herding is practiced in the region, principally in the northwest and north. High snow cover on wintering areas can cause mortality of livestock. The extent of this problem, and available statistics on livestock mortality, was beyond the scope of this investigation. 156 Table A5.7: Xinjiang—Main Crop Types and Weather Hazards Total Total Area Province Oil-bearing Orchard of selected Cultivated MAIN CROP TYPES Rice Wheat Corn Soyabean Tubers crops Cotton Fiber crops Sugar Beet Tobacco Vegetables crops crops area Provincial area ('000 ha) 66.8 686.4 518.0 83.4 23.1 215.5 1136.9 29.4 59.9 0.5 185.6 396.2 3401.7 3592.3 Peril Water stress Prolonged high temperatures Typhoon Flood Excess rain Waterlogging Prolonged low temperature Spring frost Autumn frost Winterkill Hail Wind Pest Disease Sea water storm surge = perils notified as of significant importance in the province = specific crop hazards noted Note: this table only includes crop/peril combinations advised to the mission. It is not intended to be exhaustive and can serve as a checklist for other combinations. Sources: Authors. 157 Organizational Considerations for Index Products in Xinjiang Two insurance companies operate in Xinjiang’s agricultural sector. China United was formed as the insurer from the Xinjiang Production and Construction Corps and has expanded its activities into several other provinces in China. Its client base in Xingjiang is XPCC. The Peoples Insurance Company of China (PICC) has marketed agricultural products to farmers outside of XPCC for many years. Currently insured products and perils Both China United and PICC have strong technical experience derived over many years of underwriting crop insurance in Xinjiang. China United offers multiple lines of insurance products, with agriculture insurance providing 25 percent of all premium revenue. China United offers insurance coverage for cotton, wheat, crop fire, crop hail, agriculture inputs, harvest, forest and tree fire, maize, grape, tomato, pear, and apple. Livestock insurance includes products for dairy cows, deer, bovine, and fish. The principle crop insured is cotton. Although China United offers MPCI, the company has adopted a mix of percentage-damage and yield-based loss assessment. Partial losses, which cannot be assessed soon after occurrence of the loss, are assessed at harvest on a yield basis. Since drought is not a major insured peril, most damage is from specific events, which favors percentage-damage assessment. PICC Xinjiang offers more than 60 insurance products that cover major crops and livestock (such as cotton, wheat, melons, fruits, dairy cows, sheep, and poultry) in Xinjiang. Major perils in Xinjiang include fire from April to May; hail from June to August; chilling in April or October; and drought. PICC underwrites insurance against wind damage, flood, chilling, hail, and frost. Drought damage is not covered due to the difficulties in precise loss adjustment. Insurers in Xinjiang operate a percentage-damage loss-assessment system, for which procedures have been developed in line with requirements of the loss-assessment network. Although the crop types and perils insured in Xinjiang are complex, the simplification of percentage-damage loss assessment allows the involvement of persons less highly qualified than those needed for yield assessments. The products are technically strong (for example, dividing the policy and loss-assessment procedures into growth stages in the crop cycle). Index insurance products which may be feasible in Xinjiang Several issues relevant to the role of index insurance in Xinjiang have been identified: • The climate of Xinjiang is complex, with multiple microclimates reflecting the topography and geography, ranging from desert to mountain. • Xinjiang is an extremely large region, with wide differences that make generalization difficult. 158 • Xinjiang is an irrigated region, where water management and risks associated with availability of water for agriculture are very complex. Although some weather risks such as snowfall do determine water availability, there is a very high basis risk due to the complexities of the local irrigation network. There are many variables, often human- related (such as irrigation infrastructure), which make index insurance an unfavorable option for the risk of water availability. This means that the major type of weather-index insurance that is tested in other countries (drought insurance based on rainfall-deficit index) is not feasible in Xinjiang. • The two insurance companies operating in Xinjiang have established a very sound base for the operation of traditional crop-insurance products, including the assessment of losses. Existing products would need to be evaluated against any index opportunity. In particular, the feasibility of indexation of wind damage to cotton or other crops is very difficult to foresee, and it has not been attempted elsewhere. This hazard is currently covered by insurers, including provision for a trigger of minimum wind speed before claims can be considered. • Possible opportunities for index products will require careful research and development to meet local circumstances. Cotton GDD-index insurance: A major risk to cotton growing is persistent cold temperature, giving rise to inadequate growing degree days. Because cotton agronomists have a good understanding of the need for adequate GDD, it should be possible to design an index product to pay in the event of inadequate accumulated GDD over a given period. (It is understood that this hazard is not currently covered by insurers in Xinjiang, although claims statistics suggest that it may have been covered previously.) Although this concept was not discussed, it could be considered for further research. An advantage of an agreed value payout for this hazard is that conventional measurement of damage is very difficult. However, it is not known at this point how easily a product could be designed to minimize basis risk. It is noted that basis risk for temperature is less than for other hazards, such as rainfall, because it is spatially more homogenous. It is also noted that considerable development work would be required to understand the parameters needed within such a product, considering such variables as varieties, frost- free days, and crop-production management. This product would meet the criteria of addressing a risk in the key crop of economic importance, and it was the risk said to be the most serious individual hazard facing cotton growers. Since it is not covered by existing insurance policies, there would be no conflict with the continuation of existing traditional coverage. Technical features of prototype cotton GDD-deficit index: GDD is an established index. The appropriate agronomic thresholds, limits, and increments would need to be investigated in Xinjiang. The principle of GDD is set out in Box A5.10. Assuming the establishment of correct temperature parameters and period of growth to record a day as a GDD, an index of GDD can be set up. Historical temperature data are then used to set the GDD threshold and limit, and to set options of the increment (tick size) to represent loss of production. 159 Box A5.10: Growing Degree Days—Definition GDD is a common index used in the agricultural sector, similar to HDD (heating degree day) and CDD (cooling degree day) in the energy sector. It is a measurement of the growth and development of plants (both crops and weeds) and insects during a growing season. Organisms that cannot internally regulate their own temperature are dependent on the temperature of the environment to which they are exposed. Development of an organism does not occur unless the temperature is above a minimum threshold value, known as the base temperature, and a certain amount of heat is required for development to move from one stage to the next. The base temperature varies for different organisms and is determined through research and scientific considerations. A growing degree day is calculated by the following equation: Daily GDD = max (0, T average – L ) …where T average = (T max – T min) / 2 …and where L is the baseline temperature and T average is the daily mean temperature, defined as the average of the daily maximum (T max) and minimum (T min) temperatures. If this average is greater than the threshold temperature L, the GDD accumulated for that day is the threshold temperature minus the daily average temperature. If the daily average temperature is less than the base temperature, then the GDD for that day is zero. Adding the GDD values of consecutive days gives the accumulated GDD over a specific period. Accumulated GDD are a good proxy for establishing the development stages of a crop, weed, or insect and can give an indication of the development and maturity of a crop, or the proper scheduling of pesticide or herbicide applications. Measuring the amount of heat accumulated over time provides a physiological time scale that is biologically more accurate than are calendar days (Neild and Newman 2005), and specific organisms, pest or plant, need different accumulated GDD to reach different stages of development. Comparing accumulated GDD totals with those of previous years can show whether a normal amount of heat energy has been made available to a crop. In general, assuming the availability of adequate moisture supplies, the total GDD received by the end of the growing season are often related to crop yield, and therefore GDD can be a good index for crop production. The cumulative-temperature index can be used to establish a relationship between GDD and production, and ultimately with a producer’s revenues. Sources: Authors. A prototype term sheet for a GDD index is shown in Table A5.8. 160 Table A5.8: Example Term Sheet for GDD Index Product feature Description Crop type Cotton a Index Growing Degree Day Index Policy duration (days) [to be determined] [based on the key growth and vulnerable period to be covered] Phase This index would operate with a single phase in which the GDD would be measured Threshold (“triggerâ€?) (GDD) [to be determined] Limit (GDD) [to be determined] Increment (“tickâ€?) (Yuan/GDD) [to be determined] Maximum policy limit (Yuan) [to be determined] b Reference weather station [Name of reference weather station] Maximum policy limit (Yuan) [to be determined] Comment The key variable to be researched would be the appropriate threshold temperature within the GDD formula for establishing the GDD (see box 5.11.) Notes: a. Since a total loss of the crop from inadequate GDD seems unlikely, the policy could be valued in terms of the increment, so as to compensate for loss of income rather than loss of production costs, although there are no fixed rules and production cost could be used as the basis to fix the increment. b. The reference weather station is the meteorological station where observations are recorded and on which claims payouts will be calculated Sources: Authors. Other Microlevel Index Insurance Product Opportunities in Xinjiang Opportunities for additional microlevel index insurance products, such as the following, exist in Xinjiang. Short time-window products As noted, Xinjiang includes a great diversity of crop types, microclimates, and potential exposures to weather events. An innovative insurer could research whether specific weather events could be covered by a weather-index product, which is very flexible. Specific opportunities were not explored, but possible options to be investigated further include: • Spring frost (low-temperature) index designed to meet the need to replant cotton, with the financial amount set to cover the additional costs of seed, fertilizer, and labor; 161 • Early autumn frost affecting any crop types with special quantity or quality vulnerabilities (for example, fruits and vegetables); • Excess-rainfall weather-index insurance, to protect against grape splitting at harvest. Regarding hazards and timeframe, the farmer may perceive coverage that is limited to specific windows and for a single hazard as providing only narrow coverage, but if that hazard can cause severe loss, however infrequent, a weather index can still be an effective way of covering that specific hazard. Unlikely Exposures for Index Insurance in Xinjiang Some exposures, such as the following, are very unlikely to have an application in Xinjiang. Exposures with high basis risk The index approach becomes problematic when the vulnerability of the crop is not predictable, or where the hazard is highly complex or localized. Examples of hazards that are not candidates for indexation are water stress caused by inadequate irrigation water supply, hail in any crop type, and flood. Any development of area-yield index would be dependent on assessment of the quality of district-yield data. Conclusions for Index Insurance in Xinjiang Opportunities for index insurance instruments in Xinjiang are somewhat limited due to the complexities of the climate and dependence on complex irrigation systems. In particular, the risk of drought or water stress is high, owing to dependence on irrigation water supplies that are largely under human management. However, some specific opportunities were identified. In particular, cold temperature conditions could make cotton production a candidate for a GDD-index product. Existing insurance products Many hazards (such as hail, pest, disease, and flood) cannot be addressed by a weather- index approach. In common with observations in other provinces, there continues to be an important role for traditional insurance products, and particularly named-peril crop insurance products where loss assessment can be based on percentage-damage loss assessment (for example, hail). There are potential benefits of promoting exchange of best international practices with provinces in China, through (1) organizing central training and technical-support activities in China, and (2) promoting networking with specialist agricultural insurance organizations internationally. A conclusion is that, although index insurance is likely to have an important place in the future mix of products in China, traditional crop insurance holds an equally (or more) important role. 162 China’s available manpower to carry out simplified percentage-damage loss assessment is an important consideration in supporting traditional crop insurance. Implications of offering index product in parallel with existing product Since China United and PICC have existing operations in Xinjiang and enjoy established product lines, they would need to consider how an index product might complement the existing products. A difficulty noted in Xinjiang was that the previous network for product distribution, outside of XPCC, had collapsed. Previously, marketing outlets for crops (cotton ginning companies) had acted as a sales channel for insurance, and in the case of PICC, the existing rural village infrastructure served as a collective purchasing channel for farmers, with a single policy per village. Simplified index insurance, where feasible for technical reasons, might enable distribution of the product through less-skilled intermediaries. However it was noted that even for hail insurance, which cannot be indexed, farmers were not able to buy individual-farmer insurance. Livestock-index insurance It was not possible to investigate the opportunity for an index insurance program for livestock, although it was noted that there are some regions with climatic conditions and pastoral herding similar to that of Mongolia, where such index insurance has been established. In particular, the Mongolian program relies on an official livestock-mortality surveying system, which could be investigated further for Xinjiang. Macrolevel opportunities for index reinsurance No specific opportunity was identified to transfer macrorisk for catastrophe hazards in Xinjiang. The most appropriate product opportunities for macrolevel index insurance are for drought and typhoon, which are not issues in Xinjiang. Organization of product development and implementation Meetings took place with several organizations with relevant expertise for development of index products, which could participate with insurers in Xinjiang. These are Xinjiang Meteorology Bureau; Xinjiang Agricultural Department. 5.5. Hainan Hainan Province is an island in the South China Sea off the coast of Guangdong Province. It covers an area of 34,000 square kilometers and includes 42.5 percent of the nation’s total tropical land. 163 Overview of Agriculture in the Province Hainan produces a variety of crops and is one of the agriculture centers in China. The average plot (per capita) of 0.48 hectares is used for agriculture, forestry, animal husbandry, and fisheries. Favorable weather conditions allow farmlands to be cultivated any time of the year, and many plants can yield two or three crops a year. Seven categories of land use in Hainan are farming, rubber planting, tropical crops, forestry, livestock breeding, aquaculture, and other purposes. Currently, 3,152 million hectares of land in Hainan Island are cultivated, and 260,000 hectares (about 90 percent of which are potential farm lands) remain untouched. Major tropical crops include rubber plants, coconut palm, oil palm, betel palm, pepper, sisal hemp, lemon grass, cashew, and cocoa. Crops with the widest distribution and highest yield in Hainan are rice, wheat, sweet potato, cassava, taro, maize, Chinese sorghum, millet, and beans. Among the industrial crops are sugarcane, hemp, peanut, sesame, and tea. Hainan also cultivates commodity fruits such as pineapple, litchi, longan, banana, plantain, citrus, mango, watermelon, parambola, and jackfruit. In addition, more than 120 kinds of vegetables are grown, and forestry accounts for almost half of the agricultural output. Climate and Production Systems In general, the island is characterized as tropical, with average temperatures of 22–26 degrees Celsius all year, but some parts of the province experience temperatures as cold as 0 degrees Celsius. Annual precipitation also varies by area, with the western regions averaging 1,000 millimeters per year and the southeast regions, which are subject to frequent typhoons, averaging 1,500–2,600 millimeters. Figure A5.4 shows the annual rainfall distribution in the region, and Table A5.9 shows illustrative climate characteristics in Haikou. Figure A5.4: Annual Precipitation, Hainan Source: Oregon State University, Spatial Climate Analysis Service. 164 Table A5.9: Illustrative Climate Characteristics, Hainan. (Location: Haikou) Latitude: 20.039166666667° Longitude: 110.34333333333° Elevation: 29m Prc. Tmp. Tmp. Tmp. Grnd Rel. Wind Month Prc. Prc. cv Wet mean max. min. Frost hum. Sun (2m) ETo ETo mm/m mm/d % days °C °C °C days % % m/s mm/m mm/d Jan 24.8 0.8 93.2 12.3 17.8 21.0 14.6 0.0 83.4 37.7 2.2 65.1 2.1 Feb 36.8 1.3 87.8 12.6 19.0 22.3 15.8 0.0 85.6 32.2 2.3 61.6 2.2 Mar 52.5 1.7 86.3 11.6 22.2 25.8 18.6 0.0 83.0 36.8 2.4 93.0 3.0 Apr 101.7 3.4 76.0 12.2 25.7 29.6 21.8 0.0 81.1 42.8 2.3 114.0 3.8 May 191.1 6.2 50.6 15.7 28.1 32.1 24.2 0.0 79.0 54.4 2.0 145.7 4.7 Jun 241.7 8.1 51.7 17.2 28.8 32.6 25.1 0.0 79.4 52.9 1.9 141.0 4.7 Jul 204.0 6.6 58.9 14.6 29.1 33.1 25.2 0.0 77.8 60.9 2.0 158.1 5.1 Aug 241.2 7.8 48.6 18.1 28.4 32.1 24.7 0.0 81.1 54.9 1.7 139.5 4.5 Sep 309.8 10.3 56.1 17.6 27.3 30.6 24.1 0.0 82.4 54.4 1.7 123.0 4.1 Oct 200.2 6.5 74.4 14.9 25.3 28.4 22.3 0.0 81.5 52.8 2.1 108.5 3.5 Nov 102.4 3.4 110.4 11.8 22.2 25.1 19.3 0.0 80.3 47.2 2.2 84.0 2.8 Dec 39.6 1.3 103.2 11.4 19.1 22.2 16.0 0.0 80.8 44.2 2.1 71.3 2.3 Total 1 745.8 1 304.8 Key: Prc.= precipitation; cv= coefficient of variation; Tmp= temperature; grnd= ground; rel. hum = relative humidity; ET= evapotranspiration Source: FAO Climate Information Tool (www.fao.org). Weather stations in Hainan There are five internationally reporting stations, two of which are offshore. No data were obtained on the total number of stations, but the meteorological bureau and hydrological service both operate stations. The hydrological bureau operates river gauges. Typhoon Typhoon is the single largest risk affecting Hainan island. The impact of typhoon is both from strong winds and from high levels of rainfall, leading to flooding. Because of the high flood exposure, the hydrological bureau is well developed and has responsibilities for flood control and flood warning. The disaster-control unit also maintains a database of 165 past impacts from typhoon and drought. Major investment in water infrastructure includes dams and river management. Main areas susceptible to river flooding are Qionghai, Wanning, Wenchang, Sanya, Ledong, Dongfang, Danzhou, and Lingao. Typhoon occurrence within the province is officially recorded in an annual handbook. However, at a regional level typhoon recording is the responsibility of the Shanghai Typhoon Institute, which has responsibility under WMO for data management and warnings for the Indian and Pacific oceans. The typhoon season is from May to November, with a peak in August and September. On average there are 3.1 typhoon landings on the island per year, and if tropical depression is included the figures increase to 7.7 per year. Typhoons cause major losses approximately every five years. Major typhoons occurred in 1966, 1973, 1989, and 2005. Typhoon Dawai was the worst since 1973 and caused agricultural losses estimated at RMB 8 billion. Box A5.11: Terminology of Typhoon Classification Used in China Classification—Strength of Typhoon (1986–December 1988) Strong typhoon: Maximum wind speed exceeds 32.6 meters per second (equivalent to Degree 12). Typhoon: Maximum wind speed exceeds 17.2–32.6 meters per second (equivalent to Degree 8–9). Tropical low pressure: Maximum wind speed exceeds 10.8–17.1 meters per second (equivalent to Degree 6–7). Classification—Strength of Typhoon (January 1989–December 2005) Typhoon: Maximum wind speed exceeds 32.6 meters per second (equivalent to Degree 12). Strong tropical storm: Maximum wind speed exceeds 24.5–32.6 meters per second (equivalent to Degree 10–11). Tropical storm: Maximum wind speed exceeds 17.2–24.4 meters per second (equivalent to Degree 8–9). Tropical low pressure: Maximum wind speed exceeds 10.8–17.1 meters per second (equivalent to Degree 6–7). Comparison with Terminology Used in Hurricane Zones Tropical Storm: 39–74 miles per hour (17–33 meters per second) Hurricane category 1: 74–95 miles per hour (33–42 meters per second) Hurricane category 2: 96–110 miles per hour (43–49 meters per second) Hurricane category 3: 111–130 miles per hour (50–58 meters per second) Hurricane category 4: 131–155 miles per hour (59–69 meters per second) Hurricane category 5: >155 miles per hour (>69 meters per second) Source: U.S. National Hurricane Center. 166 Identification of crops and weather hazards Key exposures to crops: Table A5.10 presents a matrix of cultivated of main crops and key peril exposures. Typhoon is the overriding hazard of importance in Hainan. In common with experience in other countries, its impact on agriculture is complex, widespread, and unpredictable, due to the following features: • Winds associated with typhoons differ in strength from the center of the eye to its periphery. They can be very intense (with rapid fall in velocity from the center), or very wide in their swath. Their forward speed and direction are variable, and they lose strength when moving from sea to land. • Rainfall associated with typhoons is highly unpredictable. Tropical depressions can have heavy rainfall but low wind speeds. Impact of heavy rains is dependent on runoff, topography, soils, flood plain, erosion characteristics, and drainage— and can cause flood or waterlogging. • Crops differ in degree of vulnerability to wind and rain, with some being more damaged by one or the other. Vulnerability is likely to depend on the stage in the cropping cycle, and for perennial crops it depends on whether the damage is to the productive asset (for example, the tree) or the annual crop. • Because multiple agricultural and other sectors are affected simultaneously, impact to the economy is cumulative. Typhoons can affect all crop types in Hainan. Typhoon winds cause damage to rubber trees and banana plantations. Typhoon Dawei, for instance, caused total agricultural loss of RMB 8 billion in September 2005, when 80 percent of bananas were destroyed, 100 percent of rubber seriously affected, grain harvest reduced by 272,400 tons, 14,100 large livestock killed; and 223,300 tons of aquaculture lost. Flooding or waterlogging caused damage to rice and vegetable crops, and many other crop types were impacted. Typhoon risk is not uniform throughout the country. The east of the island (the main direction of arrival of typhoons) has a higher frequency. Banana production has been moved to the west of the island, largely as a result of typhoon exposure. Drought is also a problem in Hainan in some years, in spite of high annual rainfall. Drought conditions prevailed in 2005 and 2006. Key exposures to livestock and aquaculture: No detailed information on livestock losses was gathered. However, as noted above, 14,100 livestock were lost in typhoon Dawei. Aquacultural losses also occurred, mainly in coastal sites. Damage to coastal bunds protecting shrimp farms is a common feature of such production systems. Freshwater aquaculture is also affected by flooding. Data on natural hazard damages: Annex 2, above, discusses the risk assessment in Hainan Province. Data on the area covered and area affected by natural hazards indicate, in decreasing order of magnitude, the hazards of drought, flood, frost, and hail. 167 Data on insured losses: PICC Hainan has insured only typhoon winds as a hazard and so does not have a record of other causes of loss. Due to the frequency of loss, and vulnerability of crops (especially banana), PICC has experienced difficulties in maintaining financial viability in its crop portfolio. Typhoon overrides all other causes of loss on the island in relative importance. 168 Table A5.10: Hainan—Main Crop Types and Weather Hazards Total Oil- Total Area Province bearing Fiber Sugar Orchard of selected Cultivated MAIN CROP TYPES Rice Wheat Corn Soyabean Tubers crops Peanuts crops Cane Tobacco Vegetables crops crops area (*1) Banana * Rubber * Provincial area ('000 ha) 334.7 0.0 13.7 13.7 109.3 47.4 43.9 0.2 69.9 0.1 161.0 163.8 957.7 826.9 35 470 Peril Drought Prolonged high temperatures Typhoon Flood Excess rain Waterlogging Prolonged low temperature Spring frost Autumn frost Winterkill Hail Wind Pest Disease Sea water storm surge = perils notified as of significant importance in the province = specific crop hazards noted Note: this table only includes crop/peril combinations advised to the mission. It is not intended to be exhaustive and can serve as a checklist for other combinations. * Plantations are believed not to be included in total provincial "cultivated area" data. Sources: Authors. 169 Organizational Considerations for Index Products in Hainan PICC has underwritten agricultural insurance in the province. As part of the 2006 initiative to revitalize agricultural insurance on the island, the new pilot projects for rubber and banana will be underwritten by a coinsurance pool, led by PICC. Currently insured products and perils The main products currently insured are banana plantations, rubber trees, and forestry. Coverage is provided on a damage-based policy form, for wind, where field loss assessment is carried out after a damaging event. These conventional damage-based insurances are well adapted to the wind hazard and are likely to be viable due to the availability of straightforward in-field procedures for counting damaged trees (in the case of rubber) and damaged plants (in the case of banana). Index Insurance Products That May Be Feasible in Hainan The following issues are relevant to the role of index insurance in Hainan: • Typhoon hazard is of overriding importance in Hainan. Exposures typical of tropical climates also include occasional drought, pest, and disease, but apart from typhoon, and associated rainfall and flooding, these other hazards are of much less significance. The tropical climate is favorable to crop production. • Many crops are grown under assisted or partial irrigation, which makes the drought risk more complex than a simple shortfall of rain in certain seasons. These crops are dependent on stored water, river water, and pumped underground water. • Hainan has a wide range of tropical crops, including tree fruits and tropical vegetables, as well as irrigated rice. The major plantation crops are banana and rubber, for which existing products exist, or are being developed. • Typhoon is complex in its impact. First, it affects a wide variety of crop types, each with its own profile of vulnerability. Second, the wind and rainfall associated with typhoon is highly variable. Third, floods associated with the typhoon are primarily dependent on the rainfall, and may be flash flood, inundation flood, or in some regions, coastal flood. Hence, modeling the localized impacts from a typhoon is very complex, and this high level of basis risk fundamentally affects the applicability of typhoon-index insurance as an instrument to offer protection to small farmers. These considerations lead to the conclusion that there is very limited scope for index insurance products at the microlevel in Hainan. Possible products for further investigation are shown below, but these products do not address the key exposures faced in Hainan, and would play only a minor role in risk management in the province. 170 Drought-index insurance Drought is a risk to rainfed crops. However, no detailed investigation was possible into the extent to which the crops at risk were fully dependent on rainfall, or the frequency and severity of drought seasons. Drought-index insurance is the most widely adopted index insurance type, however, and it is possible that there is an opportunity to develop a product for such crops in Hainan. Short time-window products As noted, there is a high diversity of crop types, micro-climates, and potential exposures to weather events, in Hainan. The weather-index product is very flexible, and an innovative insurer could research whether it could cover specific weather events. Specific opportunities were not explored, but possible options to be investigated further include excessive rainfall damage to vegetable crops. Specific windows during the growth cycle where vegetable crops are vulnerable to damage from high rainfall could be considered for index products. Rainfall triggers would be set as rainfall excess of a given threshold in a 1-day, 2-day, or longer window. Investigation of such a product would follow discussions with farmers growing the crop type, and with agronomists. Such a rainfall window could also include rainfall originating from typhoons. Unlikely Exposures for Index Insurance in Hainan Some exposures, such as the following, are very unlikely to have an index insurance application in Hainan. Exposures with high basis risk The index approach becomes problematic when the vulnerability of the crop is not predictable or when the hazard is very complex, or is localized. Examples of hazards that are not candidates for indexation are hail in any crop type and flood. Any development of area-yield index would be dependent on assessment of the quality of district-yield data. Macroindex Opportunities in Hainan Macrolevel opportunities for index reinsurance: There is some potential for transferring major typhoon risks. Macrolevel typhoon-index insurance might be used to transfer risk as a reinsurance product, in the event that conventional reinsurance is limited or not available. Such macrolevel reinsurance would carry significant basis risk for the insurance company, because the payout to the insurer would be based on measurement of the typhoon strength at its closest point of approach to the island, rather than being based on the actual claims incurred by the insurer from its policyholders. The macrolevel typhoon concept is new for agriculture, but macrolevel index for property insurance is starting to develop as an instrument in hurricane-prone areas, and where very high concentrations of property values exist. In the World Bank, the Caribbean Catastrophe Risk Insurance Facility is being developed to allow island states to obtain 171 immediate central cash-flow funding immediately after a major hurricane event. The client for such an insurance facility is an aggregator of risk, ultimately a provincial or national government. However, the client could also be an insurance company with a portfolio of agricultural risks. Conclusions for Index Insurance in Hainan Some index products have a place in agricultural insurance in Hainan, and others are problematic, as the following summaries explain. Microlevel index opportunities There are only limited opportunities for index insurance in Hainan. Typhoons are the key exposure in Hainan, and microlevel typhoon-index insurance is not considered feasible, as in other countries exposed to similar risks, due to the high local-level basis risk associated with occurrence of typhoons, variability of wind damage, and variability of rainfall (and associated flooding). At present, no product for flood indexation exists, although it is being researched. Macrolevel opportunities for index reinsurance There is some potential for transferring major typhoon risk from an insurance company. However, this would be a second-best alternative to conventional reinsurance, due to basis risk. Existing insurance products The existing insurance products being modified for banana and rubber sectors in Hainan rely on simplified damage-based field assessment, which is a viable adapted product type that can be implemented under Chinese conditions. It is important to note that this type of product overcomes the basis risk problems that would be associated with typhoon-index insurance at the micro level. Measurement of the local damage caused by winds associated with typhoon, on the crop type concerned, offers the best way of targeting claims toward those affected. Organization of product development and implementation Meetings took place with the Meteorological Bureau and Hydrological Bureau— organizations with expertise relevant to development of index products, and which could participate in developing such products. 5.6. Shanghai Shanghai, China’s most important port and commercial center, is situated on the Huangpu River at the mouth of the Yangtze River. One of four municipal cities in China, Shanghai has the same official standing as a province. It covers 6,341 square kilometers 172 and includes 3 counties and 17 urban districts (covering 2,057 square kilometers). The municipality also includes 30 islands in the Yangtze and along the coast. Overview of Agriculture in the Province In Shanghai, the agricultural GDP represents less than 2 percent of total GDP. In 2002, Shanghai generated a total of RMB 8.824 billion in agriculture, with plants, forest, livestock farming, and aquaculture representing 41.6 percent, 3.3 percent, 35.7 percent, and 19.3 percent of output, respectively. In 2002, livestock farming in Shanghai generated RMB 8.348 billion, 5.3 percent lower than the previous year. Dairy cows generated 277,200 tons of milk from 60,000 cows. The city also raised 24 million head of special poultry, among which 5.11 million are meat pigeons and 3.69 million are wild ducks. The mechanization level of agriculture has increased in Shanghai’s suburbs. More than 90 percent of fields use agricultural machinery, and 10 million hectares (1.59 million mu) of paddy rice are harvested by machinery. The rate of rice breeding by machine is 25.4 percent, and the rate of wheat breeding by machine is 29.6 percent. The municipality also produced 350,000 tonnes of tree fruit and 950,000 tonnes of watermelon, along with other high-value crops such as squash and vegetables. Vegetables are grown in all seasons, varied according to season. Vegetable production uses greenhouses as well as fields. Although agriculture is a small part of Shanghai’s overall economy, it is understood that it is a high priority of the provincial authorities to maintain and foster agricultural production, particularly of high-value products, to ensure security of supply for Shanghai’s urban population. Although interprovince trade is much increased, provincial authorities place importance on maximizing locally grown produce. Climate and Production Systems Shanghai lies in the coastal zone, with a temperate climate that rarely has freezing temperatures, and rainfall of about 1100 millimeters. The topography is flat, and production is on low lying areas of land. There is a heavy reliance on drainage of these areas, with a high investment in drainage canals and flood-control infrastructure. Heavy engineering prevents the Yangtse River from flooding. Figure A5.5 shows the annual rainfall distribution in the region, and Table A5.11 shows illustrative climate characteristics in Shanghai. 173 Figure A5.5: Annual Precipitation, Shanghai Source: Oregon State University, Spatial Climate Analysis Service Table A5.11: Illustrative Climate Characteristics, Shanghai. Latitude: 31.247777777778° Longitude: 121.4725° Elevation: 5m Prc. Tmp. Tmp. Tmp. Grnd Rel. Wind Month Prc. Prc. cv Wet mean max. min. Frost hum. Sun (2m) ETo ETo mm/m mm/d % days °C °C °C days % % m/s mm/m mm/d Jan 44.6 1.4 60.1 10.8 4.4 7.9 1.0 15.9 73.9 41.6 2.2 37.2 1.2 Feb 62.9 2.2 56.0 12.2 5.2 8.6 1.9 12.9 75.4 37.3 2.3 39.2 1.4 Mar 84.8 2.7 45.5 13.8 8.8 12.5 5.2 6.8 75.9 37.7 2.5 58.9 1.9 Apr 105.1 3.5 43.0 14.8 14.5 18.4 10.7 0.4 77.3 40.1 2.4 81.0 2.7 May 117.5 3.8 44.4 15.2 19.4 23.2 15.7 0.0 78.1 41.1 2.2 105.4 3.4 Jun 156.2 5.2 46.4 15.1 23.5 26.8 20.2 0.0 81.7 38.6 2.2 111.0 3.7 Jul 126.0 4.1 66.9 11.7 28.0 31.4 24.7 0.0 80.7 52.8 2.2 142.6 4.6 Aug 126.6 4.1 67.0 10.9 28.0 31.3 24.7 0.0 80.0 58.7 2.3 145.7 4.7 Sep 155.7 5.2 62.7 13.0 24.0 27.2 20.8 0.0 80.0 45.5 2.1 105.0 3.5 Oct 59.4 1.9 88.4 11.6 18.7 22.3 15.2 0.0 76.6 46.8 2.0 83.7 2.7 174 Nov 53.3 1.8 70.4 10.9 13.0 16.8 9.2 1.4 76.1 46.8 2.1 57.0 1.9 Dec 38.5 1.2 82.0 10.4 6.8 10.7 2.9 11.4 74.0 47.2 2.1 43.4 1.4 Total 1 130.6 1 010.1 Key: Prc.= precipitation; cv= coefficient of variation; Tmp= temperature; grnd= ground; rel. hum = relative humidity; ET= evapotranspiration Source: FAO Climate Information Tool (www.fao.org). Weather stations in Shanghai There are 10 principal weather stations, 1 located in each of the 10 counties. There are also 190 rainfall stations, of which 70 are automated, and some also report wind. These latter stations have only recently been installed. Doppler radar is used for storm detection. Identification of crops and weather hazards Key exposures to crops: A matrix of cultivated of main crops, and key-perils exposures is shown in Table A5.12. Key risk exposures in the province are the following. • Typhoon is considered the key hazard affecting the province. Typhoons may occur from June to September, bringing high rainfall, increasing water levels, and causing flooding or waterlogging. Impact of typhoons is a mix of wind damage, heavy and prolonged rainfall, flood, and waterlogging. Particularly vulnerable crops are field scale vegetables (rainfall and waterlogging) and fruits approaching harvest (wind damage to quantity of quality). Rice may be affected according to the growth stage (flowering being most sensitive) and duration of any flooding. Greenhouses are mainly affected by strong winds. Severe typhoons occur once in about every 10 years. Major damage to crops and aquaculture occurred during typhoon Maisha on August 5, 2005. • Flooding is principally caused by typhoon, but is also influenced by upriver flooding and worsened by prolonged rainfall conditions. Major improvements have been made to the drainage infrastructure (including pumping stations), substantially reducing the flooding risk. Waterlogging or flooding is now a more localized risk. • Continued low temperatures, especially if accompanied by cloudy conditions, can affect water-melon and vegetable production at critical points in the growth cycle. It could also affect the maturity of rice. • Very high temperatures were reported to cause damage to vegetable crops (or rice at milking and flowering stages), but rarely. • The rice crop is vulnerable to pest and disease. • Hail is a rare phenomenon in Shanghai. Key exposures to livestock: There is a significant portfolio of insured livestock. These risks are not considered as indexable, and are not discussed further here. 175 Data on natural hazard damages: Anxin Insurance Company’s losses since 2001 are from waterlogging and flood (63 percent, of which 53 percent originate from typhoon and 13 percent from intensive storm), 15 percent from pest and disease (in rice), 12 percent from low temperature (in fruit and vegetable), and 8 percent from other causes. It is notable that no claims have been paid in paddy rice from waterlogging or flood in this period, although some wind damage is recorded. 176 Table A5.12: Shanghai—Main Crop Types and Weather Hazards Total Total Area Province Oil-bearing Orchard of selected Cultivated MAIN CROP TYPES Rice Wheat Corn Soybean Tubers crops Cotton Fiber crops Sugar Cane Tobacco Vegetables crops crops area Provincial area ('000 ha) 111.8 21.9 4.2 8.7 0.9 31.4 1.1 0.0 2.8 0.0 139.9 28.5 351.2 404.4 Peril Drought Prolonged high temperatures Typhoon Flood Excess rain Waterlogging Prolonged low temperature Spring frost Autumn frost Winterkill Hail Wind Pest Disease Sea water storm surge = perils notified as of significant importance in the province = specific crop hazards noted Note: this table only includes crop/peril combinations advised to the mission. It is not intended to be exhaustive and can serve as a checklist for other combinations. Sources: Authors. 177 Organizational Considerations for Index Products in Shanghai Anxin Insurance Company is the sole agricultural insurer in the province. It was formerly a fund operated by the municipal government. Its staff were mainly employed by PICC, and business was transferred to Anxin from PICC when it was formed. In this respect there are very strong linkages between Shanghai municipal government and Anxin. Currently insured products and perils Products underwritten by Anxin include paddy rice, vegetables in greenhouses, and melons. Vegetables in the open are not insured. Principal hazards insured are typhoon, rainstorm, wind, hail, and lightning. Anxin reported that its clients had difficulty interpreting complex insurance policy procedures, which aroused the company’s interest in the possible application of more simplified index policies. Specific issues arose in setting of insured yields—for example, inclusion of catastrophe years in a calculation of a rolling average yield for insurance purposes. Anxin also insures a major livestock portfolio, including policies for named epidemic diseases. Government compensation for compulsory slaughter would be delivered to farmers via Anxin. Existing index insurance proposals Anxin have identified two prototype index products, following discussions with farmers. Index insurance products that may be feasible in Shanghai Typhoon hazard has an overriding importance in Shanghai, as in Hainan. Apart from typhoon, and associated rainfall and flooding, other hazards are of much less significance. Shanghai’s rice crop is automatically insured, including for typhoon damage, under existing policies. Typhoon is complex in its impact. All crops are exposed, but the main crops of concern to Anxin are vegetables grown in the open, and fruits. The impact of typhoon is from wind, rainfall, and flood. Winds associated with a typhoon vary widely. They may be intense at the central eye, dropping off rapidly away from the center, or for some typhoons, have a much wider influence. Rainfall associated with typhoon is highly variable. In Shanghai, flooding associated with typhoons is less significant, due to effective drainage, but most field crops have potential for extensive damage in a severe typhoon. Hence, modeling the localized impacts from a typhoon is very complex, and this high level of basis risk fundamentally affects the applicability of typhoon-index insurance as an microinstrument to offer protection to small farmers. These considerations lead to the conclusion that index insurance products have limited scope at the micro level in Shanghai. However, Anxin is experimenting with the design 178 of two new index products, which are adapted to specific windows of crop exposure. Other windows of exposure to specific weather hazards could be identified. A reinsurance product is one possible application of macrolevel typhoon-index insurance, to assist with the transfer of macrorisks, in the event that conventional reinsurance is limited or not available. Such macrolevel reinsurance would also carry basis risk for the insurance company. Microlevel Index Insurance Product Opportunities in Shanghai Following are descriptions of two prototype microlevel index insurance products: Anxin prototype product 1: Watermelon excess-rainfall index Anxin developed a prototype product, following discussion with watermelon farmers regarding their main exposure. The product would be targeted to a 30-day period from mid to late July onwards. Excessive rainfall (for example, more than 300 millimeters in the period, or more than 100 millimeters in any 24 hours) would be established as a threshold. Details are to be researched, but insured values could be based on either production costs or a proportion of annual revenue. Anxin is researching the appropriate thresholds and limits, and whether an incremental payment scale, or total loss payment, is appropriate. Anxin prototype product 2: Greenhouse Watermelon excess-rainfall and cloud-cover index Anxin is researching the type of damage suffered by watermelon when there is insufficient sunlight to mature the crop, which occurs when there is protracted rain. Anxin research is trying to determine whether the prolonged cloudy periods are reflected in rainfall data, or inversely, if lack of sunshine hours are correlated with lack of crop maturity. A possible payout scale could then be developed, with incremental payouts based on cloudy conditions (rainfall) or sunshine hours. Other short time-window products: The above two product examples illustrate the flexibility of the weather-index product, which an innovative insurer, such as Anxin, can research to determine whether specific weather events could be covered by a weather- index policy. Additional specific opportunities were not explored, but possible options to be investigated further include: • Prolonged cold periods impacting either vegetable production or rice production at sensitive stages of the crop calendar. • Other specific windows of excessive rainfall affecting crops, particularly sensitive types of fruit or vegetables. Macroindex Opportunities in Shanghai Macrolevel opportunities for index reinsurance: There is some potential for transferring major typhoon risks. However, as noted above, the target client could not be the 179 individual farmer, due to the high basis risk of damage caused by typhoon, rainfall, and flood at the local level. The macrolevel typhoon concept is new for agriculture, but macroindex for property insurance is starting to develop as an instrument in hurricane-prone areas, and where very high concentrations of property values exist. In the World Bank, the Caribbean Catastrophe Risk Insurance Facility is being developed, to allow island states to obtain immediate central cash-flow funding immediately after a major hurricane event. The client for such an insurance facility is an aggregator of risk, ultimately a provincial or national government. However, it could also be an insurance company with a portfolio of agricultural risks. It was also noted that Shanghai is not a typical province, in that the municipal authority’s financial capacity allows it to provide significant financial support, through premium subsidy and through reinsurance protection, to Anxin. Thus, pressures on risk transfer through conventional reinsurance, or macroindex reinsurance, is reduced. Anxin has successfully negotiated reinsurance for their portfolio. Unlikely Exposures for Index Insurance in Shanghai Some exposures, such as the following, are very unlikely to have an index insurance application in Shanghai. Exposures with high basis risk The index approach is problematic where the vulnerability of the crop is not predictable, or where the hazard is very complex, or is localized. Examples of hazards that are not candidates for indexation are hail in any crop type and flood. Any development of area- yield index would be dependent on assessment of the quality of district-yield data. Conclusions for Index Insurance in Shanghai Some index products have a place in agricultural insurance in Shanghai, and others are problematic, as the following summaries explain. Microlevel index opportunities There are only limited opportunities for index insurance in Shanghai. Typhoons are the key exposure. Micro-level typhoon-index insurance is not considered feasible, as in other countries exposed to similar risks, due to the high local-level basis risk associated with occurrence of typhoons, variability of wind damage, and variability of rainfall (and associated flooding). At present, no product for flood indexation exists, although it is being researched. 180 Macrolevel opportunities for index reinsurance There is some potential for transferring major typhoon risk. Existing insurance products Existing insurance products offered by Anxin have inherited features of the product, and the loss-assessment system, operated in the past by PICC. Anxin indicated several difficulties in the current loss-adjustment system and operation of the products, and hopes that index insurance may offer a more simplified basis of providing insurance coverage to farmers. The main portfolio consists of rice paddy insurance. Unfortunately, microlevel index products do not offer the opportunity for wide replacement of existing conventional insurance products for crops. This difficulty arises because typhoon-index insurance cannot readily be adapted to microlevel indexation, due to problems of basis risk. However, there are opportunities to develop very targeted index insurance products for key periods of exposure in particular crop types. Anxin has already identified two such index products and produced prototypes. Organization of product development and implementation Meetings occurred with several organizations, in addition to Anxin, with relevant expertise for development of index products. These are Shanghai Meteorological Bureau Department of Hydrology; and Typhoon Research Institute Center for Meteorological Research. 181 Annex 6: Reinsurance Capacity Issues This annex examines the Chinese reinsurance market and the Chinese agricultural reinsurance market in particular, providing brief profiles of four providers of commercial and public-private agricultural reinsurance programs. Later sections discuss global markets for agricultural insurance and reinsurance, and address capacity issues for agricultural reinsurance. 6.1. Chinese Reinsurance Market Prior to 2002, the Chinese reinsurance market was closed to foreign competition. The market was controlled by the state-owned China Reinsurance (Group) Company, ChinaRe. Formed in March 1999 from the former PICCRe, ChinaRe enjoyed a monopoly of the domestic reinsurance market until 2002. It prospered from the 20 percent compulsory cession of direct insurers’ premiums, which in 2001 accounted for approximately 94 percent of ChineRe’s premium income. Since China became a member of the World Trade Organization (WTO), the 20 percent compulsory cession requirement has been gradually phased out, dropping to 15 percent in 2003, 10 percent in 2004, 5 percent in 2005, and finally being abolished in 2006. Local insurance companies are now free to cede business to ChinaRe as they wish. In 2004, ChinaRe’s reinsurance premium income amounted to RMB 19.985 billion, of which 29.9 percent came from commercial reinsurance premiums. In 2004, ChinaRe’s total assets amounted to RMB 26.36 billion, an increase of 7.48 percent over the 2003 figure 21. SwissRe and MunichRe received approval to apply for a full branch licence in July 2002, thus ending ChinaRe’s monopoly. Since 2003, both of these international reinsurers have been granted full national reinsurance branch licences, permitting them to write reinsurance business throughout China. At end 2005, there were six locally licensed “professionalâ€? reinsurers in China: ChinaRe, PICC Property & Casualty Re, China Life Re, MunichRe, SwissRe, and General CologneRe 22. When the compulsory cessions to ChinaRe were phased out in December 2005, they were replaced by a series of new regulations, beginning December 1, 2005, which require that any direct insurance company wishing to place facultative or treaty reinsurances with international reinsurance markets must first offer at least 50 percent of the reinsurance cession to at least two of the locally licensed reinsurers 23 listed above. Any balance on 21 ChinaRe Annual Report 2004. 22 Other reinsurers who have sought branch approval for branch licences in Shanghai or Beijing include TransatlanticRe, EmployersRe, and Lloyd’s. 23 Willis 2005. “International Alert Reinsurance Rules Tighten in China.â€? (10) (November). 182 the company’s retention and the local reinsurance acceptance can be reinsured to reinsurers outside of China. Local foreign insurance companies are not allowed to transact reinsurance business with their affiliated companies. Before achieving approval for a full operating license in China, foreign companies must meet three primary criteria: (1) 30 years of experience as an insurer or reinsurer, (2) $5 billion in total assets, and (3) a representative office in China for two years. Separate regulations apply to foreign insurance and reinsurance brokers wishing to operate in China. 6.2. Chinese Agricultural Reinsurance Market The recent opening up of the Chinese insurance market is resulting in significant changes in agricultural reinsurance, as described below. Chinese Agricultural Reinsurance Arrangements Prior to 2005 The review shows that during the 1980s, 1990s, and early 2000s PICC, SAIC, China United Property Insurance Company (CUPIC), and Anxin implemented their agricultural insurance programs without any formal reinsurance protection except the compulsory cessions to ChinaRe. In the case of PICC Hainan and Xinjiang, agricultural underwriting losses incurred at a provincial level were absorbed by the property and casualty (P&C), division, which includes agriculture 24. In Shanghai Municipality, Anxin 25 has insured crops and livestock since 1982, again without any formal reinsurance protection. For livestock, Anxin’s coverage includes both epidemic diseases and government slaughter order. The company has operated under an agreement with the Shanghai government that provides for the municipality to compensate catastrophe losses in livestock. A similar agreement with the government applied to Anxin’s crop insurance program until 2005, when it was replaced by a formal stop-loss reinsurance treaty. In Heilongjiang Province, between 1993 and 2004 and prior to the formation of Sunlight Insurance Company, SAIC, there was no formal reinsurance program, and in years of catastrophe losses that exceeded the mutual’s premiums and claims reserves, the following mechanisms were applied: (1) the mutual adjusted down the claims for all 24 According to PICC Xinjiang, PICC Group also operates an internal reinsurance program for losses arising out of a single event that exceeded RMB 10 million. It is understood that this cover is applicable to P&C business. 25 Anxin Agricultural Insurance Company, Shanghai, was formerly PICC Shanghai. 183 farmers on a pro rata basis, and/or (2) the provincial government stepped in with disaster relief payments for the excess losses 26. In Xinjiang, CUPIC has operated for 20 years without any form of external reinsurance protection, and continues to do so (as of 2006). The company’s internal risk-management measures 27 applicable to its agricultural insurance business include: • The creation, out of operating surpluses, of an agricultural insurance fund to pay for excess losses in future years 28; • In the event of severe disasters, if the premium of the year is insufficient to cover claims, it will be supplemented by risk funds accumulated from past years; • If this is still insufficient, 30 percent of CUPIC’s profits from commercial P&C business in that year can be allocated to pay for the agricultural insurance losses; • Finally, if this is still insufficient, funds will be financed by the XPCC farms and regiments to aid the effected farmers. According to CUPIC’s results during their 20 years of operation from 1986 to 2005, operating deficits (claims plus operating expenses in excess of premiums) were experienced in 5 years (1986, 1987, 1989, 1999, and 2001), and underwriting profits were generated in 15 years. The company also reports that up to end of 2005, approximately RMB 100 million of risk funds have been accumulated to pay for catastrophe losses 29. The extent to which provincial government and the XPCC may have assisted CUPIC to indemnify excess losses during this 19 year period is not known. Currently CUPIC does not purchase agricultural reinsurance, because the price and conditions of reinsurance do not meet their needs for risk management. Nevertheless, CUPIC identifies a requirement for some form of proportional or nonproportional reinsurance protection for its agricultural portfolio, and the company has entered into preparatory talks with reinsurers 30. New Agricultural Reinsurance Arrangements Since 2004 Since 2004, the agricultural reinsurance market in China has gradually been opened up to competition, and several companies are now involved in the purchase of international stop-loss reinsurance and coreinsurance arrangements with provincial governments and their provincial finance bureaus. Key features of the new agricultural reinsurance arrangements are detailed below. 26 According to SAIC, the provincial government compensated excess losses on three occasions in the 11- year period, namely 1994, severe waterlogging; 1998, flooding; 2002 low temperature and frost losses; and that the total excess-of-loss payments amounted to RMB 280 million. 27 CUPIC (undated), An Introduction to Agricultural Insurance in Xinjiang Production and Construction Corps. 28 It is not known whether CUPIC’s insurance fund has been financed solely out of agricultural insurance operating surpluses, or whether this has received capitalization from external sources. 29 CUPIC, Ibid, page 49. 30 CUPIC Ibid, page 59. 184 International Stop Loss Treaty Reinsurance protection has been purchased by Sunlight and by Anxin on their crop-only portfolios for the past two years (2005 and 2006). In both cases, these local insurers are buying very small layers of stop-loss reinsurance protection equivalent to approximately 50 percent of losses in excess of 90 percent of gross net premium income. This stop-loss coverage is considered to be inadequate to protect against these companies’ catastrophe PML exposures, leaving them very exposed to major losses. It is understood, however, at least in the case of Anxin, that the municipality government will continue to provide ad hoc (ex post) compensation for losses in excess of their stop-loss reinsurance limits. In Zhejiang Province, the new 2006 PICC-led agricultural coinsurance pool is also buying international stop-loss reinsurance protection on its combined crop, livestock, forestry, and aquaculture portfolio, in this case for losses in excess of 100 percent up to 500 percent of GNPI, as part of a carefully structured and layered reinsurance program involving both private and public sector reinsurance. Parallel to the opening up of the international agricultural reinsurance market, CIRC is promoting pool coinsurance programs led by PICC in several provinces, including Zhejiang (since 2006) and Hainan (planned for 2007), under which the local provincial governments and their finance bureaus have agreed to provide free (no reinsurance premiums are charged) reinsurance protection under a proportional or coreinsurance arrangement for losses in excess of 200 percent GNPI up to 500 percent of GNPI. This public-private risk financing represents a significant development in the Chinese agricultural insurance and reinsurance market in 2006. In China, the leading international agricultural reinsurance brokers are providing an essential service to link local insurers with specialist international agricultural reinsurers 31. 6.3. Individual Company Agricultural Reinsurance Programs This section looks at four sources of agricultural reinsurance programs: Sunlight Insurance Company, SAIC, in Heilongjiang; Anxin Agricultural Insurance Company in Shanghai; Zhejiang Province’s public-private program; and Hainan Province’s proposed public-private program. Sunlight Insurance Company, SAIC, Heilongjiang SAIC first purchased commercial stop-loss treaty reinsurance for its crop MPCI portfolio in 2005. Coverage was for a single stop-loss layer of 30 percent of losses in excess of 90 percent of GNPI. (See the figure in Box A6.1.) In 2005, the loss ratio was 67 percent, and therefore there was no claim to the stop-loss treaty. The coverage was placed through Aon Reinsurance Brokers with a panel of five national and international reinsurers. 31 The list of international reinsurance brokers that specialize in agriculture includes, among others, Aon, Benfield, Guy Carpenter, and Willis. 185 In 2006, the crop stop-loss treaty program was modified to provide greater protection through a two-layer program: (1) 30 percent of losses in excess of 90 percent GNPI, and (2) 20 percent of losses in excess of 120 percent GNPI. The panel of reinsurers was expanded to seven national and international reinsurers. The SAIC crop stop-loss reinsurance treaty was also free of claims in 2006 32. The SAIC crop MPCI program is basically sound, is technically rated (average premium rates in 2005 were about 8 percent), and is being carefully implemented by SAIC. However, it is questionable whether the stop-loss reinsurance protection SAIC currently purchases—which is capped at 140 percent loss ratio—is adequate. The analysis in Annex 2 suggests that the 1-in-100-year probable maximum loss for crops grown in Heilongjiang may vary between 350 percent loss ratio for rice and 480 percent loss ratio for maize. It has been recommended that SAIC conduct a formal assessment of the PML exposure on its crop MPCI portfolio and use the results of this analysis to refine its stop- loss treaty limits accordingly. It is recognized, however, that if local government continues to provide SAIC with free compensation for catastrophe losses in excess of the current stop-loss treaty limits, there may be little incentive to purchase additional stop- loss treaty protection. 32 SAIC advised loss ratio of 84 percent for crops (SAIC May 2007, Agro_Insurance_in_Heilongjiang_updated). 186 Box A6.1: Sunlight Insurance Company Crop Stop-Loss Treaty (2005 and 2006) S u n lig h t In s u ra n c e C o m p a n y R e in s u ra n c e T re a ty L a y e rin g : 2 0 0 5 S in g le L a y e r fo r lo s s e s e xc e s s 9 0 % u p to 1 2 0 % G N P I* 2 0 0 6 2 la ye rs L a y e r 1 : L o s s e s e xc e s s 9 0 % G N P I u p to 1 2 0 % G N P I L a y e r 2 : L o s s e s e xc e s s 1 2 0 % G N P I u p to 1 4 0 % G N P I 2 0 0 5 S to p L o s s 2 0 0 6 S to p L o s s % GNPI % GNPI S u n lig h t S u n lig h t R e te n tio n R e te n tio n 140% 2 n d S to p L o s s R e in s u ra n c e 2 0 % xs 1 2 0 % G N P I 120% 1 s t S to p L o s s R e in s u ra n c e 120% 1 s t S to p L o s s R e in s u ra n c e 3 0 % xs 9 0 % G N P I 3 0 % xs 9 0 % G N P I 90% 90% S u n lig h t S u n lig h t P rim a ry R e te n tio n P rim a ry R e te n tio n L o s s e s u p to L o s s e s u p to 90% G N PI 90% G NP I 0% 0% *G N P I G ro s s N e t P re m iu m In c o m e Source: Authors’ calculation. Anxin Agricultural Insurance Company, Shanghai Municipality In 2005, Anxin purchased the following reinsurance protection on its agricultural portfolio: • 5 percent Quota Share Treaty with ChinaRe, • For the first time in 2005, a commercial international stop-loss reinsurance treaty on their crop-only portfolio, 95 percent retention. The crop stop-loss treaty was placed by Aon Reinsurance Brokers with a panel of local and international commercial reinsurers. The treaty was placed on a total sum insured (TSI) and loss cost basis, as opposed to the more conventional gross net premium and loss ratio basis 33. The stop-loss treaty was divided into two equal layers: the first layer, 33 If a reinsurer has reservations concerning the technical adequacy of original gross premium rates, it is more likely to structure a stop-loss reinsurance treaty on a sum-insured and loss-cost basis. The loss cost is equal to Claims divided by Total Sum Insured and expressed as a ratio of percent. 187 0.55 percent of TSI in excess of 2 percent of TSI; the second layer, 0.55 percent of TSI in excess of 2.55 percent of TSI, with each layer equivalent to about RMB 11 million. For comparative purposes, the coverage can be considered as a protection of two equal layers of about 25 percent of GNPI in excess of 90 percent of GNPI, thereby providing protection to Anxin for losses up to a loss ratio of about 140 percent. Full details of the 2005 stop-loss layering and reinsurance pricing are shown in Box A6.2. (For simplicity, it does not include ChinaRe 5 percent Quota Share.) In 2005, Anxin experienced severe typhoon losses from Typhoons Matsa (August 5, 2005) and Kahnun, with total paid claims in crops of RMB 69.3 million, equivalent to 3.36 percent of TSI, or a loss ratio of 160 percent. Consequently, both layers of the stop- loss treaty were totally consumed, giving rise to a reinsurance claim of RMB 20.2 million 34, equivalent to a loss ratio to reinsurers of about 775 percent. In 2006, Anxin was able to renew its crop stop-loss treaty with the same layering. However, the reinsurance premium rates were significantly increased, by about 80 percent. Key issues relating to Anxin’s reinsurance program include the following: • The company is buying very limited reinsurance protection on its crop treaty program and no reinsurance protection on its livestock portfolio. • The company has not conducted any formal modeling of its catastrophe typhoon exposure for crops, or the epidemic disease exposure on its livestock portfolio. For crops: Anxin charged an average rate of 2.1 percent in 2005, and its losses ranged from a low 49 percent loss ratio for paddy rice to a high 373 percent loss ratio for fruit, and overall loss ratio of 160 percent for crops. According to local experts Typhoon Matsa was a 1-in-10-year event 35. It is very likely that a 1-in-100-year typhoon could cause Anxin to lose 10 percent of its crop portfolio, which would imply a loss ratio of about 500 percent. This suggests that the company should be purchasing additional stop-loss reinsurance protection over its current limit of about 140 percent loss ratio. It is understood, however, that the municipal government may intervene to cover excess losses that exceed the stop-loss limit of 140 percent loss ratio. For livestock: Anxin charged an extremely low average rate of 1.1 percent across its 2005 portfolio, ranging from a low of 0.2 percent for canine to a high of 1.8 percent for diary cattle and freshwater aquaculture. Anxin’s livestock insurance policy insures both 34 Anxin 2006, Comprehensive Information on Anxin Agricultural Insurance Company (including underwriting manual and loss adjustment manual). 35 Typhoon Matsa (also named Maisha) reached a maximum strength of 165 kilometers per hour (105 miles per hour) on August 4 as it crossed Japan. Matsa weakened steadily to a minimal typhoon as it approached the Chinese coast near Yuhuar in Zhejiang Province on August 5 and thereafter rapidly weakened to tropical-storm status. In Shanghai Municipality the main insured damage from Matsa was caused by excess rain and flood in vegetables and fruit trees. 188 class-A epidemic diseases 36 and government slaughter order. Anxin has not conducted any formal disease modeling in livestock, but it is suggested, based on international experience, that under a catastrophic epidemic disease outbreak losses would likely exceed 10–20 percent of total sum insured, which would imply a loss ratio of at least 1,000 percent. Anxin advises that under current arrangements it has no need to purchase reinsurance protection on its livestock program, because the company has an agreement with the Shanghai government to settle losses in the event of a catastrophic epidemic disease outbreak. Anxin’s livestock program, with its current extremely comprehensive disease coverage, is unlikely to attract international reinsurance support. (See further discussion below.) It is therefore important to note that in 2007, Anxin intends to review the range of insured perils on its livestock programs with a view to bringing these into line with internationally accepted terms and conditions. 36 For example, Anxin insures against avian influenza in poultry and foot and mouth disease (FMD) in pigs and dairy cattle. 189 Box A6.2: Anxin, 2005 Crop Stop-Loss Reinsurance Treaty A n xin S h an gh ai 2005 C rop S top L oss R ein su ran ce T reaty A n xim 95% R etention E stim ated T otal Sum Insured R M B 2,000,000,000 1,900,000,000 E stim ated P rem ium Incom e (G N P I) R M B 45,000,000 42,750,000 R einsurance T reaty Layering: LAYER P R IO R IT Y P rio rity R M B L IM IT L im it R M B 1S T SL 2 % O F TSI 38 ,0 00 ,0 00 0.55 % O F TSI 10 ,45 0,00 0 2N D SL 2.5 5% O F TSI 48 ,4 50 ,0 00 0.55 % O F TSI 10 ,45 0,00 0 20 ,90 0,00 0 T otal S um Insured P rem iu m A N X IN R E T E N T IO N 3.10% 140% 2N D S T O P LO SS 0.55% T SI xs 2.55% T S I R M B 11 M IL L IO N 2.55% 115% 1ST ST O P LO SS 0.55% T SI xs 2% T SI R M B 11 M IL L IO N 2.00% 90% A N X IN R E T E N T IO N R M B 40 M IL L IO N Source: Authors’ calculation. Zhejiang Province: Public-Private Risk-Layering Reinsurance In Zhejiang, PICC is the leader of an agricultural insurance pool that was established in 2006 on a three-year trial basis to underwrite P&C business, crops, livestock, forestry, and aquaculture risks. The pool covers catastrophe losses (such as typhoon, flood, and drought) for crops, plus selected named diseases and epidemic diseases for pigs and poultry. The program attracts local government premium subsidy support of RMB 25 million. The initiative in Zhejiang provides an example. In 2006, private and public sectors in China began providing a comprehensive layered risk-management and risk-transfer program. Key features include: • CIRC required that all insurers that operate in Zhejiang Province participate in the Zhejiang Agricultural Insurance Pool. The pool is led (in 2006) by PICC, with a 60 190 percent share; the remaining 9 or 10 coinsurers take much smaller shares in the pool. This is the first agricultural insurance pool project in China, and its participants benefit from the principle of risk sharing. • PICC is appointed to manage the insurance pool and receives 20 percent ceding commission to cover operating expenses. • The provincial government, through the financial bureau, is providing catastrophe reinsurance protection to the pool on a proportional basis for losses in excess of a 200 percent loss ratio, up to 500 percent loss ratio, in two layers (as shown in Box A6.3). • Layer 1: 100 percent excess of 200 percent loss ratio, and the government is coreinsuring 50 percent of the program; • Layer 2: 200 percent excess of 300 percent loss ratio, and the government is coreinsuring 67 percent of the program. Under this program it is understood there is agreement between CIRC, the insurers, local government, and the farmers’ organizations that losses will be capped at 500 percent loss ratio, and if claims exceed this level losses will have to be adjusted down on a pro rata basis. The PICC-led pool is also purchasing private commercial stop-loss reinsurance protection on its 2006 retention for losses in excess of 100 percent loss ratio, up to 500 percent loss ratio, as shown in the second diagram in Box A6.3. The stop-loss program is placed under four layers: • Layer 1: A working layer for losses of 20 percent in excess of 100 percent GNPI; placed with a Bermudan reinsurer in 2006; • Layer 2: For losses of 80 percent in excess of 120 percent loss ratio; • Layer 3: For losses of 100 percent in excess of 200 percent loss ratio; • Layer 4: For losses of 200 percent in excess of 300 percent loss ratio. The second, third, and fourth stop-loss treaty layers are reinsured by a panel of local and international reinsurers. In 2006, the cost of the four-layer treaty coverage was 27.5 percent of the pool coinsurers retained GNPI. 191 Box A6.3: PICC Coinsurance Pool, Zhejiang Province (2006 Reinsurance Structure) PICC Led Coinsurance Pool, Zhejiang Province: 2006 Reinsurance Arrangements A. Government Proportional Reinsurance Treaty B. International Stop Loss Treaty on PICC Pool Coinsurers' retentions excess 100% upto 500% L Co-reinsurance on losses excess 200% LR up to 500% LR Losses capped at 500% loss ratio Losses capped at 500% loss ratio % GNPI by agreement provincial government % GNPI by agreement provincial government 500% 500% 4th SL on PICC PICC Pool Pool 33% share Government Retention Government 400% losses Co-reinsurance 400% Co-reinsurance 300% to 67% share 300% to 67% share 500% LR for loss ratio 500% LR for loss ratio 300% to 500% 300% to 500% 300% Government 300% Government PICC Pool Co-reinsurance 3rd Stop Loss Co-reinsurance 50% share 50% share on PICC Pool 50% share for loss ratio for loss ratio Retention for loss ratio 200% to 300% 200% to 300% 200% to 300% 200% to 300% 200% 200% PICC led 2nd Stop Loss Treaty on Coinsurance Pool PICC PICC Pool Retention Retention up to 80% xs 120% Loss ratio 200% Loss ratio 120% 1st Stop Loss Treaty 20% xs 100% 100% 100% PICC led Coinsurance Pool Net Retention up to 100% Loss ratio Source: Authors’ calculation. Hainan Province: Proposed Public-Private Risk-Layering Reinsurance in 2007 Hainan Province provides the second example of a provincial government that is proposing, through its Finance Bureau, to play a direct role in risk financing and reinsurance in 2007, with a model similar to that of Zhejiang Province. The provincial government is also planning to provide premium subsidies in 2007. Two coinsurance pools are proposed for 2007. The first, led by PICC, will insure crops (bananas and rubber trees), livestock (pigs), and forestry; the second pool, led by China Chundong Mutual 192 Insurance Association, will insure fishermen’s boats against material damage and will provide personal accident coverage to the crew of the fishing boats. The 2007 reinsurance structure is shown in Box A6.4 for banana, rubber, and swine, and for forestry, fishing boats, and personal accident. The structures are as follows: • Coinsurance pools will retain 100 percent of losses up to a loss ratio of 200 percent; • The Finance Bureau (FB) will coreinsure 50 percent of claims from 200 percent loss ratio up to 500 percent loss ratio. This reinsurance will be provided free of charge (that is, the provincial FB will subsidize 100 percent of the reinsurance premium). • For bananas, rubber, and swine, the maximum indemnity will be 500 percent loss ratio in 2007, and if losses exceed this level, all claims will have to be reduced on a pro rata basis. • For forestry, fishing boat insurance, and fishermen’s personal accident coverage, liability for settling losses exceeding 500 percent loss ratio will revert 100 percent to the pool coinsurers. Box A6.4: Hainan Province—Proposed Pools and Reinsurance Structure, 2007 Hainan Pilot Agricultural Insurance Projects 2007 Reinsurance Layering and Governm ent Co-reinsurance of Claim s 2007 Reinsurance Layering for: 2007 Reinsurance Layering for: BANANAS, RUBBER, SWINE FO REST RY, FISHING BO AT S, PERSO NAL ACCIDENT Loss Loss Ratio Finance Bureau & Coinsurers Ratio Full unlimited liability for all claims CAP on Liability at 500% loss ratio > 500% LR reverts to Co-insurers 500% 500% Pool Provincial Pool Provincial 400% Coinsurers' Finance Bureau 400% Coinsurers' Finance Bureau 50% share of 50% share of 50% share of 50% share of Claims Claims Claims Claims 300% 300% 200% 200% Co-insurer's Full Retention Co-insurer's Full Retention claims up to 200% loss ratio claims up to 200% loss ratio 100% 100% Source: Authors. 193 To date, pool coinsurers have made no decisions as to whether they will seek to purchase stop- loss reinsurance protection from international reinsurers on their retained shares in 2007. In 2007, advised premium rates on Hainan for typhoon-only coverage are 20 percent for bananas and 10 percent for rubber trees. Average rates for bananas are very high on Hainan because of the extremely high typhoon exposure on this island and the susceptibility of bananas to wind damage. The proposed 2007 stop-loss layering, with coverage up to 500 percent loss ratio, provides a high degree of protection in bananas and rubber and compares favorably with estimates of the 1-in-100-year PML for this island of about 420 percent loss ratio (range in PMLs: bananas 500 percent loss ratio; rubber 400 percent loss ratio). 6.4. Global Agricultural Insurance and Reinsurance Markets This section examines the global markets for agricultural insurance and agricultural reinsurance. Agricultural Insurance Market According to estimates provided by Guy Carpenter Reinsurance Brokers, global agricultural 37 insurance premiums were on the order of $8 billion in 2005. Of this $8 billion, North American crop MPCI and crop hail business accounted for about $5.5 billion, or 69 percent of global premium, divided between the United States ($4.6 billion, or 58 percent of total) and Canada ($0.9 billion, 11 percent of total). China is currently ranked in 10th place with about a $90 million agricultural insurance premium in 2005. (See Table A6.1.) Table A6.1: Agricultural Insurance Premiums—Top 10 Territories (2005) Territory Estimated Insurance Premium ($) % of Total U.S. 4,600,000,000 57.5% Canada 900,000,000 11.3% Spain 550,000,000 6.9% Italy 350,000,000 4.4% France 300,000,000 3.8% Indiaa 225,000,000 2.8% Germany 200,000,000 2.5% South Africa 100,000,000 1.3% Australia and NZ 100,000,000 1.3% Chinab 90,000,000 1.1% Total 7,415,000,000 92.7% a. Agricultural Insurance Corporation India estimates crop and livestock b. World Bank 2005 estimates for China are presented in the main report. Source: Guy Carpenter, July 2006. 37 This includes crop, livestock, forestry and aquaculture business. 194 Agricultural Reinsurance Market Both the traditional and the relatively new weather-index agricultural reinsurance markets are discussed below. Traditional agricultural reinsurance market Guy Carpenter estimates global agricultural reinsurance premiums at about $1.3 billion in 2005. Although this only represents about 16 percent of original premiums, bear in mind that much of the reinsurance business is placed on a nonproportional treaty basis, as opposed to a proportional quota share basis, and therefore a very much higher proportion of global agricultural insurance business is in fact reinsured. Slightly more than 57 percent of this agricultural reinsurance business originates from North America (the United States and Canada), and the remaining 43 percent is worldwide. (See Table A6.2.) For China, it is currently estimated that about $4 million of agricultural insurance premiums (less than 0.3 percent of the global market) are ceded to reinsurers. It is, however, anticipated that the volume of agricultural reinsurance originating from China will increase significantly in future years. Guy Carpenter handles approximately 50 percent of the global agricultural reinsurance premium by volume 38. Most agricultural reinsurance business is transacted on a proportional quota share treaty or nonproportional stop-loss treaty basis. The market for facultative reinsurance is small for agriculture. Table A6.2: Agricultural Reinsurance Premiums—Top 10 Territories (2005) Territory Estimated Reinsurance Premium ($) % of Total U.S. 672,000,000 50.40% Canada 95,000,000 7.13% Spain 12,000,000 0.90% Italy 150,000,000 11.25% France 60,000,000 4.50% Germany 30,000,000 2.25% India 2,000,000 0.15% South Africa 47,000,000 3.53% Australia / NZ 50,000,000 3.75% China 4,000,000 0.30% Total 1,122,000,000 84.15% Source: Guy Carpenter, July 2006; India reinsurance premium estimated by World Bank. 38 Guy Carpenter December 2006. Personal communication. 195 The agricultural reinsurance market has traditionally been dominated by a very small group of mainly European specialist reinsurers. Table A6.3 lists the leading specialist agricultural reinsurers, along with the World Bank’s best estimates of their market share of agricultural reinsurance premium. SwissRe and MunichRe are the largest agricultural reinsurers, each with about 20 percent market share, followed by three European reinsurers, HanoverRe, PartnerRe, and Scor, each with market share estimated at 5.0–7.5 percent. Overall, the eight listed companies (or markets, in the case of Lloyds and Bermuda) account for at least 70 percent of global agricultural reinsurance capacity as of 2005. A feature of recent years has been the reduction in the number of global agricultural reinsurers due to mergers and acquisitions, including SwissRe’s acquisition of GE Frankona in 2005 and Scor’s acquisition of Converium in 2007. In recent years various Bermudan reinsurers have entered the crop-reinsurance market, mainly to provide catastrophe nonproportional reinsurance. In spite of the small number of specialist global agricultural reinsurers, these companies can bring to bear significant reinsurance capacity where the terms and conditions meet their requirements. SwissRe and Munich Re are the two largest reinsurers in the world, with combined group reinsurance premiums in excess of $60 billion and AA- rating by Standard & Poor's. Other reinsurers, including PartnerRe, HanoverRe, and Lloyd’s of London, can also provide major agricultural reinsurance capacity where the business can demonstrate the required rates of return. Table A6.3: Leading Agricultural Reinsurers’ Estimated Global Share (2005)* Company % Market Share SwissRe (+ former GE Frankona portfolio) 20% MunichRe 20% Scor (+ former Converium agricultural reinsurance portfolio) 5.0–7.5% HanoverRe 5.0–7.5% PartnerRe 5.0–7.5% Bermudan Reinsurers (e.g., Axis Specialty) 5.0–7.5% EnduranceRe 5.0–7.5% Lloyds of London (e.g., Limit, Syndicate 566) 5.0–7.5% *Based on estimated reinsurance premiums $1.3–1.5 billion. Sources: Authors. These specialist agricultural reinsurers aim to earn a Return on Equity (ROE), over a full cycle, of 10–15 percent over full book of business, including proportional and nonproportional agricultural reinsurance treaties and facultative business. Typically they will seek a higher ROE of up to 20 percent on nonproportional treaty and facultative business. Equity allocations needed for agricultural business also vary among reinsurers. 196 Agricultural weather-index reinsurance market There is currently no agricultural crop weather-index insurance in China, although several companies, including SAIC and Anxin, are eager to develop weather-index insurance products. Results of a review and analysis of the opportunities for crop weather-index coverage in each selected province are presented in the main report and in Annex 5, above. The weather index reinsurance market for agriculture is quite new; four or five years ago it was largely restricted to weather-derivate instruments, as opposed to conventional reinsurance. The weather-derivatives market dates back to 1997 in the United States, since which time it has grown into a multibillion-dollar industry driven mainly by the energy sector, which typically transacts temperature indexes. According to a recent World Bank study, since 2007 more than $20 billion has been transacted through the weather-risk market and, although energy companies still constitute approximately half of this market, other potential end users include the retail, agricultural, transport, leisure, and entertainment industries 39. In the past three to five years, major advances have occurred in the design and rating of weather- index insurance products and the development of pilot commercial programs in a wide range of both developed and developing countries, including the United States, Canada, Spain, Mexico, Nicaragua, Peru, Morocco, Ethiopia, Malawi, India, and Mongolia. (Mongolia represents the first livestock-mortality index coverage to be launched to date.). Excluding the United States and Canada, the World Bank estimates the current value of global crop weather-index and livestock- index premiums at no more than $25 million. As these weather-index insurance programs have developed, the major traditional agricultural reinsurers have increasingly committed their support and reinsurance capacity to them. The list of crop weather-index reinsurers includes the world’s two largest reinsurers, MunichRe and SwissRe, followed by PartnerRe, ParisRe (formerly AxaRe), HanoverRe, Ace, XL Capital, and Tokyo Fire and Marine. 6.5. Agricultural Reinsurance Capacity Issues In China, agricultural reinsurance programs originate with the central and provincial governments and a small number of commercial providers. This section looks at both public and private sources of agricultural reinsurance. Availability of Commercial Reinsurance Capacity Due to the very small number of specialist international agricultural reinsurers, these reinsurers are able not only to set reinsurance terms and conditions, but also to influence such factors as original policy design and rating. The availability of agricultural reinsurance capacity is not a key constraint where original programs have been carefully designed, where original rates are technically sound and underwriting and loss-adjustment systems and procedures are in place, and where reinsurers can generate reasonable profits over an insurance-reinsurance cycle. In addition, reinsurers place a high value on transparency in reinsurance negotiations with their cedants and in being provided full access to original and accurate time-series insurance-history 39 World Bank 2006. Managing Agricultural Production Risk, Innovations in Developing Countries. 197 results and related insurance data and statistics. Future availability of agricultural reinsurance in China could be influenced by market experience (profitability) and rate of expansion of capacity requirements. It is likely that the initial entry of reinsurance markets into China has been influenced by strategic decisions to enter a potentially expanding market, and a desire to gain experience. However, experience in mature markets suggests that commercial reinsurance availability is not a limiting factor, provided the original programs can be demonstrated to be technically, operationally, and financially sound. Public versus Private Commercial Reinsurance The review has shown that in several provinces in the past, central and or provincial governments have provided ex post compensation for excess losses in severe claims years (for example, in Heilongjiang, Shanghai Municipality, and possibly in Xinjiang Province). The drawbacks of ad hoc compensation (recapitalization) of insurance companies following severe losses are highlighted in Paragraph 4.21 of the main report. Since 1996, the provincial governments in several provinces have also taken a direct and formal role in risk financing. In Zhejiang, both government and commercial national and international reinsurers are involved in the provision of formally structured stop-loss reinsurance to the pool agricultural crop, livestock, forestry, and aquaculture program. Conversely, in Hainan, the current proposals are for government reinsurance support alone. Pool coinsurers currently enjoy the advantage of government reinsurance protection, in that it is provided at no cost, while they have to pay commercial rates for their commercial stop-loss reinsurance protection. This review has identified a positive role for central and provincial governments in offering layered stop-loss reinsurance protection to the provincial agricultural insurance companies where a clearly defined catastrophe exposure exists (for example, typhoon coverage in Hainan, and drought and flood in Zhejiang). However, the report stresses the need for a careful balance between government-provided “freeâ€? reinsurance support and commercially priced international reinsurance, to avoid crowding out the commercial reinsurers 40. Proportional versus Nonproportional Reinsurance Currently, the three companies that purchase commercial agricultural reinsurance (SAIC in Heilongjiang, Anxin in Shanghai, and the PICC Pool in Zhejiang) are placing their business on a nonproportional stop-loss treaty basis. To date, there are no proportional (quota share) agricultural reinsurance treaties with foreign reinsurers, but some companies continue to cede small shares to ChinaRe on a proportional basis. It generally is much easier for an insurer entering into a commercial reinsurance agreement for the first time to place its reinsurance requirements under a stop-loss treaty than under a quota- share treaty. This is because the results on a stop-loss treaty do not directly follow the original program, and reinsurers are only exposed to loss when a certain threshold is reached, for example when losses exceed 100 percent of GNPI (100 percent loss ratio). A stop-loss 40 See Paragraphs 4.20 to 4.25 of the main report. 198 reinsurer’s requirements are much lower than those for a quote-share reinsurer for (1) conducting a detailed due diligence on the new company and its agricultural insurance business, and (2) providing detailed information and statistics on original rates, wordings, and loss-adjustment systems and procedures. Stop-loss treaty reinsurance issues Two companies, SAIC and Anxin, have been able to negotiate commercial stop-loss treaties with priorities of 90 percent of GNPI. This is very unusual, since international reinsurers usually set a minimum priority of 100 percent of GNPI to ensure that the ceding company is exposed to loss before the stop-loss reinsurance program starts to indemnify losses. The disadvantage for SAIC and Anxin is that they will inevitably be paying considerably higher reinsurance premium rates with a 90 percent of GNPI priority than they would if it were raised to 100 percent or 110 percent of GNPI. The SAIC crop MPCI stop-loss treaty is structured on a GNPI and loss-ratio basis, but in the case of Anxin, the crop stop-loss treaty is placed on a total sum insured and loss-cost basis. As previously noted, where a commercial reinsurer is uncertain of the adequacy of the original premium rates, it is likely to insist that any stop-loss treaty program is placed on a TSI and loss- cost basis. This report has identified a need for Chinese insurance companies to conduct more formal analyses of their crop (and livestock) PML risk exposures and to base their stop-loss reinsurance layering on such analysis. Currently, only the PICC Pool in Zhejiang has adequate commercial reinsurance and public reinsurance protection. Reinsurance for MPCI programs Many global reinsurers are reluctant to provide reinsurance for individual-grower crop MPCI programs because of the potential catastrophe exposure to drought losses, and to a lesser extent this also applies to the perils of flood and frost losses. Therefore, some reinsurers exclude crop MPCI altogether, and other reinsurers will only reinsure MPCI on a proportional or sometimes nonproportional basis if it forms part of a combined crop MPCI and crop-hail or named-peril program, where the MPCI component is restricted by imposition of sum-insured limits per crop and per risk zone or region and total sum insured limits, and where the original MPCI rates are demonstrably high and technically adequate. To date, SAIC has been able to place a small stop- loss treaty on its crop MPCI program, with restricted coverage for losses of 50 percent excess of 90 percent of GNPI. From a reinsurers’ perspective, the company’s liability is limited to a manageable and adequately priced layer, but if SAIC were to request reinsurers to provide full layering for losses in excess of 100 percent GNPI up to about 350–400 percent of GNPI, capacity requirements or reinsurance pricing could become a constraint. 199 Livestock reinsurance Currently, only the Zhejiang Agricultural Pool program includes commercial stop-loss reinsurance for livestock as part of the overall agricultural stop-loss treaty for losses in excess of 100 percent of GNPI. This review has shown that several companies, including Anxin and CUPIC, are offering Class A epidemic disease coverage in livestock as well as government slaughter order (to the extent that government compensation does not cover the full market value of animals that are culled under a slaughter order). It is noted that most commercial global reinsurers are extremely unlikely to agree to insure Class A epidemic perils at all in livestock, or will do so only with the introduction of major restrictions. It is also unlikely that any commercial reinsurer will agree to indemnify government slaughter order. In this context it should be noted that Anxin is currently reviewing the coverage provided under its livestock policies and may make some adjustments to bring these policies into line with international terms and conditions for livestock coverage. 200