MR Microinsurance_2012_03_29.indd - International Labour ...
MR Microinsurance_2012_03_29.indd - International Labour ... MR Microinsurance_2012_03_29.indd - International Labour ...
240 General insurance For reasons that are well described in the literature, agricultural index insurance works not by insuring the household directly against shortfalls in its own income or yields, 2 but instead by insuring a direct or predicted measure of the average or typical yield losses experienced by neighbouring households in a region. An index insurance contract can be represented as an indemnity schedule that links payments to an index that predicts typical losses in the zone covered by the index. To avoid problems of moral hazard and adverse selection, it should not be possible for the index to be influenced by the insured, nor should benefits depend on which particular individuals choose to purchase the insurance. Figure 11.1 illustrates the indemnity schedule that might accompany a zonelevel yield loss predictor function built around a rainfall signal. The horizontal axis shows a rainfall index (perhaps cumulative rainfall measured in millimetres) and the vertical axis shows indemnity payments. The contract is defined by a lower and an upper strike level. When the rainfall index dips below (signalling drought), indemnity payouts begin as shown by the dashed line in the figure. Similarly, when rainfall exceeds the upper strike point (signalling flood conditions), payouts to the insured farmers are triggered again. A key question faced in index insurance is the extent to which household yield shortfalls track the index of predicted shortfalls. If the index signalled exactly a 100-kilo loss every time the yields were 100 kilos below the household’s long-term average, then index insurance would perfectly cover all risks faced by the household. The problem of course is that no index will perfectly correlate with any individual’s losses in this way. The index that predicts average losses will not perfectly track individual households’ yield shortfalls for three reasons: 1. Pure idiosyncratic risk: A single farm’s crop may suffer damage from an idiosyncratic factor such as animal or bird damage, or highly localized weather events. Different levels of pure idiosyncratic risk characterize different agro-ecological zones. In the Sahel, for example, rainfall is highly localized, creating significant variation in yield losses between neighbouring villages, or even between households in the same village. 2. Noise created by the geographic scale of the index: As the geographic zone covered by a single index increases in size, household losses will correlate less well with the insurance index. For example, a weather-based index that only has to cover households within 1 kilometre of the weather station will track household outcomes better than an index that has to cover all households within 30 kilometres of the weather station. 2 A myriad of experience has shown that trying to insure all sources of variation in agricultural outcomes for small farmers is beset by a host of problems rooted in the costs of obtaining information on small farm outcomes that render such insurance infeasible (see Hazell, 1992).
Next-generation index insurance for smallholder farmers 3. Noise created by index prediction errors: Th e average loss within a defi ned geographic zone can be measured directly with high precision (as with area yield contracts in the United States where yields are measured to a tolerance of +/– 2 per cent), or it can be predicted using weather or satellite information that is likely to be cheaper to implement, but also likely to have a larger margin of error when used to predict even the average loss. Together, these three elements create what is called basis risk, yield losses experienced by the household that are not correlated with the insurance index and are therefore uninsured by the index insurance contract. As the second two sources of basis risk are infl uenced by the design of the contract (geographic scope and exact index used), we refer to them as “design eff ects” on basis risk. Th e linear contract structure in Figure 11.1 is simple, and close variations of it have been used in several index insurance pilots, including ones in Ethiopia, India, Kenya and Malawi. However, implicit in this structure is the assumption that losses are linear in the rainfall index. Empirical analysis of the sensitivity of yields to rainfall like Carter’s (1997) West Africa work suggests that yield losses respond in a non-linear way to rainfall shortages or excesses. If this is correct, then these common linear loss contracts will have large design eff ects that unnecessarily increase basis risk. Section 11.2 discusses ways to estimate statistically optimal predictor functions that can be used to design more eff ective indices and contracts. Th e stylized linear indemnity schedule represented in Figure 11.1 is highly unlikely to be the contract structure that minimizes design eff ects. Figure 11.1 A stylized rainfall index insurance contract Indemnity payments Drought Normal Flood Sl Su Sl Su Indemnities 241 Rainfall index
- Page 212 and 213: 190 Life insurance and administrati
- Page 214 and 215: 192 Life insurance 2. A surrender t
- Page 216 and 217: 194 Life insurance be earned from o
- Page 218 and 219: 196 Life insurance four of them), p
- Page 220 and 221: 198 Life insurance 9.1 What is cred
- Page 222 and 223: 200 Life insurance 9.2 Who benefits
- Page 224 and 225: 9.2.2 Lenders 202 Life insurance Fo
- Page 226 and 227: 204 Life insurance easy product. Wi
- Page 228 and 229: Table 9.2 206 Life insurance Value
- Page 230 and 231: 208 Life insurance 9.4.1 Enhanced l
- Page 232 and 233: 210 Life insurance An enhancement f
- Page 234 and 235: 212 Life insurance 9.5.1 Period of
- Page 236 and 237: 214 Life insurance The focus of ins
- Page 238 and 239: 216 Life insurance Insurers and len
- Page 240 and 241: 218 Life insurance Box 10.1 Mapping
- Page 242 and 243: 220 Life insurance Living in the sh
- Page 244 and 245: Table 10.1 222 Life insurance Copin
- Page 246 and 247: 224 Life insurance 10.2.2 Is it ins
- Page 248 and 249: 226 Life insurance such other tangi
- Page 250 and 251: 228 Life insurance “Sold, not bou
- Page 252 and 253: 230 Life insurance While funeral pa
- Page 254 and 255: 232 Life insurance case of funeral
- Page 256 and 257: 234 Life insurance 1. Protecta Clas
- Page 258 and 259: 236 Life insurance 5. It makes busi
- Page 260 and 261: 238 11 Designed for development imp
- Page 264 and 265: 242 General insurance Once an indem
- Page 266 and 267: 244 General insurance 11.1.3 Option
- Page 268 and 269: 246 General insurance possible inde
- Page 270 and 271: 248 General insurance Conducting a
- Page 272 and 273: 250 General insurance costs and the
- Page 274 and 275: 252 General insurance the most risk
- Page 276 and 277: 254 General insurance from competit
- Page 278 and 279: 256 General insurance Th is factor,
- Page 280 and 281: 258 12 Livestock insurance: Helping
- Page 282 and 283: 260 General insurance Depending on
- Page 284 and 285: 262 General insurance cultural poli
- Page 286 and 287: 264 General insurance 12.3 Difficul
- Page 288 and 289: 266 General insurance Improved iden
- Page 290 and 291: Table 12.4 268 General insurance In
- Page 292 and 293: 270 General insurance Table 12.5 Pa
- Page 294 and 295: 272 General insurance require tappi
- Page 296 and 297: 274 13 The psychology of microinsur
- Page 298 and 299: 276 Insurance and the low-income ma
- Page 300 and 301: 278 Insurance and the low-income ma
- Page 302 and 303: 280 Insurance and the low-income ma
- Page 304 and 305: 282 Insurance and the low-income ma
- Page 306 and 307: 284 Insurance and the low-income ma
- Page 308 and 309: 286 14 Emerging practices in consum
- Page 310 and 311: 288 Insurance and the low-income ma
240 General insurance<br />
For reasons that are well described in the literature, agricultural index insurance<br />
works not by insuring the household directly against shortfalls in its own<br />
income or yields, 2 but instead by insuring a direct or predicted measure of the<br />
average or typical yield losses experienced by neighbouring households in a<br />
region. An index insurance contract can be represented as an indemnity schedule<br />
that links payments to an index that predicts typical losses in the zone covered by<br />
the index. To avoid problems of moral hazard and adverse selection, it should not<br />
be possible for the index to be influenced by the insured, nor should benefits<br />
depend on which particular individuals choose to purchase the insurance.<br />
Figure 11.1 illustrates the indemnity schedule that might accompany a zonelevel<br />
yield loss predictor function built around a rainfall signal. The horizontal<br />
axis shows a rainfall index (perhaps cumulative rainfall measured in millimetres)<br />
and the vertical axis shows indemnity payments. The contract is defined by a<br />
lower and an upper strike level. When the rainfall index dips below (signalling<br />
drought), indemnity payouts begin as shown by the dashed line in the figure.<br />
Similarly, when rainfall exceeds the upper strike point (signalling flood conditions),<br />
payouts to the insured farmers are triggered again.<br />
A key question faced in index insurance is the extent to which household<br />
yield shortfalls track the index of predicted shortfalls. If the index signalled<br />
exactly a 100-kilo loss every time the yields were 100 kilos below the household’s<br />
long-term average, then index insurance would perfectly cover all risks faced by<br />
the household. The problem of course is that no index will perfectly correlate<br />
with any individual’s losses in this way.<br />
The index that predicts average losses will not perfectly track individual<br />
households’ yield shortfalls for three reasons:<br />
1. Pure idiosyncratic risk: A single farm’s crop may suffer damage from an idiosyncratic<br />
factor such as animal or bird damage, or highly localized weather events.<br />
Different levels of pure idiosyncratic risk characterize different agro-ecological<br />
zones. In the Sahel, for example, rainfall is highly localized, creating significant<br />
variation in yield losses between neighbouring villages, or even between households<br />
in the same village.<br />
2. Noise created by the geographic scale of the index: As the geographic zone<br />
covered by a single index increases in size, household losses will correlate less well<br />
with the insurance index. For example, a weather-based index that only has to<br />
cover households within 1 kilometre of the weather station will track household<br />
outcomes better than an index that has to cover all households within 30 kilometres<br />
of the weather station.<br />
2 A myriad of experience has shown that trying to insure all sources of variation in agricultural outcomes<br />
for small farmers is beset by a host of problems rooted in the costs of obtaining information on<br />
small farm outcomes that render such insurance infeasible (see Hazell, 1992).