A non-parametric approach to determine an efficient premium for drought insurance

Abstract

Insurance to deal with prolonged drought periods in rural Africa requires a practical method to estimate accurate premium values that minimize economic losses. We use non-parametric methods to determine the risk non-neutral insurer’s premium for drought insurance on rain-fed crops. Premium values are estimated on the basis of percentage of the expected yield losses over the potential yields. Expected yield losses are estimated based on data on the levels of rainfall, potential evapotranspiration and water-holding capacity of the soil, and water requirement of the crop. Maize crop in West Kenya, and rice crop in the Central High Plains of Madagascar are taken as case studies. To check if farmer’s choice of starting seasons affects the expected yields and the values of premium, we employ forecasted yields for two different sowing dates (October vs. November) for maize, and two different transplantation dates (November vs. December) for rice. The mean-variance (E-V), the First-Degree Stochastic Dominance (FSD), and the Second-Degree Stochastic Dominance (SSD) efficiency criteria are used to rank each pair of distributions. Results show that an insurer for maize production in Western Kenya would require a premium value between 43 and 55% of the potential yields to fully cover the loss caused by lack of rainfall. Under E-V and FSD, the two yield distributions cannot be ranked, but under SSD the yield distribution of the October-sown maize dominates that of November. For lowland rice in the Central High Plains of Madagascar, all three efficiency criteria indicate that the yield distribution of the December-transplanted rice dominates that of November and the premium values are less than 4 % of the potential yields.