A0608
Title: An enhanced neural network approach for agricultural index insurance design
Authors: Runqiu Xu - WUHAN UNIVERSITY (China)
Wenjun Jiang - University of Calgary (Canada)
Rogemar Mamon - University of Western Ontario (Canada)
Heng Xiong - Wuhan University (China) [presenting]
Abstract: Determining the proper payoff function and corresponding premium is essential to agricultural index insurance design. However, prevailing index insurance contracts encounter high basis risk and thus charge an unwilling premium. First, a non-linear payoff function structure utilizing a neural network with a multivariate weighted premium principle (MWPP) is proposed. Then the payoff function with a utility maximization problem is optimized to form the index insurance contract. Compared to traditional linear payoff schemes, neural networks capture non-linearity between utility-maximized payoff and indices, while MWPP provides viable premiums that offer more fair and accurate premium rates. Further, our approach is examined using China's grid-cell rice yield, weather, and soil data. Empirical results present that the proposed method reduces basis risk and improves insurers' utility with the more actual payoff and lower premiums.
