Sustainable Area-Yield Insurance Program with Optimal Risk Pooling: A Behavior-Based Machine Learning Approach

Working paper

Xu, Y.; Pan, G.; Porth, L.; Tan, K. S.; Zhu, W.

Working paper · 2021

SSRNPDF

Abstract

Area-yield insurance offers a sustainable alternative to traditional individual-loss programs by reducing moral hazard, administrative costs, and data requirements, but its efficiency is often limited by basis risk. This paper proposes a behavior- based machine learning approach to optimally form risk pools, improving contract performance. We first determine the optimal number of risk pools through farmers’ behaviour under a utility maximization framework, then apply unsupervised spectral clustering to group farmers with similar risk histories. This framework enhances the efficiency of the area-yield insurance contract and addresses challenges of high- dimensionality and computational complexity. Using corn and soybean production at both county- and farm-level in the U.S. Heartland, we show that our method significantly reduces basis risk and farmers’ tail risk, with potential profit gains of up to $2.8 billion for corn production.

Keywords: Sustainability, Area yield, Basis risk, Crop insurance, Machine learning.

Back to Selected Working Papers