Evaluating yield models for crop insurance rating

Crop insurance performance and loss rates depend directly on underlying crop yield distributions. However, there still exists much debate about how to represent the underlying crop yield distributions. Using farm-level corn and soybean yields from 1972-2008, this study examines in-sample goodness-of-fit measures of both the whole distribution and the insurance tail to compare a set of flexible parametric, semi-parametric, and non-parametric distributions in a meaningful economic context. Simulations are then conducted to investigate the out-of-sample efficiency properties of several competing distributions. The results indicate that more parameterized distributional forms fit the data better in-sample, but are generally less efficient out-of-sample - and in some cases more biased - than more parsimonious forms which also fit the data adequately, such as the Weibull. The results highlight the relative advantages of alternative distributions, in terms of the bias-efficiency tradeoff in both in- and out-of-sample frameworks.