Author(s)

J. D. Woodard, N. D. Paulson, D. Vedenov & G. J. Power

Abstract

This study assesses the performance of several alternative methods for modeling dependence between random variables in the context of pricing an agricultural insurance contract with multiple underlying risk exposures. Simulation methods are used to estimate the sampling distribution of the insurance rates generated under alternative methods. The results indicate significant variability in performance across methods, and contribute to the risk analysis and insurance literatures by quantitatively assessing out-of-sample efficiency and bias trade-off among competing methods for modeling dependence in limited data scenarios. Keywords: copulas, GRP basis risk, crop insurance rating efficiency, kernel copula, Iman and Conover procedure, Phoon, Quek, and Huang procedure. 1 Introduction Interest in the modeling of dependence structures and copulas in the fields of risk analysis, financial engineering, and actuarial mathematics has increased substantially in recent years. Copulas [1] offer a more flexible, wider set of tools for modeling dependence structures in probabilistic settings than do more conventional methods—such as Iman and Conover’s [2] (IC) or Phoon, Quek, and Huang’s [3] (PQH)—but tend to be more computationally intensive, less familiar to those in the field, and more difficult for end-users to implement. Moreover, in many fields little empirical work has actually been conducted to evaluate the performance of alternative copulas and dependence modeling methods in applied settings. The practical importance of properly implementing these methods has increased as a result of the alleged misuse of certain

Keywords

copulas, GRP basis risk, crop insurance rating efficiency, kernel copula, Iman and Conover procedure, Phoon, Quek, and Huang procedure