Using Copulas to Model Dependency Structures in Econometrics

Overview

Modeling the correlation structures of economic variables is an important part of managing financial risk. Pearson correlation matrices fully characterize joint distribution when the underlying economic variables follow a multivariate normal distribution. When the economic variables are nonnormal, copulas are needed to model the correlation structure. In fact, given a copula and the marginal distributions, one can recover the joint probability density function. This fact can be used to estimate models separately and combine them for simulation. This paper introduces advanced copula modeling capabilities in the MODEL procedure. The paper also shows how insight into the correlation structure of the copulas can be obtained by using animations produced by SAS.