Using copulas to estimate the coefficient of a binary endogenous regressor in a Poisson regression: Application to the effect of insurance on doctor visits.

Abstract

This paper presents a copula-based method for identifying and estimating the coefficient of a binary endogenous regressor in a Poisson regression. The method offers advantages over existing approaches. Most importantly, it relies upon standard maximum likelihood approaches, and it does not require numerical integration. Further, as part of its implementation, the method provides a convenient test for the presence of endogeneity. The empirical application investigates the effect of insurance status (a binary measure) on doctor visits (a count measure).