Partial copula methods for models with multiple discrete endogenous explanatory variables and sample selection

Abstract

We present a flexible parametric approach for models with multiple discrete endogenous explanatory variables (EEV) with finite support. The joint distributions of each EEV and structural error are modeled by using copulae and their marginal distributions, but the ones among the EEV’s are left unspecified. Our partial copula approach can be applied in any models with discrete EEV’s. It can be also used for correcting selection bias and finding average treatment effects.