Identification and estimation in a linear correlated random coefficients model with censoring

AbstractIn this paper, we study the identification and estimation of a linear correlated random coefficients model with censoring, namely, Y=max{B0+X′B,C}, where C is a known constant or an unknown function of regressors. Here, random coefficients (B0,B) can be correlated with one or more components of X. Under a generalized conditional median restriction similar to that in Hoderlein and Sherman, we show that both the average partial effect and the average partial effect on the treated are identified. We develop estimators for the identified parameters and analyze their large sample properties. A Monte Carlo simulation indicates that our estimators perform reasonably well with small samples. We then present an application.