Minimum Distance Estimation of Randomly Censored Regression Models with Endogeneity

This paper proposes minimum distance estimation procedures for the slope coefficients and location parameter in randomly censored regression models that are used in duration and competing risk models. The proposed procedure generalizes existing work in terms of weakening the restrictions imposed on the distribution of the error term and the censoring variable. Examples of such generalizations include allowing for conditional heteroskedasticity, covariate dependent censoring, and endogenous regressors. The estimator is shown to converge at the parametric rate with an asymptotic normal distribution. A small scale simulation study and an application using drug relapse data demonstrate satisfactory finite sample performance. JEL Classification: C14, C25, C13.