Estimation of Dynamic Programming Models with Censored Dependent Variables

Abstract

This paper considers the estimation of dynamic structural models where the decision variables are censored. We present and discuss several econometric issues and estimation methods under alternative stochastic structures of the unobservables, different potential sources of censoring, and different characteristics of the dataset (e.g., temporal dimension, frequency of corner solutions, or distribution of duration spells between two consecutive interior solutions). We use a labor demand model with kinked and lump-sum hiring and firing costs to illustrate the econometric problems and estimation methods.