A Monte Carlo investigation of the sampling behavior of conditional moment tests in Tobit and Probit models

This paper provides a Monte Carlo examination of conditional moment tests for several forms of misspecification in Tobit and Probit models. The experimental design is based on actual data taken from a study of labor supply. Our results indicate that the tests work well, in terms of size and power, for the Tobit model whereas they perform poorly for the Probit model. The major conclusion is that although there is little to choose between the various forms of the tests the degree of censoring of the latent dependent variable is crucial.