Simplified Estimation of a Nonlinear Simultaneous Equations Model with Censoring: A Monte Carlo Study

Abstract

We numerically investigate the estimation of a censored two-equation simultaneous equations model, and compare the results with maximum likelihood. The method of estimation is considerably simpler than maximum likelihood, which, in this model, is very computationally intensive. The method performs well, not only in terms of computation costs, as would be expected, but in terms of the quality of the estimates. We conclude that the method can be used to provide good starting values for maximum likelihood iterations, to evaluate model specification, or, possibly, to substitute for maximum likelihood in situations where maximum likelihood has a high rate of failure.