ESTIMATION OF THE DURATION MODEL BY NONPARAMETRIC MAXIMUM-LIKELIHOOD,MAXIMUM PENALIZED LIKELIHOOD, AND PROBABILITY SIMULATORS

Failure to properly treat heterogeneity components in longitudinal ana lyses can result in an incorrect parameterization of the duration mode l. Estimation bias is not limited to duration dependence but also exte nds to the structural parameters. Our paper uses Monte Carlo methods t o examine the finite sample behavior of three estimators for this prob lem: nonparametric maximum likelihood, maximum penalized likelihood, a nd the probability simulator. Our results on the estimators' finite sa mple behavior for this class of model add to limited experimental evid ence. They highlight the estimators' computational feasibility and poi nt to their relative strengths in empirical duration modeling.