Using Stata to estimate dynamic correlated random effectsprobit models with unbalanced panels

This paper implements the estimation of dynamic probit correlated random effects (CRE) models with unbalanced panel data. The type of models we consider include a lag of the endogenous variable and other explanatory variables that are strictly exogenous. We introduce a Stata package, xtprobitunbal; this command estimates these models allowing for the unbalancedness process to be correlated with the time-invariant unobserved heterogeneity. It reduces the computational burden of the maximum likelihood (ML) estimation, while keeping its good asymptotic properties.We also introduce the command mgf_unbal to compute the marginal effects ofthe variables of the model and its standard errors. Finally, we study the estimation of CRE unbalanced panel data probit models by ML estimation and under more restrictive assumptions than the ones considered by xtprobitunbal, discussing the main problems to implement them.