Misspecified Heteroskedasticity in the Panel Probit Model : a Small Sample Comparison of GMM and SML Estimators

摘要: Abstract This paper compares generalized method of moments (GMM) and simulated maximum-likelihood (SML) approaches to the estimation panel probit model. Both techniques circumvent multiple integration joint density functions without need restrict error term variance–covariance matrix latent normal regression Particular attention is paid a three-stage GMM estimator based on nonparametric optimal instruments for given conditional moment functions. Monte Carlo experiments are carried out which focus small sample consequences misspecification matrix. The correctly specified experiment reveals asymptotic efficiency advantages SML. estimators outperform SML in presence terms multiplicative heteroskedasticity. holds particular estimator. Allowing heteroskedasticity over time increases robustness with respect An application product innovation activities German manufacturing firms presented.