Consistent Estimation of the Fixed Effects Probit Model

This paper introduces a new estimator for the fixed-effects binary choice model. The proposed method consists of eliminating the fixed effect by reworking the inequalities implied by the binary choice model and is akin in spirit to a linear panel first-difference estimator. The new method is very simple, faster than current alternatives, as shown by Monte Carlo simulations, and, more importantly, consistent. The dynamic nonlinear panel model is also covered. As an empirical application, I consider the demand for video games in Steam