Practical Guidelines for the Estimation and Inference of a Dynamic Logistic Model with Fixed-Effects

This papers shows how to simply compute one of the estimators proposed in Honoré and Kyriazidou (2000) as well as its variance. The method involves a simple reshaping of the original dataset that is then used in a weighted logistic regression with clustering. Such procedures are widely available in standard statistics or econometrics softwares. We also extend their inference results to more than two states and more than four periods. In doing so, it appears that, with more than four periods, there are two natural candidates for the estimator of the covariance matrix, but one of these significantly outperforms the other on Monte-Carlo simulations