This study proposes transformations for dynamic fixed effects logit models. First, these transformations construct valid moment conditions for the case with neither explanatory variables nor time dummies. Using valid moment conditions, we can obtain the first-order condition of the conditional maximum likelihood estimator proposed by Chamberlain (1985). Next, we derive valid moment conditions based on the transformations for the cases with strictly exogenous continuous explanatory variables and/or time dummies when the number of time periods is four or more. Transformations of these moment conditions exactly coincide with a subset of those obtained by Honoré and Weidner (2020) using the functional differencing approach proposed by Bonhomme (2012).

本研究提出了动态固定效应 logit 模型的转换。首先，这些转换为既没有解释变量也没有时间虚拟变量的情况构建了有效的矩条件。使用有效矩条件，我们可以得到张伯伦（1985）提出的条件最大似然估计量的一阶条件。接下来，当时间段数为四个或更多时，我们基于具有严格外生连续解释变量和/或时间虚拟变量的情况的转换推导出有效矩条件。这些矩条件的变换与 Honoré 和 Weidner (2020) 使用 Bonhomme (2012) 提出的功能差分方法获得的那些条件的子集完全一致。