Corrigendum: Small Sample Methods for Cluster-Robust Variance Estimation and Hypothesis Testing in Fixed Effects Models

Abstract

Abstract Pustejovsky and Tipton considered how to implement cluster-robust variance estimators for fixed effects models estimated by weighted (or unweighted) least squares. Theorem 2 of the paper concerns a computational short cut for a certain cluster-robust variance estimator in models with cluster-specific fixed effects. It claimed that this short cut works for models estimated by generalized least squares, as long as the weights are taken to be inverse of the working model. However, the theorem is incorrect. In this corrigendum, we review the CR2 variance estimator, describe the assertion of the theorem as originally stated, and demonstrate the error with a counter-example. We then provide a revised version of the theorem, which holds for the more limited set of models estimated by ordinary least squares.