📝 SummarXiv

Abstract

We propose a Hausman test for the correct specification of unobserved heterogeneity in both linear and nonlinear fixed-effects panel data models. The null hypothesis is that heterogeneity is either time-invariant or, symmetrically, described by homogeneous time effects. We contrast the standard one-way fixed-effects estimator with the recently developed two-way grouped fixed-effects estimator, that is consistent in the presence of time-varying heterogeneity (or heterogeneous time effects) under minimal specification and distributional assumptions for the unobserved effects. The Hausman test compares jackknife corrected estimators, removing the leading term of the incidental parameters and approximation biases, and exploits bootstrap to obtain the variance of the vector of contrasts. We provide Monte Carlo evidence on the size and power properties of the test and illustrate its application in two empirical settings.