Title: A Hausman test for cross-section dependence in linear panel regression models
Authors: Simon Reese - Lund University (Sweden) [presenting]
Abstract: Testing for the presence of cross-section dependence in panel data is becoming an ever more relevant model specification issue given the increasing use of interactive fixed effects models. We contribute to this strand of the literature by proposing a Hausman test for cross-section dependence. By focusing on differences in the parameter vectors, our focus is thus different from that of dominating existent tests which are based on correlations in the model residuals. The distribution of the interactive fixed effects estimator under the null hypothesis is derived using a normalisation techniques for the factor estimates, allowing to circumvent the problem of asymptotic singularity. Simulations reveal good performance results relative to other tests.