Finite sample performance of specification tests for correlated random effects quantile panel regressions

We investigate the finite sample performance of the Abrevaya and Dahl (2008) test for coefficient heterogeneity for a correlated random effects ‘mean’ (CREM) panel quantile regression estimator. We assess size and power of the test over a range of sample sizes and panel dimensions. The test is undersized for small-to-moderate sample sizes and displays low power even with a high degree of heteroscedasticity. Size and power improve substantially in larger samples. Our results provide insight for applied researchers.