This paper constructs tests for homoskedasticity in one-way panel data error components models in line with Baltagi, Bresson and Pirotte (Journal of Econometrics 134, 2006). Our proposed tests have two robustness properties. First, we present evidence showing that the Gaussian based statistics of Baltagi et al. reject too often in the presence of asymmetric (e.g. log-normal) and heavy-tailed (e.g. t-Student) distributions. By using simple moment conditions, we derive distribution free tests statistics that are robust to these non-normalities. Second, a small sample correction makes our marginal tests insensitive to heteroskedasticity in the component not being checked, and hence help identify the source of heteroskedasticity. Additionally, they are computationally convenient since they are based on simple artificial regressions using pooled OLS residuals
