A distribution-free joint test for random individual effects and heteroscedasticity allowing for incomplete panels

This paper proposes a convenient testing procedure designed for detecting, from preliminary (pooled) OLS estimation of the model, the possible simultaneous presence of individual effects and heteroscedasticity in the error of a linear panel data regression model. The proposed testing procedure consists in (1) a distribution-free and allowing for unbalanced panel joint pseudo-LM test of the null of no individual effect and no heteroscedasticity and (2) a BMCP based on distribution-free and again allowing for unbalanced panel robust one-directional pseudo-LM tests for identifying the source(s) of departure from the joint null when it is rejected. An empirical example illustrates the usefulness of the testing procedure.