(the number of cross-sections) and T is fixed (the number of time periods). Three theoretical results emerge: (i) the standard F-test procedure will still deliver asymptotically valid inferences; (ii) under (pure) local random effects, the F-test and random effects test procedures have identical...

This paper provides evidence on problems associated with using standard tests for predictive failure when the errors of a linear regression model are not normally distributed. The ability of a simple bootstrap procedure to give a useful degree of control over the significance levels is examined.