Spurious regressions in small samples

Abstract

This paper shows another problem with simply relying on classical hypothesis testing. The size of economic tests is weak. Specially in time series analysis where the number of observations is relatively small and series are very persistent, we need to be cautious in interpreting standard t and χ^2 tests since they tend to reject the null hypothesis too often. We should examine which models are good approximations instead of simply rejecting or failing to reject the null hypothesis.