Abstract
In this paper, we examine the small sample properties of the two-stage test which consists of a pre-test for homoscedasticity followed by either an F test (i.e., the socalled Chow test) if homoscedasticity is indicated or the Wald test if homoscedasticity is not indicated for equality between individual coefficients in two linear regressions. It is shown that the exact distribution of the two-stage test can be evaluated by the formula derived by Lauer and Han [6] and the two-stage test has a better sampling performance than the Wald test which incorporates no pre-test. Further, the sampling performances of the two-stage test and the size-corrected Wald test proposed by Rothenberg [13] are examined by Monte Carlo experiments.
