Improved testing for the efficiency of asset pricing theories in linear factor models

This paper suggests a refinement of the standard T2 test statistic used in testing asset pricing theories in linear factor models. The test is designed to have improved power characteristics and to deal with the empirically important case where there are many more assets than time periods. This is necessary because the case of too few time periods invalidates the conventional T2. Furthermore, the test is shown to have reasonable power in cases where common factors are present in the residual covariance matrix