Application of the Multivariate Average F-Test to Examine Relative Performance of Asset Pricing Models with Individual Security Returns

The standard multivariate test of Gibbons et al. (1989) used in studies examining relative performance of alternative asset pricing models requires the number of stocks to be less than the number of time-series observations, which requires stocks to be grouped into portfolios. This results in a loss of disaggregate stock information. We apply a new statistical test to get around this problem. We find that the multivariate average F-test developed by Hwang and Satchell (2014) has superior power to discriminate among competing models and does not reject tested models altogether, unlike the standard multivariate test. Application of the multivariate average F-test for examination of relative performance of asset pricing models demonstrate that a parsimonious 6-factor model with the market, size, orthogonal value, profitability, investment, and momentum factors outperforms all other models.