An Investigation of the Impact of Industry Factors in Asset-Pricing Tests

Abstract

This paper seeks to better describe the characteristics that explain the cross section of average stock returns. Fama and French (FF) (1993) argue that book-to-market ratio (BE/ME) proxies for risk which investors require a premium for assuming. An alternative hypothesis proposed by Daniel and Titman (1996) (DT) suggests that BE/ME may price stocks because investors irrationally focus on firm characteristics rather than on risk factor loadings. A firm's BE/ME depends on accounting standards and other elements which vary across industries. Therefore BE/ME may be a noisy proxy either for risk factor sensitivity or for sources of potential investor irrationality. We attempt to reduce this noise by creating industry-relative factor- mimicking portfolios using both inter- and intra-industry BE/ME information. We first document that our portfolios generate significant pricing errors in FF three-factor regressions and may price the FF test assets better than FF's (1993) HML factor. We then use these new factor-mimicking portfolios to explore a hypothesis of Kenneth French, which attempts to reconcile the results of DT with rational asset-pricing. Using our intra-industry factor, we perform asset-pricing tests which may distinguish between the French and DT stories. These tests find no support for the French hypothesis.