The Extreme Bounds of the Cross-Section of Expected Stock Returns

Abstract

Several empirical studies report violations of the asset-pricing model of Sharpe (1964), Lintner (1965), and Black (1972). But, there is no consensus on specification in this literature, as such studies typically consider only a limited number of explanatory variables and do not satisfactorily control for previous findings. Extreme bound analysis (EBA), an imperfect but useful remedy for model uncertainty, suggests that comparatively few factors are robust. Given the cross-section of expected stock returns from July 1963 through December 2000, three of 23 variables - market size as well as short and medium run lagged return - pass the traditional EBA decision rule given all possible 3-, 4-, and 5-factor models of monthly stock returns. This paper also explores several potential improvements to EBA, including explicit consideration of possible multicollinearity, which largely does not affect the results, as well as sample divisions, which suggest that fewer variables are sturdy correlates of returns.