Using several multifactor models, I find strong 'betting against beta' effects - flat relations between betas and expected returns - for most non-market factors in US and international stock markets. 'Arbitrage portfolios' designed to profit from these effects earn average returns similar to those of the factors, with substantially reduced risk. Betas are persistent, indicating that the factor models successfully capture important dimensions of covariation in returns. Previously-proposed explanations for the betting against market beta effect do not explain these results. These findings raise questions about the economic content of existing empirical factor models and the covariance-based expected return paradigm.
Keywords: Betting against beta, security market line, security market plane, arbitrage portfolios, factor models, arbitrage pricing theory
JEL Classi cations: G11, G12, G15
