Using portfolios that are formed by directly sorting stocks based on their exposure to characteristics-based factors, earlier studies find that these beta-sorted portfolios have very large ex post factor beta spreads. However, the return spreads between high- and low-beta firms are typically tiny and insignificant. This study examines the time variation in the pricing of a large set of characteristics-based factors. Our evidence shows a striking two-regime pattern for most of the factor-beta-sorted portfolios: high- beta portfolios earn significantly higher returns than low-beta portfolios following high-sentiment periods, whereas the exact opposite occurs following low-sentiment periods. Remarkably, this two-regime pattern is completely reversed when macro-related factors, such as consumption growth and TFP growth, are used. The evidence based on mutual fund and hedge fund returns also confirms this two-regime pattern. Our findings suggest that the exposure to most of these characteristics-based factors is likely to be a proxy for the level of mispricing, rather than risk, especially during high-sentiment periods.

JEL Classification: G12

Keywords: Factor beta, Investor sentiment, Mispricing, Risk​​