Tail risk exposure and the cross section of expected stock returns

This paper investigates whether stocks earn a premium for their sensitivity to market tail events, referred to as tail risk exposure (TRE). We show that commonly used estimators of TRE, typically based on tail dependence between asset and market returns exhibit significant statistical biases, particularly in the presence of general market dependence. Empirically, we find that tail risk is priced only in low-correlation stocks, where average market comovement is weak. This suggests that investors underestimate TRE in low-correlation stocks and overestimate it in high-correlation stocks. To address this, we propose a novel double-sort portfolio strategy that accounts for both TRE and correlation, allowing us to isolate and accurately price TRE. This strategy consistently outperforms traditional single-sort methods in terms of predictive accuracy and risk-adjusted returns.