We study whether exposure to marketwide correlation shocks affects expected option returns, using data on S&P100 index options, options on all components, and stock returns. We find evidence of priced correlation risk based on prices of index and individual variance risk. A trading strategy exploiting priced correlation risk generates a high alpha and is attractive for CRRA investors without frictions. Correlation risk exposure explains the cross-section of index and individual option returns well. The correlation risk premium cannot be exploited with realistic trading frictions, providing a limits-to-arbitrage interpretation of our finding of a high price of correlation risk.

ABSTRACT We study whether exposure to marketwide correlation shocks affects expected option returns, using data on S&P100 index options, options on all components, and stock returns. We find evidence of priced correlation risk based on prices of index and individual variance risk. A trading strategy exploiting priced correlation risk generates a high alpha and is attractive for CRRA investors without frictions. Correlation risk exposure explains the cross‐section of index and individual option returns well. The correlation risk premium cannot be exploited with realistic trading frictions, providing a limits‐to‐arbitrage interpretation of our finding of a high price of correlation risk.

ABSTRACT We study aversion to model ambiguity and misspecification in dynamic portfolio choice. Risk‐averse investors (relative risk aversion ) fear return persistence, while risk‐tolerant investors () fear mean reversion, when confronting model misspecification concerns of identically and independently distributed (IID) returns. The intuition is that risk‐averse investors, who want to hedge intertemporally, endogenously fear return persistence, which precludes hedging. A log investor is myopic and unaffected by model misspecification, therefore only worrying about model ambiguity. Our model can generate belief scarring, nonparticipation in equity markets, and extrapolative return expectations. Extending beyond IID returns, we study model misspecification for a mean‐reverting Sharpe ratio.