I construct an equilibrium model that captures salient properties of index option prices, equity returns, variance, and the risk-free rate. A representative investor makes consumption and portfolio choice decisions that are robust to his uncertainty about the true economic model. He pays a large premium for index options because they hedge important model misspecification concerns, particularly concerning jump shocks to cash flow growth and volatility. A calibration shows that empirically consistent fundamentals and reasonable model uncertainty explain option prices and the variance premium. Time variation in uncertainty generates variance premium fluctuations, helping explain their power to predict stock returns. KNIGHTIAN UNCERTAINTY-AMBIGUITY ABOUT the true model governing fundamentals-can be an important factor influencing investors' consumption and portfolio choice decisions. Incorporating it into asset pricing models can therefore shed light on sources of asset return premiums and time variation in prices. In this paper, I construct an equilibrium model with a representative investor who has model uncertainty that varies in intensity over time. 1 A calibration of this model is shown to capture a broad set of features of equity and equity index option prices. In particular, the model generates two prominent features of index options that represent a challenge to equilibrium models: (i) the large premium embedded in their prices, called the variance premium, and (ii) the shape of their implied volatility surface, especially the "volatility skew." This is the name given to the pattern in the implied volatilities of shortmaturity puts, which are high for out-of-the-money (otm) puts and decrease sharply as put moneyness increases. At the same time, the model matches the salient properties of equity returns and the risk-free rate, the dynamics of * Itamar Drechsler is at Stern School of Business, New York University. I my grateful to my committee, Amir Yaron (Chair), Rob Stambaugh, and Stavros Panageas. I also thankThe Journal of Finance R generates the aforementioned features of index option prices, which are not captured by leading equilibrium asset pricing models. 5 The calibrated model generates a variance premium that is large and positive, as in the data, and has predictive power for one-and three-month excess stock returns that is empirically consistent. Furthermore, the calibration matches the implied volatility surface, with an implied volatility skew that is very steep for one-month options and decays steadily as the horizon increases. Finally, the model also matches the dynamic properties of conditional equity variance, such as its persistence and volatility, and the higher moments of equity returns.The calibration achieves its fit to the data using a level of model uncertainty under which the worst-case model is difficult to reject empirically; roughly speaking, it implies that one cannot reject at the 10% level the null hypothesis that the worst-case model generates the data. It is therefore reasonab...