Cumulative Prospect Theory (CPT) can explain the variance premium puzzle. We solve a simple equilibrium model with CPT investors and find that probability weighting plays a key role in generating a substantial variance premium, while loss aversion captures the equity premium. Using GMM on a sample of U.S. equity and index-option returns between 1996 and 2010, our estimate of the probability distortion parameter implies that real-world investors in option markets distort probabilities significantly, but less so than subjects in lab experiments. We also show that the CPT model prices the cross-section of out-of-the-money index options well. In a dynamic setting, probability weighting and time-varying equity return volatility combine to match the observed time-series pattern of the variance premium.