This paper examines the predictability of stock market returns using the variance risk premium, defined as the difference between implied and realized variances. The authors find that the variance risk premium explains over 15% of the ex-post variation in quarterly excess returns on the market portfolio from 1990 to 2005, with high (low) premia predicting high (low) future returns. The predictive power of the variance risk premium surpasses that of traditional predictors such as the P/E ratio, dividend yield, default spread, and consumption-wealth ratio (CAY). Combining the variance risk premium with the P/E ratio results in an R² of over 25% for quarterly returns. The results depend on the use of "model-free" implied and realized variances, which are calculated from high-frequency intraday data rather than daily data. The findings suggest that temporal variation in risk and risk aversion plays a significant role in determining stock market returns. The variance risk premium is shown to be a strong predictor of future returns, outperforming traditional variables in both accuracy and explanatory power. The study also includes robustness checks, demonstrating that the results hold even when using alternative variance measures and predictor variables. The variance risk premium is interpreted as a measure of market risk aversion, and its predictive power is consistent with the idea that changes in risk aversion influence stock market returns. The study concludes that the variance risk premium is a significant factor in explaining stock market returns, with its predictive power being largely due to time-varying risk aversion rather than systematic risk.This paper examines the predictability of stock market returns using the variance risk premium, defined as the difference between implied and realized variances. The authors find that the variance risk premium explains over 15% of the ex-post variation in quarterly excess returns on the market portfolio from 1990 to 2005, with high (low) premia predicting high (low) future returns. The predictive power of the variance risk premium surpasses that of traditional predictors such as the P/E ratio, dividend yield, default spread, and consumption-wealth ratio (CAY). Combining the variance risk premium with the P/E ratio results in an R² of over 25% for quarterly returns. The results depend on the use of "model-free" implied and realized variances, which are calculated from high-frequency intraday data rather than daily data. The findings suggest that temporal variation in risk and risk aversion plays a significant role in determining stock market returns. The variance risk premium is shown to be a strong predictor of future returns, outperforming traditional variables in both accuracy and explanatory power. The study also includes robustness checks, demonstrating that the results hold even when using alternative variance measures and predictor variables. The variance risk premium is interpreted as a measure of market risk aversion, and its predictive power is consistent with the idea that changes in risk aversion influence stock market returns. The study concludes that the variance risk premium is a significant factor in explaining stock market returns, with its predictive power being largely due to time-varying risk aversion rather than systematic risk.