In this paper, I show that the variance of Fama-French factors, the variance of the momentum factor, as well as the correlation between these factors, predict an important fraction of the time-series variation in post-1990 aggregate stock market returns. This predictability is particularly strong from one month to one year, and it dominates that afforded by the variance risk premium and other popular predictor variables such as P/D ratio, the P/E ratio, the default spread, and the consumption-wealth ratio. In a simple representative agent economy with recursive preferences, I model the portfolio weight in each asset as a function of a stock's characteristics and show that the market return can be predicted by these variances.