According to the present value model, the long-run expected return on stocks, stock yield, is an affine function of the dividend yield on stocks and a weighted average of expected future growth rates in dividends. We construct a measure of the stock yield using only the realized dividend yield and a growth rate proxy from sell-side analysts’ near-term earnings forecasts. Its out-of-sample R-square is consistently above 2% at monthly level in our sample period, better than two of the best performing models: models based on the implied cost of capital and partial least squares for forecasting stock market returns. In sample, stock yield predicts future stock market return with an adjusted R-square of 13% at one year, up to 54% at four years in US. It also performs well for predicting the return on stock markets in other G7 countries and US portfolios formed by sorting stocks based on firm characteristics such as book-to-market ratio, size, and industry. Interestingly, the forecasting ability of the stock yield is concentrated in bad times when investors’ fears are high. The out-of-sample performance of the present-value model of Binsbergen and Koijen (2010) is significantly improved when stock yield is used as an additional noisy observation of the unobserved expected return.