Anessa Custovic

Job market paper

A Mixed Frequency Approach to Return and Dividend Growth Predictability

This paper evaluates the predictive ability of a high frequency dividend price ratio. I show that we can uncover both dividend growth and return predictability across equity markets by weighting the dividends in each month differently through the application of MIDAS regressions. For US equity markets, I find strong predictability of long horizon real dividend growth with the theoretically correct sign and show that there is some predictability with nominal dividend growth as well. Out-of-sample I find that the high frequency constructed dividend price ratio can result in superior predictability for returns over the annual single frequency dividend price ratio. When combined with other predictors, the high frequency constructed dividend price ratio results in even better out-of-sample predictability for both returns and the equity premium. The out-of-sample performance is robust to horizons as well.