I show that the higher the share of institutional ownership in a stock, the more its price-dividend ratio is driven by discount rate variation rather than by changes in dividend growth expectations. Hence, the dividend yield of high institutional ownership stocks predicts returns. Conversely, for stocks held mostly by individual investors, returns are not predictable.
As a general equilibrium outcome, return predictability crucially depends on the properties of the marginal investor.
More strongly time-varying marginal utility of institutions acting as marginal investors in institutionally held stocks provides a natural explanation for the observed pattern.
In my equilibrium model, time-varying redemption risks generate the observed differences in predictability among a priori identical stocks.
My finding helps explain the weak return predictability of small and value stocks, the postwar predictability reversal, and the fact that dividend smoothing cannot explain that reversal.

We introduce Implied Volatility Duration (IVD) as a new measure for the timing of uncertainty resolution, with a high IVD corresponding to late resolution. Portfolio sorts on a large cross-section of stocks indicate that investors demand on average about seven percent return per year as a compensation for a late resolution of uncertainty. In a general equilibrium model, we show that `late' stocks can only have higher expected returns than `early' stocks if the investor exhibits a preference for early resolution of uncertainty. Our empirical analysis thus provides a purely market-based assessment of the timing preferences of the marginal investor.


I test the hypothesis that the differences in return predictability between the value and growth portfolios are indeed due to value stocks having shorter cash flow duration. 
Overall, duration acts as amplification for the change in dividend yields that is caused by discount rate variation but differences in return predictability across book-to-market sorted portfolios go beyond the effect of duration.  For comparable cash flow duration, discount rate variation explains about 40\% more of the dividend yield variation for growth stocks as opposed to value stocks. Predictive regressions across portfolios suggest commonality in discount rates and cash flows of value (growth) stocks beyond the effect of duration. This is consistent with recent evidence suggesting that the exposure to value-specific risks is not driven by duration.