Abstract:

The dividend-price ratio predicts aggregate stock market returns with higher precision during recessions than during expansions. Moreover, the difference in predictive accuracy across states is both statistically and economically significant. I show that this empirical evidence cannot be reconciled with three popular asset pricing approaches: habit-persistence (Campbell and Cochrane (1999)), long-run risk (Bansal and Yaron (2004)) and rare disasters (Gourio (2012)). To deal with this, I propose a long-run risk model which explicitly links volatility and left tail events in consumption and dividend growth, for which I find strong support in the data. In my model positive shocks to volatility lead to a higher probability of observing negative growth rates. This feature increases the variance of the dividend-price ratio during recessions (i.e. when volatility is high and growth rates are low) and helps the model to match the return predictability patterns observed in the data.