Stock Return Predictability in a Monetary Economy

In an economy where agents hold money, the short interest rate determines the trade-off between money holdings and consumption. Building on this idea, we develop a theoretical model that shows the transmission mechanism through which the short rate finds its way to stock-return predictability regressions. We construct a cointegration relation that links share prices and dividends to the short interest rate when we investigate empirically the main implications of our model. This relationship, that we denote the d pdR-ratio, strongly predict stock returns and excess returns, even when the statistical significance of long-horizon return predictability is evaluated using Hodrick (1992) t-statistics. The result that stock returns and excess returns are predictable at long horizons is different from recent findings reported by Ang & Bekaert (2006). We differ from Ang & Bekaert (2006) by explicitly accounting for the non-stationarity of the predictors in a cointegration framework. Our theoretical model also suggests that aggregate output might be better suited as a scaling variable than dividends. As a consequence, we also build a cointegration variable that relates share prices, output, and the interest rate in a cointegration framework. We denote this variable the d pyR-ratio. We show that the d pyR-ratio strongly predicts stock returns and excess returns at long horizons. The d pyR-ratio is also found to have some predictive power for dividend growth which the d pdR-ratio does not. Neither the d pyRratio nor the d pdR-ratio have predictive power for output growth and the interest rate.