On the Joint Pricing of Stocks and Bonds: Theory and Evidence

This paper examines the related questions, of the time-series behavior of expected returns and of return predictability, within the framework of the stock-bond pricing model proposed in Mamaysky (2002). The key advantage of the model-based approach adopted in this paper is that the quantities of interest (i.e. expected returns, prices of risk, and R2's of forecasting regressions of returns on their true conditional expectations) are directly observable (once the model has been fitted to the data). Furthermore, the fact that the present model accomodates jointly the pricing of both bonds and stocks allows us to derive estimates of prices of risk and of expected returns that incorporate, by construction, the relevant information from both bond and stock markets. Estimation of the model using U.S. data reveals a rich dynamic structure of prices of risk, some pro- and some countercyclical, and of expected returns. Also, the paper suggests that excess return predictability (as measured by the above mentioned R2's) for a broad market index is a hump-shaped function of the forecasting horizon, achieving a maximum value of roughly 13.5% at a time horizon of five years.