Optimal demand for long-term bonds when returns are predictable

This paper further explores the horizon effect in the optimal static and dynamic demand for risky assets under return predictability as documented by Barberis (2000). Contrary to the case of stocks, the optimal demand for long-term Government bonds of a buy-and-hold investor is not necessarily increasing in the investment horizon, and may in fact be decreasing for some initial levels of the predicting variable. The paper provides an analytical explanation based on the dependence of the mean variance ratio on the investor's time horizon. Under stationarity of the predicting variable, unusually high or unusually low levels of the predictor tend to dissapear over time inducing the mean of cumulative returns to grow less or more than linearly as the investment horizon increases. If this effect dominates that on the variance, optimal demands can either be increasing or decresing in the investment horizon. On the other hand, the solution to the investor's dynamic allocation problem in the presence of bonds indicates that long-term Government bonds do not provide a good hedge for adverse changes in the investor's opportunity set: optimal dynamic demands for bonds do not differ from static portfolio choices at any horizon.