| Many investment strategies are myopic in that they ignore everything that happens after the end of the one-period horizon. An efficient investment strategy, besides diversified across assets, should also be diversified through time. In this paper, we study fixed-income strategies in a multiperiod context using dynamic programming (DP) and Dybvig's payoff distribution pricing model (PDPM). We use DP to obtain the optimal dynamic allocation between the short bond and the long bond, and we employ the PDPM to investigate the robustness of our DP implementation and to measure the inefficiency of some fixed-income strategies.;We use the long rate and the spread (long rate - short rate) as the state variables for our DP implementation. For both the parametric and nonparametric cases, we find that the optimal proportion invested in the short bond increases when an investor becomes more risk averse or when his horizon becomes shorter, or both. Our DP results agree with those of Brennan, Schwartz, and Lagnado (1997) in that, for investors with long horizon, the short bond is not riskless because future re-investment rate is uncertain and the long bond can be used to hedge against adverse changes in future investment opportunity set.;The PDPM measures the inefficiency of any strategy whose terminal payoffs are NOT in reverse order of its terminal state price densities. Whereas Dybvig applied the PDPM to equity portfolio strategies, we extend it to fixed-income strategies. Our PDPM results confirm that the DP strategies are basically efficient. Furthermore, we find that the inefficiency amounts of some well-known fixed-income strategies (like rolling over at the short rate and rolling over short bonds) are quite substantial. In contrast, riding the yield curve and contingent immunization are rather efficient---provided the liquidity premium theory for the term structure of interest rates holds for the former. |
