Multi-period Dynamic Bond Portfolio Optimization Utilizing a Stochastic Interest Rate Model

Regardless of the asset class, applying multi-period dynamic portfolio optimization to real investment activity is challenging due to theoretical and structural complexities. In terms of a bond portfolio based on a stochastic interest rate model, some literature exists that focuses on theoretical aspects of multi-period dynamic bond portfolio optimization, such as deriving analytical solutions for optimal portfolios, to be sure, but no empirical studies analyzed the actual bond market. Additionally, a methodology that considers realistic investment constraints has not been developed thus far. In this paper, we propose a new framework for multi-period dynamic bond portfolio optimization. As bond return can be approximated by a linear combination of factors that constitute a stochastic interest rate model, we apply linear rebalancing rules that consider transaction costs, in addition to self-financing and short sales constraints. Then, as an empirical analysis, we conduct an investment backtest by analyzing discount bonds estimated from Japanese interest-bearing government bonds. The results indicate that multi-period optimization represents a relatively high performance compared to single-period optimization. Further, the performance improves as the investment horizon and investment utilization period are extended up to a certain point.