In this paper we develop a portfolio selection theory considering discrete regime shifts in the investment opportunity and conduct an empirical analysis using Japanese sector indices to verify its effectiveness. Specifically, we model the regime shifts using a first-order Markov switching model and consider a dynamic portfolio selection problem using log mean-variance criteria. The estimation result implies that the model allows us to extract stable regimes when the number of regimes is appropriately chosen. Taking advantage of these regimes, we can improve the performance of the portfolio with rebalancing frequencies kept low.
Regime switching model Dynamic portfolio selection Discrete-time Log mean-variance criteria Quadratic programming EM algorithm