MARKOWITZ VERSUS REGIME SWITCHING: AN EMPIRICAL APPROACH

Abstract. This article discusses an adjusted regime switching model in the context of port-folio optimization and compares the attained portfolio weights and the performance to a classical mean-variance set-up as introduced by Markowitz (1952). The model postulates different asset price dynamics under different regimes, and jumps between regimes are driven by a Markov process. For examples, ’bear ’ and ’bull ’ markets could be such regimes. Given a particular regime, portfolio weights are set based on the conditional means and variance-covariance structure of the asset dynamics. The model is evaluated in an out-of-sample period of the last three years with a moving window and a forecast of only one period. It is found that with the adjusted regime switching portfolio selection algorithm as applied here, the performance of the optimal portfolio is highly improved even where portfolio weights are constrained to realistic values. 1