Estimation errors in the inputs are the main problem when applying portfolio analysis. Markov regime switching models are used to reduce these errors, but they do not always improve out-of-sample portfolio performance. We investigate the levels of transaction costs and risk aversion below which the use of two regimes is superior to one regime for an investor with a CRRA utility function, allowing for skewed and kurtic returns. Our results suggest that, due to differences in risk and transactions costs, most retail investors should use one regime models, while investment banks should use two regime models.
finance, portfolio theory, regime shifting, transaction costs, risk aversion, constant relative risk aversion