Portfolio models with return forecasting and transaction costs

Abstract

In this paper, we advance portfolio models by incorporating return projection and further analyze their realized performance. To ensure practicality, the transaction costs and the optimization of short-selling weights are taken into account in portfolio rebalancing. Using the daily returns of international ETFs over a period of 14 years, the empirical results show that including return forecasting improves the realized performance due to more efficient asset allocation but not a reduction in trading costs. The models that are based on trade-off between return and volatility, such as the mean-variance and Omega models, show higher increases in performance than those mainly focus on controlling loss, such as the linearized value-at-risk, the conditional value-at-risk, and the downside risk. The superiority of forecasting risky portfolios over the equally-weighted diversification varies intertemporarily across various portfolio models. The benefit of inclusion of prediction is larger when the market is less volatile.