Do investors gain from forecasting the asymmetric return co‐movements of financial and real assets?

Recent research on asset allocation emphasizes the importance of considering non‐traditional asset classes such as commodities and real estate—the former for their diversification properties, and the latter due to its importance in the average investor's portfolio. However, modelling and forecasting asset return co‐movements is challenging because the dependence structure is dynamic, regime‐specific, and non‐elliptical. Moreover, little is known about the economic source of this time‐varying dependence or how to use this information to improve investor portfolios. We use a flexible framework to assess the economic value to investors of incorporating better forecasting information about return co‐movements between equities, bonds, commodities, and real estate. The dependence structure is allowed to be dynamic and non‐elliptical, while the state variables follow Markov‐switching stochastic volatility processes. We find that the predictability of return co‐movements is significantly improved by incorporating macro and non‐macroeconomic variables, in particular inflation uncertainty and bond illiquidity. The economic value added to investors is significant across levels of risk aversion, and the model outperforms traditional multivariate GARCH frameworks.