Are Vector Autoregressions an Accurate Model for Dynamic Asset Allocation?

Much of the growing literature on tactical and strategic asset allocation uses vector autoregressive models (VAR) for returns and predictors. Since the portfolio advice they generate may be misleading if those models are not an accurate description of reality, we evaluate the implied joint density forecasts of US monthly excess returns on stocks and bonds. From the point of view of an investor who rebalances monthly, a VAR offers a reasonable description of the data, which is not improved upon by richer models. We also study the relevance of considering time-varying risk premia and parameter uncertainty in density forecasts.