Strategic Asset Allocation for Long-Term Investors: Parameter Uncertainty and Prior Information

Abstract:     
This paper considers the strategic asset allocation of long-term investors who account for prior information about expected returns. We develop a vector autoregressive model where different investors have conflicting prior views on long-run expected returns. We distinguish two types of prior information: (i) direct views on the long-term mean of the equity and bond premium, and (ii) prior views on the long-run mean of predictor variables like the dividend yield and the nominal interest rate. Both priors have a pronounced effect on optimal portfolios. Even weak prior information on the unconditional mean of highly persistent time series like dividend yield and the nominal interest rate changes the estimated persistence of shocks and the predictability of excess returns.

For long-term investors we find that a portfolio that is optimal given one prior, often entails large utility costs when evaluated under an alternative prior distribution. The optimal portfolio for an optimistic investor is very costly (sub-optimal) in eyes of an investor with a more negative prior view. We define a robust portfolio as the portfolio of an investor with a prior that has minimal costs among all priors that we consider. Such a robust portfolio coincides with the optimal portfolio of a moderately optimistic investor. It contains a large proportion of equity, but far less than would be implied under both more optimistic as well as very diffuse priors.

Keywords: strategic asset allocation, bayesian vector autoregression, parameter uncertainty

JEL Classifications: C32, G11, C11

Working Paper Series