Asset Allocation and Long-Term Returns: An Empirical Approach

We present a discussion of the standard approaches to asset allocation, generally falling into two camps: Mean Variance Optimization and the maximization of the final value of a wealth utility function. After describing shortcomings in both of these standard approaches, we describe a heuristic, empirical approach that uses concepts of Shortfall Risk as an objective and actual data as a direct model of the stochastic market evolution. This empirical approach leads to fundamentally different conclusions than the standard approaches, and by definition fits the past data much better. Among other observations, we find that (1) for holding periods of about 15 years and greater, bonds are riskier than stocks in the practical sense that bonds are almost always expected to under perform stocks over 15-year or higher time horizons, (2) while short-term stock returns show fat tails, long-term stock returns show skinny tails, and (3) while stocks show mean reversion, bonds exhibit mean aversion.