Asset Allocation Modeling

A majority of asset allocation models and services are strictly backward looking and do not take into consideration the current market environment or beliefs on future market conditions. We are active in adjusting our assumptions based on the relative risks of asset classes, particularly their Sharpe ratios. Our adjusted assumptions are based on the rational and observable trend in increasingly efficient capital markets that volatility and return are linked.

We have all the latest tools for optimization simulations, but we are not big proponents of such exercises. First, all asset allocation models require assumptions about returns, volatility, and correlations of all relevant asset classes. No one we know has any clue about what future capital markets behavior will be to the level of specificity the models require. Second, a number of asset classes (e.g. hedge funds, private equity, and real estate) are difficult to model due to their skewed return distributions, illiquidity, and relative lack of historical data.

Third, asset allocation optimization models have the annoying tendency to maximize the effects of erroneous asset class assumptions. Any committee’s odds of being undone by input errors are, to us, unacceptably high. Fourth, committees generally pay little attention to the modeling results. For example, unconstrained models rarely give more than a token allocation to U.S. large capitalization stocks, yet committees routinely ignore the model and allocate the largest piece of the equity portion of the structure to U.S. large capitalization stocks.

We feel that an asset allocation study – rather than providing The Answer – will provide an opportunity to discuss the range of possible asset classes, the levels of risk, and the general expectations an organization might set. These studies help committees understand their plans’ sensitivities to changes in exogenous variables. Here, the absolute numbers are far less important, and the focus can be on relative movements. Thus, committees do not have to make asset allocation policy decisions in a vacuum, provided members remain conscious of the inherent weaknesses of a quantitative modeling approach, and recognize that model output is not a prescription.