Beating the Market Part 2

In Part 1, I discussed some of the structural reasons why it is easy to beat the market that stem from the incompleteness of markets. The investment policy statement limits most market participants to specific assets and factors such as liquidity, bid-ask spreads, and risk management limit the remaining market participants from maintaining constant market efficiency. If we look at the largest market participants, they are typically long only institutional investors with relatively low portfolio churn rates. A good representative portfolio consists of the following five asset classes: U.S. Equities, International Equities, Fixed Income, REITs, and Alternative Investments.

Using monthly data from January 1990 through December 2009, average correlations among asset classes range from .0942 to .7708 with the highest correlation between U.S. equities and alternatives and the lowest correlation between bonds and alternatives. A moving block bootstrap of this return series shows a optimal tactical band size of around 2% assuming a constant band size across all asset classes and maximizing the information ratio.

From January 2007 through December 2009 correlations ranged from .1304 to .9209 with the highest correlation between international and U.S. equities and the lowest correlation again between bonds and alternatives. Contrast that with the period from January 2001 through December 2003 when correlations ranged from -.3830 to .8978 and the return from asset allocation was far greater. No doubt decreasing the time interval would reveal much larger swings in correlations but few institutions have substantial daily liquidity needs relative to the size of the portfolio so large tactical allocation decisions are not needed.

If you made it this far, you are no doubt wondering why you should care about this. The biggest concern is setting tactical asset allocations for policy statements or how much to deviate from your target allocation. The full 20 year period suggests that periods of higher correlations result in an optimal tactical band size of around 2% to maximize the portfolio information ratio while shorter periods of lower correlations could be used to justify taking more risk. This should be considered when setting the investment policy statement and compensation packages for investment managers if the information ratio is a meaningful measure for you. If total return is the most important metric, larger band size always increases the potential return but has a diminishing effect as correlations increase. This means that the biggest risk takers are not going to receive the benefits of asset diversification when they need it most. In order to produce sufficient returns to justify active management expense, asset selection becomes more important as the economies of scale decrease. In other words, the smaller the portfolio, better asset selection and fewer holdings are needed to justify the expense of active management. Small portfolios cannot rely on index funds and good rebalancing discipline so hiring a portfolio manager that does not engage in asset selection is a waste of money.

Disclaimer: There are clearly many problems with this analysis such as the arbitrary time periods, disclosure of initial allocations, and the simplicity of the rebalancing rule. Unfortunately, I cannot write a complete post at this time without straining any working relationships despite the fact that these results are easily replicated, however; since I am not being compensated for my work, I consider it to be my own.