A New Look At Adaptive Asset Allocation

Adaptive Asset Allocation (AAA) was born as one of several sibling strategies for applying Modern Portfolio Theory (MPT), which was first proposed in 1967 as a way to optimize portfolio gains. Yet, many traders and financial strategists who truly believe in the math of MPT are disillusioned because the real-world results while using AAA haven’t met their calculated expectations for gains, and the volatility of such portfolios has been higher than expected.

[adrotate banner =”71″]

Recent studies of this topic have suggested that this mismatch between expectations and reality may be primarily due to the length of the time periods used for input averages and portfolio re-balancing: Apparently, when calculations are based on input data using averages obtained over much shorter periods of time, the portfolio returns are better than when those averages are calculated based on long-term numbers. And, when the portfolio re-balancing intervals are shorter, performance is better and volatility and risk are reduced.

To recap, MPT relies on 3 parameters to create ideal portfolios, typically involving a set of asset classes including stocks in the U.S., European, Japanese and emerging markets, plus U.S. and international REITs, U.S. long-term and intermediate Treasuries, as well as gold and other commodities. The parameters are:

  • Expected volatility
  • Expected returns
  • Expected correlation

It seems that using shorter-term averages for MPT scenarios leads to more accurate results. One shortcoming of the previous-generation allocation model, Strategic Asset Allocation (SAA), becomes apparent because that model applies MPT based on long-term averages regarding the above parameters. As detailed in the recent new work on this topic, using long-term averages leads to significant errors in calculated returns.

In practice, long-term averages over a 5-to-20-year time horizon are poor predictors of volatility, returns and correlation. The statistical gap between calculations using 20-year averages and those using 3-or-4-year averages with regard to stocks’ annualized returns is huge, ranging from negative returns to nearly 14%. Given the relatively short investment time horizons of most investors nowadays, it seems clear that using shorter-term parameters in the calculations will yield more realistic results.

To acknowledge reality without disavowing longer-term calculations entirely, some investors choose to tweak their calculations by applying a long-term value approach instead of a long-term average approach, which tends to weight portfolios in favor of equities when stock prices fall, and conversely to reduce weighting in equities as their prices become more expensive.

Yet, with advancing technology there are some new alternatives to using long-term valuation for “handicapping” the calculated returns. At the extreme end of the short-term horizon lie the high frequency traders, who take advantage of short-term trends, correlations and reversions-to-mean in order to generate more-realistic estimates of returns. There is currently much excitement in the trading community based on the success of traders who use HFT systems. Still, as more traders crowd into this niche, it’s possible that the spreads will thin or perhaps vanish altogether.

The predictive value of momentum

Momentum is an excellent way for investors to estimate performance over the short term. According to the old adage: The best predictor of short-term future price is the current price. And, as the investment horizon is extended from intraday or daily trading outward toward weekly periods, the effect of momentum becomes more noticeable. Perhaps due to larger, slower-moving investors, prices tend to keep moving in the same direction for several weeks. Given this probability, it’s logical to account for momentum when building a portfolio, regardless of the long-term averages already observed.


Volatility, too, has been misapplied with regard to MPT. For example, although average long-term annualized volatility is about 20% for stock prices and about 7% for 10-year Treasuries, actual volatility measured during the shorter time horizons of most investors fluctuates much more wildly, and is therefore much less accurate for projecting future conditions. So, actual volatility can have a far more adverse impact on a portfolio than the calculated volatility implies.

And, although many investors attempt to roughly balance the difference in volatility between stocks and bonds by weighting portfolios with 60% stocks and 40% bonds, still, the actual volatilities experienced can far override such a crude balancing method. Therefore, with regard to volatility assumptions it seems safest to rely on the adage mentioned above, that is, the least-biased guess of tomorrow’s price is based on today’s price. Likewise, the least-biased guess of tomorrow’s price range is the price range during the recent past, which of course represents the recent volatility.

Since recent volatility seems to offer the best guess about near-term future volatility, and most investors have a short-term horizon, it seems logical to use short-term volatility as the parameter for MPT instead of long-term volatility. As a takeaway regarding volatility, a savvy investor rebalancing a portfolio can calculate its volatility and, in order to maintain the volatility risk at a stable level over time, could reduce exposure by partly moving into cash when volatility exceeds the targeted level.

Correlation & returns

Even though long-term correlations between the prices of asset classes such as stocks and Treasuries, or stocks and gold, are low or negative, over shorter time periods the actual correlations vary greatly. So, for example, the volatility of a 50-50 stock-and-bond portfolio may decrease by 50% as the correlation decreases.

Similarly, although many traders intuitively understand that a portfolio’s risk is reduced by apportioning the volatility of its components, a less-intuitive observation from the recent studies has been that returns from risk-managed portfolios were also improved by as much as 25%. Finally, since the human nature of investors makes it difficult to focus on returns alone while disregarding risks, especially over a longer term when draw-downs may accrue, it’s also prudent to consider maximum draw-down along with volatility when seeking maximum returns.


If MPT scenarios based on near-term average values give more accurate estimates than those based on long-term values, then it seems best for HFT traders and other short-horizon investors to use current observed values for portfolio optimization. In the recent studies cited herein, the authors have advocated the monthly rebalancing of portfolios by using a true Adaptive Asset Allocation based on returns in the near term in view of their momentum, along with the appropriate short-term volatility and correlation averages.

One algorithmic approach might be to create fresh portfolios at the time of monthly rebalancing based on the top few assets according to six-month or even shorter momentum, and to allocate assets according to an algorithm specifying minimal variance in volatility, instead of apportioning each asset according to its individual volatility. This approach would account for the volatility and correlations among the top few assets in order to create a momentum portfolio with the least expected portfolio volatility, along with a palatable risk profile.

Source A New Look At Adaptive Asset Allocation

What are your thoughts on this article and AAA in general? Let us know in the comments below.

[adrotate banner =”70″]

About the Author
Eddie Flower has more than thirty years of trading experience and market insight regarding stocks and derivatives in U.S. and foreign markets. Now semi-retired from a career in financial services, he remains active as an independent trader, financial analyst and writer for onestepremoved.com and quantbar.com.

Related Posts

Leave a Reply