Tactical Asset Allocation Can Be Successful – With the Right Model

Portfolio returns in excess of an index can be achieved through active investment management in two ways: security selection, and active (or tactical) asset allocation. Research shows that about 90% of a typical balanced stock-bond portfolio risk and return comes from Policy asset allocation – see Brinson et.al. (1986, 1991), Ibbotson & Kaplan (2000). Clearly, potential for adding value through actively managing asset allocation is at least as large as from active security selection. However, while active security selection is widely practiced, tactical asset allocation (TAA) has been largely overlooked or out of favor. Here, we discuss some of the reasons for this, and describe the process that should be followed in order to successfully perform TAA.

TAA is an investment strategy that centers on altering investment proportions to take advantage of differences in expected performance and risks of broad asset classes (such as stocks and bonds) or sub-classes (such as U.S. and global equities). Several requirements to the investment process stem from this definition. First, the responsibility for TAA must be placed with the group within the investment organization that spans across asset classes – typically the office of the Chief Investment Officer. Second, it has to be based on accurate, timely asset mix information (actual and benchmark). Thirdly, the effect of these investment decisions has to be measured as part of performance evaluation. Lastly, TAA decisions have to be largely based on systematic quantitative results rather than on judgment.

Because the implications to performance are so large, many investment managers already attempt actively managing their asset allocation, even if implicitly. You may hear this in your investment strategy meeting: “We want to be positioned defensively due to anemic economic recovery in the U.S.” (or due to “debt crisis in Europe” or whatever the current concern may be), or “We would like to take advantage of the rally in equities.” However, doing this implicitly, without the proper process and structure, is dangerously likely to result in underperformance.

Clearly, successful TAA requires timely, accurate calls on expected asset class performance. “Active management is forecasting”, say Richard Grinold and Ronald Kahn in their well-known book Active Portfolio Management (1999). The authors establish the following relationship between active return (alpha) and forecasting skill, or information coefficient (IC):

α = σ * IC * Score

The key to achieving good performance from TAA, therefore, is the skill (IC) of forecasting asset class returns.

This, of course, is not easy. Qualitative judgment is likely to be affected by the prevailing sentiment in the market which will be exactly wrong at market turning points. Many managers use a set of indicators to help determine future market direction. This is a step in the right direction, but at any point in time, there usually is about the same number of indicators that give a positive signal as negative. How do we know which indicators are currently relevant, and what the proper weights are to each? In addition, a set of disjointed indicators cannot produce a history of return forecasts, which is required in order to determine if the method has any skill. One needs to combine predictor factors into a consistent statistical model that produces return forecast series, correlating which to actual returns gives IC.

Numerous published studies have tested predictability of the stock market via factor models. Generally, they find little evidence of out-of-sample forecasting ability; small excess returns that are achieved by some models often don’t justify the costs. It is common to interpret these results as not supporting the idea of actively managing asset allocation at all. But there is light at the end of the tunnel!

We attribute the lack of success of the forecasting models commonly described in academic literature to two reasons. First, while some of them are quite sophisticated from the statistical standpoint, they tend to miss important aspects of what works in investment practice. Second, researchers often limit the set of factors to only a few variables that are commonly described in macro-economic literature as drivers of business cycles. We found that using much broader set of variables selected empirically rather than fitting a pre-defined economic theory, is necessary to build a model with good forecasting ability. These variables should include economic, valuation and market factors employed by investment managers as predictive indicators. An example of a factor that is not common is the CBOE implied volatility index (“VIX”), which is known by practitioners to be inversely related to market returns.

Thus, we recommend that investment organizations develop return forecasting models that address these shortfalls, provided that the organization can devote proper resources to it. The focus should be on equities as the main source of return variability in a balanced portfolio.

Alternatively, an investment manager may wish to partner with a research film that provides return forecasting. This solution has clear advantages for many managers. First, it is cost-effective – creating an internal research team to spend considerable time (likely years) developing models would be expensive, and success would not be guaranteed. So, outsourcing this from a vendor of return forecasts may be the only viable solution for sophisticated smaller managers. Secondly, immediate access to vendor return forecasts can potentially help improve client portfolio performance much sooner than an internal solution.

Leave a Reply

Your email address will not be published.