31.4 BENCHMARKING OF CTAs

Over time, there have been three approaches to benchmarking managed futures performance. The first has been to use an index of long-only futures contracts. Because CTAs are as likely to be short as to be long, this approach has not been found to be particularly useful. Schneeweis and Spurgin (1997) note that there appears to be little connection between the absolute returns of major indices and the returns of CTA indices, and the results presented suggest that CTA-based indices provide a better benchmark for an actively managed futures portfolio than either passive or active long-only commodity-based performance indices.

The second approach is to use peer groups. In this case, managed futures are most commonly benchmarked to indices representing active or passive futures trading. Active benchmarks of futures trading reflect the actual performance of a universe of CTAs. As previously noted, there are a number of issues that one needs to be aware of when using hedge fund databases, including those reporting CTA performance. Recall that many funds report to only one of these providers and not others. Indeed, some of the best CTAs have no incentive to report to any database. A major problem with using an index reflecting the actual performance of CTAs is that the benchmark is not investable. For traditional asset classes, construction of investable indices is common; relatively inexpensive investable products such as indexed mutual funds and exchange-traded funds (ETFs) are available to investors. In contrast, construction of truly investable indices in the CTA space is complex and may not result in a truly investable index that is an accurate representation of the industry.

Within the second approach, a few indices composed only of managers who invest through managed accounts at a specific firm have recently become available. This would be true of the Newedge CTA Index or the WisdomTree Managed Futures Strategy Fund. These investable indices may suffer from access bias, which could have an adverse impact on the performance of these indices. Access bias arises because some managers may not wish to be part of an investable index due to the restrictions that such indices may impose on them. For example, a manager may be prevented from increasing or decreasing leverage outside a predetermined level, or a manager may have to agree to reduce fees. Therefore, only those managers who have difficulty raising funds outside the investable index platform may agree to be part of the platform.

Finally, CTAs may be compared to passive benchmarks of futures trading. These passive indices represent the performance of an individual algorithmic-based trading system, as opposed to the performance results of CTAs themselves.6 In managed futures, passive indices have been created primarily for trend-following strategies. The oldest and most prominent of these is the MLM Index. Mount Lucas Management has published the MLM Index since 1988. Another passively managed futures index with a reasonable track record is that of the Institute for Global Asset and Risk Management.7

The MLM Index represents the daily performance of a 252-day moving average trend-following system applied to 22 futures markets. Such an algorithmic-based index can be quite useful because it provides the return that can be achieved from a simple trend-following trading model. This return can then be compared to the performance of active trend-following managers in order to examine the value-added of more active (and expensive) strategies.

Although benchmarking trend-following CTAs can be a reasonably straightforward exercise due to the high correlation among trend-following strategies, the same is not true of other managed futures trading styles, such as non-trend-following and relative value. There is considerable heterogeneity among managers in both of these styles. For example, one fund of funds manager noted that among 30 trend-following CTAs examined, the average paired correlation was 0.56, but of 25 non-trend-following CTAs examined, the average paired correlation was zero, indicating no common trading style among these managers. Finding or creating useful indices to benchmark non-trend-following CTA styles remains a significant challenge. A recent paper by Kazemi and Li (2009) examines the performance of both trend-following and non-trend-following managers using portfolios of futures contracts. The portfolios are created using the Sharpe style analysis approach, whereby returns of individual CTAs are regressed against total returns of a set of futures contracts. Kazemi and Li report that the resulting portfolios have significant explanatory power when the performance of trend-following managers is analyzed (the R-squareds are as high as 45%). On the other hand, the performance of non-trend-following managers can hardly be explained with an average R-squared of about 6%. However, non-trend-following managers displayed higher alphas in comparison to trend-following managers.