Commodity Trading Advisor (CTA) investing is the second-biggest hedge fund category, and yet meaningful performance measurement and benchmarks remain elusive. CTAs are hedge funds that use computers to trade on pre-programmed market signals. The hedge funds may go long or short an asset and often use leverage. They are also sometimes referred to as managed futures programs because the assets traded are futures contracts, options, and swaps.

Because CTAs do not disclose their trades in detail, it is difficult to create asset-based benchmarks. So, CTA strategies have tended to be benchmarked against peers, which makes the drivers of CTA performance unclear.

The authors create a strategy, which they call adaptive time series momentum (ATSMOM). This strategy is designed to improve on the peer-group benchmark (and its limitations) and benchmarks derived from the trend-following asset strategies. Its advantages over peer-group benchmarks include increased simplicity, greater transparency, and improved accuracy.

The model is innovative in that it assesses the strength of the aggregated momentum signals created by the prices of multiple assets rather than just the direction of one signal. It also takes into account the real-life constraints that CTA managers face.

In other words, the authors create a rounded performance measurement model that closely mimics the way that trend-following managers actually allocate their portfolios and therefore better identifies the drivers of average CTA performance and the presence—or absence—of manager skill.

A momentum signal typically occurs when the price of an asset or group of assets moves above or below a moving average. There are advantages for CTA strategies — and their performance measurement — of using more than one signal. In terms of building a suitable benchmark, the authors seek to reduce data-mining and calibration concerns.

By incorporating a large number of signals over a large number of time horizons, it is possible to capture signal strength rather than simply signal direction. In this way, the authors create a trend-following benchmark based on their ATSMOM strategy, as opposed to the simpler time series momentum (TSMOM) strategy, which relies solely on binary long and short signals. To do so, they combine a large number of short- and longer-term signals using futures data from January 1990 to September 2015. These signals are compared with monthly net-of-fees returns from 433 CTA funds in the BarclayHedge database.

The ATSMOM strategy is diversified across time and asset classes (namely, commodities, equities, fixed income, and foreign exchange) and allocates more to futures contracts that displayed clear trends. The authors investigate the drivers of performance by analyzing asset classes and the relationship between returns and fund characteristics, such as size, fund flows, and strategy style.

To create an investable — rather than theoretical — benchmark, the authors incorporate market frictions and real-life limitations, such as transaction costs and trade execution delays.

The ATSMOM model is better at explaining CTA performance than existing models because it limits drawdowns and allocates more to higher-performing assets. The authors find that around half of the returns from the ATSMOM model are derived from what they term a “speed factor,” which, consistent with a model developed by previous researchers, isolates excess returns found in funds that buy longer-horizon CTA strategies and sell shorter-horizon ones.

The authors note that only 40% of the variation in individual CTAs’ returns are attributable to the ATSMOM strategy combined with the speed factor. There are a large number of funds that display considerable alpha relative to the model. Funds with positive alphas generate mean alphas of 4.77% a year.

According to the authors, smaller CTAs allocate evenly across asset classes, while larger funds overweight the more liquid futures markets, such as fixed income.

The authors find evidence that fund size is negatively correlated with risk-adjusted performance, whereas the age of a fund is positively correlated with performance.

The fund style, such as which asset classes they invest in and how they invest, also contributes to CTA alphas. For instance, funds that engage in pure trend-following approaches tend to generate higher risk-adjusted performance. Higher equity momentum also leads to superior risk-adjusted performance.

The size of fund flows does not affect risk-adjusted performance, suggesting that capacity constraints are not a big issue for CTA funds. Funds with higher management and performance fees do not achieve better performance than funds with lower fees.

Overall, the characteristics and style of funds only partly explain why some funds outperform the ATSMOM model. The rest of the explanation, the authors argue, must come down to individual manager skill.

The ATSMOM model, particularly when combined with the speed factor, is better at explaining CTA returns than existing models. The model should be able to help fund managers benchmark CTA managers and improve manager selection. An additional finding—that CTA risk-adjusted performance persists from year to year—suggests that investors might want to find a high-performing strategy (or, for better diversification, several successful strategies) and stick with it.

Incorporating frictions and real-life limitations helps ensure that the benchmark is investable, but there is the disincentive of lengthy due diligence and legal processes when changing managers.

Another caveat is that while many real-life frictions are included in the analysis, the relationship between trading volumes and market impact is not. Larger funds, the authors warn, may be forced to trade in more liquid markets and apply slower-moving signals.

Finally, although the speed factor, whereby longer-horizon strategies usually outperform shorter-term strategies, explains some CTA performance, there may be prolonged periods when shorter-term strategies outperform and longer-term strategies suffer.