A recent CAIA member contribution by Kathryn Kaminski, director of investment strategies at Campbell & Co., discusses the quantification of CTA risk management. It is worth a look, not least because it amounts to a warning about how underperformance can result for the re-jiggering of allocation for capacity constraints.
Kaminski is the co-author, with Alex Greyserman, of a treatise on trend-following with managed futures, a book published by Wiley in 2014. That book was especially concerned with “crisis alpha,” that is, with the opportunities that may be exploited via persistent trends across markets during periods of crisis. In the new paper, she builds on some of the modeling she developed with Greyserman.
Crisis alpha, by the way, is a fascinating subject, and although the work is too old for a standard book review, I hope in defiance of convention to give it some extended treatment here shortly. For the moment, though, let us focus on the new article.
Portfolio Construction
As Dr. Kaminski writes, risk management “is often cited as a key to success for CTA strategies.”
Beginning with the issue of portfolio construction, Kaminski observes that one simple way to determine position size for each market is this: multiply market conviction by market risk allocation and divide the product by market volatility. Then multiply that by a portfolio scaling factor.
Relevant definitions are as follows:
- Market conviction is the level of confidence in the trend, long or short;
- Risk allocation is the amount of risk allocated to a particular market by the portfolio’s managers;
- The use of volatility of the market as a denominator builds in a proxy for risk.
The process of portfolio construction begins, then, with valuation/model conviction. As these are understood in the CTA world, they are determined by quantitative models, often involving a moving average or channel break-out. The process works from that step toward risk management properly speaking. Most of the article zooms in on what happens after market conviction is taken as a given.
A Baseline and Four Factors
Kaminski zooms in a bit further and comes to four CTA risk management factors: liquidity, correlation, volatility, and capacity. For the allocation strategy she takes as a baseline, equal dollar risk is allocated to all included markets, all 82 of them. The liquidity factor means that the risk allocation is then tilted away from that baseline, toward the more liquid markets. Likewise, the correlation factor tilts the allocation toward markets that are the least correlated with other markets, etc. She looks at the returns attributable to each of these factors against that baseline since 2001, with data up to May 2015. See the table below, adapted from Exhibit 3 in her article:
| Risk management factors | Mean (%) | Median (%) | Standard Deviation (%) | Sharpe | Skew | Max Drawdown (%) |
| baseline | 10.33 | 13.01 | 13.10 | 0.74 | -0.39 | 27.58 |
| liquidity | 0.23 | 0.18 | 1.11 | 0.19 | 0.12 | 6.88 |
| correlation | 0.23 | -0.15 | 1.45 | 0.16 | 0.26 | 4.85 |
| volatility | -0.08 | -0.18 | 0.94 | -0.06 | 0.40 | 6.22 |
| capacity | -0.94 | -1.01 | 2.85 | -0.30 | 0.06 | 26.38 |
Source: Campbell, via Kaminski, “Quantifying CTA Risk Management”
The bottom line is that over this period the liquidity and correlation factors have been positive, the volatility and capacity factors have been negative. The volatility factor is only slightly negative. The capacity factor, though, is strikingly so. In other words, re-allocating risk due to capacity constraints results in underperformance.
Conclusions
Closing in on one year, on the performance of the capacity factor through the year 2014, Kaminski shows that it was near 0% through the first quarter of that year then became progressively negative through the second quarter, and reached its nadir, near -8%, late in the third quarter, before rebounding back to -2% at year’s end. Plotting these moves out by sector, she finds that commodities were the driving force in the negative performance.
Comparing her findings to daily performance data for a universe of managed futures ’40 Act mutual funds for 2014 and the first five months of 2015, Kaminski finds that risk management decisions do seem to explain individual manager performance. This quantitative approach to such decisions and their consequences may, she concludes, be a valuable area “for further research for applications in manager assessment and performance evaluation.”