The Performance Persistence, Flow and Survival of Systematic and Discretionary Commodity Trading Advisors (CTAs)

Abstract

This thesis studies the performance, performance persistence, survival and flow of Commodity Trading Advisors, also known as CTAs or Managed Futures Funds. One of the main contributions of this thesis is the novel classification of CTA strategies. This is obtained by hand-collecting information frequently by directly contacting the funds in the database. I thus identify two main trading styles: Systematic and Discretionary CTAs which are the main focus of this thesis. I further separate Systematic CTAs into trend-followers with differing trading horizon. This novel dataset allows me to reconsider many hitherto studied issues in the CTA space with an application to these sub-strategies. The first section investigates the differences in mortality between Systematic and Discretionary CTAs, over the longest horizon than of any in the literature. A detailed survival analysis over the full range of CTA strategies is provided. Systematic CTAs have a higher median survival than Discretionary CTAs, 12 vs. 8 years. I hand collect information on reasons for exit from the database. I propose new filters that will better identify real failures among funds in the graveyard database. Separating graveyard funds into real failure I re-examine the attrition rate of CTAs. The real failure rate is 11.1%, lower than the average yearly attrition rate of 17.3% of CTAs. The effect of various covariates including several downside risk measures is investigated in predicting CTA failure. Controlling for performance, HWM, minimum investment, fund age and lockup, funds with higher downside risk measures have a higher hazard rate. Compared to other downside risk measures, the volatility of returns is less able to predict failure. Funds that receive larger inflows are able to survive longer than funds that do not. Large Systematic CTAs have the highest probability of survival. The second part studies the performance and performance persistence of Systematic and Discretionary CTAs. Controlling for biases, after fees the average CTA is able to add value. These results are strongest for large Systematic CTAs. I extend the sevenfactor model of Fung-Hsieh (2004a) and find that this model is better able to explain the returns of Systematic rather than Discretionary CTAs. I find three structural breaks in the risk loadings of CTAs different to hedge fund breaks: September 1998, March 2003 and July 2007. Using these breaks I show that systematic CTAs were able to deliver significant alpha in every sub-period. I also find evidence of significant performance persistence. However, these findings are heavily contingent on the strategy followed: the persistence of Discretionary CTAs is driven by small funds whereas large funds drive the performance persistence of Systematic funds. These results have important implications for institutional investors who face capital allocation constraints. They also suggest that contrary to the previous findings, the CTA industry does not appear to be heading towards zero alpha. The final section looks at the relationship between fund-flows and performance. Investors chase past performance, the fund- flow -performance is significant and concave for some strategies. Although there is some long-term performance persistence of Systematic funds with the highest inflows, there is no smart money effect in the CTA literature. I find no evidence of capacity constraints among Systematic CTAs. Investors are thus not able to smartly allocate funds to future best performers and take full advantage of the liquidity that CTAs offer.