Do real live hedge funds beat mechanical trading systems designed to replicate their statistical return distribution and diversification properties? In their February 2007 paper entitled “Replication-Based Evaluation of Hedge Fund Performance”, Harry Kat and Helder Palaro update their comparison of the after-fee performances of a wide range of hedge funds to the performances of mechanical replicants. In short, they attempt to isolate true hedge fund outperformance by pitting each actual fund against a replicant that mechanically trades a basket of Eurodollar, 5-year note, 10-year note, S&P 500, Russell 2000 and GSCI futures. Their replication process assumes an existing investor portfolio (to be hedged) that is 50% S&P 500 index and 50% long-term T-bonds. Using return data for 2073 individual hedge funds and 875 funds of hedge funds through November 2006, they find that:
- 18.6% (60.7%) of funds of hedge funds convincingly outperform (underperform) their replicants.
- On average, the performance of funds of hedge funds is fairly stable over time, but there is a tendency for performance reversal on a fund by fund basis.
- 22.5% (57.0%) of individual hedge funds convincingly outperform (underperform) their replicants. These percentages vary considerably across fund categories (see table below).
- Aggregate performance of individual hedge funds deteriorates substantially over time. There is again a tendency for performance reversal over time on a fund by fund basis.
- The deterioration in performance of successful funds suggests that increasing assets under management eventually swamp attractive investment opportunities.
The following table, taken from the paper, shows the number of individual hedge funds evaluated by category, and the number and percentage from each category that clearly underperform (Inefficient) and outperform (Efficient) their mechanical replicants. Overall, only 22.5% of individual hedge funds clearly outperform their replicants. The percentage of outperforming funds ranges from 0% to 51.2% across categories. Confidence in results is, of course, lower for smaller subsamples.
In summary, only about 20% of all hedge funds have produced after-fee returns that clearly beat those of statistical replicants that mechanically trade a basket of liquid futures contracts.
The authors note that their results differ from those generated by other studies using more traditional but less precise benchmarking approaches. They note also that their sample is not large enough to determine whether outperforming 20% of hedge fund managers are essentially lucky or clearly skillful.
