Why Hedge Funds Destroy Investor Wealth
August 7, 2012
by Michael Edesess
The battle over hedge fund performance measurement
Lack’s book disposes of the belief that hedge funds have achieved superior performance, by explaining how that performance should be measured. Before taking up his approach, let me first briefly summarize some of the more blatant errors in hedge fund performance reporting that others have previously explored.
Studies of overall hedge fund performance have long yielded wildly divergent results. Unfortunately, the financial media uncritically pass on to the public the simple averages of reported hedge fund performance assessments that are spun out by database providers, without noting the deeply flawed nature of those figures.
An often-cited study is “The ABCs of Hedge Funds” by Roger G. Ibbotson, Peng Chen, and Kevin X. Zhu. In it, the authors find that when biases inherent in the hedge fund databases are corrected, average performance during the period 1995-2009 falls by more than 7% annually. But they consider only survivorship bias and backfill bias. (Survivorship bias stems from the fact that performance data for funds that were later closed are not among the database figures. Backfill bias stems from the fact that funds that try out their strategy for a while before deciding that their numbers are good enough to go live backfill their trial performance into the database once they do.)
Ibbotson, Chen, and Zhu point out, however, in an endnote, that they did not account for every source of bias. “Another bias often cited in the hedge fund literature is selection bias, which refers to not having a representative sample of funds,” they observe. “Back-end bias can also be a problem if hedge funds stop reporting after a bad month. In our study, we concentrated on survivorship bias and backfill bias.”
Hedge funds demonstrably do stop reporting after a bad month. Long-Term Capital Management’s sudden decline in 1998 did not appear in the databases, and, as Lack himself notes with respect to the Bernie Madoff fraud, “Madoff represented more than 3 percent of the entire hedge fund industry in 2008 … Madoff typically isn’t included in the returns for that year reported by most databases – the year was already quite bad enough.” (Madoff’s fund was, strictly speaking, not a hedge fund, but the point is relevant nonetheless.)
Selection bias is difficult to assess because all the major databases, of necessity, allow hedge funds to report their performance numbers only if they want to. This means that the average performance numbers reflect only averages of self-selected, willing reporters of their own performance. However, in an earlier Advisor Perspectives article, which covers all the biases in hedge fund databases, I reported the findings of another study (“Out of the dark: Hedge fund reporting biases and commercial databases,” by Adam L. Aiken of Quinnipiac University and others), which used a data source untainted by self-selection (funds-of-funds that register with the SEC and hence must report the performance of all the hedge funds they invest in) and found that self-selection bias alone slices 4% off of hedge funds’ average performance. When combined with the biases already mentioned, that’s enough to wipe out any trace of alpha.
Nevertheless, star-struck chroniclers of the hedge fund industry, such as Sebastian Mallaby, cite an isolated figure in the Ibbotson, Cheng, and Zhu article. While Ibbotson, Cheng, and Zhu find that hedge funds actually underperformed the S&P index over the period 1995-2009 on a risk-unadjusted basis, risk-adjusted they achieved on average an alpha of 3%.
But even this supposed alpha, in the first place, does not account for the 4% correction for self-selection bias found by Aiken et al. In the second place, many hedge fund investments are illiquid and not easily priced; this leads not only to some shading of prices (which are assigned by the reporting fund managers themselves) in the fund’s favor, but also to smoothing that artificially lowers risk measures like volatility and beta, resulting in spuriously high risk-adjusted returns.