Winning a sports tournament is always a major accomplishment. But it’s perhaps even more of an accomplishment when the entire tournament is a giant elimination round. Some tournaments, such as the Olympic hockey tournament or the World Cup of soccer/football allow teams to log a loss or two in round-robin play. But others, like Wimbledon or the NCAA Men’s Basketball Championship require teams to maintain totally perfect records in order to advance.
As a result, the four remaining teams in US college basketball’s “March Madness” all have records of 4-0. Does this mean that the average tournament team has a perfect record? Of course not. Sixty teams have less than perfect records but did not actually survive to this point in the tournament to be counted (actually 61 if you count the “play-in” game where the 64th and 65th seeded team fight it out for a change to get shellacked by a #1 seed). The difference between the perfect records of today’s survivors and the average records of all teams knocked out so far amounts to a “survivorship bias”.
The situation isn’t all that different in the ongoing tournament amongst hedge funds. The funds that survive to this day are more likely to be the ones with winning records. The funds that aren’t around anymore are more likely to have had losing records (otherwise, they’d likely still be around).
Survivors of 2008
A new academic paper by Xiaoqing Xu, and Anthony Loviscek of Seton Hall University (which, by the way, had a killer men’s basketball team in the early 1990’s but hasn’t been to the Big Dance since 2006) and Jiong Liu, of TIAA-CREF (yet to field a contender mainly because it’s a financial institution, not an educational one), explores the survivorship bias amongst hedge funds during the recent financial crisis.
They wondered if the global financial crisis of 2007-2009 might have led to higher attrition rates and therefore a higher survivorship bias in subsequent industry return data. This makes intuitive sense. After all, pundits described 2008 as a “culling” of weaker funds. Assuming those weaker funds had lower returns than average, the aggregate performance of the remaining funds must have been higher. In their words,
Without a careful accounting for the effects of this shift, including adjustments for survivorship bias and attrition rates, the accuracy of performance assessment and empirical modeling efforts that incorporate the financial crisis period are in doubt.
Previous research has shown the survivorship bias in hedge fund returns to be in the neighbourhood of 2% per annum. Xu, Liu, and Loviscek find that the average difference between the historical monthly returns of the funds that are currently “dead” and the historical monthly returns of the funds that are still around is 0.26% or about 3% per annum (comparing returns of the same months across time).
But curiously, that difference actually shrinks in 2008. In other words, the survivorship bias of 2008’s returns was lower than in other years – despite the fact that hedge fund attrition actually doubled or tripled that year.
What gives? According to the researchers, there is a “hidden survivorship bias” that might explain some of this paradox (think “dark matter” in hedge fund databases). They create a new metric that compares only the final 12 months of historical returns of (currently-) dead funds with the performance of the average hedge fund performance across all funds (both dead and living) in the same months.
The average “hidden survivorship bias” is a whopping 0.54% per month or around 6% per annum according to the researchers. (see chart below constructed from data contained in Table 2 of the paper).
It turns out that funds of funds fared much better that their single-manager cousins. The average survivorship bias was close to zero and the “hidden survivorship bias” was 0.32% per month or about 4% percent per annum.
Xu, Liu, and Loviscek take a dim view of hedge funds as a result, concluding:
Combining the high attribution rate of hedge funds during the recent financial crisis and this hidden survivorship bias, we believe that the reported HF return data led to an overstatement of their performance and consequently an understatement of investor losses.
In fact, despite hedge fund indices blowing away long only indices by around 20% in 2008, the trio writes that hedge fund managers…
…did not, on average, deliver significantly positive risk-adjusted excess returns during this period, suggesting that the superior performance that managers deliver in good times does not hold up in bad times.
It’s important to note, however, that they use a 10 factor model to isolate alpha produced during 5 different periods between 1994 and 2009. While hedge funds did outperform markets during the 2008 financial crisis, they find that returns can be explained using non-linear factors such as “lookback straddles” that act like puts written on various indexes (first identified as hedge fund factors by Fung and Hsieh).
So the question becomes: should hedge fund managers receive any credit for changing their factor exposures going into 2008? We’d argue they do. Since market timing of macro bets like these add value and – for the average longer-term investor – presents itself as alpha vs. the investor’s other longer-term alternatives. When you subdivide the past into several market regimes, you lose some of the ability to argue returns are passive when factor exposures are dynamic.
To their credit, the authors of this paper also try to provide some prescriptions for investors and policymakers. They re-iterate the importance of due diligence, particularly since many hedge funds have decided to “…reduce the minimum threshold and offer these products to an increasing number of small investors”. They also advocate “greater transparency”, the removal of manager discretion in reporting returns to databases, and random SEC audits.
These recommendations are indeed sound. But they don’t seem to fall from the actual research contained in this paper. While some managers do not report their full return history to hedge fund databases, the actual investors in each hedge fund are usually fully aware of the track record of their fund. Additional due diligence is warranted for operational risk mitigation – but not required in order to simply collect more historical returns. “Survivorship bias” is a reality whether or not reporting is mandated by regulators, and random SEC audits don’t really help investors understand performance any better (unless, of course, the manager is cooking the books – but that was beyond the scope of this paper).
Yet despite their interventionist recommendations, the trio actually argues against regulating hedge funds “along the lines of mutual funds” – precluding them from “dynamic trading strategies.” This position, they write, is “…in the interest of promoting price discovery, providing liquidity injections, and offering investors an alternative asset class.”
In the end, only one college basketball team will have a perfect record at the end of the NCAA tournament (say, Duke, for example). Even great teams – such as the other Final Four teams that do not begin with “D-U…” and end with “…K-E” – will end up with imperfect records. But at least those teams can take solace in an undeniable trait shared by all but one team in the tournament (that being, for example, Duke): they will have exactly one loss when all is said and done – no more, and no less.