EUROPE - Investors may like to think the performance numbers hedge funds report to databases reflect the actual returns from the assets they hold in their portfolios. At the very least, they probably expect that, once the final audit of asset returns and NAV is in, reported numbers will be revised promptly and finally.
But that is not the finding of a new empirical study presented at the 4th Lyxor/NYSE Liffe Hedge Fund Research Conference by Tarun Ramadorai of the Said
Business School and Oxford-Man Institute at the University of Oxford.
‘The reliability of voluntary disclosures: evidence from hedge funds’, co-authored by Ramadorai with Andrew Patton of Duke University and Michael Streatfield of the Said Business School in the University of Oxford, reveals that fund managers have revised their reported returns for months as far back as the late 1990s.
It also shows that revising activity is strongly predictive of performance - dud managers tend to tinker more with their historical records than good ones.
The paper considers a series of monthly downloads from five of the leading hedge fund databases, dating from 2007 to 2011.
Each of these downloaded ‘vintages’ of the databases covers performance histories stretching back to 1994.
While the majority - 60% of funds - had been subject to no revisions of past performance records, 40% had revised at least one past record by at least 1 basis point, 32% had revised an old record by as many as 10 basis points, and 22% had made revisions of as many as 50 basis points.
Some of the revisions changed a one-month return from a positive to a negative, and most revisions were downward.
These sound like small numbers, but, when the authors run the average performance of their sub-set of ‘report revisers’ before revisions and after revisions, they find that the difference in cumulative return over the 18-year time period is 55% - the actual return finally reported is 55% lower than the return according to the funds’ initial reported numbers.
And this is just an average - some funds exhibit significantly lower actual returns due to revisions.
“That’s a big result for something that’s essentially an accounting revision,” Ramadorai said.
The pattern of ‘serial revisers’ is not random, according to the paper.
Perhaps unsurprisingly, revisions are most common to months coinciding with the crises in 1998, 2000-01 and 2008-09.
Larger and longer-established funds make more revisions (Ramadorai said the relation between the revising habit and size was almost linear), as do the most volatile funds, funds pursuing illiquid strategies, and, among styles, macro, emerging market and ‘security selection’ strategies.
However, the most striking and pertinent relationship observed was that between ‘serial revisers’ and poor performance - a result the authors claim as a first for their paper.
To establish this relationship, they ran the performance of a ‘reviser’ portfolio alongside a ‘non-reviser’ portfolio, and found consistent outperformance from the ‘non-revisers’.
Moreover, recognising that there can be legitimate reasons for revising reported returns - such as pursuing an illiquid strategy in difficult-to-mark securities, running a very large portfolio, or correcting historical record following the appointment of a new auditor - the authors repeated the performance comparison correcting for these factors and found that differentiation became even larger.
While Ramadorai stopped short of saying so explicitly, he suggested that ‘serial revision’ was more often than not a “dishonest” practice.
Investors might consider possible explanations for downward revisions of returns that were reported so many years ago.
On the one hand, it is not the revisions that should concern them - but the fact the returns may have been deliberately overstated in the first place.
Do managers feel they can get away with revising very old returns because potential investors focus more on recent performance?
On the other, the importance of the act of revision itself as a predictor of fund quality should certainly give investors pause.