Measurement Biases in Hedge Fund Performance Data: An Update

Abstract

Tending to be static and single-database oriented, existing models for correcting performance measurement biases are unable to detect potential data errors arising from (1) hedge funds that migrate from one database vendor to another and (2) merged databases. In general, return measurement biases can be traced to two key events: when a hedge fund elects to enter one or more databases (backfill bias) and when a hedge fund exits a database (survivorship bias). Artificial rules (e.g., ignoring the first x number of months of performance history to minimize backfill bias) and survivorship statistics based on a single database vendor are susceptible to another form of bias as databases evolve and consolidate. The authors posit that one must be mindful of how much of the hedge fund industry one is observing before passing judgment on the performance statistics of the hedge fund industry as a whole.

Are the halcyon days of the hedge fund industry behind us? Can the long-term performance of the hedge fund industry justify the attendant high fees and its correlation with conventional asset classes during downturns? Answering these questions demands accurate performance statistics. Tending to be static and single-database oriented, existing models for correcting performance measurement biases are unable to detect potential data errors arising from (1) hedge funds that migrate from one database vendor to another and (2) merged databases. As the hedge fund industry contracts, this issue is likely to become an important consideration in compiling hedge fund performance statistics. This article summarizes the key return measurement biases in a simplified framework and alerts researchers to the types of dynamic biases that could occur as the industry consolidates.

In general, return measurement biases can be traced to two key events: when a hedge fund elects to enter one or more databases and when a hedge fund exits a database. Superficially, these biases appear to be static, requiring only a one-time adjustment. When funds migrate from one database to another, however, survivorship statistics based on a single database vendor may be biased. Missing funds need to be differentiated from liquidated funds. For example, around the end of the first quarter of 2007, fewer than 50 percent of the funds in the Hedge Fund Research graveyard database were classified as liquidated. Put another way, half the funds that were removed from the live database could have been in some other database or could have elected not to report their performance data.

The concern over backfill bias revolves around the informational content of a fund’s performance history, from its inception date to its database entry date. All artificial rules (e.g., ignoring the first x number of months of reported performance history as a proxy for excluding a fund’s incubation period) are susceptible to other forms of data error. Setting x equal to the distance from the inception date of a fund to its database entry date may cause the exclusion of valuable performance history when databases merge. In March 1999, TASS sold its database to Tremont Advisors. Some time passed before Tremont was able to identify and consolidate into a single database that portion of its hedge fund data that was incremental to TASS’s data. This activity took place mostly in 2001. This idiosyncratic event caused the number of new funds entering the TASS database to jump from 309 in 2000 to 932 in 2001, of which 544 were added in September. Clearly, not all the data prior to the first 2001 reporting date for those funds added to the TASS database are “backfill-biased.”

Overall, the literature tends to focus on the negative aspect of a hedge fund’s ability to choose when and where to report performance—namely, that poor performance is unlikely to be submitted to databases. A counterargument is to observe that close to 40 percent of the Institutional Investor Top 100 hedge fund firms do not report to the four major databases that we analyzed. Assuming that top hedge fund firms are likely to have above-average performance, this fact could represent a sizable bias in the opposite direction—namely, that good performance may also be excluded. This article posits that one must be mindful of how much of the hedge fund industry one is observing before passing judgment on the performance statistics of the hedge fund industry as a whole.