Hedge Funds: The Living and the Dead

Motivation(s)

Several studies on hedge funds have tried to document survivorship bias, but often draw conflicting conclusions due to the usage of different hedge fund data. The bias comes from the fact that data vendors collect only survived funds, which can make the resulting analysis overstate a fund’s performance. Since hedge funds report information on a voluntary basis, minimizing bias may require examining the dataset.

Proposed Solution(s)

The author propose examining survivorship biases in hedge funds by comparing different databases (e.g. TASS, HFR) under the lens of different investment styles. While the evaluation of survivorship bias is nothing new, this analysis is trying to address the reliability of commercial hedge fund data.

Evaluation(s)

The author examined the hedge fund data from TASS and HFR. The average survivorship bias is over 2% per year, which is consistent with existing studies. Contradicting reports of smaller bias is due to the relatively low number of dissolved funds in the HFR/MAR database. Note that the survivorship biases are different across styles. They are significant for 10 out of 15 styles in TASS but none is significant for HFR. As exhibited in a fund’s declining returns toward the date of liquidation, the reason for its disappearance is mainly poor performance.

Across the two databases, at least 5% of return numbers and 5% of NAV numbers differ dramatically. TASS has more return observations and NAV observations due to including more funds and a longer return history than HFR. The return numbers in TASS are consistent with the NAV numbers. TASS has more funds with incentive fee and management fee information than HFR. Mismatching between the reported returns and the percentage changes in NAVs can partially explain the difference.

The author promote the TASS data for doing hedge fund research because of its relative completeness and accuracy.

Future Direction(s)

  • How would one construct models that are resilient to inaccurate data?

  • How can unsupervised learning determine the data’s accuracy?

Question(s)

  • How much would the switch from a probit regression to a logit regression change the underlying analysis?

Analysis

When computing any statistic over data, one should verify whether the data should exhibit such characteristics before proceeding to a conclusion. The reason for conflicting survivorship biases was due to a lack of understanding of the data. Although a reasonable cause was identified, the author should have quantified how much these biases influence other statistics. At such a small percentage, is it even worthwhile to filter out the data? One interesting point, which reinforces the practice of verifying the data, was that one previous study claimed the small bias was due to onshore and offshore funds in HFR despite the data showing otherwise. The baseless use of probit regression stands out the most in the analysis; perhaps it was selected out of conventional practice in economics.

References

Lia00

Bing Liang. Hedge funds: the living and the dead. Journal of Financial and Quantitative analysis, 35(03):309–326, 2000.