Investment skill of hedge funds : a holdings-based evaluation

Abstract

In Chapter 1, I provide new compelling evidence that hedge funds possess investment skill. Using the longest-in-literature hedge fund sample with fewer biases, I show that large holdings of past winners earn 7% annual benchmark-adjusted return. This remarkable performance is consistent with the notion that large holdings represent managers' best ideas. My sample goes back to 1980 and does not miss non-surviving hedge funds, or those that do not voluntarily report to commercial databases. It consists of all investment managers that must report to the SEC, except those that I identify as managers other than hedge funds. While publicly available data is not sufficient to identify hedge funds directly, my "reverse identification" method achieves both high sensitivity and specificity. I also find weaker yet significant evidence of investment skill in standard indicators such as average fund performance and performance persistence. Additionally, I study the announcement effect of 13F holdings disclosure on the disclosed stock return and trading volume. In Chapter 2, I provide new evidence on market timing by studying ETF option holdings of hedge funds. I find that market option holdings are economically significant in terms of their impact on the market exposure of the funds. Further, I find significant time variation in market option holdings, which could be due to market timing activity. I find that market option holdings are associated with such fund characteristics as active share and market exposure of the fund due to its stock holdings; this evidence is consistent with options being used for hedging. Increases in aggregate hedge fund industry holdings of market put options predict low market returns. In the cross-section of hedge funds, the top 5% group has market volatility timing skill that is distinguished from luck with a bootstrapping test. Additionally, I measure market timing ability as the average risk-adjusted return on market option holdings, which, due to data limitations, requires additional assumptions about option prices. I find that this market timing ability is close to zero for the average fund but it is negative for heavy option users.