This paper examines the return characteristics of hedge funds. Hedge funds are unregistered private (not publicly traded) investment pools, typically in the form of limited partnerships, operated by a general partner who charges a fixed fee (usually 1-3% of assets) and an incentive fee (usually 20% of new profits). The data set in this paper was obtained from Tass Asset Management, AIG Global Investors and Paradigm LDC.
The key findings in this paper are as follows.
One, hedge fund returns have very low correlation with the returns of standard asset markets, such as short term interest rates, US stocks, non-US stocks, emerging market stocks, US government bonds, non-US government bonds, gold (as a proxy for commodities), and the traded weighted US Dollar (as a proxy for foreign currencies). This is very different from US mutual funds, whose returns have high correlation with these standard asset markets.
Two, there are five dominant hedge fund investment "styles." Generally, hedge fund consultants classify hedge funds into "styles" based on the managers' descriptions of their trading strategies. We use a quantitative classification scheme. The idea is very simple. If two managers use the same trading style on the same markets, their returns are correlated to each other, even if they are not correlated to the returns of any asset markets. Principal component analysis is a statistical tool to group funds based on their correlation with each other. We find five principal components in hedge fund returns which can jointly explain roughly 45% of the cross sectional variation. This shows that hedge funds have many different investment styles. By examining the funds mostly strongly correlated to these five principal components or styles, we associate names with these styles. Two styles are associated with futures traders or commodity trading advisors (CTAs); one style is associated with "global/macro" asset allocators such as George Soros; one style is associated with long/short equity traders; and one style is associated with funds investing in the "distressed" securities of firms who near, in, or coming out of bankruptcy.
A future research goal is to examine some niche styles which are not identified by the principal component analysis.
An important open issue is "survivorship" bias. Our data set consists of funds which were in operation at the end of 1995. We were unable to obtain any information on funds which ceased operation prior to that date. The returns of surviving funds can bias upwards the returns of all funds. We are in the process of gathering data of defunct funds to do a proper "survivorship" analysis.