Knowing and understanding the likely performance of an asset is a critical step in determining whether it has a place in the pension fund’s investment portfolio. Having effective benchmarks against which to measure manager performance is a prerequisite to proper monitoring of any investment. On both these counts, the hedge fund market falls down. Historic performance of hedge funds is flattered by methods of construction that ignore the monetary effect of failed funds. Some indices compound this by adding in the back-history of new, more successful funds. And there is no central depository for hedge fund performance data, so index providers are reliant on managers to volunteer accurate information.
There is no good reason to expect any one series of hedge fund indices to report the same information as another and this is amply demonstrated by historic monthly performance statistics (see table). Each provider has its own selection criteria for inclusion, its own construction rules and classifications for strategy sub-indices. Indices are essentially being created from completely different underlying information. Research performed by Fung and Hsieng (Benchmarks of Hedge Fund Performance: Information Content and Measurement Biases – December 2000), found that between the 1,162 funds in the Hedge Fund Research (HFR) database and the 1,627 in the CSFB/Tremont database, there were just 467 overlapping funds.
All hedge fund indices suffer from survivorship bias. This means that when a fund fails it is simply replaced by another fund in the index, and losses incurred by investors in the now defunct fund are no longer reflected in index returns. Bing Liang, in his research paper ‘Hedge Funds: The Living and the Dead’, published March 2000, estimated annual probability of failure of between one in five and one in seven, dependent on the fund strategy. This high level of attrition means that the survivorship bias is by no means trivial, Fung and Hsieng estimating that it amplifies index returns by 3% pa on average. Furthermore, some index providers rewrite history by backfilling the previous year’s performance of any new fund, and this was estimated to add 1.4% pa to performance by Fung and Hsieng.
Index providers vary in their coverage of different styles contained within the hedge fund market. For instance, HFR publishes 31 different strategy sub-indices but eschews managed futures funds, whereas CSFB/Tremont produces a managed futures index alongside eight other strategy sub-indices. Zurich Capital Markets are even more restrictive, seeing only five strategies as being sufficiently homogeneous for an index to be a meaningful concept. The managing director there, Garry Crowder, explains, “hedge fund products that depend overly on asset allocation and leverage have insufficient common ground to be categorised within an index.” To be appealing to the institutional investor, Zurich also apply stricter criteria for inclusion than other index providers, requiring a length of time in operation of three years and more than $50m under management. By comparison, CSFB/Tremont only require audited financial statements and at least $10m. And different index providers are more or less open about their index constituents. Zurich Capital Markets broadcast managers and weightings in advance, CSFB/Tremont publish lists of constituent funds, but HFR do not reveal their index constituents.
There are significant differences in index calculation, from simple arithmetic addition, favoured by HFR, to weighted by fund value, as per CSFB/Tremont, or a median performance figure as is reported by Zurich Capital Markets. Differences between these methodologies are most pronounced in periods of extremely strong or weak performance. This divergence arises because to maintain the same proportion in each fund, the simple implicitly invests and divests conversely to fund performance, whereas a weighted index automatically adjusts weightings in line with performance. The trend seems to be moving away from simple averages or medians, as investors seek a benchmark that reflects investability as well as potential returns. Established index provider MSCI is developing its own set of indices, in consultation with institutional investors, categorising funds on the basis of three primary characteristics, investment process, asset class and geography. MSCI head of hedge fund indices, Michel Serieyssol, says “the MSCI methodology strikes a fine balance between granularity and aggregation, providing users with a set of versatile tools to capture and analyse the performance of different strategies. In terms of calculating the index value we intend to provide equally-weighted index series.”
But in contrast to traditional stocks and shares, size does not necessarily translate into investability. Some of the largest, most well-established hedge funds have been closed for many years. To make an investment case for an asset class using an index where as many as 25% of the constituents are not investable is probably flawed.
Leaving aside whether it could be shown to be desirable, with minimum sizes for some funds of as much as $1m, tracking an index of hedge funds would require a small fortune. To construct a replicating basket of an AllFunds type index would need some $5bn, given that there are some 5,000+ hedge funds in operation. Creating a portfolio of sufficient size to mimic the risk characteristics of an index requires of the order of 120 funds, according to work done by Fung and Hsieng and this would cost about $100m. To reduce standard deviation down to within 10% of the index value requires significantly more hedge funds than are needed to create the same effect in an traditional asset class, for instance, equities, and reflects a significant commonality of risks between hedge funds.
Most hedge funds have a correlation of less than 0.3 to broad-based market indices and this lack of correlation is used to justify hedge fund investment on the grounds of diversification of returns. Often consultants will favour a fund of funds or portfolio of hedge funds as a means to reducing dependence on individual managers or hedge fund styles. Curiously however, Fung and Hsieng found that both the HFR and CSFB/Tremont indices had strong positive correlations to US equities, non-US equities, emerging market equities and high yield bonds. This begs the question; by diversifying hedge fund exposure to reduce manager and strategy risk, is the hedge fund investor introducing systematic exposure to traditional risk factors?
An alternative to using a global index that aggregates the returns of selected funds is a fund of funds index, which more accurately demonstrates the returns achievable from real hedge fund investment. By adding back the management charges for a fund of fund (estimated at 2% pa by Fung and Hsieng) one derives a fair estimate of the actual performance of the underlying funds held by the funds of funds contained within the indices. A fund of fund index should remove the survivorship bias that hedge fund indices suffer, as the NAV of the fund of fund will reflect losses from failed funds. Equally there is no possibility of biases arising from ‘instant history’ that comes from adding in new funds, because the fund of fund can only invest at today’s price.
Critics of this route point to the different investment philosophies and asset allocation strategies of individual funds of funds and suggest that this introduces a bias of its own. Also fund of fund index performance will be quoted net of an additional layer of fees charged by the fund of fund adviser. But for institutional investors such as pension funds, who often make their first investment in hedge funds via fund of funds, it could be the most sensible option for benchmarking, both to evaluate the success of one’s hedge fund investment, and also the skill of the fund of fund adviser.
For those pension funds who are further down the route of selecting individual hedge funds according to preferred strategies, selecting a representative strategy index against which to measure selected managers is a major challenge. This task is muddied by the different rules that each index provider operates for determining the classification into which a fund falls. Ideally a strategy sub-index should reflect the common denominator for all managers pursuing that style. One method, employed by HFR, to isolate funds for inclusion is to use statistical techniques, such as cluster and correlation analysis, which identify funds that operate closest to the purest form of the strategy.
Research has shown that almost all of the popular hedge fund styles can be broken down into combinations of options and other derivatives on traditional asset classes. David KA Mordecai, editor of the Journal of Risk Finance, member of the Investor Risk Committee of the International Association of Financial Engineers and a managing director at a large New York-based hedge fund, suggests that further fundamental research should be done to isolate the sources of return that are key to each hedge fund style. The results of this analysis could then be used as a reference against which to measure manager skill.
Some hedge fund aficionados rail against the concept of benchmarking hedge fund performance, seeing it as inconsistent with the essential eclecticism of the market. Historically, investors in hedge funds have been high net worth individuals targeting absolute rather than relative returns. As institutional investors join the party, some means of extrapolating future returns and monitoring manager performance has to be found. But the diversity and lack of homogeneity that characterises the hedge fund market means that any attempt to index or benchmark has to look outside the norm.

Topics