In this article we consider some technical questions on the construction of investable indices before briefly reviewing the main indice
In this article we consider some technical questions on the construction of investable indices before briefly reviewing the main indices commercially available.
General considerations in hedge fund index construction
An index is a means of summarising information about a broader population. In the case of public exchange traded financial securities such as equities, this population can be identified exactly and the various competing indices can be compared in respect of their ability to capture the information in the population in an efficient manner. Even so there may be no single ideal index because sometimes a capitalisation-weighed index will be preferred to an equal-weighted index but sometimes not. But everything is transparent and users can make clear choices.
Matters are less much less clear with hedge funds because there is no way of identifying the total universe or segments within it. This is because hedge funds are broadly unregulated and are not obliged to provide details to outsiders. So the starting point for any index is the database of funds the index hopes to represent. Several providers have established databases of hedge funds (usually including commodity trading advisers but not always) in the last decade or so. One might expect that they would track substantially the same set of funds, so that each database could be seen as a ‘sample’ from the underlying but unobservable population. In this case the growth of samples would lead to a growing confidence about the scale and characteristics of the underlying population. Unfortunately the various databases have surprisingly limited overlap. Lhabitant (2004 p.114) shows that the TASS, HFR and CISDM/Hedge databases only overlap for some 10% of the total funds in the combined group. Each fund accounts for little more than half of the combined total of funds. Users therefore may lack confidence in the ability of any one database – and therefore index – to capture the whole hedge fund universe at this point.
The lack of transparency of hedge funds leads to other problems that are partly familiar from the mutual fund world. These are potential biases that can creep into indices and provide an inaccurate measure of what is going on in the population. The main biases that have been identified in hedge fund indices are:
-
Survivorship bias. Many hedge funds cease operations and drop out of the universe and therefore the index; commentators normally assume that these funds have failed or suffered large redemptions although there is evidence that many funds close because their managers want to move on, in which case the funds may have been successful. But, to the extent that most funds that drop out are poorer than average performers, the index will overstate the returns an investor in a ‘typical’ hedge fund would expect. This problem is equally applicable to conventional equity indices (and to the population of stock market indices as a whole) but it matters more in the hedge fund world because of the relatively short time periods over which the effect can occur. Most index providers therefore now continue to include defunct funds in the index as of their last reported performance.
-
Self-selection bias. Owing to the private nature of the industry, hedge funds may choose whether to report to a database. If non-reporting funds are systematically different from the reporting funds then the database and any indices based on it will be biased. But the nature of the bias is ambiguous. Poorly performing funds may choose not to report, which would mean the index was biased upwards. 1 But well established and well performing but closed funds may also choose not to take part, in which case the bias could be negative. Of course if funds are closed they are not relevant in any case to an investable index, which therefore must necessarily diverge from the established universe of funds, even if that could be accurately established.
Criteria for a good index
Assuming that the database from which an index is constructed is good enough to represent the underlying population of hedge funds, a useful index ought to be:
-
Transparent: the index provider should readily provide the constituents, criteria for inclusion and process for making changes
-
Representative: the index should include funds from the full range of fund types, including various strategies, sizes and ages of fund; most importantly the fund should include both the best established and very successful funds as well as the less successful ones
-
Timely: the index needs to be published regularly, at least monthly
-
Stable: there should be no revisions to the index
The question of weighting is somewhat controversial. Whereas the original Dow Jones Industrial 30 index of leading US stocks was and remains equally weighted, the main indices of the US stock market are now capitalisation weighted eg the S&P 500. This is regarded as obviously correct by most users.
But in the hedge fund world the scale divergence between the largest and smallest funds is enormous. Most hedge funds indices are therefore equally weighted, the
main exception being MSCI. An index weighted by assets under management (AUM) would arguably be more representative of the industry, but would practically exclude the influence of many new and small funds. These funds are more likely to be open to investors so and investable index is, in effect, forced to be equally weighted.
Another argument for equal weighting is diversification. To the extent that the index intends to capture the experience of a ‘representative’ hedge fund investor, that investor would probably wish to have some degree of diversification over funds and strategies. An equally weighted index should achieve this. An AUM-weighted approach would potentially concentrate the index on more successful strategies and funds, as well as being influenced by the latest fashionable money flows. On the other hand, if the index is intended to capture the reality of what is happening in the global hedge fund universe, it should reflect these flows.
There is no ideal answer to this question but in practice equal weighting is more commonly used.2 The final criterion for an index is investibility. The Standard & Poors Hedge Fund Indices were created from the beginning as investable. Most other index providers initially provided non-investable indices and then added investable indices later on. The main reason for offering both is the conflict between investibility and some of the criteria above, chiefly that the index be representative. An index can be relatively broadly based and achieve high representativeness, but be non-investable. Or it can be investable but much more narrowly based and therefore potentially less representative. Index providers rely heavily on statistical tests to maximise the efficiency of this representative/investable trade off (see below).
Construction of an investable index
Investable index providers, starting from a database of hedge fund data, typically go through the following steps in constructing their index:
Figure 1: VAN Hedge Fund Indices:
Number of Constituents
Van Database c.6,700 funds |
|
Van Global HF Index c.2,000 funds |
|
VI2 Investable Index 60 Funds |
Source : VAN Composite Investable Index
Construction Methodology July 2005
- Quantitative screening
- Qualitative screening (due diligence) – often using a third party
- Confirmation of data and valuations – often using a third party
- Filtering for various criteria: minimum investment capacity, performance history, scale etc.
1. Quantitative screening
The key challenge for an index provider is to find a number of funds that are open to new investment and satisfy the other criteria (see below) but are representative of the strategy that the index seeks to capture. In practice this means using a very small number of funds relative to the whole database of hedge funds, or even relative to a broad non- investable index (see figure 1).
The first step is classifying the funds by strategy. It is not enough to take the fund managers’ self reported description as given. Instead index providers use correlation analysis and/or cluster analysis to identify funds that clearly behave in a group with other funds of a given strategy. Funds that appear to be too far from the group are excluded. This of course excludes any fund that achieves good performance but in an ‘undisciplined’ way. It also assumes that all strategies do fall into statistically neat groups. For conventional and well established strategies such as convertible arbitrage or merger arbitrage this is a reasonable assumption and funds employing these strategies do show a low level of strategies may not easily be classified and so may be excluded.
Assuming that a sub-universe of strategy-defined funds has been identified, the next investable and fulfil the other index criteria. The question is how many funds are needed for the index accurately to capture the performance of the overall strategy? This is simulations (Monte Carlo analysis). These simulations show that some strategies can be pretty well captured by only 10-15 funds but for others it may need to be over 20. For an aggregate index over all strategies (capturing the whole hedge fund universe) a total of 50-60 funds is normally sufficient to give very good correlation.
Clearly the more broad the index, the easier it is to find enough funds that meet the investable criteria to make it statistically robust. It is relatively straightforward to find enough funds to construct broad style indices (eg market neutral or absolute return). Finding enough for single strategy indices is much more difficult so although there are various sub-indices for detailed strategies they are not always intended for stand alone use.
2. Qualitative screening (due diligence)
Unlike say the S&P 500 index, the components of a hedge fund index require some form of qualitative assessment. All index providers include a due diligence stage in selection which would include auditing the fund’s accounts, checking the educational and other credentials of the employees, verifying the style consistency and perhaps taking up investor references. This process is often conducted by a third party, since expertise in index provision is quite separate from expertise in fund due diligence 3.
In this respect the index provider is straying into the territory of fund of hedge funds (FoF) managers. Many FoF managers would argue that much of their expertise lies in the – inevitably subjective – assessment of individual fund manager quality, integrity and business competence.
Investable Hedge Index Comparison |
||||||||||||||
nc |
nc |
Performance Measures |
Index Construction |
Constituent ManagerRequirements |
||||||||||
nc |
Historical |
CAR |
Tracking Error |
Number of |
Investable Index Weighting- |
Rebalancing |
Liquidity |
Pricing |
Separately |
AUM / Track Record |
||||
Van |
VI2 – Van Composite |
nc |
nc |
nc |
nc
|
11.80% |
0.41% |
45 |
Proportional to Strategy Weights in |
Semi-Annually |
Monthly + 30 Days Notice |
Monthly |
No |
520M / 1Year |
CSFB/ Tremont |
INVX |
HF Asset Class |
CSFB Hedge Fund Index |
August, 2003 |
Yes |
7.02% |
-3.91% |
60 |
AW/AW |
Semi-Annually |
Weekly / Monthly / Quarterly |
Weekly estimates |
No |
550M / 1 Year |
nc |
nc |
nc |
nc |
nc |
nc |
nc |
nc |
nc |
HFRX Global – |
ncnncc |
Weekly/Monthly |
nc |
nc |
nc |
MSCI |
MSCI Hedge Invest |
HF Asset Class |
MSCI Hedge Fund Index |
July, 2003 |
Yes |
4.52% |
-4.75% |
120 |
Adj Mean AW/EW |
Quarterly |
Weekly |
Daily estimates |
Yes |
Lyxor Platform Rules |
nc |
nc |
nc |
nc |
nc |
nc |
nc |
nc |
nc |
nc |
nc |
At least Quarterly |
nc |
nc |
ncnc |
FTSE |
FTSE Hedge |
HF Asset Class |
N/A |
July, 2004 |
Yes, |
6.36% |
N/A |
40 |
CW/CW |
Annually |
Weekly |
Daily GAV/NAV |
Yes |
SSOM / 2 Years |
Dow Jones |
nc |
nc |
nc |
ncnc |
Figure 2 Investable Index Comparison
Source : Index provider websites and methodology documents; tracking error calculations by Hedgequest
Screening out the clearly inept managers is a key way to ensure superior returns. FoF managers also claim to add other skills – identifying interesting new investment strategies, asset allocation among different styles and negotiating access to otherwise closed funds. Whether FoFs are in direct competition with investable indices is partly a matter of how well one judges FoFs to achieve these other goals.
3. Transparency: confirmation of pricing and valuation.
Any fund willing to be in the investable index must reach a high level of transparency and reliability in its data provision. Normally index provides want daily data and they may use additional third party specialists to verify the pricing of investments in illiquid or unusual securities. The pricing of illiquid investments can involve a significant degree of judgement and is known to provide hedge funds with some scope for smoothing their returns, so this step adds value to the index user who might not otherwise be in a position to verify pricing.
4. Filtering for various criteria.
The index providers have various criteria for fund inclusion, including the history and size of the fund. These are shown in figure 2 for some of the main investable index providers. Figure 2 shows that the index methods are not very different from one another,although the actual performance varies quite a lot. It would be wrong to conclude too much from these figures without analysing the riskiness of the indices.
Additional descriptive information on the indices is provided in figure 3.
Figure 3 Additional Information on Some Major Investable Index Providers
Broad HF Indices |
|
I d |
k l N mb f F d i |
fo |
95/J |
99/J |
|
HFR |
94/J |
MSCI |
02/J |
S&P |
02/J |
FTSE |
98/J |
D w J |
03/J |
Source: Index provider websites and methodology documents.
Click here for the full Hedgequest report on Investable Hedge Fund Indices
References:
Barry, R. (2003), ‘Hedge funds: a walk through the graveyard’, Working Paper, Ross Barry Macquarie Applied Finance Centre.
Fung, W. and D.A. Hsieh (2000),
‘The risk in hedge fund strategies: theory and evidence from trend followers’, Review of Financial Studies, 14 (2), 313-341
Lhabitant, F.S. (2004) Hedge Funds: Quantitative Insights, John Wiley & Sons, Chichester
Footnotes:
1. The fact that new funds can choose when to join a database also introduces a type of bias, especially if it comes with data on the fund’s history – the so-called instant history or ‘backfill’ bias. Index providers typically do not add the history of a new fund to their index but the database from which the index draws may show annualised return bias of around 1% according to some academic work (Fung and Hsieh, 2000 and Barry, 2003).
2. MSCI uses an asset weighting approach based on the median assets invested in a particular strategy, rather than the aggregate, combined with a six month moving average to increase stability.
3. Standard & Poor’s for example uses DPM
(Derivatives Portfolio Management), a firm established in 1993.
Author – Simon Taylor is a Senior Research Associate of the Judge Business School, Cambridge University. Prior to this he spend fourteen years in the investment banking world, as an equities analyst at BZW and Citigroup and then as Deputy Head of European Equities Research at JPMorgan. He was previously a Fellow in economics at St. Catharine’s College, Cambridge and an Overseas Development Institute Fellow in the Central Bank of Lesotho
Please click here to download the full hedgequest report The Global Reach of Investable Hedge Fund Indices