ABSTRACT
We examine hedge fund (HF) index construction methodologies, by describing and analysing case studies from two well-known database vendors and evaluating them using numerical examples on the same dataset. Despite the fact that they follow a similar due diligence process, there are great differences in the index engineering practices arising from different quantitative techniques, even for indices in the same HF category. However, those quantitative techniques provide similar results. The differences are rather due to the use of different HF universes and different inclusion criteria. This article is the first to use actual numerical case studies to illustrate and compare how HF index engineering works. Having read it, the reader will have a good understanding of how HF indices are formed.
Disclosure statement
No potential conflict of interest was reported by the authors.
Notes
1 ‘Correlation distribution’: if we have a group of HFs, then we have a number of pair correlations, i.e. the correlation of each hedge fund’s returns with the returns of each of the other HFs. Each HF has its own distribution of pair correlations, with a mean, standard deviation etc.
2 The algorithms and processes we followed for database cleaning and merging are available on request.
3 We compute the standard deviation of the pair correlations (correlation of each fund or strategy with each of the other funds or strategies) within the group.
4 ‘Distribution of their standard deviation’: If we have a group of funds within a strategy, then each member of this group has its own standard deviation of returns. Hence, we have many standard deviations in this group (one value for each fund). Thus, we can plot the overall distribution (of all fund-specific standard deviation values) represented by a mean, standard deviation etc. for this group. The lower the standard deviation of the overall distribution for the group (of funds or strategies), the better it is, because this group is more homogenous.