Selection Bias

Selection bias refers to the distortion of hedge fund index time series data because of possible selection criteria from database providers. h e distortion can occur because not all hedge funds are considered while calculating index values. Index sponsors make their database choices in the following ways:
  1. Hedge funds that fulfill the selection criteria are chosen because of their outstanding performance and hence their search for new investors and
  2. there is a relatively large number of hedge funds that have ceased to accept investment funds, and therefore refrain from reporting performance to any database.
Hence, it is possible that the performance of hedge fund indices is too low because of selection bias.

However, the empirical verii cation of net selection bias can be problematic. Fung and Hsieh note that selection bias may be indicated by the number and ident ity of hedge funds in various databases. Liang quantifies the overlapping of TASS and HFR databases, concerning existing funds with 41% and liquidated funds with 32%.

Lhabitant investigates four of the largest hedge fund databases (HFR, CS/TASS, CISDM, and MSCI) and finds that only 3% of individual hedge funds report to all four databases and only 10% report to three. This may mean that a large number of single hedge funds can only be found in one or two databases. Owing to differing construction methods, selection criteria, and data basis, the world of hedge fund indices is actually extremely heterogeneous.

Heidorn et al investigate the time series of six different index providers for the period January 1998–April 2005. They observe differences among the individual index families of up to 18.06% in yearly performance, 12.04% in volatility, and 8.5% in correlation between indices.

Regarding the selection criteria of hedge fund indices, Heidorn et al. (2006) note that 47.6% of all providers demand an average minimum fund volume of U.S. $26 million from single funds. Additionally, about 80% of the providers who demand minimum fund volume also expect a respective minimum track record from single funds.

Among the index providers, 38.1% require a minimum track record of 1.3 years. Only three providers who demand minimum requirements for volume and track record also include in their index, calculation funds that have reached capacity and have closed to new money. Out of all benchmarks, 61.9% included closed funds for index calculation.

However, the share of closed funds is relatively small, at about 10%. Several dif erent approaches exist to relate single funds to the respective strategy indices. In half the indices, the individual single funds determine the index, and thus the strategy under which they are classified.

The amount of selection criteria demanded from the databases is negatively correlated with the purity of hedge fund indices. Hence, a huge selection criteria catalog is counterproductive for the representativity of hedge fund indices.

Selection Bias
Selection Bias