TY - JOUR T1 - Hedge Fund Return–Based Style Estimation:<br/> <em>A Review of Comparison Hedge Fund Indices</em> JF - The Journal of Alternative Investments SP - 24 LP - 53 DO - 10.3905/jai.2012.15.2.024 VL - 15 IS - 2 AU - Thomas Schneeweis AU - Hossein Kazemi AU - Edward Szado Y1 - 2012/09/30 UR - https://pm-research.com/content/15/2/24.abstract N2 - The data dependency of empirical financial research is of common concern to both academics and practitioners. This is especially true for hedge funds because no single, commonly accepted database exists and because many of the databases may hold different sets of reporting managers. Each database uses current reporting managers as the basis for the construction of hedge fund indices, and these index returns reflect the characteristics of the funds reporting to the relevant database. Unlike historical returns derived from current databases, however, historical returns from most major hedge fund indices do not contain backfill or survivor bias. At the same time, performance characteristics may differ between indices because each index is constructed based on a different set of rules (e.g., equal weighted, asset weighted, and so on). In this article, the authors conduct a series of empirical tests similar to those previously conducted in academic studies. The authors use only those hedge fund indices that reflect the average returns of the entire set of reporting managers; that is, the indices representing overall industry returns. Results indicate that return-based style analyses, often used as a basis for hedge fund analysis, are impacted both by the period of analysis as well as the hedge fund index used. Moreover, results indicate that the addition of variables beyond those designed to capture underlying equity, interest rate, and credit risk have little impact on the explanatory power of these hedge fund universe indices beyond a very low level of statistical significance.TOPICS: Real assets/alternative investments/private equity, mutual funds/passive investing/indexing, statistical methods, performance measurement ER -