Skip to content
All library documents

Time-Varying Index Benchmarks for Mutual Fund Performance Evaluation

Article BigQuant
Read the original

Summary

The article examines how benchmark choice affects estimates of active mutual fund skill. It distinguishes investable index benchmarks from factor-based benchmarks, arguing that omitted common risks can bias factor-model alpha, while choosing the wrong index can increase the variance of benchmark-adjusted returns. The proposed method uses a regime-switching model to select among 17 passive US equity indexes over time, estimating transition probabilities with an expectation-maximization algorithm. It also considers beta-adjusted benchmarks and adding the risk-free rate to account for funds’ cash holdings.

The empirical study uses historical returns for active US equity funds and compares estimated time-varying benchmarks with conventional and self-designated alternatives. The article reports that the estimated benchmarks better capture fund risk exposures, improve identification of statistically significant alpha, and show greater out-of-sample alpha persistence; the identified significant alphas are more numerous but smaller on average. Results are based on a historical US sample and benchmark selection is ex post, so they may not translate directly into a benchmark chosen in advance. The analysis describes performance measurement, not a trading strategy or evidence that investors can earn persistent excess returns.

Key ideas

  • Factor-based and index-based benchmarks may diverge when common risks are omitted from the factor model.
  • A regime-switching model selects among passive indexes as a fund's style changes over time.
  • The model uses expectation-maximization estimation and can include a risk-free benchmark to represent cash.
  • The historical analysis finds better risk-exposure fit and stronger out-of-sample alpha persistence with time-varying benchmarks.
  • Ex post benchmark identification limits direct use as a prospective investment standard.

Tags

EquitiesStatisticsPortfolio constructionBacktesting

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.