Time-Varying Fund Benchmarks

Article MQL5 code base

Summary

The article reviews how benchmark choice affects estimates of active mutual fund performance. It argues that factor benchmarks and investable index benchmarks are not necessarily equivalent: omitted common risks can make factor-model alpha misleading, while a mismatched index benchmark can increase residual variation and distort alpha estimates. To address changing fund styles, it describes a regime-switching model that selects among 17 US equity indexes by minimizing fund alpha variance. An expectation-maximization procedure estimates the model and the changing benchmark probabilities; adding a risk-free asset is proposed as a way to represent cash holdings.

Using historical returns for active US equity funds, the study compares these estimated benchmarks with standard and self-designated benchmarks. It reports that time-varying benchmarks better capture fund risk exposures, identify a larger share of statistically significant alphas with smaller average magnitudes, and produce stronger out-of-sample alpha persistence. The evidence is specific to the historical US fund sample and depends on ex post benchmark selection, which may limit how the results translate into an investor's prospective benchmark choice. The article does not establish that these estimates predict future net investment returns.

Key ideas

Benchmark mismatch can increase the variance of estimated fund alpha.
A regime-switching model selects among passive indexes to track changing fund styles.
An expectation-maximization procedure estimates benchmark probabilities, with a risk-free asset available to represent cash.
The historical analysis reports improved risk-exposure fit and stronger out-of-sample alpha persistence from time-varying benchmarks.
Because benchmarks are identified ex post, the estimates may not be directly usable as prospective standards.

Time-Varying Index Benchmarks for Mutual Fund Performance Evaluation

Summary

Key ideas

Tags