Evaluating Metrics for Fund Selection – II

What works, what doesn’t?

March 2026. Reading Time: 10 Minutes. Author: Nicolas Rabener.

SUMMARY

  • Most common fund selection metrics are ineffective
  • Beating the benchmark is easier in some asset classes than others
  • Fees matter

INTRODUCTION

In June 2025, we published our first article evaluating popular fund selection metrics used by data services such as Morningstar, Lipper, and S&P Global for ranking funds (read Evaluating Metrics for Fund Selection). Our analysis concluded that most of these metrics lack predictive power – for example, identifying the top-performing funds in the past does not guarantee they will continue to outperform in the future.

That said, our initial study had limitations. It focused exclusively on U.S. equity mutual funds and ETFs between 2015 and 2025, ranking funds in 2019 based on the prior five years of data and tracking their subsequent five-year performance.

In this research article, we expand the scope by incorporating multiple in-sample and out-of-sample tests and extending the analysis to additional asset classes.

METHODOLOGY

We consider all mutual funds and ETFs trading in the U.S. market between 2000 and 2025, including both currently active and liquidated funds, to avoid survivorship bias. From an initial universe of roughly 50,000 funds, we exclude multiple share classes and funds employing leveraged, short, volatility, or option-based strategies, resulting in a final sample of approximately 11,000 funds.

Our analysis uses a universe of 45 benchmark indices spanning equities, fixed income, multi-asset, and commodities. Each fund is assigned a benchmark using our standard methodology, which evaluates the ratio of tracking error to correlation. At the end of each financial year, we rank funds within each benchmark category by percentiles across several metrics: excess return, excess Sharpe ratio, information ratio, consistent outperformance, factor alpha, R² relative to the benchmark, and total expense ratio.

We calculate these metrics using 1-year, 3-year, and 5-year lookback periods (in-sample) and then track the subsequent performance of funds in each percentile (out-of