PortfolioFuture Methodology

Research Process

PortfolioFuture performs fund substitution, investable return decomposition, and relative ranking research across ETFs and mutual funds. This section describes the methodology behind our fund comparisons, substitution rankings, replication analysis, and out-of-sample empirical testing.

For fund substitution research, we compare investments using their shared realized return history rather than relying primarily on fund categories, benchmarks, or stated investment objectives. We evaluate how closely funds have behaved, the strength of historical evidence of relative outperformance, the active risk involved in substitution, and relevant implementation costs.

Across our research, PortfolioFuture measures the investment experience delivered by actual investable products. Our results are intended to reflect how funds have behaved in the real world and how effectively one investment could have substituted for, replicated, or differentiated itself from another. Historical relationships are evaluated across available overlapping periods and, where applicable, tested out of sample.

We research both ETFs and mutual funds and evaluate relationships directly across products when sufficient comparable return history is available.

  1. A luminous wave illustrating behavioral similarity between funds
    · Vince (Qijun) Chen, CFA

    Finding Better Fund Alternatives

    A pairwise framework for comparing a fund with potential substitutes using their shared return history. Return Similarity measures how closely two funds have behaved. Tracking Error measures the distance between them. Probabilistic Information Ratio (PIR) measures the strength of historical relative-performance evidence. BAS combines Similarity and PIR into a single pairwise ranking.

    Return Similarity

    Measures pairwise comparability from realized weekly returns, combining return co-movement and volatility scale.

    Probabilistic Information Ratio (P(IR>0))

    Measures the strength of historical evidence that a candidate’s mean active return has been positive.

    Out-of-Sample Results

    Higher historical BAS ranks were associated with greater subsequent Return Similarity, lower Tracking Error, and higher subsequent outperformance rates.

    Read methodology → Read paper on SSRN ↗
  2. Colored waves illustrating a fund return decomposed into exposures
    · Vince (Qijun) Chen, CFA

    Investable Return Decomposition and Residual Persistence

    A returns-based framework for separating replicable fund behavior from the return left after replication. A constrained portfolio of liquid ETFs is fitted to each fund’s return history. The methodology measures replication quality, changes in the relationship over time, and the subsequent behavior of the residual.

    Investable Benchmark

    Builds a fully invested benchmark from 62 liquid ETFs used in the empirical study.

    Replication Persistence

    Historical replication quality was associated with next-year replication quality; forward rank correlation in Tracking Error: 0.925.

    Residual Persistence

    Historical residual Information Ratio ranked subsequent residual returns; annual top-minus-bottom decile spread: +1.22%.

    Read methodology → Read paper on SSRN ↗
  3. Luminous multi-stream wave lines illustrating fund ranking and performance persistence
    · Vince (Qijun) Chen, CFA

    Ranking and Performance Persistence Among Similar Funds

    A framework for turning pairwise fund comparisons into rankings among funds with similar return behavior. The research aggregates pairwise BAS orderings across each fund’s comparable peers and measures how those rankings behave over subsequent periods.

    Cross-Sectional BAS

    Aggregates pairwise BAS-implied directional orderings across Return-Similarity peer sets without introducing new performance parameters.

    Out-of-Sample Persistence

    Across 103 quarterly cross-sections from 2000 to 2025, current peer rankings were associated with subsequent 52-week rankings across all six lookbacks (raw ICs 0.12 to 0.16).

    Residual Peer Persistence

    After controlling for Market PIR, average residual rank IC rises from 0.074 at S50 to 0.152 at S99 as peer sets narrow.

    Read methodology → Read paper on SSRN ↗