A new research paper from ISS-STOXX and Columbia University's Andrew Ang arrives with a deceptively simple premise: diversification works, but most factor investors are not applying it far enough. In a July 2026 paper1 published through SSRN, Anthony Renshaw, Yurong Gu, Matt Candar, and Ang construct a multifactor equity portfolio that diversifies at three distinct levels: within factors, across factors, and across time. The result is an information ratio of 1.00 over a 21-year backtest against the STOXX US Universal Index, outperforming every individual sleeve the portfolio contains.
The Problem With Most Factor Portfolios
The authors open with a precise indictment: "Diversification is the only free lunch in investing, but most portfolios are less diversified than they appear to be." A portfolio holding hundreds of securities, they note, is often dominated by a single factor. Single-factor products remain the norm. The paper's contribution is to show how much that costs.
Splitting Value in Two
The most structurally novel move is dividing value into two separate sleeves. Cyclical value captures traditional metrics: book-to-price and earnings yield. Enhanced value captures something different: economic profit, intangible capital, and free cash flow. The distinction is not cosmetic. Over the full backtest, enhanced value earns an information ratio of 0.94; cyclical value finishes at -0.20. Their active return correlation is just 0.13. During the growth-led 2017-2020 market, enhanced value returned +85.71% cumulatively, keeping pace with the benchmark at +82.44%, while cyclical value trailed at +60.19%. The authors conclude that "neither sleeve dominates in all environments; holding both provides more complete exposure to the value premium."
Momentum Beyond Price
Momentum is similarly extended. The composite adds idiosyncratic momentum, analyst revisions, economic profit momentum, low short interest, and hedge fund holdings alongside standard price momentum. The result: average active exposure to the systematic momentum risk factor falls from 20.1% (price momentum alone) to 13.0%, without sacrificing alpha generation. As the authors put it, the composite "reduces reliance on any one source of momentum."
The Correlation Structure That Makes It Work
The correlation of monthly active returns across the four factor sleeves is strikingly low, ranging from -0.21 between quality and cyclical value to just 0.17 between enhanced value and quality. That structure is load-bearing. "The portfolio does not require every sleeve to work at the same time, only that the sleeves should not all fail together." The multifactor information ratio drops below zero only twice across 21 years.
Timing That Tilts But Cannot Concentrate
The third layer applies a bounded timing overlay across three models: a market similarity model weighting historical periods by their resemblance to current conditions, a factor momentum model tracking trailing 18-month information ratios, and a factor-of-factors model monitoring drift in the portfolio's own active exposures. Each model can at most double or halve a factor's weight. The three outputs are then averaged. This averaging is itself a diversification: "a model that predicts incorrectly is diluted by the other two, exactly as a noisy signal is diluted within a factor score and a struggling factor within the composite." Over the last three years, the timing overlay lifts the information ratio from 1.83 to 1.92. Over 21 years, the gain is negligible. The constraint is explicit: "timing can tilt the portfolio but cannot concentrate it."
The Factor You Would Have Cut Is the One You Need
The paper's deepest finding is structural. Because the most dispensable factor changes from period to period, "a portfolio holding all four sleeves never has to make the call and never has to be right about it." The discipline of staying diversified is not passive. It is the active decision that carries the most long-run value.
5 Key Takeaways for Advisors and Investors
- The multifactor portfolio earns an information ratio of 1.92 over the last three years, versus 0.65 to 0.94 for individual factor sleeves. Advisors can use that spread to benchmark whether a single-factor product justifies its concentration.
- Cyclical value and enhanced value carry an active return correlation of 0.13. A product blending them into one "value" score is exposing investors to a very different risk profile than one that separates them.
- Removing momentum from the multifactor portfolio over the last three years reduces the information ratio from 1.92 to 1.16. That gap is a quantifiable cost of factor concentration advisors can cite directly.
- The effective number of factors never drops below 2.6, even under the most extreme timing configuration. Advisors should ask any multifactor manager how concentrated factor weights can become under stress.
- Low volatility and size are deliberately excluded here because their beta tilts would consume the 1% tracking error budget through market exposure, not factor premia. Advisors evaluating multifactor ETFs should verify how much of the active risk comes from genuine factor exposure versus beta drift.
Footnote:
1 Renshaw, Anthony A., Yurong Gu, Matt Candar, and Andrew Ang. "Diversification All the Way Down: Multifactor Investing Within Factors, Across Factors and Across Time." ISS-STOXX / Tau Balance, July 2026. SSRN, https://papers.ssrn.com/sol3/papers.cfm?abstract_id=7156278.
