A finding is only as good as the record it is measured against. This week’s research returns to that idea from several directions: a portfolio rule whose advantage depends on the market regime, a private asset class whose calm comes partly from how it is valued, a century of returns that looks different once the rest of the world is counted, and factor libraries that shrink sharply once they are tested honestly. For a long-term investor, the lesson is about method. Before trusting a number, ask what history produced it.
Equal weighting’s edge depends on the regime. Active Portfolio Management in Concentrated Equity Markets starts from a well-documented observation: the equal-weighted portfolio has historically delivered higher returns than the capitalization-weighted market across many markets and periods. Using stochastic portfolio theory, the authors show that this relative performance is regime dependent. Equal weighting has tended to lag during periods of rising concentration and high correlation, particularly bubbles. They model an allocation between the two portfolios and report hypothetical out-of-sample backtests from 1995 to 2024 with transaction costs deducted. The durable point is not the strategy. It is that “equal weight beats the market” is a statement about particular histories, and the structure of the market at any time changes what that statement means.
Smooth reported values can hide real volatility. A research summary on buyout performance with assets valued at market takes up a measurement problem at the heart of private equity. Buyout investments are not continuously marked to market. Their reported net asset values are periodic estimates, which makes the asset class appear less volatile and more diversifying than its underlying holdings may be. The research re-estimates performance with assets valued at market prices. The general lesson carries well beyond private markets: a low measured volatility can be a property of the valuation method rather than of the asset. This blog has written before on why a hedge is not always a hedge.
The US record is one outcome, not the only one. The CFA Institute Research Foundation released a section of Roger Ibbotson’s long-run returns study that reaches back before 1925 and out to global markets, London, Japan, China, commodities and earlier international markets. Its framing is direct: the United States was an unusually successful market, not the only possible long-run outcome. The section works through survivorship bias, success bias, and the limits of extrapolating from modern US history. For anyone reasoning about diversification from the historical data, this is the reminder that the data set most often quoted is also the one most shaped by success.
Most signals do not survive a fair test. Two papers look at factor research from the testing side. Cross-Market Alpha, revised this week, tests 168 short-horizon signals built for China’s retail-dominated equity market against large US stocks from 2002 to 2022, on top of 153 established US factors. Seventeen signals showed significant loadings at first; only three stayed significant under the full set of robustness checks. Separately, Propose, Don’t Judge studies AI agents that generate investment factors and argues that the agent may propose but a frozen statistical referee, scoring only on outcomes revealed after submission, must judge. In the authors’ tests, that separation admitted far fewer false factors, at the cost of a long wait before any factor was certified. Both papers make the same point: the process that finds a pattern cannot also be the process that grades it.
A glide path defined by risk, not by age alone. For readers closer to retirement, a revised paper on target-date fund design proposes building a glide path around an explicit return objective while limiting risk through a declining Conditional Value-at-Risk constraint, a measure of losses in the worst outcomes. The authors apply it to Chile’s 2025 pension reform across nine asset classes and a 40-year horizon, and find transition age and contribution density to be critical design inputs. This is risk-budgeting showing up in pension design: the allowance for downside is set first, and the allocation follows from it.
The throughline. Each item this week is about the frame behind a number: the market regime, the valuation method, the country that happened to win, the test that happened to be run. None of these findings tells an investor what to do next. Together they reinforce a principle this blog returns to often: durable construction starts from what can be measured honestly, including risk, and treats any single historical record as one draw from a wider range of outcomes.
Curious where your portfolio’s risk structure stands? The free ETF Portfolio IQ Score is one way to see.