Diversification is portfolio theory's nearest equivalent to a free lunch. Hold enough uncorrelated positions and losses in one area can be partly offset by stability elsewhere. Decades of research support the logic in ordinary markets. The catch is the phrase "ordinary markets."
When I began working on what became Bloomarq, historical portfolio data repeatedly showed how badly standard diversification performed when investors needed it most. A portfolio that looked sound on an ordinary Tuesday in 2019 acted very differently during the March 2020 selloff. The strategy had not necessarily failed. The correlation structure supporting it had changed during the stress event.
The Correlation Instability Issue
Correlation measures how closely two securities tend to move over a historical period. Most portfolio tools calculate it across several years, blending calm and volatile markets. That produces a "typical" correlation estimate that can understate movement together at the tail.
In broad market stress, investors often become risk-averse across the board. They do not simply sell expensive technology holdings while retaining consumer staples. Margin calls, redemption pressure, and the desire to reduce uncertainty push them to cut risk broadly. Selling pressure then reaches positions that usually behave independently. A consumer staples holding and a semiconductor position with a 0.18 correlation over the prior three years might show 0.75 correlation during a sharp two-week selloff.
That is correlation instability: realized stress-period correlation differs from the historical estimate. Since stress periods are rare within any sensible window, they make up little of the data behind the estimate. Normal-period behavior therefore dominates the result.
Why 15 Holdings Do Not Equal Diversification
Investors often mistake a large list of names for genuine diversification. Fifteen technology companies spread across hardware, software, and services are not meaningfully diversified during a broad technology selloff. The shared sector factor overwhelms the individual differences between those businesses.
Factor exposure is a better diagnostic than position count. How much portfolio variance comes from one sector, one interest-rate regime, or one growth-versus-value orientation? Two portfolios may contain the same number of positions yet have very different effective concentration because their holdings load differently on shared factors.
Static portfolio views can mislead active investors here. Building each position on its own fundamentals may leave the cumulative factor exposure across positions untracked. You added strong companies, not technology sector beta. The market makes no such distinction during a technology selloff.
What Stress Correlation Looks Like in Practice
Consider a plausible early 2024 portfolio spread across businesses an active investor considers distinct: financial services names chosen for low valuations, several quality industrials, multiple technology platforms, and a couple of healthcare positions. Four sectors, and on paper, reasonable diversification.
A rate shock sends financial and technology names lower, for different reasons but at the same time. Industrials also fall as the same macro news raises demand concerns. Healthcare alone remains relatively stable, yet accounts for 15 percent of the portfolio. Effective diversification during the event was about 15 percent, not the 25-position spread that looked diversified in normal markets.
There is nothing unusual about this portfolio math. It shows what happens when macro forces dominate cross-asset correlations during stress. The diversification was genuine in normal conditions, but unavailable when the investor needed it most.
Structural and Cyclical Diversification
It helps to separate cyclical from structural diversification. Cyclical diversification means owning sectors that tend to lead in different economic cycles: financials and energy in some environments, consumer staples and healthcare in others. It diversifies across regimes but cannot shield a portfolio from a broad risk-off event affecting every equity sector at once.
Structural diversification adds assets that remain genuinely weakly correlated with equities during stress. That is difficult in a pure equity portfolio. Cash is the most accessible option: it is uncorrelated with equity stress by definition, and its "return" in a sharp selloff is the loss avoided, plus the option to deploy it after the drawdown. Shorter-duration Treasury instruments have historically offered some structural diversification within equity portfolios, although 2022 tested that relationship.
We are not saying competent equity investors must own bonds or alternatives. "Diversified across equities" and "protected against broad equity drawdowns" mean different things. Confusing them leaves investors surprised when their 20-position portfolios fall together during a stress event they expected diversification to soften.
Building a Portfolio Around Actual Risk
Portfolio construction therefore needs stress-period correlation assumptions alongside normal-period assumptions. Using only average historical correlations shows the risk picture present most of the time, not the one that appears when markets reprice quickly.
Test the portfolio against historical stress episodes, especially periods when sectors moved together in ways normal correlations would miss. That gives a more candid view of downside exposure. How did this sector mix perform in the March 2020 drawdown, the 2022 rate shock, or the 2015 to 2016 commodity selloff? Each event had its own correlation structure and affected differently built portfolios in different ways.
The aim is not a portfolio that cannot lose money. Equity investing cannot provide that. The aim is to identify when this portfolio could lose substantial value, then decide whether to hold through those conditions or change the construction. Diversification works, but it cannot replace knowing the portfolio's actual stress exposure.
