Ask most investors how diversified they are, and they will answer with a number. Twenty stocks. Thirty. A hundred ETFs across six accounts. The implicit theory is straightforward: more positions equals less risk. Spread the bets, dilute the outcome of any one of them, and you are protected.
This is wrong in a specific and consequential way. Diversification is not a property of how many things you own. It is a property of how those things move together. A portfolio of thirty stocks that all rise and fall in sync is not thirty bets. It is one bet, repeated thirty times, with thirty separate transaction costs.
What Modern Portfolio Theory Actually Said
Harry Markowitz's 1952 paper is widely remembered as the case for diversification. It was actually something narrower and more useful: a case for diversification across uncorrelated return streams. The mathematics is unforgiving on this point. The risk of a portfolio is not the average risk of its components. It is the average risk plus an interaction term — the covariance between every pair of holdings.
When pairwise correlations are low, that interaction term works in your favor. The volatility of the basket comes out lower than the volatility of the individual pieces. When pairwise correlations are high, the interaction term works against you. The basket inherits the volatility of its parts, and you have paid thirty commissions for the privilege of holding one position.
The Diversification That Disappears When You Need It
The most painful feature of correlation is that it is regime-dependent. The correlations you see during a stable period are not the correlations you get during a stress event. This is sometimes called "correlation breakdown" but the phrase understates what happens. Correlations do not break down. They snap upward toward one, exactly when the protective benefit of diversification was supposed to matter.
Diversification is a contract the market enforces in calm conditions and tears up under stress. A portfolio designed only for calm conditions is not a diversified portfolio. It is an unhedged one.
This is why a portfolio that looks well-spread across sectors, factors, or geographies can still suffer the full force of a market drawdown. The thirty positions were never thirty independent bets. They were thirty expressions of the same underlying exposure — risk-on global equity beta — and when that exposure went into drawdown, all thirty went with it.
What Real Diversification Looks Like
True diversification requires holdings whose return drivers differ. Not their ticker symbols, not their sector labels, not their factor exposures on paper — the actual economic forces that produce their returns. Long-duration government bonds are diversifying from equities not because they are in a different asset class but because they tend to rally when growth disappoints and equities sell off. Gold can diversify because its return driver is partly the inverse of real yields. Trend-following can diversify because its return driver is the persistence of price movement itself, which is partially independent of any single asset's outcome.
The test is not "do these belong to different buckets on the allocation pie chart." The test is "do these have meaningfully different return drivers, and is that difference stable across regimes." Portfolios that pass this test tend to look surprisingly concentrated by ticker count — sometimes a dozen positions across genuinely orthogonal exposures — and produce smoother return paths than the thirty-stock book they replaced.
Why This Is Hard to See
Investors anchor on what they can see directly. They can see ticker count. They can see sector pie charts. They cannot, without effort, see the covariance matrix that governs how their portfolio actually behaves. So they end up optimizing for the visible variable and being surprised by the invisible one. The portfolio looks diversified on the holdings screen and feels diversified in the brokerage app. It only fails to be diversified when the market does the one thing the holdings screen cannot show: move all positions together.
If you cannot describe the covariance structure of your portfolio, you cannot describe its risk. Ticker count is a comforting proxy for diversification. It is not diversification.
The implication is not that ticker count is bad. The implication is that ticker count is the wrong question. The right question is what economic exposures the portfolio actually represents, and whether those exposures are independent enough — across calm periods and stress periods — to produce the smoother risk-adjusted returns that diversification is supposed to deliver. That is a question the math can answer. It is also the question most investors never ask, because the answer requires giving up the comforting illusion that thirty tickers is the same as thirty bets.
This piece discusses portfolio diversification and covariance for educational purposes. Referenced research findings describe historical outcomes and do not guarantee future results. This does not constitute investment advice. This does not create an advisory relationship.