Diversification: Myth vs. Reality
Five line items on a broker statement can still be one bet, and it shows up exactly when protection is needed.
We think most retail portfolios are far less diversified than their owners believe, and the broker statement is what hides it. At the end of 2024 the ten largest companies in the S&P 500 made up 37.3% of the index by market value, the highest share on record, and by early 2025 the top ten were close to 40%, per Pensions & Investments, January 2026, using S&P Dow Jones Indices data. By the end of 2025 the largest ten had reached about 40% of the index (Lord Abbett, 2026). A European investor who holds an MSCI World ETF, a US large-cap fund, and a small technology sleeve sees three line items. We see one bet, written down three times.
Same names, different stickers
Diversification, as most retail investors practise it, is mostly a labelling exercise. The portfolio is a world equity index, an international fund, a bond ETF, perhaps a gold position. That looks like four separate bets. It usually is not, because the holdings overlap and they move together in the moments protection is needed.
An MSCI World tracker is now over 70% US equities by weight, and the same handful of large US technology companies sit near the top of the world index, the US index, and the technology sleeve at once. The label says "global, plus a tilt". The exposure is closer to one bet on a small group of US technology business models, held three times. We are not telling you to avoid those companies; our point is narrower. The label is not the protection it looks like, and a passive wrapper does not remove concentration risk, it only hides it inside an index.
Concentration shows up across asset classes too
The standard answer to equity concentration is to add bonds, on the reasoning that shares and bonds usually move in opposite directions, so when shares fall bonds rise and soften the loss. That relationship is measured by correlation, a number between minus one and plus one: minus one means they move in exactly opposite directions, zero means no link, plus one means they move together. For most of 2000 to 2021 the stock-bond correlation was negative, so bonds did cushion equity losses.
In 2022 that protection failed when it was needed. The S&P 500 returned minus 18.1% on the year. The Bloomberg US Aggregate Bond Index, a broad measure of US investment-grade bonds, returned minus 13.0%, its worst calendar year since the index began in 1976, per CNBC, January 2023. It was the first calendar year since 1977 that both US shares and US bonds fell (Vanguard, 2023). The reason is not mysterious. AQR, in The Journal of Portfolio Management, Q1 2023 shows the stock-bond correlation does not depend on the level of inflation but on whether growth or inflation is the bigger source of uncertainty; their model explains about 70% of the long-run swings in that correlation. When inflation becomes the dominant worry, as in 2022, shares and bonds price the same risk and fall together, and the correlation turned firmly positive. We do not count an asset that only protects you in one inflation regime as a hedge, because the year you needed it was the year it did not work.
Geography helps less than the brochure says
The next textbook fix is to add other countries: spread the equity across Germany, Japan, the UK, and the US, so a fall in one market is offset by the others. This works in calm periods and works far less in the moments it is supposed to matter, because in a global panic equity markets stop behaving like separate bets.
The 2012 paper The Death of Diversification Has Been Greatly Exaggerated, by Ilmanen and Kizer in The Journal of Portfolio Management, traced a portfolio spread across US shares, US bonds, international shares, emerging-market shares, and property. Through the 2008 financial crisis its equity beta, how much it moved for a given move in the stock market, rose from 0.65 to 0.95, meaning it fell almost as hard as a pure equity holding. The spread-out portfolio underperformed a plain mix of 60% US shares and 40% US bonds by about nine percentage points through the crisis. Adding country labels added almost nothing when global risk appetite collapsed; the positions all fell together.
What still diversifies, and what it means for you
The same paper found what did hold up: diversifying across return drivers rather than asset-class labels. A return driver, often called a factor, is the underlying reason an investment earns money, for example the equity risk premium (being paid to hold shares and bear their swings), the term premium (being paid to lend long), value, or momentum. Portfolios built across factors held lower correlations during crises than portfolios built across asset classes. Asset-class labels group instruments by what they are; factor labels group them by why they earn a return, and that grouping survives a regime change more often.
For a European retail investor without long-short factor funds, the practical version is narrower: hold return streams that earn money through genuinely different mechanisms, not just different tickers. Cash in a money-market fund earning the policy rate is one stream; a short-dated inflation-linked bond is another; each pays off for a different reason. Ten holdings that all pay off through the same equity risk premium are not ten bets; they are one bet held ten times. Our view is that your MSCI World ETF, your US fund, and your technology sleeve are closer to that than the broker statement suggests.
If you grouped your top ten holdings by the single reason each one earns a return, rather than by its asset-class label, how many genuinely independent bets would you actually own?
