Is the MSCI World Too American? The Real Numbers

Two ways to count a world index

A fund holding 1,250 companies across 23 countries sounds like the definition of diversification. Counting positions, though, is the least informative thing you can do with a portfolio, because a position of 0.01 % and a position of 5.7 % both count as one.

Counting by weight gives a different picture. In the largest MSCI World ETF, 72.3 % of the money sits in United States companies and 27.0 % in the ten largest positions. The single largest, NVIDIA, is 5.69 % of the whole fund — more than the entire weight of most member countries.

The comparison across index families is where this becomes useful. The ACWI, which adds emerging markets, comes to 64.2 % United States. World Small Cap, often bought as a diversifier, is still 61.4 % American. And Emerging Markets, the index usually described as the risky one, has 37.0 % in its top ten and 15.1 % in Taiwan Semiconductor alone — a concentration well beyond anything in the World.

The number that describes it best

There is a single measure that turns weights into an intuitive figure: the effective number of stocks. It asks how many equally weighted positions would produce the concentration a portfolio actually has.

For the MSCI World the answer is 88, from 1,250 holdings. The other 1,162 companies exist in the fund, but in terms of what moves the price, they are rounding. ACWI, with 2,245 holdings, comes to 107. World ex USA, with 738 holdings, comes to 230 — the least concentrated of the family precisely because the American mega caps are missing. World Small Cap, at 3,685 holdings, reaches 1,664, which is what genuine breadth looks like as a number.

So the fair statement is not that the MSCI World is undiversified. It is that its diversification is about a factor of fourteen weaker than its holdings count suggests, and that the gap comes almost entirely from one country’s largest companies.

What the concentration does in practice

Weights are a description. The question that matters is whether the index behaves like a world portfolio or like an American one, and daily returns answer it directly.

Over 524 shared trading days, the correlation between the world index and the S&P 500 is 0.978. The regression of one on the other gives a beta of 0.91 and an R² of 95.6 %: essentially all of the world index’s day-to-day movement is the US market, scaled slightly down. The correlation with developed markets outside the US is lower, at 0.859.

The returns say the same. Over that window the world index returned 49.1 %, the S&P 500 48.9 %, and developed markets outside the US 43.9 %. An investor who held the world index and an investor who held the S&P 500 had, for two years, the same experience with a difference of 0.2 percentage points.

That is the honest form of the concentration problem. It is not that the MSCI World will fall further than a US index in a crisis — it will fall slightly less. It is that buying it in the belief that you have bought the world buys you something that has been, in practice, an American index with a 28 % international tail.

How much a shock would actually cost

Mechanical arithmetic is worth doing here, because the intuition tends to overshoot in one direction and the reassurance in the other.

If the ten largest positions fell 20 % while everything else stood still, the index would lose 5.4 %. If the entire US block fell 20 %, it would lose 14.5 %. At a 30 % fall in the US block the index loses 21.7 %, and at 50 % it loses 36.1 %.

The first number is the useful one for the “NVIDIA risk” argument: even a brutal repricing of the ten biggest names costs the index single digits. The second is the useful one for the concentration argument proper: a US-specific problem is transmitted almost fully, because there is not enough of anything else to absorb it.

Sector weights point the same way. Information technology alone accounts for 29.9 % of the fund, ahead of financials at 16.7 % and industrials at 11.0 %. The concentration is not only geographic.

So is the risk real?

Three findings, each measurable and each pointing in a different direction.

The claim that the MSCI World is barely diversified is wrong in its usual form. It holds 1,250 companies, and its effective breadth of 88 positions is greater than that of a US index (48) and vastly greater than that of Emerging Markets (30). Among the widely held indices, the World sits in the middle, not at the concentrated end.

The claim that it is a world portfolio is wrong too, and more clearly so. With 72.3 % in one country, 95.6 % of its daily movement explained by that country’s index, and a two-year return within 0.2 points of it, “world” describes the holdings list rather than the behaviour.

And the claim that a handful of technology companies could sink it is the weakest of the three: a 20 % fall across the top ten costs 5.4 %. What could sink it is a country, not a company.

What none of this can tell you is where the US share is going, or whether 72.3 % is high by historical standards. Our holdings data begins in July 2026 and describes a state, not a trend. That limit is worth more than a guess: the number in this study is exact, current, and says nothing whatsoever about next year.