MSCI World vs ACWI vs FTSE All-World: structure, concentration and combinations

Basics12 min read · updated 27 October 2026 · 15 sources

MSCI World vs ACWI vs FTSE All-World: almost every global equity ETF tracks one of these three indices. They differ in which countries and how many companies they include, but hardly at all in how heavily they bet on the US and a few large technology companies. Here we compare structure and concentration according to the latest factsheets, measure concentration with the Herfindahl index, look at Japan around 1989 and work through combinations and their rebalancing consequences. How much equity belongs in a portfolio in the first place, and what share of emerging markets suits you, is covered in the article global portfolio.

Structure: countries, size segments, number of holdings

All three indices weight by freely tradable market value (free float): a company worth twice as much gets twice the weight. Differences arise in which countries count as developed or emerging markets and how far down an index reaches in each country. MSCI's standard indices cover about 85% of the free float in each country, the extended IMI around 99% (MSCI GIMI Methodology 2024).

Sources: factsheets from MSCI (MSCI World Factsheet 2026; MSCI ACWI Factsheet 2026; MSCI ACWI IMI Factsheet 2026) and FTSE Russell (FTSE All-World Factsheet 2026). A more recent FTSE All-World factsheet was not available when we checked on October 26, 2026.

The MSCI World contains developed markets only. The MSCI ACWI adds emerging markets, and the MSCI ACWI IMI also adds small companies. The FTSE All-World follows the same idea as the ACWI but has almost twice as many holdings: 1,983 from developed and 2,271 from emerging markets, 1,248 of them from China alone, which together weigh only 3.17% (FTSE All-World Factsheet 2026). The country lists differ too: FTSE classifies South Korea and Poland as developed markets, and Greece since September 21, 2026 (FTSE Russell Country Classification 2026). MSCI still classifies South Korea as an emerging market and will only upgrade Greece with its review in May 2027 (MSCI Market Classification 2026). The MSCI indices are reviewed quarterly, the FTSE All-World semi-annually in March and September, with smaller adjustments in June and December (FTSE Russell GEIS Dates 2023).

What is inside: the US, technology and the ten largest holdings

Three things stand out. First, emerging markets lower the US weight by just under nine percentage points, which does little to change the dominance of a single country. Second, the weight of information technology actually rises with emerging markets: in the ACWI, the Taiwanese chipmaker TSMC adds another technology company to the ten largest holdings. Third, the small companies in the IMI only slightly dilute the top: 5,622 additional holdings lower the weight of the ten largest by just under three percentage points. Sectors are best compared within one index family: MSCI classifies by GICS, FTSE by ICB, and the two systems draw the lines between technology, communication and consumer sectors differently.

How large is the emerging market weight in the ACWI? The factsheets do not state it, but the ratio of the weights of the same holdings gives it away: Nvidia weighs 5.11% in the ACWI and 5.81% in the World, a ratio of 0.88. Developed markets therefore make up around 88% of the ACWI and emerging markets around 12%. The ratio of the US weights gives the same result. For the FTSE All-World, the factsheet states the market value of the FTSE Developed and FTSE Emerging sub-indices: at the end of February 2026, 87.95 of 98.31 trillion US dollars, and thus 89.5%, was in developed markets and 10.5% in emerging markets. That is less emerging markets than at MSCI, partly because FTSE counts South Korea as a developed market.

Measuring concentration: the Herfindahl index and the effective number of holdings

The number of holdings says little about diversification. A better measure is the Herfindahl-Hirschman index (HHI), the sum of the squared weights. Its reciprocal is the effective number: how many equally sized holdings would have the same concentration? The basics are in the article look-through; here we apply the formula to the factsheets.

The true number lies below these upper limits because the rest is not spread evenly. An analysis of holdings lists gave around 86 for an MSCI World ETF (look-through). FTSE Russell calculates a Diversification Factor for the FTSE All-World based on the same idea: it fell from around 500 at the start of the last decade to just over 100 in autumn 2025 (Yoshimoto 2025). For a portfolio, this means: a global ETF is still broadly diversified, but its risk depends noticeably on a handful of companies and on one country. You should be aware of this concentration risk.

Concentration risk in history: Japan around 1989

A single country dominating a global index is nothing new. In September 1987, the US made up 37% of FTSE's global index, about the same size as Japan (Yoshimoto 2025). At the peak of the Japanese bubble at the end of the 1980s, Japan accounted for 45% of the market value of the 22 countries analyzed by Dimson et al. (2013); by the end of 2012 it was down to 8%, with the US at 51%. Today Japan weighs 5.91% in the MSCI World (MSCI World Factsheet 2026), and the Nikkei 225 took until 2024 to regain its 1989 high (see home bias).

Two things follow from this. Weighting by market value always holds the most of a country when it is most expensive. Concentration is a result of past price gains. On its own, however, it does not predict that a crash will follow: according to Greenwood, Shleifer and You (2019), sharp price run-ups in entire industries increase the probability of a crash, but on average they do not predict poor subsequent returns. If you want to limit the dependence, you are making an active decision that may or may not pay off. How bubbles form and how to recognize them is described in the article speculative bubbles.

Combinations: World plus emerging markets, and regions weighted equally

Instead of a single ETF, you can combine building blocks. Always compare the figures of your funds and, above all, the tracking difference over several years, because it shows what you actually receive (see ETF basics). Three typical solutions with illustrative figures for the ongoing charges (TER):

• One ETF on the ACWI or All-World: market weights, no rebalancing between regions, one TER (0.20% in the example).
• World plus emerging markets in a ratio of 88 to 12: roughly replicates the ACWI. With TERs of 0.20% and 0.18%, the mix costs 0.1976%, hardly any difference. The weights drift apart, however, and every few years the market ratio itself shifts.
• Regions weighted equally, for example one third each in the US, other developed markets and emerging markets: much less US, but in exchange a large deviation from the world market and regular reallocation.

How large is the deviation? In the ACWI, 64.21% is in the US, around 23.82% in other developed markets and around 11.97% in emerging markets. One measure of the deviation is active share, half the sum of the absolute weight differences: ½ × (|33.33 − 64.21| + |33.33 − 23.82| + |33.33 − 11.97|) = 30.88%. Almost a third of the portfolio is therefore positioned differently from the world market.

In practice, this means: if you combine funds, set rules before the weights drift, for example a tolerance band of five percentage points, and direct new contributions preferably into the underweight region. That reduces taxes and costs. More on bands and calendar dates in the article rebalancing. Combinations of a global ETF and satellite positions with their own risk budget are covered in the article core-satellite.

For advanced readers: what the metrics hide

• Security or company: Alphabet appears with two share classes in the factsheets (2.23% and 1.76% in the World). If you add them together, the sum of squares at the top rises from 103.44 to 111.29, and the upper limit for the effective number falls from 92.90 to 86.59, almost exactly the value from the holdings lists. For concentration, what counts is companies, not ticker symbols.
• Small caps in the IMI: dividing the three largest weights in the IMI by those in the ACWI gives (4.58 + 4.26 + 3.15) / (5.11 + 4.76 + 3.52) = 89.5%. The 5,622 small companies therefore make up only around 10.5% of the IMI: many holdings, little weight.
• Concentration is not the same as risk: the HHI only counts weights. Ten holdings from the same industry that are highly correlated with each other diversify less than the effective number suggests. Risk-based measures such as risk contribution take volatilities and correlations into account.
• Country classification: indices assign a company to a country according to its headquarters and stock exchange listing. A US company earns part of its revenue abroad, a European company part of its revenue in the US. The US weight in the index is therefore a measure of the price and valuation risk of the US market, not of dependence on the US economy.
• Free float and weight caps: because only the free float counts, companies with large anchor shareholders weigh less. The standard indices have no weight caps for individual holdings; capped variants follow their own rules.

What the research shows

The strong concentration of indices follows from the distribution of stock returns: Bessembinder (2018) shows that, measured over their entire lifetime, a few stocks account for most of the wealth creation in the US market. A market-cap-weighted index automatically holds these few winners, and in ever greater measure. Its theoretical justification comes from Sharpe (1991): before costs, all investors together earn the market return, and so does the average of all those who deviate from market weights; after costs, this average is lower. Market weighting is therefore the only portfolio that everyone can hold at the same time, and the cheapest.

Critics doubt that it is also the most efficient. Arnott et al. (2005) found that indices weighted by sales, book value or dividends were more efficient in their tests than market-cap-weighted indices, arguing that overvalued stocks are automatically overweight under market weighting. The other side points out that such weightings tilt toward value and smaller companies and thus collect factor premiums (see Fama-French factors). Long-run country data such as those of Dimson et al. (2013) show above all how strongly the weights shift over decades: no country stayed at the top permanently, and no current weight is a law of nature.

Limitations and common mistakes

• Looking only at the number of holdings: 4,254 holdings in the FTSE All-World sound broader than 2,414 in the ACWI. But the differences lie almost entirely in small weights; the US and top 10 weights are similar.
• Mixing data from different reference dates: factsheets are published monthly, and weights change all the time. Comparisons only hold for the same reference date. Here, the FTSE figure is seven months older than the MSCI figures.
• Buying twice: an ACWI ETF plus an emerging markets ETF counts emerging markets twice, and a World ETF plus the S&P 500 increases the US weight even further. Look-through reveals such overlaps.
• Confusing concentration with a forecast: a high weight of one country or sector is a risk, not a prediction of a crash.
• Your own weights without rules: if you weight regions equally, you need a rebalancing rule and must sit through periods in which market weighting does significantly better. Not investment advice.

How Portfolio Lab calculates it

Portfolio Lab includes templates with one ETF each on the FTSE All-World and the MSCI ACWI, as well as global portfolios made of World and emerging markets. In Look-through, you see the breakdown by country, region and sector for up to three portfolios, plus the largest single holdings across all funds, along with the metrics “Top 10” and “effective number” (1 / sum of the squared weights of the looked-through holdings). The app merges identical holdings from different funds by ISIN and name. Under Securities, you can filter ETFs by TER, fund size, replication, domicile and distribution policy. In the Backtest, you compare a single-ETF solution with combinations using calendar or tolerance band rebalancing, with trading costs; the turnover per year is shown in the results. How the country classification is derived is described in the look-through methodology.

This article explains a method; it is not investment or tax advice. Worked examples are simplified and rounded. All terms in the glossary.