Key takeaways
- MSCI World held 28.87% of its weight in information technology on 31 July 2026, against 9.09% for an even split across its 11 sectors.
- The UK's UCITS spread rules cap a single issuer at 5%, raised to 10% for up to 40% of the fund. The word sector does not appear in them.
- The Nasdaq-100 caps companies above 4.5% so their group stays under 48%, yet the index sat 68.51% in technology and 0.00% in financials on 30 June 2026.
- MSCI's sector neutral quality index tracked its parent's sector weights to within 0.83 points, while lifting its top 10 holdings to 34.48% against the parent's 26.41%.
- Equal-weighting the same 1,282 stocks cut technology to 10.44% and returned 8.33% a year since June 1994, against 9.01% for MSCI World at higher volatility.
The short answer: no capping rule in force caps a sector
You've looked at your global tracker, seen close to a third of it in one sector, and gone looking for a capped or sector-neutral version that fixes it. Here's the answer, and it isn't the one the labels suggest.
Every diversification rule that actually binds a fund constrains issuers. None of them constrains sectors. The UK's UCITS spread rules cap a single company. The US tax code's fund test caps a single issuer. The Nasdaq-100's own weighting rules cap a single company. A fund can obey all three and still hold two thirds of its money in one industry, and one of them does.
So the concentration in a plain cap-weight index survives almost intact through the constructions marketed as fixes for it. The one construction that did move sector weights hard was never designed as a sector control at all, and it charged for the privilege. What follows is the arithmetic, from the index factsheets and rulebooks themselves.
Sector concentration in a cap-weight index: 28.87% in one bucket, 6.5 effective sectors
Start with the reference case. A cap-weight index is sector-agnostic in the strict sense: it holds no view on sectors, and takes whatever mix the market's prices hand it.
MSCI World's factsheet dated 31 July 2026 shows 1,282 constituents across 23 developed markets. Its sector weights ran information technology 28.87%, financials 16.81%, industrials 11.45%, health care 9.17%, consumer discretionary 9.02%, communication services 8.04%, consumer staples 5.1%, energy 4.02%, materials 3.26%, utilities 2.54% and real estate 1.71%. The top 10 constituents came to 26.41% of the index.
An even split across 11 sectors would be 9.09% each. Technology at 28.87% is 3.18 times that. Real estate at 1.71% is under a fifth of it.
To put one number on the spread, we measure concentration as the sum of squared weights, the Herfindahl calculation. Applied to those 11 sector weights it returns 1,539. An even split returns 909. The reciprocal is easier to read: MSCI World was carrying the sector concentration of roughly 6.5 equally sized sectors, not 11. That figure and the ones that follow are ours, computed from MSCI's published weights, not MSCI's.
The rules that bind are issuer rules, and they were written that way
It's worth reading what the rulebooks actually say, because the gap between issuer caps and sector caps is where the whole question lives.
The FCA Handbook's COLL 5.2.11R, the spread rule for UK UCITS schemes, in the version dated 30 April 2026, reads: "Not more than 5% in value of the scheme property is to consist of transferable securities or approved money-market instruments issued by any single body." The next paragraph raises that: "The limit of 5% in (4) is raised to 10% in respect of up to 40% in value of the scheme property." That is the 5/10/40 rule European fund investors know, and the Handbook notes it as article 52 of the UCITS Directive. Read the whole of COLL 5.2 and the word "sector" appears nowhere in it.
The US equivalent works the same way. Section 851(b)(3) of the Internal Revenue Code, the diversification test a fund passes to be taxed as a regulated investment company, requires that "not more than 25 percent of the value of its total assets is invested in the securities of any one issuer", and that at least half the assets sit in positions each capped "in respect of any one issuer to an amount not greater in value than 5 percent of the value of the total assets of the taxpayer". The statute does mention "the same or similar trades or businesses", but only for issuers "which the taxpayer controls". That clause reaches issuers a fund controls, not sectors a fund merely holds.
MSCI built an index family to that tax rule. Its 25/50 methodology, dated October 2024, states the constraints plainly: "No group entity exceeds 25% of index weight" and "All group entities with weight above 5% cannot exceed 50% of the index weight." A 10% buffer means the rebalance targets 22.5% and 45% instead, quarterly, in February, May, August and November. Every one of those numbers is a company limit.
The Nasdaq-100 caps companies hard and sits 68.51% in one industry
If you want the cleanest demonstration that an issuer cap is not a sector cap, it's the Nasdaq-100.
Its methodology, in the 2026 edition, sets a two-stage constraint at each reconstitution. Stage one: if any company's weight exceeds 24%, "the weights are adjusted such that no company's weight exceeds 20%". Stage two: "Any resulting company weights that exceed 4.5% are added together. If the sum of those weights is 48% or greater, then that group of companies will have its aggregate weight adjusted down to 40%." Breaching either constraint on end-of-day values can trigger a special rebalance outside the calendar.
That is a strict cage by the standards of index construction. Now look at what it permits. The Nasdaq-100 factsheet dated 30 June 2026 shows the industry breakdown: technology 68.51% across 47 securities, consumer discretionary 16.43%, health care 3.56%, telecommunications 3.46%, industrials 3.13%, consumer staples 2.04%, basic materials 1.28%, utilities 1.14%, energy 0.45%, financials 0.00% and real estate 0.00%. On the same sum-of-squared-weights measure, that is 5,005, or about 2 effective industries out of 11.
Two of the 11 buckets hold nothing at all, and one holds more than two thirds. The company cap did its job. Nobody asked it to do the other one. Nasdaq classifies by industry rather than by the sector scheme MSCI uses, so the two concentration figures are not strictly comparable, but the shape of the point survives either taxonomy.
"Sector neutral" means neutral to the market's bet, not neutral between sectors
This is the label that misleads most, and the numbers make the point better than any definition.
MSCI runs two versions of its quality screen. Both select 301 stocks from MSCI World on three fundamentals: return on equity, leverage and the stability of earnings. The plain version is sector-agnostic: it takes the highest scorers wherever they sit. The sector neutral version, in MSCI's own words, ranks securities "relative to their peers within the same GICS sector", which is what holds each sector's weight close to the parent index.
The sector-agnostic version, on 31 July 2026, ran information technology at 37.1%, health care 14.64%, industrials 13.68%, communication services 10.41%, financials 9.5%, consumer staples 7.9%, consumer discretionary 3.98%, materials 2.56%, real estate 0.15% and energy 0.09%. Utilities were absent entirely. Its sum-of-squared-weights concentration comes to 2,061, or 4.85 effective sectors. Measured as sector active share, half the sum of the absolute weight differences, it sat 21.09 points away from its parent.
The sector neutral version, on the same date, ran technology 29.7%, financials 16.22%, industrials 11.53%, consumer discretionary 9.67%, health care 8.99%, communication services 7.71%, consumer staples 4.89%, energy 3.83%, materials 3.22%, utilities 2.54% and real estate 1.7%. Compare that line against the parent's. The largest single deviation anywhere is 0.83 points, in technology. Sector active share is 1.55 points. Its concentration measure is 1,570 against the parent's 1,539.
So the sector-neutral index is fractionally more sector-concentrated than the cap-weight index it neutralises against. That is not a flaw in the product. It is what the product says on the tin, once you read "neutral" as "neutral versus the benchmark" rather than "even-handed between sectors". Neutrality is defined against the market's sector bet, so it reproduces that bet by construction.
There's a second-order cost that the sector label hides. Holding sector weights fixed while selecting only 301 of 1,282 stocks forces the weight into fewer names inside each sector. The sector neutral index's top 10 came to 34.48%, against 18.83% for those same 10 constituents in the parent, and against the parent's own top 10 of 26.41%. Constrain the sectors and the concentration moves into the stocks. The same tension shows up when funds are stacked on top of each other, which is the subject of our piece on fund overlap and mega-cap concentration.
The one construction that moved sector weights was not built to
Equal weighting is a stock-level rule with nothing to say about sectors. It just holds every constituent at the same weight and rebalances back. Yet it is the only construction here that changed the sector mix substantially.
MSCI World Equal Weighted holds the identical 1,282 constituents. Its largest position was 0.13% against 5.18% in the parent, and its top 10 came to 1.17%. On 31 July 2026 its sector weights ran industrials 19.35%, financials 19.09%, information technology 10.44%, consumer discretionary 9.82%, health care 9.19%, materials 6.98%, consumer staples 6.79%, utilities 5.58%, real estate 4.59%, communication services 4.37% and energy 3.8%.
Technology fell by 18.43 points. Real estate rose from 1.71% to 4.59%. Utilities more than doubled. The concentration measure drops to 1,209, or 8.27 effective sectors against the parent's 6.5. Sector active share is 22.32 points, slightly wider than the sector-agnostic quality screen managed in the opposite direction.
The chart above plots those five concentration readings on one scale: 909 for an even sector split, 1,209 equal-weighted, 1,539 for MSCI World, 1,570 sector neutral, 2,061 sector-agnostic quality. The index with "sector neutral" in its name sits on top of the cap-weight index. The one with no sector rule at all sits furthest from it.
What the de-concentration cost, measured on each index's own basis
Lower concentration is not free, and the factsheets price it.
MSCI World Equal Weighted returned 8.33% a year gross since 30 June 1994, against 9.01% for MSCI World over the same window. Its annualised standard deviation since June 1994 was 15.40% against 14.85%, and its Sharpe ratio 0.41 against 0.47. Over 32 years it delivered 0.68 points a year less return at higher volatility. Its trailing 12-month turnover was 31.16% against 2.95%, which is the trading bill behind the gap.
The quality pair tells a subtler story, and the two factsheets are quoted on different bases, so each is compared only with the MSCI World line printed alongside it. Over 10 years to 31 July 2026, the sector-agnostic quality index returned 14.74% a year gross against MSCI World's 13.29%, an edge of 1.45 points. The sector neutral version returned 12.38% a year net against MSCI World's 12.73% on the same net basis, trailing by 0.35 points. Same stock-selection rule, same 301 names, and the version that kept the sector bet is the version that beat the market over that decade.
That's the uncomfortable symmetry. Sector concentration is not only a risk you are carrying. Over this particular decade it was also where the return came from, which is why the pieces on equal weight vs market cap keep landing on such narrow margins.
The strongest objection: concentration is an outcome, not a design choice
The serious counter-argument to everything above is that a cap-weight index is doing exactly the right thing, and the concentration is the market's answer rather than the index's error.
The case has force. A cap-weight index needs almost no trading, and MSCI World's 2.95% annual turnover against 31.16% for the equal-weighted version is the price of the alternative. Sector weights move because prices move. Holding them down means selling companies for being large, and over the 32 years from June 1994 that trade came out 0.68 points a year behind. The 18.43-point technology underweight equal weighting produced was not an insight. It was a side effect, and it was not free.
There's a fair objection to our concentration measure too. Sum-of-squared-weights treats the 11 sector buckets as equally distinct, and they aren't. A firm classified in communication services and one in information technology can share more risk than two technology firms do. The measure counts labels, not correlations, and 6.5 effective sectors overstates the diversification if the labels have drifted from the economics.
What these numbers cannot tell you
Every figure here is a snapshot. The MSCI weights are dated 31 July 2026 and the Nasdaq-100 breakdown 30 June 2026. Sector weights change with prices, so the 28.87% technology reading has no permanence and neither does the 1,539.
The return comparisons carry heavier limitations. MSCI World Equal Weighted was launched on 22 January 2008 and the sector neutral quality index on 11 August 2014, so returns before those dates are back-tested, which MSCI states on both factsheets. A back-test is not a forecast, and index families are built knowing which back-tests looked good.
The 10-year comparison also covers a single decade, and a different window ranks these constructions differently. From 31 October 2007 to 9 March 2009 the equal-weighted index fell 59.66%, MSCI World fell 57.46% and the sector-agnostic quality index fell 48.01%. On that episode the most sector-concentrated of the three fell least. The data here cannot tell you which pattern the next decade resembles. One further caution: the gross and net figures are not interchangeable, which is why each index above is compared only against the MSCI World line printed on its own factsheet.
What would change the conclusion
The finding is narrow and it has a clear failure condition. It holds only while the diversification rules stay written at the issuer level. If the FCA added a sector limit to COLL 5.2, or a major provider launched a genuinely capped-sector version of a mainstream index with real assets behind it, the sentence "no rule caps a sector" stops being true and the arithmetic here has to be redone.
The second thing that would change it is the market itself. Sector concentration reads as a problem because one sector is near 29%. If prices spread out and the largest sector fell back toward that 9.09% even split, the cap-weight index would converge on the capped alternatives and the whole comparison would collapse into noise.
The measure to watch isn't the label on the index. It's the gap between the largest sector weight and an even split, currently 28.87% against 9.09%, and the top 10 weight that moves with it. When a construction claims to address concentration, those two numbers say whether it did. LedgerTouch reports both across the funds you actually hold, which is where a sector bet usually hides. The related question of what happens when one company inside that sector gets big enough to matter on its own is covered in our piece on NVIDIA index concentration.