ETF List by Sector: 2026 Classification Guide

What does "ETF list by sector" actually mean?

An ETF list by sector is a mapping of fund identifiers to sector labels, and the label is assigned by whoever built the list, not by the fund itself. That single sentence is the whole problem: the sector column is data, it has a source, and almost no published list tells you what the source is.

There are three places a sector label on an ETF can come from. It can come from a classification standard applied to the fund's underlying holdings, most commonly GICS or ICB. It can come from the index provider, whose index construction rules define an eligible universe that may or may not follow a public standard. Or it can come from the issuer's own marketing name, which is a product decision made before the fund launched and is never revised when the portfolio drifts.

Those three sources disagree, routinely and predictably. A fund marketed as an infrastructure ETF can be majority utilities and industrials under GICS, sit under a proprietary thematic taxonomy at its index provider, and appear on three different directories under three different headings.

None of this matters if you are reading a sector list as a menu. It matters a great deal if you are using it as an input: joining it to a risk model, computing aggregate sector exposure across a book, screening a universe for a mandate, or reconciling a portfolio against a benchmark. A label with unknown provenance behaves like a number with unknown units.

What are the 11 ETF sectors?

The eleven sectors are the GICS sectors: energy, materials, industrials, consumer discretionary, consumer staples, health care, financials, information technology, communication services, utilities, and real estate. GICS is maintained jointly by S&P Dow Jones Indices and MSCI, and it is the framework most US-listed sector ETF families are built against.

Two of those eleven are recent enough to break older files. Real estate was separated out of financials in 2016, and communication services replaced the old telecommunication services sector in 2018, absorbing a set of media and interactive services companies that had previously classified as information technology or consumer discretionary. An archived constituent file from 2014 has ten sectors and a different membership for at least two of the ten that survived.

ICB, the alternative standard, does not line up one for one. Its top level is industries rather than sectors, and it carves the economy differently: financials and real estate, technology and telecommunications sit in relationships that do not map cleanly onto the GICS eleven. A European-domiciled or FTSE-linked sector list will therefore not join row for row against a GICS list, and forcing the join with a lookup table introduces silent errors at the boundaries.

State Street's sector page on ssga.com, read on 1 September 2026, presents its sector fund range against the GICS framework and separately lists industry-level funds as their own group. That structure is worth noticing: the issuer itself distinguishes between a sector fund and an industry fund, even though a directory that publishes one sector column per ticker usually collapses the two.

Why does the same ETF appear in different sectors on different sites?

Because a directory has to assign one primary sector to a fund whose holdings span several, and each directory picks differently. The disagreement is not an error in either list; it is the visible edge of a modelling choice that neither list documents.

Thematic funds are the clearest case. A clean energy fund typically holds solar and wind equipment manufacturers that classify as information technology or industrials, utilities that own generation assets, and materials companies supplying polysilicon or components. There is no GICS sector that contains the theme, so one site files it under utilities, another under energy because of the name, and a third under industrials because that is the largest computed weight.

Industry-level funds create a second, quieter disagreement. A biotechnology fund is an industry within health care, and a semiconductor fund is an industry within information technology. One directory shows them at the parent sector, which makes the parent sector row look enormous, and another gives them standalone rows, which makes the same fund invisible to a sector-level filter.

Weighting variants inherit whatever the parent got. Equal-weight, revenue-weighted and fundamentally weighted versions of the same sector exposure are usually assigned the label of the cap-weighted original, even though their computed sector weights can differ by several percentage points once a large constituent is de-emphasised.

The directory on etfdb.com, as read on 1 September 2026, explicitly places funds under multiple themes at the same time rather than forcing a single sector column. That is structurally more honest than a one-label list, because it admits the fund is multi-sector. It is also harder to join against, since a many-to-many mapping cannot be used as a primary key and has to be resolved with a rule you write yourself.

How do you build a sector ETF list from holdings instead of labels?

You classify the constituents and aggregate up, so the sector "label" becomes a computed weight vector rather than a single string. The published label then becomes something you test against, not something you trust.

The sequence is mechanical:

1. Pull the fund's constituent file for a specific date, with weights.
2. Resolve every holding to a stable identifier that is not the ticker.
3. Attach a sector code to that identifier from a classification source you control.
4. Multiply sector membership by portfolio weight and sum by sector.
5. Compare the dominant computed sector to the published label and flag the gap.

Step two is where this fails in production. Tickers collide across venues and are reused after delistings, security names arrive in the holdings file in whatever form the issuer's fund accountant used that morning, and CUSIP or ISIN may be absent for non-US lines. When Bloomberg FIGIs are carried on each holding, as they are in AmericanETP's constituent list holdings files, the ticker-collision problem disappears at the join rather than being patched downstream with fuzzy name matching. The Bloomberg Extensions page covers the identifier fields available on the holdings side.

Here is the arithmetic, using round numbers to make the mechanics visible rather than sourced holdings from any named fund. Take a thematic infrastructure ETF with 60 holdings. Classify each line and aggregate: utilities 34 percent, industrials 31 percent, energy 18 percent, materials 9 percent, real estate 5 percent, cash 3 percent. The dominant sector is utilities at 34 percent, which means two thirds of the fund is not utilities. Any risk system that treats this fund as a utilities position is mis-stating 66 percent of the exposure, and any sector roll-up that adds it to a utilities bucket at full notional is wrong by construction.

The same aggregation trap shows up in valuation work. Summing a ratio across constituents is not the same operation as computing the ratio of the sums, which is why the P/E ratio of the S&P 500 is built differently by different vendors and why published PE figures for the same index disagree. Sector weights are the friendlier version of that problem, because weights at least sum to one, but the discipline is identical: state the aggregation rule, or the number is not reproducible.

Inverse ETF list by sector: why the look-through breaks

For an inverse or leveraged sector fund, the holdings file shows swaps, futures and cash rather than sector equities, so constituent-level aggregation returns nothing useful. The fund's sector exposure has to be read from its stated benchmark and its leverage factor, not from what is in the portfolio.

Open the holdings file for a -1x energy fund and you will typically find a handful of total return swap lines against counterparty banks, some index futures, and a large cash or treasury position collateralising them. Classify those constituents and you get financials and government, which is a true description of what the fund holds and a completely false description of what it is exposed to.

The sign matters as much as the magnitude. A -1x energy fund and a +1x energy fund do not cancel one for one in a sector roll-up unless you have adopted an explicit sign convention and applied it before summing. Without one, an account holding $10 million of each nets to $20 million of energy exposure in a naive weight aggregation, when the economic exposure is close to zero.

Daily-reset leverage adds a second complication that a static sector list cannot express. A 2x fund delivers twice the daily return of its benchmark, not twice the period return, so its effective sector beta over a month is path dependent and is not the number printed on the label. Treating it as a constant 2.0 multiplier is an approximation, and it should be documented as one.

The directory on etfdb.com, as read on 1 September 2026, carries inverse as a category separate from its sector taxonomy. That is the correct structural choice: leverage and direction are orthogonal to sector, and collapsing them into one column loses information that cannot be recovered. The machine-readable version of the same distinction is a leverage field on the fund record, which is what AmericanETP's fundamentals file carries alongside the ETF description and constituent counts.

What an institutional ETF holdings database has to carry

Sector work needs five things: the constituent file, a stable identifier on every holding, a leverage flag, a constituent count, and all of the above for past dates as well as today. Drop any one and a specific class of sector calculation stops being reproducible.

AmericanETP covers 3878 US and global indexes and US-traded ETFs, with constituent lists updated twice daily, a primary run at 6pm EST and a secondary run at approximately noon EST. Holdings files carry Bloomberg FIGIs. A separate fundamentals file carries the ETF description, leverage and constituent counts. Archived constituent list datafiles go back to July 2009, delivery includes FTP access, and pricing is $1500 per year or $150 per month for an individual and $2500 per year or $250 per month firm-wide.

The field-level view is more useful than the feature list:

The constituent count deserves its own sentence, because it is the cheapest integrity check available. A fund that reported 503 holdings yesterday and 61 today has almost certainly delivered a truncated file, and a roll-up that ignores the count will happily normalise those 61 weights to 100 percent and produce a sector vector that is arithmetically clean and factually wrong. The Constituent Change Report is the day-over-day view of the same signal, and sample files are on the Downloads page.

How does this compare to FactSet, Morningstar, ETF Global and the rest?

The market splits into two shapes, and the AmericanETP vs ETF Global comparison is a concrete way to see the split. Analytics platforms hand you a finished sector classification inside their own interface and their own data model; flat-file feeds hand you constituents and identifiers and let you classify against whatever standard your mandate requires.

Each side genuinely wins at something. Platforms such as FactSet, Morningstar, Refinitiv and S&P Global Market Intelligence carry their own classification research, entity and ownership hierarchies, corporate action handling and screening interfaces, and a CSV feed does not attempt any of that. If you want a sector answer without building anything, that is what those products are for, and building an equivalent internally is a multi-year commitment that most teams should not make.

A flat-file feed wins on a narrower set of things. It is cheaper to join into a pipeline that already exists, it does not force your sector definition to match a vendor's, and it does not require your calculation to run inside someone else's environment. If your risk system already has a GICS or ICB mapping and a security master, what you are missing is holdings with resolvable identifiers, not another classification opinion.

No competitor pricing appears in this article, because none of it was sourced today and a figure quoted from memory about a named vendor is worse than no figure. Where numbers were sourced from a vendor's own pages, they are published with their date in the FactSet pricing breakdown. For everything else, no claim is made here about what these vendors do or do not publish, since that was not verified today and is checkable on their own pricing pages in a couple of minutes; the comparison worth making is about positioning and workflow shape rather than list price.

Rebuilding sector history: why archived constituent files matter

Sector weights are only comparable over time if you use the sector membership that was in force on each date. Applying today's classification to a past portfolio produces a series that looks continuous and is not, in the same way a survivorship-biased universe looks clean and is not.

Two dates make this concrete. In 2016, real estate was separated from financials, so a financials sector weight computed before and after that date is measuring two different things unless you hold the mapping fixed at the historical version. In 2018, the communication services rework moved a group of large media and interactive services names out of information technology and consumer discretionary, which changed the technology sector's composition materially at the top of the cap range.

A back-test that maps historical holdings through the 2026 GICS structure will show a technology sector that never contained those names and a real estate sector that existed in 2012. Neither statement is true, and the resulting sector attribution is not wrong by a rounding error, it is wrong by whole positions.

Point-in-time reconstruction is the fix, and it needs dated constituent files rather than a current-membership table with an "as of" column bolted on. Archives reaching back to July 2009 cover the 2016 and 2018 changes with the pre-change files intact, which is what lets you compute a sector series that a reviewer can reproduce.

Snapshot timing is the adjacent problem. A daily constituent file is a statement about a specific moment, and as trading hours extend, that moment stops coinciding cleanly with a session boundary; the consequences for daily data are covered in what 24×5 trading breaks in data. Get the timestamp convention wrong and your sector history is off by a day at every rebalance, which is small until it is not.

If you want to talk through how a specific sector roll-up should be wired against the constituent and fundamentals files, contact the team.

FAQ

What are the 11 ETF sectors?

Under GICS, the eleven sectors are energy, materials, industrials, consumer discretionary, consumer staples, health care, financials, information technology, communication services, utilities and real estate. Sector ETF families listed in the US are usually built against this framework. ICB, used more widely in Europe, organises its top level as industries and does not map one-to-one onto these eleven.

What are the best ETFs for each sector?

There is no vendor-neutral answer, and any list claiming one is applying selection criteria it has not stated. The comparable questions are answerable from data: which fund tracks the sector definition your mandate uses, what its computed sector purity is under look-through, whether it is cap-weighted or equal-weighted, and how many constituents it actually holds. Those are all fields in a constituent and fundamentals file, and they let you rank funds against your own criteria rather than someone else's.

Which sector of ETF is best?

This is a market view, not a data question, and this article does not offer one. What data can tell you is how much of a given fund's exposure is actually in the sector on its label, which is frequently less than the name implies for thematic and multi-industry products. Compute the weight vector from holdings before treating any fund as a pure sector expression.

What is the 7% rule in ETF?

The phrase circulates in retail trading commentary as a stop-loss convention, generally meaning an exit after a position falls about 7 percent below entry. It is a position-sizing heuristic, not a regulation, a classification rule, or anything defined by an index provider or an issuer. It has no bearing on how sector labels are assigned or how holdings files are constructed.

Where can I download daily ETF holdings by sector?

Issuers publish holdings for their own funds on their fund pages, usually as a daily file in a format specific to that issuer. Aggregating across issuers into one schema, with consistent identifiers and a consistent update time, is what a constituent feed does. AmericanETP delivers constituent lists for 3878 US and global indexes and US-traded ETFs twice daily, at 6pm EST and approximately noon EST, with FIGIs on the holdings and FTP access available.

Is there a free ETF database with sector classifications?

Free directories exist and are useful for browsing, but they publish a label rather than the derivation behind it, and the labelling rules are generally not documented. For screening a universe by hand that is fine. For anything that feeds a calculation, the constraint is not cost, it is provenance: you need to know which classification standard produced the label and on what date, and you need the holdings to check it against.

How often should a sector ETF list be refreshed?

Fund-level sector labels change rarely, but constituent weights change every day, and a computed sector vector is only as current as the holdings behind it. Daily is the working minimum for anything used in risk or attribution, with an intraday secondary run useful when you need a same-day view before the primary evening file lands. Classification standard changes are infrequent but disruptive, so the mapping version should be stored with the output rather than assumed.

AmericanETP delivers daily updated ETF and index constituent lists with FIGIs, leverage flags, constituent counts and archives back to July 2009, as flat CSV files over FTP. See the coverage and file formats.