Before Calling a Move Company-Specific, Check Which Rulebook Assigned That Stock Its Sector

A single name moves four percent on a quiet tape. The first instinct is to look for the reason inside the company — a filing, a guidance revision, a broker note that landed before the open. Before that search starts, there is a cheaper question worth asking: how much of that four percent was being done to every name sitting beside it?

Which names sit beside it is not a matter of taste, and it is not whatever a screener happens to sort by. Industry membership is assigned by published rulebooks. Two of them are built for markets and one is built for federal statistics, and the three do not agree with each other. All three publish their methodology, all three revise it on a calendar that is announced in advance, and none of them ask the person holding the position for an opinion.

The consequence is narrow but useful. If a move cannot be split into the part that belonged to the market, the part that belonged to the sector, and the part left over, the post-mortem written about that trade is describing the wrong thing.

The sector label is assigned, not chosen

The Global Industry Classification Standard, developed jointly by S&P Dow Jones Indices and MSCI and launched in 1999, is a four-tier hierarchy. As documented in the GICS methodology, it currently comprises 11 sectors, 25 industry groups, 74 industries and 163 sub-industries. Sector codes run from 10 for Energy through 60 for Real Estate.

Assignment is rule-driven rather than discretionary. The GICS methodology sets a revenue test first: a company is classified in the sub-industry whose definition most closely describes the business activities generating more than 60% of the company’s revenues. Where no single activity clears that bar, the classification falls to whichever business provides the majority of both revenues and earnings, with market perception recognised as a further relevant input. Companies diversified across three or more sectors land in Industrial Conglomerates under Industrials, or Multi-Sector Holdings under Financials.

The competing standard, the Industry Classification Benchmark maintained by FTSE Russell, is also four-tier and also revenue-anchored, but the shape is different: 11 industries, 20 supersectors, 45 sectors and 173 subsectors. FTSE Russell documentation describes company classifications as reviewed annually and commits to a minimum of six months’ notice before structural changes to the taxonomy itself.

Two rulebooks, both defensible, both public, and not identical. A name benchmarked against an ICB supersector and the same name benchmarked against a GICS industry group are not being measured against the same peer set.

How much of the move the sector explains

The honest answer is that it is usually a minority share, and that the minority share is large enough to matter. S&P Dow Jones Indices has published an attribution on exactly this question. Measured as the ratio of the squares of inter-sector and total effects, sector effects accounted for an average of 22% of the S&P 500’s monthly dispersion since 1989.

Two readings follow from that single number, and both are worth holding at once.

• Roughly a fifth of the spread between winners and losers within the index, on a monthly view, traces to which sector a name was in rather than to anything the company did. That is not noise. A position sized as though it carried only company-specific risk is carrying more correlated exposure than the sizing assumed.
• The other roughly four-fifths did not come from the sector. Sector attribution is a filter, not an explanation. Treating a sector story as a full account of a single-name move discards most of what moved it.

The same S&P DJI work notes that sector effects were not evenly distributed through the calendar: in U.S. presidential election years, sector effects were elevated in the November months, contributing a greater-than-average proportion in 75% of instances. That is a statement about a historical distribution and about specific months, not a claim about what any future November will do.

The decomposition above is arithmetic, not forecasting. What makes it fragile is the middle row. Subtracting “the sector” requires choosing a sector index, and that choice is inherited from a classification decision made by a committee that publishes its reasoning.

Two taxonomies, two different units of measurement

The gap that catches people out is not GICS against ICB. It is either of those against the classification system that federal economic data uses.

The North American Industry Classification System is the framework behind Bureau of Labor Statistics, Census Bureau and Federal Reserve industry series. The Census Bureau describes NAICS as based on a production-oriented concept, grouping establishments into industries according to similarity in the processes used to produce goods or services. Its hierarchy has five levels — sector at two digits, subsector at three, industry group at four, NAICS industry at five, national industry at six — and 20 sectors at the top, coded 11 through 92.

Three differences do the damage.

1. Unit. GICS and ICB classify a company. NAICS classifies establishments. One company can be spread across several NAICS codes, so no single NAICS number describes it the way a single GICS code does.
2. Test. GICS asks where the revenue comes from. NAICS asks how the output is produced. A firm selling a service and a firm manufacturing the hardware that delivers it can share a revenue-based bucket while sitting in different production-based ones.
3. Clock. BLS states that reviews of NAICS are scheduled every five years, with NAICS 2022 the current version; the Census Bureau has indicated that recommendations for the 2027 revision were expected in the Federal Register in early 2026. GICS is reviewed annually. The two taxonomies therefore drift apart between revisions and then re-converge unevenly.

The conversion itself is dated and public. The BLS Current Employment Statistics program moved from NAICS 2017 to NAICS 2022 with the January 2023 data released on February 3, 2023, applying the conversion retroactively to history, with revisions described as minor in most sectors and major in retail trade and information.

So when a macro release describes strength in an “industry” and a sector index prints a move in a similar-sounding “sector,” the two labels came from different rulebooks answering different questions about different units. Whether they point the same way is something to check, not assume.

The peer group is mostly not on the screen

There is a second, quieter mismatch. A sector is a global construct; a trading screen usually is not. S&P Dow Jones Indices publishes the share of each GICS sector, measured inside the S&P Global BMI, that is listed in the United States.

As of December 31, 2025, the United States accounted for 61.4% of the S&P Global BMI overall, but the sector-level figures ranged from 78.5% for Information Technology down to 30.5% for Materials, with Industrials at 48.3% and Financials at 48.7%. For a Materials name, close to seven-tenths of the global sector complex prices in sessions that a U.S.-hours screen never shows. For an Information Technology name, roughly four-fifths of it prices in the same session as the position.

That changes what an overnight gap means. Where a sector’s global float is majority non-U.S., part of what looks like an opening gap is a peer group that already traded while the U.S. book was shut. Where it is concentrated in U.S. listings, the same gap is more likely about the market or the company.

Reclassification arrives on a published calendar

Classification changes are not surprises. They are announced, consulted on, and dated.

The most disruptive recent example is documented end to end. S&P DJI and MSCI announced revisions to the GICS structure on March 31, 2022, effective after the close on Friday, March 17, 2023 — close to a full year of notice. The Data Processing & Outsourced Services sub-industry was discontinued and its constituents dispersed: transaction and payment processing companies moved into a newly created Transaction and Payment Processing Services sub-industry under Financials, payroll processors moved to Industrials, and travel data processing moved to Consumer Discretionary. Broadline Retail was created by merging the former General Merchandise and Department Stores categories. Affected large-cap companies were identified by June 30, 2022, with the full company list published by December 15, 2022. Notably, proposed renewable energy and cannabis-related changes were not adopted, cited as lacking market consensus — a consultation does not guarantee a change.

A live consultation is running now on the same pattern. S&P DJI and MSCI opened a GICS consultation on July 17, 2026 that closes on October 30, 2026, covering the classification of artificial-intelligence-related activity — including a restructuring of semiconductors, definitions for high-performance computing services and data-lifecycle services, and the treatment of foundation model developers — along with the Application Software sub-industry and listed investment companies. Any resulting changes are to be announced by November 2026. The announcement states plainly that the consultation may or may not result in changes.

The index mechanics that follow a reclassification are equally scheduled. In the S&P North American sector index methodology dated March 2026, membership is reviewed semi-annually, effective after the close on the third Friday of June and December, weight capping is applied quarterly in March, June, September and December, and a company whose classification change removes it from a qualifying sector is dropped at the next reconstitution rather than immediately.

Capping matters more than it sounds. Select Sector indices use a capped market-capitalisation weighting designed to reflect diversification requirements imposed on investment companies under the U.S. Internal Revenue Code and the Investment Company Act of 1940, with an iterative capping approach introduced at the September 2024 rebalance. The gap this opens is measurable: as of December 31, 2025, the S&P 500 Information Technology index showed a three-year annualised return of 38.81% against 33.34% for the capped Technology Select Sector index — same label, same names, different rules, 5.47 percentage points a year over that window.

Anyone comparing a single name to “its sector” is therefore comparing it to a specific rules package, not to an abstraction.

What Would Invalidate This

This frame is a filter with clear boundaries. Several conditions push it out of usefulness.

• Wrong horizon. The 22% figure describes monthly dispersion within the S&P 500 since 1989. It is not an intraday statistic, it is not a statistic about small caps, and it is not a statistic about any individual name. A five-minute move attributed to a sector effect using a monthly-frequency justification is a misuse of the number.
• Concentrated sector index. If a handful of names dominate a sector index by weight, subtracting “the sector” can subtract what is itself company-specific risk from one or two constituents. In that case the residual is not cleanly idiosyncratic, and the capped-versus-uncapped gap above is a direct measure of how much concentration is present.
• Genuinely diversified issuers. A company classified into Industrial Conglomerates or Multi-Sector Holdings has been placed there precisely because no sector describes it. Its sector benchmark carries little information about it, and the decomposition should be treated as uninformative rather than as evidence of a large idiosyncratic component.
• No sector vehicle. Part of why sector membership transmits into price is the flow through index-tracking products. Where a listing sits outside the indices that such products track, the mechanical channel is weaker and the classification is closer to a label than a force.
• Recent reclassification. In the window between an announced classification change and the reconstitution that implements it, historical sector series and current membership describe different baskets. Attribution computed across that boundary is comparing two different peer groups.
• Cross-listing and currency. For sectors with a large non-U.S. float share, part of the apparent sector move is a currency move and a different session’s news, neither of which is a statement about the company.

Concrete Framework

1. Write down the classification before writing down the thesis. Record the GICS sector, industry group, industry and sub-industry, and separately the ICB industry and supersector if the name is benchmarked outside the U.S. If the two disagree, note which one the relevant index and index-tracking products follow — that is the one with the mechanical channel.
2. Check the 60% test. If revenue is split so that no activity clears 60%, the classification rested on the revenue-and-earnings majority test and on market perception. That is a weaker anchor, and the sector benchmark should be given correspondingly less weight in the post-mortem.
3. Decompose before diagnosing. For any move worth explaining, record three numbers in the journal: the broad index move, the sector index move, and the residual. Do this before searching for company news. Searching first tends to find a story for the market component.
4. Match the frequency of the benchmark to the frequency of the claim. A monthly attribution statistic supports monthly claims. For an intraday decision, compute an intraday decomposition or state explicitly that the attribution is unmeasured.
5. Look up the sector’s non-U.S. float share once per sector. Above roughly half non-U.S., treat overnight gaps as partly a peer-session event and check whether the overseas complex moved before attributing the gap to the name.
6. Put reconstitution dates in the calendar. The third Friday of June and December for membership reviews, and March, June, September and December for capping, in the sector index framework cited above. Confirm the dates in the methodology that governs the specific index being used, since families differ.
7. Track open consultations, not just completed changes. The current GICS consultation closes October 30, 2026, with any changes announced by November 2026. A consultation is a forward notice that the peer group definition may move; it is not a signal about price direction and should not be traded as one.
8. Size for the correlated share. If sector effects have historically explained a meaningful minority of dispersion, then several positions inside one sector are less diversified than a position count implies. Set a sector-level exposure limit alongside the per-position limit, and check it against the classification recorded in step one rather than against an impression of what the names do.
9. Re-check the classification after any corporate restructuring. GICS reviews a company when a significant corporate restructuring occurs or when a new annual report becomes available. A merger, spin-off or a material shift in revenue mix can move the bucket, and with it the benchmark that every prior attribution in the journal assumed.

None of this makes a single-name move predictable. It makes the record of that move accurate, which is a lower bar and a more achievable one. The sector component of a move is defined by a document that anyone can read, revised on a schedule that anyone can look up, and implemented on dates that are announced months ahead. Treating it as background noise is a choice to leave public information on the table.