7.6% Then, 0.8% Now: The S&P 500's September 2026 Rebalance Lands 4 September — and Why Index Inclusion Stopped Paying

S&P Dow Jones Indices announces its September quarterly rebalance after the close on Friday 4 September 2026, with the changes effective prior to the open on Monday 21 September — the session after the 18 September quarterly expiry. The trade that headline used to imply is the one worth examining before the announcement lands, because it has been measured, and it is gone: the average abnormal return on an S&P 500 addition fell from 7.6% in the 1990s to 0.8% in the decade to 2020, and deletions fell from −16.6% to −0.6%. That happened while the money tracking the index kept growing, which is the opposite of what a downward-sloping demand curve predicts, and the reason why is visible in plain sight in the June 2026 announcement.

The calendar, and why the announcement date is the tradeable one

The S&P 500 is reconstituted on a schedule that has been stable all year. Changes are announced after the close, and take effect prior to the open of trading on the Monday following the third Friday of the rebalance month.

Both 2026 announcements so far confirm it. On 6 March, S&P DJI said Vertiv Holdings, Lumentum Holdings, Coherent and EchoStar would join the S&P 500 effective prior to the open on Monday 23 March. On 5 June, it announced that Marvell Technology and Flex would join the S&P 500, replacing Pool and The Campbell's Company, effective prior to the open on Monday 22 June, with the stated rationale that "the changes ensure that each index is more representative of its market capitalization range."

Two features of that table matter more than the names that will fill it. First, there are roughly eleven trading sessions between announcement and effect — the window in which every index fund on earth knows exactly what it must buy and sell, and exactly when. Second, the effective date is always the session after a quarterly options and futures expiry, which is already the highest-volume close of the quarter before a single index share changes hands.

One caveat on the convention: the first-Friday pattern describes the scheduled review. S&P DJI changes index constituents between quarters whenever a company is acquired, goes private or otherwise stops qualifying, and those announcements arrive on their own timetable — the March cycle itself carried a separate SmallCap 600 substitution effective 13 March.

What actually has to be true to be eligible

The published screens are a filter, not a selection rule, and the distinction does most of the work in this story.

Since 1 July 2025 the unadjusted company-level market-capitalisation floor for an S&P 500 addition has been US$22.7bn, raised from US$20.5bn; the same update moved the MidCap 400 band to US$8.0bn–US$22.7bn and the SmallCap 600 band to US$1.2bn–US$8.0bn. Those numbers are calibrated to percentiles of the S&P Total Market Index universe — approximately the top 85th percentile for the S&P 500, the 85th to 93rd for the MidCap 400, the 93rd to 99th for the SmallCap 600 — which is why they ratchet upward as the market capitalises rather than being set by judgment.

Alongside the size test: US domicile, listing on an eligible exchange, liquidity screens, at least 50% of shares outstanding publicly floated with security-level float-adjusted capitalisation of at least half the company-level minimum — about US$11.35bn on the current threshold — and positive GAAP earnings both in the most recent quarter and summed across the trailing four quarters.

The index effect, measured

The folklore is old and it was once true. Shleifer (1986) and Harris and Gurel (1986) documented abnormal returns of roughly 3% around S&P 500 addition announcements, and the finding became one of the standard challenges to the efficient markets hypothesis: prices moved on demand unrelated to fundamentals.

Greenwood and Sammon extended the series and found it had collapsed.

Both figures in the final row are described in the paper as statistically indistinguishable from zero. The authors estimate that S&P 500-tracking mutual funds and ETFs grew from essentially nothing in the 1980s to roughly 7% of the market in recent years, with other estimates higher still. A constant-elasticity demand curve hit by a growing shock predicts a growing price impact. The data delivered the reverse.

Why it decayed: the migration mechanism, visible in June

The largest identified reason is deceptively simple. Most additions are no longer additions to the passive complex at all — they are transfers within it.

The June 2026 announcement is the mechanism in one document. Flex appears twice: as an S&P 500 addition and as an S&P MidCap 400 deletion. Pool and Campbell's appear twice: as S&P 500 deletions and as SmallCap 600 additions. Nothing entered the S&P complex and nothing left it — four companies moved between its shelves, and the tracking funds on either side of each move were on opposite sides of the same trade.

The return data follows the structure. In the mid-1990s, migration and non-migration additions returned about the same — 6.7% and 6.4% respectively. By the late 2010s they had split: direct additions returned 2.2%, migrations returned −2.3%.

Three further forces sit alongside it. Value-weighted bid-ask spreads fell by roughly a factor of ten between the early 1990s and the late 2010s, so the cost of absorbing any demand shock collapsed. Index changes became more predictable, letting arbitrageurs position before the announcement rather than after it. And Wall Street built the other side of the trade on purpose: dedicated index-trading desks at large dealers now stand ready to supply shares into the auction. The paper's most striking evidence for that last point is that although trackers buy 7–8% of an added company on inclusion, total institutional ownership barely moves — active managers sell them the stock.

What the Marvell tape actually showed — and the trap in reading it

None of this means individual names sit still. Marvell rose roughly 6% in after-hours trade on 5 June when the announcement landed, and CNBC's coverage of the following session headlined a 10% jump. Point at that and the 0.8% average looks wrong.

It is not wrong; it is an average, and the Marvell tape is the cleanest available demonstration of why the average is the honest number. In the month before the announcement, Marvell had already risen about 57% on AI infrastructure demand — the same run that took its capitalisation across the eligibility threshold in the first place. The causal arrow runs from the fundamental repricing to the index inclusion, not the other way around, and any post-announcement move blends a genuine index demand shock with the continuation of the news that created eligibility. Marvell's own numbers have kept moving since, as its Q2 FY2027 print showed. Coherent, added in the March cycle, has traded on an entirely separate regulatory storyline since.

That is the whole analytical problem in one sentence: the selection is not random, so the event study is contaminated unless you control for what made the company selectable. Greenwood and Sammon's population-level estimate does that work; a single headline percentage does not. It also explains why "which names get added" is the wrong question to spend the next week on. The committee weighs sector balance and representativeness on top of the screens, and the authors' own conclusion is that while a rule of picking the largest eligible firm has become a better indicator over time, which precise stocks get added remains difficult to predict.

Where the volume actually goes: 18 September

For anyone whose exposure is to US500 or NAS100 rather than to single names, the rebalance is not a directional event. Two or three constituents changing inside a capitalisation-weighted index of roughly 500 companies moves the index level by an amount lost in the second decimal place — the added and dropped names sit at the small end of the distribution by construction.

What does change is the shape of the tape. The share of trading volume in the two months surrounding an index change that prints on the effective date itself rose from about 15% in the 1990s to almost 30% in the 2010s, and it concentrates in the closing auction because that is where index funds transact at the official close they are benchmarked to.

This quarter that auction has company. The 18 September expiry is one of the four quarterly options and futures expiries — the others in 2026 fell on 20 March, 19 June and 18 December — and it arrives two sessions after the 15–16 September FOMC meeting, which carries a Summary of Economic Projections. The rate backdrop is doing real work: the S&P 500 closed Friday 28 August at 7,712, around 1.6% below its 13 August record, after Chair Warsh's Jackson Hole remarks pushed September hike pricing sharply higher — the sequence traced in the Jackson Hole post, and the same rate channel that dominates the dollar's factor read.

The practical distinction is between information and mechanics. Unusual size in the final minutes of 18 September is index funds and expiring open interest finding each other at a printed price. It is not a revision to anything, and the two should not be read as one signal — the general point about how equity flows and currency flows interact is set out in stocks and the dollar.

What would change the picture

Three things would revive the effect the folklore still assumes.

A genuinely direct addition — a large company entering the S&P complex from outside it rather than migrating up from the MidCap 400 — restores the unmatched demand shock. Large recent listings that skipped the mid-cap indices entirely are the category to watch, and they are also the category the data says still moves: 2.2% for direct adds in the late 2010s versus −2.3% for migrations.

A change to the announcement convention would matter more than a change to the constituents. The pre-announcement window is what lets liquidity assemble; shorten it, or make changes unpredictable, and the arbitrage capital that currently absorbs the shock cannot position for it.

And a deletion of an unusually large constituent would test the asymmetry directly. Deletions were historically the bigger effect, they are the harder side to anticipate because continued membership has no mechanical test, and the −0.6% average of the 2010s was earned in a decade when almost every drop was a small company migrating down to the SmallCap 600 into waiting buyers.

Until one of those changes, the September rebalance is what the last decade of data says it is: a scheduled, fully telegraphed transfer of shares between two sets of index funds, executed in one auction, with an average price impact indistinguishable from zero — and a great deal of commentary attached to it that stopped being true around 2010.

To learn how Pip Theory builds its fundamental scores, see the methodology overview.

Educational macro context only — not investment advice.