The Rout the Average Stock Missed
The S&P 500 fell 3% over a month in which the average company inside it rose about 2%. We think the gap says more about what a concentrated benchmark measures, and who sits downstream of it, than about artificial intelligence.
By the team at Banyantree Investment Group
This essay is adapted from analysis first shared in our Monthly Investment Letter of 29 June 2026.
June read as a poor month for US equities, and for the index it was. Over the month to 26 June, the S&P 500 fell 3.0%.¹ Now take the same list, 500 companies held across 503 listed securities, and change nothing except the position sizes. The S&P 500 Equal Weight Index, which holds every constituent of the S&P 500 at roughly the same weight and resets to that weight each quarter, rose 2.1% over the same period.² One list of companies, two portfolios, five percentage points apart in a single month. The entire difference is how much of each company was held.
We think that gap is the most informative number June produced, and that much of the commentary read it backwards. The month was widely described as a technology rout, and a few louder voices called it the AI bubble bursting. Before reaching for either story, it's worth asking a narrower question: which weights produced the result?
Put rough numbers on it. The ten largest companies ended 2025 carrying just over 40% of the index.³ Use the December figure as a rough anchor only: at any top-ten weight from 35% to 45%, the conclusion that follows barely moves. Suppose the other 493 averaged about what the equal-weight index printed, call it 2%. The arithmetic then forces the remainder: for the whole index to fall 3.0%, the ten largest must have lost 9% to 12% between them, on average, inside one month; call it roughly 10% at the December weight. The figures are approximate and the method is blunt. The shape doesn't move. The decline was concentrated in the handful of companies carrying the greatest weight, led by chipmakers and the technology names around them,⁴ while the average stock in the index finished the month higher.
That shape narrows the readings available, which is what makes it useful. A broad retreat from equities is harder to square with it: uniform selling across the index would tend to weigh on both portfolios, and an average constituent finishing higher sits awkwardly beside it. What the prints are consistent with is rotation within equities rather than an exit from them. Index returns don't identify whose money moved where, and an average doesn't count how many stocks rose; both limits belong on the record. Within those limits, the prints show the average constituent return was positive while the largest weights fell.²
The bubble reading fails the same test, at least on June's evidence. A repricing of the largest AI-adjacent securities and a failure of the AI adoption story are different events, and the tape can only speak to the first. Revenue can grow, adoption can broaden, and a stock can still fall a long way if the price had already assumed more than the growth delivered. Equally, nothing in one month's rotation proves the adoption story intact; the evidence for that lives in earnings, capacity, and usage, none of which reprints daily. What June actually shows is narrower than either camp wants it to be. The market's largest positions fell in price while the average stock rose, inside the same index. As at late June, that is the whole of what the divergence says.
The reason it says so much anyway is what those weights have become. At this level of concentration, "the market" has turned into a loose word. It can mean the total value of large US companies, in which case June was bad. It can mean the experience of the average listed business, in which case June was mildly good. It can mean the breadth of participation, or it can mean the benchmark a manager is judged against. Those four used to travel together closely enough that the distinction was pedantic. Concentration has pulled them apart.
Capitalisation weighting has real virtues, but one property matters here. The method carries yesterday's performance into tomorrow's position sizes. As a company outperforms, its weight rises automatically; in an existing cap-weighted portfolio the price move itself produces the larger position, with no purchase required and no committee resolution recording the decision. Run that process through a long stretch of narrow leadership and the index gradually becomes a concentrated portfolio that no discrete decision ever enlarged. Choosing the benchmark was a decision, and retaining it is one; what nobody ever approved is taking the largest holding from 5% of the portfolio to 8%, because the price did that on its own. Every individual weight is defensible as the market's current appraisal of one company. The portfolio they add up to is sized by what has already run hardest, wearing an index label.
The drift is worth quantifying, because it runs faster than intuition suggests. A company at 5% of the index that outperforms the rest by 20 points a year sits near 8% of it three years later, and nobody tracking the index needed to buy a single incremental share along the way. The institution's largest exposure has grown by roughly two-thirds through price action alone, and the minutes record no decision, because none was made. The figures are illustrative. The compounding is not.
Which is where the governance problem starts, because wherever benchmark weight is treated as the position requiring no explanation, that portfolio is also the working definition of neutral, and every deviation from it is the thing that has to be defended. Under dispersed leadership the convention is harmless. Under concentrated leadership it means the default exposure is being set by past outperformance, and a default, by its nature, is the one setting nobody is asked to justify at the table.
Watch it operate on a mandate with an ordinary risk control. A committee caps any single holding at 5% of the portfolio; a sensible rule, written to limit dependence on one company. While every benchmark weight sits below 5%, the cap does nothing. Once a constituent grows past it, the rule manufactures an automatic underweight, and the arithmetic of that underweight is worth writing down. A company at 8% of the benchmark, held at the 5% cap, is a 3 percentage point active position in a single name. In a year when that company beats the index by 30 points, the cap alone costs about 90 basis points of relative return, before the manager has made a single discretionary call. Run June through it using the decomposition above: a name falling roughly 10% against an index down 3.0% underperforms by about 7 points, so a 3 point underweight adds roughly 20 basis points of relative return in a single month. Same rule, opposite verdicts, and neither verdict says anything about the manager's skill.
All of this can be deliberate. A committee can adopt the S&P 500, impose the cap, and understand exactly what follows: a diversification benefit purchased with tracking error, and stretches of relative underperformance accepted as the intended cost. In that case, the process is working as designed.
The failure mode sits at the next review. The manager is asked why the portfolio owned less of the benchmark's best-performing company. If the review stops there, a rule written to limit dependence on one company is being judged by the weight that company's own outperformance produced, and the committee has let the benchmark's verdict stand in for its own. Continued leadership then makes the original trade-off dearer each year, and a committee can respond coherently in either direction: keep the cap and own the active risk, or change the mandate because its tolerance for that risk has genuinely changed. What it should not do is treat the change as a repair for disappointing numbers, because that converts a risk decision into a performance chase with governance paperwork attached. The question that keeps the review honest is short. Are we judging this against the risk the rule was written to control, or against the weight the winner happens to have reached?
The tension has also escaped the committee room, into documents anyone can check. Regulated funds run under diversification limits drafted when today's weights would have seemed implausible, from the US tax code's tests to Europe's UCITS 5/10/40 framework, sketched here in simplified form, and index providers now publish capped benchmark variants in part so that funds tracking them can stay inside such limits.⁵ One trigger has already fired. In July 2023, Nasdaq invoked the special-rebalance provision of its own methodology and cut the weights of the Nasdaq-100's largest constituents, after their combined weight exceeded the concentration threshold the methodology monitors.⁶ An index provider reached into its own rulebook and rebuilt its benchmark mid-year, against the market's live verdict on those companies. Whatever else the episode was, it is documentary evidence that "the market portfolio" and "a portfolio the rules permit" had stopped being the same thing.
An Australian committee doesn't need to look at New York to run any of this. The largest bank in the ASX 200 ended 2025 at just over 10% of that index on its own,⁷ with the ten largest constituents together approaching half of it, a more concentrated top line than the S&P 500's. Any capped domestic mandate, and any fund run under UCITS-style limits against an Australian benchmark, has been living the identical arithmetic for years.
The strongest objection to all of this is a good one. Capitalisation weighting is the only weighting every investor can hold at once; in aggregate, we all hold the cap-weighted portfolio, because that is what the market is. On that reading, concentration is an outcome, and calling it a flaw in the index is blaming the mirror. The objection continues into the alternative. Equal weighting is an active bet, and a specific one: it tilts systematically toward smaller companies, it leans against momentum, its quarterly rebalance sells relative winners to buy laggards, it turns over more, and at institutional scale it meets real capacity limits in the smallest constituents.⁸ And the leaders may simply deserve their weights. Strong margins, deep balance sheets, and the capacity to fund enormous investment programmes can justify both superior earnings and greater index share. If the largest companies' share of index earnings has kept pace with their share of index weight, the concentration is being paid for, and the inertia worry is mostly theoretical.
Nearly all of that is right, and the argument here doesn't need it to be wrong. Over long stretches the evidence will embarrass anyone confident that cap weighting is broken or that equal weighting is better. The claim here is narrower: it concerns the institutional role the cap-weighted portfolio has acquired. A market outcome is a fine thing to measure. It is a stranger thing to treat as the position requiring no justification once the outcome has become this concentrated, because at that point "no view" and "a large view on ten companies" have become the same portfolio. The earnings test inside the objection is the right test, and it cuts both ways: concentration is easiest to defend while the leaders' share of profits keeps pace with their share of weight, and hardest from the moment those two shares part company, because from then on the index's structure depends on investors paying an ever-larger premium for the same earnings leadership.
That premium can be read off two public numbers, so read it. As at the end of 2025, the ten largest companies were about 40.7% of the index's value and roughly 32% of its expected 2025 earnings.⁹ Dividing the first share by the second puts the leaders' collective price-to-earnings at about 1.3 times the index's own. The sharper comparison sets the leaders against the remaining 490 companies directly, each side's weight share over its earnings share, and at those figures the market is paying close to one and a half times as much for a dollar of the leaders' expected earnings as for a dollar of everyone else's. Nothing in either ratio says the premium is wrong; durable growth deserves one. What the ratios do is convert a mood into a measurement. A committee that writes them down each quarter will notice when the premium is widening while the earnings share stands still, which is the precise condition under which concentration stops being paid for and starts being financed by the next buyer.
June is one month, and one month proves nothing about which weighting wins from here. What history offers is a base rate for how large these gaps can run, and for which conditions switch their sign. In 2000, the year the last great concentration episode broke, the S&P 500 returned minus 9.1% while its equal-weighted counterpart returned plus 9.6%: a gap of nearly 19 percentage points in a single year, in the same direction as June's.¹⁰ In 2023 the gap ran the other way, the cap-weighted index returning 26.3% against 13.9% for equal weight, as a handful of leaders did nearly all of the work.¹⁰ One honesty note belongs beside the 2000 figure: S&P launched the equal-weight index in January 2003 and calculated its earlier history retrospectively,¹¹ so nobody actually managed money against that print in real time. Neither episode maps cleanly onto now. 2000 doesn't tell you the leaders are about to break, and 2023 doesn't tell you they're about to resume. What carries over is that the divergence can run enormous in both directions. Whether earnings decide the direction is what the readings below are built to test.
Which is why June begins a test rather than settling one, and the test has two recognisably different outcomes.
Breadth by addition looks like this: the largest companies keep delivering, earnings revisions improve across more sectors, and both indices rise, with equal weight closing part of the gap because more businesses are contributing rather than because the leaders are failing. Breadth by subtraction is the other pattern: equal weight outperforms mainly in months when the cap-weighted index falls, with weakness in the largest positions funding the relative improvement elsewhere. The first is the benign resolution of a concentrated market. The second may still be a healthy reversal after a narrow run, but it turns dangerous if earnings outside the leaders fail to strengthen before the selling spreads.
Three readings will separate them, and all three are public. First, the pattern of equal-weight outperformance itself: whether it arrives in rising months or only in falling ones. By that test, June sits on the subtraction side of the ledger: the cap-weighted index fell and the relative improvement was funded from the top. The equal-weight portfolio rising in absolute terms makes it a comparatively benign version, and one month settles nothing. Second, revisions breadth: whether earnings estimates outside the ten largest companies are being revised up as the rotation proceeds, which would give the broadening actual support. Third, the two shares behind the premium above, watched over quarters rather than weeks. We'd put the most weight on the second reading, because flows can rotate for months on positioning alone, and prices without earnings behind them have a way of giving the move back. And we'd revise our reading of June quickly on the evidence: a fast return to concentrated leadership, with broadening revisions underneath it, would say the month was a positioning reset and nothing more.
The index fell 3.0%. The average company inside it went up. Both numbers are true, and only one of them gets quoted as the market. Before the next review reaches for either, it is worth deciding which one the mandate was actually written about.
General information only. Not personal advice. Past performance is not indicative of future performance. Examples are illustrative. This material is intended for wholesale and professional investors.
Notes
1. S&P Dow Jones Indices, index performance data: over the month from 29 May to 26 June 2026, the S&P 500 declined 3.0%.
2. S&P Dow Jones Indices, index performance data: the S&P 500 Equal Weight Index rose 2.1% over the same period. The index holds each S&P 500 constituent at an equal weight, rebalanced quarterly, and the S&P 500 comprised 500 companies across 503 constituent securities owing to multiple listed share classes.
3. S&P Dow Jones Indices, Why Does the S&P 500 Matter to Australia?, Index Education, March 2026, Exhibit 5: the ten largest S&P 500 constituents carried a combined weight of 40.73% as at 31 December 2025.
4. S&P Dow Jones Indices sector index data, June 2026: information technology and semiconductor-related constituents led the decline in the capitalisation-weighted index.
5. US Internal Revenue Code section 851(b)(3): for at least 50% of a regulated investment company's assets, positions must satisfy diversification conditions including 5% single-issuer and 10% voting-securities tests, with separate 25% limits applying to larger concentrations; see also Revenue Ruling 2003-84. Directive 2009/65/EC (UCITS), Article 52: single-issuer exposures above 5% of fund assets may not in aggregate exceed 40%; Article 53 permits higher single-issuer limits, generally 20%, for qualifying index-replicating funds. The two regimes differ in structure and are summarised here only to the extent the essay relies on them. Index providers including S&P Dow Jones Indices, MSCI, and Nasdaq publish capped variants of headline benchmarks to accommodate such limits among other purposes.
6. Nasdaq, The Nasdaq-100 Index Special Rebalance, and The Nasdaq-100 Index Special Rebalance to Be Effective July 24, 2023, both 7 July 2023: a special rebalance conducted under the index methodology's discretionary special-rebalance provision to address overconcentration, reducing the aggregate weight of the largest constituents.
7. S&P Dow Jones Indices, Why Does the S&P 500 Matter to Australia?, Index Education, March 2026, Exhibit 5, as at 31 December 2025: the S&P/ASX 200's largest constituent carried a weight of 10.15%, with the ten largest constituents' combined weight at 46.68%.
8. S&P Dow Jones Indices research on the S&P 500 Equal Weight Index documents its tilt toward smaller constituents, its anti-momentum characteristic, and its higher turnover relative to the capitalisation-weighted index.
9. RBC Wealth Management, The "Great Narrowing": S&P 500 Concentration, 22 January 2026, citing FactSet data as at 31 December 2025: the ten largest S&P 500 constituents represented approximately 40.7% of index market capitalisation and approximately 32% of expected 2025 earnings.
10. Calendar-year 2000 total returns of the S&P 500 (minus 9.1%) and the S&P 500 Equal Weight Index (plus 9.6%) per ClearBridge Investments, Is the S&P 500 Still a Diversified Index?, July 2024, citing FactSet; the equal-weight figure for 2000 is back-tested. Calendar-year 2023 total returns of 26.3% and 13.9% respectively per S&P Dow Jones Indices.
11. S&P Dow Jones Indices, S&P 500 Equal Weight Index methodology and index dashboard disclosures: the index launched on 8 January 2003, with history prior to launch calculated retrospectively on a back-tested basis.
