Load-bearing claims in this essay passed 32 groups × 3 refute-by-default adversarial votes (96 votes: 1 group clean, 3 mostly clean, 28 with calibration corrections, 0 overturned), with contradiction-search and methods-audit seats added for 4 single-source empirics (8 seats). The audit result is the most damaging in this site's run history: not one of the four can bear the weight at which it is routinely cited — one vetoed, three restricted, two of them direction-contested — while the mechanism layer of two was upgraded to multi-source. Verification produced 130+ calibration corrections, eight of which reversed this essay's own original wording; all eight are in the text and the appendix, not buried in footnotes. This run also caught three fabrications at the web-scraping summarizer layer (invented quotations for one paper, an invented statistic for another, an invented model description for a third) plus one search-echo contamination; every quotation here was read verbatim from the primary document. Evidence grades: [multi-source] ≥2 independent sources agreeing; [single-source verified] traceable to a primary document; [direction contested] independent sources conflict; [interested party] self-reported by a party with a stake — direction usable, magnitude not load-bearing; [unverified]. All figures as of August 2026. This is an evidence review, not investment advice.
The most-quoted "fact" in this debate is that passive investing has crossed half the market. It is true — on one particular ruler, and even then the crossing year depends on whose ruler you pick.
Ruler one: fund assets. Morningstar's data (reported by CNBC in January 2024) put US long-term funds — mutual funds plus ETFs, excluding money market funds and funds of funds — at $13.29 trillion passive versus $13.23 trillion active at the end of 2023. A $60 billion margin on a $26.5 trillion base is 0.2%, inside any plausible classification error [multi-source: Morningstar states the year three separate times, CNBC carries the pair]. Under the Investment Company Institute's own definition, index funds were only 48% of long-term fund assets at end-2023 and crossed in 2024 — so when State Street wrote in January 2026 that "in 2024, the AUM of passive funds surpassed that of active funds," it was not wrong; it was using the other ruler [multi-source]. Earlier still: for US equity funds alone, the crossover came at the end of August 2019, when Morningstar's preliminary figures put passive at 50.15% [single-source verified, Morningstar's August 2019 commentary]. Latest readings: ICI monthly for June 2026 shows index funds at $21.88 trillion against active at $18.83 trillion, or 53.7%; Morningstar's ruler puts year-end 2025 at roughly 55% — its own report prints "55%," "over 55%," and chart labels that compute to 54.8%, an internal inconsistency recorded here as found.
Ruler two: market capitalization. Switch to "what share of US stock market value do passive funds own," and the number collapses to about a fifth: ICI says 19% at year-end 2025 (numerator: index domestic equity funds only; the series runs 11% in 2015 to 19% in 2025, and it dipped in 2020); Vanguard, using a wider numerator that counts worldwide funds' US-equity allocations, says 23% (end-2024); Federal Reserve staff estimated 14% as of March 2020 [multi-source in direction, but the three numerators and denominators differ and must not be strung into one series]. Note that all three "small" figures declare themselves lower bounds — the Fed paper's very next sentence cites BlackRock's estimate that passive investors owned 18% of global equity at end-2016, "with most of the holdings outside the MF and ETF sector."
How big is shadow passive? This is the widest disagreement in the debate, and the one direct measurement cannot carry it alone. Chinco and Sammon (Journal of Financial Economics, July 2024) invert reconstitution-day trading volume to estimate total US passive ownership — including internally managed institutional indexing and closet indexing — and what they produce is not a number but a family: between 13% and 34% of market cap in 2021, depending on which slice of reconstitution-day volume you count as passive rebalancing. The widely quoted 33.5% is the loosest member (all-day volume); the authors themselves call the after-4pm measure (24.62%) "our preferred intraday measure"; and the tightest — closing-auction volume only — is 13.31% (±1.96%), which the authors state in print is statistically indistinguishable from the ICI's fund-based 16%, with a lower point estimate.
The methods-audit seat restricted this item for reasons that go further: the paper's entire identification argument (passive rebalancing is prearranged and prints at the close) supports at most 24.62%, while the extra 8.88 percentage points in the headline are intraday volume — precisely the volume the mechanism says passive does not generate; three of the five indexes reconstitute on triple-witching days and contribute 65% of the headline, and the authors' own 2022 working paper excluded such indexes for exactly that reason ("we would tend to overstate the passive-ownership share... we are trying to be as conservative as possible"); and the series does not behave like an ownership stock (31.85% in 2009, 24.32% in 2010, 23.21% in 2013) but like reconstitution-day trading intensity, which rises and falls with volatility.
The contradiction-search seat, however, rescued half of it: the direction — that true passive ownership substantially exceeds the fund-based number — is multi-source, corroborated by four independent teams on four different datasets. The most powerful bypasses volume inference entirely: Escobar and coauthors (2026) use transaction-level records from the Colombian exchange with investor-type identifiers to observe directly who trades MSCI index rebalancings, and find that "non-passive benchmarked investors generate rebalancing-driven inflows that are 1.7 times larger than those of explicitly passive funds" — a total-to-explicit multiplier of 2.72×, higher than Chinco-Sammon's 2×. S&P Dow Jones's own asset survey shows roughly $13 trillion indexed to the S&P 500 alone at end-2024, about 20.6% of a ~$63 trillion US market — one index family exceeding ICI's count of all index funds. And BlackRock's 2017 white paper, written to argue that indexing is small (its own headline: under 20% of global equities), broke out global indexing at 17.5% including 7.9 percentage points of institutional indexing and 2.1 of internal indexing, against a fund-only 7.4% — a 2.36× multiplier from a party biased toward a small number.
So this essay's usage is: direction multi-source, level single-source. Independent measurements of the same quantity span 17.5% to 41%; 33.5% sits inside that range with no second team pinning it. Cite the range, not the point. One trap to avoid en route: Chinco and Sammon's text states that "a 2017 BlackRock report put the passive-ownership share at 25.6%" — checked line by line against the original, 25.6% is the active share in that exhibit; the index share is 17.5%. It does not touch their own estimate, but "BlackRock said 25.6%" must not be repeated as corroboration.
There is an irony worth noting: this paper is no friendlier to the explosive-growth narrative. On its own ruler, passive was already 13.4% in 2000 — a 2.5× rise over 21 years, against roughly 5× on the fund-based ruler. "The share is understated" and "the growth is a regime change" cannot both be supported by the same paper.
Ruler three: trading volume. Vanguard (August 2025) puts index funds' portfolio-management trading at 1.2% of US exchange dollar volume in 2024 (active funds about 2%) — the denominator is all Cboe volume, the numerator deliberately counts only funds' own rebalancing. ICI puts ETF secondary-market trading at 28% of US stock trading (2025) — but that is trading in ETF shares, most of which nets out without touching constituents, and it includes active ETFs. State Street reports the closing auction rising from an 8.40% average since 2018 to 15.39% in December 2025 — with no passive attribution, and noting that active managers migrating to the close push it up too. Michael Green's estimate, given to Institutional Investor, is 20% passive direct trading plus 8% index derivatives plus 36% "passive market making" — about 64%, though that sum is ours; the article never prints it, the 36% component has no primary source, and it may double-count against the 20%. Vanguard's reply in the same article puts all-in indexing under 5%. From 1.2% to 64% is not a measurement dispute; these are four different objects — the bubble thesis needs the large framings, the defense needs the small one, and neither side reports the other's.
These three rulers are the floor of everything that follows. For each piece of evidence below, ask first: which ruler, and which segment of it?
"Passive is breaking markets" is not a paper; it is a set of people. Putting each one's exact words, offered evidence, commercial position, and subsequent track record side by side is the most basic physical this kind of narrative can get — and the result is more damaging than expected, in both directions.
On September 3, 2019, Burry answered Bloomberg by email (published September 4, with an editor's note that comments were "lightly edited and condensed"): "And now passive investing has removed price discovery from the equity markets... This is very much like the bubble in synthetic asset-backed CDOs before the Great Financial Crisis in that price-setting in that market was not done by fundamental security-level analysis, but by massive capital flows based on Nobel-approved models of risk that proved to be untrue." His testable content was a liquidity count: that day, 1,049 Russell 2000 stocks traded under $5 million ("over half"), 456 of those under $1 million; 266 S&P 500 stocks traded under $150 million. "The theater keeps getting more crowded, but the exit door is the same as it always was." He also wrote: "However, I just don't know what the timeline will be" [single-source verified, verbatim].
The track record: from the publication-day close of 2,937.78, the S&P 500 reached 7,798.99 on August 13, 2026 — a record close, +165.5% on price alone over roughly seven years [multi-source, FRED]. The small-cap exit stampede he described did not appear in any of the three stress tests since (March 2020, August 2024, April 2025). Scion, which managed about $340 million then, was down to $155 million by March 2025 and formally deregistered with the SEC on November 10, 2025. Burry now writes a paid Substack ("Cassandra Unchained," $379/year) whose main theme is an AI bubble — hyperscaler depreciation accounting, not index funds. Across 79 published posts there is not one restating the 2019 indexing thesis (most are paywalled; this describes the visible record) [multi-source]. A contemporaneous rebuttal is also on the record: Elm Wealth's Haghani and White pointed out in October 2019 that the thesis undercuts itself — if passive flows really did distort small-cap prices, that is an opportunity for active managers, not a catastrophe [single-source verified].
In February 2024, Einhorn delivered this cycle's sharpest sentence on Bloomberg's Masters in Business: "I view the markets as fundamentally broken... value is just not a consideration for most investment money that's out there... almost the best way to get your stock to go up is to start by being overvalued." His mechanism is a value-manager redemption spiral: money moves from active to passive, value managers get redeemed, value stocks fall, more redemptions follow.
In the same interview, he supplied the other half: "it's actually much more exciting now... there's way fewer people competing with us for ideas... the opportunity is actually probably as good or better than it's ever been." Hence buying at four to five times earnings with 15-20% cash flow returned via buybacks, rather than waiting for anyone to re-rate him. By CNBC's Delivering Alpha in November 2024, the phrasing had softened to "a bit of a broken market" [single-source verified, machine transcript, verbatim]. "The market is broken" and "the opportunity is the best ever" are, in the Grossman-Stiglitz framework, the same sentence (Section 2) — Einhorn's testimony reads more like an illustration of that framework than a counterexample to it. He credits the whole framework to Michael Green, so he is not independent corroboration [single-source verified].
Green is the theorist of this position and the only person offering a falsifiable timetable. On the Excess Returns podcast of January 20, 2026, he said that at roughly 83% passive "the market becomes so impossibly volatile that it becomes an inevitable event that it will eventually cause its closure," estimated that "we're about 54% passive by market share and we picked up about 4% last year," and concluded "somewhere in the neighborhood of five years out the world comes to an end." Checked against primaries [single-source verified throughout]:
Commercial position: in August 2026 Green announced he was leaving Simplify to found Tier1 Alpha Asset Management as CEO and CIO — a strategy that measures and trades the price pressure of passive flows. The warner turned the thesis into a product, and in a two-way conflict check that belongs on the record beside Vanguard's [single-source verified, his own announcement].
What must be recorded with equal care: Green's directional call — "until this stops, you can't beat the indices... markets will continue to go up" — has been right. He is not a crash-caller in the Burry mode but a "this keeps working until it doesn't" caller, so the 2019-2026 melt-up is consistent with his view rather than a refutation of it. That makes his thesis harder to falsify and means rising markets are not evidence against him.
Fundsmith's semi-annual letter of July 2026 is the warning side's most recent and most personally expensive text: the market is "dominated by so-called passive or index funds (of which the majority are ETFs)" and the AI boom, "produc[ing] a market dominated by momentum rather than any fundamental factors"; momentum's relative performance "is at a 30 year high and more extreme than in late 1999 just before the Dotcom bubble burst"; "I profess no insight into how or when this passive-led momentum market will end, other than to say badly." The cost: the fund fell 2.9% in H1 2026 against its comparator's +11.2%, a 14.1 percentage point shortfall (the letter's own figure), with portfolio turnover at 51.8% for the half (against 3.2% for all of 2024) and an explicit decision to incorporate momentum into the process [single-source verified, primary letter].
The physical cuts both ways here too. In November 2025 he told Expansión that passive is "more than 50% of assets" and that in 2000 it was "less than 10% of funds" — switching denominators inside one exchange; eight months later the letter says "AUM in index funds is now >60%." None of the three figures carries a source. The "30-year high" rests on a Bloomberg chart reproduced in the letter (US momentum index versus US market, quarter-over-quarter, series starting 1995) — his characterization of a chart, not an independently verifiable statistic. The letter also contains hard errors: "In 2007-08, the S&P fell 57% in five months" (it was roughly 17 months, October 2007 to March 2009); "index funds outperformed active funds on average by 24%" is 24 percentage points, and sets an index against an equal-weighted fund average carrying survivorship exposure. And the letter's own five largest detractors — LVMH, Zoetis, IDEXX, Microsoft, Coloplast — are stock-specific calls, which sits in visible tension with attributing 14.1 points to indexation [single-source verified].
The authority the warning side cites most stands on the other side. Bogle's November 2018 Wall Street Journal op-ed warns about ownership concentration — "If historical trends continue, a handful of giant institutional investors will one day hold voting control of virtually every large U.S. corporation" — contains no claim about price discovery or bubbles, and closes by defending index funds, quoting Coates: "'A ban on indexing would clearly not be a good idea.' I can only say, 'Amen' to those words." At Berkshire's 2017 meeting he did say "If everybody indexed, the only word you could use is chaos, catastrophe" — the very next sentence is "Now, what are the chances of everybody indexing? It's zero," and later in the same answer, "So, I'm not concerned about it." A reporter paraphrased him as expecting indexing "could exceed 75% of the market and not have it become dangerous." Where did 75% come from? Asness left the provenance in a 2024 footnote: he pressed Bogle on it and was told, "I made it up" [multi-source: original op-ed, interview, and Asness footnote 40, all verbatim].
The other camp is not clean either. Vanguard's 1.2% comes from the largest beneficiary of debunking; SPIVA is published by an index provider; the flagship material for "bond ETFs provided price discovery in the crisis" is a BlackRock conference deck — hosted on the SEC's website, and therefore routinely mistaken for an SEC finding.
Three voices refuse to read from their camp's script, and they are the most useful anchors in the whole file. Asness (AQR) concedes that markets have become less efficient but judges indexing "a relatively small part of the story," with social media and gamified trading the larger suspects. Lamont (Acadian) is an active manager — the party the bubble narrative would excuse — and writes the sharpest rebuttal of the cycle: "Since passive investors mostly do not trade, it is mostly true that they do not impact market prices... As we say in the homicide squad, 'no body, no crime,' and here we have 'no trade, no crime'"; and "There is no magic number, other than 100%, at which the market suddenly flips to being dysfunctional due to passive ownership... What matters is who stays." GMO, a famous bear, judges passive's impact "overblown but not negligible" and argues future impact will be smaller because the defined-benefit-to-defined-contribution transition that drove the flows is ending — the same demographic fact Green uses, with the opposite conclusion [multi-source, all verbatim].
The 1980 paper has one core claim: "an equilibrium degree of disequilibrium" — prices reflect informed traders' information only partially, so that those who pay for information earn a return on it. The much-cited corollary — more passive means bigger mispricings, so the remaining active managers earn more, so the system self-corrects — is the contrapositive of the paper's equilibrium condition, and the paper never writes it (what it writes is the other direction: the more informed there are, the lower the relative payoff to being informed). The inference is legitimate; it is an inference, not a quotation. Two misuse warnings matter more. The famous non-existence result — no competitive equilibrium — is driven by noise going to zero (prices becoming too informative), which is the mirror image of "everyone indexes so nobody is informed"; invoking it for a 100%-passive breakdown is a misapplication. And the paper's actual prediction for the high-passive corner is Conjecture 7: markets become thin — a volume claim, not a mispricing claim. The authors also state plainly that outside the CARA-normal example, "we have not been able to obtain a general proof of any of these propositions" [single-source verified, verbatim, 3/3 votes clean].
Gârleanu and Pedersen (Review of Asset Pricing Studies 2022) embed asset managers in the Grossman-Stiglitz world. Their Proposition 7: when the cost of passive investing falls, overall mispricing rises, and macro (index-level) mispricing rises more. Their Proposition 6: when information costs fall, mispricing falls, and macro mispricing falls more. Both costs have collapsed over recent decades (the authors associate the information-cost decline with the 1980s-90s and the passive-cost decline with the 2000s), so quoting only Proposition 7 is selective; the net prediction under both forces is indeterminate — that synthesis is ours, not the paper's sentence. This is the strongest peer-reviewed theory card the bubble thesis holds, and one author is a principal at AQR (disclosed; directionally convenient for active management). Bond and García (Review of Financial Studies 2022) reverse Grossman-Stiglitz's strategic structure: model participation rather than information acquisition and indexing becomes a strategic complement — more indexing makes indexing more attractive, with no restoring force. Yet in the same model indexing raises welfare, including for the least informed, and relative price efficiency across individual stocks improves. Self-correction is a property of one modeling choice, not a theorem [single-source verified]. A counterintuitive footnote: in Chabakauri and Rytchkov's dynamic model (JFE 2021), indexing lowers market volatility — and this run caught a web summarizer fabricating quotations attributing the opposite conclusion to that paper.
This number, treated by Green and much of the press as the academic bedrock, comes from Gabaix and Koijen's Inelastic Markets Hypothesis. Verbatim checking places it as follows: (a) it estimates the multiplier for aggregate flows between stocks and bonds on the whole market's valuation, not passive versus active — across 123 pages there is exactly one substantive mention of passive holding, a paraphrase of De Long et al. (1990); (b) the point-estimate family is 5.28 / 7.08 / 7.69 / 10.93 with standard errors of 1.10-2.64, and the abstract's "about $5" is the conservative end, with the paper's own stated range being "$3 to $8"; (c) five years after first circulation it remains unrefereed (the companion methods paper cleared the Journal of Political Economy in 2024; the application did not); (d) the 2024-26 elasticity literature is splitting — van der Beck (JFE 2026) finds stock-level multipliers of about 3 in the short run decaying to about 1 in the long run (a 3-4× ratio), though that is a micro object and cannot simply be divided into Gabaix-Koijen's macro multiplier, while Gabaix-Koijen's own horizon estimates show aggregate shocks "fairly stable" over four quarters. Whether the aggregate multiplier is large-but-transitory or large-and-persistent is unsettled; the Hartley-Gerding rebuttal paper could not be retrieved at all, so only the authors' public description is on the record [multi-source verification; grade: the GK multiplier is unrefereed single-source, its persistence direction-contested].
The micro layer has the number people reach for most. Haddad, Huebner and Loualiche (American Economic Review, March 2025) estimate from 13F holdings that when peers become less price-sensitive, active investors trade more aggressively — but this strategic response offsets only two-thirds of the initial change, so the rise of passive over 20 years made demand for individual stocks about 11% more inelastic. This is usually presented as the cleanest causal magnitude in the debate. Both audit seats disagreed, in different ways.
Start with magnitude. The range is wide: the fund-based version in the same paper, on the same page, gives 6.3%; Behmaram (2026), running a parallel counterfactual in the same framework, implies about 43%. 11% is the middle-to-low member, and it is the one everyone quotes.
What the methods-audit seat found is more fundamental: the number rests entirely on a model-generated instrument. The paper's own robustness table has two rows: drop the instrument for aggregate elasticity and the strategic-response parameter falls from 2.97 to 0.04 (standard error 0.22) — a statistical zero. At that value the pass-through is 97%, the same drop in active share yields about 31% more inelastic (nearly triple the headline), and the sentence "competition offsets two-thirds" evaporates entirely. The authors read those rows as showing that correcting endogeneity matters; a hostile referee reads a 75-fold OLS-versus-IV gap with no external instrument to arbitrate — because the instrument is not an external design but a quantity the model itself solves for under a counterfactual portfolio, containing the very parameters being estimated.
One more thing must be stated plainly: "passive" here is model-defined, and the sign of the result follows from the definition. An institution counts as passive if and only if its estimated elasticity falls below a threshold, set at 0.06 because it "calibrate[s] the level of passive investing"; passive investors are then assigned elasticity exactly zero; and the dependent variable is a position-weighted average of those same individual elasticities. "More passive means less elastic" is therefore true by construction, and the only empirical content is the size of the offset. That 0.06 threshold has no sensitivity analysis anywhere in the paper or its appendix — the entire 19%-to-41% passive series, the 32% log decline, and the headline magnitude all float on an untested tuning parameter.
Version drift is nontrivial too: the 2022 working paper said 15% (parameter 2.15), the AER version says 11% (parameter 2.97), on the same 2001-2020 13F data — a 27% move from re-estimation alone, with 2.15 falling outside the published version's own robustness band.
And one funding relationship must be disclosed: the lead author acknowledges support from UCLA Anderson's Fink Center for Finance and Investments, endowed in 2008 by BlackRock founder and CEO Laurence Fink. This is not an indictment of the work — the paper's other headline ("the stock market is far from the competitive ideal") does not flatter passive giants, and both camps cite it — but the sentence most often deployed in passive's defense, "competition offsets two-thirds," comes from a paper funded by a center endowed by the largest passive manager's founder, and that deserves the same label as Vanguard's 1.2%.
There is also an object mismatch that matters for how the number is used. HHL estimate a quarterly 13F demand slope — a model-internal object. The sentence it gets deployed to support, "passive changed how prices respond to trading," has been measured directly: Greenwood and Sammon (Journal of Finance 2025) measure how much index-driven trading actually moves prices, and find the multiplier fell from 6.75 in the late 1990s to 0.37 in 2010-2020 — a factor of 7 to 20, over an overlapping window with independent data. One says the demand curve steepened 11%; the other says the ability of the same index buying to move prices collapsed by an order of magnitude. They can technically coexist (a slope and a realized impact are not the same quantity), but if the claim is "buying pressure moves prices more now," HHL is the wrong evidence and the right evidence points the other way [direction contested, contradiction-search seat].
Three calibration points to nail down: the object is relative pricing across individual stocks (footnote 37 explicitly distinguishes it from aggregate market elasticity, for which "neither our model nor any other empirical model to our knowledge" applies); HHL's own text concedes that Coles, Heath and Ringgenberg (JFE 2022) "show a full degree of strategic substitution" (pass-through near zero); and in the paper's own decomposition, measured aggregate elasticity fell 0.41 to 0.27 (about 34%), of which passive explains roughly 11 percentage points, with a separate investor-specific trend driving the rest — passive is a minority contributor on the paper's own accounting, and that measured series is non-monotonic while passive growth was monotonic.
So what survives? The audit seat's recommendation: cite the fraction, not the total. The pass-through of about one-third (competition offsetting about two-thirds) is the paper's real contribution and is roughly corroborated by independent work (Behmaram's independent reduced-form bridge implies about 27% against HHL's 33%). The 11% is that fraction multiplied by a contested measurement of how much passive grew.
Another paper from the same demand-system machinery cuts the other way: Koijen, Richmond and Yogo (Review of Economic Studies 2024) find that the active-to-passive transition had a large effect on prices (14% cross-sectional repricing in a 2007-to-2016 counterfactual — a sum of absolute changes, not a market-level move) but a small effect on price informativeness, because capital did not on average flow from better-informed to less-informed investors. Koijen sits in both chairs of this debate: the inelasticity paper is cited by the bubble camp, the informativeness paper by the defense.
"Price discovery" in this literature covers at least six non-substitutable objects. Lay them out and each gives a different answer.
Bai, Philippon and Savov (JFE 2016): the power of prices to forecast S&P 500 nonfinancial firms' earnings three and five years out rose about 60% and 80% respectively on a 1960-2014 trend regression — markets became more informative. But footnote 2 of the same paper: outside the S&P 500, informativeness "appears to decline" (concentrated at short horizons), and the authors attribute it to dramatic compositional change in the listed universe while warning the comparison is "potentially misleading." And footnote 7: the paper's own earlier drafts (the 2013 NBER version, and a 2012 New York Fed staff report before it) concluded informativeness had remained stable — the conclusion reversed after three specification choices (inflation adjustment, delisting correction, adding the five-year horizon). People who cite "BPS proved markets got more efficient" have generally not read the footnotes. Farboodi, Matray, Veldkamp and Venkateswaran (RFS 2022) use the same measure and find the split: +70% at the five-year horizon for the S&P 500 sample, −80% for non-S&P firms; the rise is confined to firms that are large and high-growth; and they find the trend is a size effect rather than an index-membership effect (matched non-members look the same) — which weakens the passive mechanism and points toward a data-economics one. The fatal clock: both papers' informativeness series end in 2010, before passive's main ascent, and cannot adjudicate what happened afterward. The methodological ground is shifting too: Dávila and Parlatore (RFS 2025) argue the BPS measure cannot separate information aggregation from fundamental volatility [multi-source, split presented as found].
Coles, Heath and Ringgenberg (JFE, lead article, September 2022) use the Russell 1000/2000 reconstitution to get exogenous variation in passive ownership (about 1.5 percentage points of market cap) — the cleanest causal experiment available. The results come in two halves, and citing either without the other misrepresents the paper. The zero half: information production falls (EDGAR page views −24.8 log points and analyst reports −10.8 on entering the Russell 2000, both only p<0.10; the reverse panel is better powered), but the three families of price-efficiency measures — 4- and 8-day variance ratios, an 11-anomaly mispricing score, post-earnings drift — are precisely estimated zeros, with a minimum detectable effect of roughly one-sixth of a standard deviation; and in the aggregate time series, index-fund ownership of Russell 3000 stocks went from 2% to 11% of market cap between 2007 and 2016 while the variance ratio improved from 1.16 to 1.10. The non-zero half: volatility (weakly), comovement with both indexes, market beta, turnover (+12.1%) and short interest (+1.9 percentage points of shares outstanding, about +19% relative) are all significantly positive — people citing this paper as "passive is harmless" omit Section D. Three honest annotations: the paper's own introduction prints "EDGAR page views fall by 14.1%," which does not match its own table's −24.8 log points (this essay originally accused citers of misquoting; verification reversed it — the inconsistency is internal to the paper); the 2017 working-paper version of this same paper concluded the opposite ("index investing introduces noise into stock prices," variance ratios worsening by about 30%), with the published version being the revised result; and the authors state their local-average-treatment-effect has unknown external validity, then immediately add that their model's predictions "hold for any feasible level of passive investing." Testifying before Congress in June 2026, Ringgenberg restated the finding without retraction: information production falls, overall price efficiency does not [single-source verified by design uniqueness, both halves presented].
Sammon (Management Science 2025) measures a different link in the chain: how much of an earnings report is already in the price before it is published. Four measures deteriorate over 1990-2019: the absolute earnings-day return rises from about 2% to 4%; the price-jump ratio (the share of the total move that happens after the release) from 0.2 to 0.5. In the cross-section, firms at the 90th percentile of passive ownership in 2019 (27%) show earnings-day absolute returns 152 basis points higher than those at the 10th (3%).
This essay's original draft called it "the most solid evidence that price discovery deteriorated." The methods-audit seat vetoed that framing — and note that this seat ran twice, returning "restricted use" the first time and "veto" the second; the seat's own instability is recorded here, and the stricter verdict adopted. The reasons are checkable:
So this grid splits in two. What can bear weight is the descriptive fact: the share of earnings-window return variation occurring after the announcement rose substantially between 1990 and 2019 — independently corroborated by Beaver, McNichols and Wang. What cannot is the attribution to passive ownership, or any quantified magnitude.
The mechanism layer, though, is multi-source: "higher passive ownership, lower firm-specific/private information content in prices" is confirmed by three independent teams on three different measures, none of them using an earnings window. Höfler, Schlag and Schmeling (2025) decompose return variance and find that per standard deviation of passive ETF ownership the private-information share falls 10.2 (t = −3.42) and the noise share rises 6.09, while the market-information share is unchanged — precisely Sammon's hypothesized channel, measured with entirely different machinery. Bennett, Stulz and Wang find the same direction using GPIN around S&P 500 additions. Israeli, Lee and Sridharan (2017) find it using future earnings response coefficients.
But the same facts admit an opposite reading, which must be presented alongside. Martineau (Critical Finance Review, 1984-2019) independently documents Sammon's core descriptive fact — pre-announcement drift weakening over time — and reads the whole configuration as markets becoming more efficient: earnings-day prices now approximately a martingale, post-earnings-announcement drift dead since 2006 for all but microcaps, and announcement-day responses several times more sensitive to earnings surprises than before. The same facts, the opposite welfare sign. Two further timing counterpoints: Bennett-Stulz-Wang's own GPIN effect weakens significantly after 2008 — exactly when passive grew fastest; and Farboodi et al. (RFS 2022) find informativeness about future earnings rising for large growth firms, where passive ownership is most concentrated [direction contested].
The relationship with Coles-Heath-Ringgenberg needs stating precisely: they find significant post-earnings drift on average but a zero treatment effect from indexing; Sammon finds the timing of information entering prices shifting later. "Passive moved price discovery from anticipatory to reactive" is one possible reconciliation, but it is our synthesis, not either paper's conclusion — and Martineau's reading says the same shift may simply mean earnings reports have become more informative.
Laid on one ruler, Section 3 is more awkward than either side's marketing. The "harmless" side: CHR's precisely estimated zeros (three measure families, explicit minimum detectable effects) and KRY's small informativeness effect are real, reproducible results. The "damage" side: the two most-cited findings were both downgraded by the dual-seat audit — HHL's −11% elasticity ruled direction-contested (direct measurement of price response shows the multiplier collapsing by an order of magnitude), and Sammon's pre-announcement informativeness had its causal claim and magnitude vetoed (the "one quarter" is an equal-weighted composite that becomes one sixth under a different normalization, and the effect is absent in the high-passive data after 2012).
What is left standing on the damage side is a weaker and more honest sentence: the firm-specific information content of prices falls with passive ownership — supported by three independent teams on three different measures — while whether that constitutes "worse price discovery" has a live opposite reading (Martineau reads the same facts as earnings-day pricing becoming more accurate and drift disappearing). This is not fence-sitting: the two readings imply different welfare judgments, and the available measures cannot separate them.
The decisive shared fact is this: not one paper in this literature has an independent variable anywhere near 50%. Sammon's passive measure runs from under 1% to about 15% of market cap over 1990-2019; CHR's runs 2% to 11%; JVZ's reaches only 4.95% of index market cap by 2020; HHL's broad measure reaches 41%. The 50-55% figure is a fund-asset number; the empirical literature's independent variable is a market-cap number. Any sentence of the form "now that passive is over half, the research implies X" extrapolates 3-5× beyond sample — and CHR, Sammon and Höfler all say so explicitly. A 2026 working paper (Choi) suggests why extrapolation may fail in both directions: passive's efficiency effect appears to be a function of the level of active ownership — improving efficiency where active ownership is substantial, degrading it only where active is thin — which would mean the marginal harm is concentrated in the tail and worsens as passive expands [unverified, working paper, recorded as a lead only].
S&P 500 inclusion is the most direct litmus test for passive's mechanical impact. Greenwood and Sammon (Journal of Finance, April 2025, published Table 1): cumulative abnormal returns on addition fell from 3.42% in the 1980s, 7.39% in the 1990s and 5.16% in the 2000s to 0.83% in 2010-2020, statistically insignificant; deletions converged from −16.11% in the 1990s to −0.62%, also insignificant. This happened while funds tracking the S&P 500 grew from roughly zero to about 7% of market cap: for a typical 2020 addition, trackers had to buy 7.53% of shares outstanding (median), while total 13F institutional ownership moved only 0.65 percentage points — somebody was handing over the shares. Who? Not active mutual funds (their net demand is about zero); the paper concludes it is institutions outside the 13F mutual-fund universe — hedge funds, pension funds, endowments. The authors' explanation ranks migrations (additions that simultaneously exit the MidCap 400, netting the forced buying against forced selling) and a general rise in the market's capacity to provide liquidity, "in the spirit of Lo's adaptive markets hypothesis," with predictability playing a small role. Watch the composition: migrations as a share of additions were roughly flat (about 69% in the 1990s to 63% in the 2010s); what exploded was the migration share of deletions (5.88% to 87.36%). Direct additions still earned +5.4% in the 2010s — and that figure is driven by Tesla alone (+53.35% from announcement to effective date); excluding Tesla, the average 2020 inclusion effect was −3 basis points [single-source verified, published version; note the paper's abstract contradicts its own Table 1, saying deletions averaged +0.1% where the table says −0.62% — cite the table].
Then the out-of-sample evidence: in 2025, the effect came back. Coinbase, announced after the close on May 12, 2025, rose 23.97% the following session (about +23.3% market-adjusted); Robinhood rose 15.83% and AppLovin 11.59% on September 8, 2025. All three were direct additions, all outside the paper's 1980-2020 sample, and none has a peer-reviewed re-estimation. They do not overturn "the effect converged to zero," but they put a boundary marker on "disappeared": for direct additions with high retail attention, double-digit effects can return [multi-source, verified directly against market data].
"Index inclusion makes a stock move with the index" was once the flagship evidence that passive changes pricing (Barberis, Shleifer and Wurgler, 2005: 1988-2000 additions raised daily S&P beta by an average 0.214). Chen, Singal and Whitelaw (JFE 2016) removed its foundation: the bivariate specification is uninformative, and in a univariate difference-of-differences the excess comovement for 2001-2012 is −0.004 (t = −0.14) — note that raw beta still rises (+0.071, significant), it just no longer rises more than the non-S&P beta. The Fed staff survey adopted that verdict ("comovement effects have declined significantly since 2001"), though its own body text calls the evidence "mixed." Then in 2025, DeCoste (Global Finance Journal) reversed it back with a fuzzy regression discontinuity: significant positive excess comovement in the long term, with CSW's null attributed to heterogeneity between newly included and established members. Add the ECB's 2024 euro-area estimate (each additional percentage point of passive ownership associated with +0.005 correlation with the index — a panel fixed-effects estimate with an autoregressive term, where the instrumental-variables version is only a robustness check). The honest verdict on this line is [direction contested], and no US series extends past 2012 [multi-source verification].
Jiang, Vayanos and Zheng (RFS, December 2025) is the most credentialed evidence for "passive inflates megacaps": a one-standard-deviation increase in flows into S&P 500 index funds is associated with quarterly excess returns (relative to the index) rising monotonically with size — +0.145% for the top 200 (marginal), +0.528% for the top 50 (t = 3.66), +0.687% for the top 10. The methods-audit seat restricted this to associational statements and set a hard condition on its most popular number:
The widely circulated "25 years of passive flows raised the top 50 by 29%" is the authors' own linear cumulation of a contemporaneous coefficient, and their own calibrated model produces only 0-4% for the same object — the authors state in the text that the reduced-form estimate is "larger than in our calibration" and "can partly mean-revert." That number cannot bear weight, and where it appears it must carry the 0-4% band in the same sentence. The regression itself is contemporaneous OLS with no instrument; the paper's own results section says "is associated with," switching to "caused" only in the extrapolation sentence — so the correct phrasing is "quarters of index-fund inflows coincided with megacap outperformance," not "passive flows raised megacap prices."
Three further limits are hard: the sample is 1996Q2-2020Q4, and the entire Magnificent Seven era is out of sample (the window was never extended across four and a half years of revision), so it is a mechanism hypothesis that may extend to today, not evidence about today's concentration; the size gradient was never statistically tested (no interaction, no joint monotonicity test), the top-10 cell is the table's weakest (t = 2.46, adjusted R² = 0.049), and the top-200 endpoint is insignificant; and the passive measure is narrow — S&P 500-tracking index mutual funds plus three ETFs, a cumulative 4.95 percentage points of ownership-share change — which cannot be silently transferred to "passive investing" broadly. Two appendix tests often cited as corroboration are not independent: one is a near-algebraic restatement of the main result on the same 99 observations, the other is confounded with the turn-of-the-month calendar effect documented since 1987. Funding: the LSE's Paul Woolley Centre for the Study of Capital Market Dysfunctionality, which one author directs — academic funding, recorded at light weight [single-source verified, restricted use].
The contradiction-search seat produced a layered picture. The mechanism is multi-source: HHL find that the same investor behaves more inelastically in large stocks, with extremely low elasticities for the very largest (a size-elasticity gradient about 25% steeper than the fixed-elasticity benchmark); Behmaram (2026), using a holdings-based measure, finds the elasticity decline concentrated in the largest size quintile; and the cleanest natural experiment in the world sits in Japan, where the Bank of Japan's ETF purchases (Barbon and Gianinazzi, RAPS 2019; Charoenwong, Morck and Wiwattanakantang) did raise constituent prices without near-term reversal while lowering price informativeness and liquidity. But none of these is evidence for a megacap-specific effect: Japan's cross-sectional dose is Nikkei 225 purchase intensity relative to float, not firm size.
And the 29% has zero independent measurement, with the nearest independent evidence pointing the other way. Two out-of-sample sign failures are on the record: in 2022, ETF inflows exceeded $600 billion while the Magnificent Seven mean-reverted and the equal-weight index beat cap-weight by roughly 34% through the drawdown; in 2026 to date, US ETF inflows have reached $1.2 trillion (about 40% ahead of 2025's record pace) while the S&P 500 Equal Weight index beat cap-weight in 7 of 12 months and technology was the worst sector. One framing must also be deleted: "passive buys more of the megacaps" — cap-weighted index buying is proportional by construction, and Vanguard's FactSet data (Russell 3000, end-2025) shows index funds' proportional ownership peaking in small and mid caps, with the top 200 owned less than most mid-caps. JVZ's channel is elasticity and idiosyncratic risk, not dose; calling it dose gets the mechanism wrong. Lamont, in the Financial Analysts Journal (2026), calls the mechanism "certainly complicated and arguably far-fetched" — recorded as argument, not as a competing estimate.
The bubble thesis's other pillar is that passive structure will snap under stress. Across five repeatedly cited episodes, no official post-mortem attributes the cause to growth in passive ownership or index-fund flows — though two do name mechanisms in the ETF trading and arbitrage layer, and that distinction must be preserved:
One closing fact: the Federal Reserve, which published the canonical 2018 paper on passive-investing risk, mentions passive, index fund, indexing and ETF exactly zero times in its May 2026 Financial Stability Report [single-source verified by full-text search of the official PDF]. The most recent regulatory-side verdict comes from the FCA-commissioned survey of July 2025: "passive investing increases the volatility of returns... On the other hand, the literature has not settled on the question of passive investing's impact of price efficiency with many influential articles arguing for no impact at all." The calibrated magnitude it cites is worth copying down: shifting the active:passive split from 90:10 to 40:60 raises annualized volatility from 11.13% to 11.84% — a real effect and a rounding error at the same time [multi-source, official verdicts verbatim].
The testable form of the Grossman-Stiglitz corollary is: active money leaves, mispricing accumulates, the survivors earn more. The data decline to supply that script.
SPIVA (published by S&P Dow Jones Indices — an index provider; the interest label stays): as of June 2025, large-cap active funds underperformed over 10 years in 85.98% of cases and over 20 years in 91.03% (risk-adjusted is worse: 94.39% over 10 years); for all domestic funds against the S&P Composite 1500, 20-year underperformance is 93.81%. Across the 2001-2024 annual series, a majority of large-cap funds beat the index in only three years — 2005, 2007 and 2009 (two of them within rounding) — and none since 2010; the trend is essentially flat (OLS slope +0.235 points a year, r² = 0.025), with no correspondence to the passive ramp. Calendar 2025 came in at 79% underperforming, which S&P calls the fourth-worst in 25 years — while the first half of that same year was 54% and "on track for the best year since 2022." That within-year reversal is itself evidence that these annual figures are driven by megacap concentration episodes: 2025 was precisely the year when the top five stocks' contribution to index return fell linearly from 78% in Q1 to 11% in Q4 — concentration eased, and active still had its fourth-worst year [single-source verified, vendor, full series presented].
Two calibration corrections must travel with those numbers. Survivorship decomposition: of the 88.29% 15-year underperformance rate, 50.05 percentage points is "did not survive" (SPIVA scores death as underperformance — whether the dead funds were actually lagging while alive is exactly the disputed point), 38.24 points survived and lost, and among funds that did survive only 23.4% beat the index. Asset weighting: the 20-year annualized gap is −1.96 points equal-weighted but −1.25 asset-weighted — where the money actually sits did about a third better than the fund-count number implies. Morningstar's independent scoreboard (different method: the hurdle is a portfolio of real, investable passive funds) puts 10-year success rates at 21% for active funds overall, 8.1% for US large blend and 3.6% for large growth; on a like-for-like category comparison it is actually more lenient than SPIVA (whose large-cap core 10-year underperformance rate against the S&P 500 is 96.57%) — this essay's original draft claimed the opposite, and verification reversed it: a net-of-fee passive-fund hurdle is a lower bar than the index. Morningstar's own counterpoints are on the record too: over the past decade the average dollar beat the average fund in 17 of 20 categories, and the cheapest quintile succeeded 31% of the time against 15% for the priciest [multi-source, two vendors on both sides of the trade, corrected].
Persistence: SPIVA's scorecard (year-end 2024) shows 0.00% of top-quartile domestic funds from 2020 still in the top quartile four years later — but pure chance predicts only 0.25⁴ ≈ 0.39% (an expectation of 2 funds out of 512, with a 13.5% probability of observing zero), so "zero" is not statistically unusual. The informative test is top-half persistence: 4.21% against a random 6.25% — real but modest anti-persistence. The same report contains a counterexample: five-year peer-ranking top-quartile retention is 54.76% (against 25% by chance) — long-window peer persistence does exist. Bottom-quartile funds are merged or liquidated within five years at 24.89% against 6.81% for the top — death does skew toward losers, though this is a peer-ranking measure, not a benchmark one [single-source verified, vendor, baselines and counterexamples presented].
The active industry's counterattack arrived in May 2026: Cremers, Fulkerson and Riley (unpublished working paper, supported by the Investment Adviser Association's Active Managers Council; Cremers is an independent director at Ariel Investments and an independent academic advisor to Touchstone — and, in the other direction, S&P Dow Jones awarded his pro-indexing paper a $15,000 SPIVA prize in 2014; both interests recorded). Crediting exiting funds with their actual pre-exit performance, asset-weighting, and benchmarking against real passive funds moves 20-year underperformance from a replicated 92.04% to 55.31% of assets, with fixed income reversing (71.34% to 37.09%). Morningstar's Ptak hit back: nearly one fifth of the 37-point improvement in the investment-grade intermediate bond category came from replication error, and he concedes only one of the three adjustments (the benchmark — Morningstar's own barometer already makes two of the three). S&P's substantive reply is a scope argument: SPIVA measures the proportion of funds against a market benchmark, because the claim being tested is "beat the market." There is no referee here (the paper is unrefereed and S&P has issued no methodological response), but what is in dispute is magnitude, not direction: no version of anyone's numbers shows active outperformance as the norm [direction contested on magnitude, multi-source on direction].
So the paradox stands: passive multiplied, and active's life did not improve — it got worse after 2010 (Pástor-Stambaugh-Taylor's industry diseconomies imply that a shrinking active industry should have raised gross alpha; note their sample ends in 2011, that whether active mutual funds' share of market cap actually fell needs its own citation, and that the result has an unresolved exchange in Critical Finance Review). Two explanations are on the table: (i) the level story — passive ownership degrades price discovery, so mispricing is more abundant but harder to harvest; (ii) the flow story (Unterberg, 2026 working paper, all figures relayed via Morningstar) — persistent passive inflows apply price pressure against off-benchmark holdings, so what active managers own cheapens in relative terms; that mechanism switches itself off when inflows plateau, and its author explicitly says it is not skill deterioration and not Grossman-Stiglitz-style informational inefficiency. Which explanation holds determines whether the scoreboard paradox is evidence for the bubble thesis or a transitional friction [direction contested; secondhand status flagged].
Bebchuk and Hirst's ledger (law review, a different refereeing standard than economics): the Big Three's average combined stake in S&P 500 companies rose from 5.2% in 1998 to 20.5% in 2017 — note this includes their active assets (index funds were 66% of BlackRock's, 74% of Vanguard's, 79% of SSGA's) — and they cast an average 25.4% of all votes at 2017 annual meetings (the wedge arises because the Big Three vote nearly everything while other holders voted 68% in 2017). The famous projection — 34.3% of votes in ten years, 40.8% in twenty — extrapolates a constant 0.84% compounding annual decline in non-Big-Three holdings; the authors quote a BlackRock executive conceding the projections are "arithmetically correct" while disputing the growth assumption. Verification added two things: the 2022 update revised nothing (the 27.6%/33.4% pair often mistaken for a revised vote projection is the 2019 paper's ownership projection); and their own table shows median combined holdings plateauing at 21.5%, 21.1%, 21.9% across 2019-2021 — three years with no progress along a path that requires 27.6% by 2028 [single-source verified, including the paper's internal inconsistency: it prints both 24.9% and 27.5/27.6% for 2021 votes cast]. "The Big Three are the largest single shareholder in 88% of S&P 500 companies" is a law firm's rank statistic from December 2018 (combined stake 18.7%) — being first is not control, and the figure is nearly eight years old, yet still quoted as current.
The classic capacity indictment (Bebchuk-Hirst: BlackRock spends under 4 person-days and $5,000 a year per billion-dollar position) meets Kahan and Rock's reply: the right comparison is not a hypothetical concentrated owner but the funds' own economic owners — millions of retail investors whose own governance input was near zero — and they call the alternative benchmark a "category mistake." The 2025-26 updates cut both ways: BlackRock's stewardship team grew from 45 to 60+ (plus 13 in a separate active stewardship team), though assets grew faster; in the 2024-25 proxy year its stewardship arm voted 152,000+ proposals at 12,700+ companies, supporting management about 89%, and engaged 1,950 companies (about 15% by count — but roughly 70% by client asset value). In the 2026 proxy season, each of the three split stewardship into two independent teams (BlackRock's BIS and BAIS — the latter already existing since early 2025; Vanguard's VCM and VPM; State Street's main team plus an opt-in service), with one motive reported as keeping a single team below the 5% filing threshold. But EDGAR filings show this is not a uniform design (Vanguard split to 4.96%/4.94% at Macy's; BlackRock still filed 13.9% as a single entity at NETSTREIT), the splits are highly asymmetric (BIS covers about 90% of assets), and the practitioner observation is that "both teams often vote in parallel" — the legal form has split, the voting behavior has not. Pass-through voting tells the same story: BlackRock's eligible assets are $3.76 trillion with roughly $885 billion enrolled (end-2025); Vanguard had 82,000 participating investors in 2025, which on its own metrics is about 10% of direct-brokerage clients in participating funds and 3.4% by assets — the "this is already fixed" narrative uses the eligibility number, not the participation number, and the practitioner verdict is that pass-through voting "has not meaningfully impacted stewardship voting" [multi-source, both directions].
"Common ownership softens competition" was once the loudest empirical claim on the antitrust side: Azar, Schmalz and Tecu (Journal of Finance 2018) reported US airline ticket prices 3-7% higher than under separate ownership (panel), or 10-12% using the BlackRock-BGI merger as an instrument (the merger's direct effect was only about 0.5%); the published abstract contains no price percentage at all. A failed replication appeared in the same journal (Dennis, Gerardi and Schenone, 2022): the correlation loads on the market-share half of the measure and is sensitive to how investor control and bankruptcy are treated. Everything since — a "refutation" and a "surrebuttal" — remains in working-paper limbo. The one peer-reviewed survey (Gerardi, Lowry and Schenone, Annual Review of Financial Economics 2024 — note two of the three authors wrote the replication critique) concludes: "Across the newest papers employing the most credible identification techniques, there is relatively little evidence that common ownership causes lower competition. However, further research is necessary before broad conclusions can be reached." The structural blow comes from Backus, Conlon and Sinkinson (AEJ: Micro 2021): between a third and a half of the variation in the profit-weight measure comes not from overlapping ownership but from relative investor concentration, and taking the theory literally implies more than 10% of S&P 500 firms were tunneling by 2017 — which they treat as a reductio. The best twist: Azar and Vives' own 2022 empirical work finds that Big Three common ownership is associated with lower airline prices, with higher prices associated with non-Big-Three shareholders [direction contested, full chain presented].
Regulatory reality (2025-26): the 13-state suit alleging the Big Three used their holdings to coordinate a reduction in coal output survived dismissal in part on August 1, 2025 (the core antitrust counts stand; the judge noted that inferring conspiracy from purely circumstantial evidence was "a close call"); Vanguard settled in February 2026 for $29.5 million with commitments to expand pass-through voting and abstain from climate alliances; and the FTC/DOJ statement of interest permits no over-reading in either direction — the same document reaffirms the 2017 position that broad common-ownership limits lack compelling evidence, argues that using horizontal shareholdings to suppress output is actionable, and expressly declines to take a position on acquisition-only theories [multi-source, official filings].
If "the power of indexation" has a location best supported by evidence, it is not the fund layer but the index construction layer. The SEC's own 2022 request for comment states that three index providers account for over two-thirds of the index market (industry-wide revenue of about $5.0 billion in 2021 — note that is the whole industry, with the top three taking roughly $3.4 billion), and that constructing an index "leave[s] room for significant discretion," sometimes "without publicly disclosing their index methodologies or rules." The academic pricing: An, Benetton and Song (JFE 2023) find that over a third of ETF expense ratios is paid to index providers as licensing fees (SPY's 9 basis points include 3 basis points plus $600,000 a year to S&P Dow Jones), and a structural counterfactual suggests improving competition could cut expense ratios "by up to 30%" — but the same paper finds the market cannot heal itself: entry by a new low-fee index provider leaves equilibrium fees almost unaffected (with Morningstar's 2016 Open Indexes Project as real-world confirmation). The regulatory ending: that request for comment never became a proposed rule and was formally withdrawn from the regulatory agenda on April 21, 2025. US index providers remain outside the Advisers Act while EU and UK benchmark administrators are registered under the Benchmarks Regulation — a live jurisdictional asymmetry [multi-source plus official documents].
The symmetric fact — that "passive" funds do exercise discretion — is also on the record: Molk and Robertson (Yale Journal on Regulation 2024) measure round-trip deviations from the index of 1.7-7.5% of assets for the largest S&P 500 funds in Q4 2022 (0.85-3.75% one-sided), about $61.5 billion round-trip across all S&P 500 funds — but 98.9% of fund-quarters show deviations under 20%, so at the megafund level this looks more like tracking management than hidden activism. The paper cuts both camps: index funds are neither mindless price-takers nor literally passive — the party actually being delegated to is the index provider, which is Robertson's older "Passive in Name Only" point [single-source verified].
Putting all of it back on the three rulers, here is a verdict that flatters nobody:
What stands: (i) [multi-source] passive is past half on the fund-asset ruler and about a fifth on the market-cap ruler; the direction that true passive exceeds the fund-based number is multi-source (independent measurements 17.5%-41%), though no specific level is pinned by a second team; (ii) [multi-source] active large-cap long-horizon underperformance is robust, softened but not reversed on an asset-weighted basis; (iii) [multi-source] there is rough consensus on a volatility channel, with a calibrated magnitude under one percentage point annualized; (iv) [multi-source] the firm-specific information content of prices falls with passive ownership (three independent teams, three measures); (v) [multi-source] index-linked, price-insensitive demand does raise prices, with the Bank of Japan's ETF program as the cleanest natural experiment; (vi) [multi-source] the governance concentration numbers are real, and the index-provider layer's rents and discretion are the best-evidenced part of the whole file.
What does not stand: (i) "price discovery has been removed" — CHR's precise zeros, the index effect converging to zero while passive doubled, and KRY's small informativeness effect all point the other way; (ii) "the index bubble will burst" — +165% over seven years, no official post-mortem attributing anything to passive ownership, and the warning side's own model unable to produce a consistent threshold (83% versus 65%/90%); (iii) "Bogle warned about this too" — he warned about governance concentration and defended index funds; (iv) "passive is more than half of trading volume" — the largest component (36% "passive market making") has no primary source.
What is unsettled: (i) [direction contested] whether passive made demand less elastic — HHL's structural estimate says yes (about 11%, though the same paper's fund-based measure gives 6.3%), but that magnitude rests entirely on a model-generated instrument that, when removed, drives the strategic-response parameter to zero and eliminates the "competition offsets two-thirds" claim; meanwhile direct measurement of index trading's price impact shows it collapsing from 6.75 to 0.37; (ii) [direction contested] whether "less firm-specific information" equals worse price discovery — Martineau reads the same facts as markets becoming more efficient; (iii) [direction contested] inclusion-driven excess comovement (CSW's zero versus DeCoste's 2025 positive); (iv) [direction contested] the magnitude of active underperformance (fund-count versus asset-weighted), though not its direction; (v) [vetoed] the quantified claim that passive ownership cut pre-announcement informativeness by about a quarter — a composite that becomes a sixth under a different normalization, and insignificant in the high-passive data after 2012.
The real blanks (these matter more than either camp's talking points): no peer-reviewed empirical work covers a world where passive exceeds 20% of market cap; no re-estimation of the index effect on 2021-2026 data exists, while three double-digit revivals from 2025 sit unexamined; the comovement series stops in 2012; there is no data on voting divergence after the Big Three's stewardship splits; and the shadow-passive measurement dispute (13.31% versus 33.5%) has no third-party arbitration — that paper has no independent replication and no published comment to date. The structural asymmetry of this field is that passive's growth is continuous and monotonic while every empirical study of it looks backward — each "harmless" finding testifies about a level of passive that no longer exists, and each "harmful" prophecy is still waiting for its sample.
One meta-conclusion from this run deserves its own line: of the four single-source empirics this debate leans on hardest, not one can bear the weight at which it is cited — one vetoed, three restricted, two direction-contested. This is not to say they are bad research; most are methodologically transparent, publish their data, and state their own limitations. The failure is at the transmission layer: the loosest member of a proxy family became the headline, a model-internal demand slope became evidence that "price response changed," a reduced-form correlation stripped of its mechanism section and insignificant in recent data became causal, and an extrapolation the authors themselves call seven times their own calibration became "passive inflated the megacaps by 29%."
And once the audit seats pried those four numbers open, what remained came into sharper focus: almost every finding that survives independent scrutiny lives at the mechanism layer, not the magnitude layer. The firm-specific information content of prices does fall with passive ownership (three teams, three measures); price-insensitive index buying does move prices persistently (Japan); shadow passive is far larger than the fund-based count (observed directly in transaction data). No independent evidence answers the magnitude question — how much distortion has actually resulted — and the entire force of the bubble thesis comes from magnitude.
Ordered by strength of evidence; each with its test.
1. [multi-source] "Passive passed half" holds only on the US fund-asset ruler (Morningstar end-2023; ICI's definition, 2024); the market-cap ruler is 19-23%. The direction that true passive exceeds the fund-based number has four independent teams behind it (measurements spanning 17.5%-41%), but no specific level is pinned by a second team. Test: ICI Fact Book and Morningstar annual updates; whether any independent replication of the reconstitution-volume estimator appears (there is none to date).
2. [multi-source] Active large-cap funds underperform over the long run (20-year rates: 91-94% by fund count, 55-78% by assets), and this did not improve as passive expanded after 2010. Test: the annual SPIVA and Morningstar Barometer series; whether Cremers-Fulkerson-Riley clears peer review and whether S&P issues a methodological response.
3. [direction contested + restricted use] Whether passive made individual-stock demand less elastic. The structural estimate says yes (HHL: about −11%; but the same paper's fund-based measure gives 6.3%, and Behmaram's parallel counterfactual about 43%), while direct measurement of index trading's price impact reports it falling from 6.75 to 0.37. More seriously, HHL's magnitude rests entirely on a model-generated instrument — remove it and the strategic-response parameter becomes a statistical zero, taking "competition offsets two-thirds" with it — and the sign itself follows from the model's definition (passive investors are assigned zero elasticity). Test: re-estimate the pass-through without the same demand system; run a sensitivity analysis on the 0.06 threshold that classifies institutions as passive — the paper still has none.
4. [multi-source on mechanism / vetoed on magnitude] The firm-specific information content of prices falls with passive ownership — three independent teams agree using return-variance decomposition, GPIN, and future earnings response coefficients; but the quantified claim that "pre-announcement informativeness therefore fell about a quarter" is vetoed (an equal-weighted average of four collinear ratios, a sixth under a different denominator, insignificant in the 2012-2019 high-passive subsample). Test: extend Höfler et al.'s variance decomposition to 2021-2026; test directly whether "less anticipation" or "more bundled disclosure at the announcement" drives the rise in announcement-day response.
5. [direction contested] The same "information enters prices later" facts admit an opposite welfare reading: Martineau documents weakening pre-announcement drift, earnings-day prices approximating a martingale, and PEAD dead since 2006 for all but microcaps — and reads it as markets becoming more efficient. Test: a measure that separates "less anticipation" from "more information released at the announcement"; a test where the two readings make different predictions on the same data.
6. [multi-source on mechanism / not load-bearing on magnitude] Index-linked, price-insensitive demand raises prices, and does so more for less elastic large stocks (HHL's size-elasticity gradient, Behmaram, the Bank of Japan natural experiment); but JVZ's "+29% over 25 years" is the authors' linear cumulation against their own 0-4% calibration, with two out-of-sample sign failures (2022 and 2026 to date: record ETF inflows alongside equal-weight beating cap-weight). Test: re-estimate the size gradient on 2021-2026 data with a joint monotonicity test. Note the "passive buys more of the megacaps" dose framing is wrong — buying is proportional, and index funds' proportional ownership peaks in small and mid caps.
7. [multi-source] The S&P 500 inclusion effect converged from 7.4% in the 1990s to an insignificant 0.83% in 2010-2020, but three 2025 direct additions produced double-digit moves (COIN +24%, HOOD +15.8%, APP +11.6%). Test: any peer-reviewed re-estimation covering 2021-2026; the next large direct addition.
8. [direction contested] Inclusion-driven excess comovement: CSW (2016) finds zero, DeCoste (2025) finds significantly positive using a discontinuity design. Test: extend the beta event study to the most recent sample; run both designs on the same data.
9. [multi-source] Official post-mortems of five "passive fragility" episodes attribute none of them to passive ownership or index-fund flows; two name ETF trading/arbitrage-layer mechanisms. Test: whether the next stress event's official report names passive ownership for the first time.
10. [multi-source] There is directional consensus on a volatility channel, with a calibrated magnitude of roughly 0.7 percentage points annualized — real and small. Test: the volatility structure once passive exceeds 25% of market cap.
11. [interested party × both directions] The 1.2% and 64% figures for "passive share of trading" are both construction choices, with different numerators and denominators, and the largest component (36% passive market making) has no primary source. Test: either side publishing a reconciliation table; an independent academic estimate of passive-attributable volume.
12. [multi-source] The Big Three concentration numbers are real but plateaued over 2019-2021; the 2026 stewardship splits are a legal form, with practitioners observing that the two teams often vote in parallel; and pass-through voting participation (3-11% by assets) is far below eligibility. Test: the first data on voting divergence between split teams; whether the 27.6% figure the projection requires for 2028 is on track.
13. [multi-source] The index-provider layer is the most solidly evidenced concentration in the file: three firms hold two-thirds of the market, take about a third of ETF expense ratios, entry fails to discipline pricing, and the SEC formally withdrew its regulatory agenda item in April 2025. Test: whether the SEC reopens it; the fee effects of any large fund family moving to self-indexing.
14. [single-source verified] The warning side's most testable threshold (83% in five years) conflicts with its own preprint (65%/90%), and its demonstration arithmetic (23% a year for Apple) is off by nearly a factor of two against its own input (0.167% a day). Test: the formal threshold once Green-Krishnan-Sturm clears review; the passive share and market state around 2031 on his own measure.
The dual-seat audit restricted all four single-source empirics, vetoed one, and ruled two direction-contested — the most damaging result in this site's run history, with all four verdicts written into the relevant sections rather than repeated here. Beyond that, verification reversed eight of this essay's own statements, each rewritten in the text:
Also logged: three fabrications at the tool layer — a summarizer inventing quotations for Chabakauri-Rytchkov claiming indexing raises volatility and lowers informativeness (the paper says the opposite on volatility and makes no informativeness claim), inventing a "flows explain one-third of fluctuations" statistic for Gabaix-Koijen (not in the paper), and inventing a model description for an unreachable Hartley-Gerding paper. Plus one search-echo contamination: a 0.46% persistence figure was fed into a query and returned verbatim by a summarizer, and was dropped for lack of independent verification.
Share measurement: Morningstar US Fund Flows (Aug 2019, Jan 2024, Dec 2025); ICI 2026 Investment Company Fact Book ch.2 and monthly Active and Index Investing (June 2026); Anadu, Kruttli, McCabe & Osambela, FEDS 2018-060r1 (published in Financial Analysts Journal 2020); Chinco & Sammon, JFE 157 (2024) 103860; Escobar, Pandolfi, Pedraza & Williams, GWU WP 2026-008; S&P Dow Jones Annual Survey of Assets; BlackRock ViewPoint (Oct 2017); Rowley, Lawrence & Ryder-Green, Vanguard Research (Aug 2025); SSGA "Closing time" (Jan 2026).
Warning and defense primaries: Bloomberg (Sept 4, 2019, Burry email interview); Ritholtz Masters in Business (Feb 2024, Einhorn transcript); Excess Returns (Jan 20, 2026, Green transcript) and Green, Krishnan & Sturm, SSRN 6438678; yesigiveafig.com (Aug 9, 2026); Fundsmith Equity Fund Semi-Annual Letter (July 2026) and Expansión (Nov 3, 2025); Bogle, WSJ (Nov 29, 2018) and Yahoo Finance (May 6, 2017); Asness, SSRN 4942046; Lamont, Acadian Owenomics (Feb 2024) and Financial Analysts Journal (2026); Institutional Investor (Jan 3, 2025).
Theory: Grossman & Stiglitz, AER 1980; Gârleanu & Pedersen, RAPS 2022; Bond & García, RFS 2022; Chabakauri & Rytchkov, JFE 2021; Gabaix & Koijen, NBER WP 28967; van der Beck, JFE 184 (2026); Haddad, Huebner & Loualiche, AER 115(3) 2025; Koijen, Richmond & Yogo, RES 91(4) 2024; Behmaram (UQAM, 2026).
Price discovery and index effects: Bai, Philippon & Savov, JFE 2016; Farboodi, Matray, Veldkamp & Venkateswaran, RFS 2022; Dávila & Parlatore, RFS 2025; Coles, Heath & Ringgenberg, JFE 145(3) 2022; Sammon, Management Science 71(6) 2025; Höfler, Schlag & Schmeling (2025); Bennett, Stulz & Wang, NBER w27593; Israeli, Lee & Sridharan, Rev Acct Stud 2017; Martineau, Critical Finance Review; Beaver, McNichols & Wang, JAE 2020; Ben-David, Franzoni & Moussawi, JF 73(6) 2018 and NBER WP 20071; Glosten, Nallareddy & Zou, Mgmt Sci 2021; Jiang, Vayanos & Zheng, RFS 38(12) 2025; Greenwood & Sammon, JF 80(2) 2025; Barberis, Shleifer & Wurgler, JFE 2005; Chen, Singal & Whitelaw, JFE 2016; DeCoste, Global Finance Journal 65 (2025); Kasch & Sarkar, FRBNY SR 484; Barbon & Gianinazzi, RAPS 2019; Charoenwong, Morck & Wiwattanakantang.
Fragility and official assessments: SEC/CFTC flash crash report (Sept 30, 2010); SEC research note (Dec 2015); BIS Quarterly Review March 2018 (two features), March 2021, September 2025; Federal Reserve Financial Stability Report, May 2020 and May 2026; FCA/Bristol survey (July 8, 2025); Bank of Japan statement (Sept 19, 2025) and Nikkei reporting.
Fund performance: SPIVA US Scorecard Mid-Year 2025 and Year-End 2025 (the latter via two independent media reproductions); SPIVA Persistence Scorecard YE2024; Morningstar US Active/Passive Barometer YE2025; Cremers, Fulkerson & Riley working paper (May 4, 2026); Ptak, Morningstar (June 8, 2026); Fama & French, JF 2010; Harvey & Liu, JF 2022; Berk & van Binsbergen, JFE 2015; Pástor, Stambaugh & Taylor, JFE 2015 and the Critical Finance Review exchange (2022); Cremers, Ferreira, Matos & Starks, JFE 2016; Unterberg, SSRN 6902444 (relayed via Morningstar); ICI Research Perspective 32-1; Morningstar 2026 US Fund Fee Study.
Governance and index providers: Bebchuk & Hirst, BU L Rev 2019 and 2022, Colum L Rev 2019; Kahan & Rock, BU L Rev 2020; Azar, Schmalz & Tecu, JF 2018; Dennis, Gerardi & Schenone, JF 2022; Backus, Conlon & Sinkinson, AEJ: Micro 2021; Gerardi, Lowry & Schenone, ARFE 16 (2024); Azar & Vives, ECGI 827/2022; Texas v. BlackRock filings (ECF 113, settlement documents) and the FTC/DOJ Statement of Interest (ECF 99); SEC IA-6050 and the Unified Agenda (RIN 3235-AM95); An, Benetton & Song, JFE 149(3) 2023; Molk & Robertson, Yale J. Reg. 41 (2024); BlackRock 2025 Global Voting Spotlight and BIS Annual Report 2025; Innisfree/HLS Forum (June 11, 2026).
Related research on this site: the yardstick-ladder method used here (one "share" differing 40-fold across three denominators) is the same instrument applied in A Physical for "95% of AI Pilots Fail"; for the gap between capital-market narratives and primary numbers, see Is This AI Capex Boom Another 1999?; for the structure in which both sides' signature numbers take equal damage, see Screen Time and Teen Mental Health.