\section{Related literature} \label{sec:literature} % Citations verified at DOI level via OpenAlex/Crossref on 2026-07-13 and % re-verified against Crossref on 2026-10-17 (docs/CHANGELOG-REVISION.md). The question sits at the intersection of several literatures: theories of cross-asset shock transmission; empirical work on fire sales, flows, and ownership linkages; the measurement of comovement and liquidity commonality; lead--lag effects and the Epps effect; network approaches to contagion; the transformation of U.S.\ market structure and the high-frequency trading debate; and the econometrics of jumps, co-jumps, and scheduled announcements. We take each in turn and close with the paper's positioning. \subsection{Theories of cross-asset shock transmission} Theory identifies two broad families of channels through which a shock to one security moves others. The first is informational. \citet{king1990transmission} model agents who infer information from other markets' price changes, so that an idiosyncratic ``mistake'' in one market transmits broadly, and \citet{kodres2002rational} show that cross-market rebalancing of shared macro exposures transmits idiosyncratic shocks even without common fundamentals. The second family operates through balance sheets and funding. \citet{kyle2001contagion} generate contagion as a wealth effect of convergence traders; \citet{yuan2005asymmetric} shows that borrowing constraints make contagion and asymmetric price movements likelier after losses; and \citet{brunnermeier2009market} link market liquidity to funding liquidity in margin spirals that explain sudden, correlated liquidity dry-ups. Most directly related to our setting, \citet{cespa2014illiquidity} show that when liquidity providers learn from other securities' prices, illiquidity becomes contagious and the market admits liquidity-crash equilibria, a mechanism they connect to May~6, 2010. These theories jointly predict that transmission should be state-dependent, asymmetric, and increasing in the degree to which pricing and liquidity provision are cross-security activities. Our measures of propagation breadth, network concentration, sign asymmetry, and market-state dependence are designed to test these comparative statics at the minute scale. \subsection{Fire sales, flows, and ownership networks} A complementary empirical literature documents non-fundamental transmission through investor portfolios: forced sales by mutual funds facing outflows generate price pressure and reversals \citep{coval2007asset}, predictable flows generate predictable pressure \citep{lou2012flow}, and uninformative imbalances in one stock spill over to correlated stocks through risk-averse liquidity providers \citep{andrade2008trading}. Ownership structure predicts fragility \citep{greenwood2011stock}, common active ownership forecasts correlation beyond fundamentals \citep{anton2014connected}, and overlapping portfolios aggregate into measurable systemic vulnerability \citep{greenwood2015vulnerable,duarte2021fire}. This work establishes that idiosyncratic demand shocks do propagate, but at monthly or quarterly resolution. Whether the same propagation operates within the hour, and whether its speed changed as intermediation became algorithmic, is the margin our data address. \subsection{Comovement, commonality, and liquidity dynamics} Liquidity has a systematic component: spreads and depths co-move with market- and industry-wide liquidity \citep{chordia2000commonality}, common factors in order flow explain much of the commonality in returns \citep{hasbrouck2001common}, and commonality is supplied in part by common intermediaries \citep{coughenour2004common}. Liquidity falls after market declines, especially when funding is tight \citep{hameed2010stock}, specialist losses forecast spread widening \citep{comertonforde2010time}, and returns to liquidity provision spike as liquidity ``evaporates'' in turmoil \citep{nagel2012evaporating}; aggregate liquidity risk is priced \citep{pastor2003liquidity,acharya2005asset} and commonality strengthens in downturns across markets \citep{karolyi2012understanding}. Relative to this literature we do not measure liquidity levels: bars contain no quotes, and the only liquidity information we use is prior-month dollar volume as a ranking device, in the spirit of \citet{amihud2002illiquidity}, and the high--low range as a descriptive volatility proxy \citep{parkinson1980extreme, garman1980estimation}. Such OHLC-based proxies have a long pedigree \citep{roll1984simple,corwin2012simple,goyenko2009liquidity}, but they do not identify the quote-level mechanisms that the commonality literature studies. \subsection{Lead--lag effects, asynchronous trading, and the Epps effect} Cross-autocorrelation in returns is among the oldest facts in empirical asset pricing: large stocks lead small stocks \citep{lo1990contrarian}, predominantly within industries \citep{hou2007industry}, along customer--supplier links \citep{cohen2008economic,menzly2010market}, and, at tick resolution, from liquid to illiquid assets with intraday-seasonal structure \citep{huth2014high}; \citet{chinco2019sparse} show that one-minute-ahead cross-stock predictability is sparse and short-lived. At the shortest horizons measured comovement is attenuated mechanically. \citet{epps1979comovements} first documented that correlations among stocks fall toward zero as the sampling interval shrinks. Part of this ``Epps effect'' is a pure artifact of nonsynchronous trading and price discreteness \citep{scholes1977estimating,lo1990nonsynchronous,harris1991stock}, which \citet{reno2003closer} shows can reproduce the decay by simulation and which estimators built on actual transaction times remove \citep{hayashi2005covariance,zhang2011estimating}; microstructure noise further contaminates high-frequency second moments \citep{hansen2006realized}. Part is genuine delay in the incorporation of common information. \citet{toth2009epps} document that the Epps curve among NYSE stocks flattened steadily between 1993 and 2003 and argue that denser trading explains only part of the change. Our Epps evidence extends this series to 2024 on previous-tick synchronized grids and must be read with both mechanisms in mind; we return to the point in Section~\ref{sec:results}. The same mechanics matter for our design, because neighbor sets are identified from one-minute residual correlations and are therefore identified less precisely in eras where the Epps effect was strong. \subsection{Networks, connectedness, and contagion measurement} Network theory shows when idiosyncratic shocks fail to diversify away \citep{acemoglu2012network,elliott2014financial}; \citet{glasserman2016contagion} survey the contagion-network literature; real-economy links propagate firm-level shocks measurably \citep{barrot2016input} and network positions carry asset-pricing implications \citep{herskovic2018networks}. On the measurement side, connectedness is estimated from variance decompositions \citep{diebold2012better,diebold2014network}, Granger-causality graphs \citep{billio2012econometric}, sparse VARs \citep{barigozzi2019nets}, and tail-risk networks \citep{hautsch2015financial}. The contagion-measurement tradition warns that unconditional correlation comparisons are biased in volatile regimes \citep{forbes2002no} and defines contagion as comovement beyond fundamentals \citep{bekaert2014global}; classic volatility-spillover evidence \citep{engle1990meteor,hamao1990correlations} anticipates our design at coarser resolution and for a handful of markets. These networks are estimated at daily or lower frequency on small panels. We estimate pre-shock correlation neighborhoods on minute data for the five hundred most active securities each month and use them as lagged conditioning information, with the same neighborhood contrast computed on placebo events as the benchmark. \subsection{Market-structure evolution and the high-frequency trading debate} Our sample spans the modern transformation of U.S.\ equity trading. Decimalization in 2001 cut spreads \citep{bessembinder2003trade}; liquidity promotes efficiency \citep{chordia2008liquidity}; turnover exploded in the 2000s alongside algorithmic trading \citep{chordia2011recent}; NYSE automation made prices more efficient while raising the cost of immediacy \citep{hendershott2011automation}; and fragmentation did not measurably harm quality \citep{ohara2011fragmentation}. The normal-times evidence on algorithmic and high-frequency trading is broadly positive \citep{hendershott2011algorithmic,brogaard2014high,menkveld2013high, hasbrouck2013low,chaboud2014rise,boehmer2021algorithmic}, with \citet{menkveld2016economics} surveying the channel-by-channel evidence, and markets absorb even large predictable trades with limited lasting impact \citep{bessembinder2016liquidity}. Theory tempers the optimism: speed investment can be socially excessive \citep{biais2015equilibrium}, fast news trading concentrates volume and short-run correlation \citep{foucault2016news}, toxic arbitrage degrades liquidity \citep{foucault2017toxic}, the continuous limit order book hard-wires a latency arms race \citep{budish2015high,aquilina2022quantifying}, and speed can crowd out information acquisition \citep{weller2018does,baldauf2020high}, with fast traders eventually trading with informed institutional orders \citep{vankervel2019high}; \citet{ohara2015high} argues the machine age changes what microstructure concepts mean. This literature motivates our central hypothesis but measures liquidity levels, price discovery, or single events. We contribute the missing time series: propagation breadth and network concentration of extreme shocks measured uniformly across three decades of structural change, with year-by-year estimates rather than imposed trend breaks. \subsection{Fragility and flash events} Episodic evidence shows that modern liquidity can vanish abruptly. On May~6, 2010, high-frequency traders did not trigger the crash but amplified selling once it began \citep{kirilenko2017flash}; cross-market arbitrage broke down \citep{menkveld2019flash}; and fragmentation shaped which ETFs were hit \citep{madhavan2012exchange}. Sub-second ``fractures'' in individual stocks proliferated as machine ecologies moved beyond human response times \citep{johnson2013abrupt}. More systematically, high-frequency traders supply liquidity in single-stock extreme moves but their provision weakens when many stocks move together \citep{brogaard2018high}---precisely the margin our breadth measure tracks. Proposed early-warning metrics remain contested: VPIN signaled deteriorating conditions before the Flash Crash \citep{easley2012flow}, but its power appears mechanically tied to volatility \citep{andersen2014vpin}. This debate cautions against fragility indicators that repackage volatility; our propagation measures are therefore defined relative to lagged conditional distributions and benchmarked against matched placebo minutes. \subsection{ETFs, indexing, and passive intermediation} A structural change specific to the later sample is the growth of index-linked intermediation. ETF ownership raises non-fundamental volatility of underlying stocks through the arbitrage channel \citep{bendavid2018etfs} and contributes to return comovement \citep{da2018exchange}; index-product arbitrage flipped index return serial dependence from positive to negative across twenty markets \citep{baltussen2019indexing}. Effects on efficiency are contested \citep{israeli2017dark,glosten2021etf,coles2022index}. For us, the ETF layer is a candidate mechanism for the sharpening of propagation along correlation neighborhoods, since basket arbitrage converts idiosyncratic shocks into correlated flows. We cannot test it directly: the data carry no security-type labels, so ETFs are neither identified nor excluded, a limitation we state rather than paper over. \subsection{Jumps, co-jumps, and scheduled announcements} Our shock measurement builds on high-frequency econometrics. Intraday volatility periodicity must be filtered before anything else \citep{andersen1997intraday}, and robust periodicity estimators are needed because the periodicity estimate is itself contaminated by the jumps one wishes to detect \citep{boudt2011robust}; our minute-of-day factor is a cross-sectional median for that reason. Bipower variation separates jumps from diffusion \citep{barndorffnielsen2004power}, formal jump tests exist at the observation level \citep{lee2008jumps,aitsahalia2009testing}, and jump tails carry extreme cross-sectional dependence \citep{bollerslev2013jump}. Two cautions discipline our thresholds: at ultra-high frequency much of what coarser methods call jumps is friction \citep{christensen2014fact} or a ``drift burst'' extending over several minutes \citep{christensen2022drift}, and multiple-testing corrections eliminate a large share of spurious detections \citep{bajgrowicz2016jumps}. Our events are threshold exceedances of standardized residuals, not test-based jump detections, and we use the term ``extreme move'' rather than ``jump'' wherever the distinction matters. The co-jump literature is the nearest empirical neighbor to our propagation object. Index-level jumps are systematic co-jumps linked to macro news while idiosyncratic jumps diversify away \citep{bollerslev2008risk}; formal tests for common jump arrivals exist \citep{jacod2009testing}; co-jumps among large U.S.\ stocks are frequent and concentrated around announcements \citep{gilder2014cojumps}; simultaneous multivariate jumps are associated with liquidity dry-ups and priced \citep{caporin2017systemic}; jumps coincide with liquidity-demand shocks and news \citep{boudt2014intraday}; and simultaneous jumps across countries carry portfolio consequences \citep{das2004systemic}. Closest to our co-jump series, \citet{bormetti2015modelling} model systemic price co-jumps among many stocks with Hawkes factor processes and \citet{calcagnile2018collective} document that the number of stocks jumping in the same minute is far larger than independence or a single factor would imply, with collective synchronization events increasing in frequency in the high-frequency era. Self- and cross-excitation of jumps has a canonical language in Hawkes processes \citep{hawkes1971spectra,bacry2015hawkes}, estimated on international indices by \citet{aitsahalia2015modeling} and tested at high frequency by \citet{dungey2018testing}. Scheduled macro announcements provide the identified common shocks against which idiosyncratic propagation must be benchmarked: announcement surprises produce jump-like responses \citep{andersen2003micro,andersen2007real}, generate co-jumps \citep{lahaye2011jumps}, structure volume and volatility \citep{bollerslev2018volume}, and structure expected returns around announcement times \citep{savor2013how,lucca2015pre}. Our robustness battery accordingly excludes the main announcement windows. Where the co-jump literature tests for simultaneity, our propagation measures trace the sequence---breadth at one through 120 minutes---and our co-jump series traces the secular evolution of simultaneity itself. \subsection{Factor structure at the minute scale} Separating idiosyncratic from systematic movement requires a factor structure: common factors exist in high-frequency returns and order flow \citep{hasbrouck2001common}, high-frequency principal components are consistent with low-frequency covariance structure \citep{aitsahalia2019principal}, latent continuous and jump factors are estimable \citep{pelger2020understanding}, and betas decompose into continuous, jump, and overnight components with distinct pricing \citep{bollerslev2016roughing,bollerslev2020realized}; jump-tail risk commands a large, time-varying premium \citep{bollerslev2011tails}. We deliberately use a single, strictly lagged market loading per stock-month rather than these richer structures, because any loading estimated on the event month would overfit the very minutes it is used to classify; the cost, a coarser residual, is absorbed by the placebo contrast. \subsection{Positioning and contribution} Each literature contributes a piece: theory predicts state-dependent, asymmetric transmission; the flows literature demonstrates non-fundamental propagation at monthly resolution; the network literature measures connectedness on small panels at daily frequency; the market-structure literature documents the transformation of trading but measures liquidity levels and single events; the co-jump literature tests simultaneity at high frequency over short samples. To our knowledge---based on a documented search dated 2026-07-13 and repeated on 2026-10-17---no prior study combines minute-level resolution, the full U.S.\ equity cross-section including delisted securities, and a 1992--2026 sample spanning floor trading, decimalization, Reg~NMS fragmentation, and ETF-intermediated algorithmic markets. The joint analysis of market-wide and network-concentrated propagation, its secular evolution, and the evolution of synchronized extreme minutes at that resolution and length---with placebo-timing, placebo-network, announcement-exclusion, and tick-regime falsifications---is the contribution.