Financial markets can be liquid and stable for long stretches, then suddenly seize up. The 2010 US equity market ‘flash crash’, the 2014 US Treasury turbulence, and the March 2020 pandemic sell-off in US Treasuries all share a disturbing pattern: liquidity vanished precisely when it was most urgently needed. Existing explanations tend to focus on inventory constraints, information asymmetries, or regulatory frictions. We offer a different, complementary diagnosis. We argue that the culprit is opacity — the inability of market participants to observe each other’s positions and order flow. When markets are sufficiently opaque, a robust mechanism generates fragility as an equilibrium property, not as an accident or a dysfunction.

Flash crash events have been recurrent in equity, bond, and foreign exchange markets, raising concerns among the regulatory authorities and central banks (e.g. BIS 2017, Easley et al. 2011). As the experience of the US stock market crash of 19 October 1987 makes clear, crashes unrelated to fundamentals have been a regular feature of financial markets (see also Gehrig et al. 2015 on historical evidence on how market opacity amplifies market stress). Gennotte and Leland (1990), in a model with multiple equilibria, argue that this event is the outcome of portfolio insurance strategies not anticipated by traders. However, in their model programme trading is exogenous. In Cespa and Vives (2026), we identify a mechanism that generates fragility in a model where all traders are risk-averse expected utility maximisers.

The key mechanism

Modern financial markets are not populated solely by traditional market-makers. A large and growing share of liquidity supply comes from ‘non-standard’ liquidity providers: hedge funds taking contrarian positions, high-frequency traders (HFTs) providing two-sided quotes, and portfolio managers who act as willing counterparties when prices overshoot (Biais et al. 2017, Brogaard et al. 2014, Anand et al. 2021) These participants are discretionary. Unlike designated market-makers with formal obligations, they choose whether to supply liquidity based on the profit opportunities they perceive. Our central insight in Cespa and Vives (2026) is that their willingness to do so depends critically on how much they can infer about the aggregate order flow hitting the market.

When markets are transparent — when investors who trade late can observe or reliably infer the direction of earlier order flows — contrarian trading is attractive. Seeing that prices have moved because investors needed to rebalance, late traders can infer the move is likely to revert, and it pays to step in and absorb it. Opacity destroys this signal. Unable to read the direction of earlier order flows, late traders speculate more heavily on their own noisy private signals, injecting price noise precisely when the market is under stress and the signal-to-noise ratio is lowest. This makes execution unpredictable for early rebalancers and drives them to scale back. But the mechanism turns perverse at this point: as early rebalancers withdraw, execution conditions for late traders improve — with less early-round order flow to contend with, late traders face lower execution risk and respond by expanding their positions. Opacity thus sets off a self-reinforcing cycle: the speculative noise that late traders inject crowds out early rebalancers, whose retreat then rewards late traders with better trading conditions and draws them in further. In this respect, digital technology is no panacea, as it may generate opacity by making different traders have different access to public information (Ding et al. 2014, Duffie et al. 2022, and Foucault et al. 2025).

Strategic complementarity

We formalise this intuition through a feedback loop. When aggregate selling pressure is heavy, dealers demand wider spreads to bear the inventory risk it creates. This is not merely a static cost: under market opacity it sets off a chain reaction. Investors who need to rebalance early, unable to disentangle the sources of the price movements they face, scale back their orders. Their withdrawal makes prices less reactive to early trading pressures — which improves conditions for investors who trade later: with less noise from early-round orders in the price signal, later traders can read the market more clearly and act more aggressively. But heavier late-round trading loads more inventory risk onto dealers, who respond by demanding even more compensation. Therefore, high price impact begets higher price impact.

This is the hallmark of strategic complementarity, creating the potential for multiple equilibria. The same fundamentals — the same asset, the same investors, the same distribution of shocks — can support either a good equilibrium with deep markets and narrow spreads, or a bad equilibrium with thin markets and wide spreads. Which one prevails can depend on beliefs, recent history, or small perturbations. Opacity is what enables the loop: under transparency, early investors could identify the sources of price movements and trade selectively rather than pulling back uniformly, breaking the chain. It is the inability to make that distinction that turns high price impact into an undiversifiable execution risk and locks in fragility as an equilibrium property rather than a pathology (see also Engle and Acharya (2009) on a related point). Consistent with this, we show that higher opacity degrades liquidity even in the good equilibrium — not just making crashes more likely but lowering average market quality throughout.

Traders who trade most, pay most

One striking equilibrium implication is that in our model, “traders who trade most, face worse execution.” In the bad equilibrium, the traders who most need to trade — those with large positions to unwind, those facing margin calls, those responding to large fundamental shocks — face the worst execution quality. This is precisely backwards from a well-functioning market, where large orders can be broken up and absorbed over time. In the fragility equilibrium, urgency and size are penalised rather than accommodated. This result has direct relevance for understanding the experience of leveraged institutions during stress episodes and the amplification dynamics seen in forced-selling cascades.

Policy implications

The analysis yields a clear and actionable policy conclusion: transparency raises welfare. If regulators can increase the observability of order flow and positions — reducing the opacity that enables strategic complementarity — they can shift the market toward the good equilibrium and reduce the probability of a fragile outcome. This is not a generic endorsement of disclosure; it is a targeted argument about the specific channel through which opacity generates systemic risk in liquidity provision.

There is an important nuance, however. The beneficial effect of transparency operates primarily through its impact on liquidity suppliers’ inference problem. Transparency that reduces uncertainty about aggregate order flow is welfare-improving; transparency that primarily exposes individual positions to strategic exploitation could have the opposite effect. Policy should therefore be targeted: the goal is to give non-standard liquidity suppliers better information about market-wide conditions, not necessarily to mandate full disclosure of proprietary positions.

Our paper speaks directly to three live policy debates. First, the resilience of the US Treasury market: the 2014 Joint Staff Report identified high-frequency traders and principal trading firms as key liquidity providers, yet their participation evaporated during the October 2014 intraday spike. Our framework explains this as a foreseeable consequence of opacity in a market dominated by non-standard suppliers, and it suggests that reforms to post-trade transparency — such as wider dissemination of TRACE data for Treasuries — could reduce fragility risk. The Fed warned in its 2022 Financial Stability report about fragile liquidity provision. Second, the Securities and Exchange Commission’s proposed dealer registration rule, which would extend regulated dealer status to high-frequency traders and other active principals, is evaluated in the model’s terms: requiring such entities to register as dealers would impose affirmative obligations that reduce their ability to withdraw strategically, but it would also raise their costs and potentially reduce participation, so the net effect is ambiguous and depends on the degree to which their participation is welfare-improving at the margin. Third, the paper bears on the EU and UK consolidated tape implementation. A consolidated tape that makes post-trade data more widely and rapidly available would, in the model’s logic, reduce opacity and support the good equilibrium — a concrete welfare rationale for a reform that has often been justified on vaguer grounds of market efficiency. A consolidated tape is set to go live in June 2026 for bonds in the UK, and ESMA has already selected providers for bonds and equities in the EU.

Taken together, we provide a rigorous foundation for why modern financial markets are structurally prone to liquidity crises, and why the standard toolkit — circuit breakers, leverage limits, mandatory market-making — addresses symptoms rather than causes. The root cause is opacity. The prescription is carefully designed transparency measures.

References

Anand, A, C Jotikasthira and K Venkataraman (2021), “Mutual fund trading style and bond market fragility”, Review of Financial Studies 34(6): 2993–3044.

Bank for International Settlements (BIS) (2017), “The Sterling ‘flash event’ of 7 October 2016”, BIS Working Paper.

Biais, B, F Declerck and S Moinas (2017), “Who supplies liquidity, how and when?”, Working Paper, Toulouse School of Economics.

Board of Governors of the Federal Reserve System (2022), Financial Stability Report, November.

Brogaard, J, T Hendershott and R Riordan (2014), “High-frequency trading and price discovery”, Review of Financial Studies 27(8): 2267–2306.

Cespa, G and X Vives (2026), “Market opacity and fragility: why liquidity evaporates when it is most needed”, American Economic Review, forthcoming. 

Ding, S, J Hanna and T Hendershott (2014), “How slow is the NBBO? a comparison with direct exchange feeds”, Financial Review 49: 313–332.

Duffie, D, T Foucault, L Veldkamp and X Vives (2022), Technology and finance, The Future of Banking 4, CEPR Press.

Easley, D, M O’Hara and M López de Prado (2011), “The microstructure of the Flash-Crash: flow toxicity, liquidity crashes, and the probability of informed trading”, Journal of Portfolio Management 37: 118–128.

Engle, R and V Acharya (2009), “A case for (even) more transparency in the OTC markets”, VoxEU.org, 29 August.

Foucault, T, L Gambacorta, W Jiang and X Vives (2025), Artificial Intelligence in Finance, The Future of Banking 7, CEPR Press.

Gehrig, T, M Haas and C Fohlin (2015), “Rumors and runs in opaque markets: Evidence from the Panic of 1907”, VoxEU.org, 7 May.

Gennotte, G and H Leland (1990), “Market liquidity, hedging, and crashes”, American Economic Review 80(5): 999–1021.