Flash crash (May 6, 2010)

Definition

On May 6, 2010 the US equity market fell ~6% and largely recovered within roughly half an hour; hundreds of individual securities printed absurd prices — some at a penny, some at $100,000 — before trades were cancelled. It is the KB’s case study in MODERN MICROSTRUCTURE FRAGILITY: electronic liquidity is dense in calm and can evaporate in seconds, market orders and stops execute into the vacuum, and ETF pricing detaches when its arbitrage machinery pauses.

How it works / structure

• The sequence (joint report + Kirilenko et al): a large automated sell program executed E-mini futures with a volume-participation algorithm insensitive to price (ms-execution-algos — parameterization matters); high-frequency intermediaries absorbed inventory, hit limits, and began passing it among themselves (“hot potato” volume without net absorption); liquidity withdrew across correlated instruments; the cascade propagated from futures into cash equities and ETFs.
• The single-stock absurdities: with books empty, market orders filled against STUB QUOTES (placeholder penny/extreme quotes) — the origin of one of the KB’s standing rules: market orders and unattended stop-market orders carry unbounded slippage in dislocations (mgmt-stop-loss gap mechanics, ms-liquidity).
• ETF specificity: a large share of cancelled trades were ETFs — creation/redemption arbitrage paused with unpriceable underlying baskets (instrument-etf NAV-band failure mode in its purest form).
• The structural response (engine-relevant): single- stock circuit breakers, then the LULD limit-band regime (ms-halts-luld), market-maker quoting obligations, and clearly-erroneous-trade rules — today’s halt mechanics are this episode’s direct output.

When it applies

Cited for order-type discipline (the platform’s default to limit orders in stress descends from this record); for intraday strategy risk (strategy-day-trading-styles sessions can contain microstructure vacuums); for ETF premium/discount behavior in dislocations; for execution- algorithm parameter honesty (participation without price limits was the trigger’s shape).

Risk profile & failure modes

• The liquidity mirage: displayed depth is not committed depth; calm-period book statistics say nothing about dislocations — sizing and order types must assume the vacuum case.
• Stop-market cascades: resting stops became market orders into emptiness — the documented mechanism behind the KB’s stop-limit and alert-based alternatives.
• Recovery-speed misread: the V-shaped recovery tempts “it self-corrected” complacency; the correction required trade cancellations and rule changes — the system did not self-correct at fair prices.
• Misuse: blaming “HFT” generically — the academic record (Kirilenko et al) distinguishes liquidity-taking from intermediation; the mechanism was interaction, not a villain.

Evidence & limits

The CFTC-SEC joint report is the official record; Kirilenko et al (2017) is the peer-reviewed microdata analysis. The initiating program’s identity and mechanics are documented; broader attribution debates (market fragmentation’s role) continue. LULD mechanics are current, documented rules.

Falsifiable-thesis examples

Illustrations only, not signals:

• “This strategy’s replay, re-run with a 30-minute liquidity-vacuum overlay (spreads 10x, depth 1/10th), keeps its worst trade loss under 3× normal (vacuum audit)” — falsified by the stressed replay.
• “No live order this book rests is a stop-market order in a name without LULD-band protection (order-type audit)” — falsified by the order-log scan.

Cross-references

• The rules it created: ms-halts-luld; the plumbing it tested: ms-liquidity, instrument-etf
• The order-type doctrine: mgmt-stop-loss, ms-execution-algos
• The style exposed: strategy-day-trading-styles
• The successor stress test: episode-2015-etf-dislocation