Dark pools & off-exchange trading

Definition

A substantial share of US equity volume — routinely 40%+ in recent years (FINRA transparency data) — executes OFF the lit exchanges: in dark pools (ATSs that match orders without displaying quotes), via wholesaler internalization (ms-payment-for-order-flow), and in broker internal crosses. Prices still reference the lit market’s NBBO, but the DEPTH visible on screens is a minority of true liquidity — a structural fact that changes how volume, liquidity, and price levels should be read.

How it works / structure

• The venue taxonomy: lit exchanges (displayed quotes, the NBBO’s source — ms-consolidated-tape); dark ATSs (regulated under Reg ATS, matching typically at the midpoint, no pre-trade display — institutional blocks hiding their size); wholesalers (retail internalization); single-dealer platforms. All off-exchange prints report to the tape with a delay suited to their type (FINRA transparency data publishes per-venue weekly volumes).
• Why darkness exists: displaying a large order IS market impact (ms-execution-algos — the information leakage problem); dark venues let size meet size without pre-announcing; the price they use is parasitic on lit discovery — the documented tension regulators manage (fair-access and volume thresholds in Reg ATS).
• Reading consequences (engine-relevant): displayed book depth understates true liquidity by a large, name-varying factor (ms-liquidity calibration); volume-profile levels built from tape prints include dark volume, but INTRADAY tape sequence under-represents where resting institutional interest sits (indicator-volume-profile caveat); unusually high off-exchange share in a name is a measurable institutional-activity fingerprint.
• The short-sale data connection: off-exchange prints drive the daily “dark pool short volume” statistics that circulate retail — mechanically mostly market-maker hedging prints, a documented misreading source (sent-short-interest data-hygiene note extended).

When it applies

Liquidity estimation (screen depth × a venue-share adjustment, not screen depth); execution routing literacy (midpoint dark fills as spread savings for patient orders); institutional-footprint reading (off-exchange share shifts); data hygiene against dark-pool-mythology indicators circulating socially.

Risk profile & failure modes

• Discovery erosion (the policy debate): if too much volume prices off a lit book too thin to trust, the reference degrades — documented threshold debates, unresolved; labeled as such.
• Adverse-selection asymmetry in pools: informed flow seeks dark fills too — uninformed resting orders in pools get picked off around moves (documented venue-toxicity variation; institutional concern, minor at retail scale).
• Myth-based indicators: retail dashboards selling “dark pool positioning” as directional signal over-read mechanical prints — the KB treats venue data as structure, not signal, absent validated evidence.
• Regime shifts: venue shares move with volatility (stress pushes volume lit) — liquidity models calibrated on calm venue mix mis-estimate stress depth.

Evidence & limits

Reg ATS and FINRA transparency data document the structure and volumes; the price-discovery debate is extensively studied without consensus — labeled. Venue toxicity research exists but is largely proprietary or institutional; the entry carries the public architecture.

Falsifiable-thesis examples

Illustrations only, not signals:

• “Names in the top decile of off-exchange share show displayed-depth-to-realized-volume ratios below the market median (hidden-liquidity structure check)” — falsified by the cross-sectional measurement.
• “A sustained 10-point rise in a name’s off-exchange share precedes 13F-visible ownership change the following quarter (institutional-footprint thesis)” — falsified by the paired data.

Cross-references

• The reference price: ms-consolidated-tape; the retail segment: ms-payment-for-order-flow
• The liquidity calibration: ms-liquidity, indicator-volume-profile
• The impact-hiding rationale: ms-execution-algos
• The data-hygiene sibling: sent-short-interest