Equity market makers (how they behave)

Definition

Equity market makers earn the bid-ask spread by standing ready to trade against incoming orders — and everything in their observable behavior follows from one problem: INVENTORY RISK. Grossman-Miller’s model formalizes it: market makers absorb temporary imbalances between buyers and sellers who don’t arrive simultaneously, holding unwanted inventory until the other side shows up, and charging (via the spread and price concessions) for the risk carried in between. Modern market making is electronic and concentrated (a handful of firms — Citadel Securities, Virtu, and peers — handle most US retail flow), but the inventory logic is unchanged, and it explains the behaviors traders actually encounter: fading quotes, widening spreads into uncertainty, and price pressure after large one-sided flow.

How it works / structure

• The core economics: quote both sides; profit = spread capture × volume − adverse-selection losses − inventory-risk costs; the maker’s enemy is INFORMED flow (trading against someone who knows something loses money — the Glosten-Milgrom adverse-selection logic), so spreads WIDEN when information risk rises (pre-earnings, macro releases — event-earnings behavior traders observe directly).
• Inventory management in the tape: after absorbing size, makers skew quotes to shed inventory (bid lower/offer lower after buying too much) — producing the temporary price PRESSURE and REVERSAL pattern documented as the transitory component of market impact (ms-market-impact); mean-reversion scalping strategies are, functionally, amateur inventory-absorption in the same business.
• Segmentation behavior: retail flow is prized (uninformed on average — the economic basis of ms-payment-for-order-flow), institutional flow is feared; makers internalize the former off-exchange and quote thinner on-exchange where informed flow concentrates (ms-dark-pools-ats venue geography follows this logic).
• Stress behavior (the documented pattern): with no affirmative quoting obligations of consequence, makers WITHDRAW when inventory risk explodes — quote fading and stub quotes in the episode-flash-crash-2010 record; liquidity is a fair-weather service priced continuously, not a utility (the KB’s standing liquidity caveat traces here).

When it applies

Interpreting spreads and depth as risk gauges (maker behavior is the transmission — spread widening IS their risk repricing); execution planning (large orders pay the inventory premium — ms-market-impact sizing); understanding post-flow reversals (strategy-mean-reversion overlap); event-window tactics (spreads widen into scheduled information — documented, plan crossing costs accordingly).

Risk profile & failure modes

• Assuming liquidity persistence: the 2010 lesson — displayed depth is an option makers can pull, not a commitment; stress-scenario execution plans must price its disappearance.
• Fighting the skew blind: entering right after visible one-sided sweeps often buys the maker’s inventory problem at the worst price — the reversal pattern is documented but not reliable enough to trade naively (regime-dependent magnitudes).
• Anthropomorphizing: “market makers are hunting my stop” — maker behavior is mostly mechanical inventory and adverse-selection management at portfolio scale; conspiracy readings generate bad models (though stop-clustering zones ARE liquidity pools any large trader can see — the mechanical version of the intuition).
• Era drift: floor-specialist behavioral lore predates electronic making — quoting behavior, speeds, and obligations changed structurally (quant-data-hygiene era discipline).

Evidence & limits

Grossman-Miller (1988) anchors the inventory model; Glosten-Milgrom the adverse-selection component; the flash-crash record documents stress behavior (SEC/CFTC joint report, cited in that entry). Firm-level behavior is proprietary — the KB carries the documented mechanics and labels finer-grained flow lore as practitioner inference.

Falsifiable-thesis examples

Illustrations only, not signals:

• “Effective spreads in name X widen >50% in the 30 minutes before scheduled earnings vs matched non-event windows (information-risk repricing check)” — falsified by the paired spread data.
• “Large one-sided sweep sequences are followed by partial price reversal within 30 minutes at above-chance rates in liquid large-caps (inventory-pressure thesis)” — falsified by the event-study distribution.

Cross-references

• The priced surface: ms-bid-ask-spread, ms-liquidity; the cost model: ms-market-impact
• The flow economics: ms-payment-for-order-flow, ms-dark-pools-ats
• The speed cohort: inst-hft-behavior; the stress record: episode-flash-crash-2010