Bid-ask spread

Definition

The bid-ask spread is the gap between the highest price buyers are posting (bid) and the lowest price sellers are posting (ask). It is the price of immediacy: a marketable order pays roughly half the spread relative to the midpoint on entry and again on exit, making the spread the unavoidable floor of round-trip trading cost.

How it works / structure

• Quoted vs effective: the quoted spread is the posted gap; the effective spread measures actual executions against the midpoint (captures price improvement and walking the book).
• Why it exists: classical decomposition — order-processing costs, inventory risk, and adverse selection: Glosten and Milgrom (1985) showed spreads arise because some counterparties are better informed, so the market maker’s quote must charge for that risk.
• Estimation without quotes: Roll (1984) derived a spread estimate from the negative autocovariance of price changes — useful for historical data lacking quote records.
• Determinants: volatility (up → wider), volume/competition (up → tighter), price level (tick-size constraints), and session time (ms-sessions-auctions — wide at the open, wide after hours).
• Simulation parameters: per-instrument spread model (constant, time-of-day, or volatility-scaled); the platform charges half the modeled spread per marketable fill by default.

When it applies

Every cost model, every strategy-viability check (high-turnover strategies live or die on spread), every options analysis (option spreads are proportionally far wider than equity spreads — ms-option-chain), and every execution-style choice (limit orders avoid paying the spread but accept non-fill risk).

Risk profile & failure modes

• Turnover multiplication: a strategy trading daily pays the spread ~250 times a year; a paper edge smaller than cumulative spread cost is not an edge.
• Quote fading: displayed spreads can vanish before a marketable order arrives; stressed effective spreads exceed calm quoted ones.
• Mid-price fictions: marking illiquid positions at mid overstates liquidation value — a 40%-wide options market has no meaningful mid.
• Limit-order adverse selection: resting orders fill preferentially when the market is moving through them — the fills arrive exactly when they are least wanted (the Glosten-Milgrom logic from the trader’s side).

Evidence & limits

The decomposition literature (Glosten-Milgrom 1985 and successors) and the Roll (1984) estimator are canonical and replicated. US equity spreads narrowed dramatically after decimalization (2001) — long-horizon cost backtests must use period-appropriate spread assumptions rather than current ones.

Falsifiable-thesis examples

Illustrations only, not signals:

• “X’s time-weighted quoted spread will average under 5 basis points during regular sessions next month” — falsified by the quote record.
• “Strategy S remains profitable in replay when charged 1.5x its modeled spread cost” — falsified by the stressed-cost replay run.

Cross-references

• Broader cost stack: ms-slippage-friction (impact, fees)
• Liquidity context: ms-liquidity, ms-sessions-auctions
• Options-specific spread issues: ms-option-chain