On This Page
Key Takeaways
- Adverse selection arises when better-informed traders select quotes that favor their information.
- In the worked binary-value model, a 20% informed-order share produces a $49.90 bid and $50.10 ask.
- The market maker earns zero expected profit across the specified order population.
- Effective spread and realized spread compare a fill with the arrival midpoint and a selected later midpoint.
- Compare fill price, subsequent midpoint changes and completion rate when evaluating passive orders.
What is adverse selection?
Adverse selection is the exposure a quoting firm faces when a trader with better information chooses the resting quote that favors that information. A market maker posts a bid and an ask, and informed counterparties take the side that offers them the stronger trade: they buy below, or sell above, the value they expect. The spread has to cover those expected losses.
The Glosten–Milgrom model isolates this mechanism. A competitive, risk-neutral quoting firm faces informed and uninformed orders. With zero processing and inventory costs, it still needs a positive spread to earn zero expected profit across that population.
Formula
Effective spread = 2q(P − M₀); realized spread at h = 2q(P − Mₕ); midpoint impact at h = 2q(Mₕ − M₀)- q is +1 for a customer buy and −1 for a customer sell. P is execution price, M₀ is the midpoint at order receipt and Mₕ is the midpoint at the selected post-execution horizon.
- Effective spread minus realized spread equals the direction-adjusted midpoint change, multiplied by two.
- Realized spread is the liquidity provider’s edge after the quote moves, the fill measured against the later midpoint Mₕ before fees and rebates.
- Rule 605 uses realized-spread horizons of 50 milliseconds, 1 second, 15 seconds, 1 minute and 5 minutes.
Adverse selection example: how informed orders set the spread
A market maker quotes stock XYZ before a binary announcement. The value will be either $50.50 or $49.50 with equal probability, so the unconditional value is $50.00. Twenty percent of incoming orders come from traders who already know the outcome and always trade in the profitable direction. The other eighty percent trade for reasons unrelated to the news and buy or sell with equal probability. The market maker has no processing cost, no inventory constraint, and competes to zero expected profit.
- P(buy | good news) = 20% + 80% × 50% = 60%; P(buy | bad news) = 40%
- P(good news | buy) = (50% × 60%) / 50% = 60%
- Ask = 60% × $50.50 + 40% × $49.50 = $50.10; bid = $49.90
- Quoted spread = $50.10 − $49.90 = $0.20
- For a 1,000-share sale to an informed buyer, terminal profit = 1,000 × ($50.10 − $50.50) = −$400
- For a 1,000-share sale to an uninformed buyer, expected terminal profit = 1,000 × ($50.10 − $50.00) = $100
- Population-weighted expected profit = 20% × (−$400) + 80% × $100 = $0
The model’s twenty-cent spread funds expected losses to informed traders. Raising the informed-order share to 40% moves the initial ask to $50.20 and bid to $49.80, doubling the spread to forty cents.
Selling 1,000 shares to an informed buyer loses $400 at the terminal value. Selling the same quantity to an uninformed buyer earns $100 in expectation. Weighting these results by 20% and 80% gives zero expected profit.
Increase the informed share to 40% and the initial quote becomes $49.80 by $50.20. The larger information imbalance doubles the spread in this example.
How effective and realized spreads measure adverse selection
Effective spread compares execution with the midpoint at order receipt. Realized spread compares the same execution with the midpoint at a specified later horizon. Their difference measures twice the direction-adjusted midpoint change over that interval.
For a buy, execution above the later midpoint produces a positive realized spread for the liquidity provider. Execution below that midpoint produces a negative one. Evaluate the measure alongside actual position exits and trading costs when reviewing profitability.
The SEC’s amended Rule 605 for NMS-stock execution reports specifies average realized spread at 50 milliseconds, 1 second, 15 seconds, 1 minute and 5 minutes after execution. The rule covers stock executions, and older reports follow the earlier version of the rule, so match the rule version before comparing two reports. Each horizon shows a different point in the quote’s development.
How traders check their fills for adverse selection
Compare passive fills with subsequent quotes, completed quantity and the orders left unfilled. Segment the results by ticker, order size and trading conditions to identify the setups with the strongest execution.
The same reading applies to options: in options flow, compare repeated offer-taking or bid-hitting across strikes with changes in spread and displayed size, then test entry prices across the quoted spread in the options profit calculator to see their effect on breakevens and payoff.
Limitations of adverse selection models and measures
- The model’s $0.20 spread prices information alone. Inventory and processing costs come on top in a live quote, so read any spread as the sum of all three.
- The informed share never prints on a live quote, so measure it in the markouts. Judge it across many fills, since one fill’s later move fits public news as well as an informed counterparty.
- Realized spread is gross. Subtract fees, add rebates and use the actual exit price to turn it into trading profit.
- Read the markout at several horizons. A 50-millisecond markout and a 5-minute markout tell different stories about the same fill, and public news inside the window moves the midpoint whoever the counterparty was.
Bottom line
Adverse selection is the cost of trading against better-informed counterparties, and it is one reason a quoting firm needs a positive spread even with no other costs. Assess the exposure on your own fills by comparing each execution with the arrival midpoint and with the midpoint at a fixed later horizon, and judge passive orders by what the quote does after they fill.
Clear answers
Frequently Asked Questions
Sources and Further Reading
- Glosten and Milgrom: Bid, Ask and Transaction Prices in a Specialist Market with Heterogeneously Informed Traders (Journal of Financial Economics, 1985)
- SEC Staff: Equity Market Structure Literature Review Part II, High Frequency Trading
- Huh: Machines vs. Machines, High Frequency Trading and Hard Information (Federal Reserve FEDS 2014-33)
- SEC Final Rule: Disclosure of Order Execution Information (Release 34-99679)
How entries are sourced, calculated and updated: editorial methodology.