Adverse Selection

The core mechanism of adverse selection occurs when one party possesses material non-public information or a superior predictive signal not reflected in the current market price. This asymmetry allows the informed trader to buy below or sell above the asset's fundamental value. The uninformed counterparty, typically a market maker or passive liquidity provider, systematically loses to this informed flow. The degree of asymmetry is often proxied by the probability of informed trading (PIN), which estimates the fraction of orders originating from traders with private information.

Market makers defend against adverse selection by widening the bid-ask spread. The spread is not merely a transaction cost; it is an insurance premium against informed flow. The spread decomposes into three components:

• Order processing cost: The fixed cost of executing a trade.
• Inventory holding cost: The risk of holding a position until it can be unwound.
• Adverse selection cost: The expected loss to informed traders. When toxicity is high, the adverse selection component dominates, causing spreads to widen significantly. This is modeled by the Glosten-Milgrom model, which shows spreads increase with the proportion of informed traders in the market.

Order flow toxicity quantifies the degree to which incoming marketable orders are informed. A toxic order flow is one that consistently moves the market price against the liquidity provider after execution. The Volume-Synchronized Probability of Informed Trading (VPIN) metric measures toxicity by tracking volume imbalances in short time buckets. High VPIN values signal that market makers are being adversely selected and often precede volatility events. Liquidity providers monitor toxicity in real-time to dynamically adjust quotes or withdraw from the market entirely.

The empirical signature of adverse selection is adverse post-trade price movement. When a liquidity provider sells to an informed buyer, the price immediately rises, causing the provider to miss the gain. Conversely, buying from an informed seller results in an immediate price decline. This is measured by the effective spread, which compares the execution price to the midpoint of the bid-ask spread at a future time horizon (e.g., 5 seconds or 1 minute post-trade). A large effective spread relative to the quoted spread indicates severe adverse selection.

In electronic limit order books operating under price-time priority, adverse selection manifests as queue position risk. A liquidity provider placing a resting limit order at the best bid or offer is exposed to being 'picked off' when new information arrives. Informed traders will aggressively sweep all available liquidity at a stale price before the passive provider can cancel and reprice their order. This forces market makers to invest in low-latency infrastructure and quote cancellation logic to manage their queue position dynamically.

Venues and algorithms deploy specific defenses to neutralize adverse selection:

• Speed bumps: Intentional microsecond delays (e.g., IEX's 350-microsecond coil) that neutralize latency arbitrageurs attempting to pick off stale quotes.
• Minimum resting times: Requiring orders to remain in the book for a set duration before cancellation, preventing fleeting orders that exploit fleeting information.
• Anti-gaming logic: Smart order routers randomize order timing, size, and venue selection to prevent predatory algorithms from detecting and front-running a large institutional order's execution pattern.
• Midpoint peg orders: Hiding orders at the midpoint of the NBBO to avoid displaying a price that can be picked off by informed flow.

A quantitative metric that measures the probability that incoming marketable orders are informed—meaning the counterparty possesses superior information about the asset's true value. High toxicity signals that a market maker is likely on the wrong side of the trade.

• VPIN (Volume-Synchronized Probability of Informed Trading) : A real-time toxicity metric that updates with volume buckets rather than clock time
• Bias-Free Toxicity: Measures whether buy or sell order flow is systematically more informed
• Market makers use toxicity signals to dynamically widen spreads or reduce quoted depth

The risk that a market maker accumulates a directional position after repeatedly being picked off by informed traders. Unlike adverse selection on a single trade, inventory risk compounds across multiple transactions.

• Skewed inventory: When a market maker is disproportionately long or short due to toxic flow
• Hedging costs: The expense of offsetting unwanted inventory in the broader market
• Inventory mean reversion: Strategies that adjust quotes to encourage flow that reduces position imbalance
• Market makers manage this by adjusting bid-ask spreads based on current inventory levels

The ordinal rank of a resting limit order within the price-time priority stack at a specific price level. Adverse selection risk increases the longer an order sits at the front of the queue, as informed traders target visible, executable liquidity.

• Queue priority: Earlier timestamps execute first when contra-side flow arrives
• Last look: Some venues give liquidity providers a final opportunity to reject trades
• Queue jumping: Paying higher fees or using special order types to gain priority
• Adverse selection is highest for orders at the front of the queue in fast-moving markets

A non-displayed order type that automatically reprices to the midpoint of the NBBO, allowing traders to access liquidity without revealing their trading intention. This is a primary defense against adverse selection in dark pools.

• Pegged to primary: Tracks the midpoint of the listing exchange only
• Pegged to market: Tracks the consolidated midpoint across all venues
• Minimum quantity: Requires contra-side orders to meet a size threshold before execution
• Midpoint orders avoid the spread capture that informed traders exploit