We empirically identify a group of stock-exchange accounts that profit from 11 years of earnings surprises. Their trading behavior is consistent with privately informed trading, yet they have negative and temporary price impacts. We then empirically identify a second group of accounts that have positive and permanent price impacts. The trading behavior of the second group is more consistent with trading on public information, and they trade the wrong way before earnings surprises. The behavior of both account groups contrasts with models that associate permanent price impact with privately informed trading.

Theory on high-frequency traders (HFTs) predicts that market liquidity for a security decreases in the number of HFTs trading the security. We test this prediction by studying a new Canadian stock exchange, Alpha, that experienced the entry of 11 HFTs over 4 years. We find that bid–ask spreads on Alpha converge to those at the Toronto Stock Exchange as more HFTs trade on Alpha. Effective and realized spreads for non-HFTs improve as HFTs enter the market. To explain the contrast with theory, which models the HFT as a price competitor, we provide evidence more consistent with HFTs fitting a quantity-competitor framework.

We model how securities dealers respond to regulations on leverage, position, and liquidity such as those imposed by the Basel III framework. The dealers respond by endogenously moving to make markets on an agency basis, matching buyers to sellers rather than taking client positions on the balance sheet. Agency-based market making creates a cost-risk tradeoff in which investor welfare declines but dealers become less risky. The costs to investors do not show up in all liquidity metrics: While asset prices exhibit greater price impact, bid-ask spreads do not change and trading volumes can even increase, which can help explain the varying findings from the empirical literature.

Limit order markets use a queuing system in which limit orders must wait in line to execute. We show that the queue position of a limit order influences its adverse selection risk and inhibits inventory risk management. Trade may worsen market maker risk sharing, unlike many protocols without queuing. We uncover a crowding-out effect: An inventory shock reduces liquidity provision by market makers later in the queue. Using futures data, we confirm both low risk sharing and the crowding-out effect. These two results imply a trade-off, as the queuing sequence that optimizes risk sharing decreases quoted depth up to 8.4%. • Queue position affects adverse-selection risk and inventory management. • Market-maker risk sharing may worsen due to queuing. • Inventory shocks reduce liquidity provision later in the queue. • Canadian futures data confirm low risk sharing and crowding-out effects. • Optimizing risk sharing lowers quoted depth by up to 8.4%.