I study how arbitrage affects liquidity by analyzing several billion trades in the American Depositary Receipt (ADR) market from 2001 to 2016. Price deviations persist, on average, for 12 min, and mainly arise because of price pressure. Impulse response functions estimated at 1 min intervals indicate that a positive shock to arbitrage—simultaneous trades of the ADR and the home-market share in the opposite direction—decreases deviations and bid-ask spreads. I confirm these findings by exploiting institutional details that create exogenous variation in the impediments to arbitrage across days. Overall, these results suggest that arbitrage decreases price pressure and provides liquidity.

The capital asset pricing model (CAPM) performs poorly overall, as market risk (beta) is weakly related to 24-h returns. This is because stock prices behave very differently with respect to their sensitivity to beta when markets are open for trading versus when they are closed. Stock returns are positively related to beta overnight, whereas returns are negatively related to beta during the trading day. These day-night relations hold for beta-sorted portfolios and individual stocks in the US and internationally as well as for industry and book-to-market portfolios and cash flow and discount rate beta-sorted portfolios. In addition to the change in slope of returns with respect to beta, the implied risk-free rate differs significantly between night and day. Consistent with this, returns on US Treasury futures differ significantly between night and day.

Using limit order books across all US exchanges, we show that while liquidity for small orders (e.g., the quoted and effective spreads) decreases, liquidity for large orders (e.g., the cumulative depth and the price impact of multiple trades) improves after the implementation of the Tick Size Pilot Program. We find significant spillover effects on liquidity for small and large orders that extend beyond the top of the book. Finally, we show that the pilot program results in an improvement in pricing efficiency, an increase in trade size, and a decrease in the number of trades.