Tomy Lee

I examine the impact of cross-venue latency on market quality using a model of informed trader competition in a fragmented market. As cross-venue latency decreases, liquidity and price discovery improve while the expected profits of informed traders decline. Moreover, a fall in the latency of one venue can harm liquidity at the other venue. An extension predicts that, as the informed traders consolidate or outsource trading, benefits of shorter cross-venue latency are attenuated and its harmful effects intensify. My model generates testable predictions about the effects of changes in cross-venue latency on market quality.

Over-the-counter (OTC) trading thrives despite competition from exchanges. We let OTC dealers cream skim from exchanges in an otherwise standard Glosten and Milgrom (1985) framework. Restricting the dealer's ability to cream skim induces ``cheap substitution'': some traders exit while others with larger gains from trade enter. Cheap substitution implies trading costs, trade volumes, and market shares are poor policy indicators. In a benchmark case, restricting the dealer raises welfare only if trading cost increases, volume falls, and OTC market share is high. By contrast, the restriction improves welfare when adverse selection risk is low. A simple procedure implements the optimal Pigouvian tax.

Conflicts of interest are inherent to banking conglomerates. Regulators increasingly manage these conflicts by enforcing China Walls-internal information barriers around key affiliates, dealers in particular. We map information sharing among dealers and funds using a near universe of foreign exchange transactions involving the Israeli Shekel to evaluate if today's China Walls are effectively enforced. Our difference-in-differences design compares the trading activities of affiliates and entirely unrelated firms around exceptionally large trades to detect information sharing. We document islands of informational autarky between dealers and their affiliate funds surrounded by a sea of information sharing: (1) The affiliate dealers and funds never trade and do not share information with each other. (2) The dealers and funds connected via trading relationships consistently share information, including on days when a dealer and its connected fund do not trade with each other. (3) Affiliates without China Walls intensely share information among themselves. From a back-of-the-envelope calculation, extending the China Walls to the non-walled affiliates would eliminate $16.1 billion in transactions, comprising 37% of their trades on the event dates with exceptionally large trades. Our results hold during crisis and noncrisis periods, and across granular cells of firm and asset characteristics. Our results reveal remarkable regulatory capacity to control information flows.

We introduce a tractable framework for bank runs with socially responsible depositors who care about nonpecuniary externalities. Although each depositor is vanishingly small, she rationally internalizes the impact of her withdrawal on the risk of bank failure. We apply this framework to examine the effects of divestment on financial gains and social surplus. Any divestment that positively assorts bank investment portfolios and depositor social responsibility generates financial gains. However, the divestment harms social surplus if depositors are insufficiently responsible. Our results explain the widespread adoption of exclusion and tilting among financial firms, and caution against their potentially harmful social impact.