Research

Most foreign-exchange transactions route through the U.S. dollar even when neither party wants dollars: a trade between two other currencies is executed as two dollar legs. That concentrates volume in dollar markets and keeps their spreads tight, but every cross-currency order pays twice. We ask whether a single pool that makes markets in many currencies at once can do better, and derive the condition under which it does. The answer turns on how correlated the currencies' returns are and how much of the order flow is cross-currency. The framework applies beyond FX to any market in which one asset serves as the common settlement asset, for instance tokenized stocks trading against cash.

Most foreign exchange transactions route through the U.S. dollar, even when the parties want to exchange two other currencies. Vehicle routing concentrates volume in dollar markets and lowers their trading costs, but a cross-currency trade pays fees and price impact on two dollar legs. We ask whether a single liquidity pool that makes markets in several currencies can lower aggregate trading costs. The designer chooses the capital weights across currencies and the fee on each pair. With competitive liquidity provision and aggregate fee revenue set optimally, the weights determine aggregate trading costs and each pair's price impact; the division of fees across pairs determines how those costs are shared by traders. We characterize the optimal weights and derive a closed-form condition under which some choice of weights yields lower aggregate costs than vehicle routing: the volume-weighted sum of pairwise return correlations and cross-currency flow intensities must be positive. The routing cost is a boundary value of the pool's cost function as its dollar share approaches one. We also characterize the fee schedules that deter entry by a dedicated dollar market and show that aggregate dominance does not guarantee robustness to entry. Foreign exchange is our main application, but the results apply to any set of assets traded against a common settlement asset.

Trading venues sell two things: execution and the market data their trading produces. Because a venue's data is only worth buying if enough other traders also subscribe, data fees gate access to a network externality. We model competition between an incumbent and an entrant venue and find three things. First, an entrant faces a coordination failure: nobody subscribing to its data is an equilibrium at any non-negative fee, and only a subsidy removes it. Second, the incumbent can deter entry by overpricing its own data above the revenue-maximizing level, which squeezes the fee the entrant can sustain and raises the critical mass it needs. Third, brokers who pass only part of the data cost to clients behave as if fees were even higher, and past a sharp threshold that agency friction alone forecloses the entrant. Paper coming soon.

We study whether automated market makers (AMMs), trading mechanisms prominent in crypto markets, can improve liquidity provision in equities, a question made salient by the SEC's Project Crypto and exchanges exploring tokenized trading. We develop an equilibrium model of AMM liquidity, derive the stock-specific fee that minimizes trading cost in closed form, and show it reflects the stock's adverse selection. The optimal fee scales with return volatility and the square root of order size relative to volume, consistent with the price-impact law. Calibrated to U.S. equities, AMM trading costs are 38–55 percent below half-spreads, implying billions in annual savings.