Research

Abstract: Over 20% of U.S. options volume trades on physical floors, almost exclusively in S&P 500 index options. We exploit the COVID-19 closure of the Cboe floor to identify the impact of floor trading. The floor attracts complex trades, particularly the riskiest trades. During the closure, Cboe activated electronic auctions to execute complex orders. When electronic auctions were withdrawn, the cost of complex trades triples, while costs for simple trades are mostly unchanged. During the closure, the riskiest trades become much less prevalent. The floor facilitates the most difficult trades, but raises the cost of less difficult ones. 

Abstract: We study the optimal length of an FX benchmark fixing window. A client trades with a dealer at a benchmark averaged over the window, and the dealer hedges before and during it under permanent and temporary impact. A longer window dilutes temporary impact but raises the client's benchmark risk and the dealer's tracking risk, which competitive fees return to the client, so the client-optimal window is finite and interior. The result survives endogenous pre-hedging, window-dependent impact, transient impact, and dealer competition. Calibrated to GBP/USD tick data, the optimum is state dependent, and the five-minute standard is a robust cross-regime compromise.

Abstract: Payment for order flow and its auctions rest on market fragmentation. We separate the auction from this structure using two markets on one index. Options on the S&P 500 ETF (SPY) are multi-listed and intermediated through payment for order flow, while options on the S&P 500 index (SPX) trade only on Cboe, without fragmentation or order-flow payments. When Cboe introduced auctions for small SPX orders in February 2021, effective spreads fell by four to six cents and price improvement rose, while market-maker revenue turned negative. SPY did not change. Auctions raise execution quality through competition, but only where the spread exceeds one tick.

Abstract: We study whether concentrating order flow in fragmented bond markets improves market quality. Centralisation concentrates order flow on designated venues and improves public information efficiency by 4.3 percentage points for treated bonds. The improvement is driven primarily by stronger cross-venue price synchronization, while variance ratios and autocorrelation also indicate more efficient pricing. Price impact results are mixed, suggesting that market-design interventions can improve information incorporation while generating transition costs as liquidity reallocates across venues. The results show that centralising a decentralised bond market does some good: it improves public price discovery, but not without frictions.

Abstract: We study how an interstate war affects equity market functioning using the Russian invasion of Ukraine as a precisely timed shock. Using millisecond-level trades and quotes, we show that liquidity deteriorates sharply and disproportionately for firms headquartered closer to the conflict. Markets do not anticipate the exact timing of the invasion. Abnormal quoted spreads for nearby and distant firms are near zero in the closing hours of the days before, but at the market open following the invasion, quoted spreads for nearby firms jump by about 6.6 standard deviations. In difference-in-differences estimates, first-degree neighbors experience higher quoted and effective spreads, and trading activity. Continuous distance measures show even larger effects, with spreads increasing by 16-24 bp and trading intensity by 25-37% for more exposed firms. By contrast, war-induced price changes and firm-specific volatility explain little of the liquidity response.

Abstract: We study how fee pricing and allocation rules affect market quality in decentralized payment systems. Using Ethereum's transition from first-price to uniform-fee pricing (EIP-1559) as a natural experiment and Bitcoin as a control in a difference-in-differences design, we find that algorithmic fee and elastic allocation mechanisms reduce fee volatility but increase fee dispersion and inflation, with average transaction costs rising by approximately 120%. Tips paid to miners fell by about one third, yet miner per-transaction payoffs rose by about 75%. On the welfare side, confirmed transactions on Ethereum declined by 22% relative to Bitcoin, dollar value settled on-chain dropped by 40%, smart-contract deployment fell by nearly half, and network congestion dropped by 44%. A placebo test using Ethereum's Merge to Proof-of-Stake produces no comparable discontinuities. Replacing the problematic first-price auction with a uniform posted fee traded lower price uncertainty for greater rent extraction and reduced aggregate network welfare.