How do Extreme Price Movements End? with J. Brogaard and J. Zhang
We test competing theories on liquidity dynamics during extreme price movements (EPMs). Our findings indicate that market makers strategically allow for price pressures and earn compensation from pricing errors. As a result, liquidity provision intensifies towards the end of an average EPM. This goes counter to a widespread concern that market making constraints cause the deterioration of liquidity as EPMs develop. Finally, we demonstrate that limit order book dynamics during EPMs is in line with a socially beneficial equilibrium.
Ransomware Activity and Blockchain Congestion, forthcoming at the Journal of Financial Economics
I examine blockchain congestion episodes caused by more than 4,400 triggers for ransomware attacks over a two-year period. When demand for settlement exceeds blockchain capacity, blockchain users engage in fee competition to prioritize their transaction settlement. A typical surge in ransomware activity causes transaction fees to increase by 2.1% and up to 28% in extreme cases. Consistent with theory literature, some users forego blockchain settlement when transaction fees increase. An event study around an extreme spike in ransomware activity supports the findings of the main analysis.
Every Cloud Has a Silver Lining: Fast Trading, Microwave Connectivity and Trading Costs, with A. Shkilko, forthcoming at the Journal of Finance
The modern marketplace is characterized by speed differentials, with some traders being fractions of a second faster than others. Theory models suggest that such differentials may have both positive and negative effects on liquidity. We examine these effects empirically by studying a series of exogenous weather-related episodes that temporarily remove the speed advantages of the fastest traders by disrupting their microwave networks. The disruptions are associated with lower adverse selection, lower trading costs and lower volatility. Limit orders that improve the best quotes are submitted more frequently. The results are consistent with theory models suggesting that speed differentials result in greater adverse selection of liquidity suppliers.
Unfiltered Market Access and Liquidity: Evidence from the SEC Rule 15c3-5, with B. Chakrabarty, P.K. Jain and A. Shkilko, forthcoming at the Management Science
In November 2011, the SEC implemented the final provision of Rule 15c3-5 curbing unfiltered market access. The provision mandated that brokers verify their clients’ order flow for compliance with credit and capital thresholds before routing to market centers. We find that the new checks introduce latency to order flow and force some latency-sensitive strategies out of the market. As a result, liquidity providers are better able to revise their quotes in response to new information, adverse selection declines and liquidity improves. Consistent with the notion that the market for liquidity provision is competitive, our results show that the benefit of lower adverse selection is transferred entirely to liquidity demanders in the form of lower trading costs. We test for but do not find evidence for more direct effects such as counterparty risk reduction or binding effects of capital limits.
High-Frequency Trading and Extreme Price Movements, with J. Brogaard, A. Carrion, T. Moyaert, R. Riordan and A. Shkilko, 2018, Journal of Financial Economics 128, 253-265
Are endogenous liquidity providers (ELPs) reliable in times of market stress? We examine the activity of a common ELP type – high frequency traders (HFTs) – around extreme price movements (EPMs). We find that on average HFTs provide liquidity during EPMs by absorbing imbalances created by non-high frequency traders (nHFTs). Yet HFT liquidity provision is limited to EPMs in single stocks. When several stocks experience simultaneous EPMs, HFT liquidity demand dominates their supply. There is little evidence of HFTs causing EPMs.
Factor Models for Binary Financial Data, with F. Perez and A. Shkilko, 2015, Journal of Banking and Finance 61, S177-S188
Researchers are often interested in modeling binary decisions made by firms (e.g., the yes or no decisions to split the shares, initiate a dividend, or acquire another firm) as functions of economy-wide variables (common factors). Although factor models for continuous dependent variables are used widely, the toolkit of a financial researcher does not contain a generally accepted methodology that allows estimating factor models for binary dependent variables. In this paper, we study such a methodology. Using simulations, we identify data characteristics that allow for reliable estimates of factor parameters and conclude that the methodology is appropriate for the panel datasets of the type often used in finance. As an illustration, we use the methodology to address a currently debated issue of common factors in firms’ decisions to split their shares.