Working Paper

[1] Cost of Market Fragmentation (Job Market Paper)

Abstract: Traders use Intermarket Sweep Orders (ISO) to trade "as-if" the market were single-venued and avoid the fragmented environment. Using a sample of over 2,600 securities over the period January 2019 to April 2021, I find that traders are willing to pay a premium of roughly 40% of the effective spread to trade with ISOs. The premium is explained by market fragmentation as measured by an exchange-based Herndahl{Hirschman Index, fraction of off-exchange trade, and dispersion in realized variation (RV) across exchanges. A 1% increase in each of these measures are associated with a 1.28%, 0.29%, and 0.18% rise in the premium respectively. This relationship is very robust, it is greater for securities traded across more exchanges; it persists across different measures, methodologies, and sub-samples.

[2] The Profitability of Liquidity Provision (with Lingyan Yang)

Abstract: By tracking the cumulative inventory position of all passive liquidity providers in the US equity market and matching each position with its offsetting trade, we construct a measure of profits to liquidity provision (realized profitability) and assess how profitability varies with the average time to offset. Using a sample of all common stocks from 2017 to 2020, we show that there is substantial variation in the horizon at which trades are turned around even for the same stock. As a mark-to-market profit, the conventional realized spread—measured with a prespecified horizon—can deviate significantly from the realized profits to liquidity provision both in the cross-section and in the time series. We further show that, consistent with the risk-return tradeoff faced by liquidity providers as a whole, realized profitability is low for trades that are quickly turned around and high for trades that take longer to reverse.

Work In Progress

[3] Source of Gain from Vertical Integration (with Lingyan Yang)

[4] Market Making Horizon and Financial Market Stability (with Lingyan Yang)