Publications
Trade-time measures of liquidity, with Dan Bernhardt and Ryan J. Davies (2019), Review of Financial Studies 32(1): 126--179.
Dramatic microstructure changes in equity markets have made standard liquidity measures less accurate proxies of trading costs. We develop trade-time based liquidity measures that reflect per-dollar price impacts of fixed-dollar volumes. Our measures better capture institutional trading costs and better explain the cross-section of returns than do standard measures, especially in recent years. Despite improvements in measures of market quality, expected trading costs still have explanatory power for the cross-section of expected returns: we obtain monthly liquidity premium estimates of 5.3bp for expected returns and 2.4bp for risk-adjusted returns. Estimated premia rise after the financial crisis and remain high thereafter.
The Night and Day of Amihud's (2002) liquidity measure, with Dan Bernhardt, Thomas Ruchti, and Marc Weidenmier, The Review of Asset Pricing Studies,
Amihud's (2002) stock (il)liquidity measure averages daily ratios of absolute close-to-close return to dollar volume, including overnight returns. Our modified measure uses open-to-close returns, matching return and trading volume measurement windows. It is more strongly correlated with trading-cost measures (by 8--37%); moreover, it better explains cross-sections of returns, doubling estimated liquidity premia. Using non-synchronous trading near close, we show overnight returns are primarily information-driven: including them in Amihud's proxy for price impacts of trading magnifies measurement error, understating liquidity premia. Our modification helps wherever Amihud's measure is required. Our measures are publicly available for 1964-2019, and can be updated.
Selected working papers
The Information in Industry-Neutral Self-Financed Trades, with Zhi Da and Mitch Warachka.
We identify Industry-Neutral Self-Financed Informed Trading (INSFIT) by long only fund managers who possess a positive short-lived private signal and self finance informed stock purchases by selling an equivalent dollar amount of stock in the same industry. INSFIT, which constitutes less than 1% of trading, produces a cumulative abnormal return spread of nearly 0.90% over the subsequent ten days. INSFIT also precedes the release of positive public information. The prevalence of relative valuation as well as the need to maintain industry allocations and hedge industry exposure motivate INSFIT’s industry neutrality. Furthermore, INSFIT occurs more frequently among cash-constrained managers but is uncorrelated across fund managers. Although INSFIT involves relatively large dollar-denominated trades, transaction costs cannot account for its profitability.
Uncovering the Impacts of Endogenous Liquidity Consumption in Intraday Trading Patterns, with Dan Bernhardt.
We document new intraday trading patterns indicative of the key roles of endogenous trading responses of investors to variations in imperfectly-competitive liquidity provision. When measured in trade times of fixed dollar values, price impacts and volatility fall sharply from open to close, and as trading activity rises. We also document reversions in trade-time returns in inactive markets, and priced, heavily-forecastable, order flow imbalances in active markets. Standard calendar-time aggregation approaches conceal these primitive trading patterns by matching up overly-balanced signed-trade observations with large price movements in active markets. Once one controls for over-aggregation, calendar-time patterns align with trade-time patterns.
Are Short Selling Restrictions Effective?, with Andrew Bird, Stephen A. Karolyi, and Thomas Ruchti.
Despite strong theoretical predictions based on disagreement, limited empirical evidence has linked short selling restrictions to higher prices. We test this relationship using quasi-experimental methods based on Rule 201, a threshold-based policy that restricts aggressive short selling when intraday returns cross -10%. When comparing stocks on either side of the threshold in the same hour of trading, we find that the restriction leads to 8% lower short sale volume and 35 bps higher daily returns. These price effects do not reverse and are not associated with information events, suggesting that Rule 201 restricts short selling based on transient opportunities unrelated to fundamentals. Although these persistent direct effects align with policymaker objectives, we find evidence of offsetting transient spillover effects on peer stocks consistent with cross-stock substitution by short sellers.
A Test of Speculative Arbitrage: Is the Cross-section of Volatility Invariant, with Dan Bernhardt and Thomas Ruchti.
We derive testable implications of Kyle and Obizhaeva's (2016) notion of "bet invariance'' for the cross-section of trade-time volatilities. We jointly develop theoretical foundations of "no speculative arbitrage'' whose implications incorporate those of bet invariance. Our proposed test circumvents the unobservable nature of "bets.'' Utilizing a large sample of U.S. stocks post decimilization, we show that using realized volatilities rather than expected volatilities introduces noise that substantially biases the tests. This leads us to use estimates of normalized volatilities based on running 24 month windows. We find strong support for no speculative arbitrage at a moment in time, but not across time.
Selected works in progress
Auto-regressive Conditional Durations in modern financial markets, with Jiaying Gu.
Auto-regressive Conditional Duration models of Engle and Russell (1998) were designed to estimate the time durations between individual transactions. Their fundamental assumption for tractability of the log-likelihood functions is that transactions are i.i.d. The temporal correlations in individual transactions driven by strategic order splitting and the behaviors of high-frequency traders grossly violates the $i.i.d$ assumption. My other work, including "Trade-time based liquidity measures" highlight the ability of trade-time aggregation in addressing such temporal dependencies. We propose trade time intervals of fixed dollar volumes as inputs to ACD models to largely retrieve the i.i.d assumption. We do find that the residuals from fitting ADC models using these trade time intervals satisfy the ACD models' underlying assumptions. The goal of the project is to provide econometric insight about conditions under which the i.i.d assumption of ACD models is retrieved by aggregating temporally-dependent transactions over trade-time intervals.
Estimating a dynamic structural model of parasitic trading, with Dan Bernhardt and Alexie Boulatov.
This project structurally estimates a theoretical model of intraday trading to match the findings from my other paper "The dynamics of intraday trade time outcomes'' that (i) trading costs decline over the trading day; and (ii) trading costs decline as markets become more active. The existing theoretical models are able to match one set of these facts at the expense of violating the other. We develop a unifying framework that incorporates both evolving trading incentives over the trading day and strategic responses of investors to stochastic variations in extant liquidity. In our model, an uninformed institutional investor observes a position shock at open, and will execute "optimally'' to avoid price risk that may realize by close. However, parasitic traders (some HFT) will exploit his price impact near open, raising trading costs. By close, parasites must rebalance their accumulated positions driven by their trade against the institutional investor in earlier hours. Thus, liquidity improves in later trading hours. Extant liquidity at each point in the trading day features observable stochastic shocks, and traders react to these shocks by changing their intensity of endogenous liquidity consumption. We estimate our model using MCMC type approaches.