Unscheduled News and Market Dynamics

• Published date: 01 December 2018
• Date: 01 December 2018
• DOI: http://doi.org/10.1111/jofi.12717
• Author: JÉRÔME DUGAST

THE JOURNAL OF FINANCE •VOL. LXXIII, NO. 6 •DECEMBER 2018

Unscheduled News and Market Dynamics

J´

ER ˆ

OME DUGAST∗

ABSTRACT

When unscheduled news arrives, investors react with a stochastic delay yet still may

exploit new information. In this context, I study the equilibrium dynamics of limit

order markets. Continuous idiosyncratic liquidity shocks result in trades on both

sides of the order book. News therefore arrives at random times. Following news,

order flows become unbalanced and market depth is consumed, leading to positive

covariance between price variability, trading volume, and order book unbalances.

Holding the unconditional price variability constant, news frequency has a negative

effect on both market depth and the variability-volume covariance.

IN MODERN FINANCIAL MARKETS,MOST stocks, as well as bonds and derivatives,

are traded via electronic limit order books. In such a trading venue, any par-

ticipant can provide new quotes, with limit orders, or trade against existing

ones, with market orders. Limit orders allow the investor to improve the exe-

cution price but potentially expose her to a winner’s curse problem when public

information arrives. In this paper I explore what underlies the placement of

limit and market orders when information is released, at random times, by

unscheduled news (e.g., on Bloomberg, Reuters Newswire, etc.). I also exam-

ine how news arrivals, and their frequency, impact market liquidity, trading

volume, and volatility. This paper is the first to tackle these questions in the

context of an equilibrium model in which all agents are rational and can use

fully dynamic strategies that involve market orders as well as the placement

and cancelation of limit orders.

∗J´

erˆ

ome Dugast is with Universit´

e Paris-Dauphine, PSL Research University, CNRS, UMR

7088, DRM, Finance, 75016 Paris, France. I am grateful to the Editor, Bruno Biais; the anony-

mous Associate Editor; and two anonymous referees for constructive feedback. I am indebted to

my advisor, Thierry Foucault, for his unconditional support, stimulating discussions, and helpful

suggestions. I am also grateful to William Cong, WilliamFuchs, Ming Guo, Johan Hombert, Sophie

Moinas, Michael Moore, Christine Parlour, Ioanid Rosu, and Pierre-Olivier Weill for advice and

detailed feedback. I further thank seminar and conference participants at LBS Transatlantic Doc-

toral Conference 2012, Frontiers of Finance 2012 Conference, 4th INSEAD PhD Finance Workshop,

HEC Lausanne, Banque de France, ToulouseSchool of Economics, WFA Annual Meeting 2013, 13th

SAET Conference, ESSEC, Universit¨

at Z¨

urich, Market Microstructure: Confronting Many View-

points #3, TSE Tradingand Post-Trading Conference, Rotterdam School of Management, Frankfurt

School of Finance and Management, Universit´

e Paris-Dauphine, Luxembourg School of Finance,

and EFA Annual Meeting 2016. I have read the Journal of Finance’sdisclosure policy and have no

conflicts of interest to disclose.

DOI: 10.1111/jofi.12717

2537

2538 The Journal of Finance R

My model of a limit order market builds on the over-the-counter (OTC) mar-

ket framework of Duffie, Gˆ

arleanu, and Pedersen (2005,2007), hereafter re-

ferred to as “DGP.” My model differs from DGP in that investors trade on a

centralized limit order market, and it introduces unscheduled news arrival,

which is novel in the literature. In the spirit of DGP, all sources of uncertainty

are modeled as Poisson processes, in particular, the time at which news may

arrive and the delay in each investor’s reaction following that news. These

modeling choices offer tractability and allow me to describe the joint dynam-

ics around news arrival of the order book, order flows, trading volume, and

price.

Prior to the arrival of news, the market is in a steady state and trading takes

place. Both market and limit orders are used in equilibrium. As in DGP, at

every instant a constant fraction of the investors’ pool incurs a shock in private

valuation and thus needs to trade. From the steady-state condition it follows

that a constant fraction of these investors trades with market orders and the

remaining fraction places limit orders. As a result, the flows of buy and sell

market orders are continuous and nonvarying. Under these circumstances, the

equilibrium bid-ask spread must be equal to its minimal value of one tick.

Indeed, if there were an available trading price between the best bid and ask

prices, then an investor who places a market order in equilibrium would prefer

to submit a limit order at this price—and that order would be immediately

executed by the contemporaneous flow of market orders. In equilibrium, such

a profitable deviation cannot exist. Nowadays, a one-tick bid-ask spread is

empirically realistic. For instance, O’Hara, Saar, and Zhong (2015) document

that stocks with a large tick size relative to their price exhibit a one-tick bid-ask

spread more than 50% of the time.

In my model, investors face a trade-off between limit and market orders.

Consider an investor who wants to buy. If he uses a buy limit order, upon

execution he saves the one-tick bid-ask spread compared to a buy market order.

However, the order incurs an execution delay that results in adverse selection

risk if news arrives in the meanwhile. Indeed, the order is exposed to picking-off

risk following negative news and to nonexecution risk following positive news,

as in Foucault (1999). Yet agents have the possibility of reacting to news. With

some probability, in the case of negative news the former investor can cancel

his buy order and avoid being picked off, or, in the case of positive news, can

cancel his order, send a buy market order before the price change, and avoid

nonexecution. Thus, the possibility of reacting to news mitigates the risks

associated with limit orders. This reaction to news is also beneficial to market

order users. Indeed, if the investor uses a buy market order, then following

negative news, with some probability he can resell his asset share and wait

for the price to adjust down to buy again, while following positive news, the

investor keeps the asset. Importantly,in the case of negative news, regardless of

whether the investor has used a buy limit order or a buy market order,outcomes

are similar: ex post, he has overpaid for the asset unless he has managed to

react and has either canceled his limit order or undone his previous trade.

Thus, these outcomes cancel out in the trade-off between limit and market

Unscheduled News and Market Dynamics 2539

buy orders. The trade-off is symmetric for sell orders. Overall, compared to a

market order, a limit order suffers from a relative adverse selection cost that

arises because of nonexecution risk due to news arrival.

Limit orders queue before execution. In equilibrium, the limit order execution

speed adjusts so that likely saving the one-tick bid-ask spread compensates for

the limit order’s relative adverse selection cost. If this cost increases, then the

limit order execution speed must increase, so that investors remain indifferent

between market and limit orders. As this execution speed increases, so does

the flow of limit orders leaving the book, which reduces market depth.

When news hits the market, investors observe it only after some delay. This

generates a transition phase during which investors react to the news by can-

celing their own stale limit orders and picking off those remaining in the book.

Once all stale limit orders have been canceled or executed, the new equilib-

rium bid and ask prices are established, which ends the transition phase. The

greater the market depth prior to news arrival, the more stale limit orders

must be canceled or executed and the longer the transition phase.

More frequent news arrivals imply a larger adverse selection cost for limit

orders. This results in lower depth and shortens the transition phase. During

the transition phase, the flow of limit order cancelations is proportional to the

market depth, and thus this flow decreases as news arrives more frequently.At

the same time, however, the flow of market orders does not depend on the news

arrival frequency. So in the transition phase, the trades’ share of the orders’

outflow (i.e., trades +cancelations) increases when the news arrival frequency

increases.

In equilibrium, when news hits the market, trading volume goes up, as

stale limit orders get picked off. Thus, the model offers a microfoundation for

the time-series correlation between volatility and volume. Investors, however,

rationally anticipate the cost of stale limit orders. Hence, in the cross-section, as

the frequency of news increases, limit orders are deterred and trading volume

is reduced.

Following the arrival of positive news, investors hit stale limit sell orders.

Consequently, there is more activity (cancelations and executions) on the ask

side than on the bid side of the book, and market depth is depleted on the

ask side relative to the bid side. This continues until the end of the transition

phase, at which point new and higher ask and bid quotes are established.

Symmetric effects take place after negative news. Thus, empirically, a flurry

of activity on the ask (bid) side of the book predicts an increase (decrease) in

prices.

My paper contributes to theoretical literature on dynamic limit order mar-

kets and is the first to my knowledge to study the effect of unscheduled news.

This model has the advantage of tractability without being too restrictive on

the agents’ action set. The model is tractable because (i) I use a setup `

ala

DGP in which order flows are continuous and (ii) I focus on equilibria in which

aggregate preferences are in a steady state, as doing so reduces the dimension-

ality of the state space. Biais, Hombert, and Weill (2014) offer a dynamic limit

order market model that also builds on DGP. Their paper differs from mine in