News Trading and Speed

• Author: JOHAN HOMBERT,IOANID ROŞU,THIERRY FOUCAULT
• Published date: 01 February 2016
• Date: 01 February 2016
• DOI: http://doi.org/10.1111/jofi.12302

THE JOURNAL OF FINANCE •VOL. LXXI, NO. 1 •FEBRUARY 2016

News Trading and Speed

THIERRY FOUCAULT, JOHAN HOMBERT, and IOANID ROS¸U

∗

ABSTRACT

We compare the optimal trading strategy of an informed speculator when he can

trade ahead of incoming news (is “fast”), versus when he cannot (is “slow”). We find

that speed matters: the fast speculator’s trades account for a larger fraction of trad-

ing volume, and are more correlated with short-run price changes. Nevertheless, he

realizes a large fraction of his profits from trading on long-term price changes. The

fast speculator’s behavior matches evidence about high-frequency traders. We predict

that stocks with more informative news are more liquid even though they attract

more activity from informed high-frequency traders.

High-frequency traders do not care if information is accurate or inaccurate.

They just want to know what is coming out on the market that might sway

public sentiment. So this is very different than traditional insider trading

[...]. This isall just about what might move the market, because they

are in and out in milliseconds. They don’t really care about the long-term

effects of the information.

Atty. Gen. Schneiderman’s speech, “High-Frequency Trading and Insider

Trading 2.0.”1

TODAY’S FINANCIAL MARKETS are characterized by an almost continuous flow of

“news.” Every quote update or trade in one asset (e.g., a stock index futures

∗Thierry Foucault, Johan Hombert, and Ioanid Ros¸u are with HEC, Paris. Wethank two anony-

mous referees, Ken Singleton (the Editor), Terrence Hendershott, Jennifer Huang, Leonid Kogan,

Pete Kyle, Stefano Lovo, Victor Martinez, Albert Menkveld, Han Ozsoylev, Marco Pagano, Tarun

Ramadorai, Vincent van Kervel, Dimitri Vayanos,Xavier Vives, and Mao Ye for their suggestions.

We are also grateful to finance seminar participants at Copenhagen Business School, Duisenberg

School of Finance, Frankfurt School of Management, University Carlos III in Madrid, EIEF, ES-

SEC, Lugano, IESE, INSEAD, Oxford, Paris Dauphine, University of Illinois, and University of

Leicester, as well as conference participants at the 2014 American Finance Association Meetings,

the UBC Finance Winter Conference, the 2014 SFS Finance Cavalcade, the 2014 SwissQuotes

conference, the 2013 Gerzensee Symposium, the 2013 European Finance Association meetings,

the 2012 NYU Stern Microstructure Meeting, the Newton Institute in Cambridge, CNMV Inter-

national Conference on Securities Markets, the 9th Central Bank Microstructure Workshop at

the ECB, the 12th Colloquium on Financial Markets in Cologne, the 2nd Market Microstructure

Many Viewpoints Conference in Paris, the 5th Paris Hedge Fund Conference, the High Frequency

Trading conference in Paris, and the Dauphine-Amundi Chaire in Asset Management Workshop

for valuable comments. The authors acknowledge financial support from the Amundi-Dauphine

Foundation Chair in Asset Management.

1March 18, 2014 (available at: http://www.ag.ny.gov/pdfs/HFT_and_market_structure.pdf).

DOI: 10.1111/jofi.12302

335

336 The Journal of Finance R

or an exchange-traded fund) is a source of information for pricing other assets.

Furthermore, traders increasingly rely on machine-readable text in tweets,

Facebook pages, blogs, newswires, economic and corporate reports, company

websites, etc., which greatly expands their information set because the arrival

rate of such news is very high.2

News thus plays an increasing role in shaping trade and price patterns in

financial markets. High-frequency trading is a case in point. High-frequency

traders’ (HFTs) strategies are diverse (see SEC (2014)): some specialize in mar-

ket making whereas others follow directional strategies, establishing positions

in anticipation of future price movements, mainly using aggressive (i.e., mar-

ketable) orders.3

Academic evidence suggests that high-frequency news plays an important

role in directional HFTs’ strategies.4First, HFTs’ aggressive orders anticipate

short-term price movements and contribute significantly to trading volume.

For instance, Brogaard, Hendershott, and Riordan (2014) find that HFTs’ ag-

gressive orders predict price changes over very short horizons and account

for 25% to 42% of trading volume depending on market capitalization (see also

Baron, Brogaard, and Kirilenko (2014), Benos and Sagade (2013), and Kirilenko

et al. (2014) for similar evidence). Second, HFTs’ aggressive orders are corre-

lated with news such as market-wide returns, quote updates, macroeconomic

announcements, E-mini price changes, and newswires items (see Brogaard,

Hendershott, and Riordan (2014) and Zhang (2012)). These observations sug-

gest that directional HFTs trade on soon-to-be-released information. However,

directional HFTs realize a large fraction of their profits on aggressive orders

over relatively long horizons (e.g., over the day; see Carrion (2013), table 5, and

Baron, Brogaard, and Kirilenko (2014), table 6). This last finding is difficult to

reconcile with the view that directional HFTs trade only on short-term price

reactions to news. Carrion (2013, p. 710) thus concludes that “models where

HFTs solely profit from very short-term activities [...] may beincomplete.”

In this paper, we propose a model of trading on news that explains the afore-

mentioned facts and generates new predictions, especially about the effect of

news informativeness on HFTs’ trading strategy, the sources of their profitabil-

ity (speculation on short-term versus long-term price movements), and liq-

uidity. We therefore contribute to the theoretical literature on high-frequency

trading, which thus far has not considered dynamic models of trading on news.

2See “Trading via Twitter”, Traders Magazine, June 2014. This article takes the example of

a prop trading firm that “everyday scans 400 to 500 million tweets looking for a breaking news

event.”

3SEC (2014) provides a survey of empirical findings on HFTs. This survey notes, on page 9, that:

“Perhaps the most noteworthy finding of the HFT dataset papers is that HFT is not a monolithic

phenomenon, but rather encompasses a diverse range of trading strategies. In particular, HFT

is not solely, or even primarily, characterized by passive market making strategies [ . . . ]. For

example, Carrion (2013) and Brogaard, Hendershott, and Riordan (2014) [ . . . ] find that more than

50% of HFT activity is attributable to aggressive, liquidity taking orders.” See also Hagstr¨

omer

and Nord´

en (2013) and Benos and Sagade (2013) for evidence that HFTs’ strategies are diverse.

4Numerous media articles also emphasize the importance of news in HFTs’ strategies. See,

for instance, “Computers that Trade on the News”, the New York Times, May 22, 2012 or “Speed

Traders Get an Edge”, the WallStreet Journal, February 7, 2014.

News Trading and Speed 337

Our model builds on Kyle (1985). One speculator and one competitive dealer

continuously trade while receiving a flow of signals about the payoff of a risky

asset (its “long-run” value).5The dealer’s signals are public information. We

interpret these signals as high-frequency news. In contrast, the speculator’s

signals are private and informative about the long-run value of the asset. Since

news is also informative about the long-run value of the asset, the speculator’s

signals can also be used to predict short-run price reactions to (the surprise

component of) news.6

We say that there is news trading if the speculator’s signals affect his trades

above and beyond their effects on the speculator’s estimate of the long-run

change in the asset. We show that news trading arises in equilibrium only

when the speculator is fast relative to the dealer, that is, if he can trade on his

forecast of short-run price movements before the dealer reacts to (or receives)

news. In this case, the speculator’s optimal position in the risky asset follows

a stochastic process with a drift proportional to the speculator’s forecast of

the long-run change in the asset value (as in Kyle (1985) and others) and an

instantaneous volatility proportional to the speculator’s forecast of news. This

volatility component is a novel feature of our model and is key for our pre-

dictions. This component drives short-run changes in the speculator’s position

while the drift component determines the long-run change in this position.

To develop intuition, suppose that the speculator’s latest signal is positive,

and yet his forecast of the asset payoff (which depends on his history of signals,

not just the latest signal) is lower than the asset price. In this case, the specula-

tor expects the price to increase in the short-run, due to news arrival (because

the speculator’s signal is positively correlated with news), but to decrease in

the long run. This calls for two different trades: a buy in anticipation of the

short-run price increase and a sell in anticipation of the longer-run price de-

cline. The drift component of the speculator’s position is his desired trade given

his estimate of the long-run price change, while the volatility component is his

desired trade given his forecast of impending news. The speculator’s actual

trade is the sum of these two—possibly conflicting—desired trades.

The volatility component always swamps the drift component in explaining

short-term variations in the speculator’s position. Thus, short-run changes in

the speculator’s position are driven by news, that is, the speculator trades in

the direction of incoming news. However, over a longer period of time, the

speculator’s position changes in the direction of his long-run forecast of the

asset value. Hence, in the previous example, the speculator buys the asset just

ahead of news arrival, even though he estimates the asset to be overpriced

relative to its long-run value; then, in the longer run, he sells the asset to

exploit this mispricing. Hence, when he is fast, the speculator trades on what

moves prices in the short run but he also cares about the long-run implications

of his information.

5“Long-run” in our model should be interpreted as, say,an hourly or daily horizon. Forecasts at

this horizon are long-run relative to forecasts of price changes over the next second.

6The model nests the particular case in which the speculator can perfectly forecast news. This

corresponds to the case of advance access to news content.