The Sound of Silence

• Date: 01 May 2014
• DOI: http://doi.org/10.1111/fire.12032
• Published date: 01 May 2014
• Author: Mohsen Saad,Jeffrey H. Harris

The Financial Review 49 (2014) 203–230

The Sound of Silence

Jeffrey H. Harris∗

American University

Mohsen Saad

American University of Sharjah

Abstract

Using comprehensiveelectronic data collected directly from NASDAQ systems, we assess

the impact of changes in electronic message traffic on predicting short-term changes in prices,

spreads and quoted depth levels. We document evidence that message traffic at, and nearby,

the inside quotes predicts upcoming price and quoted depth changes as much as 75 seconds

in advance. Controlling for the time series properties of silent information, past price, volume,

electronic communication network volume, time-of-day, and firm-specific fixed effects, we

find that message traffic is strongly related to short-term returns. Our results demonstrate that

modern electronic trading systems can be employed by high-frequency traders to effectively

forecast short-term market conditions.

Keywords: finance, stock market

JEL Classifications: G14, G17, G18

∗Corresponding author: Kogod School of Business, American University, 4400 Massachusetts Avenue,

NW,Washington, DC 20016; Phone: (202) 885-6669; Fax:(202) 885-1992; E-mail: jharris@american.edu.

Wethank an anonymous referee, Kirsten Anderson, Tom Bates, Shane Corwin, Jay Coughenour, Michael

Goldstein (the guest editor), Oliver Hansch, Tim McCormick, George Wang, seminar participants at the

2005 Financial Management Association, the U.S. Commodities Futures TradingCommission (CFTC) and

brown bag seminar participants at the University of Delawarefor their helpful comments and suggestions.

We are especially grateful to Claude Courbois, Frank Hatheway, and Jeff Smith at the NASDAQ Stock

Market for their insightful comments on this work. All remaining errors are our own.

C2014 The Eastern Finance Association 203

204 J. H. Harris and M. Saad/The Financial Review 49 (2014) 203–230

“...People talking without speaking,

People hearing without listening....”

Paul Simon (1964), “The Sound of Silence,” Wednesday morning, 3 am

1. Introduction

Technologyhas dramatically transformed the world’s financial markets.1Indeed,

new trading technology has fueled competition and created opportunities for high-

frequency traders to link directly with markets electronically. While open-outcry pits

and physical floor traders shouted messages and thereby relayed information audibly,

high-frequency traders use electronic signals coupled with trading algorithms to

execute orders silently, with minimal human intervention.

While Coval and Shumway (2001) examine the effects of audible noise from

futures pits on the costs of transacting bond futures, in this paper we test whether

changes in “silent” electronic message traffic on an equity market can predict very

short-term market conditions, including price and liquidity changes. This effort is

analogous to those of high-frequency traders who also attempt to use silent message

trafficto trade profitably. Wefind strong evidence that de-trended and de-seasonalized

changes in silent message traffic(of many types) on equity markets are able to forecast

short-term market conditions (within one minute). This predictive power also applies

to estimations in a simultaneous equation framework and in out-of-sample tests,

suggesting that our results are robust for practical application to high-frequency

trading strategies.

Our message traffic,although silent, includes visible signals, such as inside quote

and depth updates, quote and depth updates at nearby prices from both electronic

communication networks (ECNs) and NASDAQ’s (nowdefunct) SuperMontage sys-

tem.2In addition, we examine nonvisible messages including total message arrivals

and changes to reserve depth from SuperMontage. We show that different messages

(changes to price updates, depth updates at and away from the inside, as well as re-

serve size message traffic) significantly presage NASDAQ midpoint and trade price

changes. Moreover, visible message trafficis also useful in predicting future changes

in market quoted depth and effective spreads only for larges trades.

The SuperMontage and ECN data we employ are special, in that our 2003

data are more comprehensive than current data (such as NASDAQ’s current ITCH

1NASDAQ (now NASDAQ OMX), the world’s first electronic stock market, has grown to become (by

some measures) the largest stock exchange in the world: since 2010, NASDAQ-listed volume has averaged

more than two billion shares per day, with NASDAQ systems executing more than three billion shares

per day. The New York Stock Exchange (now NYSE Euronext, or NYSE) spent more than $3 billion on

technology in the decade prior to 2007.

2SuperMontage provided the first comprehensive data with common timestamps for all types of message

arrivals, a crucial component for evaluating short-term market dynamics. Data integrity from the system

is high. NASDAQ’s ITCH data feed contains much of this same information for more recent periods.