Asymmetric news responses of high‐frequency and non‐high‐frequency traders

• Date: 01 August 2019
• Published date: 01 August 2019
• Author: S. Sarah Zhang
• DOI: http://doi.org/10.1111/fire.12200

DOI: 10.1111/fire.12200

ORIGINAL ARTICLE

Asymmetric news responses of high-frequency and

non-high-frequency traders

S. Sarah Zhang

Alliance Manchester Business School, University

of Manchester, Manchester, UK

Correspondence

S.Sarah Zhang, Alliance Manchester Business

School,University of Manchester, Booth Street

East,Manchester M15 6PB, UK.

Email:sarah.zhang@manchester.ac.uk

Abstract

Using NASDAQ trade and Reuters news data, I show that the

response of aggressive non-high-frequency traders (nHFTs)to news

is stronger than that of aggressive high-frequency traders (HFTs).

Classifying news into quantitative (“hard”) and less quantitative

(“softer”) news, the trading response of aggressive nHFTs to softer

news exceeds HFTs’ response. Positive news elicits greater return

and nHFT responses than negative news during the 2008 financial

crisis period. As this phenomenon persists even after excluding the

2008 short-sale ban, the results support the hypothesis of nHFTs

exhibiting stronger asymmetric responses during crisis periods.

KEYWORDS

financial crisis, high-frequency trading, information processing,

news

JEL CLASSIFICATIONS

G10, G14

1INTRODUCTION

Thelast decade has witnessed a significant increase in the use of computer algorithms in financial markets (cf. Deutsche

Bank, 2016). High-frequency traders (HFTs)1utilizespecific groups of computer trading algorithms and contribute to

price discovery through the application of different information processing strategies. While some people claim that

computerized traders follow complexand sophisticated trading strategies that surpass human traders in “scale, speed,

and complexity,”2others argue that HFTs and algorithmic traders (ATs) merely focus on simplistic strategies and do

not possess the equivalent abilities to human traders for processing complex information.3Regulatory concerns, such

as those of the U.S. Securities and Exchange Commission (SEC) and the European Securities and Markets Authority

(ESMA), focus on the effects of HFTson market quality and price discovery (cf. ESMA, 2011; SEC, 2010).

To analyze the regulatory concerns related to price discoveryand the question of whether HFTs are better able

to process complex information than non-high-frequency traders (nHFTs),I investigate the role of HFTs and nHFTsin

1HFTis a subcategory of AT,which is commonly defined as the use of computer algorithms to support the trading process (cf. Hendershott, Jones, & Menkveld,

2011).HFTs distinguish themselvesfrom other groups of traders through their use of high-speed trading and information processing, their high trading volume,

andtheir sophisticated algorithms for use in a proprietary capacity rather than for agency purposes (see SEC, 2010).

2See“The Irreversible Rise of the Investing Machines,” FT View, 2018 (https://www.ft.com/content/e1def550-f600-11e7-8715-e94187b3017e).

3See“Human Traders Can Still Beat Computers,” FT Opinion Smart Money, 2015 (https://www.ft.com/content/dc9e3dbc-57c9-11e5-a28b-50226830d644).

Financial Review.2019;54:451–475. wileyonlinelibrary.com/journal/fire c

2019 The Eastern Finance Association 451

452 ZHANG

processing stock-specific news information. Using an HFT data set from NASDAQ and news data from Reuters News-

Scope Sentiment Engine (RNSE), I provide insight into the intradaynews responses of stock returns and net trading of

HFTs and nHFTs. I further analyze whether responses differ according to news types, such as “hard” quantitative or

“softer” less quantitative news, and news direction, such as positive or negative news.

The results show that stock-specific news has significant effects on stock returns and net trading and that the

strength of response of aggressive nHFTs (liquidity-demanding traders initiating a transaction bysubmitting a mar-

ketableorder) to news releases exceeds that of aggressive HFTs.Classifying news announcements based on their quan-

titative (hard) and less quantitative (softer) content, I show that nHFTsrespond more strongly to softer news. There is

some evidence of a more pronounced response of nHFTsto positive rather than negative news during financial crisis

periods, which is in contrast to previous findings in the literature (e.g.,Tetlock, 2007). The analysis of time variation of

the effect in different periods during the 2008–2009 financial crisis reveals an enhanced asymmetry effect during the

peak of the financial crisis in 2008, supporting the hypothesis of differential behavior in recession periods by Garcia

(2013). As the short-sale ban of 2008 cannot explain those effects, thedifferential processing and responses of nHFTs

to positive and negative news are the more likelycause of an asymmetry effect rather than short-selling restrictions.

The results have important implications for our understanding of information processing and price discoveryin the

U.S. stock market. Theyserve to alleviate regulatory concerns that HFTs may initiate an overreaction to news. In con-

trast, nHFTsreact more to news than HFTs and nHFTsappear more prone to asymmetric responses. Furthermore, this

paper shows that nHFTs and HFTs specialize in different types of information releases, and HFTs provide enhanced

liquidity after news releases, which facilitates price discovery.This study is the first to document HFT behavior around

stock-specific news releases as well as the differential responses of HFTs and nHFTs to different types of news. The

paper's contributions are threefold.

First, I contribute to the growing body of literature that analyzes HFT and nHFT behavior around news announce-

ments.4I show that nHFTsoverall exhibit a stronger response to news information than HFTs,which parallels the find-

ings of Hautsch, Noe, and Zhang (2017) for macroeconomic announcements. Furthermore, this paper providesa more

detailed insight into the types of information processed by HFTs. Jovanovic and Menkveld (2016) suggest that the

advantage of HFTslies in analyzing “hard” quantitative information. Following Liberti and Petersen (2018), I distinguish

further between hard quantitativeand softer less quantitative information events.5I show a slightly stronger response

of HFTsto hard news, whereas nHFTs respond more strongly to softer news than HFTs. These results are in line with

the theoretical predictions of Jovanovic and Menkveld (2016) and Foucault, Hombert, and Rosu (2016), which state

that HFTs are better able to process quantitative information and might trade on more “precise”information. Using

the NASDAQHFT data set, I compare the trading responses of HFTs and nHFTsand test the predictions for both types

of traders and different kinds of news.

Second, I extend the literature on the intradayeffects of stock-specific news announcements using the same RNSE

data set as Kyle, Obizhaeva, Sinha, and Tuzun (2012),6Groß-Klußmann and Hautsch (2011), and Riordan, Storken-

maier, Wagener, and Zhang (2013). Specifically, I revealsignificant return and trading responses to news. My results

therefore complement the results of Groß-Klußmann and Hautsch (2011) and Riordan et al. (2013), who find an

increase in trading activity and higher adverse selection costs around news arrivals. I further show that the classifi-

cation of news based on its hardness or sentiment can yield significantly different results for return and net trading

responses. Therefore, I provide deeper insight into the cause for the documented differential effects of news on liquid-

ity,as demonstrated by Riordan et al. (2013).7

4Ageneral overview of the literature on HFT and its impact on financial markets is providedby Biais and Foucault (2014), Goldstein, Kumar, and Graves(2014),

Lattemannet al. (2012), Menkveld (2016), and O'Hara (2015), among others.

5Asdiscussed by Liberti and Peterson (2018, p. 2), the classification of information into hard and soft is a simplification and it is possible to change the nature

ofinformation, such as through the “hardening of soft information.” For example, Reuters has made considerable effort to “harden” news information and make

itusable for practitioners and academics. Therefore, the classification applied here regard the hardness of news information rather than distinctly classifying

informationas purely “hard” or “soft.”

6Kyleet al. (2012) use Reuters news as a proxy for public information flow and relate it to the invariance hypothesis of trading games.

7Inthis context, Foucault et al. (2016) predict that stocks with higher news informativeness can be more liquid despite the attraction of informed HFTs.Other

studies on HFT and price discoveryshow positive effects of HFTs on price discovery and efficiency in general (see Brogaard, Hendershott, & Riordan, 2014;