High‐Frequency Trading around Large Institutional Orders

• DOI: http://doi.org/10.1111/jofi.12759
• Date: 01 June 2019
• Author: VINCENT VAN KERVEL,ALBERT J. MENKVELD
• Published date: 01 June 2019

THE JOURNAL OF FINANCE •VOL. LXXIV, NO. 3 •JUNE 2019

High-Frequency Trading around Large

Institutional Orders

VINCENT VAN KERVEL and ALBERT J. MENKVELD∗

ABSTRACT

Liquidity suppliers lean against the wind. Weanalyze whether high-frequency traders

(HFTs) lean against large institutional orders that execute through a series of child

orders. The alternative is HFTs trading with the wind, that is, in the same direction.

We find that HFTsinitially lean against these orders but eventually change direction

and take positions in the same direction for the most informed institutional orders.

Our empirical findings are consistent with investors trading strategically on their

information. When deciding trade intensity,they seem to trade off higher speculative

profits against higher risk of being detected and preyed on by HFTs.

MIGRATION TO ELECTRONIC TRADING CREATED a new type of market participant:

high-frequency traders (HFTs). The U.S. Securities and Exchange Commission

(SEC) characterized such traders as “professional traders acting in proprietary

capacity” who use “extraordinarily high-speed and sophisticated computer pro-

grams for generating, routing, and executing orders” and end the trading day

“in as close to a flat position as possible.” HFTs entered securities markets in

the late 2000s, appearing first in equity markets. Their market participation,

in percentage terms, is typically a few deciles (SEC (2010), European Securities

and Markets Authority (2014)).

High-frequency trading has triggered a great deal of academic study, par-

ticularly after NASDAQ released data that label HFT trades. The evidence

is by and large favorable on HFTs, emphasizing reduced bid-ask spreads and

increased price efficiency, although evidence on how HFTs relate to “excess”

∗Vincent van Kervel is at Pontificia Universidad Cat´

olica de Chile. Albert J. Menkveld is at

Vrije Universiteit Amsterdam and TinbergenInstitute. We thank Andrew Ainsworth, Bruno Biais

(the Editor), Jonathan Brogaard, Bj¨

orn Hagstr¨

omer, Terrence Hendershott, Ron Kaniel, Robert

Korajzyck, Charles-Albert Lehalle, Anna Obizhaeva, Neil Pearson, Marc Ponsen, Talis Putnins,

Yazid Sharaiha, George Sofianos, Ren´

e Wells, Avi Wohl, Shihao Yu, an anonymous Associate Edi-

tor,and two anonymous referees for their very useful comments. We are also grateful for comments

received during conferences at the Alan Turing Institute, NBIM, and Tel Aviv University and at

seminars at AFM, CFTC, Universit´

e Paris Dauphine, Goldman Sachs, SEC, University of Syd-

ney, UNSW, and UTS. We thank Bernard Hosman and Sailendra Prasanna Mishra for excellent

research assistance. Menkveld gratefully acknowledges NWO for a Vidi grant and van Kervel

gratefully acknowledges the financial support of the Fondecyt Iniciaci´

on (project 11150485). The

authors further acknowledge support from APG, DNB, NBIM, Swedbank Robur, and QuantVal-

ley/FdR Quantitative Management Initiative. The authors have signed nondisclosure agreements

for the proprietary data used in this study.

DOI: 10.1111/jofi.12759

1091

1092 The Journal of Finance R

volatility, such as in flash crashes, is mixed. Biais and Foucault (2014), O’Hara

(2015), and Menkveld (2016) provide survey articles on the young and rapidly

growing HFT literature.1

Notwithstanding, we know little about how HFT relates to trading by an

important group of end users of securities markets: institutional investors.

Retail investors benefit from a smaller bid-ask spread, since there is generally

enough depth at the best quote to execute their order. Institutional investors,

however, need to “work their order” by splitting a parent order into small child

orders that are fed to the market sequentially. Institutional investors care

about “implementation shortfall,” that is, the average price at which the entire

order executes relative to the price that prevailed when it started executing.

In other words, they care about how far they pushed the price away from them

while executing. Importantly, institutional investors care about cumulative

price impact rather than the half-spread paid on, for example, a single market

order. Some have expressed the concern that trading costs have increased and

attribute this to the presence of HFTs.2

Objective. In this paper, our objective is to document how HFTs trade, while

an institutional order executes through a series of child trades, and to provide

an economic narrative grounded in theoretical studies. Three such studies

speak directly to our empirical setting. First, as suggested by Grossman and

Miller (1988), HFTscould bemarket makers who take the other side of the order.

For example, they would be selling in the period when a buy order executes.

Second, as Brunnermeier and Pedersen (2005) suggest, they could engage in

predatory trading, buying along with the institutional order initially to then

sell when the buy order is almost finished executing, essentially riding the

(transitory) price-pressure wave. Third, as Yang and Zhu (2017) suggest, they

might need time to learn that the buy order is executing to back-run on it later

by also buying.3Taking such a long position relatively late only makes money,

however, if the order is information-motivated (as opposed to an uninformed

order that seeks liquidity). These three studies offer distinct predictions for

how HFTs trade as intermediaries while an institutional order executes and

for the way in which prices respond (see Section I).

1Several empirical studies find that HFT activity reduces bid-ask spreads (Hendershottt, Jones,

and Menkveld (2011), Hasbrouck and Saar (2013), Menkveld (2013), Brogaard et al. (2015),

van Kervel (2015)) and improves price efficiency (Boehmer, Fong, and Wu (2014), Brogaard,

Hendershott, and Riordan (2014)). The effect of high-frequency trading activity on short-term

volatility and crashes is mixed: Some studies document a negative correlation (Hasbrouck and

Saar (2013), Chaboud et al. (2014), Hasbrouck (2015)), whereas others document a positive corre-

lation (Gao and Mizrach (2013), Ye,Yao, and Gai (2013), Boehmer, Fong, and Wu(2014), Kirilenko

et al. (2017)).

2See, for example, “Institutional Investors Air HFT Concerns,” Financial Times, September 12,

2011, “Wealth Fund Cautions against Costs Exacted by High-Speed Trading,”The New York Times,

October 20, 2013, and “Berkshire’s Munger: High-Frequency Tradings’ Basically Evil,” Berkshire

Munger, May 3, 2013.

3Boulatov, Bernhardt, and Larionov (2016) propose a model that also generates back-running

but considers the exogenous price impact function of Almgren and Chriss (1999). However, they go

beyond two periods and analyze Nash strategies in a continuous-time setting.

High-Frequency Trading around Large Institutional Orders 1093

We take the predicted patterns to the data. The empirical analysis is based on

a sample that combines proprietary data on order executions by institutional

investors with public HFT trade data. The sample runs from January 1, 2011

through March 31, 2013 and covers trading in Swedish index stocks. The order

execution data were provided by four large institutional investors (APG, DNB,

NBIM, and Swedbank Robur) and include 801,341 child trades. We construct

parent orders, which we refer to as meta orders, by stringing together child

trades by a single institution in a single stock, possibly over multiple days (for

details, see Section II.B). The final sample consists of 5,136 such meta orders

that, on average, contain 156 child trades that execute over a bit more than

four hours. Not surprisingly, we find that these orders are directional in the

sense that all child trades in a single order are either almost exclusively buys or

almost exclusively sells. Finally,institutional orders tend to be large averaging

$2.2 million, or 3.88% when expressed relative to average daily volume.

An important benefit of this particular sample is that during the period

considered, HFTs had to reveal their trades on NASDAQ OMX,4the domi-

nant market at the time with a market share of two-thirds for the local index

stocks.5We select Europe’s largest high-frequency trading firms according to

Financial News: Citadel, Flow Traders,Getco, IAT, IMC, Knight, Optiver,Spire,

Susquehanna, and Virtu.6Collectively, the participation rate of these firms in

NASDAQ OMX trades is almost a third in our sample.

Findings. The empirical analysis yields three main findings. First, the pat-

tern of HFT trading fits none of the three theoretical studies perfectly. We find

that HFTs as a group appear to lean against the wind when an order starts ex-

ecuting, but if execution lasts more than seven hours (possibly multiple days),

they reverse course and trade with the wind. Indeed, they not only close out

their position, but they also actually build a speculative position in the direc-

tion of the order. For buy orders, for example, HFTs initially sell but after seven

hours start buying, and they buy lots more than they initially sold. A potential

concern with this result is that HFTs could have entered offsetting positions

in alternative markets or in highly correlated securities. We consider this con-

cern unlikely, however, because perfectly correlated securities are hard to find

for single stocks, and NASDAQ OMX is by far the largest equity exchange for

Swedish stocks.

A more pressing concern is that this result could be mechanical: to the extent

that institutional investors and HFTs both respond to certain market condi-

tions, any correlation in their trading patterns would be driven by a third fac-

tor. To rule out such an explanation, we create a placebo sample in which none

of the institutional investors were active yet market conditions were similar

4This changed on March 23, 2014, when NASDAQ OMX changed to voluntary reports. Many

HFT firms opted to go under the radar and not report their trades. See “Changes to Post Trade

Counterparty Visibility in NASDAQ OMX Nordic Blue Chip Shares,” GlobeNewswire,February

6, 2014.

5These numbers come from Fidessa, a trade reporting company (http://tinyurl.com/ozo8ytm).

6See “Europe’s Top10 High-Frequency Kingmakers,” Financial News, October 3, 2011.