CORRELATED BEHAVIOR IN LIMIT ORDER CANCELLATIONS, COMOVEMENT IN ASSET RETURNS, AND COMMONALITY IN LIQUIDITY

• Date: 01 March 2020
• Author: Jared Egginton,Ethan D. Watson
• DOI: http://doi.org/10.1111/jfir.12200
• Published date: 01 March 2020

The Journal of Financial Research Vol. XLIII, No. 1 Pages 37–62 Spring 2020

DOI: 10.1111/jfir.12200

CORRELATED BEHAVIOR IN LIMIT ORDER CANCELLATIONS,

COMOVEMENT IN ASSET RETURNS, AND COMMONALITY IN LIQUIDITY

Jared F. Egginton

Boise State University

Ethan D. Watson

University of North Carolina Wilmington

Abstract

We examine whether there is common behavior in limit order cancellation activity,

that is, commonality in cancellation activity, on U.S. exchanges. We then examine

whether this commonality in cancellation activity is associated with increased levels

of return comovement and commonality in liquidity. We document strong evidence

of limit order traders exhibiting exchange, industry, marketwide, and stock‐level

commonality with regard to cancellation activity, which is consistent with limit order

traders exhibiting correlated trading behavior. We also find that this correlated

behavior in cancellation activity is associated with increased levels of return

comovement and commonality in liquidity.

JEL Classification: G10

I. Introduction

As a central part of the trading process, limit order traders receive much attention in the

academic literature, traditionally being cited as suppliers of liquidity. Aided by

technological advances, limit order traders can now access markets at incredibly fast

speeds (Angel 2014). One byproduct of these technological advances is that limit order

traders are shown to submit and subsequently cancel a large percentage of orders, with

some orders living in the limit order a book only short amounts of time (see Ellul et al.

2007; Hasbrouck and Saar 2009).

Recently, some researchers are investigating the impact that this relatively new

innovation of frequent cancellations has on market quality. For example, Angel (2014)

questions whether the increase in cancellations affects the matching process for trades.

Van Ness, Van Ness, and Watson (2015) use an instrumental variables approach and

find that high levels of limit order cancellations are associated with larger spreads and

price impacts and with less depth at the inside quotes. Furthermore, Griffith and Van

Ness (2019) use a natural experiment, where the Philadelphia Options Exchange

(PHLX) introduced a fee on cancellations, and find evidence that liquidity providers

left the exchange, resulting in a detrimental effect on spreads. Griffith and Van Ness

Ethan Watson acknowledges funding from the University of North Carolina at Wilmington’s 2015 BB&T

Research Grant.

37

© 2019 The Southern Finance Association and the Southwestern Finance Association

also find that the fee is effective in reducing cancellations and is associated with a

higher arrival rate of marketable orders, a higher nonmarketable order‐fill rate, and

faster order execution speeds. Thus, the studies indicate that limit order traders

canceling many of their orders has an impact on the market.

In addition to considering what impact the level of cancellation activity has on

the market, the literature considers how certain subsets of limit order traders exhibit

correlated behavior with regard to cancellation of limit orders. For example Van Kervel

(2015) builds a model where limit order traders can face an adverse selection problem

from “fast traders.”The limit order traders (market makers) post limit orders in two

venues and can update and cancel orders on other markets after observing a trade in

one of the venues. Van Kervel’s model provides justification for how cancellation

activity may be correlated across exchanges.

1

Thus, it is reasonable to believe that limit

order cancellations may exhibit correlated behavior, in this manner as well as in other

manners as we explain later.

In this article, we extend the literature that examines the effects of limit order

cancellations by studying correlated behavior of order cancellations, and then examine

whether this has potentially deleterious effects on the market by examining whether

correlated cancellation activity is associated with return comovement and liquidity

comovement (usually referred to as commonality in liquidity).

2

This research question

is motivated by two threads of the literature that examine the impact of correlated

behavior of various market participants on markets: the return comovement literature

and the commonality in liquidity literature.

There is now a lengthy literature that examines how correlated trading

behavior among various market participants leads to return comovement. For example,

Barberis and Shleifer (2003) build a model that considers how return comovement can

arise as traders chase returns and switch to styles (value, growth, etc.) that perform well

in previous periods. Further evidence documents that comovement of returns increases

as correlated trading behavior increases because of additions/deletions of stocks to an

index (Barberis, Shleifer, and Wurgler 2005; Greenwood and Sosner 2007), stock splits

(Green and Hwang 2009), similar informational environments due to geography

(Pirinsky and Wang 2006), and similar information sets due to investment banking

relationships (Grullon, Underwood, and Weston 2014). Additionally, correlated

behavior among investor groups is shown to lead to return comovement, such as

retail investors (Kumar and Lee 2006) and institutional investors (Pirinsky and Wang

2004; Sun 2008; Antón and Polk 2014).

1

As we detail later, we use the motivation of Van Kervel (2015) to define one of the forms of commonality

that we consider in our analysis (stock‐level commonality in cancellations). However, we also consider much

broader types of commonality such as across stocks, across exchanges, and across industries.

2

The term “commonality”was first coined by Chordia, Roll, and Subrahmanyam (2000) and is used to

indicate that common, marketwide factors determine individual security characteristics. For example, Chordia,

Roll, and Subrahmanyam find there are common, marketwide factors that determine bid–ask spreads and depths

for an individual security. There is now a fairly extensive literature that investigates commonality of phenomena

such as global liquidity and short‐selling behavior (see, e.g., Brockman, Chung, and Pérignon 2009; Lynch

et al. 2014).

38 The Journal of Financial Research