Examining drivers of trading volume in European markets

• Published date: 01 April 2018
• Author: Philip C. Treleaven,Bogdan Batrinca,Christian W. Hesse
• DOI: http://doi.org/10.1002/ijfe.1608
• Date: 01 April 2018

RESEARCH ARTICLE

Examining drivers of trading volume in European markets

Bogdan Batrinca | Christian W. Hesse | Philip C. Treleaven

Department of Computer Science,

University College London, Gower Street,

London WC1E 6BT, UK

Correspondence

Bogdan Batrinca, Department of

Computer Science, University College

London, Gower Street, London WC1E

6BT, UK.

Email: bogdan.batrinca.09@ucl.ac.uk

Funding information

Engineering and Physical Sciences

Research Council (EPSRC) UK

JEL Classification: C32; C52; C58; G12;

G15; G17

Abstract

This study presents an in‐depth exploration of market dynamics and analyses

potential drivers of trading volume. The study considers established facts from

the literature, such as calendar anomalies, the correlation between volume and

price change, and this relation's asymmetry, while proposing a variety of time

series models. The results identified some key volume predictors, such as the

lagged time series volume dataand historical price indicators (e.g.intraday range,

intradayreturn, and overnightreturn). Moreover,the study providesempirical evi-

dence for the price–volume relation asymmetry, finding an overall price asymme-

try in over 70%of the analysed stocks, whichis observed in the formof a moderate

overnight asymmetry and a more salient intraday asymmetry. We conclude that

volatility features, more recent data, and day‐of‐the‐week features, with a notable

negative effect on Mondays and Fridays, improve the volume prediction model.

KEYWORDS

asymmetric models,behavioural finance, European stockmarket, feature selection, price volume

relation, tradingvolume

1|INTRODUCTION

This study investigates the drivers affecting the trading

volume with an in‐sample analysis. We explore the inter-

action between truly exogenous determinants and trading

volume. Several hypotheses are evaluated while looking

at the previous literature, where various factors are

discussed in isolation, and we propose a liquidity extrac-

tion model by placing these findings in a unified context.

Identifying the drivers of trading volume is crucial in

order to anticipate and minimize market impact, by accu-

rately sizing and executing orders. Achieving optimal

order sizing relies on precise volume prediction, that is,

planning trades and deciding how much to trade given

the current market context and the predicted volume

levels. To better illustrate the importance of trading vol-

ume, some recent facts include the total turnover value,

which was $63tn in 2011 (World Federation of Exchanges,

2012) and $49tn in 2012 (World Federation of Exchanges,

2013). The NYSE's turnover averaged more than 100%

between 2004 and 2009, with 138% in 2008 (NYSE

Euronext, 2016), meaning that the entire market value

has changed hands once a year, although it has decreased

to significantly lower levels during the following years,

averaging 72% for the 2010–2015 period.

In order to better understand the factors affecting the

trading volume, it is necessary to survey and combine

apparently disjoint literature concepts. We start by

reviewing the relevant areas of the behavioural finance

literature. Here, a large amount of research has mainly

investigated the calendar effects on price returns, and

there is very little emphasis on the calendar effects on

trading volume. We particularly focus on the day‐of‐the‐

week effect, which, once investigated, can formulate

------------------------------------------------------- -- --- -- -- --- -- --- -- -- --- -- --- -- -- --- -- --- -- -- --- -- --- -- --- -- -- --- -- --- -- -

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the

original work is properly cited.

© 2018 The Authors. International Journal of Finance & Economics Published by John Wiley & Sons Ltd

Received: 28 July 2016 Accepted: 19 December 2017

DOI: 10.1002/ijfe.1608

134 Int J Fin Econ. 2018;23:134–154.wileyonlinelibrary.com/journal/ijfe

several hypotheses to analyse other calendar effects (e.g.

the effect of stock index futures expiries and cross‐market

holidays). We then connect the behavioural finance find-

ings with evidence from the literature on the relation

between price changes and volume (i.e. the price–volume

lead–lag effect). Following this reverse path, we test the

direct relation between calendar effects, represented in

this study by the day‐of‐the‐week effect, and volume.

Behavioural finance mainly consists of regression

models built on a collection of indicator variables, imply-

ing a certain limitation with regard to its statistical signif-

icance. We propose a model based on lagged time series

and lagged smoothed time series in order to explain

observed volumes in terms of recent time series; this fol-

lows the behavioural finance paradigm and represents

market dynamics on the run, while assuming stationarity

and disregarding outliers. However, the financial data are

a strong nonstationary and nonconstant mean time series,

due to the existence of notable event dates (e.g. MSCI

rebalance dates, futures expiry dates, and company earn-

ings announcement dates). This analysis aims to bridge

the gap between behavioural finance and traditional

finance and explores the feasibility of a potential special

event effect (e.g. futures expiries or cross‐market holidays)

on trading volume by starting with an analysis of the day‐

of‐the‐week effect on trading volume. The financial mar-

kets are event‐driven, and their dynamics are perma-

nently shifting. Therefore, it is important to predict the

trading volume at different time horizons.

The main motivations of this study include the follow-

ing: the insufficiency of literature looking at the calendar

effects on trading volume (and not on returns), the incon-

clusive results of the price–volume relation and whether it

is characterized by asymmetry, and the abundance of

studies investigating certain volume determinants in

complete isolation from other types of volume drivers.

Out of a total number of 55 surveyed articles, which

are all cited in this study and investigate the price–volume

relation and the day‐of‐the‐week effect, only seven of

them use data sets after 2000 and none of the cited papers

employs market data after 2006. Moreover, only seven

studies include a few European stocks or indices among

their international data sets, and only two papers are

based on European data sets exclusively. Given the lack

of a broad European stock universe and post‐2000 data

sets, we employ an extensive pan‐European stock uni-

verse consisting of 2,353 stocks, for which we use daily

market data between 1 January 2000 and 10 May 2015,

and we also test for structural breaks by comparing the

results before and after the financial crisis of 2007–2008.

The aim of this study is to define a unified volume pre-

diction model, while exploring the endogenous variables

in conjunction with exogenous variables and performing

feature selection. We investigate a pan‐European stock

universe for a sample period of over 15 years in order to

test the improvement of an autoregressive volume model,

by sequentially adding features, such as volatility, more

recent data, and day‐of‐the‐week, and test additional

hypotheses such as the existence of an asymmetric

price–volume relation. The rest of the paper is organized

as follows: Section 2 provides a literature review of the

main research topics addressed by this study: volume

dynamics, price actions, and volume–price correlations,

along with a survey of the relevant calendar effects;

Section 3 introduces the sample data set, whereas Section

4 outlines the main models and the analytical approach;

Section 5 describes the methodology of the trading

volume study, while gradually introducing the various var-

iables we are examining in order to better predict the vol-

ume; Section 6 exhibits the main results of the various

volume prediction models and the types of price–volume

asymmetry; Section 7 provides a conclusion of this study,

together with a discussion on the results and potential sug-

gestions for other researchers to further extend this study.

2|BACKGROUND

The literature review starts by setting the context of this

study, that is, why volume prediction is important,

followed by a review of studies on the types of the

price–volume relation and its potential asymmetry. We

then switch to the behavioural finance literature by

outlining the main calendar effects and elaborate on the

day‐of‐the‐week effect.

2.1 |Trading volume historical dynamics

Trading volume is extraordinarily large across developed

stock exchanges, and many interesting patterns in prices

and returns are closely related to the volume movement;

volume is highly used in conjunction with price actions.

For instance, the volume of high‐priced “glamour

stocks”tends to be larger than the volume of low‐priced

value stocks, and a stock with higher trading volume

tends to have lower future returns (Hong & Stein,

2007). Trading volume is a strong indicator of economic

activity.

Auctions account for a high trading volume, and there

are three types of auctions: opening auctions, intraday

auctions, and closing auctions. A normal day starts with

pretrading auctions or opening auctions, in order to set

the price after the nontrading hours during the night,

when news came out, and is followed by continuous

trading. In Europe, this phase can be temporarily halted

by volatility interruptions, which trigger a 2‐to 5‐min

BATRINCA ET AL.135