Springer Proceedings in Business and Economics

Krzysztof Jajuga
Hermann Locarek-Junge
Lucjan T. Orlowski
Karsten Staehr Editors

Contemporary
Trends and
Challenges in
Finance
Proceedings from the 6th Wroclaw
International Conference in Finance
Springer Proceedings in Business and Economics
Springer Proceedings in Business and Economics brings the most current research
presented at conferences and workshops to a global readership. The series features
volumes (in electronic and print formats) of selected contributions from conferences
in all areas of economics, business, management, and finance. In addition to an
overall evaluation by the publisher of the topical interest, scientific quality, and
timeliness of each volume, each contribution is refereed to standards comparable to
those of leading journals, resulting in authoritative contributions to the respective
fields. Springer’s production and distribution infrastructure ensures rapid publication
and wide circulation of the latest developments in the most compelling and prom-
ising areas of research today.
The editorial development of volumes may be managed using Springer’s inno-
vative Online Conference Service (OCS), a proven online manuscript management
and review system. This system is designed to ensure an efficient timeline for your
publication, making Springer Proceedings in Business and Economics the premier
series to publish your workshop or conference volume.

More information about this series at http://www.springer.com/series/11960

Krzysztof Jajuga • Hermann Locarek-Junge •
Lucjan T. Orlowski • Karsten Staehr
Editors

Contemporary Trends and

Challenges in Finance
Proceedings from the 6th Wroclaw
International Conference in Finance
Editors
Krzysztof Jajuga Hermann Locarek-Junge
Department of Financial Investments and Fakultät Wirtschaftswissenschaften
Risk Management TU Dresden
Wroclaw University of Economics & Dresden, Sachsen, Germany
Business
Wroclaw, Poland

Lucjan T. Orlowski Karsten Staehr

Department of Economics and Finance Department of Economics & Finance
Sacred Heart University Tallinn University of Technology
Fairfield, CT, USA Tallinn, Estonia

ISSN 2198-7246 ISSN 2198-7254 (electronic)

Springer Proceedings in Business and Economics
ISBN 978-3-030-73666-8 ISBN 978-3-030-73667-5 (eBook)
https://doi.org/10.1007/978-3-030-73667-5

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland
AG 2021
This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether
the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of
illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and
transmission or information storage and retrieval, electronic adaptation, computer software, or by
similar or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors, and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, expressed or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Switzerland AG.
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface

This volume presents papers from the 6th Wrocław International Conference in
Finance held at Wrocław University of Economics and Business via video confer-
encing on September 22–23, 2020. We have assembled a set of studies addressing a
broad spectrum of recent trends and issues in finance, particularly those concerning
markets and institutions in Central and Eastern European countries. In the final
selection, we had accepted 15 of the papers that were presented at the conference.
Each of the submissions has been reviewed by at least two anonymous referees and
the authors have subsequently revised their original manuscripts and incorporated
the comments and suggestions of the referees. The selection criteria focused on the
contribution of the papers to the modern finance literature and the use of advanced
analytical techniques.
The chapters have been organized along the major fields and themes in finance:
financial markets, corporate finance, banking, and personal finance.
The section on financial markets contains seven papers. The paper by Saqib Amin
investigates the relationship between diversity and stock market development by
using data of 187 countries. Anna Białek-Jaworska in her paper answers the question
whether withholding tax reduces income shifting with the use of debt and equity
FDI. The paper by Lesław Markowski examines the relationship between condi-
tional volatility of individual stock returns and trading volume on the Warsaw Stock
Exchange. Dorika Mwamtambulo in her paper determines the factors behind the low
individual investor participation in Dar es Salaam Stock Exchange. The paper by
Aleksandra Pasieczna analyzes the model risk of Expected Shortfall and Value at
Risk using different variants of Monte Carlo approach. Pham Khang in his paper
presents an analysis of the tick size adjustment in Vietnamese stock exchange.
Gopinath Ramkumar studies the portfolio of nine most important cryptocurrencies
constructed using several types of strategies.
The section on corporate finance contains two papers. The paper by Julia
Koralun-Bereźnicka studies the relation between corporate material and financial
decisions, based on 12 EU countries. Katarzyna Prędkiewicz, Paweł Prędkiewicz,

v
vi Preface

and Marek Pauka in their paper examine whether Warsaw Stock Exchange is an
effective means of alleviating financial constraints for high technology companies.
The section on banking contains three papers. The paper by Monika Kołodziej is
aimed at critically analyzing the impact of blockchain technology on banking
industry. Małgorzata Olszak and Anna Kowalska in their paper analyze what is
the role of competition in the effects of macroprudential policy. The paper by Witold
Szczepaniak and Marta Karaś presents the results of the empirical measurement of
systemic risk levels using several quantile-based measures.
The section on personal finance contains three papers. The paper by Agnieszka
Huterska is aimed at assessing the disproportions in the use of loan products by
young people in the countries—old and the new members of the European Union.
Katarzyna Kochaniak and Paweł Ulman analyze the gap between subjective and
objective financial risk tolerance. The paper by Ergun Kutlu is aimed at finding the
relationship between financial behavior and socio-demographic variables for Italian
and Turkish students.
We wish to thank the authors for making their studies available for our volume.
Their scholarly efforts and research inquiries made this volume possible. We are also
indebted to the anonymous referees for providing insightful reviews with many
useful comments and suggestions. In spite of our intention to address a wide range of
problems pertaining to theoretical concept and empirical trends in finance, there are
issues that still need to be researched. We hope that the studies included in our
volume will encourage further research and analyses in modern finance.

Wroclaw, Poland Krzysztof Jajuga

Dresden, Germany Hermann Locarek-Junge
Fairfield, CT, USA Lucjan T. Orlowski
Tallinn, Estonia Karsten Staehr
February 25, 2021
Contents

Part I Financial Market

The Relationship Between Ethnic Diversity and Stock
Market Development: A Global Perspective . . . . . . . . . . . . . . . . . . . . . . 3
Saqib Amin
Does Withholding Tax Reduce International Income-Shifting
by FDI? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
Anna Białek-Jaworska
The Relationship Between Trading Volume and Returns Volatility
on Warsaw Stock Exchange . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
Lesław Markowski
Factors Influencing Individual Investor Participation in Stock
Market . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
Dorika Jeremiah Mwamtambulo
Model Risk of VaR and ES Using Monte Carlo: Study on Financial
Institutions from Paris and Frankfurt Stock Exchanges . . . . . . . . . . . . . 75
Aleksandra Helena Pasieczna
Tick Size Reduction and Liquidity Dimensions: Evidence from
an Emerging Market . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
Quoc-Khang Pham
Cryptocurrency Portfolio Construction Using Machine Learning
Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
Gopinath Ramkumar

vii
viii Contents

Part II Banking
Development Factors of Blockchain Technology Within Banking
Sector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
Monika Kołodziej
Does Competition Matter for the Effects of Macroprudential
Policy on Bank Asset Growth? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
Małgorzata Olszak and Iwona Kowalska
Systemic Risk in Selected Countries of Western and Central Europe . . . 169
Marta Karaś and Witold Szczepaniak

Part III Corporate Finance

Industry and Size Effect in the Relation Between Corporate Material
and Financial Decisions: Findings from the EU Countries . . . . . . . . . . . 189
Julia Koralun-Bereźnicka
Technology Level and Financial Constraints of Public Listed
Companies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
Katarzyna Prędkiewicz, Paweł Prędkiewicz, and Marek Pauka

Part IV Personal Finance

Differences in Use of Credit Products Between the Old and New
Member States of the European Union . . . . . . . . . . . . . . . . . . . . . . . . . . 217
Agnieszka Huterska
Diversified Risky Financial Assets in Portfolios of Risk-Averse
Households: What Determines Their Occurrence? . . . . . . . . . . . . . . . . . 229
Katarzyna Kochaniak and Paweł Ulman
Financial Behavior: Preliminary Survey Results . . . . . . . . . . . . . . . . . . . 241
Kutlu Ergün
About the Editors

Krzysztof Jajuga is a professor of finance at Wrocław University of Economics

and Business, Poland. He holds master, doctoral, and habilitation degrees from
Wrocław University of Economics, Poland, the title of professor given by President
of Poland, an honorary doctorate from Cracow University of Economics, and an
honorary professorship from Warsaw University of Technology. He carries out
research within financial markets, risk management, household finance, and multi-
variate statistics.

Hermann Locarek-Junge is Professor of Finance and Financial Services at TU

Dresden, Faculty of Management and Economics. He graduated in the field of
business and economics and earned his Ph.D. at University of Augsburg (Germany);
he also studied business informatics and has been appointed professor in that field at
Essen University. Since then, he has been visiting professor and research fellow at
several international institutions and universities. During his academic career, he
undertook research work for several banks.

Lucjan T. Orlowski is a professor of economics and finance and a director for the
Doctor of Business Administration (DBA) in Finance Program at Sacred Heart
University in Fairfield, Connecticut. His research interests include monetary eco-
nomics and stability of financial markets and institutions. He has authored numerous
books, chapters in edited volumes, and over 80 articles in scholarly journals. He is a
Doctor Honoris Causa recipient from Wrocław University of Economics.

Karsten Staehr is professor at Tallinn University of Technology, Estonia. He is

also employed part-time as a research supervisor at Eesti Pank, the central bank of
Estonia. He holds a master’s degree from the Massachusetts Institute of Technology
and a master’s degree and a Ph.D. from the University of Copenhagen. He carries out
research and policy analysis within macroeconomics, international finance, mone-
tary economics, and public economics. He is an associate editor of the Baltic Journal
of Economics and on the editorial board of several other journals.

ix
 Part I
Financial Market
The Relationship Between Ethnic Diversity
and Stock Market Development: A Global
Perspective

Saqib Amin

1 Introduction

No doubt, stock market is a key determinants for economic prosperity of any country
(Pan and Mishra 2018; Rousseau and Wachtel 2000). In order to find the determi-
nants of stock market development, diversity is found as one of the key indicators,
which directly affect the behaviour of investor’s mentor. Limited debates in litera-
ture show the linkages between diversity and stock market development and some-
how complex and puzzle (Forti et al. 2011; Tang et al. 2016; Zulfiqar and Weller
2013). However, diversity plays a vital role in drastically changes the stock market
development that has directly and indirectly effects the economic development
(Bove and Elia 2017; Montalvo and Reynal-Querol 2005; Nettle et al. 2007). This
type of financial shocks leads to a catastrophically changes in country’s economic
prosperity. Most of the researchers argued that number of financial shocks i.e. great
depression of 1929, financial shocks of 2008 and 2010 caused by the wreck people’s
behaviours, markets and the economy which creates the burden for local as well as
global economy (Jain and Jordan 2009; Sacasa 2008; Verick and Islam 2010). On
the other hand, some argued that the reasons behind these episodes of widespread
economic bubbles, financial havoc and great depression are remains unclear. In their
opinions, these types of bubbles happen when people mindlessly trust the behaviour
of others particularly surrounded by ethnic peers. In this context, diverse societies
are largely creators of bubbles because it creates changes on how people think, feel
and behave.

S. Amin (*)
Department of Economics, National College of Business Administration and Economics,
Lahore, Pakistan
Department of Financial Investments and Risk Management, Wroclaw University of Economics
and Business, Wroclaw, Poland

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 3

K. Jajuga et al. (eds.), Contemporary Trends and Challenges in Finance, Springer
Proceedings in Business and Economics,
https://doi.org/10.1007/978-3-030-73667-5_1
4 S. Amin

Every country is trying to create a good investment environment to building up a

strong, healthy and prosperous stock market. Traditional economists believed that
financial market behaviour is affected by the key macroeconomic variables such as
GDP growth (Levine and Zervos 1998), trade volume (Chinn and Ito 2006) inflation,
exchange rate, etc., They argued that diversity is importance only in the context of
economic advancement (Collier 1998; Elbadawi and Sambanis 2000; Watkins and
Ferrara 2005) and investment (Montalvo and Reynal-Querol 2005). But, modern
economist argued that despite the importance of macroeconomic variables, micro-
economic variables are also important for the financial market development under
the behavioural finance theory. They converted the directions most of policy makers
and researchers towards socioeconomic factors in order to explain the stock market
behaviours in contrast to the traditional economists (Alesina and Giuliano 2015).
Patrick (1966), Hall (2001) and Guiso et al. (2006) argued that social factors are
important as well as economic factors into the stock market development. They also
argued that diversity explain more of the difference in finance, development and
optimal utilization of resources within the society and economy (Stulz and
Williamson 2003). Most of the researcher revealed that socioeconomic factors
such as diversity has reposted impact on financial market behaviours (Bond et al.
2012; D. J. Miller 2006) financial development (Kwok and Tadesse 2006) and the
investors behaviours (Breuer et al. 2014).
In this contemporary world and modern society the nature of markets always
remained ubiquitous (North 1991). The stock market development has become a
main central financial institution for development and prosperity of any country
because it escalating the information, expectation and better use of resources other
than the government or individual (Hayek 1945). Whereas stock market also cause
of catastrophically changes such as price bubbles and share values are drastically
changes upon the behaviours of entrepreneurs in market. Market herd behaviours
emerge from the homogenous to heterogeneity society collectively changes the
failure of market and prices changes such as prices bubbles. These types of changes
destroy and destabilize the economy not just to the home country but also to the
entire world. In this context, most of the psychologist, sociologist and economists
have shown unsurprising regarding the bubbles behind the behavioural biases of
individuals.
Attempting to pinpoint the cause of bubbles, some researchers have shown that
bubbles are related to financial conditions such as excess cash, but also to behaviour
that exhibits of irrationality. Indeed, bubbles have been long ascribed to collective
delusions, implied in terms as “herd behaviours”, but their exact causes remain
nebulous. In modern markets, traders place greater confidence in the actions of
others. They are more likely to accept their co-ethnics decisions as reasonable, and
therefore more likely to act alike. In a homogenous market, individuals are less likely
to scrutinize on others behaviour than the diverse market. People tend to be more
trusting of the perspectives, actions, and intentions of ethnically similar society
(Brown 2000; Hogg 2007; Pettigrew et al. 2011). Therefore, those markets are
more vigilant and overreliance on others decisions which may be more risky.
Diverse society may be more harmful towards the modern markets due to the
The Relationship Between Ethnic Diversity and Stock Market Development: A. . . 5

confidence on other decisions. This study suggests that price bubbles arise not only
from individual errors or financial conditions but also from the social context of
decision-making.
Recently, in the financial economics debate, many researchers investigate the
various determinants of stock market development but limited literature shows the
relationship between diversity and stock market development. This study explores a
very interesting relationship between diversity and stock market development using
data of 187-countries of the world (for more detail see in Appendix list of the
countries).

2 Theoretical Framework

In the behavioural finance theory, investor’s decisions directly impact the stock
market performance, whereas investor’s decision-making are depends upon
behavioural changes and adaptive diversification of societies. Literature shows gap
to find the relationship between diverse society and stock market development. It
may be one reason of this neglecting aspect is too difficult and complex in measuring
diversity in quantification. However, this problem has resolved by the construction
of fractionalization index by Alesina et al. (2003). In the macro perspective, various
discipline and theories such as sociology, psychology, economics and finance are
merging under this context. The theory of social conflict, rational choice theory,
anomie theory, rational choice theory, social exchange theory, agency theory,
rational expectation theory are well discuss directly or indirectly the relationship
between diversity and behaviours of financial markets.
Ethnic diversity has been studied in multiple spheres, including economic growth
(Alesina and Ferrara 2005; Florida 2002), social capital (Putnam 2007), cities and
neighbourhoods (Pettigrew et al. 2011), organizational performance (Herring 2009;
Richard 2000), work teams (DiTomaso et al. 2007; Van Knippenberg and Schippers
2007; Williams and O’Reilly III 1998), and jury deliberations (Sommers 2006). As
intergroup contact theory and social identity theory, ethnic identity is mainly
establishing the trust among strangers. Moreover, empirical evidence shows specif-
ically that people surrounded by ethnic peers tend to process information more
superficially (Antonio et al. 2004; Sommers 2006; Sommers et al. 2008). In markets,
where information is incomplete and decisions are uncertain (Kahneman 2003),
traders may be particularly reliant on ethnicity as a group-level heuristic for
establishing confidence in others decisions. Such superficial information processing
can engender conformity, herding, and price bubbles. This is not an individual
idiosyncrasy, but a collective phenomenon pricing errors of traders in homogenous
markets are more likely correlated than those of traders in diverse markets. The
culmination of these processes leads to bigger bubbles.
Breuer et al. (2014) investigated the impact of cultural diversity on time prefer-
ence and founded that cultural diversity is the time preference indicator: culture and
time preference are closely related, and different cultures will lead to different time
6 S. Amin

preferences. The intertemporal investment consumption theory indicated that the

different time preference of investors shaped by cultural diversity will deduce
different trading strategy choice, thus affect the financial market. Whereas, cognitive
processes get influenced by culture, thus the cultural diversity could bring hetero-
geneous beliefs, which promote the prosperity of the stock market. Behavioural
finance studies have indicated that the heterogeneity of the expectations and beliefs
caused by cognitive differences has an important impact on financial market (Chui
et al. 2010). Tang et al. (2016) revealed that diversity could promote stock market
prosperity and the effect varies from country to country. In particular, it was found
that cultural diversity has positive impact on stock market prosperity in high
development-level and legal system quality countries, and vice versa. Weller and
Zulfiqar (2013) argued that greater diversity is associated with faster growth, larger
credit markets, a broader deposit base, and a smaller chance of asset bubbles, all of
which could contribute to more stability.
Researcher wants to know, does increase in the level of diversity will promote
prosperity of the stock market? Diversity plays different role in different countries.
Compared to the common ethnic and cultural conflicts in African countries, different
ethnic groups in the European and American multinational countries get along well,
and there is frequent diversity communication between them. Taydas et al. (2010)
found that countries with higher levels of legal system have less social internal
conflict. The good legal system level can provide protection of basic rights and a fair
trading environment for investors. Only under the common standards and rules, the
diversity of will not lead to market confusion, and the stock market can develop in an
orderly manner.
Frame and White (2004) showed that financial innovation (new products, new
services, new processes and new forms of organization) could improve the function
of financial sector. Diverse society promote technological innovation which ultimate
promote economic growth, and onwards lead to the prosperity of stock market
(Ashraf and Galor 2011). Limited literature was found about relationship between
diversity and financial innovation, but the positive effect of diversity on innovation
became the academic consensus by Herbig and Dunphy (1998).
Head and Ries (1998) define this knowledge as more ethnic groups from different
countries, more the export market can be expanded which stimulates trading volume.
Egger et al. (2012) specify that ethnic groups engage in the market creation and make
be able to open up to other foreign markets. The turnover ratio and trading volume
can manipulate the stock market degree of activeness. Gozzi et al. (2008) suggested
accounting for excessive measurement and data availability, turnover ratio is more
suitable. Therefore, turnover ratio was used as a proxy variable for the prosperity of
stock market, noted as in addition, in order to test the robustness of the model,
trading volume as another proxy variable for the stock market prosperity level
was used.
Stock market prosperity is usually defined as activity of buying and selling
behaviour in the market and liquidity characteristics. This paper initially looks at
the linkages between diversity and stock market development using panel data
regression of 187-countries of worldwide (see list in Appendix) for the period of
The Relationship Between Ethnic Diversity and Stock Market Development: A. . . 7

1990–2010 (with 05-years of interval). To analyse the direct relationship of ethnic

and religious diversity with stock market development, we adopt the standard
specification and followed the model proposed by Alesina et al. (2003) and Bove
and Elia (2017).

MCit ¼ ai þ β1 FRACEDit þ β2 FRACRDit þ β3 IC it þ β4 FDI it

ð1Þ
þβ5 GDPPC it þ β6 IST it þ β7 CI it þ β8 GE it þ ai þ εit

VSTGit ¼ ai þ β1 FRACEDit þ β2 FRACRDit þ β3 IC it þ β4 FDI it

þ β5 GDPPC it þ β6 IST it þ β7 CI it þ β8 GE it þ ai þ εit ð2Þ
VSTM it ¼ ai þ β1 FRACEDit þ β2 FRACRDit þ β3 IC it þ β4 FDI it
þ β5 GDPPC it þ β6 IST it þ β7 CI it þ β8 GE it þ ai þ εit ð3Þ

DC it ¼ ai þ β1 FRACEDit þ β2 FRACRDit þ β3 ICit þ β4 FDI it

ð4Þ
þβ5 GDPPC it þ β6 IST it þ β7 CI it þ β8 GE it þ ai þ εit

GEI it ¼ ai þ β1 FRACEDit þ β2 FRACRDit þ β3 IC it þ β4 FDI it

þ β5 GDPPC it þ β6 IST it þ β7 CI it þ β8 GE it þ ai þ εit ð5Þ

Whereas:
MC ¼ market capitalization (% of GDP), αi is the unknown intercept, VSTG ¼
value of share traded (% of GDP), VSTM ¼ value of share trade (% of market
capitalization), DC ¼ No. of listed domestic companies, GEI ¼ S&P/Global equity
index, FRACED ¼ Fractionalization of Ethnic Diversity, FRACRD¼ Fractionaliza-
tion of Religious Diversity, IC¼ Intergroup Cohesion, FDI ¼ Foreign direct invest-
ment, GDPPC¼ GDP per capita PPP, IST¼ Intergroup Safety and Trust, CI¼
corruption perception index and GE¼ Gender equality, at is the unobserved time-
invariant individual effect, and εit is the error term.

3 Methodology

In the context of diversity and stock market development, this study used panel data
methodology to explore this nexus and dynamics for empirical analysis. Basically,
panel data methodology is the mixture of cross-sectional and time series data which
not just increase the power and size of data but also restructuring this effect that is
difficult to distinctive with only cross-sections or time series data (Hsiao 1986).
Baltagi et al. (2003) describes the key advantages of using panel data, such as how
heterogeneity in individual (firms, regions or countries) is absent when using
aggregate time series data.
Whereas in linear unobserved effect model under fixed effect can be expressed as
8 S. Amin

Y it ¼ αi þ X 0it β þ at þ vit

Whereas t ¼ 1,. . .,T, i ¼ 1,. . .,N (i ¼ entity and t ¼ time), αi is the unknown
intercept, also called the individual effect of the individual heterogeneity, it reflects
the unobservable variable that explains the inherent differences between the different
individuals, which are indexed by i. at is the unobserved time-invariant individual
effect. For example, the innate ability for individuals or historical and institutional
factors for countries and vit is the error term. In a fixed effects model, the
un-observed variables are permitted to have any relations whatever with the
observed variables. Fixed effect models organize for or partial out the properties
of time-invariant variables with time-invariant effects. This is correct, whether the
variables are explicitly measured or not, and how do varies by the statistical
technique applied. Unfortunately, the effects of time-invariant variables are cannot
be estimated.
To use fixed effect models are only interested in analysing the impact of variables
that differ over time. When using fixed effect model the individual may affect or bias
the outcome variables and it is necessary to control for this. One of the important
assumptions about fixed effect model is that time-invariant characteristics are unique
and should not be correlated with other explanatory variables for the entity i over
time. Each entity is different so to error term and the constant not to be correlated
with the others variables. If this assumption is violated, we face omitted variables
bias. The second assumption ensures that variables are i.i.d. across entities i ¼ 1,. . .,
n. This does not require the observations to be uncorrelated within an entity. This is
basically a common property of time series data and the same is allowed for errors
vit. In order to control the third factors that do not change with time and the time
effect, a panel model containing individual random effects and time fixed effects
was used.
Data all the variables used in this paper free to access (see Table 1 for data
description). The data of diversity (ethnic and religious) was copied from the
database of Cline Centre for Democracy, University of Illinois, USA. This study
followed the same methodology of Alesina et al. (2003) for diversity calculation
(on the basis of ethnic and religious) by using the following formula.

X
N
FRACj = 1 2 S2ij
i=1

Whereas, Sij is the share of group i, (i ¼ 1,. . .,N) in the country j. The range of the
fractionalization index is between 0 and 1. Zero “0” means homogenous country and
“1” shows total heterogeneous country.
The Relationship Between Ethnic Diversity and Stock Market Development: A. . . 9

Table 1 Description of variables and expected signs

Variables Expected
category Symb. Description sign Data source
Dependent variables
Market capitali- MC Market capitalization (also known WDI
zation (% of as market value) is the share price
GDP) times the number of shares out-
standing (including their several
classes) for listed domestic com-
panies. Investment funds, unit
trusts, and companies whose only
business goal is to hold shares of
other listed companies are
excluded.
Value of share VSTG The value of shares traded is the WDI
traded (% of total number of shares traded,
GDP) both domestic and foreign, multi-
plied by their respective matching
prices. Figures are single counted
(only one side of the transaction is
considered). Companies admitted
to listing and admitted to trading
are included in the data.
Value of share VSTM Turnover ratio is the value of WDI
trade (% of mar- domestic shares traded divided by
ket their market capitalization. The
capitalization) value is annualized by multiply-
ing the monthly average by 12.
No. of listed DC Listed domestic companies, WDI
domestic including foreign companies
companies which are exclusively listed, are
those which have shares listed on
an exchange at the end of the year.
Investment funds, unit trusts, and
companies whose only business
goal is to hold shares of other
listed companies, such as holding
companies and investment com-
panies, regardless of their legal
status, are excluded. A company
with several classes of shares is
counted once. Only companies
admitted to listing on the
exchange are included.
S&P/Global GEI S&P Global Equity Indices mea- WDI
equity index sure the U.S. dollar price change
in the stock markets covered by
the S&P/IFCI and S&P/Frontier
BMI country indices.
(continued)
10 S. Amin

Table 1 (continued)
Variables Expected
category Symb. Description sign Data source
Independent variables
Fractionalization FRACED % of population with ethnic  Cline Centre
of ethnic groups groups (out of total population) of Democracy
and used formulation of Alesina
et al. (2003) for its calculations.
Fractionalization FRACRD % of population with religious  Cline Centre
of religious groups (out of total population) of Democracy
groups and used formulation of Alesina
et al. (2003) for its calculations.
Intergroup IC Intergroup Cohesion measures + ISS
Cohesion ethnic and sectarian tensions, and
discrimination
Intergroup Safety IST Interpersonal Safety and Trust, + ISS
and Trust focusing on perceptions and inci-
dences of crime and personal
transgressions
Gender Equality GE Gender Equality reflecting gender + ISS
discrimination in home, work and
public life.
Foreign Director FDI Foreign direct investment is the  WDI
Investment net inflows of investment to
acquire a lasting management
interest in an enterprise operating
in an economy other than that of
the investor.
GDP per capita GDPPC GDP per capita (current US$)  WDI
Corruption per- CI The CPI scores and ranks coun-  Transparency
ception Index tries/territories based on how cor- International
rupt a country’s public sector is
perceived to be by experts and
business executives.
Note: WDI means world development indictors, database archives. ISS stand for institute of social
development. The indices CI, IST and GE are composed from 21 reputable data sources for
195 countries, over the period from 1990 to 2015, and are updated as new data become available.
The indices are aggregated using the innovative method of ‘matching percentiles’. The S&P/IFCI
Composite is a liquid and investable subset of the S&P Emerging market indices; BMI

4 Empirical Results

The focus of this study is to reveal the impact of diversity, i.e. ethnic and religious,
on dependent variable, i.e. stock market development (including market capitaliza-
tion, market liquidity, turnover ratio, listed domestic companies, S&P/global equity
index). The results of all Tables 2, 3, 4, 5 and 6 shows diversity (i.e. ethnic and
religious) has a significantly positive impact on all the stock market development
Table 2 Diversity and stock market development (Market capitalization in % of GDP). Dependent variable: Market capitalization
Market capitalization (% of GDP)
Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
(12) (11)
FRACEG 0.2299*** 0.1105** 0.3702***
0.1787*** 0.9229**
(0.2797) (0.2838) (0.3533)
(0.5024) (0.2920)
FRACRG 1.0874*** 1.1674*** 0.9487**
0.0947** 0.1251*
(0.3146) (0.3291) (0.3944)
(0.5810) (0.3308)
IC 1.3149** 1.7236*
2.6054***
(0.7884) (0.9378)
(0.7235)
FDI 0.0075** 0.0014 0.0056
(0.0031) (0.0083) (0.0084)
GDPPC 3.2300*** 3.0100*** 1.5600***
(3.4800) (5.7900) (5.5100)
IST 3.5696*** 1.0301 0.6470
(0.8215) (1.0026) (0.7965)
CI 0.0259*** 0.0281 0.0226**
(0.0027) (0.0034) (0.0060)
GE 2.2339* 0.2371 0.340495
(0.6986) (0.7328) (0.7539)
C 3.416*** 3.0914*** 2.7481*** 3.5087*** 2.9204*** 1.7860*** 2.2950*** 1.9574*** 3.0610*** 2.9783*** 1.8945 3.6367***
(0.1331) (0.1398) (0.4853) (0.0709) (0.0928) (0.4408) (0.1586) (0.5107) (0.1644) (0.7503) (0.5217) (0.7692)
R2 0.8239 0.8389 0.9110 0.7182 0.9296 0.8660 0.8512 0.9312 0.7463 0.7271 0.8859 0.91062
Model FE FE FE FE FE FE FE FE FE FE FE FE
Cross- 81 84 78 89 89 86 85 89 79 69 75 68
The Relationship Between Ethnic Diversity and Stock Market Development: A. . .

sections
Obs. 283 297 252 315 320 269 277 319 279 195 241 186
Notes: Regressions are estimated with white cross-section standard errors correction. Heteroscedasticity robust standard error estimates are reported in parentheses.
The asterisks *, ** and*** denotes statistical significance at 10%, 5% and 1% level respectively
11
 12

Table 3 Diversity and stock market development (Market Liquidity). Dependent variable: Value of share trade
Value of share trade (% of GDP)
Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (12) (10) (11)
FRACEG 1.2548*** 1.2372** 0.2934***
0.0792** 0.6149*
(0.5014) (0.5221) (0.5870)
(0.8207) (0.5011)
FRACRG 0.2837*** 0.5014*
1.0003*1.1734** 1.2904**
(0.5759) (0.6022) (0.6456)
(0.9437) (0.5667)
IC 1.4323** 2.7573
3.1503***
(1.3805) (1.5816)
(1.0411)
FDI 0.0002 0.0227 0.0075
(0.0059) (0.0147) (0.0116)
GDPPC 4.6300*** 4.0100** 1.1200**
(6.5400) (9.0200) (8.2500)
IST 8.1851*** 4.1421 1.2159*
(1.4222) (1.6611) (1.0723)
CI 0.0422 0.0570*** 0.0548***
(0.0047) (0.0060) (0.0091)
GE 3.8315 0.7164 1.1704
(1.3233) (1.2685) (1.1157)
C 2.3711*** 1.7777*** 1.0816*** 1.8514*** 0.9503*** 2.1932*** 0.0222*** 0.9034*** 2.1465*** 0.5023*** 0.0215*** 0.3428
(0.2394) (0.2524) (0.8437) (0.1302) (0.1719) (0.7664) (0.2767) (0.9613) (0.2992) (1.2845) (0.8977) (1.1736)
R2 0.8203 0.9077 0.9395 0.8605 0.8278 0.9047 0.9145 0.9739 0.9689 0.9282 0.9083 0.9551
Model FE FE FE FE FE FE FE FE FE FE FE FE
Cross- 84 87 78 93 93 87 86 93 82 68 75 67
sections
Obs. 304 316 273 340 345 285 289 344 298 205 250 194
Notes: Regressions are estimated with white cross-section standard errors correction. Heteroscedasticity robust standard error estimates are reported in parentheses. The
asterisks *, ** and*** denotes statistical significance at 10%, 5% and 1% level respectively
S. Amin
Table 4 Diversity and stock market development (Turnover Ratio). Dependent variable: Value of Share traded
Value of Share traded (% of market capitalization)
Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)(12)
FRACEG 1.2048*** 1.0149*** 0.4524** 0.0083**
0.0876**
(0.3796) (0.3898) (0.4379) (0.4110)
(0.6419)
FRACRG 0.6431*** 0.4963** 0.4174 1.4128***
0.9102**
(0.4490) (0.4575) (0.4895) (0.4714)
(0.7545)
IC 0.0341*** 1.7601* 0.4521
(1.0660) (1.1768) (0.9568)
FDI 0.0079* 0.0224 0.0072
(0.0042) (0.0115) (0.0113)
GDPPC 1.6700*** 1.7900** 2.9700
(5.4300) (7.0200) (7.0000)
IST 4.3477*** 2.4662** 1.5058
(1.0740) (1.2389) (1.0355)
CI 0.0188*** 0.0294*** 0.0296**
(0.0040) (0.0049) (0.0077)
GE 2.2623*** 0.1373 1.3860
(1.0159) (1.0398) (0.9995)
C 3.4518*** 3.2208*** 3.0178*** 2.9689*** 2.5622*** 0.8127*** 2.0621*** 1.2634*** 3.5409*** 2.5981*** 2.1047*** 0.4614
(0.1818) (0.1994) (0.6527) (0.1002) (0.1449) (0.5774) (0.2406) (0.7412) (0.2293) (0.9348) (0.7418) (1.0278)
R2 0.8364 0.7272 0.8504 0.8139 0.9303 0.9635 0.9250 0.8185 0.9362 0.9235 0.9345 0.965291
Model FE FE FE FE FE FE FE FE FE FE FE FE
Cross- 80 84 77 89 89 83 84 89 79 66 74 65
The Relationship Between Ethnic Diversity and Stock Market Development: A. . .

sections
Obs. 268 284 241 299 304 257 263 303 265 186 228 177
Notes: Regressions are estimated with white cross-section standard errors correction. Heteroscedasticity robust standard error estimates are reported in parentheses.
The asterisks *, ** and*** denotes statistical significance at 10%, 5% and 1% level respectively
13
 14

Table 5 Diversity and stock market development (Listed domestic companies). Dependent variable: Listed domestic companies number
No. of listed domestic companies
Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (12) (11)
FRACEG 0.3504*** 0.1833*** 0.1010*** 0.6908**
1.5840***
(0.3283) (0.3400) (0.4267) (0.3698)
(1.2345)
FRACRG 0.1876*** 0.1552** 0.7643** 0.8954*
1.8966**
(0.3872) (0.3936) (0.4762) (0.4248)
(0.8925)
IC 1.7535*** 2.9335**
0.3000**
(0.9828) (1.1341)
(0.4261)
FDI 0.0077** 0.0328 0.0001
(0.0040) (0.0112) (0.0047)
GDPPC 1.4100*** 1.08 8.9300*
(4.6500) (6.6500) (4.8600)
IST 3.9580*** 4.0155*** 0.5862
(0.9611) (1.1598) (0.4792)
CI 0.0158 0.0285 0.0221***
(0.0037) (0.0044) (0.0058)
GE 1.1661 0.0571 0.1149
(0.8660) (0.9257) (0.5014)
C 5.1215*** 5.1110*** 6.2417*** 5.0284*** 4.7109*** 3.1065*** 4.3494*** 4.1442*** 5.0875*** 4.5032*** 3.8009*** 3.9872***
(0.1542) (0.1689) (0.6068) (0.0862) (0.1207) (0.5148) (0.2151) (0.6322) (0.1899) (0.9298) (0.6557) (0.7214)
R2 0.8353 0.8060 0.7969 0.7063 0.8238 0.8510 0.9092 0.9150 0.9313 0.9359 0.9456 0.9746
Model FE FE FE FE FE FE FE FE FE FE FE FE
Cross- 85 88 80 95 95 88 88 95 83 69 76 68
sections
Obs. 326 342 298 370 375 308 316 374 320 222 271 212
Notes: Regressions are estimated with white cross-section standard errors correction. Heteroscedasticity robust standard error estimates are reported in parentheses.
The asterisks *, ** and*** denotes statistical significance at 10%, 5% and 1% level respectively
S. Amin
Table 6 Diversity and stock market development (S&P/Global Equity Indices). Dependent variable: S&P/global equity index
S&P/global equity index (% change)
Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (12) (10) (11)
FRACEG 0.6757** 0.9607** 0.4621***
0.4052** 0.5695**
(0.3824) (0.4014) (0.5177)
(0.5585) (0.4608)
FRACRG 0.7839** 1.0291** 1.0282*
1.5443*** 0.9774**
(0.4120) (0.4343) (0.5337)
(0.6276) (0.4847)
IC 4.6745*** 2.5950*
2.1041**
(1.0326) (1.3838)
(1.6904)
FDI 0.0068 0.0003 0.0026
(0.0114) (0.0121) (0.0144)
GDPPC 1.6700* 3.2400 2.9200
(5.2500) (7.7600) (8.7700)
IST 2.0293** 1.3034 2.4164*
(0.9146) (1.2299) (1.4272)
Corr 0.0102*** 0.0025 0.0091*
(0.0035) (0.0048) (0.0077)
GE 3.9898*** 2.9391** 2.8214
(0.9198) (1.2377) (1.6870)
C 2.4456*** 2.9786*** 5.4040*** 2.6862*** 3.0099*** 3.6904*** 3.1993*** 5.5795*** 2.7479*** 5.1816*** 5.1168 7.2127***
(0.1732) (0.1900) (0.6172) (0.1055) (0.1401) (0.4884) (0.2193) (0.6816) (0.2132) (1.0609) (0.8572) (1.5072)
R2 0.8019 0.8204 0.7255 0.8204 0.8533 0.9281 0.7417 0.79516 0.7799 0.9179 0.9277 0.9873
Model FE FE FE FE FE FE FE FE FE FE FE FE
Cross- 70 74 73 76 76 76 73 76 69 67 66 64
The Relationship Between Ethnic Diversity and Stock Market Development: A. . .

sections
Obs. 163 177 149 180 182 172 170 181 161 126 148 119
Notes: Regressions are estimated with white cross-section standard errors correction. Heteroscedasticity robust standard error estimates are reported in parentheses. The
asterisks *, ** and *** denotes statistical significance at 10%, 5% and 1% level respectively
15
16 S. Amin

indicators. The robustness is represented in all tables, column 1–11. As Table 2

shows at 1% level of significance, ethnic and religious diversity has positive impact
on market capitalization. The coefficient shows 1 unit change in ethnic and religious
diversity, market capitalization boost-up with 0.17% and 0.09% respectively. As
Table 3 shows at 5% and 10% level of significance, ethnic and religious diversity
enhance the market liquidity with 0.79% and 1% respectively. Similarly, as Table 4
shows at 5% level of significance, ethnic and religious diversity enhance the turnover
ratio with 0.08% and 0.9% respectively. As Table 5 shows at 1% and 5% level of
significance, ethnic and religious diversity increment the no. of domestic companies
with 1.5 and 1.8 units respectively. As Table 6 shows at 5% and 1% level of
significance, ethnic and religious diversity enhance the global equality index with
0.40 and 1.5 units respectively. Religious groups also present same behavior in this
case.
All the Tables 2–6 confirmed that ethnic and religious diversity both are impor-
tant indicators, which enhance the stock market development. It means that, if the
diversity is well established, shareholder engagements focus on persuading compa-
nies to take advantage of the benefits associated with gender diversity, including
improved decision-making, oversight and financial performance. These results are
similar with the literature such as diversity enhances the financial market develop-
ment (Baier and Bergstrand 2001; Collier et al. 2000; Watkins and Ferrara 2005;
Yanikkaya 2003). However, in some extent literature also found the relationship
between diversity and stock market development through their socio-economic cost.
They argued that ethnic diversification could be sometimes an engine for produc-
tivity and innovation, because different ethnic groups have different productive
skills, which are complement to each other. Ethnic mix also embodied abilities,
experiences and cultures when turns productive then may lead to innovation and
creativity. Diversity may be helpful to ensure sustainable development because
diversity itself demands for the intergroup cohesion process. Groups coming with
different interests but equal influence will compete with each other, thus forcing
conciliation on the growth-oriented policies from which all could lead to earn profit.
Furthermore, rise in both types of diversity do not create conflict inside and
outside the society because of cohesive measures and collective actions. Fractional-
ization has the aptitude to make societies safe and emphasized that ethnic diversity
has a positive role on stock market because it creates a non-coercive, stable,
development-oriented state through intensive public engagements, it minimizes the
possibility of civil war due to conflict. It improves varied productivity and innova-
tion and lays a foundation for better performance of private sector. Ultimately, the
evidence suggests ethnic fractionalization enhances stock market development. The
results, therefore, strengthen the emerging positive consequences of ethnic diversity
on stock market.
Roberson and Park (2007) affirmed a positive relationship between diversity and
book-to-market equity, and a curvilinear U-shaped relationship between diversity
and revenues, net income and book-to-market equity. Dutta and Mukherjee (2012)
shows that as culture evolve in the form of greater trust, control and other traits,
individuals attitudes towards financial market change, and they engage in greater
The Relationship Between Ethnic Diversity and Stock Market Development: A. . . 17

financial transactions. Considering multiple dimensions of culture, trust is a key

cultural trait, should positively influence stock market development; uncertainty
avoidance, Hofstede’s cultural dimension should negatively influence the develop-
ment of the stock market; and individualism, an alternate cultural dimension of
Hofstede’s measures, should be positively correlated with stock market development
(Dutta and Mukherjee 2015).
According to the World Bank report (2000) emphasized that diversity can reduce
civil war since maintaining the unity of a rebel movement composed of diverse
groups tends to become harder over time. World Bank report concluded that ethnic
diversity is a prevention mechanism rather than a cause of civil war. Collier (1998),
tries to differentiate between ethnic fractionalization and ethnic dominance and
conclude that fractionalization in democratic nations is not a problem by itself
because it creates a reasonable competition among the private sector even though
societies will have worse public sector performance.
In addition to number of control variables such as intergroup cohesion, corruption
perception index, intergroup safety and trust used in empirical analysis. In all the
above-mentioned Tables 2–6, corruption index has found a strong indicator, which
shows significantly positive relationship with stock market development. Our results
also similar with Mouselli et al. (2016) and Shahbaz et al. (2013) by confirmed a
positive impact of corruption on stock market development, such as corruption
greases the wheels of economy by expediting transactions and allowing private
firms to overcome governmental imposed inefficiencies. As well as the corruption
increases, the circulation of money within and outside the countries increases. This
surplus money enhances the financial markets activities, in another aspect increase
the underground economy. These results are opposite to Bolgorian (2011), Yartey
(2010) and Cherif and Gazdar (2010) about negative relationship between corruption
and stock market development. However, GDP per capita and gender equality shows
insignificant relationship with stock market development. The literature shows that
GDP per capita and gender equality has indirectly relationship with financial activ-
ities, basically it’s directly link toward the wellbeing of society (Kabeer and Natali
2013; Masoud 2013).

5 Conclusion and Policy Implication

This study shows the relationship between diversity and stock market behavior by
using data of 187-countries of the world. Based on panel data methodology this
study concluded that diversity has significant positive impact on stock market
development. These results are similar with literature that diversity enhances the
financial market development (Baier and Bergstrand 2001; Collier et al. 2000;
Watkins and Ferrara 2005; Yanikkaya 2003). We live in a deeply connected and
global world. It should come as no surprise that more diverse companies and
institutions are achieving better performance. The unequal performance of compa-
nies in the same industry and the same country implies that diversity is a competitive
18 S. Amin

differentiator shifting market share toward more diverse companies. Companies in

the top quartile for racial and ethnic diversity are 35 percent more likely to have
financial returns above their respective national industry medians. That is particu-
larly true for their talent pipelines: attracting, developing, mentoring, sponsoring,
and retaining the next generations of global leaders at all levels of organizations.
At now, most of the researchers intuitively affirmed that diversity matters. They
believed that diverse markets are better able to win top talent and improve their
customer orientation, employee satisfaction, and decision-making, and all that leads
to a virtuous cycle of increasing returns. Literature also found the relationship
between diversity and stock market development through their socio-economic
cost. Tang et al. (2016) revealed that diversity could promote stock market prosper-
ity and the effect varies from country to country. In particular, it was found that
cultural diversity has positive impact on stock market prosperity in high
development-level and legal system quality countries, and vice versa. Weller and
Zulfiqar (2013) argued that greater diversity is associated with faster growth, larger
credit markets, a broader deposit base, and a smaller chance of asset bubbles, all of
which could contribute to more stability. According to the UN report 2017, if the
diversity is well established, shareholder engagements focus on persuading compa-
nies to take advantage of the benefits associated with gender diversity, including
improved decision-making, oversight and financial performance. Whereas,
intergroup cohesion, GDP per capita and corruption index shows positive relation-
ship with stock market development. Such as the circulation of money increased,
more investors take part in financial activities, which ultimately enhance the stock
market development.
This study showed limitation in this regards that other kinds of diversity for
example, in age, sexual orientation, and experience (such as a global mind-set and
cultural fluency) are also likely to bring some level of competitive advantage for
companies that can attract and retain such diverse talent. This study suggests that
diversity cannot be reduced it is a natural phenomenon; however, positive effect can
be obtained by providing equal opportunity, secure and peaceful society through
cohesiveness.

Appendix 1: Construction of the Fractionalization Index

1. Approach. This paper used ethnic fractionalization index, developed by the

Alesina et al. (2003) for diversity calculation (on the basis of ethnic and religious)
using the following formula.

X
N
FRACj ¼ 1  S2ij
i¼1
The Relationship Between Ethnic Diversity and Stock Market Development: A. . . 19

Whereas, Sij is the share of group i, (i ¼ 1,. . .,N ) in the country j. The range of the
fractionalization index is between 0 and 1. Zero “0” means homogenous country
and “1” shows total heterogeneous country.
2. Data. The data of diversity (ethnic and religious) has been taken from database of
Cline Centre for Democracy, University of Illinois, USA. The dataset contain
annual data from 1990 to 2010 for 187-countries. Data at the Cline Centre based
on various projects that document the changing varieties of social identity around
the world (composition of religious and ethnic groups, CREG). In addition, they
also identified the causes of conflict between religious and ethnic groups. The
Composition of Religious and Ethnic Groups Project (CREG) has started to
create a set of time-varying measures that gauge the nature and depth of
country-specific socio-cultural cleavages. It focused on the largest countries in
the world (all countries with a population above 500,000 (in 2014) during the
post-WWII era to create country-specific projections on the relative sizes of the
different groups during the post-war era.
3. Final index. The data at CREG project shows various types of groups each and
every country at annual basis on the ethnic and religious identity. We have found
the number of ethnic and religious groups of each country out of total
populations. Therefore, it easy to apply the Alesina’s fractionalization index
formula to calculate the diversity for ethnic and religious basis. We constructed
the index values for the year 1990, 1995, 2000, 2005 and 2010 because diversity
is not change over time. We have also compiled the index for 187-countries (out
of 195 totals) due to limitation of data. The result of index can be used for relative
ranking of countries on basis of ethnic and religious identity in current scenarios.
(Please see results at Appendix 3).
 20

Appendix 2: Summary Statistics

Descriptive statistics
VSTG VSTM MC DC GEI FRACED FRACRD IC IST GE FDI GDPPC CI
Mean 34.02 48.77 65.49 498.3 5.159 0.457 0.400 0.599 0.499 0.698 5.151 1285 42.82
Median 8.713 26.22 36.39 158.0 3.789 0.455 0.423 0.606 0.521 0.706 2.687 6631 34.50
Maximum 668.5 1721 1185 7487 158.0 0.989 0.925 0.788 0.773 1.021 341.0 1277 98.14
Minimum 0.000 0.015 0.010 2.000 61.04 0.013 0.003 0.031 0.231 0.211 16.58 257.7 8.400
Std. Dev. 71.31 113.0 104.5 1049 28.97 0.258 0.218 0.100 0.100 0.100 14.76 1632 22.17
Skewness 5.117 11.21 6.942 3.839 1.153 0.037 0.130 0.921 0.414 0.171 15.41 2.547 0.934
Kurtosis 38.72 160.4 66.62 19.16 6.617 1.696 1.746 5.936 2.874 4.537 325.5 11.66 2.696
Jarque-Bera 1985 3204 5654 5004 252.2 51.53 51.83 291.9 15.44 87.89 3714 3678 93.79
Probability 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Sum 1173 1482 2095 1868 1697 331.9 304.0 349.7 263.6 593.5 4373 1123 2689
Sum sq. dev. 1749 3871 3486 4.124 2753 48.35 36.24 5.839 5.320 8.592 1847 2.334 3081
S. Amin
The Relationship Between Ethnic Diversity and Stock Market Development: A. . . 21

The above Appendix 2 shows the descriptive statistics of dependent and inde-
pendents variables. VSTG, VSTM, MC, Dc and GEI have a maximum values such
as 668.5, 1721, 1185, 7487 and 158 and minimum values 0.00, 0.015, 0.010, 2.00
and 61.04 respectively. Maximum values of FRACED and FRACRD are 0.98 and
0.92 and minima are 0.01 and 0.03 respectively. In the line of empirical analysis,
Table 1 shows that data is well organized respective to dependent and independent
variables.

Appendix 3: List of Countries

Andorra Ireland Qatar Samoa Eritrea

Antigua and Israel San Marino Sao and Principe Ethiopia
Barbuda
Australia Italy Saudi Arabia Serbia and Gambia, The
Monten.
Austria Japan Seychelles Solomon Islands Guinea
Bahamas, The Kuwait Singapore South Africa Guinea-Bissau
Bahrain Latvia Slovak Rep. Sri Lanka Haiti
Barbados Liechtenstein Slovenia St. Lucia Zimbabwe
Belgium Lithuania Spain Sudan Senegal
Brunei Darussalam Luxembourg Sweden Afghanistan Sierra Leone
Canada Malta HK, China Benin Somalia
Chile Monaco Hungary Burkina Faso Tanzania
Cyprus Netherlands Iceland Burundi Togo
Czech Republic New Caledonia Switzerland Central African Uganda
Rep.
Denmark New Zealand Trinidad & Chad Rwanda
Tob.
Estonia Norway UAE Comoros Liberia
Finland Oman UK Congo, Dem. Rep. Madagascar
France Palau United States Nepal Malawi
Germany Poland Uruguay Niger Mali
Greece Portugal Fiji Kiribati Mozambique
Albania Cameroon Gabon Kyrgyz Republic Pakistan
Algeria Cape Verde Georgia Lao PDR Panama
Angola China Ghana Lebanon Pap. New
Guinea
Argentina Colombia Grenada Lesotho
Armenia Congo, Rep. Guatemala Libya
Azerbaijan Costa Rica Guyana Macedonia, FYR
Bangladesh Cote d’Ivoire Honduras Malaysia
Belarus Croatia India Maldives
Belize Cuba Indonesia Marshall Islands
(continued)
22 S. Amin

Bhutan Djibouti Iran, Rep. Mauritania

Bolivia Dominica Iraq Mauritius
Bosnia and Herz Dominican Rep. Jamaica Mexico
Botswana Ecuador Jordan Micronesia, Fed.
Sts.
Brazil Egypt, Arab Kazakhstan Moldova
Rep.
Bulgaria El Salvador Kenya Mongolia
Cambodia Equatorial Timor-Leste Turkey
Guinea
Suriname Taiwan, China Tonga Turkmenistan
Swaziland Tajikistan Tunisia Ukraine
Syrian Arab Thailand Vanuatu Uzbekistan
Republic
Venezuela, RB Yemen, Rep. Morocco Paraguay
Vietnam Zambia Namibia Peru
Nicaragua Philippines Russia

References

Alesina A, Ferrara EL (2005) Ethnic diversity and economic performance. J Econ Lit 43
(3):762–800
Alesina A, Giuliano P (2015) Culture and institutions. J Econ Lit 53(4):898–944
Alesina A, Devleeschauwer A, Easterly W, Kurlat S, Wacziarg R (2003) Fractionalization. J Econ
Growth 8(2):155–194
Antonio AL, Chang MJ, Hakuta K, Kenny DA, Levin S, Milem JF (2004) Effects of racial diversity
on complex thinking in college students. Psychol Sci 15(8):507–510
Ashraf Q, Galor O (2011) Cultural diversity, geographical isolation, and the origin of the wealth of
nations. National Bureau of Economic Research
Baier SL, Bergstrand JH (2001) The growth of world trade: tariffs, transport costs, and income
similarity. J Int Econ 53(1):1–27
Baltagi BH, Song SH, Koh W (2003) Testing panel data regression models with spatial error
correlation. J Econ 117(1):123–150
Bolgorian M (2011) Corruption and stock market development: a quantitative approach. Physica A
Stat Mech Appl 390(23–24):4514–4521
Bond P, Edmans A, Goldstein I (2012) The real effects of financial markets. Annu Rev Financ Econ
4(1):339–360
Bove V, Elia L (2017) Migration, diversity, and economic growth. World Dev 89:227–239
Breuer W, Riesener M, Salzmann AJ (2014) Risk aversion vs. individualism: what drives risk
taking in household finance? Eur J Financ 20(5):446–462
Brown R (2000) Social identity theory: past achievements, current problems and future challenges.
Eur J Soc Psychol 30(6):745–778
Cherif M, Gazdar K (2010) Macroeconomic and institutional determinants of stock market devel-
opment in MENA region: new results from a panel data analysis. Int J Bank Financ 7(1):8
Chinn MD, Ito H (2006) What matters for financial development? Capital controls, institutions, and
interactions. J Dev Econ 81(1):163–192
The Relationship Between Ethnic Diversity and Stock Market Development: A. . . 23

Chui AC, Titman S, Wei KJ (2010) Individualism and momentum around the world. J Financ 65
(1):361–392
Collier P (1998) The political economy of ethnicity. World Bank, Washington, DC
Collier P, Elbadawi I, Sambanis N (2000) How much war will we see? Estimating the likelihood
and amount of war in 161 countries, 1960-1998. The World Bank (January), Unpublished
mimeo
DiTomaso N, Post C, Parks-Yancy R (2007) Workforce diversity and inequality: power, status, and
numbers. Annu Rev Sociol 33:473–501
Dutta N, Mukherjee D (2012) Is culture a determinant of financial development? Appl Econ Lett 19
(6):585–590
Dutta N, Mukherjee D (2015) Cultural traits and stock market development: an empirical analysis. J
Entrepreneurship Public Policy 4(1):33–49
Egger PH, Von Ehrlich M, Nelson DR (2012) Migration and trade. World Econ 35(2):216–241
Elbadawi E, Sambanis N (2000) Why are there so many civil wars in Africa? Understanding and
preventing violent conflict. J Afr Econ 9(3):244–269
Florida R (2002) The economic geography of talent. Ann Assoc Am Geogr 92(4):743–755
Forti CAB, Yen-Tsang C, Peixoto FM (2011) Stock market development: an analysis from a
multilevel and multi-country perspective. BAR-Brazilian Admin Rev 8(4):351–375
Frame WS, White LJ (2004) Empirical studies of financial innovation: lots of talk, little action? J
Econ Lit 42(1):116–144
Gozzi JC, Levine R, Schmukler SL (2008) Internationalization and the evolution of corporate
valuation. J Financ Econ 88(3):607–632
Guiso L, Sapienza P, Zingales L (2006) Does culture affect economic outcomes? J Econ Perspect 20
(2):23–48
Hall RE (2001) The stock market and capital accumulation. Am Econ Rev 91(5):1185–1202
Hayek FA (1945) The use of knowledge in society. Am Econ Rev 35(4):519–530
Head K, Ries J (1998) Immigration and trade creation: econometric evidence from Canada. Can J
Econ:47–62
Herbig P, Dunphy S (1998) Culture and innovation. Cross Cult Manag 5(4):13–21
Herring C (2009) Does diversity pay?: Race, gender, and the business case for diversity. Am Sociol
Rev 74(2):208–224
Hogg MA (2007) Uncertainty–identity theory. Adv Exp Soc Psychol 39:69–126
Hsiao C (1986) Analysis of panel data, Econometric Society Monograph No. 11. Cambridge
University Press, Cambridge
Jain A, Jordan C (2009) Diversity and resilience: lessons from the financial crisis. UNSWLJ 32:416
Kabeer N, Natali L (2013) Gender equality and economic growth: is there a win-win? IDS Working
Papers 2013(417):1–58
Kahneman D (2003) A perspective on judgment and choice: mapping bounded rationality. Am
Psychol 58(9):697
Kwok CC, Tadesse S (2006) National culture and financial systems. J Int Bus Stud 37(2):227–247
Levine R, Zervos S (1998) Stock markets, banks, and economic growth. Am Econ Rev:537–558
Masoud NM (2013) The impact of stock market performance upon economic growth. Int J Econ
Financ Issues 3(4):788
Miller DJ (2006) Technological diversity, related diversification, and firm performance. Strateg
Manag J 27(7):601–619
Montalvo JG, Reynal-Querol M (2005) Ethnic diversity and economic development. J Dev Econ 76
(2):293–323
Mouselli S, Aljazaerli MA, Sirop R (2016) Corruption and stoCK marKet development: neW
evidenCe from gCC Countries. Bus Theory Pract 17:117
Nettle D, Grace JB, Choisy M, Cornell HV, Guégan J-F, Hochberg ME (2007) Cultural diversity,
economic development and societal instability. PLoS One 2(9):e929
North DC (1991) Institutions. J Econ Perspect 5(1):97–112
24 S. Amin

Pan L, Mishra V (2018) Stock market development and economic growth: empirical evidence from
China. Econ Model 68:661–673
Patrick HT (1966) Financial development and economic growth in underdeveloped countries. Econ
Dev Cult Chang 14(2):174–189
Pettigrew TF, Tropp LR, Wagner U, Christ O (2011) Recent advances in intergroup contact theory.
Int J Intercult Relat 35(3):271–280
Putnam RD (2007) E pluribus unum: Diversity and community in the twenty-first century the 2006
Johan Skytte Prize Lecture. Scand Polit Stud 30(2):137–174
Richard OC (2000) Racial diversity, business strategy, and firm performance: a resource-based
view. Acad Manag J 43(2):164–177
Roberson QM, Park HJ (2007) Examining the link between diversity and firm performance: the
effects of diversity reputation and leader racial diversity. Group Org Manag 32(5):548–568
Rousseau PL, Wachtel P (2000) Equity markets and growth: cross-country evidence on timing and
outcomes, 1980–1995. J Bank Financ 24(12):1933–1957
Sacasa N (2008) Preventing future crises. Financ Dev 45(4):11–14
Shahbaz M, Hye QMA, Shabbir MS (2013) Does corruption increase financial development? A
time series analysis in Pakistan. Int J Econ Empirical Res 1(10):113–124
Sommers SR (2006) On racial diversity and group decision making: identifying multiple effects of
racial composition on jury deliberations. J Pers Soc Psychol 90(4):597
Sommers SR, Warp LS, Mahoney CC (2008) Cognitive effects of racial diversity: white individ-
uals’ information processing in heterogeneous groups. J Exp Soc Psychol 44(4):1129–1136
Stulz RM, Williamson R (2003) Culture, openness, and finance. J Financ Econ 70(3):313–349
Tang Z, Liu Q, Li Z (2016) Cultural diversity and stock market prosperity
Taydas Z, Peksen D, James P (2010) Why do civil wars occur? Understanding the importance of
institutional quality. Civil Wars 12(3):195–217
Van Knippenberg D, Schippers MC (2007) Work group diversity. Annu Rev Psychol 58:515–541
Verick S, Islam I (2010) The great recession of 2008–2009: causes, consequences and policy
responses
Watkins A, Ferrara EL (2005) Ethnic diversity and economic performance. J Econ Lit 43
(3):762–800
Weller CE, Zulfiqar G (2013) Financial market diversity and macroeconomic stability. Political
Economy Research Institute Working Paper 332
Williams KY, O’Reilly CA III (1998) Demography and diversity in organizations. Res Organ
Behav 20:77–140
World Bank (2000) World development report 2000/2001: attacking poverty. The World Bank,
Oxford University Press, New York
Yanikkaya H (2003) Trade openness and economic growth: a cross-country empirical investigation.
J Dev Econ 72(1):57–89
Yartey CA (2010) The institutional and macroeconomic determinants of stock market development
in emerging economies. Appl Financ Econ 20(21):1615–1625
Zulfiqar GM, Weller C (2013) Financial market diversity and macroeconomic stability
Does Withholding Tax Reduce
International Income-Shifting by FDI?

Anna Białek-Jaworska

1 Introduction

On the one hand, foreign direct investment (FDI) inflows solve problems of limited
investment issues and firm growth related to financing constraints (Harrison et al.
2004). FDI inflows bring such scarce capital to recipient firms in emerging countries
and relax constraints to firm growth caused by limited access to debt or equity
financing. FDI inflows augment investment resources and facilitate development
(Kose et al. 2006; Henry 2007). FDI also brings knowledge to the host countries,
improving both efficiencies of capital allocation and productivity (Bonfiglioli 2008;
Bekaert et al. 2011; Benigno and Fornaro 2014; Benigno et al. 2015).
On the other hand, MNEs use debt FDI and equity financing to shift the profits
from affiliates or subsidiaries in host countries with higher taxes through dividends
or, even better, interests which can be deducted from taxable income (and allow to
benefit from interest tax shield). This issue gradually becomes more apparent in the
literature (Kudła 2018; Schimanski 2018; Polish Economic Institute 2020), the EU is
looking for instruments to limit such unfair practice. Several initiatives were made:
Base Erosion and Profit Shifting by OECD (2015), Anti-Tax Avoidance Directive
by European Commission, and an allowance for corporate equity (ACE) rule in
European countries (European Committee 2018).
Among factors that influence the FDI inflow into the host country, an important
role is played by comparative advantages described in trade theories (Gudowski and
Piasecki 2020) and tax competition (Kudła et al. 2015). After 1989, a financial
capital movement’s liberalization led to implementing the withholding tax (WHT)
rate for FDI related to income-shifting (royalties, dividends, and interests). The

A. Białek-Jaworska (*)
Faculty of Economic Sciences, University of Warsaw, Warsaw, Poland
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 25

K. Jajuga et al. (eds.), Contemporary Trends and Challenges in Finance, Springer
Proceedings in Business and Economics,
https://doi.org/10.1007/978-3-030-73667-5_2
26 A. Białek-Jaworska

efficiency of this instrument raises a new research question on the role of WHT. Our
study is even more essential because of the recent amendments to the Polish tax law
that restrict thin capitalization and limit taxable debt costs since 2018 (Bialek-
Jaworska 2018). Next, it introduces a new mechanism of collecting WHT tax for
dividend payments (in Poland and abroad) over 2 million PLN annually per one
taxpayer. For lower annual payouts, the WHT agent will have to meet additional
requirements to apply for a reduced WHT.
Moreover, interests and discounts on corporate bonds can be free of the WHT
when the specific conditions are met. The legitimacy of applying a reduced WHT is
limited to confirming that the recipient is a business partner engaged in actual
business activity in the country of its registered office. A WHT taxpayer may
apply to the tax authority for an opinion on applying for a WHT collection exemp-
tion. The tax authority has 6 months to respond, but this opinion is valid for
36 consecutive months. However, the tax authority may also reject the application
if tax avoidance provisions apply (PWC 30.12.2020). We shed light on this urgent
issue by analyzing WHT’s impact on debt and equity FDI and related interest and
dividend payments. This research design allows us to contribute to the literature.
The chapter aims to check whether WHT on interests and dividends reduces
income-shifting by FDI. Thus, we intend to assess this tax policy tool’s efficiency by
measuring the sign and significance of WHT’s impact on debt FDI and equity FDI
inflow to Poland in 2010–2018 and their withdrawal. We target identifying which
type of MNEs (vertically or horizontally integrated) use debt FDI instruments rather
than traditional equity FDI.
The remainder of the chapter is organized as follows. Section 2 briefly reviews the
literature and formulates the research hypotheses. Section 3 describes the method-
ology and research design, and Sect. 4 presents the results of our analysis. We
conclude in Sect. 5.

2 Literature Review

The causes and role of FDI in the development of emerging markets, including
Poland, have been studied extensively in the literature (i.e., Cieślik 2019a, b;
Gorynia et al. 2015; Igan et al. 2020). The previous empirical analyses show that
multinational companies (MNEs) are motivated by different advantages (market
access motive, efficiency-seeking by vertically or horizontally integrated MNEs)
in investment decisions. However, FDI may also be used for income-shifting to
avoid taxes. Kałdoński (2016) studied in depth tax avoidance of European compa-
nies using the micro-econometric approach.
In tax avoidance and tax planning literature, debt and equity FDI are known as
income-shifting instruments from high-tax to low-tax countries. This issue is
discussed in the literature reviewed by Fonseca and Juca (2020). Borrowing from
members of business groups located in low-tax home countries and lending to
affiliates in high-tax host countries allows borrowers to deduct interest payments
Does Withholding Tax Reduce International Income-Shifting by FDI? 27

from taxable income and benefit from debt tax shield (Mintz and Smart 2004). The
difference in tax rates between home and host countries stimulates vertical integra-
tion due to transfer pricing for MNEs (Egger and Seidel 2013). Damgaard et al.
(2019) highlight that the possibility of tax payment saving guides FDI, and a share of
such foreign investment is directed to tax havens. For German MNEs, Buttner and
Wamser (2007) show that tax-rate differences within the multinational business
group influence income-shifting with intercompany loans. The authors assessed
the implied tax revenue effects as relatively small. This effect can be caused by
substantial costs related to adjusting the capital structure for income-shifting pur-
poses. However, their research uses micro-level panel data made available by the
Bundesbank. Therefore, our study can contribute to the literature by analyzing
macro-level FDI data retrieved from the Balance of Payments.
Buettner et al. (2018) support our assumptions by pointing out that MNEs may
use foreign operations’ financial structure to shift taxable income. Intercompany
loans may be used to create interest payments, which are deducted in countries with
high taxes (Fuest et al. 2011; Dischinger et al. 2014; Hansson and Olofsdotter 2014;
Devereux et al. 2018). MNEs that operate in a developed capital market optimize
their tax payments by intercompany loans provided inside their business groups,
from parent companies to their subsidiaries and among subsidiaries or affiliates
(Desai et al. 2004; Bénassy-Quéré et al. 2005; Azémar 2010; Barrios et al. 2012;
Egger et al. 2014; Hebous and Ruf 2017).
The EU aims to prevent profit-shifting via extraordinary debt financing by
intercompany loans. Therefore the EU limits such opportunities increasing the
restrictiveness of thin-capitalization rules. On the one hand, this has positive effects
via curbing tax planning, but on the other—it negatively impacts investment (Merlo
et al. 2019). Thus, the EU tries to solve this inefficiency by introducing an allowance
for corporate equity (ACE) to achieve tax neutrality between debt and equity
financing. In some EU countries, the ACE decreased debt financing in total but
generated a rise in intra-group lending within MNEs for tax planning purposes
(Hebous and Ruf 2017). Moreover, Buettner et al. (2018) show that anti-profit
shifting legislation intensifies the adverse tax effects on FDI if imposed by host
countries with relatively high tax rates.
Polish Economic Institute (2020) shows that 17 billion PLN of corporate profit
was transferred abroad, adding to Poland’s corporate income tax (CIT) gap. Among
the recipient countries, there are the following tax havens: Ireland, Netherlands,
Luxemburg, Switzerland, Cyprus, Belgium, and Malta. Because the Polish Eco-
nomic Institute (2020) omits possible tax policy tools preventing such an income-
shifting and focuses only on the size of the CIT gap, there is a place to contribute. By
analyzing WHT’s impact on debt and equity FDI and related interest and dividend
payments, we shed light on this urgent issue and contribute to the literature. So far,
Arena and Roper (2010) have provided evidence that differences in international tax
rates and tax regimes affect MNEs’ debt location decisions. They analyze micro-
level data of 8287 debt issues from 2437 firms headquartered in 23 different
countries with debt-issuing subsidiaries in 59 countries, considering differences in
personal and corporate (PIT and CIT) tax rates, tax credit and exemption systems,
28 A. Białek-Jaworska

and bilateral cross-country WHT on interest and dividend payments. These

researchers show that MNEs issue debt through subsidiaries located in countries
with higher WHT on dividend payments. Simultaneously, MNEs issue more con-
siderable debt in foreign countries with low or non-existing WHT on interest
payments. Our chapter adds to the literature by focusing on WHT’s impact on the
debt FDI and equity FDI instruments’ income-shifting role.
We contribute to the state-of-the-art of FDI by a separate analysis of debt FDI and
equity FDI instruments, and an extension of the knowledge-capital (KC) model
analyzed by Cieślik (2019a) for total FDI inflow to Poland till 2015. Cieślik (2019a)
shows that the horizontal dimension explains FDI inflow to Poland from 15 EU
members. Besides identifying integration type of MNEs that provide debt FDI or
equity FDI to Poland, we check whether bilateral cross-country WHT is a tool to
reduce the income-shifting via debt or equity FDI. Thus, we state the following
hypotheses:
H1: Withholding tax (WHT) on interest payments reduces income-shifting by debt
FDI instruments.
H1A: Higher WHT on interests motivates MNEs to withdraw debt FDI instruments.
H2: WHT on dividend payments reduces income-shifting by equity FDI.
H2A: Higher WHT on dividends increases the withdrawal of equity FDI.
H3: Vertically integrated MNEs are likely to use debt FDI instruments rather than
traditional equity FDI.

3 Research Design

This chapter studies the determinants of debt and equity FDI to Poland in
2010–2018, and their withdrawals. We use inflow FDI data retrieved from the
Central Bank of Poland (Narodowy Bank Polski, NBP) website. They allow us to
separate debt from equity FDI instruments. Earlier, separate debt FDI and equity FDI
data (a component of the Balance of Payments) were analyzed by Igan et al. (2020)
to study their impact on the growth of industries in the emerging markets (including
Poland) in 1998–2010. Contrary to that study, our research focuses on the profit
shifting with the use of interests or dividend transfers from Poland to other countries,
including 11 tax heavens (Cyprus, Hong Kong, Ireland, Liechtenstein, Luxembourg,
Malta, the Netherlands, Singapore, British Virgin Islands, Andorra, Gibraltar,
United Arab Emirates, Jersey, Cayman Island, Panama, Saint Kitts and Nevis, and
the Isle of Man).
Table 1 presents the definition of variables. Four positions retrieved from the
Balance of Payments are used as a dependent variable: debt FDI inflow to Poland,
equity FDI inflow to Poland, and both of their withdrawals. We focus on the WHT’s
impact on interests (dividends) on income shifting using FDI.
We apply the theoretical knowledge-capital (KC) model. Because we distinguish
between debt and equity FDI inflows and withdrawals, we replace the dependent
Does Withholding Tax Reduce International Income-Shifting by FDI? 29

Table 1 Definition of variables

Variables Definition of variables
Explained variable(s) (y_FDI)
debt fdi ln(debt foreign direct investment (FDI) inflow from country i to Poland)
withdraw debt ln(withdrawal of debt FDI from country i to Poland)
fdi
equity fdi ln(equity foreign direct investment (FDI) inflow from country i to Poland)
withdraw ln(withdrawal of equity FDI from country i to Poland)
equity fdi
Explanatory variables
kdiff Capital per worker difference between host and home countries calculated as ln
(a ratio of the national capital stocks expressed in PPPs in constant 2011 USD
to total employees)
hdiff Differences in human capital endowments between host and home countries
calculated as ln(a ratio of the human capital indexes that are based on the mean
years of schooling and returns to education to total employees)
sum ln(sum of country-lender’s and Poland’s GDPs)
loans ln(intercompany loans as a part of debt instruments of FDI)
borrowings
sdi Helpman’s size dispersion index that is calculated using data on output-side
real GDP at chained purchasing power parity (PPP) rates and expressed in
constant 2011 US dollars for Poland and a country which companies provide
loans to Polish firms

2

gdpi gdp 2
sdiij ¼ 1  sum i,j
 sumi,jj
gdpi/sumi, j Country i’s share in the sum of GDPs calculated using data on output-side real
GDP at chained purchasing power parity (PPP) rates and expressed in constant
2011 US dollars
gdpj/sumi, j Country j’s share in the sum of GDPs calculated using data on output-side real
GDP at chained purchasing power parity (PPP) rates and expressed in constant
2011 US dollars
distance A proxy for trade costs equals to ln(a geographical distance of each parent
country’s capital city from Warsaw that is measured “as the crow flies”
distance between the capital cities of the host and home country, and it is
expressed in kilometers)
wht_int Bilateral cross-country withholding tax rate on interest payments (Deloitte
2019)
wht_div Bilateral cross-country withholding tax rate on dividend payments (Deloitte
2019)
kaufman A mean of six Kaufmann’s Worldwide Governance Indicators’ estimates
(Kaufmann et al. 2010)

variable used in the well-known KC model with the separate components of FDI
inflows. The main explanatory variables of FDI inflows in the KC model are the
similarity in economic size (sdi), the summation of Poland’s and the home country’s
GDPs (sum), relative factor endowments based on human and physical capital per
worker (kdiff, hdiff), and a geographical distance of each pair home and host
countries’ capital cities. The relative factor endowments describe Poland’s
30 A. Białek-Jaworska

differential and the countries investing in Polish firms by equity injections or giving
them loans. The sdi, kdiff, and hdiff explanatory variables allow us to identify how
MNEs investing in Polish firms are integrated: horizontally or vertically, depending
on investment: debt or equity. The Penn World Table 9.1 is the primary data source
used to measure the KC model components. The research limitations deal with the
availability of the Penn World data only until 2017. Therefore we used 1-year lagged
data for the sdi, kdiff, hdiff, sum explanatory variables used in the theoretical KC
model. This solution was previously applied by Cieślik (2019a). We extend the
classic KC model by adding additional explanatory variables like bilateral cross-
country WHT rate on interest wht_int (dividend wht_div) payments and Kaufmann’s
Worldwide Governance Indicator of home countries’ governance quality (kaufman).
The aggregate Kaufmann’s indicators are based on several hundred individual
underlying variables, taken from a wide variety of existing data sources that are
grouped in six dimensions: Voice and Accountability, Political Stability and
Absence of Violence/Terrorism, Government Effectiveness, Regulatory Quality,
Rule of Law, and Control of Corruption.
By assessing the statistical significance and signs of the estimated coefficients on
the sdi, kdiff, hdiff explanatory variables, we identify what cross-country pattern of
MNEs integration explains the debt FDI or equity FDI investments better. A positive
coefficient at the similarity in the economic size of a pair of countries confirms the
privilege of horizontal integration and the importance of the market access motive,
whereas its insignificance proves the vertical integration favor. We use the
Helpman’s size dispersion index (sdi) to measure the similarity in relative country
size (see definition provided by Table 1). Negative coefficients at the differences in
relative physical and human capital factor endowments support horizontally inte-
grated MNEs, while a significant positive relationship proves the favor of the
efficiency-seeking motive over the market access motive. We control institutional
governance using a mean of six Kaufmann’s indices (regulatory quality, voice &
accountability, political stability & absence of violence, government effectiveness,
the rule of law, and control for corruption). The geographical distance between
capital cities of host and home countries controls for trade costs. We focus on the
WHT on interests and dividends impact on the income-shifting as an international
tax avoidance tool separately. Therefore, we consider a withdrawal of both types of
FDI instruments too.
Table 1 provides detailed definitions of all explained and explanatory, both test
and control variables.
Table 2 shows descriptive statistics, whereas Table 3 illustrates the correlation
coefficients of explanatory variables. Missings are caused by lack of data in the Penn
World Table 9.1 for the Isle of Man, Jersey, Gibraltar, Andorra, Liechtenstein,
Guernsey in total, and a lack of human capital and employment data for Saint
Kitts and Navis, Cayman Islands, British Virgin Islands. There is a lack of
Kaufmann Governance Indicators in the case of Guernsey, Gibraltar, and the Isle
of Man.
We study the effectiveness of bilateral cross-country WHT instruments in reduc-
ing income-shifting by interests on foreign loans or dividends paid to foreign
Does Withholding Tax Reduce International Income-Shifting by FDI? 31

Table 2 Descriptive statistics of the variables

Variable N Mean Std. err. Min Max
debt fdi 595 2.24 2.66 0.00 9.55
withdraw debt fdi 595 1.60 2.41 0.00 9.24
equity fdi 595 1.92 2.74 0.00 9.91
withdraw equity fdi 595 1.17 2.42 0.00 9.72
kdiff 503 11.76 1.03 6.18 13.21
hdiff 503 0.72 1.56 7.84 2.92
sum 567 14.20 0.68 13.51 16.78
loans borrowings 595 3.32 3.77 0.00 10.77
sdi 567 0.26 0.17 0.00 0.50
distance 586 7.71 1.06 5.85 9.78
wht_int 586 0.11 0.06 0.00 0.20
wht_div 586 0.14 0.04 0.05 0.20
kaufman 559 0.76 0.76 0.97 1.87

shareholders using the KC model. We estimate the models for four different depen-
dent variables using the one-step Arellano-Bond dynamic panel-data estimator,
adjusting standard errors for clustering on the country. We focus on the impact of
WHT on interests and dividends on income-shifting using debt or equity FDI. This
approach bases on the following equation of the extended KC model:

X
p
y FDI ijt ¼ α0 y FDI ij,tk þ β1 sumijt þ β2 sdiijt þ β3 kdiff ijt þ β4 hdiff ijt
k¼1

þ β5 distanceij þ β6 wht int ijt þ β7 wht divijt þ β8 controls þ vij

þ εijt ð1Þ

Next, we apply the generalized method of moments GMM Arellano-Bond

dynamic panel-data estimator. We estimate separate models for four different
dependent variables: debt FDI inflow, equity FDI inflow, and both of their with-
drawals (y_FDI). The GMM approach uses instruments for the differenced equation.
Next, it eliminates the disadvantages of reduced sample size (Arellano and Bond
1991). The dynamic panel-data system allows a better understanding of adjustment
dynamics when current behavior depends on past behavior.

4 Results

Table 4 presents results obtained from the estimations of models using the GMM
Arellano-Bond dynamic panel-data estimator. The WHT on interests (dividends) is
the leading test variable, while Kaufman’s governance indices is a control variable.
 32

Table 3 Correlation matrix

kdiff hdiff sum loans sdi dist whtint whtdiv kaufman
kdiff 1
hdiff 0.083 1
sum 0.083 0.54 1
loans 0.530 0.018 0.228 1
sdi 0.219 0.69 0.111 0.109 1
distance 0.21 0.198 0.218 0.329 0.014 1
wht_int 0.46 0.025 0.204 0.548 0.202 0.459 1
wht_div 0.23 0.07 0.145 0.168 0.003 0.169 0.434 1
kaufman 0.64 0.282 0.09 0.564 0.031 0.156 0.378 0.008 1
A. Białek-Jaworska
Table 4 Results of one-step Arellano-Bond dynamic panel-data estimation (Std. Err. adjusted for clustering on country)
debt fdi debt fdi withdraw debt fdi equity fdi withdraw equity fdi
(1) (2) (3) (4) (5)
debt fdi 0.024
0.066
L1. 0.2386 *** 0.2386 *** 0.090
(0.086) (0.086) (0.070)
L2. 0.2730 *** 0.2730 *** 0.157 **
(0.104) (0.104) (0.066)
L3. 0.055
(0.087)
withdraw debt
L1. 0.1164 *
(0.0692)
L2. 0.1368
(0.1023)
equity fdi
L1. 0.126 #
(0.0962)
L2. 0.110
Does Withholding Tax Reduce International Income-Shifting by FDI?

(0.0963)
withdrawequity
L1. 0.236 ***
(0.071)
L2. 0.234 ***
(0.065)
sum 9.835 9.8352 7.9990 12.5463 3.1120
(continued)
33
 34

Table 4 (continued)
debt fdi debt fdi withdraw debt fdi equity fdi withdraw equity fdi
(1) (2) (3) (4) (5)
(7.610) (7.6101) (8.7905) (7.7909) (6.017)
L1. 1.5216 * 1.5216 * 1.4044 * 0.3545 2.9004
(0.872) (0.8722) (0.8141) (0.7202) (6.576)
L2. 1.187 1.1872 1.8876 ** 3.0190 ** 2.3974 ***
(0.930) (0.9296) (0.8082) (1.3862) (0.474)
sdi 14.543 14.5426 2.7216 25.480 ** 9.4824
(11.70) (11.697) (14.612) (12.208) (12.62)
L1. 17.93 *** 17.928 *** 18.4831 *** 1.7164
(5.631) (5.6313) (4.9585) (11.90)
L2. 9.6014 ** 9.6014 ** 6.1432 6.8648 ***
(4.843) (4.8426) (4.0012) (1.713)
kdiff 0.683 *** 0.6827 *** 0.7969 ## 0.0100 0.748 ***
(0.246) (0.2457) (0.5032) (0.1404) (0.239)
L1. 0.2404 0.2404 0.1101 0.3633 0.459
(0.432) (0.4322) (0.5718) (0.2226) (0.334)
L2. 0.640 *
(0.3504)
hdiff 4.1875 ** 4.1875 ** 5.034 *** 1.3032 0.302
(1.681) (1.6808) (1.6891) (1.2981) (1.116)
L1. 2.391 *** 2.3908 *** 2.3711 *** 0.1604 1.6053
(0.849) (0.8486) (0.8093) (0.3387) (2.040)
L2. 0.9641 ** 0.9641 ** 0.2924 0.2019
(0.448) (0.4476) (0.3790) (0.4166)
distance 63.929 * 58.3601 #
(38.476) (44.362)
A. Białek-Jaworska
wht_int 102.4 *** 102.4 *** 113.4 *** 91.62 ***
(36.59) (36.590) (27.786) (31.67)
wht_div 94.247 ** 94.2468 ** 123.36 *** 35.1598 ## 123.22 ***
(43.30) (43.304) (35.015) (23.68) (42.67)
equity fdi 0.0387 0.0387 0.0412
(0.070) (0.0700) (0.0639)
L1. 0.1245 * 0.1245 * 0.1010
(0.068) (0.0685) (0.0628)
L2. 0.1440 * 0.1440 * 0.1541 **
(0.080) (0.0801) (0.0712)
loansborrowings 0.2816 *
(0.1606)
L1. 0.1953
(0.2232)
kaufman 6.1854 ** 6.1854 ** 8.135 ** 2.1244
(2.745) (2.7453) (3.6940) (2.080)
year 0.3084 * 0.3084 * 0.115 0.325 ## 0.046
(0.191) (0.1913) (0.2317) (0.2168) (0.165)
Sargan test 32.278 32.278 27.424 53.46 52.423
p-value 0.1501 0.1501 0.3351 0.001 0.0004
Does Withholding Tax Reduce International Income-Shifting by FDI?

Arellano-Bond test
AR1 9.074 *** 9.074 *** 8.739 *** 4.51 *** 7.729 ***
p-value 0.0000 0.0000 0.0000 0.0000 0.0000
AR2 0.8857 0.8857 1.7451 0.957 0.283
p-value 0.3758 0.3758 0.0810 0.339 0.7774
(continued)
35
 36

Table 4 (continued)
debt fdi debt fdi withdraw debt fdi equity fdi withdraw equity fdi
(1) (2) (3) (4) (5)
N observations 332 332 332 332 277
N groups 56 56 56 56 56
N instruments 45 46 45 42 43
Wald test 132.19 *** 139.02 *** 84.64 *** 61.96 *** 213.48 ***
***p < 0.01, **p < 0.05, and *p < 0.1, #p < 0.15, ##p < 0.2
A. Białek-Jaworska
Does Withholding Tax Reduce International Income-Shifting by FDI? 37

Models in the first and second columns explain debt FDI inflow to Poland. In the
model in the third column, we analyze its withdrawal. Thus, we control equity FDI
inflow in these models. In the model that describes equity FDI inflow to Poland, we
handle intercompany loans and borrowings (debt FDI instruments) as a control
variable.
Our findings show a negative correlation between WHT on interests and debt FDI
inflow to Polish firms (models 1 and 2), which align with the H1 hypothesis. Thus,
we confirm that WHT on interest payments reduces international income-shifting by
debt FDI instruments. The positive impact of the WHT on interests on the with-
drawal of debt FDI (a dependent variable in model 3) and the departure of equity FDI
from Polish firms (model 5) follow our expectations and confirm the supportive
hypothesis H1A. Therefore, we provide evidence that higher WHT on interest
payments motivates MNEs to withdraw debt FDI instruments from Polish firms.
The positive coefficient at the equity FDI inflow to Polish firms (explained variable
of the model 4) supports the H2 hypothesis but at a low (15%) level of significance.
This result means that WHT on dividends reduces international income-shifting
by equity FDI, but the impact is less significant than in the case of WHT on interests.
Moreover, the results of model 5, shown in the last column, confirm that an increase
in the WHT on interests forces MNEs to withdraw equity FDI from Poland.
However, this finding is contrary to the impact of WHT on dividends because higher
WHT on dividend payments reduces the withdrawal of equity FDI instruments (see
models 5). Therefore, we reject the H2A hypothesis. We identify that higher WHT
on dividends decreases withdrawal of equity FDI, discordant to our expectations.
However, higher WHT on dividends stimulates debt FDI inflow to Poland according
to the results of models in columns (1) and (2). This relationship is in line with the
negative impact of WHT on dividends on withdrawal of debt FDI presented in
column (3). Nonetheless, WHT on interests positively influences the withdrawal of
equity FDI instruments. To summarize, these results indicate that higher WHT on
dividends motivates MNEs to provide intercompany loans and other debt FDI
instruments instead of equity FDI.
We provide evidence that debt FDI inflow to Poland increases (its withdrawal
decreases) with the size of the combined market of both lending and borrowing
countries (sum). The similarity of pair-countries’ economic size positively affects the
withdrawal of both debt and equity FDI instruments. This coefficient shows that
market access matters only for investment exits. However, in work-intensive indus-
tries, the efficiency-seeking motive is more important than the market access in debt
FDI. Positive coefficients at the differences in relative factor endowments (hdiff) in
models 1 and 2 confirm this finding. Thus, the debt FDI determinants’ results support
the H3 hypothesis on the privilege of vertical integration of MNEs that provide debt
FDI to Polish firms in work-intensive industries. The negative coefficient at the
distance variable in column (2) supports the H3 hypothesis that vertically integrated
MNEs likely provide debt FDI to Poland rather than equity FDI. Higher trade costs
(proxied by a larger geographical distance between capital cities of host and home
countries) discourage MNEs from providing debt FDI to Polish firms. Thus, closer
neighbor countries provide more debt FDI to Polish firms. The negative influence of
38 A. Białek-Jaworska

the human capital relative factor endowments differential (model 3) identifies that
horizontally integrated MNEs likely withdraw debt FDI, while vertically integrated
MNEs renew debt FDI provided to Polish firms.
Regarding control variables, higher debt FDI, including intercompany loans,
comes from countries with good quality governance and better institutional envi-
ronment measured by a mean of Kaufmann’s Governance Indicators.
Interestingly, the governance quality (including voice and accountability, politi-
cal stability and absence of violence, governance efficiency, regulatory quality, the
rule of law, and control for corruption) of host countries matters neither for equity
FDI nor their withdrawal.

5 Conclusions

The literature on business groups underlines the role of loans provided within the
internal capital market in liquidity management and financing investment projects of
enterprises with limited bank loan access. Loans guarantee a higher and more certain
rate of return than the dividend yield (Białek-Jaworska et al. 2019). The internal
capital market may shift lending from less effective projects to more effective ones
(Buchuk et al. 2014). However, restrictive thin capitalization rules limit borrowers’
internal borrowing benefits since 2018 in Poland, although it is still more attractive
than equity infusion. Intragroup loans’ reallocation is more common and plays a
more critical role in countries with less developed capital markets (Stein 1997).
Intra-group loans are used to manage cash excesses in one firm and cash shortages in
another (Bialek-Jaworska et al. 2020). Non-financial companies step into banks’
shoes by lending using money gained from their cash flow, especially when they
have long-term investments (Bialek-Jaworska 2017), including shares in affiliates.
Next, business groups use internal revenues to set up or acquire capital-intensive
firms (Almeida et al. 2011), as corporate investments are partly financed by the
dividends paid out by the group members (Gopalan et al. 2014).
On the supply side, lenders may benefit from inter-corporate loans by lending at a
higher interest rate than in the case of alternative investment (e.g., bank deposits).
Intercorporate loans result from the redistribution of cash holdings and money
borrowed from banks (Bialek-Jaworska et al. 2020). The redistribution effect is
better known for the trade credit (Bialek-Jaworska and Nehrebecka 2016) that is also
a component of the debt FDI. While on the demand side, inter-corporate loans are
emergency financing for small firms with limited access to bank loans, suffering
from financial constraints, and lacking creditworthiness (Bialek-Jaworska et al.
2019).
This chapter contributes to the literature with new findings that WHT taxation of
interest payments effectively limits income-shifting by debt FDI instruments, con-
trary to WHT on dividends. However, negative externalities are observed because
higher WHT on dividends motivates MNEs to shift from equity to debt FDI
financing.
Does Withholding Tax Reduce International Income-Shifting by FDI? 39

Our results should raise the policymakers’ attention as debt and equity FDI are
used as a tool for income-shifting from firms located in countries with higher taxes to
low (or null) tax countries (including tax heavens). We show that an effective
solution is to increase WHT on interests to reduce profit-shifting activity and stop
the tax base’s erosion in Poland. Under regulation valid till 2018, WHT on dividends
seems inefficient, contrary to WHT on interests. A rise in WHT on dividend payouts
motivates MNEs to shift to debt FDI instruments and provide intercompany loans
instead of equity FDI.
WHT aside, this chapter finds vertically integrated MNEs likely to provide debt
FDI to Polish firms. We adopt and extend the knowledge-capital model to study
separate debt and equity FDI focusing on WHT on interest and dividend payouts.
Using the Arellano-Bond dynamic panel data approach and controlling home coun-
tries’ governance policy, we add to Cieślik (2019a)’s study on total FDI inflow from
15 EU member countries to Poland. We learn that home countries with better
institutional environments provide more debt FDI to Poland. We add to the actual
discussion on the FDI instruments’ income-shifting that causes Poland’s income tax
gap. This issue was raised by Polish Economic Institute (2020). Our study’s practical
implications suggest how income-shifting practices can be reduced via FDI by
increasing WHT on interests transferred to the home countries.

Acknowledgments The article was prepared as a part of the project financed under the Bekker
Programme by the Polish National Agency for Academic Exchange, Contract No. PPN/BEK/2018/
1/00426/U/00001 (research scholarship at Queensland University).

References

Almeida H, Campello M, Weisbach M (2011) Corporate financial and investment policies when
future financing is not frictionless. J Corp Finan 17:675–693
Arellano M, Bond S (1991) Some tests of specification for panel data: Monte Carlo evidence and an
application to employment equations. R Econ Stud 58(2):277–297. https://doi.org/10.2307/
2297968
Arena MP, Roper AH (2010) The effect of taxes on multinational debt location. J Corp Finan 16
(5):637–654. https://doi.org/10.1016/j.jcorpfin.2010.07.006
Azémar C (2010) International corporate taxation and U.S. multinationals’ behaviour: an integrated
approach. Can J Econ/R Can D'écon 43(1):232–253. https://doi.org/10.1111/j.1540-5982.2009.
01570.x
Barrios S, Huizinga H, Laeven L et al (2012) International taxation and multinational firm location
decisions. J Public Econ 96(11–12):946–958. https://doi.org/10.1016/j.jpubeco.2012.06.004
Bekaert G, Harvey CR, Lundblad C (2011) Financial openness and productivity. World Dev 39
(1):1–19
Bénassy-Quéré A, Fontagné L, Lahrèche-Révil A (2005) How does FDI react to corporate taxation?
I Tax Pub Fin 12(5):583–603. https://doi.org/10.1007/s10797-005-2652-4
Benigno G, Fornaro L (2014) The financial resource curse. S J Econ 116(1):58–86
Benigno G, Converse N, Fornaro L (2015) Large capital inflows, sectoral allocation, and economic
performance. J Int Money Financ 55(C):60–87
40 A. Białek-Jaworska

Bialek-Jaworska A (2017) Do Polish non-financial listed companies hold cash to lend money to
other firms? EBR 3(17):87–110. https://doi.org/10.18559/ebr.2017.4.6
Białek-Jaworska A (2018) Uwarunkowania, mechanizmy i efekty udzielania pożyczek przez
przedsiębiorstwa niefinansowe (Conditions, mechanisms and effects of lending money by
non-financial companies). CeDeWu, Warsaw
Bialek-Jaworska A, Nehrebecka N (2016) The role of trade credit in business operations. AOE 2
(37):189–231. http://www.dbc.wroc.pl/dlibra/editions-content?id¼38481
Bialek-Jaworska A, Gadowska-dos Santos D, Faff R (2019) Conceptual framework for lending
money outside business groups: evidence from Poland. In: Jajuga K, Locarek-Junge H,
Orłowski L, Staehr K (eds) Contemporary trends and challenges in finance. Springer, Cham,
pp 71–80
Bialek-Jaworska A, Faff R, Zieba D (2020) A liquidity redistribution effect in intercorporate
lending: evidence from private firms in Poland. ERSJ 23(1):151–175. https://doi.org/10.
35808/ersj/1543
Bonfiglioli A (2008) Financial integration, productivity and capital accumulation. J Int Econ 76
(2):337–355
Buchuk D, Larrain B, Muñoz F, Urzúa F (2014) The internal capital markets of business groups:
evidence from intra-group loans. J Financ Econ 112(2):190–212
Buettner T, Overesch M, Wamser G (2018) Anti profit-shifting rules and foreign direct investment.
Int Tax Public Financ 25(3):553–580. https://doi.org/10.1007/s10797-017-9457-0
Buttner T, Wamser G (2007) Intercompany loans and profit shifting – Evidence from company-
level data. CESifo Working Paper Series WP 1959. CESifo, Munich
Cieślik A (2019a) Determinants of foreign direct investment from EU-15 Countries in Poland.
CEEJ 6(53):39–52. https://doi.org/10.2478/ceej-2019-0007
Cieślik A (2019b) What attracts multinational enterprises from the new EU member states to
Poland? EBR 10(2):253–269. https://doi.org/10.1007/s40821-019-00122-z
Damgaard J, Elkjaer T, Johannesen N (2019) The rise of phantom investments. https://www.imf.
org/external/pubs/ft/fandd/2019/09/pdf/the-rise-of-phantom-FDI-in-tax-havens-damgaard.pdf.
Accessed 1 July 2020
Deloitte (2019) Withholding tax rates. https://www2.deloitte.com/content/dam/Deloitte/global/
Documents/Tax/dttl-tax-withholding-tax-rates.pdf. Accessed 20 Nov 2019
Desai MA, Foley C, Hines JR (2004) Foreign direct investment in a world of multiple taxes. J
Public Econ 88(12):2727–2744. https://doi.org/10.1016/j.jpubeco.2003.08.004
Devereux MP, Maffini G, Xing J (2018) Corporate tax incentives and capital structure: new
evidence from UK firm-level tax returns. J Bank Financ 88:250–266. https://doi.org/10.1016/
j.jbankfin.2017.12.004
Dischinger M, Knoll B, Riedel N (2014) The role of headquarters in multinational profit shifting
strategies. Int Tax Public Financ 21(2):248–271
Egger P, Seidel T (2013) Corporate taxes and intra-firm trade. Eur Econ Rev 63:225–242
Egger P, Keuschnigg C, Merlo V et al (2014) Corporate taxes and internal borrowing within
multinational firms. AEJ: Econ Policy 6(2):54–93. https://doi.org/10.1257/pol.6.2.54
European Committee (2018) Tax policies in the European Union. 2018 Survey. https://eceuropaeu/
taxation_customs/sites/taxation/files/tax_policies_survey_2018.pdf. Accessed 5 July 2020
Fonseca P, Juca M (2020) The influence of taxes on foreign direct investment: systematic literature
review and bibliometric analysis. ERSJ 23(2):55–77. https://doi.org/10.35808/ersj/1580
Fuest C, Hebous S, Riedel N (2011) International debt shifting and multinational firms in devel-
oping economies. Econ Lett 113(2):135–138. https://doi.org/10.1016/j.econlet.2011.06.012
Gopalan R, Nanda V, Seru A (2014) Internal capital market and dividend policies: evidence from
business groups. Rev Financ Stud 27(4):1102–1142. https://doi.org/10.1093/rfs/hhu004
Gorynia M, Nowak J, Trapczyński P et al (2015) Should governments support outward FDI? The
case of Poland. In: Marinova S (ed) Institutional impacts on firm internationalization. Palgrave
Macmillan, London, pp 120–145
Does Withholding Tax Reduce International Income-Shifting by FDI? 41

Gudowski J, Piasecki R (2020) Foreign direct investment from emerging markets. Theory and
practice. Comp Econ Res 23(2):7–19. https://doi.org/10.18778/1508-2008.23.09
Hansson Å, Olofsdotter K (2014) Labor taxation and FDI decisions in the European Union. Open
Econ Rev 25(2):263–287. https://doi.org/10.1007/s11079-013-9282-8
Harrison AE, Love I, McMillian MS (2004) Global capital flows and financing constraints. J Dev
Econ 75(1):269–301
Hebous S, Ruf M (2017) Evaluating the effects of ACE systems on multinational debt financing and
investment. J Public Econ 156:131–149. https://doi.org/10.1016/j.jpubeco.2017.02.011
Henry PB (2007) Capital account liberalization: theory, evidence, and speculation. J Econ Lit
45:887–935
Igan D, Kutan AM, Mirzaei A (2020) The real effects of capital inflows in emerging markets. J
Bank Fin (in press). https://doi.org/10.1016/j.jbankfin.2020.105933
Kałdoński M (2016) Unikanie opodatkowania z perspektywy nadzoru korporacyjnego. Badania
nad odrębnością zjawiska w firmach rodzinnych. Wydawnictwo UEP, Poznań
Kaufmann D, Kraay A, Mastruzzi M (2010) The worldwide governance indicators: methodology
and analytical issues. World Bank Policy Research Working Paper No. 5430. https://ssrn.com/
abstract¼1682130
Kose MA, Prasad ES, Rogoff K et al (2006) Financial globalization: a reappraisal. IMF Working
Paper 06/189
Kudła J (2018) Anti-avoidance regulatory policy for international transfer pricing and excess debt
financing. SSRN Electron J. https://doi.org/10.2139/ssrn.3169976
Kudła J, Kocia A, Kopczewska K et al (2015) Optimal fiscal policy in an open economy with capital
income shifting and consumer cross-border purchases. Equilibrium 10(2):9–30. https://doi.org/
10.12775/equil.2015.011
Merlo V, Riedel N, Wamser G (2019) The impact of thin-capitalization rules on the location of
multinational firms’ foreign affiliates. ROIE 28(1). https://doi.org/10.1111/roie.12440
Mintz J, Smart M (2004) Income shifting, investment, and tax competition: theory and evidence
from provincial taxation in Canada. J Public Econ 88(6):1149–1168. https://doi.org/10.1016/
s0047-2727(03)00060-4
OECD (2015) Measuring and monitoring BEPS. Action 11. https://www.oecd-ilibrary.org/taxation/
measuring-and-monitoring-beps-action-11-2015-final-report_9789264241343-en. Accessed
30 June 2020
Polish Economic Institute (2020) The CIT gap in Poland in 2004–2018. Warsaw
PWC. Withholding tax in Poland. https://www.pwc.pl/en/services/tax-services/international-tax-
law/withholding-tax-in-poland.html. Accessed 30 Dec 2020
Schimanski C (2018) Do multinational companies shift profits out of developing countries? How
data availability may hide the evidence 2018/52. WIDER Working Paper
Stein JC (1997) Internal capital markets and the competition for corporate resources. J Financ
52:111–133
The Relationship Between Trading Volume
and Returns Volatility on Warsaw Stock
Exchange

Lesław Markowski

1 Introduction

Investors throughout the trading day receive a lot of information at different times.
This information is reflected in intermediate equilibrium prices and when investors
glean the new information the final informational equilibrium in prices is reached.
The flow and adoption of information is defined as sequential information arrival
hypothesis (SIAH) (Copeland 1976). It can be assumed that with the simultaneous
arrival of information, the trading volume of any asset will increase. Therefore, the
volume can be a variable predicting variability in price changes. The Volume can
represent the stochastic information flows process and therefore it can be used for
construction estimators of liquidity and volatility of financial instruments
(Będowska-Sójka and Kliber 2019, 2021; Olbryś 2018). There should be a positive
correlation between trading volume and absolute returns of price changes (Copeland
and Friedman 1987). This statement, in turn, is the subject of the mixture of
distribution hypothesis (MDH) and proposed by Clark (1973). Price changes and
trading volume as an information arrival rate are independently distributed but they
can also be considered as a joint distribution. This implies positive contemporaneous
relation between price volatility and volume. Conditional variance of price change
and trading volume are positive function of information arrival. The potential impact
of trading volume on price volatility can be tested using GARCH model in which the
variance of returns is heteroskedasticity due to the information that arrive to the
market. Using the contemporaneous trading volume should decrease the volatility
persistence in conditional variance equation compared to the equation without that

L. Markowski (*)
Department of Finance, Faculty of Economic Sciences, University of Warmia and Mazury,
Olsztyn, Poland
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 43

K. Jajuga et al. (eds.), Contemporary Trends and Challenges in Finance, Springer
Proceedings in Business and Economics,
https://doi.org/10.1007/978-3-030-73667-5_3
44 L. Markowski

volume. This statement has found support in many studies. Lamoureux and
Lastrapes (1990a) discovered that phenomenon of variance clustering diminishes
when the contemporaneous trading volume is included to the variance equation.
They also confirmed that GARCH measures of persistence in variance are sensitive
to type of model misspecification (Lamoureux and Lastrapes 1990b). Gallo and
Pacini (2000) also observed reduction of the estimated persistence in the GARCH
(1) and EGARCH(1) for US stocks. Bohl and Henke (2003) in Polish stock market
showed a reduction of persistence in volatility after including trading volume in the
model. Gorgul and Wójtowicz (2006) demonstrated that for the equities listed in the
DJIA index volume and returns volatility exhibit long memory. Louhichi (2011) on
Euronext proved that trading volume decreases the persistence of volatility when the
volume is introduced in the conditional variance equation. Ezzat and Kirkulak-
Uludag (2014) in the Saudi Exchange provided strong evidence for the validity of
the mixture of distribution hypothesis. They demonstrated that volatility persistence
decreases when the trading volume is included in the conditional variance equation.
The same effect Tan et al. (2015) revealed using the sample in Australian stock
market. MDH and SIAH are verified by Tseng et al. (2015) and Shen et al. (2018).
They supported the sequential information arrival using lagged volume to predict
volatility. Wang et al. (2019) in the Bitcoin market documented significant relation-
ship between trading volume and return volatility and supported the SIAH incorpo-
rated different kinds of measurements of volume and sample periods.
In this paper, the main problem is to define the role of volume as such, and in the
context of price changes in the perception of information, often private information,
by traders. The existence of strong simultaneous relations between the volume and
the volatility of rates of return denotes that investors’ decisions depend on the level
of volume. In addition to the impact of trading volume on volatility, a problem of the
relationship was raised in connection with the sign of arrival rate. The question arises
whether good or bad information strengthened the volume and whether there is an
asymmetry in the impact of a given type of information. A broad study on this topic
was carried out by Tan et al. (2015). They proposed price movement indicators and
revealed that model with these variables reduce persistence in returns’ volatility
more than models with volume alone. Furthermore, they observed that upward price
movement influence conditional variance more than a downward price movement.
The objective of this study is to verify the relationship between return volatility
and mixing variable as volume using GARCH(1) specification in Polish capital
market. Changes of persistence in volatility are investigate considering trading
volume in positive and negative price movement.
The Relationship Between Trading Volume and Returns Volatility on Warsaw. . . 45

2 Methodology

2.1 Conditional Variance–Volume Relation

The explanation that stock returns are generated according with a mixture distribu-
tion hypothesis can be verify in using of proper ARCH model. That model can
describe of the rate of information arrivals as a generating process of mixing vari-
ables (Ahmed et al. 2005). One of that variables is trading volume and it is
incorporating to explain of conditional variance stock returns. The volume plays
an important role from investors’ point of view because it reflects their expectations.
It seems important to study the joint distribution of returns and volume to confirm the
mixture of distribution hypothesis.
The returns in the research period for trading day are calculated as follows:



Rt ¼ ln Pclose
t  ln Popen
t ð1Þ

The adopted methodology for determining the rates of return and price-volume
indicators was ordered by several reasons. We can consider trading as a trading
period and overnight non-trading period. In this article only price changes during the
trading period were calculated. Moreover, this approach enables investors to avoid
the influence of noise traders on prices. The role of noise traders in the context of the
arbitrage approach to financial markets can be found at work (Shleifer and Summers
1990).
The GARCH(1) model is the tool that use contemporaneous trading volume and
allow to measure the level of persistence’ reduction. Including trading volume
generalised variance specification with the mean equation considering autocorrela-
tion in returns is given by (Tan et al. 2015):

Rt ¼ φ0 þ φ1 Rt1 þ εt ð2Þ
σ 2t ¼ α0 þ α1 ε2t1 þ β1 σ 2t1 þ δVolt , ð3Þ
~ ð0, 1Þ
εt ¼ σ t ϑt ; ϑt IID ð4Þ

where Volt is return and trading volume in trading period t. The volume reflects an
appropriate proxy of information arrival rate, where news tends to increase the
trading volume. Therefore, the estimation of δ is expected to be positive, which
means increasing conditional variance of stock returns. The sum of parameters α1and
β1 measures the level of volatility persistence. Including the mixing variable in the
form of volume should reduce of the persistence in volatility that is eliminate the
ARCH effect in time periods of returns. Nevertheless, the volume should be treated
as a weakly exogenous variable (Lamoureux and Lastrapes 1990a). Karpoff (1987)
argued that, if volume is not exogenous the regression of return volatility on volume
can be biased.
46 L. Markowski

2.2 Trading Volume and Price Movement Direction

in Conditional Variance

In this paragraph, the study is expanded to include the impact of the type of
information (good and bad news) on volatility-volume relation. Investors may
react in distinct way to different types of information. Taking into account growing
and descending price movement the conditional variance equation with a trading
volume is as follows:

σ 2t ¼ α0 þ α1 ξ2t1 þ β1 σ 2t1 þ δ1 Volþ 

t þ δ2 Volt ð5Þ

where
(
Volt if Popen  Pclose <0
Volþ
t t
t ¼ ð6Þ
0 if otherwise
(
Volt if Popen
t  Pclose
t 0
Vol
t ¼ ð7Þ
0 if otherwise

and Popen
t , Pclose
t denote open and close trading price of any asset. Obviously, the
price-volume indicators as variables (6) and (7) aggregate to the trading volume.
Equation (5) leads to the following two research questions. Firstly, whether volume
in upward and downward price movement periods more reduce a persistence in
conditional variance than contemporaneous trading volume in Eq. (3). Secondly,
assuming that investors react much more strongly and immediately to good news
than to bad information, good arrival rate persuade conditional variance more than
bad arrival rate.

3 Data

A dataset includes a series of daily log returns of 10 stocks quoted on the Warsaw
Stock Exchange and belonging to the index WIG20, of which seven stocks are
characterized by full time series. The trading volume is the daily transaction volume
(number of shares traded during the day).
The sample period is from January 4, 2010 to December 30, 2019 what spans
2497 observations. Descriptive statistics of returns of stocks are given in Table 1.
The skewness of a half of stocks is positive indicating right tail in distributions. The
excess kurtosis for all shares is leptokurtic. The statistics of the Jarque-Bera test
reject the hypothesis that the distributions of returns for all companies are normal.
The results of the Ljung-Box test (are not present in the table) indicate that the rates
of return exhibit autocorrelation in volatility, which justifies the use of ARCH class
The Relationship Between Trading Volume and Returns Volatility on Warsaw. . . 47

Table 1 Summary statistics of daily returns

Stock Mean SD Skew. Kurtosis J-B ADF
Alior Bank 0.00087 0.019 0.015 1.402 142.87 31.2 [0.000]
Bank Pekao 0.00067 0.016 0.609 7.666 6268.49 37.1 [0.000]
CCC 0.00044 0.020 0.170 3.071 992.99 35.6 [0.000]
CD Projekt 0.00065 0.027 0.308 7.911 6551.44 38.5 [0.000]
Cyfrowy Polsat 0.00028 0.018 0.058 1.338 187.70 37.5 [0.000]
Dino Polska 0.00047 0.021 0.357 3.140 289.45 25.9 [0.000]
Grupa Lotos 0.00019 0.019 0.008 1.475 226.23 48.1 [0.000]
JSW 0.00099 0.029 0.682 6.678 4097.33 43.6 [0.000]
KGHM 0.00137 0.019 0.349 3.150 1082.94 46.8 [0.000]
LPP 0.00008 0.020 0.127 2.787 814.89 35.5 [0.000]
Notes: J-B-Jarque-Bera test of normality with critical value as χ 2(2) ¼ 5.991; ADF-Augmented
Dickey-Fuller test of stationarity in the version of intercept and trend
Source: Own study

Table 2 Stationarity and autocorrelation of daily volume

Stock ADF LB-Q (4) LB-Q (8) LB-Q (12)
Alior Bank 22.5 [0.000] 386.2 [0.000] 631.8 [0.000] 927.3 [0.000]
Bank Pekao 26.3 [0.000] 492.0 [0.000] 636.3 [0.000] 722.8 [0.000]
CCC 27.8 [0.000] 1211.1 [0.000] 2110.6 [0.000] 3004.7 [0.000]
CD Projekt 17.9 [0.000] 3554.9 [0.000] 5623.6 [0.000] 7709.2 [0.000]
Cyfrowy Polsat 29.0 [0.000] 311.9 [0.000] 423.4 [0.000] 534.7 [0.000]
Dino Polska 14,3 [0.000] 146.9 [0.000] 185.7 [0.000] 243.6 [0.000]
Grupa Lotos 22.9 [0.000] 933.9 [0.000] 1232.3 [0.000] 1432.3 [0.000]
JSW 21.1 [0.000] 1356.7 [0.000] 2337.1 [0.000] 3209.0 [0.000]
KGHM 24.3 [0.000] 984.1 [0.000] 1367.5 [0.000] 1657.3 [0.000]
LPP 30.1 [0.000] 289.2 [0.000] 477.0 [0.000] 676.9 [0.000]
Notes: ADF-Augmented Dickey-Fuller test of stationarity in the version of intercept and trend;
LB-Q(k)-Ljung-Box test for the join significance of autocorrelation
Source: Own study

models. Augmented Dickey-Fuller test indicates that daily returns are stationary time
series. For modelling conditional variance of returns is used GARCH(1) model
which provides a good fit. The GARCH model will be used despite the limitations
caused by the weakness of this model in situation with rapid changes in the level of
volatility (Fiszeder and Perczak 2016).
Some diagnostic tests were performed for trading volume and the results are
displayed in Table 2.
The Ljung-Box test was applied to investigate serial autocorrelation in the
volume series. The statistics of LB-Q test for different lags provide a conclusive
rejection of the null hypothesis that the volume time series is not autocorrelated. This
serial correlation appearance is necessary in implementing the mixture of distribu-
tion hypothesis (MDH) with autoregressive conditional heteroskedasticity model
specification. The MDH implies positive contemporaneous relation between price
48 L. Markowski

and volume and the volume create the conditional heteroscedasticity of returns (Pati
and Rajib 2010). The ADF test confirms the stationarity of volume time series.

4 Results

The study of the impact of volume on conditional variance compares the estimates of
GARCH(1) model and the same model with trading volume. Results in Table 3
indicate the high level of persistence in nine of ten stocks which varies from 0.6287
to 0.9999. On the other hand, estimates of models with contemporaneous trading
volume in Table 4 reveal a substantial decreasing of volatility persistence in nine

Table 3 Estimates of Model σ 2t ¼ α0 þ α1 ε2t1 þ β1 σ 2t1

GARCH(1) model
Stock b
α1 b
β1 α1 þ b
b β1
Alior Bank 0.1517*** 0.6265*** 0.7782
Bank Pekao 0.0508*** 0.9271*** 0.9779
CCC 0.0418*** 0.9217*** 0.9635
CD Projekt 0.1088*** 0.8439*** 0.9527
Cyfrowy Polsat 0.0871*** 0.5496*** 0.6367
Dino Polska 0.1248*** 0.5039*** 0.6287
Grupa Lotos 0.0646*** 0.8913*** 0.9559
JSW 0.0761*** 0.9097*** 0.9858
KGHM 0.0518*** 0.9252*** 0.9770
LPP 0.2054*** 0.0000 0.2054
Notes: ***, **, * indicates significance at the 1%, 5% and 10%
respectively
Source: Own study

Table 4 Estimates of GARCH(1) model with trading volume

Model σ 2t ¼ α0 þ α1 ε2t1 þ β1 σ 2t1 þ δVolt
Stock b
α1 b
β1 α1 þ b
b β1 b
δ
Alior Bank 0.1361 ***
0.0687 **
0.0674 8.59  107***
Bank Pekao 0.0495*** 0.9288*** 0.9783 7.99  1010
CCC 0.1389*** 0.2000*** 0.3389 1.71  106***
CD Projekt 0.0031 0.0609*** 0.0578 1.69  106***
Cyfrowy Polsat 0.0762*** 0.0631* 0.0131 3.46  107***
Dino Polska 0.0092 0.0533** 0.0441 2.25  106***
Grupa Lotos 0.0591*** 0.0569* 0.0022 1.09  106***
JSW 0.0490*** 0.0472*** 0.0018 2.35  106***
KGHM 0.0404** 0.0795*** 0.0391 4.71  107***
LPP 0.1542*** 0.0332 0.1208 8.27  105***
Notes: ***, **, * indicates significance at the 1%, 5% and 10% respectively
Source: Own study
The Relationship Between Trading Volume and Returns Volatility on Warsaw. . . 49

6000

5000
Volume in thousands of shares

4000

3000

2000

1000

0
2010 2012 2014 2016 2018 2020
Year

Fig. 1 The volume of KGHM in the period of 2010–2019. Source: Own study

stocks. The results of estimated models demonstrate that GARCH effects become
much smaller when trading volume is included to the variance equation. These
results are consistent with those obtained by Lamoureux and Lastrapes (1990a)
and Koulakiotis et al. (2007). Furthermore, trading volume has statistically positive
influence on conditional variance of these stocks.
Despite the general conclusions consistent with expectations, the negative esti-
mates of parameter β1 pay attention because they exceed the restrictions for this
coefficient in the conditional variance equation. In the presence of volume with
positive parameter δ, the parameters α1 and β1 are expected positive and insignificant
if daily volume is serially correlated. Lifting restriction on these parameters should
satisfy the requirement of a positive volatility process. Verification of the volatility
process gave positive values of the conditional variance for all t. Nelson and Cao
(1992) showed is not always the case that negative coefficients in GARCH models
may result from misspecification or sampling error.
Similarities of contemporaneous changes in trading volume and conditional
variance are given by plots in Figs. 1 and 2 for example for transactions of
KGHM shares. Periods of increased investor activity are marked with graphic
symbols in Fig. 1. They correspond in most cases to periods of relatively higher
values of conditional variance in Fig. 2. It follows that if information arrival rate
increases and thereby trading volume, then the volume becomes the explanatory
variable for volatility clustering.
50 L. Markowski

0,0016

0,0014

0,0012
Conditional variance

0,001

0,0008

0,0006

0,0004

0,0002

0
2010 2012 2014 2016 2018 2020
Year

Fig. 2 The conditional variance from GARCH(1) model for KGHM in the period of 2010–2019.
Source: Own study

The next stage of the study of volatility-volume relationships is the estimation of

GARCH (1) models covering the impact of both volume and a price movement. The
results are presented in Table 5.
The model outperforms the model only with trading volume in the variance
equation in case of three out of ten stocks as far as persistence reduction. For all
stocks, the sum of b α1 þ b
β1 is lower than the sum of these parameters for variance
equation without any exogenous variables. Variables considering price changes
make an additional contribution to the explanation of conditional variance. As
expected, the impact is mostly positive and statistically significant in the case of
nine stocks both in positive and negative price movement. However, the above effect
is asymmetric. Using the Wald test in seven stocks the combined variable (upward
price movement and volume) has a greater impact on volatility than the variable
connecting downward price movement with trading volume. This shows a stronger
and less immediate response of investors to good news than bad news. These results
are consistent with the results obtained by Tan et al. (2015). The explanation for such
results can be volume levels in different price movement periods. On the example of
KGHM, Fig. 3 gives the plots of trading volume of KGHM shares in upward and
downward price movement, respectively.
For observations with positive price movement the volume was characterized by
periods of increased investor activity throughout the sample period, while for
Table 5 Estimates of GARCH(1) model with trading period indicators

Model σ 2t ¼ α0 þ α1 ε2t1 þ β1 σ 2t1 þ δ1 Volþ
t þ δ2 Volt
Stock b
α1 b
β1 b
α1 þ b
β1 b
δ1 b
δ2 δ1 > δ2
Alior Bank 0.1204*** 0.0305*** 0.0899 2.95  106 *** 3.87  107 *** Y
Bank Pekao 0.0487*** 0.9289*** 0.9776 2.87  109 4.26  109 N
CCC 0.1399*** 0.2209 0.3608 1.37  106 ** 1.95  106 *** N
CD Projekt 0.0015 0.0156** 0.0171 3.08  106 *** 9.27  107 *** Y
Cyfrowy Polsat 0.0404*** 0.0199** 0.0205 1.29  106 *** 1.54  107 *** Y
Dino Polska 0.0149 0.0208 0.0059 4.30  106 *** 1.22  106 *** Y
Grupa Lotos 0.0177** 0.0272*** 0.0095 2.43  106 *** 6.62  107 *** Y
JSW 0.0340*** 0.0284*** 0.0056 4.45  106 *** 1.36  106 *** Y
KGHM 0.0524*** 0.0829*** 0.0305 5.85  107 *** 3.32  107 *** Y
LPP 0.1539*** 0.0335 0.1204 8.21  105 *** 8.34  105 *** N
Notes: ***, **, * indicates significance at the 1%, 5% and 10% respectively
Source: Own study
The Relationship Between Trading Volume and Returns Volatility on Warsaw. . .
51
52 L. Markowski

6000
Volume in thou. of shares in upward price movement

5000

4000

3000

2000

1000

0
2010 2012 2014 2016 2018 2020
Year

6000
Volume in thou. of shares in downward price movement

5000

4000

3000

2000

1000

0
2010 2012 2014 2016 2018 2020
Year

Fig. 3 The volume of KGHM in upward and downward price movements in the period of
2010–2019. Source: Own study
The Relationship Between Trading Volume and Returns Volatility on Warsaw. . . 53

Table 6 The average trading volume of KGHM in thou. of shares

Subsample period
Direction of price Subsample 2010– Subsample 2015– Whole sample 2010–
movement 2014 2019 2019
Positive price movement 922.2 855.0 889.7
Negative price 999.3 756.1 874.2
movement
Source: Own study

periods with negative price movement, increased volume can only be seen in the first
part of the sample period. This is confirmed by the values in the Table 6.
The average values of volume for upward price movement in the second sub-
sample and whole sample are higher than in downward price movement.

5 Conclusions

The paper presents research the relationship between return volatility and trading
volume using GARCH(1) model on the example of ten companies listed on the
Warsaw Stock Exchange and belonging to the WIG20 index. Changes of persistence
in volatility are investigate considering trading volume in upward and downward
price movement.
The study provides some important findings. The higher trading volume the
higher volatility of stock returns. Next, the results of variance equation reveal that
contemporaneous trading volume causes a substantial decreasing of volatility per-
sistence in nine out of stocks and it has a statistically positive influence on condi-
tional variance of these stocks. This findings with the results of Doman (2008),
which showed a significant influence of the volume on the volatility for the most
liquid companies. Wójtowicz (2008) confirmed the significant linear relationships
between the conditional variance of returns and the residuals of the ARFIMA-
FIGARCH model for log-volume. The model with indicator variables considering
both the volume and price movement demonstrate a reduction of the persistence in
three stocks compared to the model with trading volume alone. Moreover, as
expected, for more stocks the upward price movement has a greater impact on
conditional variance than the downward price movement. The combined variables,
volume and indicator of price movement direction have positive and statistically
significant influence on returns’ variation.
The work analysed the relationship between instrumental variables for the infor-
mation process volume and price change versus volatility. The research verifies
investors’ behaviour in the trading period due to the type of information related to
movement in price. Therefore, traders can consider the volume as an additional
source of information. Observation of the trading volume by investors can increase
54 L. Markowski

the advantage over other market players. Furthermore, it expands the capabilities of
modelling and forecasting variation of traded stocks returns.

References

Ahmed HJA, Hassan A, Nasir AMD (2005) The relationship between trading volume, volatility and
stock market returns: a test of mixed distribution hypothesis for a pre- and post crisis on Kuala
Lumpur Stock Exchange. Investment Manag Financ Innov 2(3):146–158
Będowska-Sójka B, Kliber A (2019) The causality between liquidity and volatility in the Polish
Stock Market. Financ Res Lett 30:110–115
Będowska-Sójka B, Kliber A (2021) Information content of liquidity and volatility measures.
Physica A Stat Mech Appl 563:125436. https://doi.org/10.1016/j.physa.2020.125436
Bohl MT, Henke H (2003) Trading volume and stock market volatility: the Polish case. Int Rev
Financ Anal 12(5):513–525. https://doi.org/10.1016/S1057-5219(03)00066-8
Clark PK (1973) A subordinated stochastic process model with finite variance for speculative
prices. Econometrica 41(1):135–155. https://doi.org/10.2307/1913889
Copeland TE (1976) A model of asset trading under the assumption of sequential information
arrival. J Financ 31(4):1149–1167. https://doi.org/10.1111/j.1540-6261.1976.tb01966.x
Copeland TE, Friedman D (1987) The effect of sequential information arrival on asset prices: an
experimental study. J Financ 42(3):763–797. https://doi.org/10.1111/j.1540-6261.1987.
tb04585.x
Doman M (2008) Zależności pomiędzy zmiennością, wolumenem i czasem trwania ceny na
Giełdzie Papierów Wartościowych w Warszawie. Studia i Prace Wydziału Nauk
Ekonomicznych i Zarządzania 9:185–199
Ezzat H, Kirkulak-Uludag B (2014) Information arrival and volatility: evidence from the Saudi
Arabia Stock Exchange (Tadawul). MPRA Paper No. 61160
Fiszeder P, Perczak G (2016) Low and high prices can improve volatility forecasts during periods of
turmoil. Int J Forecast 32(2):398–410
Gallo G, Pacini B (2000) The effects of trading activity on market volatility. Eur J Financ
6:163–175. https://doi.org/10.1080/13518470050020824
Gorgul H, Wójtowicz T (2006) Długookresowe własności wolumenu obrotów i zmienności cen
akcji na przykładzie spółek z indeksu DJIA. Badania operacyjne i decyzje 3–4:29–56
Karpoff JM (1987) The relation between price changes and trading volume: a survey. J Financ
Quant Anal 22(1):109–126
Koulakiotis A, Dasilas A, Molyneux P (2007) Does trading volume influence GARCH effects? –
Some evidence from the Greek market with special reference to banking sector. Investment
Manag Financ Innov 4(3):33–38
Lamourex CG, Lastrapes WD (1990a) Heteroskedasticity in stock return data: volume versus
GARCH effects. J Financ 45(1):221–229. https://doi.org/10.2307/2328817
Lamourex CG, Lastrapes WD (1990b) Persistence in variance, structural change, and the GARCH
model. J Bus Econ Stud 8(2):225–234. https://doi.org/10.2307/1391985
Louhichi W (2011) What drives the volume-volatility relationship on Euronext Paris? Int Rev
Financ Anal 20(4):200–206
Nelson DB, Cao CQ (1992) Inequality Constraints in the Univariate GARCH Model. J Bus Econ
Stat 10(2):229–235. https://doi.org/10.1080/07350015.1992.10509902
Olbryś J (2018) Testing stability of correlations between liquidity proxies derived from intraday
data on the Warsaw Stock Exchange. In: Jajuga K, Locarek-Junge H, Orłowski LT, Staehr K
(eds) Contemporary trends and challenges in finance, Proceedings from the 4th Wroclaw
International Conference in Finance. Springer, Cham, pp 67–79
The Relationship Between Trading Volume and Returns Volatility on Warsaw. . . 55

Pati PC, Rajib P (2010) Volatility persistence and trading volume in the emerging futures market. J
Risk Financ 11(3):296–309
Shen D, Li X, Zhang W (2018) Baidu news information flow and return volatility: evidence for the
Sequential Information Arrival Hypothesis. Econ Model 69:127–133. https://doi.org/10.1016/j.
econmod.2017.09.012
Shleifer A, Summers LH (1990) The noise trader approach to finance. J Econ Perspect 4(2):19–33.
https://doi.org/10.1257/jep.4.2.19
Tan PP, DUA G, Ting SS (2015) Modelling price movement in trading volume-volatility relations.
Malays J Econ Stud 52(2):135–156
Tseng TC, Lee CC, Chen MP (2015) Volatility forecast of country ETF: the sequential information
arrival hypothesis. Econ Model 47:228–234. https://doi.org/10.1016/j.econmod.2015.02.031
Wang P, Zhang W, Li X, Shen D (2019) Trading volume and return volatility of Bitcoin market:
evidence for the sequential information arrival hypothesis. J Econ Interac Coord 14:377–418.
https://doi.org/10.1007/s11403-019-00250-9
Wójtowicz T (2008) Wpływ wielkości obrotów na ocenę warunkowej wariancji stóp zwrotu akcji
na GPW w Warszawie. Studia i Prace Wydziału Nauk Ekonomicznych i Zarządzania
10:685–696
Factors Influencing Individual Investor
Participation in Stock Market

Dorika Jeremiah Mwamtambulo

1 Introduction

In 1969 an outstanding article related to an optimal selection of a lifetime portfolio

was developed by Robert C. Merton. In the study, Merton (1969) examined how an
individual investor should allocate his/her wealthy between risky and non-risky
assets. In determining the optimal mix between risky and non-risky assets, he
observed that both poor and rich individuals with a Constant Relative Risk Aversion
(CRRA) would allocate equal weights on both risky and non-risky assets. In the case
of a Decreasing Relative Risk Aversion (DRRA) utility; rich individuals assigned
higher weights on risky assets in their portfolio as compared to poor individuals.
From the findings, Merton (1969) argued that to maximise their utility through the
optimal portfolio holding, both poor and rich individuals must hold some risky assets
in their portfolio. Merton (1969) conclusion is on the need of individuals to partic-
ipate in investing in some risky assets such as shares and bonds, which are signif-
icantly traded in capital markets.
In 1995 the findings of Merton (1969) were questioned on what is termed as the
stock market participation puzzle. This is a puzzle observed by Haliassos and Bertaut
(1995) in which despite the requirements of the Modern Portfolio Theories and the
findings by Merton (1969), individual households are still reluctant in investing in
risky assets. Very few individual households are observed to be willing to participate
in investing in capital markets. The findings of Hunnicutt (2017) supported these
observations, where only 17.5% of the global equity market is observed to be owned

D. J. Mwamtambulo (*)
Wroclaw University of Economics and Business, Wroclaw, Poland
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 57

K. Jajuga et al. (eds.), Contemporary Trends and Challenges in Finance, Springer
Proceedings in Business and Economics,
https://doi.org/10.1007/978-3-030-73667-5_4
58 D. J. Mwamtambulo

by index investors,1 25.6% by institutional investors2 and the remaining 57% is

owned by the government, pension funds, insurance and companies. According to
Carlson (2018), only 19% of individual households in the United States of America
(USA) are directly participating in the stock in the year 2018. In the UK, by 2018,3
only 13.5% of individual investors owned equity type of investments (Statistics UK
2020). According to Hankur (2016), only 14% of Germans are participating in
capital markets in 2016 despite savings tendencies in the country. Giannetti and
Koskinen (2010) observed similar findings, where less than 50% of direct partici-
pation by individual households is observed in more than 20 countries around the
world.
Many studies have been conducted on trying to find the reasons for low individ-
uals’ participation and hence solving the stock market participation puzzle. Factors
of financial literacy, Intelligence Quotient (IQ), awareness and numeracy have been
observed to be significant on influencing the participation decision (Grinblatt et al.
2011; van Rooij et al. 2011; Sivaramakrishnan et al. 2017). Factors of high transac-
tion costs which included information, entry and recurrent costs are also observed to
be associated in keeping away individual households from investing in stock
markets (Bogan 2008; Leung 2013). Other factors include that of trust, social,
community influence, income and financial wealth (Brown et al. 2004, 2008;
Balloch et al. 2015). Demographic factors such as age, gender, marital status,
education and behavioural factors of overconfidence, loss and regret aversion are
also observed to be significant in some studies.
Tanzania like many other countries around the world is facing a problem of low
participation of individual household in its capital market and in this case the Dar es
Salaam Stock Exchange (DSE) market (Massele et al. 2013; Viswanadham et al.
2014; King and Milanzi 2015; Komba 2016). The low participation by individual
households has been a persistent problem since its start of operations in April of
1998. The low participation in this market does not show any sign to change in the
near future. Due to the persistent nature of the problem, there is a constant need for
determining the factors behind the reasons for low individual households’ partici-
pation in the stock market. Literature has provided an account of factors such as age,
gender, marital status, education, financial literacy, income or wealth, transaction
costs and behavioural factors such as overconfidence, loss and regret aversion to be
significant in influencing the decision on whether to participate in the capital market.
There is a significant need for testing these factors in Tanzania capital market to
observe their applicability in accounting for low participation by individual house-
holds in Tanzania. The findings of the study are necessary for trying to answer the
participation puzzle in Tanzania, where there is persistent low participation by
individual households. Moreover, the findings are essential for setting up new

1
This includes mutual funds, exchange-traded funds, institutional accounts and private investors
tracking on the index.
2
This includes the actively managed hedge funds, mutual funds and institutional accounts.
3
This is an increase from 12.6 reported by Statistics UK (2017).
Factors Influencing Individual Investor Participation in Stock Market 59

policies that are necessary to increase the level of individual households’ participa-
tion in the stock market. The outcome of solving the problem of the participation
puzzle is an increase in the number of individual households in the capital market.
The increase in the number of individual households in the market are expected to
bring into the economy a significant amount of savings which can be used for
production and investment purpose.

2 Literature Review

A number of factors have been accounted for the low individual investors’ partic-
ipation in the capital market. In the study of Aroni et al. (2014b) using a sample of
311 individual investors out of 836,250 prevailing during 2013 in Nairobi stock
market exchange observed that the amount of dividend pay-out by a company attract
the individual investors to invest in shares. In the absence of dividend pay-out, fewer
individuals are observed to participate in share investments. The findings of
Nyamute and Maina (2011), Amisi (2012), Olima (2013) and Mwangi (2015)
observed that financial literacy is a driving factor for individual investors to choose
to invest in the capital market. In the study of Nyamute and Maina (2011), individual
investors with financial training have better financial management on savings,
expenditure, debt management, money, retirement and contingency practices.
Olima (2013) observed a significant relationship between individuals with a high
level of financial literacy with the decision to save and social security or retirement
plans. In the study of Amisi (2012) pension fund managers with a high level of
financial knowledge are observed to take a significant consideration on factors such
as return, risk, portfolio composition, inflation, past performance. Other factors
include the market index tracking, market, liquidity, operation, regulatory, strategy
and counterparty risk when making the investment decision. Mwangi (2015) noted
that although individuals with the high financial literacy are more likely to partici-
pate in the capital market but they are not limited only to invest in shares they also
choose to invest in real assets. The choice on whether to invest in shares or real assets
depends on individual investors dependency factors on expected dividends, capital
appreciation and factors such as the price of the shares, the volatility of the market,
gender, age and income.
Aroni et al. (2014a) observed the availability of financial information do influence
the decision to invest in risky assets. Individual investors who can acquire all the
necessary financial information are more willing to participate in the capital market
as compared to those who are lacking access to financial information. Kimani (2011)
on the other hand it is observed that behaviour factors of herding, loss aversion,
regret aversion, mental accounting, overconfidence, gambler’s fallacy, anchoring
and availability bias influence the decision to participate in the capital market by
individual investors. Market factors such as price changes, market information and
past trends are also observed to influence the participation of the individual investors
in risky assets. In the study of Ndiege (2012) economic factors of expected
60 D. J. Mwamtambulo

dividends, capital appreciation and price of a share and behavioural factors of

herding, advocacy and overconfidence are observed to be significant in influencing
the decision as to whether individual investors in Kenya should participate in the
capital market. Murungi (2011), on the other hand, observed that the factor of market
awareness and trust in the stock market significantly influenced the individual
Ugandan investors’ decision to participate in the capital market. Sindambiwe
(2015) observed factors of financial literacy and market awareness to be significant
in influencing the decision to participate in Rwanda stock market. Increase in the
individuals’ saving behaviours in countries such as Botswana Kenya, Namibia,
Swaziland, Zimbabwe, Ivory Coast, Malawi, Nigeria, Tanzania, Egypt, Morocco,
Uganda, Tunisia, Ghana, Mauritius, South Africa and Zambia are observed by
Andrianaivo and Yartey (2010) to increase the individuals participation in capital
market. The increase in the participation by individual investors through saving are
also attributed to the growth of the capital markets in these countries.
The USA holds the top two largest stock markets in the world. These are the
New York Stock Exchange and the NASDAQ. Given this reputation, it can be
expected that a large number of individual households are participating in the capital
market. Nevertheless, the USA also accounted for a small proportion of individual
households’ participation in its capital markets. According to Carlson (2018), 30%
of individual investors were directly participating in the USA capital market in 1999,
and by 2018, the account was only 19%. The Federal Reserve Survey of Consumer
Finance and Pisani (2017) accounted for a total of 62% of individual households
who were, directly and indirectly, participating in the capital market before the
financial crisis, and this fell to only 54% by 2017. A number of studies have
accounted for low individual participation in the US stock market. This includes
the studies of Brown et al. (2004, 2008), Hong et al. (2004), Shum and Faig (2006),
Bogan (2008, 2013), Yoong (2011), Vestman (2012), Leung (2013), Balloch et al.
(2015) Giannetti and Wang (2016), Bilias et al. (2017), and Chien and
Morris (2017).
According to Brown et al. (2004), the influence of the surrounding communities
and the presence of listed firms both local and public traded influences the individual
households’ participation in the stock market. Suppose the individual household
surrounding neighbourhood (within 50 miles radius) included a number of individ-
uals participating in the capital market. In that case, this increases the likelihood of
that individual household to participate in the capital market. An increase in partic-
ipation of neighbour households by only 2% increases individual investors’ partic-
ipation by 10% when there is no control of income. In the case of control of income,
this will lead to an increase of 2%. Individual households are more motivated to
participate in the market provided that their neighbours or surrounding communities
do participate in the capital market. They choose to do more so when these
household participating in the capital have similar income or wealth.
Similarly, Brown et al. (2004) observed that whenever there is a presence of at
least one local firm that is traded in the stock market in the vicinity of 50 miles range
increases the household participation by 1%. They do more so when a large traded
company is within their community. In the light of Brown et al. (2004) findings,
Factors Influencing Individual Investor Participation in Stock Market 61

individual households that are less surrounded by individuals participating in the

stock market will lead to low participation in the capital market. The results are
supported by Brown et al. (2008), who observed an increase of 10% in the average
ownership of the community increase the individual participation stock market by
4%. The influence of communal interactions is also observed by Hong et al. (2004)
where individuals who have a good relationship with their friends and relative are
observed to participate in investing in capital markets significantly.
A similar finding is observed by Chien and Morris (2017). They observed that the
participation in the USA capital market to be affected by the geographical location,
i.e. State an individual is located. In their study, they examined different households’
participation in the stock market in different states. They observed that the difference
in income between the states leads to a difference in participation in the stock
market. Those states with higher income are more willing to participate in the
stock market than those with lower income. States with higher average income
such as New York, Connecticut, are observed to have a large number of individuals
participating in the capital market compared to state such as Mississippi. When they
controlled for income and examined the individuals in the two-state with lower and
higher average income, still those state with higher income had more individuals
participating in the capital market. Although Chien and Morris (2017) did not
investigate the factors behind this, one speculation can relate to the community
and local firm influence. In cities like New York and California, many individuals
participate in capital markets because a number of largest traded companies are
registered in these areas. Both the community and local firm effect can account for
the higher individuals’ participation if control for income. In the study Balloch et al.
(2015) they argue that communal effects are related to the level of financial literacy
individual investors. Those with high financial knowledge are more likely to be
influenced by advice from other members of the community. However, Balloch et al.
(2015) emphasised that the factor of financial literacy and trust rather than commu-
nity effect do influence the decision to participate in capital markets.
Using the USA Consumer Finance Survey for the years’ 1992, 1995, 1998 and
2001 Shum and Faig (2006) observed that the selection of investing in stock is
related to the factors of wealth, age, retirement savings and financial advice. Indi-
vidual investors with a significant amount of wealth are willing to participate in the
capital market as compared to those without wealth. This notion is confirmed by the
findings of Wolff (2017) where 83% of the stock ownership is owned by only 10%
of the most affluent families while 93% of the stock ownership is owned by 20% of
the most affluent USA households. Age is observed to have a concave relationship
with the investment decision to invest in shares. The concave relationship between
age and participation in the stock market is also observed by Gardini and Magi
(2007) in Italy; Fujiki et al. (2012) in Japan; Thomas and Spataro (2015) in Europe;
and Fagereng et al. (2017) in Norway. Individual managing the retirement savings
are more likely to invest in stock compared to those without savings. Individual
receiving professional investments advice from brokers, financial adviser shows
significant willingness to invest in the stock market.
62 D. J. Mwamtambulo

The study of Bogan (2008) showed that when the individual investors are relieved
from the cost burden of the information and transaction costs through the availability
of internet are more likely to choose to invest in the capital market. This is observed
not to be same as those individual investors with little or no access to the internet.
Bogan (2008) argued that the cost burden that individual incurred for acquiring
information and entry cost for participating in the capital market discourage many
individual investors from participating in investing in risky assets. Leung (2013) also
observed that the probability of senior generation using the internet in the USA to
participate in the capital market has increased by 21.8%. This increase to 38.7%
when advised by professional investors regarding the investment process. The
findings of both Bogan (2008) and Leung (2013) showed that the information cost
involving the process of learning the investment activities and obtaining the market
information is managed or kept at minimal. Individual investors show more will-
ingness to participate in capital markets. Bogan (2013), on the other hand, examined
the influence of gender of the dependants to the participation in the stock market. He
observed that family with female children are more likely to participate in the stock
market as compared to parents with more male children. It is only single female
parents with male children who are more willing to invest in the stock market as
compared to any other single-parent family.
Yoong (2011), on the other hand, examined how financial literacy influenced
participation in the capital market. Yoong (2011) observed that financial illiteracy
individual investors are less willing to participate in investing in risky assets for fear
of their lack of knowledge regarding the management of the assets. Vestman (2012)
showed that house ownership is related to an increase in participation in the stock
markets. His findings showed that house owners are twice (61%) more likely to
invest in the stock market as compared to renters (26%). Vestman (2012) argued that
the increase in participation in the stock market is a result of an increase in wealth
rather than house ownership. The wealth is accumulated from the rental savings that
individual households are no longer needed to pay. However, in the study of Bilias
et al. (2017), individual investors participating in the capital market are observed not
to accumulate a significant amount of wealth from their investment. In their study
Bilias et al. (2017) observed that wealth inequalities between households did not
change by participating in the capital markets. Giannetti and Wang (2016) examine
the influence of corporate scandals on individual participation in stock markets. In
the presence of corporate scandals such as Enron and WorldCom, individual inves-
tors are observed to decrease their level of participation in the capital market. The
effect did not have only a negative impact on those companies caught up in the
scandals; instead, it had a ripple effect on other companies that are not directly
involved with scandals. This can be seen in the case of Arthur Andersen auditing
firm that was related to the scandals of Enron and WorldCom, many traded compa-
nies audited by the firm experienced a drop in the participation of the individuals
owning their shares.
The findings of Xia et al. (2014) in China showed that the individuals participat-
ing in the Chinese market are more overconfident regarding the level of skills and
knowledge regarding investment. Their overconfident behaviour resulted in making
Factors Influencing Individual Investor Participation in Stock Market 63

poor investment decisions. According to the findings, the factor of financial literacy
had a significant role in increasing participation in stock markets. Individual inves-
tors who are financially informed are more likely to invest in a risky asset than
individuals with little or no skills regarding the capital markets.
Fujiki et al. (2012) examined the influence of factors of age, income, financial
assets and education on the participation Japanese in the capital market. Using
survey data for the period between 2007 and 2010, they observed factor age to
weakly influence the participation in the capital market after controlling for factors of
income, education and financial assets. In Japan, the older generation is observed to
participate more in the capital market than the young generation. However, when
controlling for the factor of income and financial assets, young Japanese are
observed to participate more in the risky market as compared to the old generation.
With an increase in income, the concavity relationship between age and participation
in the capital market is observed in the Japanese market.
Sivaramakrishnan et al. (2017) and Vohra and Kaur (2016) examined the factors
influencing individual participation in India’s capital market. In the study of
Sivaramakrishnan et al. (2017) they observed that the attitude of individual investors
towards risk, financial well-being, regulatory perception, social influence and invest-
ment hassles significantly influence the individual investors’ decision to participate
in the capital market. The level of financial literacy is also observed to be the factor
behind the decision where individual investors with a high level of financial literacy
are more likely to invest in the risky asset. In the study of Vohra and Kaur (2016)
observed that factor of capital market awareness affects the level of participation
between the group of women investors and non-investors. Women with a high level
of market awareness choose to participate in the capital market than those lacking the
general awareness of the market.
Hunkar (2016) argues that the increase in participation in the Germany stock
market is a result of the lowering of the interest rate by the European Central Bank to
nearly zero rates. The opportunity cost of individual investors of keeping their
investment in fixed deposit is higher than investing in share markets. Using data
from fifteen (15)4 European countries, Hagman (2015) examined the influence of the
level of trust on individual investors’ participation in the stock market. Two types of
trust are analysed, the trust with other individuals and the trust with institutions. The
findings showed that the trust towards other individuals increases individual partic-
ipation in the capital market. Only the trust of the government institutions led to an
increase in individual investors’ participation in capital markets. The influence of
individual investors trust is also examined by Guiso et al. (2008) and they observed
that less trusting individuals in both countries Italy and Netherlands are less likely to
choose to participate in the capital market as compared to trusting individuals. The
lack of trust in the stock market increased when the individual investors are observed

4
The data was collected from a Survey of Health, Ageing and Retirement in Europe (SHARE) and it
included the following countries Austria, Belgium, Czech, Denmark, Estonia, France, Germany,
Italy, Israel, Luxembourg, Netherlands, Slovenia, Spain, Sweden and Switzerland.
64 D. J. Mwamtambulo

to have little knowledge or lack of familiarity regarding the operations of the capital
markets. In the study of Georgarakos and Pasini (2011), the factors of trust and
sociability are observed to influence the individual investors’ participation in the
capital market. Using data from countries with high individual participation such as
Sweden, Denmark and Switzerland and low participation such as Austria, Spain and
Italy they observed that both individuals’ trust and sociability increase their partic-
ipation in the capital market. Social effects are observed to restore the individual
investors’ trust in participating in capital markets.
When examining the influence of financial literacy in the participations of Dutch
in stock market Rooij, Lusardi and Alessie (van Rooij et al. 2011) observed that
though many Dutch households showed to possessed basic financial knowledge,
they lacked in terms of advanced financial knowledge. They are not able to show the
difference between stocks and bonds. The analysis of results showed that the level of
capital market participation is related to the level of financial literacy. Individual
possessing high levels of financial knowledge are more likely to choose to invest in
the capital market. Similar results are obtained by Guiso and Jappeli (2005) in Italy,
Almenberg and Widmark (2012) in Sweden, Georgarakos and Inderst (2012) in 15
European countries, Brown and Graf (2013) in Switzerland, Almenberg and Dreber
(2015) in Sweden. Guiso and Jappeli (2005), on the other hand, observed that lack of
awareness regarding investment such as shares, bonds, mutual funds lead to less
participation in capital markets. They also observed awareness regarding the capital
market’s investments increases with factors of education, house resources, social
interactions and long-term relationship with the bank. Almenberg and Widmark
(2012) examined both the influence of numeracy and financial literacy on the
participation of Swedish households’ in the capital market or house ownership.
They observed that high numeracy ability is associated with both a high level of
participation in the capital markets and house ownership. Financial literacy is only
associated with house ownership. Brown and Graf (2013) observed that out of 1500
individual respondents half of them could answer the question regarding basic
financial knowledge while individuals having a young family, low income and
immigrant had trouble answering the questions. The failure regarding their financial
knowledge is related to the lack of their participation in the capital market. In the
study, Almenberg and Dreber (2015) observed that when controlling for basic
financial knowledge, both male and female are more willing to participate in
investing in the capital market. Georgarakos and Inderst (2012) on the other hand
examined the influence of financial advice to participants in the stock market in
Finland, Sweden, UK, Ireland, Denmark, Germany, Netherlands, Belgium, Luxem-
burg, France, Austria, Italy, Spain, Portugal, and Greece. Individuals who have
received financial advice are more likely to participate in the capital market since
the advice is related to investment that led to the maximisation of their expected
utility.
Kaustia and Torstila (2011) examined the influence of political affiliation on
participation in capital markets. Looking at whether individual investors are either a
right- or left-wing voter or politician, they examined their likelihood to invest in
share markets. They observed that when controlling for income, wealth and educated
Factors Influencing Individual Investor Participation in Stock Market 65

left wings, voters and politicians show less interest in participating in the share
markets. The decline is between 5–6% by an increase in 1 scale left and 17–20%
when a scale is increased by 3. While Gardini and Magi (2007) examined the factors
influencing participation in the stock market in Italy, they observed that difference in
both financial and real wealth leads to a difference in participation. Those with
higher financial and real wealth are more likely to participate in the stock market as
compared to those will low wealth. Increase in wealth is also observed to play a
significant role in influencing individual participation in the Swedish market.
According to Briggs et al. (2015), increase in Swedish household wealth is associ-
ated with an increase in the level of participation in the capital market. Increase in
wealth allowed the individual household to cover for both entry and periodic costs
associated with investing in the capital market. The participation level is also
observed by Gardini and Magi (2007) to be influenced by the concavity of the factor
of age. The concavity relates to the likelihood of young people invest in shares as
compared to older people. Meaning the participation in stock market increase with
the factor of age at a decreasing margin. In the findings of Vestman (2012) Swedish
individuals owning a house are more likely to choose to invest in the capital market
that those renting a house. A same observation is observed in the USA market.
In the findings of Thomas and Spataro (2015) in Austria, Belgium, Denmark,
Germany, Italy, France, Switzerland, Sweden and Netherlands observed factors of
age, financial literacy, education, number of children, marital status, income and
house ownership to influence the participation in the capital market. Investment in
the capital market is observed to increase positively with the factor of age. As
individuals get older, they choose to participate in investing in capital markets.
They mostly do so in the early younger age as compared to old age accounting for
the concavity relationship between participation decision and age. Increase in both
financial literacy and education are observed to influence the participation in capital
market positively. Financial literacy relates to having either basic or advanced
financial knowledge, education look on if the individual has a basic knowledge of
investment and decision making. Individual investors having both a high level of
financial literacy and education are more likely to choose to participate in the capital
market. Married individuals are observed to choose to participate in the capital
market, but they choose not to do so when they have a large size number of children
or dependants. Increase in income allows individual investors to save leading for
investment. House ownership increases the wealth of individual investors through
rental savings. Both the increase in income and house ownership is observed to
increase the level of participation in capital markets. Fagereng et al. (2017) examined
the influence of factors of age and education in Norway for the years between 1995
and 2009. They observed that a high level of education leads to the participation of
the capital market at the early stage of the life cycle as compared to the later stages.
This confirms the concavity influence of the factor of age, which is also associated
with the level of education. These findings are confirmed by Thomas and Spataro
(2015) with 2010 data.
Grinblatt et al. (2011) were interested in the idea of the influence of the Intelli-
gence Quotient (IQ) on the decision to participate in the capital market of the Finland
66 D. J. Mwamtambulo

residents. Using trading data between 1995 and 2002, they observed individual with
high IQ are more likely to participate in the capital market after controlling for
factors of age, occupation, income and factor of wealth. The individual with IQ when
they participate in capital markets or mutual funds are observed to be facing a
significantly low amount of risks and having a very high Sharpe ratio. On the
other hand, Andersen and Nielsen (2011) examined the influence of wealth that
has been suddenly acquired after death into individuals’ participation in capital
markets. The study aimed to check if individual investors are discouraged from
investing in risky assets due to fixed costs which can be afforded by the newfound of
wealth. They observed that the level of participation in the capital market increases
with an increase in a windfall type of wealth. However, not all Danish who suddenly
inherited the wealth chooses to participate in the capital market. The majority did not
choose this option, and many who had inherited any share investment actively sell
the whole investment rather than holding it. In the findings of Hurd et al. (2011) in
Netherlands factor of return was an influence toward investing in the capital market
individual was observed to be more pessimistic regarding the idea of a return from
capital markets.
The analysis of the literature has accounted for the factors of risk, expected return,
demographic, behavioural, economical, social and organisation factor to positively
influence the stock market participation by an individual investor. Demographics
factors include factors of age, gender, marital status, family size, level of income and
education while behavioural factors include factors of overconfidence, anchoring,
gambler’s fallacy, availability and representative bias, mental accounting, herding
behaviour, regret and loss aversion. Economical factors including the amount of
accounting information disclosed by a company and its past and future performance.
This also includes the amount of dividend and bonused declared by the company. On
the other hand, social and organisational factors included the firm image and the
influence of friends, relative and brokers on individual decision. This study aims at
examining the influence of these factors on the participation of individual investors
in Tanzania. The interest is on examining if these factors have the same positive
influence as observed by the cited literature in this study.

3 Research Methodology

The population of interest to this study includes both individual households partic-
ipating and not participating in the DSE. The individual households participating in
the stock market consist of the involved in investing directly into share and bonds.5
Non-participants include the population of individual households participating in
investment other than bonds and share and those not investing at all. The study
intends to use two population groups, the participators and non-participators in the

5
Bond investment include both investment in government and corporate bonds.
Factors Influencing Individual Investor Participation in Stock Market 67

DSE. The population of interest is generally the entire population of individual

households of Tanzania. As of 2019, the Tanzania population is estimated at around
58.1 million people. As it is difficult to collect data from the whole population, a
sample is drawn to present the population. As the sample needs to be a representation
of the population, there is a need for a sample to constituents all the characteristics of
the population. This requires the sample size to be large. As the population of interest
is above 100,000 individuals, using the Lincoln University (2006) formula for
determining the required sample size for the level of precision of
3% and confi-
dence interval of 95% and a P-value of 0.5 the sample size can be calculated from the
formula:
 
zs 2
n>
e

Where n is the sample size, z ¼ 1.95, s ¼ 0.5 and e ¼
3% the required

representative sample is over one thousand one hundred individuals (1100). In this
study, a total of 1600 individuals’ household are used in the data collection. A
stratified6 random sampling method is used in selecting the individual households in
the sample. The stratified random sampling method allowed each household in the
population to have an equal chance of being selected into the sample. This helps in
ensuring that the data collected is independent and identically distributed (i.i.d), a
characteristic that is necessary for modelling the results.

3.1 Data

Primary data is used for the analysis of factors determining the individual house-
hold’s decision participation in the DSE. Primary data, rather than secondary data, is
used due to the nature of the study aimed to be conducted. In the case where sec-
ondary data is used, it will only be limited to the individual households already
participating in DSE. This will lead to failure of capturing in the model the factors
influencing those individual households’ not participating in the DSE. Moreover,
primary rather than secondary data is collected due to information nature needed to
be collected from individual households. Information such as demographic factors
can be obtained from secondary data such as the Household Budgetary Survey
(HBS), which is conducted in the country. The problem is on other factors such as
behavioural factors, risk and returns attitudes as they are related to the individual
attributes. The personality traits are hidden characteristics. In many cases, this kind
of information attributable to personality traits are not readily available; thus, the
need to shift from secondary to primary data. There is also a need to generate a

6
Two strata are formed for the population of individuals participating and not participating in the
capital market.
68 D. J. Mwamtambulo

measure that will allow individuals to reveal their personalities through their
response to a number of questions.
An interview-based questionnaire method of data collection is used on the
collection of the data. Questionnaire rather than observation or interview methods
are used because of its ability to cover a large group of individuals in a limited
amount of time. It gets rid of the Hawthorne effect, which significantly occurs when
an observation type of data collection method is in use. Moreover, the nature of the
question, which involves the use of a number of Likert items that identify a specific
Likert scale encouraged the use of the questionnaire method. On the other hand, to
ensure a uniformly understanding of the requirements in the questionnaire, an
interview-type rather than the self-administering method is used. An interview-
type method of data collection is also used to increase the response rate a problem
commonly occurring when a self-filling type of questionnaire is used.
In ensuring that the questionnaire measure the required attributes intend for the
study, a Cronbach’s Alpha test measure of reliability is used. A Cronbach Alpha test
of greater than 0.7 is acceptable. A pilot study was conducted before the general
survey to initially test the reliability of the questionnaire as an instrument for data
collection. A sample of four hundred (400) individuals was studied under the pilot
study which is above the required number of three hundred (300) needed to run a
Principal Component Analysis (Comrey 1973). After the Pilot study variables that
were observed to be less significant were dropped and only those who were
considered to be significant constituent the final questionnaire.
The principal component analysis is used to group the Likert items into their
respective Likert scale as argued by Clason and Dormody (1984), Hodge and
Gillespie (2003), and Sullivan and Artino (2013). The component extracted from
the PCA analysis will form the Likert scale, which is treated as a continuous variable.
In this study, only those components extracted from PCA with communalities of
greater than 0.5 are accepted. Moreover, Kaiser-Meyer-Olkin (KMO) measure of
sample adequacy of greater than 0.5 and Bartlett’s test of sphericity of less than 5%
are considered.
Logit regression is used in modelling the decision on whether or not to participate
in DSE. Logit regression than the Ordinary Least Square (OLS) is being used as the
study is dealing with the limited dependent variable. The dependent variable the
study needs to model that is the decision on whether to participate in the stock
market take only two forms, 1 if an individual decided to participate in the stock
market and 0 if otherwise. The maximum likelihood method is used to estimate the
parameters in the models. The diagnostic test, including the goodness of fit, hypoth-
eses testing, also performed.

3.2 Findings and Discussion

Table 1 below present the findings for the logistic regression on the participation
decision. Demographic factor such as age, gender, marital status education, income
Factors Influencing Individual Investor Participation in Stock Market 69

Table 1 Logistic regression results on the participation decision

Participation decision
Coefficient Coefficient
Gender 0.0604 Anchoringa (0.1290)
Age (0.0308) Representative bias 0.1119
Age square 0.0006 Gambler’s fallacyb 0.0305
Marital status 0.4502 Pattern gamblerc (0.0038)
Number of children (0.0411) Availability bias (0.0637)
Number of children in school 0.1893** Loss aversion (0.0016)
Household size (0.0252) Regret aversion (0.3631) **
Household income earners (0.0292) Mental accountingd (0.4852) ***
House ownership (0.2241) Market factor 0.3439**
Occupation (0.2658) Herding behaviour (0.1583)
Working experience 0.0028 Past performance (0.2204)
Education 0.0278 Accounting information (0.4280) ***
Income 0.8660*** Expected earning 0.1854
Return 0.1434 Dividend/bonus pay-out 0.4060**
Risk (0.2651) ** Advocacy 0.2708**
Risk aversion 0.2095 Image (0.2552) *
Overconfidencee 0.3100** Constant (14.6437) ***
Pseudo R_squared 0.2987
Model prediction 90.68%
Source: Logit regression results. Note ***, **, * are 1%, 5% and 10% significance levels
respectively
a
The behaviour of relying heavily on single rather than multiple attributes of the investment
b
This is the anticipation by investors that price patterns are more predictable than they are really are
c
The behaviour of depending on the price pattern in making investment decision
d
The behaviour of human of holding into a particular information irrespective of changes taking
place
e
The tendence of overestimating one’s ability, attributing gains to one’s skills and failure to bad luck

and family size are included in the analysis. Factors measuring the level of depen-
dency per household, including the number of children per household and those who
are in school are also included. The remaining factor measures the attitude of
investors toward risk, return, available information and the influence of behavioural
factors. It should be noted that during the model specification stage, the interaction
factors were added to observe their increase in the explanation power of the model.
This was rejected by both the Likelihood ratio and Wald tests; thus, they were
dropped from the model. The results below are obtained from the model, which
significantly explained the data in the analysis.
The likelihood of an individual investor to participate in DSE is observed to
increase with the demographic factor of a number of children in school and income.
An increase in the number of children per household going to school increase the
need for an extra source of income to cover for the education expenses. Individual
are observed to more likely to choose to invest in the DSE to obtain the extra income
to support the household. Increase in the household income increase the level of
70 D. J. Mwamtambulo

participation in DSE. Individual earning high income are more likely to save and
opting to invest. Increase in income allows the individual household to explore
different types of investments available, leading to their participation in DSE.
Risk aversion behaviour decreases the likelihood of individuals to choose to
participate in the DSE. This is with the implication that many individual investors
choosing to participate in the capital market are more risk-takers than an average
investor. Individual choosing to participate in DSE are more overconfident with their
level of skills. They are more likely to attribute their success with their skills and
failure as bad luck. On the other hand, these skills allow them to make better
investment decisions. This includes selling non-performing assets irrespective of
their past performance while continue holding performing assets. This is observed
by a significant negative relationship between the likelihood of participating in the
DSE and factors of regret aversion and mental account. Although individual inves-
tors participating in DSE show significant decision-making skills, they are observed
to be significantly affected by the noises from the market. This is observed by a
significant positive influence of the market factor on the decision to participate
in DSE.
Individual investors choosing to participate in the DSE are less likely to be
dependent on the accounting information of companies and institutions trading
investments in the DSE. The results showed that individuals who are significantly
influenced by the amount of information disclosed by the financial statements of the
companies are less likely to participate in the DSE. On the other hand, the amount of
dividend or bonus declared in the market significantly increase the level of partic-
ipation by individual investors. Individual investors are observed to be attracted to
invest in companies paying out a significant amount of dividend per share owned by
the investors. Individual investors decision to participate in DSE is observed to be
significantly influenced by peers, relatives, friends and brokers. Individual investors
depend on the advice given to them by their office peers, relatives, friends and
brokers on whether or not to participate in DSE. This is with the implication that
their decision on whether to participate in DSE is significantly influenced by
how they highly weigh advices given to them by third parties rather than a decision
made solely by the individual investor’s judgment.

4 Conclusion

Since its discovery by Haliassos and Bertaut (1995) participation puzzle has been of
interest to many researchers, especially the during the current situation where a low
number of individual investors’ participation in the capital market are being
observed. The interest of the researcher is on findings the factors behind the low
individual investors’ participation in capital markets and thus solving the puzzle.
Over the years, a number of factors have been identified to influence the participation
decision and tended to vary in different countries. This study aimed at examining the
factors behind the low individual household participation in Tanzania’s capital
market.
Factors Influencing Individual Investor Participation in Stock Market 71

Data collected from a total of 1600 individual is used in logistic regression. It is

observed that the willingness of individual investors to participate in the capital
market increases with the increase in the number of children in school and income.
Individuals with low-risk appetite are less likely to participate in capital markets as
compared to an individual with high risk-taking ability. The high-risk taking behav-
iours of the individual investors in Dar es Salaam stock market conform with typical
features of a frontier equity market Overconfidence individuals are more likely to
participate in the capital markets. Similarly, individuals participating in the capital
market are more likely to sell non-performing assets and keep on holding a
performing asset and are not caught up with past performances. They significantly
influence the noised in the market, although they do not follow the actions of others;
rather, they make their judgments. They are less dependant to the accounting the
information of companies trading their financial product in the markets but are
interested in the dividend or bonuses announcement news. They also consider the
advice from friends, relative or brokers.
In order to increase individual investors’ participation in the stock market income
level should be increased. This is through the central government, raising the
minimum income level in the country and growing entrepreneurship in the country.
Education on optimal decision making should be made available to individual
investors interested to participate in the capital market. This includes different
training organised by the Dar es Salaam Stock Exchange. Education should also
be on the different ways of processing information from friends, relatives and
brokers to ensure utilisation of information which is on the line of maximising
their utility. Companies trading their financial products in the capital markets should
disclose all relevant information, including past and future performances. The
companies should ensure they maintain a dividend growth strategy to attract more
individual investors participation in share markets.
In this study, very few factors from the literature are accounting for the lack of
individual investors’ participation in Tanzania capital market. This study suggests a
further examination of the factors in the Tanzania context influencing the participa-
tion decision. This includes examining the influence of investment infrastructures;
government policies, economic condition, law and regulations pertaining to invest-
ment and their participations on investment activities. The study also suggests the
possibility of examining the study over a long period of time in examining the
altering factors that influence each investment decision during a specific period.

References

Almenberg, J. and Dreber, A. (2015) Gender, stock market participation and financial literacy: Econ
Lett 137:140–142. htpps://doi.org/10.1016/j.econlet.2015.10.009
Almenberg J, Widmark O (2012) Numeracy, financial literacy and participation in asset markets.
SSRN Electron J:1–40. https://doi.org/10.2139/ssrn.1756674
Amisi S (2012) The effect of financial literacy on investment decision making by pension fund
managers in Kenya. The University of Nairobi
72 D. J. Mwamtambulo

Andersen S, Nielsen KM (2011) Participation constraints in the stock market: evidence from
unexpected inheritance due to sudden death. Rev Financ Stud 24(5):1667–1697. https://doi.
org/10.1093/rfs/hhq146
Andrianaivo M, Yartey CA (2010) Understanding the growth of African financial markets. Afr Dev
Rev 22(3):394–418
Aroni J, Gregory N, Sakwa M (2014a) The effect of financial information on investment shares – a
survey of retail investors in Kenya. Int J Bus Soc Sci 3(8):58–69
Aroni J et al (2014b) Influence of dividend payout on investment in shares – a survey of retail
investors in Kenya. Int J Bus Soc Sci 3(8):58–69
Balloch A, Nicolae A, Philip D (2015) Stock market literacy, trust, and participation. Rev Financ 19
(5):1925–1963. https://doi.org/10.1093/rof/rfu040
Bilias Y, Georgarakos D, Haliassos M (2017) Has greater stock market participation increased
wealth inequality in the US? Rev Income Wealth 63(1):169–188. https://doi.org/10.1111/roiw.
12225
Bogan V (2008) Stock market participation and the internet. J Financ Quant Anal 43(1):191–212.
https://doi.org/10.1109/WPMC.2014.7014917
Bogan VL (2013) Household investment decisions and offspring gender: parental accounting. Appl
Econ 45(31):4393–4406. https://doi.org/10.1080/00036846.2013.788782
Briggs J et al (2015) Wealth and stock market participation: estimating the causal effect from
Swedish lotteries *. NBER Working Paper Series, May(3), pp 1–45. http://w4.stern.nyu.edu/
finance/docs/pdfs/Seminars/1501f-cesarini.pdf
Brown M, Graf R (2013) Financial literacy, household investment and household debt: evidence
from Switzerland. Working Papers on Finance No.13/1. St.Gallen
Brown JR et al (2004) The geography of stock market participation: the influence of communities
and local firms. Working Paper 10235. Massachusetts. http://www.nber.org/papers/w10235
Brown JR et al (2008) Neighbors matter: causal community effects and stock market participation. J
Financ 63(3):1509–1531. https://doi.org/10.1111/j.1540-6261.2008.01364.x
Carlson B (2018) Inequality in the stock market. Bloomberg View, 20 February, pp 1–6
Chien BY, Morris P (2017) Household participation in stock market varies widely by state. The
Regional Economist, Third Quar, pp 1–6
Clason DL, Dormody TJ (1984) Analyzing data measured by individual likert-type items. J Agric
Educ 35(4):31–35
Comrey AL (1973) A first course in factor analysis. Academic Press, New York
Fagereng A, Gottlieb C, Guiso L (2017) Asset market participation and portfolio choice over the
life-cycle. J Financ 72(2):705–750. https://doi.org/10.1111/jofi.12484
Fujiki H, Hirakata N, Shioji E (2012) Aging and household stockholdings: evidence from Japanese
household survey data. p 41
Gardini A, Magi A (2007) Stock Market participation: new empirical evidence from Italian
households’ behavior. Giornale degli Economisti e Annali di Economia 66(1):93–114
Georgarakos D, Inderst R (2012) Financial advice and stock market participation. SSRN Electron J
May. https://doi.org/10.2139/ssrn.1641302
Georgarakos D, Pasini G (2011) Trust, sociability, and stock market participation. Rev Financ 15
(4):693–725. https://doi.org/10.1093/rof/rfr028
Giannetti M, Koskinen Y (2010) Investor protection, equity returns, and financial globalization. J
Financ Quant Anal 45(1):135–168. https://doi.org/10.1017/S0022109009990524
Giannetti M, Wang TY (2016) Corporate scandals and household stock market participation. J
Financ 71(6):2591–2636. https://doi.org/10.1111/jofi.12399
Grinblatt M, Keloharju M, Linnainmaa J (2011) IQ and stock market participation. J Financ 66
(6):2121–2164. https://doi.org/10.1111/j.1540-6261.2011.01701.x
Guiso L, Jappelli T (2005) Awareness and stock market participation. Rev Financ 9(4):537–567.
https://doi.org/10.1007/s10679-005-5000-8
Guiso L, Sapienza P, Zingales L (2008) Trusting the stock market. J Financ 63(6):2557–2600.
https://doi.org/10.1111/j.1540-6261.2008.01408.x
Factors Influencing Individual Investor Participation in Stock Market 73

Hagman M (2015) The effects of trust on stock market participation: a cross-sectional study based
on 15 countries. Umea University
Haliassos M, Bertaut CC (1995) Why do so few hold stocks? Econ J 105(September):1110–1129.
https://doi.org/10.2307/2235407
Hodge DR, Gillespie D (2003) Phrase completions: an alternative to likert scales. Soc Work Res 27
(1):45–55. https://doi.org/10.1093/swr/27.1.45
Hong H, Kubik JD, Stein JC (2004) Social, interaction and stock-market participation. 59
(1):137–163
Hunkar D (2016) On the stock market participation of Germans. TopForeignstocks.com. https://
topforeignstocks.com/2016/02/01/on-the-stock-market-participation-of-germans. Accessed
6 Jan 2020
Hunnicutt T (2017) Less than 18% of global stocks owned by index investors. Business Insider
Hurd M, Van Rooij M, Winter J (2011) Stock martert expectations of Dutch households. J Appl
Econ 26:416–436. https://doi.org/10.1002/jae
Kaustia M, Torstila S (2011) Stock market aversion? Political preferences and stock market
participation. J Financ Econ 100:98–112. https://doi.org/10.1016/j.jfineco.2010.10.017
Kimani VW (2011) A survey of behavioural factors influencing individual investors choices of
securities at the Nairobi Securities Exchange. The University of Nairobi
King NAS, Milanzi MC (2015) Factors considered for investment decision making in capital
markets in Tanzania. Int J Econ Bus Rev 3(6):100–104
Komba GV (2016) Essays on investor behaviour and corporate governance in sub-Saharan African
frontier markets. University of Hull
Leung C (2013) Determinants of stock market participation among elderly US households with
internet access. Haverford College
Massele J et al (2013) Challenges faced by Dar-es-Salaam Stock Exchange market in Tanzania. Res
J Financ Account 4(15):36–43
Merton RC (1969) Lifetime portfolio selection under uncertainty: the continuous-time case. Rev
Econ Stat 51(3):247–257. https://doi.org/10.2307/1926560
Murungi LJ (2011) Awareness, trust and stock market efficiency: the case of Uganda securities
market. Makerere University
Mwangi PM (2015) The relationship between financial literacy and stock market participation by
retail investors in Kenya. The University of Nairobi
Ndiege CO (2012) Factors influencing investment decision in equity stocks at the Nairobi securities
exchange among teachers in Kisumu Municipality, Kenya. The University of Nairobi
Nyamute W, Maina JKM (2011) Effect of financial literacy on personal financial management
practices: a case study of employees of finance and banking institutions. In: Africa International
Business and Management (AIBUMA), Nairobi, pp 1–15
Olima B (2013) Effect of financial literacy on personal financial management on Kenya Revenue
Authority employees in Nairobi. The University of Nairobi
Pisani B (2017) Stocks are high, but investor numbers are low. CNBC News, pp 1–5. https://www.
cnbc.com/2017/11/02/stocks-are-high-but-investor-numbers-are-low.html
Shum P, Faig M (2006) What explains household stock holdings? J Bank Financ 30:2579–2597.
https://doi.org/10.1016/j.jbankfin.2005.11.006
Sindambiwe P (2015) Financial literacy, stock market awareness and capital market participation of
an emerging stock market. Int J Multidiscipl Approach Stud 1(5):366–406
Sivaramakrishnan S, Srivastava M, Rastogi A (2017) Attitudinal factors, financial literacy, and
stock market participation. Int J Bank Mark 35(5):819–841. https://doi.org/10.1108/IJBM-01-
2016-0012
Statistics UK, N (2017) Ownership of UK quoted shares: 2016, London
Statistics UK, N (2020) Ownership of UK quoted shares: 2018, London
Sullivan GM, Artino AR (2013) Analyzing and interpreting data from likert-type scales. J Grad
Med Educ December:541–542. https://doi.org/10.4300/JGME-5-4-18
74 D. J. Mwamtambulo

Thomas A, Spataro L (2015) Financial literacy, human capital and stock market participation in
Europe: an empirical exercise under endogenous framework. Dipartimento di Economia e
Management–Università di Pisa Discussion. Discussion Paper n.194. https://doi.org/10.
13140/RG.2.1.4458.7360
University, L (2006) Sample size. Semester 2. Lincoln
van Rooij M et al (2011) Financial literacy and stock market participation. J Financ Econ 101
(2):449–472. https://doi.org/10.1016/j.jfineco.2011.03.006
Vestman R (2012) Limited stock market participation among renters and home owners. SSRN
Electron J. https://doi.org/10.2139/ssrn.2082376
Viswanadham N et al (2014) A study of perceptual factors influencing investors buying behavior in
Tanzanian equity market. J Financ Invest Anal 3(2):99–108
Vohra T, Kaur M (2016) Awareness and stock market participation of women: a comparative study
of stock investors and non-investors. IUP J Manag Res 15(4):22–38
Wolff EN (2017) Household wealth trends in the United States, 1962 to 2016 has the middle-class
wealth recovered? NBER Working Paper Series 24085. Massachusetts
Xia T, Wang Z, Li K (2014) Financial literacy overconfidence and stock market participation. Soc
Indic Res 119(3):1233–1245. https://doi.org/10.1007/s11205-013-0555-9
Yoong J (2011) Financial illiteracy and stock market participation: evidence from the RAND
American life panel. Pension Research Council Working Paper. PRC WP 2010-29. Philadephia.
https://doi.org/10.1093/acprof:oso/9780199696819.003.0005
Model Risk of VaR and ES Using Monte
Carlo: Study on Financial Institutions from
Paris and Frankfurt Stock Exchanges

Aleksandra Helena Pasieczna

1 Introduction

Mathematical modeling has proven to be an efficient tool for contemporary finance

practitioners. It becomes vitally important to judge these mathematical constructs
before they are used in practice. Two intuitive ways exist to judge their validity—
accuracy and precision. The first refers to the performance of the model, which can
be understood as the distance of the obtained model result (prediction) from the ideal
result. By ideal we mean a benchmark, or the actual result obtained from observa-
tion, depending on the type of model. The second corresponds to the reproducibility
or the sensitivity of the model across multiple measurements or observations. A good
model is a model that is both, accurate (not far from a benchmark), and precise (gives
similar results repetitively).
Precision provides an estimate of the consistency of a model to give a reliable
output on multiple simulations with slightly noisy input parameters. Unlike accu-
racy, there is no reference model or benchmark against which this quality can be
judged, and only the repeatability (although under slightly different inputs) can be
checked. In essence, the precision reflects the risk of the model. If the model is the
risk measure, the precision will be the risk of the risk measure, henceforth called
model risk.
In an earlier study (Pasieczna 2019), we presented the accuracy analysis of Value
at Risk (VaR) and Expected Shortfall (ES) models with an application on the Polish
WIG20 and mWIG40 indices. Here, we focus on the precision aspects of the models
on 116 financial institutions (out of 126 after data cleaning) from the Paris and
Frankfurt stock exchanges, while restricting the study to Monte Carlo-based

A. H. Pasieczna (*)
Koźmiński University, Warsaw, Poland
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 75

K. Jajuga et al. (eds.), Contemporary Trends and Challenges in Finance, Springer
Proceedings in Business and Economics,
https://doi.org/10.1007/978-3-030-73667-5_5
76 A. H. Pasieczna

approaches only. VaR, recommended by the Basel II agreement (Basel 2004), is

frequently used to estimate potential losses with a certain confidence, while ES was
recommended by the Basel III agreement (Basel 2010), as it captures tail risk more
efficiently. Both, VaR and ES, are frequently used by practitioners to analyze
market risk.
This paper focuses on measuring the model risk for VaR and ES for different
configurations (confidence intervals). Four models for each configuration are con-
sidered, based on the Monte Carlo technique (Pasieczna 2019), and attempt to
analyze the effect of including inter-institutions covariance structure and the effect
of filtering past data using rolling and exponential-weighted windows. For studying
their model risks, we focused on the changes to the input parameters, which were
made by choosing different historical periods. The work was applied on over
hundred financial institutions traded on the Paris and Frankfurt stock exchanges.

2 Data Collection and Processing

The dataset consisted of daily close prices between 2000 and 2020 year, for
126 financial institutions traded on the Paris and Frankfurt stock exchanges. The
list of companies was obtained by using the stock screener from Investing.com. The
price data was then obtained using the Python package investpy (Investpy 2020).
In the simulations relative returns were used to compute the market risk measures.
By relative returns we mean percentage price change with respect to the previous
day’s price. These returns are unitless and do not depend on the currency in which
the stock might be potentially traded and form the basis for the uncertainty in our
Monte Carlo simulations. However, using relative returns resulted in certain banks
(whose prices did not change for several days) having a small standard deviation,
and consequently—a small market risk, leading to high model risk, when compared
across different historical time periods. To understand this, consider a bank whose
returns are zero for large part of 2019. If the historical period considers data for
1 year, the estimated standard deviation will be small, and consequently the VaR and
ES. If the historical time considers data for 2 years (2018 and 2019) then the
estimated standard deviation will be larger than in the first case. Comparing these
two standard deviations would give a low precision score due to this effect, thus the
model risk is high. Additionally, these infrequent price movements might lead to
incorrect covariance estimations between banks.
It should be pointed out that this indeed is a source of model risk and requires
special treatment, which is out of the scope of this work. To deal with this, instead of
rejecting the bank completely, the prices were forward-filled to a limit of five
business days and not more, leading to missing values in the price time series. If
within the simulation period for a day, a bank had missing data after this procedure,
the bank was not considered for that particular day. In our dataset after this process,
116 institutions remained out of 126.
Model Risk of VaR and ES Using Monte Carlo: Study on Financial. . . 77

3 Theory and Computational Methods

3.1 Value at Risk

VaR is defined as the maximum possible loss that can be incurred over a particular
time horizon within a particular confidence level. In this work we treat any given
combination of a time horizon and confidence level as a “VaR configuration”. Since
this definition of VaR does not provide a computational algorithm, but only a
property that it must have, multiple ways exist to compute it (Holton 2014). For
our purposes, the time horizon is fixed to 1 day and the confidence levels are chosen
to be 95% and 99%. Furthermore, we use the Monte Carlo approaches to estimate
VaR on a basket of 116 financial institutions from the Paris and Frankfurt stock
exchanges.

3.2 Expected Shortfall

Expected Shortfall (ES) is a market risk measure linked to VaR, which is defined as
the expected average tail loss within a certain confidence level, but unlike VaR, it
accounts for worse scenarios for an institution. It is considered to be more useful than
VaR in terms of capturing market risk in the tail, and hence recommended in the
Basel III agreement. It is calculated for a given confidence level (also called quantile
level) and is defined as the expected portfolio loss when the loss is occurring at or
below this level. As an example, if the average loss on the worst 5% of possible
outcomes is 100 EUR, the expected shortfall is 100 EUR at 5% tail (95% confidence
level).

3.3 Monte Carlo Methods

We chose the MC method (Glasserman 2003) to simulate the relative returns for a
single trading day across all institutions in our dataset. The main advantage of the
MC method is that one can simulate the different sources of uncertainty that affect
the stocks by drawing random numbers from predetermined probability distribu-
tions. As such, the model limitations are mainly due to the choice of the distributions
and the computational costs associated with the generation of statistics. MC methods
have been applied in various areas of finance, such as portfolio optimization and risk
analysis.
Our approach uses MC to simulate the uncertain relative price changes and the
uncertainty is described through the mean, standard deviation and inter-institute
covariance structure, determined by real past historical data. The VaR is then simply
the quantile of these simulated price changes across multiple MC runs corresponding
78 A. H. Pasieczna

to the configuration (e.g. 95%). The ES is the average value of the returns that
crossed the VaR threshold. Four MC models were studied, based on the estimation
methods of the mean and standard deviation, and the choice of inclusion of the
covariance between banks. Two variants to compute the mean and standard devia-
tion were used:
• Rolling approach: In this approach, equal weight is assigned to all points in a
given window, with the mean for a bank at time t as,

1 XN
μt ¼ r 0
t 0 ¼0 tt
N

where N is the historical time window.

• Exponentially weighted approach (EWM): In this approach, more weight is
assigned to more recent events with an exponentially decaying weight curve.
The mean at time t is given as,
X1 0
μt ¼ t 0 ¼0
α ð1  αÞt r tt0

where α is expressed in terms of a center of mass (com), with the com indicating an
equivalent historical window α ¼ comþ1
1
.
Similar expressions exist for estimating standard deviations for each bank and
covariance between two banks. Two additional variants were then constructed, one
which includes the covariance structure of the banks, and the second which assumes
their independence. The algorithm is as follows:
1. For all the banks and for each day, estimate the distribution quantities (mean,
standard deviation, and covariance structure), based on the past historical data on
relative price changes. Two variants were studied: using rolling quantities and
using exponentially weighted quantities. Three historical periods were used:
125, 250, 500 trading days (approximately 0.5, 1 and 2 years).
2. Draw random numbers from a Gaussian distribution with the precalculated
quantities (mean, standard deviation, covariance). These random numbers repre-
sent the next day’s relative returns. Two variants were built: classical approach
that ignores the covariance structure (each bank is independent), and a multivar-
iate approach that samples random numbers using the pre-calculated covariance
structure. For each day and each institution 20,000 numbers (MC iterations) were
drawn based on the predetermined distributions.
3. Rank the simulated absolute returns in a descending manner and choose the
quantile corresponding to the confidence level as the VaR. The ES is the average
of the simulated returns below the VaR. Two confidence levels were tested—95%
and 99%.
Once the VaR and ES were computed for the banks at the given confidence levels
with the three historical periods with different MC approaches, model risk was
estimated using precision-based metrics, described in the next subsection.
Model Risk of VaR and ES Using Monte Carlo: Study on Financial. . . 79

Normal distribution has been assumed for the returns (multivariate, as well as
classical) in our work. This assumption is usually good enough to work with for the
following reasons:
• Simplicity—Normal distribution is well studied and can be applied in a large
range of applications. For this reason, it is also the starting choice for stock
returns.
• Comparability—It can act as a benchmark for models with more complex
distributions. While using models with other distributions, it is difficult to know
whether certain failures are due to the choice of distribution, parametrization, or
simply for computational reasons. This, in fact, increases model risk.
• Interpretability—Deviation from the normal distribution for stock returns hap-
pens mainly in the tails and less so in the bulk. Using a normal distribution
simplifies computational code and allows us easy interpretation of the bulk of the
days, where tail events do not happen. Since our work focuses on trying to
understand the precision of the risk estimates, the actual risk estimate is less
important, and we can focus on the bulk of the days.
Nonetheless, there are some shortcomings in this choice:
• Tail events—Fat tails are not considered within a normal distribution approxi-
mation, and so might cause an underestimation of the market risk. However, since
we look at the spread of multiple measurements and not the actual value of the
market risk, we expect a small impact from this.
• Skew, kurtosis and higher moments—Stock returns might have a tendency to
deviate from the normal distribution, when higher moments are considered. These
can have a significant impact on risk estimation and hence the model risk. In our
case we look at VaR and ES solely based on the mean and standard deviation
(first two moments), and so our model risk estimates are accurate up to these
moments.
• Multi-modal distributions—Non-unimodal distributions have been shown to be
important in VaR and ES estimations (Guegan et al. 2017). However, these
approaches involve many more parameters and are less frequently used than
those with unimodal distributions.

3.4 Model Risk Measures

Model risk refers to the risk that practitioners are exposed to when using a particular
model in their work. In a broad sense, it signifies the uncertainty due to human error
in model application (e.g. parametrization, inapplicability), or intrinsic model short-
comings (such as instability, sensitivity). According to Derman (1996), model risk is
a consequence of general model construction and uncertainty in the field of finance.
This is a view shared by Crouhy et al. (1998). Thus, everything related to a model
can be part of the corresponding model risk, including data contaminations, wrong
80 A. H. Pasieczna

implementations, badly approximated solutions, software or hardware bugs, and

even the practitioners themselves.
Clearly it is then not possible to isolate individual sources of risk for risk control.
Furthermore, the optimal solution or benchmark might simply not exist. Thus, it is
not possible to truly mitigate all these risks, especially in the complex domain of
finance. Other authors look at model risk with more focused points of views. For
example, some research works define model risk as inaccuracy arising from estima-
tion errors and uses of incorrect models (Boucher et al. 2014; Glasserman and Xu
2014; Hendricks 1996). Yet other authors might consider model risk induced by the
probability space used for statistical modeling, choice of tests for the data and
estimation of the model parameters (Sibbertsen et al. 2008). In our work, we focus
on the model risk that arises due to the choice of input parametrization of the model.
Additionally, model risk is represented by the concept of precision, specifically we
look at the similarity of the model estimates across various input parameters.
Two measures were developed based on precision, the spread across ES and VaR
estimates, and the ratio of the highest to lowest ES and VaR estimates. Their
mathematical expressions for VaR are shown below (similar expressions for ES):

max ðVaR125 , VaR250 , VaR500 Þ  min ðVaR125 , VaR250 , VaR500 Þ

spread ¼
meanðVaR125 , VaR250 , VaR500 Þ

max ðVaR125 , VaR250 , VaR500 Þ

ratio ¼
min ðVaR125 , VaR250 , VaR500 Þ

Note that the estimates used in these measures always correspond to the same
configuration (prediction time period equal to 1 day and confidence levels 95%,
99%), but different historical periods. In our case, the spread is the difference
between the maximum and minimum VaR or ES estimates obtained from the three
historical periods (125, 250, 500 trading days) in units of the average estimate,
whereas the ratio is the ratio of the maximum to minimum estimates. Ratio as a
model risk measure has already been used to analyze market risk and systemic risk
measures in literature (Danielsson et al. 2016).

4 Results and Discussion

We present here our simulation results for the 116 financial institutions, where the
1-day VaR and ES were computed at 95% and 99% confidence intervals. Estimates
from three historical periods (125, 250, 500 days) for given confidence intervals
were used to judge both precision metrics. As a recap, four models were built,
classical rolling (rolling window filter with independence of institutions assumed),
multivariate rolling (rolling filter with covariance of institutions included), classical
Model Risk of VaR and ES Using Monte Carlo: Study on Financial. . . 81

Table 1 Precision metrics—spread and ratio for the VaR and ES under different Monte Carlo
approaches with the different confidence levels
Precision (avg ratio) Precision (avg spread/mean)
Rolling EWM Rolling EWM
VAR 95 Classical 1.439 1.364 0.306 0.279
Multivariate 1.438 1.466 0.306 0.337
VAR 99 Classical 1.432 1.348 0.302 0.270
Multivariate 1.431 1.450 0.301 0.328
ES 95 Classical 1.434 1.352 0.303 0.272
Multivariate 1.433 1.454 0.302 0.331
ES 99 Classical 1.430 1.344 0.300 0.267
Multivariate 1.430 1.445 0.300 0.326

EWM (exponential weighted moving window with independence of institutions

assumed), and multivariate EWM (EWM filter with covariance included).
Our results are summarized in Table 1. To analyze the spread, we divided the
difference between the maximum and minimum estimates by the mean to make it
comparable across time (unitless metric). This unit allows us to look at the model
risk (maximum variation in the estimates) in terms of the average estimate providing
for a more intuitive feel of the quantity. For example, if the spread is 0.3, then the
variation can be up to 30% of the average estimate. The ratio on the other hand
provides a different view on the variation in the estimates. For example, if the ratio is
1.4, then the variation in the estimates lies between values that are 1.4 times larger
and 1.4 times smaller than any single estimate measurement. We observe that though
the four models have similar VaR and ES model risks, the classical EWM approach
performed slightly better than the rest, whereas the multivariate EWM performed
slightly worse than the remaining models.
To compare temporal evolution of the precision metrics, we plot the average VaR
and ES estimates across all financial institutions along with the spread and ratio
metrics in Figs. 1, 2, 3 and 4. These plots allow us to study how the various quantities
evolve during periods of financial crises (e.g. 2008, 2020) and post crises which are
characterized by changes in volatility of relative returns. Intuitively we expect model
risk to increase when volatility regimes change. To understand this further, consider
the case where the volatility increases (typically associated with a crisis period or
negative returns). Here, a VaR or ES estimate based on 125 days will react faster
than an estimate based on 500 days, causing a jump in the model risk metrics.
Interestingly, if the volatility drops, the same effect occurs causing yet another jump
in the model risk.
Figures 1 and 2 present the VaR at 99% and 95% plots with the corresponding
risk measures, respectively. We observe that the VaR tends to increase in magnitude
during periods of turmoil. The model is considered precise when the spread and ratio
are as close as possible to 0 and 1 respectively, which would imply that the model is
not highly dependent on the input parameters (here, historical time period). How-
ever, in our study we observe that the ratio is never 1, and the spread is always
82 A. H. Pasieczna

Fig. 1 Results for 1-Day VaR computed at 99% confidence level

Fig. 2 Results for 1-Day VaR computed at 95% confidence level

greater than 0. This means that the model risk is never completely nullified.
Interestingly we observe that though both model risk metrics provide different
quantitative viewpoints on variability of the estimates, they are visually very similar
across time.
Furthermore, we see an increase in these model risk measures during crisis and
post crisis periods. However, we see a maximal peak in the ratio and spread for the
rolling approaches that seem to be a lagged effect of the 2008 crisis. We interpret this
as the consequence of one estimate having a memory of 500 days, compared to
Model Risk of VaR and ES Using Monte Carlo: Study on Financial. . . 83

Fig. 3 Results for 1-Day ES computed at 99% confidence level

Fig. 4 Results for 1-Day ES computed at 95% confidence level

another estimate with a memory of 125 days. EWM approaches do not show that
strong an effect, mainly because they are much more smooth filter functions on past
data, which leads to a slightly lower model risk during these periods.
Figures 3 and 4 shows the results for the ES at 99% and 95% confidence levels,
respectively. We see similar behavior as for the VaR, which we interpret to be a
consequence of the fact that the ES is derived from the VaR. Indeed, the ES can be
represented as an integral of the VaR at all quantiles above the threshold. Comparing
the Figs. 1–4 with the Table 1 we conclude that though the model risk was not very
84 A. H. Pasieczna

high overall, the main inconsistency is observed in times when the measures like
VaR and ES are most needed.

5 Conclusions and Perspectives

We presented our analysis for the VaR and ES calculations of financial institutions
traded on the Paris and Frankfurt stock exchanges. In particular, the precision-based
model risk measures of ratio and spread were looked at based on different input
parameters. Four Monte Carlo approaches were studied, based on how distribution
characteristics were estimated. Our results indicate that the classical EWM approach
performed best in terms of both precision metrics, whereas the multivariate EWM
performed worst. The rolling-based approaches were similar for both, the classical
and multivariate cases and had performance in between those of the EWM
approaches. The temporal evolution of the precision metrics showed that model
risk was high during and just after market crises, i.e. during periods when market risk
measures matter most.
For future study we recommend considering different distributions, such as
Weibull and Student distributions which might allow us to capture tail events.
Additionally, it might be insightful to compare these models in a multidimensional
setting, across different distributions and their parametrizations with the intention of
understanding the importance of each component of model risk. Multivariate (cop-
ula) and multimodal distributions might also be studied with our approach to analyze
the model risk.

References

Basel II (2004) Basel II: International convergence of capital measurement and capital standards: a
revised framework. https://www.bis.org/publ/bcbs107.htm. Accessed 20 Dec 2019
Basel III (2010). https://www.bis.org/press/p100912.pdf. Accessed 20 Aug 2020
Boucher CM, Danielsson J, Kouontchou PS, Maillet BB (2014) Risk models-at-risk. J Bank Financ
44:72–92
Crouhy M, Galai D, Mark R (1998) Model risk. J Financ Eng 7:267–288
Danielsson J et al (2016) Model risk of risk model. J Financ Stability 23(C):79–91
Derman E (1996) Model risk: what are the assumptions made in using models to value securities
and what are the consequent risks? Risk-London-Risk Magazine Limited 9:34–38
Glasserman P (2003) Monte Carlo methods in financial engineering. Springer Science+Business
Media, New York, NY
Glasserman P, Xu X (2014) Robust risk measurement and model risk. Quantitative Financ 14
(1):29–58
Guegan D, Hassani B, Li K (2017) Impact of multimodality of distributions on VaR and ES
calculations. Documents de travail du Centre d’Economie de la Sorbonne 17019, Université
Panthéon-Sorbonne (Paris 1), Centre d’Economie de la Sorbonne
Hendricks D (1996) Evaluation of value-at-risk models using historical data. Econ Policy Rev 2(1)
Model Risk of VaR and ES Using Monte Carlo: Study on Financial. . . 85

Holton GA (2014) Value-at-risk theory and practice, 2nd edn (self-published). https://www.value-
at-risk.net. Accessed 20 Aug 2020
Investpy (2020) investpy – Financial Data Extraction from Investing.com with Python. developed
by Alvaro Bartolome del Canto. https://github.com/alvarobartt/investpy. Accessed 01 Sept 2020
Pasieczna AH (2019) Monte Carlo Simulation Approach to Calculate Value at Risk: Application to
WIG20 and MWIG40. Financial Sciences, Wrocław University of Economics, Wrocław 24
(2):61–75
Sibbertsen P, Stahl G, Luedtke C (2008) Measuring model risk. J Risk Model Valid 2(4):65–82
Tick Size Reduction and Liquidity
Dimensions: Evidence from an Emerging
Market

Quoc-Khang Pham

1 Introduction

Tick size on the stock market is the minimum price movement of a trading stock.
Price movements vary significantly across stock exchanges. Tick size represents the
minimum amount a stock price can move either up or down on an exchange.
Recently, stock exchanges around the world have used a new tick size system.
The purpose of the variation is to decrease transaction costs and improve market
liquidity. For instance, the New York Stock Exchange and NASDAQ reduced the
price increment from $US1/8 to $US1/16 in 1997; then, they reduced tick size to
$US0.01 in 2001. The Stock Exchange of Thailand (SET) reduced tick size on
5 November 2001. The Warsaw Stock Exchange (GPW) employed a new tick size
system on 4 March 2019. The tick size of shares on the GPW is determined by stock
prices and the liquidity measure (turnover).
Several articles tell of empirical investigation that determined the impact of tick
size changes on market structures. A smaller tick size decreases the transaction costs
and spread. Harris (1991) authored the first study which examined the effects of a
tick size reduction. Harris (1991) and Chordia et al. (2001) indicate that a decrease in
a pricing grid leads to a reduction in the bid-ask spread. On the other hand, the U.S
Securities and Exchange Commission (SEC) implemented a changing tick size pilot
program. The pilot program would increase the minimum tick size from 1 cent to
5 cents for small market capitalization stocks. The tick size change took place over
2 years, from 3 October 2016 to 28 September 2018. The tick size pilot program
inspired several studies (e.g., Rindi and Werner 2017; Griffith and Roseman 2019;
Chung et al. 2020) to examine the impact of tick size on the financial market’s

Q.-K. Pham (*)

Wroclaw University of Economics and Business, Wroclaw, Poland
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 87

K. Jajuga et al. (eds.), Contemporary Trends and Challenges in Finance, Springer
Proceedings in Business and Economics,
https://doi.org/10.1007/978-3-030-73667-5_6
88 Q.-K. Pham

aspects. Rindi and Werner (2017) argue that liquidity demanders would have higher
costs and earn higher profits. Griffith and Roseman (2019) analyze the three dimen-
sions of liquidity following a tick size increase on the NASDAQ. The tick size
increase results in a higher spread, less resiliency, and lower depth. Albuquerque
et al. (2020) study the relationship between a widening in tick size and stock prices in
the U.S after the pilot program. The results show a negative impact of tick size
increase on stock prices. Albuquerque et al. (2020) find a growth in two proxies of
transaction costs (quoted spreads and effective spreads) and price-impact proxy, but
a decline in trading volume. Chung et al. (2020) explore the changes in liquidity in
small and large orders followed by a widening of tick size on the U.S stock market.
Chung et al. (2020) find that liquidity decreases for small orders but increases for
large orders.
The stock exchanges adjust tick size to increase competitiveness among investors
in an order-driven market. The previous empirical studies that examined the influ-
ence of tick size reduction on the order-driven markets show conflicting results.
Aitken and Comerton-Forde (2005) found that tick size reduction improved liquidity
on the Australian Securities Exchange. The average spread for the lowest price range
group fell by 26%. However, it reduced liquidity in a group with a small tick size and
a low trading volume group. Hsieh et al. (2008) indicate that a reduction in tick size
increases market efficiency and decreases trading costs on the Taiwanese Stock
Market. However, Pan et al. (2012) find a different result; aggregate liquidity for a
liquid stock group declined in an order-driven market, the Hong Kong Stock
Exchange. Anderson and Peng (2014) examine the impact of a tick size change
experienced in 2011 by 17 eligible stocks in New Zealand. Anderson and Peng
(2014) found that both spreads and depth decrease significantly during the post-
period. This study has a limitation in that it could not determine significant statistical
changes in liquidity. Bacidore (1997) analyzes the tick size reduction on the Toronto
Stock Exchange, stating that spreads and depth declined, but there was no change in
trading volumes. Ahn et al. (2007) find a significant decrease in spreads, resulting in
a tick size reduction on the Tokyo Stock Exchange. However, Ahn et al. (2007) do
not see a significant change in trading volume. Kuo et al. (2010) explore the impact
of minimum price changes on the market liquidity on the Taiwan Stock Exchange
(TWSE), which occurred on 1 March 2005. The results indicate that a decrease in
tick size on the TWSE led to a decline in three transaction cost proxies: quoted
spread, liquidity premium, and execution costs. However, two market depth proxies
presented a significant decrease, including the best-quoted depth and the cumulative
depth, because the market participants spent less time or executed a larger quantity of
shares. In light of these findings, the study hypothesizes that the reduction in tick size
leads to decreased quoted spread and increased trading volume.
Liquidity is a multidimensional concept; it includes tightness, depth and resil-
ience dimensions (Kyle 1985). The tightness dimensions refer to the transaction
costs, which are measured by various proxies based on bid/ask spread. The depth
dimensions determine a possible trade size with a minimum price impact, captured
by the volume-based or turnover-based measures. The resilience dimensions refer to
Tick Size Reduction and Liquidity Dimensions: Evidence from an Emerging Market 89

a characteristic of markets in which new orders flow quickly to correct order

imbalances and return the prices to fundamental values (Pham-Quoc 2020).
The Ho Chi Minh Stock Exchange (HOSE) in Vietnam implemented a tick size
reduction on 12 September 2016. The purpose of policymakers was to enhance
market liquidity following the decline of tick size on the HOSE. This study aims to
explore the effects of tick size reduction on the various liquidity aspects on the
HOSE. To the best of my knowledge, the article is the first one that to investigate the
relationship between tick sizes and liquidity aspects in the context of the emerging
Vietnamese market.
Changes in the minimum tick size lead to changes in market participants’ trading
activities. Market participants are more flexible to offer a bid or ask prices. Regard-
ing the empirical evidence in the literature, the study expects that the market liquidity
is higher in the tightness, depth and resilience dimensions after the tick size
reduction.
The remainder of the paper is organized as follows. In Sect. 2, the paper presents a
brief description of the stock exchange features in Vietnam. Section 3 indicates the
selection of data in the sample and introduces the methodology used in this study.
Section 4 presents and discusses the empirical findings in liquidity dimensions after
tick size reduction. The paper includes some concluding remarks in Sect. 5.

2 Institutional Features of the Ho Chi Minh Stock

Exchange

The Vietnamese stock market includes two stock exchanges: The Ho Chi Minh
Stock Exchange (HOSE) and The Hanoi Stock Exchange (HNX). According to the
State Securities Commission of Vietnam, the Vietnamese stock market’s participants
contained about 28.5 thousand accounts of foreign investors and approximately 1.7
million domestic investors account as of the end of 2019. The HOSE has officially
operated since 2000, while the HNX was established later in 2005. The HOSE is the
larger of the two stock exchanges in Vietnam, with a market capitalization of about
$US200 billion, over ten times more than on the HNX. The HOSE operates the
trading as an order-driven market; without market makers as is the case with the
stock exchanges in London or New York, investors execute their transactions
through an automated order-matching system. The HOSE implements multiple
tick size systems, with different tick sizes for different price ranges. Trading time
on the HOSE takes place 5 days a week from Monday to Friday. It includes two
kinds of the trading auction, i.e. continuous auction and periodic auction. The
investors on the HOSE can make transactions in four different trading sessions
during a trading day. The stock exchange determines the open and close-price
each day using a single-price auction in the first and last session with types of
order: At the opening order matching price (ATO), At the closing order matching
price (ATC), limit order (L.O.). For the remainder of the day’s sessions, the HOSE is
90 Q.-K. Pham

Fig. 1 Changes in the VNIndex. Notes: The figure shows the changes of the market indicator,
VNIndex from 24/06/2016 to 28/11/2016. The vertical line, centered on 12 September, 2016,
indicates the date that the HOSE implemented the tick-size reduction. Source: Ho Chi Minh Stock
Exchange

operated in a continuous auction with two types of order: limit order (L.O.), market
order (M.P.). The matching principles are implemented according to the priority of
price, then time.
Figure 1 presents the market performance changes, named VNIndex, from
24 June 2016 to 28 November 2016. The VNIndex on the HOSE showed an upward
trend during this period. The market index was 620.7 at the beginning of the period
and then peaked at 688.89 at the end of September 2016. The VNIndex decreased
slightly in the following months.

3 Data and Methodology

The HOSE launched a new system of tick size on 12 September 2016. The empirical
sample consists of listed companies on the HOSE, covering the 3-month periods
before and after the event date. Following Aitken and Comerton-Forde (2005),
Hsieh et al. (2008), the data exclude the five trading days before and after the
event date. This method is to avoid any unusual trading behavior surrounding the
event date. The study uses a set of daily trading data, which is obtained from
Thomson Reuter Datastream.
The literature investigates the influence of the tick size reduction by separating
the sample into two periods: the pre-tick-size-reduction and the post-tick-size-
reduction period. For instance, Chen and Hsieh (2013) examine the changes in
liquidity before and after implementing a new tick-size rule on 1 March 2005.
They compare the market liquidity between 10-day trading in the pre-event periods
and 10-day trading in the post-event periods on the Taiwan Stock Exchange. Kuo
Tick Size Reduction and Liquidity Dimensions: Evidence from an Emerging Market 91

Table 1 Tick size rule before and after the tick size reduction on the HOSE
Number Tick size in Tick size in
of Price range pre-event post-event (%)
Group companies (VND) (VND) (VND) Difference Difference
1 108 <10,000 100 10
90
90
2 173 10,000–49,950 100 50
50
50
3 28 50,000–99,500 500 100
400
80
4 3 >100,000 1000 100
900
90
Notes: The event date of tick size reduction was on 12 September 2016. The pre-event period was
from 24 June 2016 to 1 September 2016, and the post-event period was from 20 September 2016 to
28 November 2016
Source: Ho Chi Minh Stock Exchange

et al. (2010) use a sample that includes 52 trading days and 59 trading days before
and after the change of tick size rule on the Taiwan Stock Exchange. Aitken and
Comerton-Forde (2005) use a sample covered 80-day trading, including before and
after the tick size reduction periods, to study market liquidity changes on the Jakarta
Stock Exchange. Ahn et al. (2007) conduct a study to compare market liquidity
between 68 trading days in the pre-event period and 76 trading days in the post-event
period. Porter and Weaver (1997) have a comparison 1 month before and 1 month
after the tick size reduction on the Toronto Stock Exchange, which was implemented
on 1 April 1996. So, the 50-day trading period in the pre- and post-event period is
sufficient to observe the traders’ trading behavior changes comprehensively. After
tick size reduction, traders would change their trading behavior immediately to
become acquainted with new changes. The paper will analyze the changes in
liquidity before and after implementing a reduction in tick size. The time frame
includes pre- and post-event periods. The sample covers 50 trading days in the
pre-event period (from 24 June 2016 to 1 September 2016) and 50 trading days in the
post-event period (from 20 September 2016 to 28 November 2016). The sample
stocks need to meet the following criteria: (1) stocks are still listing on the HOSE
until the end of 2016, (2) the spread was not negative or missing. The number of
companies listed on the HOSE at the end of 2016 was 320. However, the sample
includes 312 companies listed on the HOSE that are eligible.
The HOSE has a multiple tick size system, in which tick sizes vary in different
price regimes. After implementing a new tick size system on 12 September 2016, the
price grids on the HOSE are separated into four groups (Table 1). Group 1 contains
stocks whose prices are below VND 10,000.1 Tick size for this group is reduced by
90%, from VND 100 to VND 10. The second group includes stocks whose price
range from VND 10,000 to VND 49,950. Tick size for this group drops by 50% to
VND 50. The price range in the third group is from VND 50,000 to VND 99,500.
The HOSE adjusted tick size from VND 500 to VND 100. The HOSE implemented

1
Exchange rate of USD/VND in 2016 was 1 USD ¼ 22,300 VND.
92 Q.-K. Pham

a tick size reduction for the last group, which decreased by 90% to VND100. The
sample, including 312 companies, was split into four subsamples as four price
regimes to analyze the tick size reduction effect. The number of companies in the
groups is 108, 173, 28, and 3, respectively.
The effects of the tick size reduction in this study are examined in different
dimensions, i.e., tightness, depth and resiliency. The study analyzes the stock
liquidity dimensions in a numerous liquidity proxy for an emerging market. Thus,
the study constructs a set of low-frequency measures for empirical research.
Low-frequency proxies are advantages to measure liquidity efficiently on every
stock market. The daily data is available for not only developed but also emerging
markets. The recent studies provide enormous literature on low-frequency (daily and
monthly) liquidity proxies (Le and Gregoriou 2020). This section presents the
low-frequency proxies regarding the liquidity dimensions, such as depth, tightness,
and resilience.
Tightness dimensions, or known as transaction costs, are often measured by the
spread measures. The most common proxies in the tick size studies are quoted spread
and effective spread. Quoted spread, QSPRi, d, is the difference between the best ask
price and the best bid price as

QSPRi,d ¼ PAi,d
PBi,d , ð1Þ

where PAi,d , PBi,d are the best ask price and the best bid price for stock i on day d.
The second proxy of the tightness dimensions is effective spread, EFSPi, d,
calculated as two times the absolute difference between the transaction prices and
the midpoint of the quoted spread, as




PAi,d þ PBi,d

EFSPi,d

¼ 2 
PRICE i,d

, ð2Þ
2

where PRICEi, dis the closing price for i on day d.

The depth dimensions are captured by four common measures: traded volume,
traded value, turnover ratio and Amihud (2002) measure. Traded volume, VOLi, d, is
the number of traded shares for stock i on day d. Traded value, VALi, d, is the amount
of traded value for stock i on day d. And the turnover ratio is a ratio of the traded
volume over the number of shares outstanding as

VOLi,d
TURN i,d ¼ , ð3Þ
NOST i,m

where TURNi, d is the turnover ratio for stock i on day d; NOSTi,m is the number of
shares outstanding for stock i in month m.
The last measure in the depth dimensions is Amihud (2002) measure, which is the
best proxy in the price-based measures (Fong et al. 2017). This illiquid measure is
calculated as
Tick Size Reduction and Liquidity Dimensions: Evidence from an Emerging Market 93

jRi,d j
AMIHUDi,d ¼ , ð4Þ
VALi,d

where Ri, d, VALi, d are the return, and the trading value for stock i on day d,
respectively.
Regarding the resilience dimensions, the study follows Hasbrouck and Schwartz
(1988), using the market efficiency coefficient (MEC) to proxy resiliency. The MEC
indicates the fact that price movements continuous in liquid markets. Hasbrouck and
Schwartz (1988) propose that MEC is defined as the ratio of observed long-variances
and short-variances.

1 Var ðLRt Þ
MEC ¼   ð5Þ
T Var SRtÞ ,

where Var(LRt) is the variance of the logarithm of long-period returns, Var(SRt) is

the variance of the logarithm of short-period returns, T is the number of short periods
in each long-period. The resilient markets are closer but slightly below one (Sarr and
Lybek 2002).
In my research, the study calculates the MEC over a week of trading (five trading
days period) as a long-period and a given trading day as a short-period. The long-
period variance (weekly), Var(LRt) is calculated following Martens and Van Dijk
(2007) as Eq. (6). The short-period variance, Var(SRt) calculated follows the study
introduced by Parkinson (1980) as Eq. (7).

1 X5
Var ðLRt Þ ¼ ðh
ld,t Þ2 , ð6Þ
4 log ð2Þ d¼1 d,t

1 X hd,t
ld,t
5
Var ðSRt Þ ¼  , ð7Þ
5 d¼1 4 log ð2Þ

where hd, t, ld, t are the logarithm of the high and low prices for day d within week
t (five trading days period).
The study is inspired by Ahn et al. (2007), Anderson and Peng (2014) as the
standard procedure in examining the changes in liquidity dimensions after tick size
reduction on the HOSE in September 2016. For this purpose, the study constructs
aggregate liquidity measures using firm-level liquidity. The study first calculated
daily measures for all stocks in the sample except for the MEC, which is calculated
weekly. Then the study applied the value-weighted average based on market capi-
talization to estimate the aggregate liquidity.
The statistical significance in the difference during the pre- and post-event periods
is tested using the Wilcoxon Signed-Rank Test, known as the Wilcoxon matched-
pairs test. The Wilcoxon test is one of the most common tests to evaluate the same
subjects under two sets of conditions. The Wilcoxon Signed Rank Test, based on
94 Q.-K. Pham

Table 2 Summary statistics for stock trading in four tick size categories
Mean Minimum Maximum

Group 1 (n ¼ 108 companies)

Market capitalization (VND billion) 1218.4 27.1 25,283.3
Daily trading volume (thousand shares) 578.6 0.0 47,752.8
Daily trading value (VND million) 3701.2 0.7 299,250.9

Group 2 (n ¼ 173 companies)

Market capitalization (VND billion) 5131.4 62.4 153,238.7
Daily trading volume (thousand shares) 483.1 0.0 49,007.9
Daily trading value (VND million) 8968.2 0.0 552,959.0

Group 3 (n ¼ 28 companies)
Market capitalization (VND billion) 17,227.6 312.8 226,426.6
Daily trading volume (thousand shares) 162.4 0.0 9539.5
Daily trading value (VND million) 13,329.9 1.0 919,009.0

Group 4 (n ¼ 3 companies)
Market capitalization (VND billion) 6112.6 914.2 13,533.7
Daily trading volume (thousand shares) 95.9 0.0 515.8
Daily trading value (VND million) 12,580.4 1.0 3500.0
Notes: Group 1, 2, 3 and 4 indicate the first group with a price range under VND 10,000, the second
group with a price range from VND 10,000 to VND 49,950, the third group with a price range from
VND 50,000 to VND 99,500, the fourth group with price range above VND 100,000, respectively

different scores, analyzes the signs of difference through the magnitude of observed
differences. The Wilcoxon Signed-Rank Test reveals the null hypotheses’ statistic
results, where the median difference is zero. The Wilcoxon test is applied commonly
in the literature, for instance, Ahn et al. (2007), Ascioglu et al. (2010), Pan et al.
(2012), Chen and Hsieh (2013), Anderson and Peng (2014).
Table 2 reports the descriptive statistics on the market capitalization and trading
activities of stocks in four groups on the HOSE from 24 June 2016 to 28 November
2016.
Table 2 presents the difference in market capitalization on the HOSE in the
sample. The first and second groups include stocks with a price under VND
50,000, having the smallest average market capitalization with VND 1218.4 billion
and VND 5131.4 billion. The highest average market value is in the third group at
VND 17,227.6 billion. The market value in this group varies from VND 312.8
billion to VND 226,426.6 billion. The fourth group contains stocks with the highest
stock price, above VND 100,000. However, the market capitalization is only VND
6112.6 billion. Investors on the HOSE traded the most on stocks in the third and
fourth groups. Specifically, they spend VND 13,329.9 million and VND 12,580.4
million for trading stocks in these groups, respectively. The highest trading value per
day is also in the third group at VND 919,009.0 million. The third and fourth groups’
Tick Size Reduction and Liquidity Dimensions: Evidence from an Emerging Market 95

daily average trading volume is smaller than the two first groups with 162.4 and 95.9
thousand shares, respectively. However, the first group has the highest average
trading volume at 578.6 shares per day. Because the stock price in this group is
minimal, the trading value is the smallest at VND 3701.2 million.

4 Empirical Results

This section presents the description of stock liquidity before and after the event date
and analyzes the empirical results. The paper will discuss the changes in liquidity
dimensions, i.e., the tightness, depth and resilience.

4.1 Tick Size Reduction and Tightness Dimensions

This section explores the impact of changes in minimum tick size on the tightness
dimensions. The study calculates two proxies for the tightness dimensions as quoted
spread and effective spread.
Table 3 provides a comparison in the quoted spread between the pre- and post-
event periods. In general, the quoted spread decrease after the reduction in tick size
in four subsamples. The stocks have a price under VND 10,000 in group 1, illustrate
a reduction in the quoted spread. The average quoted spread for group 1 fell from
VND 159.12 to VND 146.55. Groups 2 and 3 also have a quoted spread decline,
which accounts for about 3.8%. However, the pre- and post-event differences do not
present significant statistics in the Wilcoxon test. The most change in quoted spread
occurs in group 4, where stock prices are above VND 100,000. The quoted spread
falls by 8.87%, from VND 1536.19 to VND 1399.91. The results prove significant

Table 3 Comparison of pre- and post-event in the quoted spread

Group Pre-event Post-event Difference (%) Difference Signed-rank p-value
1 159.12 146.55
12.57
7.90 0.00***
2 361.76 347.98
13.78
3.81 0.05**
3 986.03 949.17
36.86
3.74 0.26
4 1536.19 1399.91
136.29
8.87 0.04**
Notes: The quoted spread is the difference in the best bid and best ask prices. This proxy is
calculated for the pre-event period from 24 June 2016 to 1 September 2016, and the post-event
period from 20 September 2016 to 28 November 2016. The whole sample includes all stocks in the
sample. Group 1, 2, 3 and 4 indicate the first group with a price range under VND 10,000, the
second group with a price range from VND 10,000 to VND 49,950, the third group with a price
range from VND 50,000 to VND 99,500, the fourth group with a price range above VND 100,000,
respectively. The study tests the null hypothesis that a tick size reduction does not decrease the
quoted spread. *, **, and *** denote the statistical significance p-value from the Wilcoxon
one-tailed test between the pre- and post-periods at the 10%, 5%, and 1% levels, respectively
96 Q.-K. Pham

Table 4 Comparison of pre- and post-event in the effective spread

Group Pre-event Post-event Difference (%) Difference Signed-rank p-value
1 75.76 65.88
9.88
13.04 0.00***
2 203.58 211.12 7.54 3.70 0.96
3 513.67 495.32
18.35
3.57 0.04**
4 1,175.00 978.50
196.50
16.72 0.04**
Notes: The effective spread is two times the absolute difference between the trade price and the
bid-ask midpoint. This proxy is calculated for the pre-event period from 24 June 2016 to
1 September 2016, and the post-event period from 20 September 2016 to 28 November 2016.
The whole sample includes all stocks in the sample. Group 1, 2, 3 and 4 indicate the first group with
a price range under VND 10,000, the second group with a price range from VND 10,000 to VND
49,950, the third group with a price range from VND 50,000 to VND 99,500, the fourth group with
a price range above VND 100,000, respectively. The study tests the null hypothesis that a tick size
reduction does not decrease the effective spread. *, **, and *** denote the statistical significance
p-value from the Wilcoxon one-tailed test between the pre- and post-periods at the 10%, 5%, and
1% levels, respectively

statistical evidence for decreasing the quoted spread in groups 1, 2 and 4 after the
tick size reduction. However, the decrease of quoted spread in group 3 is not
statistically significant.
Table 4 shows the Wilcoxon test’s results to examine the difference in the
effective spread. The liquidity proxy reduces during the post-event, except for
group 2. The effective spread decreases the most in groups 3 and 4 with stock prices
above VND 50,000. The Wilcoxon test results indicate that the study can reject the
null hypothesis in groups 1, 3 and 4. Tick size reduction decreases the effective
spread and significant at the 0.05 level. The liquidity improves in the post-periods in
these groups. However, the Wilcoxon test indicates that the changes in the effective
spread are not statistically significant. The results support that tick size reduction
decreases the effective spread.

4.2 Tick Size Reduction and Depth Dimensions

A market is deep when there is a massive flow of trading orders on both the buy and
sell-side frequently. The depth dimensions are captured through four proxies: trading
volume, trading value, turnover ratio, and Amihud (2002) measure. This section
examines the difference in the liquidity proxies after the tick size reduction on
the HOSE.
Table 5 provides the detailed changes in the trading volume for four groups
before and after the event. The post-periods exhibit a decline in trading volume in the
three first groups. Specifically, the trading volume in group 3 decreases sharply,
about 88% in the post-periods. The Wilcoxon tests present that cannot reject the null
hypothesis in all groups. Trading volume does not increase during the post-events
following the tick size reduction.
Tick Size Reduction and Liquidity Dimensions: Evidence from an Emerging Market 97

Table 5 Comparison of pre- and post-event in the trading volume

Group Pre-event Post-event Difference (%) Difference Signed-rank p-value
1 518,180.8 441,512.0
76,669.2
14.80 0.99
2 385,686.7 317,993.0
67,693.9
17.55 0.99
3 465,487.9 57,141.7
40,8346.0
87.72 0.99
4 236,867.9 238,411.0 1543.2 0.65 0.58
Notes: Trading volume is the number of traded shares for stocks. This proxy is calculated for the
pre-event period from 24 June 2016 to 1 September 2016, and the post-event period from
20 September 2016 to 28 November 2016. The whole sample includes all stocks in the sample.
Group 1, 2, 3 and 4 indicate the first group with a price range under VND 10,000, the second group
with a price range from VND 10,000 to VND 49,950, the third group with a price range from VND
50,000 to VND 99,500, the fourth group with a price range above VND 100,000, respectively. The
study tests the null hypothesis that a tick size reduction does not increase the trading volume. *, **,
and *** denote the statistical significance p-value from the Wilcoxon one-tailed test between the
pre- and post-periods at the 10%, 5%, and 1% levels, respectively

Table 6 Comparison of pre- and post-event in the trading value

Group Pre-event Post-event Difference (%) Difference Signed-rank p-value
1 3251.53 2232.02
1019.51
31.35% 0.99
2 10,868.47 8900.44
1968.03
18.11% 0.99
3 5896.67 3871.99
2024.68
34.34% 0.99
4 36,262.70 34,786.38
1476.32
4.07% 0.33
Notes: Trading value is the amount of traded value for stocks. This proxy is calculated for the
pre-event period from 24 June 2016 to 1 September 2016 and the post-event period from
20 September 2016 to 28 November 2016. The whole sample includes all stocks in the sample.
Group 1, 2, 3 and 4 indicate the first group with a price range under VND 10,000, the second group
with a price range from VND 10,000 to VND 49,950, the third group with a price range from VND
50,000 to VND 99,500, the fourth group with a price range above VND 100,000, respectively. The
study tests the null hypothesis that a tick size reduction does not increase the trading value. *, **,
and *** denote the statistical significance p-value from the Wilcoxon one-tailed test between the
pre- and post-periods at the 10%, 5%, and 1% levels, respectively

Table 6 details the changes in the trading value during the pre- and post-periods.
Unfortunately, the trading value decreases in the post-event periods in four sub-
samples. Groups 1 and 3 have the most declines, above 30%. In contrast, the highest
stock price group has the least changes in trading value, about 4%. The Wilcoxon
tests prove that they cannot reject the null hypothesis. The tick size reduction does
not increase the trading value in the post-periods.
Table 7 presents the changes in another proxy of the depth dimensions, turnover
ratio during the pre- and post-periods. The results present declines in the turnover
ratio on the HOSE after the tick size reduction. The first group has the highest
turnover ratio in the pre-periods with 3.82  10
3. The proxy declines the most by

1.6  10
3 following the tick size reduction. On the other hand, the last group has
the least difference in the post-periods, which occurs similarly in remained proxies of
the depth dimensions. The Wilcoxon tests do not show significant statistical evi-
dence to reject the null hypothesis. They suggest that the tick size reduction does not
98 Q.-K. Pham

Table 7 Comparison of pre- and post-event in the turnover ratio

Group Pre-event Post-event Difference (%) Difference Signed-rank p-value
1 3.8210
3 2.210
3
1.610
3
41.05 0.99
2 3.6210
3 2.610
3
110
3
26.65 0.99
3 2.210
3 1.710
3
510
4
23.85 0.99
4 710
4 610
4
3.410
5
5.07 0.16
Notes: Turnover ratio is a ratio of the traded volume over the number of shares outstanding. This
proxy is calculated for the pre-event period from 24 June 2016 to 1 September 2016 and the post-
event period from 20 September 2016 to 28 November 2016. The whole sample includes all stocks
in the sample. Group 1, 2, 3 and 4 indicate the first group with a price range under VND 10,000, the
second group with a price range from VND 10,000 to VND 49,950, the third group with a price
range from VND 50,000 to VND 99,500, the fourth group with a price range above VND 100,000,
respectively. The study tests the null hypothesis that a tick size reduction does not increase the
turnover ratio. *, **, and *** denote the statistical significance p-value from the Wilcoxon
one-tailed test between the pre- and post-periods at the 10%, 5%, and 1% levels, respectively

Table 8 Comparison of pre- and post-event in the Amihud (2002) measure

Group Pre-event Post-event Difference (%) Difference Signed-rank p-value
1 410
4 510
4 110
4 30.58 0.99
2 110
3 1.0410
3
410
5
4.36 0.22
3 710
4 210
3 1.310
3 148.68 0.99
4 310
4 210
4
810
5
31.23 0.37
Notes: Amihud (2002) measure is a ratio of the absolute stock returns over trading value. This proxy
is calculated for the pre-event period from 24 June 2016 to 1 September 2016, and the post-event
period from 20 September 2016 to 28 November 2016. The whole sample includes all stocks in the
sample. Group 1, 2, 3 and 4 indicate the first group with a price range under VND 10,000, the
second group with a price range from VND 10,000 to VND 49,950, the third group with a price
range from VND 50,000 to VND 99,500, the fourth group with a price range above VND 100,000,
respectively. The study tests the null hypothesis that a tick size reduction does not decrease Amihud
(2002) measure. *, **, and *** denote the statistical significance p-value from the Wilcoxon
one-tailed test between the pre- and post-periods at the 10%, 5%, and 1% levels, respectively

increase the turnover ratio, and market liquidity is not enhanced relating to the
decline of tick size minimum.
Table 8 provides the changes in Amihud (2002) measure, an illiquid proxy of the
depth dimensions in the post-event periods. The results indicate that the measure in
groups 1 and 3 become higher than before the tick size reduction. It suggests that
market liquidity decreases groups 1 and 3 during the post-periods. Stocks in group
3 have the most remarkable change in this measure that increases by 2  10
3. In
contrast, the proxy declines slightly by 1  10
3 and 2  10
4 in groups 2 and
4. The liquidity improves modestly in these groups after the changes of tick size
minimum. However, the Wilcoxon tests prove that Amihud (2002) measure
decreases in the post-periods. It suggests that market liquidity is not improved
following the tick size reduction.
To sum up, the Wilcoxon tests do not illustrate statistical evidence that three
liquid proxies (trading volume, trading value and turnover ratio) increase, and the
Tick Size Reduction and Liquidity Dimensions: Evidence from an Emerging Market 99

Table 9 Comparison of pre- and post-event in the MEC measure

Group Pre-event Post-event Difference (%) Difference Signed-rank p-value
1 1.3410
2 1.2810
2
610
4
4.37 0.99
2 1.0610
2 1.0510
2
110
4
0.60 0.32
3 8.1010
2 7.2010
3
910
4
11.04 0.99
4 8.4010
2 910
3 610
4 6.79 0.29
Notes: MEC measure is the ratio of observed long-variances and short-variances. This proxy is
calculated for the pre-event period from 24 June 2016 to 1 September 2016, and the post-event
period from 20 September 2016 to 28 November 2016. The whole sample includes all stocks in the
sample. Group 1, 2, 3 and 4 indicate the first group with a price range under VND 10,000, the
second group with a price range from VND 10,000 to VND 49,950, the third group with a price
range from VND 50,000 to VND 99,500, the fourth group with a price range above VND 100,000,
respectively. The study tests the null hypothesis that a tick size reduction does not decrease the
MEC measure. *, **, and *** denote the statistical significance p-value from the Wilcoxon
one-tailed test between the pre- and post-periods at the 10%, 5%, and 1% levels, respectively

illiquid proxy (Amihud (2002) measure) decrease in the post-periods. Thus, the
depth proxies present that tick size reduction does not improve market liquidity.

4.3 Tick Size Reduction and Resilience Dimensions

Table 9 provides a comparison between the pre- and post-event periods in a proxy of
the resilience dimensions, the market efficiency coefficient (MEC). Stocks in groups
1, 2 and 3 have declines in this liquidity measure after the event date. The MEC
decreases by 4.37%, 0.60% and 11.04%, respectively. On the other hand, the MEC
in group 4 increases by 6.79% after the tick size reduction. The results indicate that
the Wilcoxon tests do not prove specific evidence to reject the null hypothesis. It
suggests that the declines of the MEC are not statistically significant. The tick size
reduction does not enhance the market liquidity through resilience dimensions in the
post-event periods.

5 Conclusions

This study has explored the changes in liquidity dimensions on an emerging stock
exchange, following a tick size reduction. The main findings are as follows.
The study finds that market liquidity significantly decreases in the post-event of
the tick size reduction on 12 September 2016 on the HOSE. The Wilcoxon tests
examine the statistical hypothesis and indicate that the market liquidity is not
enhanced on the HOSE, except for the tightness proxies. The tick size reduction
positively affected liquidity related to the tightness dimensions because the transac-
tion costs are narrower. In contrast, the results suggested that the depth dimensions
100 Q.-K. Pham

are reduced following the smaller minimum tick size implementation. The study
does not find an improvement in the market resiliency during the post-periods. These
findings are consistent with Jones and Lipson (2001), Kuo et al. (2010), Pan et al.
(2012), Anderson and Peng (2014). The article proves empirical evidence for the
policymakers in Vietnam that tick size reduction has not enhanced the HOSE market
liquidity.
The study illustrates several contributions relating to the first change of tick size
on the HOSE. The article provides a comprehensive understanding of stock market
liquidity, particularly on an emerging South East Asia market. The market efficiency
coefficient, which has not been applied in the literature for the Vietnamese stock
market, is analyzed first in this study. Finally, the results shed light on the stock
market liquidity through the tick size reduction on the HOSE. Further research is
needed to find more direct evidence of tick-size reduction effects on the HOSE
market liquidity relating to the macroeconomy, stock returns and firm
characteristics.

References

Ahn HJ, Cai J, Chan K, Hamao Y (2007) Tick size change and liquidity provision on the Tokyo
Stock Exchange. J Jap Int Econ 21(2):173–194
Aitken M, Comerton-Forde C (2005) Do reductions in tick size influence liquidity? J Account
Financ 45(2):171–184
Albuquerque R, Song S, Yao C (2020) The price effects of liquidity shocks: a study of the SEC’s
tick size experiment. J Financ Econ. https://doi.org/10.1016/j.jfineco.2020.07.002. Accessed
20 September 2020
Amihud Y (2002) Illiquidity and stock returns: cross-section and time-series effects. J Financ Mark
5(1):31–56
Anderson HD, Peng Y (2014) From cents to half-cents and its liquidity impact. Pac Account Rev 26
(3):160–176
Ascioglu A, Comerton-Forde C, McInish TH (2010) An examination of minimum tick sizes on the
Tokyo Stock Exchange. Japan World Econ 22(1):40–48
Bacidore JM (1997) The impact of decimalization on market quality: an empirical investigation of
the Toronto Stock Exchange. J Financ Intermed 6(2):92–120
Chen HK, Hsieh WL (2013) The impacts of tick size reduction in a market with multiple tick sizes.
Available at SSRN 2629524
Chordia T, Roll R, Subrahmanyam A (2001) Market liquidity and trading activity. J Financ 56
(2):501–530
Chung KH, Lee AJ, Rösch D (2020) Tick size, liquidity for small and large orders, and price
informativeness: evidence from the Tick Size Pilot Program. J Financ Econ 136(3):879–899
Fong KYL, Holden CW, Trzcinka CA (2017) What are the best liquidity proxies for global
research? Rev Financ 21(4):1355–1401
Griffith TG, Roseman BS (2019) Making cents of tick sizes: the effect of the 2016 U.S. SEC tick
size pilot on limit order book liquidity. J Bank Financ 101:104–121
Harris LE (1991) Stock price clustering and discreteness. Rev Financ Stud 4(1991):389–415
Hasbrouck J, Schwartz, RA (1988) Liquidity and execution costs in equity markets. J Portf Manag
14(3):10
Ho Chi Minh Stock Exchange. www.hsx.vn. Accessed 20 June 2020
Tick Size Reduction and Liquidity Dimensions: Evidence from an Emerging Market 101

Hsieh TY, Chuang SS, Lin CC (2008) Impact of tick-size reduction on the market liquidity —
Evidence from the emerging order-driven market. Rev Pac Basin Financ Mark Policies 11
(04):591–616
Jones CM, Lipson ML (2001) Sixteenths: direct evidence on institutional execution costs. J Financ
Econ 59(2):253–278
Kuo SW, Huang CS, Chen CC (2010) Impact of the change in tick size on transaction costs and
liquidity: an empirical investigation of the Taiwan stock exchange. Asia Pac J Financ Stud 39
(4):524–551
Kyle AS (1985) Continuous auctions and insider trading. Econometrica 53(6):1315–1335
Le H, Gregoriou A (2020) How do you capture liquidity? A review of the literature on
low-frequency stock liquidity. J Econ Surv. https://doi.org/10.1111/joes.12385. Accessed
20 Sept 2020
Martens M, Van Dijk D (2007) Measuring volatility with the realized range. J Econ 138(1):181–207
Pan W, Song FM, Tao L (2012) The effects of a tick-size reduction on the liquidity in a pure limit
order market: evidence from Hong Kong. Appl Econ Lett 19(16):1639–1642
Parkinson M (1980) The extreme value method for estimating the variance of the rate of return. J
Bus 53(1):61–65
Pham-Quoc K (2020) Dimensions of stock market liquidity: empirical evidence of a frontier
market. Res Pap Wrocław Univ Econ 64(2):72–80
Porter DC, Weaver DG (1997) Tick size and market quality. Financ Manag 26(4):5–26
Rindi B, Werner IM (2017) U.S. Tick size pilot. Fisher College of Business Working Paper
No. 2017-18. https://doi.org/10.2139/ssrn.3041644. Accessed 20 Sept 2020
Sarr A, Lybek T (2002) Measuring liquidity in financial markets. IMF Working Paper. https://www.
imf.org/en/Publications/WP/Issues/2016/12/30/Measuring-Liquidity-in-Financial-Markets-
16211. Accessed 20 Sept 2020
State Securities Commission of Vietnam. https://www.ssc.gov.vn. Accessed 20 June 2020
Cryptocurrency Portfolio Construction
Using Machine Learning Models

Gopinath Ramkumar

1 Introduction

Time series forecasting is challenging. Unlike the simpler problems of classification

and regression, time series problems add the complexity of order or temporal
dependence between observations. Traditional timeseries linear methods like
Autoregressive integrated moving average (ARIMA), Autoregressive moving aver-
age (ARMA) and vector auto-regression (VAR) are popular and well understood.
However, these methods do suffer from limitations like completeness of data, focus
on linear relationship, fixed temporal dependence, univariate data and are one step
forecasts. Meanwhile, deep learning neural networks of convolutional neural net-
works (CNNs) and recurrent neural networks (RNNs) have been widely applied in
multi-step time series forecasting and are able to learn arbitrary complex mappings
from inputs to outputs and support multiple inputs and outputs. These methods are
attributed to the open source deep learning frameworks, such as keras1 and
Tensorflow,2 including flexible and sophisticated mathematical libraries.
Cryptocurrencies are one of the recent innovations in a long list of financial
market developments that were generally aimed at accumulating, concentrating, and
redistributing financial resources and risk. The market of cryptocurrencies is
undoubtedly dominated by Bitcoin, a decentralized digital currency. However,
alternative currencies (altcoins) among which Ethereum, Ripple, Litecoin are pop-
ular, boldly gaining attention as well as market shares. As on 25th Sep 2020, the

1
https://keras.io/.
2
https://www.tensorflow.org/.

G. Ramkumar (*)
Wroclaw University of Economics and Business, Wroclaw, Poland

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 103
K. Jajuga et al. (eds.), Contemporary Trends and Challenges in Finance, Springer
Proceedings in Business and Economics,
https://doi.org/10.1007/978-3-030-73667-5_7
104 G. Ramkumar

Fig. 1 Structure of the paper

global market cap for cryptocurrencies is $340.55 B.3 Because of the high percent-
age of daily traded volume of cryptocurrencies and control the risk better, machine
learning methods are considered as the most powerful tool to boost investment
return.
In general, investment portfolios should adhere to following stages: setting
investment goals, choice of investment portfolio, implementing a portfolio trading
strategy and evaluating the effectiveness of strategy. In this paper, nine
cryptocurrencies based on availability of data and market capitalization are identi-
fied. The cryptocurrency closing prices are forecasted using machine learning
methods in the likes of CNN and LSTM using four factors that will be discussed
Sect. 3.2. The forecasted methods with least RMSE (Root Mean Squared Error) are
taken as final model for prediction. Different investment portfolios are created using
various techniques and will be discussed in Sect. 3.3. The optimal investment
portfolio is formulated using portfolio performance measures discussed in Sect.
3.4. An empirical analysis using pair trading strategy is performed to confirm the
increase in profitability of the optimal investment portfolio.
The remainder of the paper is structured as follows. Section 2 summarizes related
work. Section 3 proposes relevant methodology of the forecasted methods, portfolio
construction methods and portfolio performance measures. Section 4 presents the
data set used and results. Section 5 concludes with a brief discussion. Figure 1 shows
the structure of the paper.

2 Related Work

Many research works have been published in terms of time series forecasting using
machine learning methods and cryptocurrency as an investment option is becoming
the hot topic in investment banks, hedge funds and brokerage houses. For example,
Vanstone et al. (2012) used a neural network to decide about the buying and selling
signal of the stock. The inputs are variables from the fundamental analysis: return on
equity (ROE), price-earnings ratio (PER), dividend payout ratio (DPR) and price
book-value ratio (PBR) and expected returns of the predicted stock served as the
output. Abe and Nakagawa (2020) predicted the cross-sectional daily stock prices in
Japanese stock market using deep learning for actual investment management. Wan

3
https://coinmarketcap.com/.
Cryptocurrency Portfolio Construction Using Machine Learning Models 105

et al. (2019) has used multivariate temporal convolutional network (M-TCN) model
to improve prediction accuracy and data dependence on aperiodic data for Beijing
PM2.5 and ISO-NE dataset. Bohte and Rossini (2019) has compared the forecasting
of cryptocurrencies by Bayesian time-varying volatility models. It has been shown
that the stochastic volatility is significantly outperforming the benchmark of VAR in
point and density forecasting. Ta et al. (2020) has proposed long short-term memory
to predict stock prices and constructed an efficient portfolio using multiple portfolio
optimization techniques including equal-weighted modeling (EQ), simulation
modeling monte carlo simulation (MCS) and mean variance optimization (MVO)
thus improving portfolio performance. Chen and He (2018) has proposed deep
learning method based on CNN to predict stock price movement of Chinese stock
market and concluded CNN is reliable for stock price prediction. Yang et al. (2020)
has proposed a deep learning framework to predict price movement direction based
on historical information in financial time series. The paper has combined CNN and
LSTM network for stock price prediction and concluded the CNN and LSTM model
outperforms state of art models in predicting stock price movement direction. On the
prior researches in cryptocurrency and portfolio management, Elendner et al. (2017)
found that top 10 cryptocurrencies by market capitalisation have low linear depen-
dency with traditional assets. Cheun et al. (2017) investigated performance of such
portfolio when adding CRIX. Other notable literature like Elendner et al. (2017)
introduced Liquidity bounded risk-return optimization (LIBRO) and considered
including a large sample of cryptocurrencies into a portfolio consisting of
S&P100, US Bonds and Commodities. Jiang and Liang (2017) has proposed
model less convolutional neural network can be effectively used with set of
cryptocurrency assets as its input, outputting portfolio weights of the set. Leung
and Nguyen (2019) has analysed the process of constructing cointegrated portfolios
of cryptocurrencies. Platanakis and Sutcliffe (2019) has compared the performance
of seven heuristics in forming a portfolio of six popular cryptocurrencies.
However, some of the studies did not justify the input choices and some studies
did not consider stressed period for the portfolio construction. The main idea behind
this paper is to construct an efficient portfolio with simple techniques using
forecasted data.

3 Problem Formulation and Methodology

In this section problem formulation is explained first and time series forecasting
techniques is explained second, portfolio construction methods are formulated next
and portfolio performance measures are explained last.
106 G. Ramkumar

3.1 Problem Formulation

Let us consider closing price of cryptocurrency i at day t represented as Ut and the

four technical factors in Table 2 referred in Sect. 4.1 by xi, tER4 serving as input
values. The variable dimension used in this project is 4. The output values are,
yi, t + 1, yi, t + 2. . . . yi, t + w 2 R, where N is number of days to train the dataset, K being
the size of training data and w being the length of forecasted output. The problem is
to find the predictor f and θT is the parameter calculated by solving the below Eq. (1).
In this project, root mean squared error (RMSE) is used as a loss function and we can
define RMSE when training the model at T as,
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 XT X  2
RMSE T ¼ yi,tþw  f ðxi,t ; θT Þ ð1Þ
K t¼TN i2U t

RMSE is a quadratic scoring rule that also measures the average magnitude of the
error. It’s the square root of the average of squared differences between prediction
and actual observation as seen in Eq. (1). Also, one of the salient features of RMSE is
sensitivity to outliers. The need to measure forecast accuracy using scale-dependent
measure like RMSE is to choose forecasted price from CNN or LSTM to create
different portfolios.

3.2 Forecasting Methods

CNN and LSTM are used as prediction models of the function f and ARIMA (p, q, d)
as a comparison model to measure if prediction model accuracy outperforms com-
parison model.
ARIMA Model
ARIMA methods are widely used approach to timeseries forecasting. It takes past
values of time series plus previous error terms containing information for the
purposes of forecasting. Though it is popular and one of the common approaches
used as reference testing more complex problems, it also has its own limitations such
as focus on missing data, linear relationships, fixed temporal dependence, univariate
data and one step forecasts. Makridakis et al. (2018) has argued ARIMA models
have better prediction accuracy compared to other machine learning methods.
Figure 2 illustrates the general ARIMA modelling and forecasting strategy and it
is self-explanatory.
An ARIMA model can be created using python library “statsmodels”4 by calling
ARIMA () function and passing p, q and d parameters. The training data is used to
prepare the model by calling fit () function. Predictions can be made by calling the

4
https://www.statsmodels.org/stable/index.html.
Cryptocurrency Portfolio Construction Using Machine Learning Models 107

Fig. 2 ARIMA forecasting procedure

predict () function. In this project, (p, q, d) order considered as (5,1,0) for forecasting
cryptocurrencies closing prices.
Convolutional Neural Network Model
Convolutional Neural Network (CNN) on other hand is extremely popular artificial
neural network technique initially developed for image recognition tasks. However,
they can be used to predict cryptocurrency prices after data has been pre-processed.
Chen et al. (2016) has used proposed planar feature representation methods and
CNN to improved algorithmic trading framework. Notable key benefits using CNN
is it uses fewer parameters to learn than a fully connected network and can auto-
matically learn and generalize features from the input domain. Figure 3 shows the
CNN feature extraction and classification. CNN has three types of layers including
convolution layer comprising of filters and feature maps, pooling layer and fully-
connected layer. Filters have both weighted inputs and generate output value like a
neuron. Feature map is the output of one filter applied to previous layer. Each
position results in an activation of neuron and output are collected in feature map.
Pooling layers may be considered as technique to generalize or compress feature
representations and reduce overfitting of the training data. Pooling layers takes the
average or maximum of input value to create its own feature map. Fully connected
layer is normal flat feedforward neural network layer and has non-linear activation
function to output probabilities of class predictions.
As the CNN expects the data to have shape of [samples, timesteps, features], the
shape of the training dataset used is [400, 100, 6]. Then we iterate over timesteps
such that each timestamp predicts next 100 ticks of data and then divide data into
overlapping windows. Because we have parametrized our inputs and outputs as
100 ticks, we can keep track of start and end indices. We then fit the model on
training data. The convolution layer has 10 filters and kernel size is 5 meaning the
input sequence will read with convolution operation 5 times steps at a time and the
operation is performed 10 times. Then pooling layer reduces feature maps and fully
connected layer interprets it before output layer predicts 100 ticks in sequence. The
efficient “Adam implementation”5 is used and fit model with 20 epochs with batch
size of 4. We are using walk forward validation for predicting the 100 ticks meaning
we have prior 100 ticks of data.

5
https://machinelearningmastery.com/adam-optimization-algorithm-for-deep-learning/.
108 G. Ramkumar

Input Layer Convolution 1 Sampling 1 Convolution 2 Sampling 2

(Feature maps) Fully Connected
(Feature maps)

Input Data Output

Fig. 3 CNN feature extraction and classification

Fig. 4 LSTM memory cell illustration

Long Short-Term Memory Model

Long short-term memory (LSTM) is special kind of RNN capable of learning long
term dependencies. One of the default behaviours of LSTM is remembering the
information for long period of time. The LSTM network has an input layer, a hidden
layer including memory cells and an output layer. Each of the memory cells has three
gates for maintaining and adjusting its cell state st: a forgot gate ( ft) to decide the
fraction of information to be allowed, an input gate (it) consisting the input and
output gate (ot) consisting the output generated by LSTM. Figure 4 illustrates
computation carried out in LSTM memory cell.
LSTM model maps a sequence of past observations as inputs to an output
observation. split_sequence () function is used to split the input series to output
samples where each sample will have input timesteps and output timestep. In this
project, we have used the efficient “Adam implementation” and fit model with
20 epochs with batch size of 32. For the timesteps, hidden neurons used is
100 and dropout ratio is 0.2.
From the ARIMA, CNN and LSTM methods, forecasted prices of different
cryptocurrency assets of choice with least average RMSE scores is considered for
portfolio construction.
Cryptocurrency Portfolio Construction Using Machine Learning Models 109

3.3 Portfolio Construction Methods

Portfolio construction refers to process of selecting the optimum mix of assets for the
purposes of achieving maximum returns by minimizing risk. Mean variance
approach has been standard approach in portfolio construction. Despite its rationality
and theoretical appeal, it does not hold well in practice (DeMiguel et al. 2009;
Broadie 1993). Michaud (1989) has also referred mean-variance optimization
approach as “Error Maximization” procedure as it is shown a small change in
expected return assumptions lead to different efficient portfolios. Hence in this
project apart from constructing a minimum variance portfolio and maximum sharpe
ratio portfolio, additional portfolios are constructed by equal weight method, risk
parity, kelly criteria, cointegrated pairs and apply portfolio performance measures to
validate the results.
Equal weight method is one of naive portfolio methods where investors allocate
capital and every asset has weight w ¼ 1/N where N being number of assets.
DeMiguel et al. (2009) suggests portfolio manager is not required to make assump-
tions on the distribution of the assets returns. These equal weighted portfolios are
widely used in practice (Benartzi and Thaler 2001; Windcliff and Boyle 2004) and
have shown promising out-of-sample results (DeMiguel et al. 2009).
MPT developed by Harry Markowitz and published under the title “Portfolio
Selection” in the Journal of Finance in 1952. As per mean-variance optimization
concept, a portfolio is constructedP by means of vector of weights w ¼ (w1, w2, . . .
wN), with the constraint given Ni¼1 wi ¼ 1. With N dimensional vector denoted as I,
the constraint can be written as wTI ¼ 1. Let the random returns of crypto assets
denoted as r1, r2, . . .rN and the vector of expected return denoted as μ ¼ (μ1, μ2, . . .μN)
with μi ¼ E(ri) for i ¼ 1, 2. . .N. The covariances are denoted by σ ii ¼ σ 2i ¼ Var ðr i Þ.
The N  N covariance matrix ϕ.
Expected returnof portfolio
 is given by μp ¼ E(Rp) and variance of portfolio is
given by σ 2p ¼ Var Rp .
XN
μp ¼ i¼1
wi μi ¼ wT μ ð2Þ
  XN
σ 2p ¼ Var Rp ¼ w w σ ¼ wT ϕ w
i,j¼1 i j ij
ð3Þ

Then classical mean-variance problem can be written as,

Minimize ðwÞ ¼ wT ϕ w, wT μ ¼ r target , wT I ¼ 1

The choice of minimum variance portfolio is it can be easily derived from mean-
variance optimization approach (Jagannathan and Ma 2003) and its property of not
requiring information on expected returns makes it easy to compute. Behr et al.
(2008) find that minimum variance portfolio outperforms higher returns and better
110 G. Ramkumar

risk adjusted. In this project, we use scipy ‘minimize’ function6 to calculate portfolio
choice with maximum sharpe ratio and minimum volatility.
Kelly Criteria is a formula for bet sizing that has been very successful for trading
in long run. The Kelly bet size is found by maximising the expected logarithm of
wealth which is equivalent to maximising the expected geometric growth rate. In this
project, the daily percentage change in a portfolio’s value can be calculated by
multiplying asset weight that is percentage capital allocation and forecasted crypto
asset rate of return that is percentage change of price. Once the daily percentage
change is calculated, final portfolio value is then calculated by compounding the
daily values. While building a portfolio, the past daily returns of each crypto asset in
the portfolio is known but the weights are unknown and need to be optimised. To
simplify, we can take the logarithm of the final portfolio value and then optimize
it. Maximise (A *B) ¼ Maximise (Log (A *B)) ¼ Maximise (Log(A) + Log (B))
While optimising for the best weights of crypto assets in a portfolio, we use the
logarithmic sum of returns. Since the maximisation of the logarithm of portfolio
returns would give the same results as the maximisation of compounded returns, one
will be using the sum of logarithmic returns to solve for the best weight combination.
Hence Kelly criteria can be shown as,
" !#
X X
j
log 1 þ wi  r i ð4Þ
i

∑ is the sum of the logarithm of the daily portfolio values; w is asset weights; r is
forecasted crypto asset rate of return.
Risk parity is an investment management approach coined by Edward Qian of
PanAgora Asset management.7 It is a conceptual approach to investing which
attempts to provide a lower risk and lower fee alternative to the traditional portfolio
allocation. The main goal of risk parity portfolio is no asset contribute more to total
risk of portfolio than any other asset. Maillard et al. (2010) and Qian (2005) have
extensively investigated the theoretical foundations in their literature. Portfolio
weights for a two-asset portfolio used in this project is calculated as,

1
σ1
W1 ¼ ð5Þ
1
σ1 þ σ12

σ is the standard deviation of asset return.

Mean-reversion trading strategies are widely researched by practitioners to under-
stand long term co-movement of different asset prices and arbitrage from mean-
reversion property of price spread. Theoretical framework for constructing
cointegrated pairs are created from the forecasted crypto asset dataset using Engle-

6
https://docs.scipy.org/doc/scipy/reference/generated/scipy.optimize.minimize.html.
7
https://www.panagora.com/members/edward-qian-ph-d/.
Cryptocurrency Portfolio Construction Using Machine Learning Models 111

Table 1 Portfolio perfor- Portfolio performance measures Formula used

mance measures pffiffiffiffiffiffiffiffipffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Annualized volatility 252 variance
Sharpe ratio Rp R f
σp
Sortino ratio Rp R f
σd
Beta CovðRp , Rm Þ
βp ¼ varðRm Þ
Rp R f
Treynor ratio Tp ¼ β
p
Rp Rm
Information ratio Ip ¼ σ p,m
Maximum drawdown MD ¼ LP
P
Rp,Rf, σ p, σ d, Rm, σ p, m, L, P represents Portfolio returns, Risk free
returns, standard deviation of portfolio returns, standard deviation
of negative asset returns, market returns, Standard deviation of the
difference between portfolio and market returns, Lowest value
before the new high and Maximum value before the largest drop

Granger test. As per Engle-Granger test, we can calculate the spread by using linear
regression to get the coefficient for linear combination between the cointegrated pair.
Alternatively, we can also use ratio of the cointegrated pair. In other words, we
construct the spread in such way that yields highest profit when trading the mean-
reverting crypto-portfolio.

3.4 Portfolio Performance Measures

Portfolio analysis is the guide to investors to study certain portfolio regarding its
performance. The main objective of portfolio performance is to minimize the risk
and maximize the returns. Some of the performance analysis measures used in this
project is provided in Table 1.

4 Dataset and Results

This section covers the dataset used first and results on forecasting methods using
CNN and LSTM. Then different portfolios are constructed using techniques in the
likes of equal weighted portfolio, minimum variance, maximum sharpe ratio,
cointegrated pairs, Kelly criterion and risk parity are elaborated next. The results
are discussed using portfolio performance measures.

4.1 Dataset

In general, tick by tick data includes every changed, added or removed bid and ask to
an order book thus helping end users to reconstruct the market states at any given
112 G. Ramkumar

time. Tick by tick data for this project is sourced from binance exchange for the
month of March 2020. The choice of selecting tick by tick data for the month of
March is because cryptocurrency markets suddenly collapsed with bitcoin prices
getting halved in less than a day on 12th March as shown in Fig. 5. This phenomenon
dubbed as “Black Thursday” raised several questions on valuation mechanisms as
well as absence of circuit breakers in financial markets, although distributed denial-
of-services (DDoS) that brought BitMEX down (BitMEX 2020) acted as an implicit
circuit breaker. The base assets used are binancecoin, bitcoin, bitcoincash, chainlink,
EOS, ETH, Litecoin, MCO and XRP and quotation is in USD. The choice of
cryptocurrencies is purely based on availability of data and average transaction
volume increase including crash period. Only 10% snapshot is considered for this
project meaning bids/asks placed within 10% of the mid-price at the time the order
book snapshot was taken was considered and aggregated for 1 min.8 It is well known
that technical indicators are heuristic signals produced from price, volume and/or
open interest and is widely popular in analyzing future prize movements. Table 2
explains selected technical factors9 and deterministic trend signals calculated from
the exchange data.

4.2 Results

Table 3 explains average RMSE scores for different cryptocurrencies using CNN
and LSTM compared with ARIMA.
As per Table 3, we can see the forecasting techniques have achieved better results
and very low RMSE scores implies the accuracy and precision of the prediction is
high. The choice of epochs used in CNN and LSTM is just to ensure the curve is
close to optimal (not overfitting) and reduces the error. It is also worth noticing that
CNN with fewer batch size has outperformed LSTM in few cryptocurrencies. The
overall performance of neural network is better than traditional methods like
ARIMA, the results of ARIMA are low and it is self sufficient to forecasting problem
for this project. Because of the close results from three machine learning techniques,
we can take CNN forecasted data for portfolio construction analysis. Figure 6
represent equal weighted portfolio return with each cryptocurrencies assigned a
weight ¼1/9.
As per histogram in Fig. 6, we can see the average returns is hovering around
0. For this project let us assume market returns for comparing other portfolios.
Figure 7 explains the calculated portfolio optimization based on efficient frontier.
In Scipy10 optimise funtion, there is no ‘maximize’, so objective function used in this
project is to minimize the “negative Sharpe ratio”.

8
The data was sourced from kaiko and aggregated 1 min bucket with 10% snapshot of different
cryptocurrencies are calculated by data provider.
9
https://school.stockcharts.com/doku.php.
10
https://www.scipy.org/.
Cryptocurrency Portfolio Construction Using Machine Learning Models 113

Fig. 5 Bitcoin price volume chart for month of March 2020

114 G. Ramkumar

Table 2 Technical factors with deterministic trend signal

Technical factors Formulas Trend signals
Simple moving average SMAt ¼ Ct + Ct  1 + ⋯ + Ct  n/n If Ct, signal “1”; other-
(SMA) wise signal “0”.
Commodity channel TPt ¼ Ht + Lt + Ct/3 If CCIt, signal “1”; other-
Index (CCI) CCI t ¼ TPt SMAnðTPt Þ wise signal “0”.
Pt 
j¼tnþ1
TPt SMAn ðTPt Þj
0:015 n

Momentum (M) Mt ¼ Ct + Ct  n If Mt, signal “1”; other-

wise signal “0”
Exponential moving EMAt ¼ EMAt  1 + α  (xt  EMAt  1) If Ct, signal “1”; other-
average (EMA) 2
α = nþ1 wise signal “0”.
Ht, Lt, Ct being the High, Low and close price. xt is past value. n is the period for which calculations
are done

Table 3 Average RMSE Crypto asset ARIMA CNN LSTM

scores
Binance coin 0.416 0.175 0.028
Bitcoin 26.923 47.12 12.764
Bitcoin cash 2.248 1.941 0.033
Chainllink 0.192 0.018 0.041
EOS 0.411 0.007 0.031
ETH 0.96 0.857 0.029
Litecoln 1.241 0.55 6.229
MCO 0.114 0.104 0.05
XRP 0.043 0.001 0.029

Equal weighted portfolio returns

10

8
freq

0
–0.0010 –0.0005 0.0000 0.0005 0.0010
portfolio returns

Fig. 6 Equal weighted portfolio return

Cryptocurrency Portfolio Construction Using Machine Learning Models 115

Fig. 7 Portfolio optimization based on efficient frontier

Table 4 Portfolio 1 with Cryptocurrencies Max Sharpe Portfolio (PORTFOLIO 1)

associated weights
Binance Coin 0
Bitcoin 0
Bitcoin Cash 0.005
Chainlink 0
EOS 0.315
ETH 0
Litecoin 0.201
MCO 0
XRP 0.477

As per the above Fig. 7, annualized return is slightly over 0 for maximum sharpe
ratio portfolio allocation and slightly less than 0 for minimum volatility portfolio
allocation. Tables 4 and 5 are the maximum Sharpe ratio portfolio labelled portfolio
1 with allocated weights and minimum volatility portfolio allocation with allocated
weights labelled portfolio 2.
Figure 8 explains three cointegrated pairs of cryptocurrencies with pvalue less
than 0.5. Lower pvalues mean high cointegration and Engle-Granger test is used to
check cointegrated timeseries. Portfolio 3, portfolio 4 and portfolio 5 is created in the
likes of [binance coin, EOS], [binance coin, litecoin] and [EOS, litecoin] with equal
weights allocated for each cryptocurrencies in pairs.
Figure 9 explains the portfolio optimization based on Kelly criterion. As seen in
Fig. 9 it is evident Kelly criterion has outperformed the equal weighted portfolio.
116 G. Ramkumar

Table 5 Portfolio 2 with Cryptocurrencies Min Vol Portfolio (PORTFOLIO 2)

associated weights
Binance coin 0.364
Bitcoin 0.142
Bitcoin cash 0.001
Chainlink 0.002
EOS 0.006
ETH 0.005
Litecoin 0.004
MCO 0.002
XRP 0.267

Fig. 8 Cointegrated pairs of cryptocurrencies

Thus portfolio 6 is created using kelly criterion and the weights are allocated using
“cvxpy” function11 in Python.
Portfolio 7 and portfolio 8 are formed by selecting random cryptoassets and
portfolio weights are allocated using risk parity approach. Figures 10 and 11
explains portfolio 7 in the likes of [XRP, bitcoin cash] and portfolio 8 in the likes
of [ETH, litecoin].

11
https://www.cvxpy.org/install/.
Cryptocurrency Portfolio Construction Using Machine Learning Models 117

Portfolio Optimization based on Kelly Criterion

Kelly Portfolio Performance
Equal Weight Portfolio Performance
100.4

100.2
Portfolio returns

100.0

99.8

99.6

99.4

0 25 50 75 100 125 150 175 200

Ticks

Fig. 9 Portfolio performance (portfolio 6 vs equal weighted portfolio)

Fig. 10 Portfolio 7 with associated weights

Fig. 11 Portfolio 8 with associated weights

Portfolio performance measures are then used in the eight portfolios as shown in
Fig. 12. The methodology is mentioned in Sect. 3.3.
As seen in Fig. 12, we can see the portfolio 5 and portfolio 6 has highest
annualised returns. Portfolio 5 has higher Sharpe ratio than other portfolio as Sharpe
ratio tells whether the returns on a portfolio are due to good investment decision or
the result of excessive risk taken. Sortino ratio determines an investment’s risk-
adjusted returns as it relates to downside risk. Portfolio 5 has highest Sortino ratio
implying higher returns per unit of downside risk. Beta captures the relationship
between the benchmark returns and the portfolio returns. Treynor Ratio is the
variation in the denominator of the Sharpe ratio that tells investors how good the
118 G. Ramkumar

Fig. 12 Portfolio performance measures on all eight portfolios

Running Maximum Drawdown (PORTFOLIO 5)

0.000

–0.001
Maximum Drawdown

–0.002

–0.003

–0.004

Running Maximum Markdown

Fig. 13 Running maximum drawdown for portfolio 5

Fig. 14 Pair trading strategy

investment though it does not quantify how much good the investment. Information
ratio tells the portfolio’s return in excess of the benchmark’s return with respect to
the volatility of these returns. Higher information ratio implying consistency and
better performance. A positively skewed investment in the portfolio indicates fre-
quent small losses and few large gains and vice versa. Kurtosis, like skewness, is a
measure of distribution. kurtosis tells about the heaviness in the tails while skewness
talks about the symmetry. Maximum Drawdown measures the peak-to-trough
decline in the value of the portfolio and is quoted as the percentage of the peak
Cryptocurrency Portfolio Construction Using Machine Learning Models 119

Fig. 15 Rolling ratio z score

120 G. Ramkumar

Ratio
Buy Signal
Sell Signal

0 20 40 60 80 100 120

Fig. 16 Buy and sell signal on ratio

value. Maximum Drawdown doesn’t say how frequently the losses are occurring and
how much time it took to recover from those losses. It only measures the size of the
largest loss. A low value of maximum drawdown is preferred. As seen in Fig. 12,
almost all the portfolios have negative drawdown. Figure 13 shows the running
maximum drawdown for portfolio 5. Overall, Portfolio 5 has comparatively fared
better than other portfolios.
As portfolio 5 is created using cointegrated pairs, it is possible to perform a pair
trading strategy to maximise the returns. Pair trading is a high alpha strategy that has
distinct advantage being hedged against market movements. It is a form of mean-
reversion and based on mathematical analysis. Figure 14 illustrates the pair trading
strategy. Once the spread or ratio of the cointegrated pair is calculated, Z score is
calculated to standardize the ratio.

xi  x
Zi ¼ ð6Þ
s

A z-score is the number of standard deviations a datapoint is from the mean. More
importantly, the number of standard deviations above or below the population mean
is from the raw score.
Trading signals are indicators to trading strategy to buy or sell assets. In this
project, once the z score is calculated, we will create “buy” signal on calculated ratio
when z score is below 0.5 and “sell” signal when z score is above 0.5. The training
and testing data is split 60/40. Additional features like 3 day moving average, 7 day
moving average and z score is added to determine the direction of z-score move-
ment. Then returns are calculated for all cointegrated cryptocurrency pairs. Fig-
ures 15 and 16 represent Rolling ratio Z score and buy and sell signal on the ratio.
As seen in Fig. 15, if the timeseries moves beyond 0.5 standard deviation beyond
the mean, it tends to revert back and Fig. 16 shows the trading signals on the ratio
Cryptocurrency Portfolio Construction Using Machine Learning Models 121

over the timeseries.The annual returns for this portfolio has improved to 1.309 when
using pair trading strategy for portfolio 5 implies that any right trading strategy with
trading signal can maximise the return. Thus the paper will be extended in the future
to include multiple trading strategies.

5 Conclusion

This paper presented deep learning neural network models like CNN and LSTM to
predict cryptocurrency prices. Though, the overall performance of neural network
models are better than traditional methods like ARIMA, the results of ARIMA are
comparitively low and it is sufficient. Different portfolios are created using equal
weighted portfolio, modern portfolio theory, cointegrated pairs, Kelly criterion and
risk parity. Portfolio performance measures like annualized returns, annualized
volatility, Sharpe ratio, Sortino ratio, beta, treynor ratio, information ratio and
maximum drawdown for all the portfolios are analysed and best portfolio is selected.
To maximise the return, a high alpha strategy like pair trading is used.
In the future work, it will be a challenge to consider proprietary factors from order
book analytics to forecast the data and implement different trading strategies and use
sophisticated portfolio optimization techniques to maximise the return.

References

Abe M, Nakagawa K (2020) Cross-sectional price prediction using deep learning for actual
investment management
Behr P, Güttler A, Miebs F (2008) Is minimum-variance investing really worth the while? An
analysis with robust performance inference. EDHEC-Risk working paper
Benartzi S, Thaler RH (2001) Naive diversification strategies in defined contribution saving plans.
Am Econ Rev 91(1):79–98
BitMEX (2020) How we are responding to last week’s DDoS attacks. https://blog.bitmex.com/
how-we-are-responding-to-last-weeks-ddos-attacks/
Bohte R, Rossini L (2019) Comparing the forecasting of cryptocurrencies by Bayesian time-varying
volatility models. J Risk Financ Manag
Broadie M (1993) Computing efficient frontiers using estimated parameters. Ann Oper Res 45
(1):21–58
Chen S, He H (2018) Stock prediction using convolutional neural network. IOP Conference Series:
Material science and engineering 435, University of Shanghai
Chen JF, Chen WL, Huang CP, Huang SH, Chen AP (2016) Financial time-series data analysis
using deep convolutional neural networks. In: 2016 7th International Conference on Cloud
Computing and Big Data (CCBD), Macau, pp 87–92
Cheun et al (2017) Handbook of digital finance and financial inclusion: cryptocurrency, FinTech,
InsurTech, Regulation, ChinaTech, Mobile Security, and Distributed Ledger, 1st edn
DeMiguel V, Garlappi L, Uppal R (2009) Optimal versus naive diversification: How inefficient is
the 1/N portfolio strategy? Rev Financ Stud 22(5):1915–1953
122 G. Ramkumar

Elendner H, Trimborn S, Ong B, Lee TM (2017) The cross-section of cryptocurrencies as financial

assets. In: Lee Kuo Chen D, Deng R (eds) Handbook of digital finance and financial inclusion:
cryptocurrency, FinTech, InsurTech, Regulation, ChinaTech, Mobile Security, and Distributed
Ledger, 1st edn
Jagannathan R, Ma T (2003) Risk reduction in large portfolios: why imposing the wrong constraints
helps. J Financ 58(4):1651–1683
Jiang Z, Liang J (2017) Cryptocurrency portfolio management with deep reinforcement learning.
IntelliSys. https://doi.org/10.1109/IntelliSys.2017.8324237
Leung T, Nguyen H (2019) Constructing cointegrated cryptocurrency portfolios for statistical
arbitrage. Applied Mathematics Department, University of Washington
Maillard S, Roncalli T, Teïletche J (2010) The properties of equally weighted risk contribution
portfolios. J Portfolio Manag 36(4):60–70
Makridakis S, Spiliotis E, Assimakopoulos V (2018) Statistical and machine learning forecasting
methods: concerns and ways forward. PLoS One 13(3):e0194889
Michaud RO (1989) The Markowitz optimization enigma: is ‘optimized’ optimal? Financ Analysts
J 45(1):31–42
Platanakis E, Sutcliffe C (2019) Cryptocurrency portfolios using heuristics. School of Management,
University of Bath
Qian E (2005) Risk parity portfolios: efficient portfolios through true diversification. Panagora
Asset Management
Ta V, Liu C, Tadesse D (2020) Portfolio optimization-based stock prediction using long-short term
memory network in quantitative trading. MDPI J Appl Sci
Vanstone BJ, Hahn T, Finnie G (2012) Developing high-frequency foreign exchange trading
systems. In: 25th Australasian Finance and Banking Conference
Wan R, Mei S, Wang J, Liu M, Yang F (2019) Multivariate Temporal Convolutional Network: a
deep neural networks approach for multivariate time series forecasting. MDPI J Electron
Windcliff H, Boyle PP (2004) The 1/n pension investment puzzle. N Am Actuarial J 8(3):32–45
Yang C, Zhai J, Tao G (2020) Deep Learning for price movement prediction using convolutional
neural network and Long short-term memory. Article ID 2746845. Mathematical Problems in
Engineering, Hindawi
 Part II
Banking
Development Factors of Blockchain
Technology Within Banking Sector

Monika Kołodziej

1 Introduction

Blockchain is currently considered as disruptive technological concept with huge

implementation potential and numerous of applications. Especially for banking and
global financial ecosystem, blockchain can be treated as a promise of safety, trust,
stability, transparency, immaturity, availability, cost reduction and efficiency. How-
ever, in the global environment exists a lack of understanding where and how
technology is applicable and where it can produce practical and measurable effects.
Over 10 years ago, when Satoshi Nakamoto presented White Paper with
blockchain’s thesis, the technology is still in growth phase. Although blockchain
technology is considered to be potential driver within digital economy, applications
built on blockchain are still not commercially available (Axios 2018). To complete
process of adoption, implementation and diffusion it is require rebuilding business
models and rules, create governance architecture, identify the proper blockchain
application sectors that adds value, prepare methods to describe the technology and
educate new human resources.
Currently blockchain is receiving a lot of public attention as advocates argue that
it constitutes the foundation for truly trust-fee economic transactions based on its
unique technological characteristics (Glaser 2017). Blockchain technology can be
applied to a certain application scenario provided the scenario has one of the
following properties: multiparty interaction, creditability, disintermediation,
atomicity and privacy (Zheng et al. 2017). Much of the attention on blockchain
technology is focused on its ability to redefine currently meaning of banking
industry. As per study described by Cocco, Pina and Marchesi blockchain possesses

M. Kołodziej (*)
Wroclaw University of Economics and Business, Wroclaw, Poland
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 125
K. Jajuga et al. (eds.), Contemporary Trends and Challenges in Finance, Springer
Proceedings in Business and Economics,
https://doi.org/10.1007/978-3-030-73667-5_8
126 M. Kołodziej

a potential to create value to several financial service activities, from payments to

compliance. They also highlighted a role of blockchain in overcoming some tradi-
tional banking inefficiencies (Cocco et al. 2017). Nevertheless, each organization
taking the patch to blockchain adoption must confront the business need, require
resources, time range (scope), risk and potential benefits.1 Implementation potential
of blockchain technology for banking sector can be outline by establish a credit
mechanism in a situation where there is a lack of mutual trust among parties, thereby
resolving the high costs caused by the non-technical aspects of centralization (Guo
and Liang 2016). A recent World Economic Forum report showed that over 40 cen-
tral banks are researching distributed ledger technology for a variety of use-cases
(WEF 2019).
The article shows the flow of blockchain implementation process dedicated for
banking sector. In this paper has been prepared literature review with focus on the
newest publications and case study. The main aim of the article is to outline financial
determinates of blockchain technology development within banking industry and to
clarify meaning of blockchain technology implementation for banking sector.

2 Blockchain Beyond Solution to Support Cryptocurrencies

Blockchain is emerging technology with huge implementation potential. However,

expectations without researches and tests can caused a risk of failure in implemen-
tation process. Every disruptive technology (like blockchain) makes first impact in
this area of industry that stands at the leading edge of adoption. Only particular
group of companies and customers understand that technology’s long-term value
and the ways it might upend current thinking. Based on Deloitte’s 2019 Global
Blockchain Survey, blockchain is going through a path of diffusion far beyond
cryptocurrencies and initial fintech application (Deloitte 2019). In the moment of
article’s preparation, the applications are modest so far. Reach further than
cryptocurrencies (Eyal 2017), there are many investments in blockchain within
banking sector (including Initial Coin Offerings), supply chain documentation
(including logistic), asset registries (land ownership), health care (medical records),
proof of origin (diamonds market), smart contracts and build on blockchain basis
platform (Ethereum). Second generation of blockchain technology allows the carry-
ing out of computation of network and includes third party data ledgers (Peters and
Panayi 2016). Blockchain should be treated as a transformative technology, which
conform trust as the crucial lubricant to economic transactions and empower an era

1
Based on Accenture’s “Building Value with Blockchain” survey, more than 64% of blockchain
projects are being funded by IT or research/innovation budgets—implying that the focus is on
technology, rather than on aligning with the main areas of opportunity for the organization. For
further information: http://www3.weforum.org/docs/WEF_Building_Value_with_Blockchain.pdf
(access: 22.06.2020).
Development Factors of Blockchain Technology Within Banking Sector 127

14 13.2
12.1
12 Services 65%
10.2 40%
10
7.8 Applications 30%
8 10%
5.6
6 30%
Infrastructure 25%
4 2.9
2 0% 20% 40% 60% 80%
0
2019 2020 2021 2022 2023 2024 2020 2019

Fig. 1 Blockchain market growth, 2019–2024 and market share, 2019–2014. Source: Pelz-Sharpe
(2019)

of decentralized and independent business models (Välikangas 2020). Figure 1

represents market scope for blockchain technology for years 2019–2024.

2.1 Genesis of Blockchain Technology

and Development Path

To outline blockchain’s genesis, it must be emphasized that blockchain is not

innovation, rather combination of well-known technological concepts with innova-
tive character. A blockchain era had begun in 2008 with financial crisis. Most
probably the moment of publication blockchain thesis was an attempt to refute
financial systems rules. Global financial crisis exposed instability and infectivity of
financial sector. As a key issue related to financial crisis should be provided the
asymmetry of information that characterized the cooperation line client/investor and
financial industry. While crisis circumstances, many criticisms have formulated
against the existing financial practices leading to a call for more transparency and
higher inclusion of investors into banking or financial ecosystem (Schinckus 2020).
In this context blockchain technology appears to cover requirements claimed by
financial market. The technology should outweigh challenges associated with tradi-
tional meaning of business models and processes with third party related
to. Nevertheless, blockchain is only technological concept, a tool dedicated to
improving, support and probably (in the future) to replace currently existing business
models. Blockchain technology is the resultant of many IT concepts. It has been built
on the basis of Merkle tress, blind signatures, consensus protocol, Proof-of-Work
mechanism, e-cash systems, peer-to-peer networks and cryptography rules to main-
tain safety. The core ideas of blockchain have been developed in the late 1980s and
early 1990s (Drescher 2018). The growth in the interest among researches and tests
128 M. Kołodziej

DEVELOPMENT PHASES
2013 Ideas and concepts

2014 Researches and development

2015 Prototypes

2016 Further development

2017 Tests

2018 Commercial solutions

2019 Evaluation and regulations

2020 Production phase

2021 Implementation in daily business

2022-2025 Further development

2026 Maturity or further development

Fig. 2 Blockchain technology maturity flow. Source: https://www.researchgate.net/figure/
Blockchain-development-and-utilization-timeline-with-power-system-concertation_fig3_
333627407 (accessed: 20.12.2020); https://medium.com/thundercore/thundercore-blockchain-
roadmap-aaddb8b3d2ca (accessed: 20.12.2020)

has been noticed since 2013, when stakeholders (industries, banks, financial insti-
tution etc.) understood implementation potential of blockchain technology. The flow
of blockchain technology development has been presented on Fig. 2.
Around implementation of blockchain technology into banking services have
been created many problems. First of all, risk which has been compound by
immaturity of this disruptive technology and instability related to fluctuations of
Bitcoin’s value. Then many inconsistencies caused by lack of knowledge, clarity and
relevant governments rules (Hassani et al. 2018). This caused many difficulties in
identifying and developing applications dedicated for banking industry. Moreover,
the way in which blockchain technology works in short- and long-term horizon is not
clear. From banking sector perspective security and privacy are crucial. In general,
blockchain technology delivers safety due to cryptography and real-time stamping.
However, security risk is significant when miners (users) control more than 51% of
the computing power. Counterproductive issue is immutability, which not allows
remove data upon client’s request (Upadhyay 2020). The last important thing related
to blockchain implementation in banking industry is capacity and performance. On
the other hand, blockchain promises following spectrum of benefits: cutting
timeframes in financing, cost reduction, increasing processes efficiency, simplifica-
tions of services. As it has been mentioned by Guo and Liang, banking industry
needs urgent transformation and is seeking new growth path (Guo and Liang 2016).
Development Factors of Blockchain Technology Within Banking Sector 129

2.2 Principles of Blockchain Technology

Blockchain technology can introduce really disruptions to the currently existing

models due to decentralized way of processing with the same level of certainly. This
technology can be treated as a huge, distributed database that is organized as a list of
ordered blocks, where the committed blocks are immutable. Some researchers
consider blockchain technology as ideal solution in the banking industry because
banks can cooperate under the same blockchain and push their customers’ trans-
actions. On this way, beyond transparency blockchain can facilitate transactions’
auditing (Casino et al. 2018).
In general, a blockchain technology is a distributed database of records or a public
ledger of all transactions or digital events that have been processed and shared
among authorized users. Each transaction in the public ledger have to be verified
by consensus mechanism resulted by the majority of users. Data inputted into the
ledger cannot be modified (Crosby 2015). Blockchain is a peer-to-peer (P2P),
distributed data structure which allows transactions to be recorded chronologically
and stored securely in a sequence or chain of blocks via cryptography rules (Li et al.
2018). Due to combination of P2P network and the distributed nature of server that
marks the timeshare transactions, has been prepared a database that is autonomous
and shared among all network members (Knezevic 2018). Blockchain puts focus
around distribution, sharing and encryption (Guimarães et al. 2020). Blockchain is
an encrypted digital ledger stored in a private or public network on numerous
computers. A structure of blockchains includes nodes located on those networks
that use a common communication protocol. Each node in the network contains a
complete copy of the transactions saved in chain. A consensus mechanism is used to
validate transactions to ensure the immutability of the chain (Bashir 2018). Nodes
within blockchain contain a copy of encrypted data blocks (records) chained by hash
codes to each other (Swan 2015). Blockchain technology can be seen as a variant of
distributed ledger technology (DLT)2 where the data are presented in a linear chain
of blocks that are cryptographically linked to make them resilient against
unintentional or malicious manipulation (Huth et al. 2020).
From business perspective, blockchain technology can be described as a platform
which allows value and money transfer using transactions and any central trust party
is needed. For instance, by money transfer, third party (intermediary) is bank.
Elimination of intermediaries makes that blockchain can be decentralized

2
Distributed ledger technology (DLT) essentially can be described as a database which works
across several locations or among multiple participants. An advantage of DLT means an elimination
of trusted third party from transaction ecosystem. DLT use independent computers to build network
in which transactions are recorded, shared and synchronized in their respective electronic ledgers.
Sometimes DLT an blockchain are treated synonymously, however blockchain is only a type of
DLT. An impact of DLT for financial sector has been described by The World Bank in: Blockchain
& Distributed Ledger Technology (DLT), 2018 available in: https://www.worldbank.org/en/topic/
financialsector/brief/blockchain-dlt (access: 02.07.2020).
130 M. Kołodziej

Table 1 Elements of blockchain

Name Description
Block A selection of transactions bundled together and organized logically. Blocks
are made from transactions. Each block contains a copy of database. First
block in the blockchain is called a genesis block.
Transaction A recorded event (for example: transferring cash from sender’s account to a
beneficiary’s account). It is a representation of value transfer from one address
to another one.
Nonce A unique number generated only once. Nonce is used extensively in many
cryptographic operations to provide replay protection, authentication and
encryption.
Merkle root A hash of all of the nodes of a Merkle tree. Merkle trees are used to validate
the large structures of data securely and efficiently. A main aim of the Merkle
root is efficient verification of transactions in a block.
Peer-to-peer A network topology means that all peers can communicate with each other
network and send or receive information.
Node A node is responsible for proposing and validation of transactions and for
performing mining to facilitate consensus and secure the blockchain. A way to
achieve this goal is a consensus protocol. Transactions are first created by
nodes and digitally signed using private keys to proof the legitimation of
owner of the asset transferring to another blockchain user. This asset is usually
token or virtual currency.
Source: https://users.cs.fiu.edu/~prabakar/cen5079/Common/textbooks/Mastering_Blockchain_
2nd_Edition.pdf (access: 03.07.2020)

mechanism ruled by consensus where no one is responsible for database (Bashir

2018). Key elements of blockchain has been presented in the Table 1.

2.3 Advantages of Blockchain Technology Crucial

for Banking Sector

Blockchain technology can be applied as a digital backbone for projects, models and
operations within banking sector. The potential of blockchain technology in the
financial sector can be described by smart technology solutions, applications as well
as new business opportunities. Blockchain delivers a technology platform dedicated
for stock trading, record keeping, smart contacts and cryptocurrencies. Harris and
Wonglimpiyarant (2019). The technology can improve efficiency and transparency
in global banking systems. Nevertheless, in global business exist many technological
concepts there are not blockchains but only benefits some from blockchains rules. To
clarify, the key features proving that the solution is strictly blockchain have been
gathering and systematizing in Table 2.
Despite growing interest in the technology, stakeholders and organizations raise a
problem related to lack of knowledge, rules and methods. The issue has been also
highlighted by Zhou et al. (2020). The research team noticed limitations in existing
Development Factors of Blockchain Technology Within Banking Sector 131

Table 2 Key features of blockchain technology

Feature Clarifications
Distributed Each participant in a blockchain has access to the entire database and no one
database controls the data. Every single transaction can be verified directly without a
need for third-party intermediaries.
Peer-to-peer Communication occurs directly between peers without the need for central
transaction coordination. Peer nodes simultaneously functioning as both “clients” and
“servers” to the other nodes on the network.
Immutability of Once stored data cannot be modified. Various computational algorithms are
data developed to ensure that the recording on the database is permanent and
available to all network users.
Consensus Each record of data is updated based on consensus mechanism. That means
mechanism no central authority. Every data entry are verified by criteriums defined in
protocol.
Real time review Documents accessible through blockchain are reviewed and approved in real
time. All parties work on same ledger, all online and instant. No risk of
duplication or loss.
Cryptography A private key is known only to its owner and a public key is shared with the
world. A private key is first generated in random method and is then used to
create a public key. The private key is used to encrypt the transaction which
can then be decrypted by the intended recipient using the sender’s public
key.
Source: Iansiti and Lakhani (2017)

studies, literature and models. Still exists too many unclearnesses therefore further
works and tests are required. Blockchain faces many opportunities and many
challenges. According to the studies conducted by Ajrun and Suprabha in the catalog
of emerging areas of research should be included: business model transformations,
market reaction of blockchain, digital platforms as infrastructure, regulatory land-
scape, infrastructure requirements, socio-economic drivers and social media lever-
aged predictions (Arjun and Suprabha 2020). The potential of blockchain
technology means an added value created by implementation of blockchain tech-
nology for customers, stakeholders, workers and management.

3 Blockchain’s Implementation Potential in Banking

Industry

The value generated by implementation of blockchain technology within banking

industry is related to the following determinates: ability to ensure data has not been
tampered with, full traceability of any information on the blockchain, smart contracts
and distribution, increased security and quality and new business products or
services (WEF 2019). The key advantage of blockchain technology is data-sharing
mechanism. Implementation potential of blockchain within banking sector has been
132 M. Kołodziej

Table 3 Comparison of traditional banking and blockchain based banking

Factor Traditional banking Blockchain based banking
Customer experience Uniform scenarios Wide spectrum of possibilities
Homogenous service Personalized service
Poor customer experience Good customer experience
Efficiency Many intermediaries Disintermediation
Complexity of clearing process Distributed ledger, simplification
Low efficiency High efficiency
Costs Manual processes caused issues Automated
High costs Low costs
Safety Data storage in centralized way Distributed data storage
High probability of attack Asymmetric encryption
Low level of safety High level of safety
Source: Guo Y. and Liang Ch., Blockchain application and outlook in the banking industry. https://
link.springer.com/article/10.1186/s40854-016-0034-9/tables/2 (access: 05.07.2020)

widely discussed as disruptive and consumptive (Deloitte 2019). However, this trend
has been changed and banks and major financial institution are working on
blockchain projects.
The implementation of blockchain technology within banking sector is essential
for trade or credit financing, fractionalized asset trading, process automation and
interbank or cross-border payments. Through looking at blockchain impact for
banking sector there are some areas where added value is significant:
• Improving access to markets and funding.
• Increasing transparency, standardization, performance and quality.
• Maintain compliance with standards.
• Improving infrastructure operations, transactions and record keeping.
• Enhancing technological integration (OECD 2019).
From banking sector perspective implementation of blockchain technology can
be a part of big changes and following a development path which contains standard-
ization in infrastructure, stressing the importance of data transparency, risk manage-
ment, cost reduction and increase efficiency of charges (G20 2018). Table 3 includes
a quick comparison of traditional banking and new model of banking (includes
blockchain).
Implementation potential of blockchain technology in banking sector is focused
on the payment clearing and credit information systems. Additionally, banks are
exploring an opportunity delivered by smart contracts. In huge finance ecosystem is
required an extensive number of manual inspections and paper-based transactions.
There are many processes in banking that require numerous of intermediaries, a high
level of illegal transactions risk, high costs and low efficiency. The inclusion of
blockchain can reduce manual work and employ smart contracts to improve pro-
cedures by reduction of paperwork. Whereas smart contracts confirm that payment
has been processed automatically once a predetermined time. A key benefit of
blockchain technology implementation is costs reduction to banks and trade
Development Factors of Blockchain Technology Within Banking Sector 133

financing enterprises (Guo and Liang 2016). Blockchain can be also applied in the
context of anti-money laundering and customer identification programs. The tech-
nology can minimalize frauds through establishing complete history of transactions
within a single source of truth. A concept is very useful in increasing transparency
between market participants (Attaran and Gunasekaran 2019). According to
Buitenhek (2016) blockchain technology has been developed to resolve the follow-
ing problems: double spending, the issue of trust, consensus on the latest correct
version of the transaction history and maintenance of blockchain unchanged. Sum-
marized, the value of blockchain technology can be described in four main areas:
operational efficiency, risk mitigation, new business models and products and client
experience.

4 Methodology

To understood, define and analyze impact of blockchain technology for banking

sector have been created a procedure presented on Fig. 3. The main aim of the study
was to find factors related to blockchain technology implementation process in
banking sector. The research is based on literature review and case study analyses.
Working on this material, the following research questions have been set up:
1. What does implementation potential of blockchain technology mean?
2. What are the key benefits and challenges related to implementation process?
3. What factors are particularly important in banking sector?
The study is explorative in nature however it has been based on scientific
skepticism, therefore an author mentioned about opportunities and challenges
resulting from implementation and diffusion process. Literature review combined
with case study analyses is in a time range 2015–2020. In terms of case study an
author analyzed 20 use cases in banking sector.

Stage 1 – blockchain technology basis

Stage 2a – literature review Stage 2b – case study

Knowledge base to answer 3 questions

Fig. 3 The flow of research. Source: Own elaboration

134 M. Kołodziej

4.1 Stages of the Research’s Flow

The first stage of the research presented above includes blockchain’s basis. This step
was required to outline key advantages of blockchain technology which are impor-
tant from banking sector perspective. In the catalog of blockchain features essential
for blockchain implementation process are as follows: open distributed ledger,
consensus mechanism, immutability of data and cost reduction. The second stage
contains literature review and case study analyses. In this phase has been noticed that
authors highlight a huge implementation potential of blockchain technology espe-
cially for finance sector. Most of articles raise that to adopt blockchain technology it
is necessary to understand not only benefits but also challenges as blockchain is still
immature (Fig. 3).

4.2 Extraction of Impact Factors from Case Studies

Case study analysis was carried out on 20 use cases, which are explored and
popularized within banking sector on a global and local scale. Chosen use cases
represent also possible way for blockchain technology’s implementation. Use cases
included to research have been presented in Table 4 and have been selected using the
following criteria:
• capability: a tool can create in future the significant impact on financial sector’s
entities and services
• duration: a selected technological concept has been in existence for more
than year
• further development: a concept possess a potential for further development and
can be rebuild in the future
• interoperability: a use case can support and works with another systems
• transparency: available data and description reliably present the functioning
mechanism
Each use case has been checked for the factors listed in Table 5. For each impact
factor has been assigned the importance level which means number of cases in which
a given in Table 5 impact factor occurs. The similar study has been conducted by
Mohamad Osmani, Ramzi El-Haddadeh and Nitham Hindi. In this study have been
following factors mentioned: privacy, transparency, security, efficiency, immutabil-
ity and trust (Osmani et al. 2020). This stage gave a rise to the answers to the
questions from phase 3.
In the last stage there are answers for additional questions prepared. All questions
have been designed to gain a comprehensive overview of blockchain. Answering the
first question it has been noticed that it does not exist a strict definition of imple-
mentation potential. This term can be described as combination of blockchain
features, which properly implemented generate additional value for clients,
Development Factors of Blockchain Technology Within Banking Sector 135

Table 4 Use cases within banking sector

Use case Description
Abra (2017) Abra is a secure mobile-based app which offers instant P2P money
transfers without transaction fees.
Airfox (2016) Airfox is solution offering inclusive financial services for emerging
markets.
Atlas Money (2014) Atlas Money is banking platform developed as P2P using
blockchain technology.
Binance (2017) Binance is the operator of a blockchain-based platform developed to
facilitate cryptocurrency exchange.
Binkabi (2017) Binkabi is a tool recommended for issuing, trading, and financing
commodities on the blockchain.
Bitpesa (2013) Bitpesa is dedicated for cross-border payments for business and
individual level between African countries.
Blockchain Exchange The Blockchain Exchange Alliance is a solution to develop a
Alliance (2018) cryptocurrency trading platform designed to provide comprehensive
financial services.
Bloom (2017) Bloom is an end-to-end protocol created for identity attestation, risk
assessment and credit scoring.
Colendi (2016) Colendi is a protocol providing credit scoring evaluation.
CoMakery (2016) CoMakery accelerates the value of blockchain projects. It is a
community-based platform that helps contributors to be paid in
tokens directly from the projects they have provided their expertise
to.
Gauntlet (2018) Gauntlet is a developer of a simulation platform created to monitor
network activity and the fluctuation of asset values.
ICON (2017) ICON is a decentralized network that allows independent
blockchains with different governances perform transactions with-
out intermediaries.
Inclusivity (2015) Inclusivity delivers the inclusive banking and financial ecosystem
based on blockchain technology.
MonetaGo (2014) MonetaGo cooperates with financial institutions and central banks
to provide private permissioned blockchain solutions.
Ripple (2012) Ripple is a real-time gross settlement system, currency exchange,
and remittance blockchain network.
Standard Kepler (2016) Standard Kepler is the provider of a crypto-exchange platform that
offers blockchain-based investment banking services (asset man-
agement, token financing, other customized financial products).
The Next Ventures (2018) The Next Ventures is a solution which help startups get financial
support.
Tiger Trade [Redds] Tiger Trade is a full-service platform to buy and sell overstock
(2014) merchandise worldwide.
Uulala (2017) Uulala helps the underbanked communities of the world by giving
an access to the financial tools they need and the entertainment they
desire.
Wala (2015) Wala helps emerging market consumers reach financial prosperity
by eliminating barriers to banking.
Source: https://www.gsb.stanford.edu/faculty-research/publications/2019-blockchain-social-
impact (access: 20.12.2020)
136 M. Kołodziej

Table 5 Impact factors

Factor Importance level
Process efficiency 18—in 18 of analyzed cases this factor was mentioned
Distributed systems 20—in 20 of analyzed cases this factor was mentioned
Manual work reduction 13—in 13 of analyzed cases this factor was mentioned
Transparency 10—in 10 of analyzed cases this factor was mentioned
Costs reduction 20—in 20 of analyzed cases this factor was mentioned
Stability 14—in 14 of analyzed cases this factor was mentioned
Improving access 17—in 17 of analyzed cases this factor was mentioned
Safety and trust 19—in 19 of analyzed cases this factor was mentioned
Source: Own elaboration

stakeholders and management. Key benefits result from blockchain technology

implementation are as follows: cost reduction, disintermediation, transparency,
possibility to register transactions in real-time dimension. Main challenges related
to implementation of blockchain technology: lack of knowledge, insufficient infra-
structure, scalability, reliability, lack of universal rules and standards. In the banking
sector especially important are the following factors: efficiency, cost reduction,
transparency, using cryptography and accessibility to data, resources etc.

5 Conclusion

The article began with a blockchain technology explanation. The main contribution
of this paper lies in presenting the impact of blockchain technology on banking
sector. As an emerging and disruptive concept, blockchain can redefine (in long
term) currently existing business models. Nevertheless, the implementation and
diffusion blockchain technology in banking require changes, time and investment
expenditures. It must be highlighted that blockchain is still relatively immature
solution. The implementation potential of blockchain technology in banking indus-
try means—transparency, speed, safety, cost reduction, processes efficiency, data
integrity, authentication and stability. Banks see in blockchain technology a solution
of many organizational issues such as: manual work, complexity of processes,
unclear procedures. Implementation of blockchain technology can be treated as a
must due to competition from Fintech and another organization side. Being an early
adopter can be risky, but it can also generate a huge competitive advantage.
The study described in the article includes literature review and case study. Based
on that, it has been found that blockchain projects in banking are focused on deliver
an added value for clients (end users) and processes efficiency. Very important are
also cost reduction and data immutability. A literature review shows that authors are
seeing and potential of blockchain technology in decentralized structure of data base
and combination of technologies which are a guarantee of trust and stability
(e.g. cryptography, consensus mechanism).
Development Factors of Blockchain Technology Within Banking Sector 137

References

Arjun R, Suprabha KR (2020) Innovation and challenges of blockchain in banking: a scientometric

view. In: Special Issue on Artificial Intelligence and Blockchain. https://doi.org/10.9781/ijimai.
2020.03.004
Attaran M, Gunasekaran A (2019) Applications of blockchain technology in business. Springer.
https://doi.org/10.1007/978-3-030-27798-7
Axios (2018) Corporate America’s blockchain and bitcoin fever is over. https://www.axios.com/
corporate-america-blockchain-bitcoin-fervor-over-fb13bc5c-81fd-4c12-8a7b-07ad107817ca.
html. Accessed 01 Jul 2020
Bashir I (2018) Mastering blockchain. Helion, Gliwice
Buitenhek M (2016) Understanding and applying Blockchain technology in banking: evolution or
revolution? J Digital Bank 1(2)
Casino F, Daskalis TK, Patsakis C (2018) A systematic literature review of blockchain-based
applications: current status, classification and open issues. Telemat Informat 36:55–81
Cocco L, Pinna A, Marchesi M (2017) Banking on blockchain: costs savings thanks to the
blockchain technology. Future Internet 9:25
Crosby M (ed) (2015) Blockchain technology: beyond bitcoin. Sutardja Center for Entrepreneur-
ship & Technology Technical Report
Deloitte (2019) Blockchain. Enigma. Paradox. Opportunity. https://www2.deloitte.com/content/
dam/Deloitte/uk/Documents/Innovation/deloitte-uk-blockchain-full-report.pdf. Accessed
27 Jun 2020
Drescher D (2018), Blockchain. Podstawy technologii łańcucha bloków w 25 krokach. Helion,
Gliwice
Eyal I (2017) Blockchain technology: transforming libertarian cryptocurrency dreams to finance
and banking realities. Computer 50(9):38–49. https://doi.org/10.1109/MC.2017.3571042
G20 (2018) Communiqué, Finance Ministers & Central Bank Governors Meeting, 19–20 March
2018. https://back-g20.argentina.gob.ar/sites/default/files/media/communique_g20.pdf.
Accessed 02 Jul 2020
Glaser F (2017) Pervasive decentralization of digital infrastructures: a framework for block-chain
enabled system and use case analysis. Paper Presented at the Proceedings of the 50th Hawaii
International Conference on System Sciences (HICSS 2017). https://doi.org/10.24251/HICSS.
2017.186
Guimarães T, Silva H, Peixoto H, Santos M (2020) Modular blockchain implementation in
intensive medicine. Procedia Comput Sci 170:1059–1064. https://doi.org/10.1016/j.procs.
2020.03.073
Guo Y, Liang C (2016) Blockchain application and outlook in the banking industry. Financ Innov
2:24. https://doi.org/10.1186/s40854-016-0034-9
Harris W, Wonglimpiyarat J (2019) Blockchain platform and future bank competition. Foresight
21:625–639
Hassani H, Huang X, Silva E (2018) Banking with blockchained big data. J Manag Analytics 5
(4):256–275. https://doi.org/10.1080/23270012.2018.1528900
Huth M, Karame G, Vishik C (2020) An overview of blockchain science and engineering. Roy Soc
Open Sci 7:200168. https://doi.org/10.1098/rsos.200168
Iansiti M, Lakhani KR (2017) The truth about blockchain. Harv Bus Rev 95(1):118–127
Knezevic D (2018) Impact of blockchain technology platform in changing the financial sector and
other industries. Montenegrin J Econ 14(1):109–120
Li Z, Wang WM, Liu G, Liu L, He J, Huang GQ (2018) Toward open manufacturing a cross-
enterprises knowledge and services exchange framework based on blockchain and edge com-
puting. Indus Manag Data Syst 118(1):303–320
OECD (2019) Blockchain technologies as a digital enabler for sustainable infrastructure. OECD
Environment Policy Paper No. 16
138 M. Kołodziej

Osmani M, El-Haddadeh R, Hindi N, Janssen M, Weerakkody V (2020) Blockchain for next

generation services in banking and finance: cost, benefit, risk and opportunity analysis. J
Enterprise Inf Manag. https://doi.org/10.1108/JEIM-02-2020-0044
Pelz-Sharpe A (2019) Enterprise blockchain-market forecast & scenarios 2019-2024. Deep
Analysis
Peters GW, Panayi E (2016) Understanding modern banking ledgers through blockchain technol-
ogies: future of transaction processing and smart contracts on the internet of money. In: Tasca P,
Aste T, Pelizzon L, Perony N (eds) Banking beyond banks and money. New economic
windows. Springer, Cham. https://doi.org/10.1007/978-3-319-42448-4_13
Schinckus C (2020) The good, the bad and the ugly: an overview of the sustainability of blockchain
technology. Energy Res Social Sci 69
Stanford Business (n.d.) Blockchain for social impact 2019. https://www.gsb.stanford.edu/faculty-
research/publications/2019-blockchain-social-impact. Accessed 20 Dec 2020
Swan M (2015) Blockchain: blueprint for a new economy, 1st edn. O’Reilly Media, Sebastopol, CA
Upadhyay N (2020) Demystifying blockchain: a critical analysis of challenges, applications and
opportunities. Int J Inf Manag 54(2020):102120
Välikangas L (2020) Put your head on a blockchain? a few notes on the emergence of blockchain
technologies. Manag Organ Rev 16:199–201. https://doi.org/10.1017/mor.2019.68
WEF, Accenture (2019) Building value with blockchain technology: how to evaluate blockchain’s
benefits. http://www3.weforum.org/docs/WEF_Building_Value_with_Blockchain.pdf.
Accessed 22 Jun 2020
Zheng Z, Xie S, Dai H N, Chen X, Wang H (2017) An overview of blockchain technology:
architecture, consensus, and future trends. Presented at the 2017 IEEE International Congress on
Big Data (BigData Congress)
Zhou L, Zhang L, Zhao Y et al (2020) A scientometric review of blockchain research. Inf Syst
E-Bus Manag. https://doi.org/10.1007/s10257-020-00461-9

Internet sources accessed by research:

https://www.abra.com
https://airfox.com
https://positiveblockchain.io/database/atlas/
https://www.binance.com/en
https://www.bitpesa.com
https://bloom.co
https://www.bxa.com
https://www.binkabi.io
https://www.colendi.com
https://media.comakery.com/about
https://gauntlet.network
https://icon.foundation/?lang¼en
https://inclusivity.network/en/
https://www.monetago.com
https://ripple.com
https://www.standardkepler.com
https://www.next-ventures.com
https://tigertrade.com
https://uulala.io
https://getwala.com
Does Competition Matter for the Effects
of Macroprudential Policy on Bank Asset
Growth?

Małgorzata Olszak and Iwona Kowalska

1 Introduction

The aim of this paper is to determine to what extent competition in the banking sector
affects effectiveness of countercyclical macroprudential policy. This research thus
focuses on the role of competition in government economic (and regulatory) policy
aimed at reduction of procyclicality of the banking activity. And it deserves
highlighting that this procyclicality is considered to be one of major sources of
financial instability of the banking sector, and of financial sector (Borio 2003; CGFS
2012; BIS, FSB, IMF 2011; ESRB 2014a, b). Therefore, it becomes the main driving
force of financial crises which bring about undesirable real costs, in terms of reduced
economic growth and thus diminished societal welfare (CGFS 2012).
To resolve this problem, we shall focus on the impact of competition intensity on
the effect of macroprudential policy instruments on two areas of the banking activity,
which are of huge significance from the point of view of financial stability, i.e. the
general activity of banks proxied with assets growth and the sensitivity of asset
growth to business cycle. As previous research shows, macroprudential policy
instruments (see e.g. Lim et al. 2011; Cerutti et al. 2015; Claessens et al. 2014;
Olszak et al. 2018) have been found to affect both banking growth and sensitivity of
this growth to business cycle, with heterogenous results. We aim to test two sets of
hypotheses, which explain the role of competition in the relationship between:

M. Olszak (*)
Department of Banking and Money Markets, Faculty of Management, University of Warsaw,
Warsaw, Poland
e-mail: [email protected]
I. Kowalska
Department of Quantitative Methods, Faculty of Management, University of Warsaw, Warsaw,
Poland

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 139
K. Jajuga et al. (eds.), Contemporary Trends and Challenges in Finance, Springer
Proceedings in Business and Economics,
https://doi.org/10.1007/978-3-030-73667-5_9
140 M. Olszak and I. Kowalska

(1) bank asset growth and macroprudential policy; and (2) sensitivity of asset growth
to business cycle (procyclicality of the banking activity) and macroprudential policy.
According to the first hypothesis (H1), intense bank competition increases bank
incentives to invest in risky assets by increasing the costs of bank funding, and
makes bank activity to grow more and to be more sensitive to business cycle.
Consequently, effectiveness of countercyclical macroprudential policy is reduced.
The explanation for such a role of bank competition in bank risk-taking is termed as
“competition-fragility hypothesis” and is well established in the literature (see
Keeley 1990; Beck 2008). Considering the fact that more intense competition in
the deposit market results in greater risk-taking by banks (or inadequate risk-taking
across the business cycle), we expect that in countries with more competitive
banking markets, effectiveness of countercyclical macroprudential policy is reduced.
As for the second hypothesis (H2), intense bank competition reduces the risk of
bank assets, and thus increases the effectiveness of countercyclical macroprudential
policy. This expectation has its roots in the idea also named as the “competition-
stability hypothesis” and has been developed by Boyd and De Nicoló (2005). The
basic notion behind this explanation is that the classical argument that more bank
market power in the deposit market reduces bank risk—taking ignores the potential
impact of bank’s lending rates on firms and consumers behavior (i.e. it does not
consider the role of the demand side of the loan market). Boyd’s and De Nicoló
(2005) model shows that higher interest rates charged by banks may induce bor-
rowers to assume greater risk, which results in a higher probability that bank loans
turn non-performing. Therefore in countries with more intense competition in the
loan market, banks engage in less risk-taking, consequently the effectiveness of
countercyclical macroprudential policy is enhanced. In the opposite case, when
banks operate in countries exhibiting less competition in the loan market, they
tend to make more risky investments, therefore it is macroprudential policy that is
the very salient factor behind reduced procyclicality. However, the effectiveness of
this policy may be reduced by weak competition.
Contemporary research on macroprudential policy shows diversity of effects
(Claessens et al. 2014; Cerutti et al. 2017; Gambacorta and Murcia 2020; Gomez
et al. 2020). Several lines of research show that use of macroprudential policies may
be even associated with increased procyclicality in banking activity (Cerutti et al.
2017; Gambacorta and Murcia 2020). Such results seem to be in line with suggestion
of Danielsson et al. (2016), that the use of macroprudential policy instruments, in
particular those oriented at taming financial cycle, may result in more procyclicality.
Therefore, we expect that the support for particular hypotheses will be affected by
the type of macroprudential policy instrument and its effects on procyclicality.
This paper makes three significant contributions to the literature. Firstly, this
study is the first to relate competition to effects of macroprudential policy on bank
asset growth. Second, there is no research on the role of competition in the sensitivity
of asset growth to business cycle in countries which have applied macroprudential
policy instruments. Previous literature shows that research on the role of banking
sector competition in bank risk-taking focuses only on the level of bank-risk taking.
There is only one study directly focusing on the role of competition in procyclicality
Does Competition Matter for the Effects of Macroprudential Policy on Bank. . . 141

of credit by Leroy and Lucotte (2019). This study is however, considering only the
sample of individual banks from 12 European countries. Finally, this research will
put together the literature on countercyclical macroprudential policy, procyclicality
and competition. Contemporary research shows that in countries in which
macroprudential policy instruments have been applied in the pre-crisis period,
balance sheet and leverage growth as well as procyclicality have been reduced to
some extent (Lim et al. 2011; Cerutti et al. 2015, 2017; Claessens et al. 2014;
Gambacorta and Murcia 2020). This research, however, does not consider that
competition may be a factor affecting procyclicality, and thus, the force that has
an impact on the effectiveness of countercyclical macroprudential policy. Conse-
quently, this research should extend our knowledge on the nexus between compe-
tition and effectiveness of macroprudential policy, bridging the gap in the existing
state of the arts.
The intensity of competition in the banking industry, as it is in other industries, is
likely to have far-reaching implications for economic growth, productivity, financial
stability and, consequently, consumer welfare. Theoretical and empirical research
that can assess the extent to which competition in the banking sector affects
procyclicality and effectiveness of countercyclical macroprudential policy, has
important implications for government agencies responsible for the effective regu-
lation and supervision of the financial system (Beck et al. 2004; Boyd and De Nicoló
2005; Berger et al. 2009). However, as noted above, the effects of competition, in
particular on risk-taking and procyclicality of banks, are not obvious and potentially
complex (see e.g. Acharya and Richardson 2009; Beck et al. 2013).
By affecting bank risk-taking competition has the potential to make pressure on
the effects of macroprudential policies aimed at procyclicality. This motivates us to
contribute to the long-lasting debate on the role of competition in bank risk-taking,
by extending previous research in the area by testing the role of competition in
procyclicality of the banking activity and the impact of competition on effectiveness
of countercyclical macroprudential policy.
In this study we apply robust random effects estimator (as well as additional
estimation techniques to test the sensitivity of results, i.e. OLS, FE and GMM
two-step robust estimator) to a set of individual financial data of commercial
banks operating in over 90 countries in 2004–2016.
The rest of the paper is structured as follows. Section 2 provides the review of
relevant literature and develops our hypotheses. We describe our sample and
research methodology in Sect. 3. We discuss results and robustness checks in Sect.
4. Section 5 concludes our work.

2 Literature Review

This study is related to two significant streams in the finance literature. The first is the
stream which focuses on the links between competition and bank risk. The other, is
the contemporary stream concentrating on the links between bank risk-taking and
142 M. Olszak and I. Kowalska

procyclicality and the role of macroprudential polity in this link. The literature
focusing on the link between competition and bank risk can be divided into three
streams. As for the theoretical models, they have made contrasting predictions on the
relationship between bank competition and bank risk (for in depth literature analysis
see e.g. Carletti and Hartmann 2003; Allen and Gale 2004). In general, the theoret-
ical literature may be summarized under two headings. The first, is called
competition-fragility view, and predicts a positive relationship between competition
and bank risk. The other, named competition-stability view, predicts a negative
relationship between competition and bank risk.
A large academic literature provides support to the “competition-fragility” nexus.
The argument goes that competition in deposit market erodes banks’ profit margins
and hence charter values, which increases risk-taking incentives because banks have
less to lose in an insolvency (Marcus 1984; Keeley 1990; Allen and Gale 2004).
Keeley (1990) shows that in a situation in which a large number of banks compete,
profit margins are eroded and banks might take excessive risks to increase returns.
As more poor quality loan applicants receive financing, the quality of the loan
portfolio is likely to deteriorate and thereby increase bank fragility. Extended
versions of the Keeley’s framework also provide theoretical support for the
competition-instability hypothesis (see e.g. Allen and Gale 2000; Hellmann et al.
2000; Repullo 2004). But, Repullo (2004) also shows that for intermediate levels of
competition, both outcomes can occur. Empirical research on competition-fragility
hypotheses suggests a negative relationship between competition and bank stability
(see Keeley 1990; Dick 2006; Salas and Saurina 2003; for a thorough review refer to
Kowalska et al. 2016). More recently, Cipollini and Fiordelisi (2012) find a negative
link between bank market power and distress. Support for the competition-fragility
view is also found in Beck et al. (2013), who analyze the heterogeneity of compe-
tition and stability nexus in a sample of banks from 79 countries. Also Mirzaei et al.
(2013) find evidence that stability increases in less competitive environment in
emerging markets. Craig and Dinger (2013) find a robust positive link between
deposit market competition and asset risk, which they interpret as evidence for the
risk-increasing effects of deposit market competition. Thus, the authors suggest that
banks with less deposit market power are more likely to choose riskier strategies.
The competition-stability hypothesis contends that financial instability increases
as the degree of competitiveness is lessened. Banks with market power will earn
more rents by charging higher interest rates on business loans. In an important paper,
Boyd and De Nicoló (2005) have shown that the competition-fragility trade-off is
not robust to the introduction of competition in the loan market. Boyd and de Nicoló
(2005) argue that banks which compete on deposits also provide loans and set prices
for loans by taking into account the total amount of loans provided in the market. In
line with Stiglitz and Weiss (1981), Boyd and de Nicoló (2005) assume that the risk
of these loans is increasing along with the loan interest rate charged by banks. They
argue it is the borrowers who choose the riskiness of their investment undertaken
with bank loans. Generally, within the Boyd and de Nicoló’s framework, lower
levels of banking competition do not decrease bank risk but rather increase bank risk
through the enterprises risk shifting channel. Empirical papers testing the
Does Competition Matter for the Effects of Macroprudential Policy on Bank. . . 143

competition–stability hypothesis produce ambiguous results. Boyd et al. (2009)

investigate the relationship between competition and bank risk in South East Asian
banking sectors and find that competition does not increase or affect bank risk-
taking. Another work by Schaeck et al. (2009) concludes that more competitive
banking systems are more stable than monopolistic systems because of a lower
likelihood of bank failure and a longer time to crisis. Some support for competition–
stability view is found in Berger et al. (2009), as they show that increased compe-
tition in banking sectors operating in 23 industrial nations does not increase loan risk
in these nations. Schaeck and Cihák (2010) in a study comprising European and US
banks find that competition may have positive impact on bank soundness.
Martinez-Miera and Repullo (2010) extend the Boyd and de Nicoló (2005)
framework by introducing an idiosyncratic risk factor for firms and build a model
which predicts that the effect of bank competition on bank risk taking is non-linear.
Results of simulations which account for this risk factor indicate that the risk-shifting
effect of enterprises identified by Boyd and de Nicoló (2005) prevails, but another—
the so-called margin effect—emerges. Thus, more competition leads to lower loan
rates, and consequently lower margins from non-defaulting loans. This decreases
bank’s operating profits before taxes and provisions, which can be buffers against
loan losses. In such circumstances, banks become riskier and more vulnerable for
corporate failures which again increases bank risk. Generally, Martinez-Miera and
Repullo (2010) model shows that, as competition increases, the margin effect
eventually dominates the risk-shifting effect and leads to a U-shaped relationship
between probability of bank failure and the number of banks (see Lee and Chiang
2012). Several papers challenge empirically the concept of U-shaped relationship
between competition and bank risk. Liu et al. (2010) find that a U-shaped relation-
ship exists between regional bank competition and stability. Bretschger et al. (2012)
argue that the “concentration-stability” and the “concentration-fragility” hypotheses
suggest opposing effects operating through specific channels. Their findings provide
support for the assumption of channel effects in general and both the concentration-
stability and the concentration-fragility hypothesis in particular. The effects are
found to vary between high and low income countries. Tabak et al. (2012) also
find that competition affects risk-taking behavior in a non-linear way as both high
and low competition levels enhance financial stability. They find the opposite effect
for average competition. In a recent paper Cuestas et al. (2020) find an inverse
U-shaped relationship between competition and financial stability in Baltic coun-
tries. This contradicts the Martinez-Mierra and Repullo (2010) theory, which argues
that both high and low competition endanger financial stability.
Contemporary research on procyclicality of the banking activity suggests that its
driving force is inadequate response of banks towards risk-taking during the whole
business cycle (Olszak et al. 2018; Borio et al. 2001; Borio 2014). This phenomenon
may be explained to some extent with the economic theory (market-frictions and
asymmetric information, moral hazard as a side effect of the information asymmetry,
risk illusion, coordination problems) and using the theoretical background of psy-
chology and behavioral finance. In the boom periods and in favorable macroeco-
nomic conditions, banks perceive risk as marginally low, and therefore make more
144 M. Olszak and I. Kowalska

investments (e.g. loans) ignoring prudent standards, e.g. using less restrictive rules
for extending loans. The opposite approach is employed by banks during recessions
and unfavourable conditions in the economy. Such changes in risk-taking result in
amplification of lending in boom periods and excessive unwillingness to extend
loans in recessionary periods (see also Minsky 1986), when bank lending is most
needed to stimulate economic growth. Due to the fact that many economies around
the globe have suffered from the effects of the recent financial crisis of 2007/2008,
both academic researchers as well as regulatory standard setters, have started looking
for policy instruments which could reduce the potential of the banking sector to be
excessively (see Borio and Zhu 2012) procyclical. These instruments are currently
named as countercyclical macroprudential policy tools, and include ratios which aim
to reduce borrower risk taken by a bank (e.g. loan to value, LTV; debt to income,
DTI), as well as bank risk-taking in general (e.g. countercyclical macroprudential
policy instruments or dynamic provisions) (see Claseesns 2014). The empirical
research on the effects of macroprudential policy in countries which applied
macroprudential policy instruments before the financial crisis of 2007/2008, show
that macroprudential policy has the potential to affect bank risk-taking and to some
extent procyclicality of the banking activity (Lim et al. 2011; Cerutti et al. 2015).
The empirical evidence on the effectiveness of macroprudential policies in
managing the resilience of the banking (and financial) sector and the credit cycle,
and thus financial stability, is still preliminary. The literature presenting this evi-
dence can be grouped into cross-country studies and micro-level evidence (mostly
based on the use of one, or a few, macroprudential policy instruments). We shall
focus on cross-country studies because this study is conducted in a cross-country
context (for analysis of individual country evidence refer to Cerutti et al. 2015). One
of the first cross-country studies was a paper by Lim et al. (2011), exploring the links
between macroprudential policy instruments (LTV caps, DP, DTI caps, limits on FC,
countercyclical capital; buffers, limits on credit growth) and developments in lever-
age and credit, using aggregated annual data from 49 countries in years 2000–2010.
They document that the presence of ratios such as LTV and DTI limits, ceilings on
credit growth, reserve requirements and dynamic provisioning rules can mitigate the
procyclicality of credit and leverage (i.e. they reduce the positive sensitivity of credit
and leverage to the business cycle, proxied by real GDP growth). This study also
shows that reserve requirements and dynamic provisions are effective in reducing
credit growth during booms. Caps on LTV are associated with generally higher loans
growth. As for the leverage growth, they document that only reserve requirements
reduce it in a significant way, both generally and in boom periods. In another study,
IMF (2013) investigates the impact of changes in the use of macroprudential policy
instruments on financial vulnerabilities (i.e. credit growth, house price inflation, and
portfolio capital inflows) and on the real economy (real output growth and the share
of residential investment). This study implies that both capital requirements and RR
strongly influence credit growth. LTV limits and capital requirements are found to
strongly affect house-price inflation rates. RR evidently reduce portfolio inflows in
emerging markets with floating exchange rates. This study also considers whether
Does Competition Matter for the Effects of Macroprudential Policy on Bank. . . 145

the effects are asymmetric between tightening and loosening, but finds no significant
indication of such asymmetry.
Vadenbushce et al. (2012) find that capital ratio requirements and non-standard
liquidity measures (such as marginal reserve requirements on foreign funding or
linked to credit growth) helped slow down house-price inflation in Central, Eastern
and Southeastern Europe. Dell’ Ariccia et al. (2012) show that macroprudential
instruments can reduce the incidence of general credit booms and decrease the
probability that booms end badly. Specific policies, such as credit and interest
controls and open foreign exchange position limits, are found to be effective in
reducing the probability that booms ends up in a financial crisis or subsequent
economic underperformance. This study implies that due to the fact that these
policies reduce the risk of a bust, they simultaneously make the whole economy
resilient to the disruptions in the financial system.
Claessens et al. (2014) investigate how changes in balance sheets—i.e. in lever-
age, assets and non-core liabilities growth, of some 2800 banks in 48 countries over
2000–2010, respond to specific macroprudential policy instruments. They find that
borrower-targeted instruments—LTV and DTI caps, and CG and FC limits—are
effective in reducing the growth in bank’s leverage, asset and non-core liabilities.
Countercyclical instruments (such as RR and DP) also help mitigate increases in
bank leverage, but they are of little effect thorough the cycle. Some of policies are
counterproductive during downswing, serving to amplify declines. In a related
study, Kuttner and Shim (2013) using data from 57 countries find that housing credit
growth is significantly affected by changes in the maximum debt-service-to-income
(DSTI) ratio, the maximum loan-to-value ratio, limits on exposure to the housing
sector and housing related taxes. However, only the DSTI ratio limit has a significant
effect on housing credit growth when they apply mean group and panel event study
methods.
Cerutti et al. (2015) document the use of macroprudential policies for 119 coun-
tries over the 2000–2013 period, covering many instruments. This study shows that
usage of macroprudential policies is generally associated with lower growth in
aggregated credit, notably in household credit. These effects are, however, less
evident in more developed and open economies, in which the usage of
macroprudential policies comes with greater cross-border borrowing, suggesting
that these countries face issues of avoidance. Generally this study implies that
macroprudential policies can help manage financial cycles, but they work better in
the boom than in the bust phase of the financial cycle.
In a recent paper, Olszak et al. (2018) show that effectiveness of various
macroprudential policy instruments in reducing the procyclicality of loan-loss pro-
visions (LLPs) using individual bank information from over 65 countries and
applying the two-step GMM Blundell and Bond (1998) approach with robust
standard errors. They identify several new facts. Firstly, borrower restrictions are
definitely more effective in reducing the procyclicality of loan-loss provisions than
other macroprudential policy instruments. This effect is supported in both uncon-
solidated and consolidated data and is robust to several robustness checks. Secondly,
dynamic provisions, large exposure concentration limits and taxes on specific assets
146 M. Olszak and I. Kowalska

are effective in reducing the procyclicality of loan-loss provisions. They also find
that both loan-to-value caps and debt-to-income ratios, are especially effective in
reducing the procyclicality of LLP of large banks. Concentration limits and taxes are
also effective in reducing the procyclicality of LLP of large banks. Dynamic pro-
visions reduce the procyclicality of LLP independently of bank size.

3 Methodology and Data

3.1 Methodology

To measure the impact of macroprudential policy instruments on the banking

activity and on procyclicality of the banking activity we need to have access to
data on the actual use of macroprudential policy instruments across countries.
Information on the actual use and effects of macroprudential policies and some
data have nevertheless been collected recently for 119 countries by the IMF (see
Cerutti et al. 2015, 2017) for 2000–2016 period, which is the best and most
comprehensive database on these instruments, applied in recent banking research
(e.g. Gaganis et al. 2020). Analysis of the data-set developed in Cerutti et al. (2015,
2017) shows that many of instruments whose nature is macroprudential were applied
in years 2000–2015, which covers the economic and financial boom period of
2001–2006/2007 and the crisis and its direct side-effects period (2008–2010). The
number and diversity of these macroprudential policy instruments has evolved in
recent years. Some of them have been in use in many countries definitely before the
crisis (see Cerutti et al. 2015, 2017 database), whereas several are relatively new
tools implemented after the crisis, i.e. in 2011. In our study, we consider the
instruments which were in effective use in the period of our analysis (2004–2016)
and following Altunbas et al. (2018) we divide them into two groups: cyclical
(affecting cyclicality of banking activity) and resilience oriented (affecting directly
balance-sheets and income statements, and therefore the bank-capital base). Cyclical
instruments include: loan to value caps (denoted as ltv_cap), debt to income ratios
(denoted as dti), foreign currency limits (fc), credit growth limits (cg) and FX and/or
countercyclical reserve requirements (rr_rev). Resilience oriented tools include:
dynamic provisions (denoted as dp), leverage ratios (denoted as lev), interbank
operations limits (inter), concentration limits (conc), levies/taxes on financial insti-
tutions (tax). All instruments are binary variables, which take the value of 1 if the
instrument is applied in a country, and 0 if it is not applied. Some of them have been
applied for the whole period in some countries (e.g. ltv, dti, conc, lev, see the Cerutti
et al. 2017, database) and are therefore time invariant. We also use aggregated
macroprudential policy indices as defined in Cerutti et al. (2017). Borrower covers
ltv_cap and dti, and takes values of 0, 1 or 2. Financial covers macroprudential
instruments affecting the balance sheets of financial institutions, with values ranging
between 0 and 8. In our sample the highest value is 5, showing that some countries
have applied 5 balance-sheet oriented macroprudential policy instruments.
Does Competition Matter for the Effects of Macroprudential Policy on Bank. . . 147

To identify competition intensity, we use traditional Lerner index (available in the

World Bank Global Financial Development database). To measure relative intensity
of competition across countries we divide countries in two sets, of which the first
covers countries with less market power, and thus with stronger competition, and the
other covers countries with relatively weak competition. To identify banking sectors
with intense competition we introduce the dummy variable taking the value of 1 in
countries in which the Lerner index is below the 30% of the countries. We denote
this measure as High competition. To identify low competition countries we use the
dummy equal to 1 in the 30% of countries in which the Lerner index takes highest
values. This binary variable is denoted as Low competition. The rest of the sample
includes 40% of countries with medium intensity of competition.
To test our hypotheses we apply the following model:

Asset growthi,j,t ¼ α þ β1 MPI j,t þ β2 CompetitionLevel j þ β3 MPI j,t

 CompetitionLevel j þ β4 GDPG j,t þ β5 MPI j,t

 GDPG j,t þ β6 CompetitionLevel j  GDPGk,t

þ β7 CompetitionLevel j  MPI k,t  GDPGk,t
þ γBSOC i,k,t1 þ εi,t ð1Þ

where: Asset growthi,j,t—asset growth of the i-th bank in country j in moment t;

CompetitionLevelj—competition intensity in the banking sector in country j. We
will use two types of this measure, in separate models, High competition and Low
competition, as defined in the previous paragraph; BC—the measure of the business
cycle, proxied with real GDP growth rate; MPI—dummy variable denoting whether
the macroprudential policy instrument is applied (or not) in a given country or the
number of macroprudential policy instruments applied (in models with aggregated
measure of MPI); BSOC—bank-specific and other control variables, which include
loans growth rate, deposits growth rate, loans to customer deposits (as a proxy for
liquidity risk), growth of leverage and logarithm of assets (to proxy bank size),
lagged asset growth and the country level Lerner index.
The β1 regression coefficient on MPI informs about asset growth in countries
using macroprudential policy instruments. Following previous evidence, mostly on
credit growth (Aiyar et al. 2014; Gomez et al. 2020) and some on asset growth
(Claessens et al. 2014), we expect a negative coefficient on this coefficient as
denoting reduced growth in countries using macroprudential instruments. In models
in which we will include interaction term between MPI and Competition level this
coefficient informs about the role of MPI in countries in which competition level as
defined in the interaction is equal to 0. If we consider High competition in the
interaction, this coefficient on MPI informs about asset growth in low and medium
level competition countries.
The β2 is our measure of the role of competition level in asset growth. A positive
(negative) sign on high (low) competition denotes increased asset growth in more
competitive banking sectors. In models including interaction term between MPI and
148 M. Olszak and I. Kowalska

Competition level this coefficient informs about the role of High (low) competition in
countries in which macroprudential instruments are not applied (i.e. MPI ¼ 0).
The β3 on the interaction term between MPI and Competition level informs about
the asset growth in countries using macroprudential instruments (MPI ¼ 1) under
given competition intensity. Thus, if competition is high and this coefficient is
positive, it implies that in countries using macroprudential instruments asset growth
is increased in high competition countries. This would be our test of hypothesis H1.
In contrast, a negative coefficient on this term informs about decreased asset growth
in countries applying macroprudential policy instruments in highly competitive
banking industry. Such a result is a support for hypothesis H2.
The β4 on GDPG informs about sensitivity of asset growth to business cycle. A
positive (negative) sign denotes procyclicality (countercyclicality) of asset growth.
Following previous research (mostly) on bank credit growth (Lim et al. 2011; Cerutti
et al. 2017; Gambacorta and Murcia 2020) we expect a positive effect of business
cycle on asset growth.
The β5 coefficient on double interaction term between MPI and GDPG informs
about the sensitivity of asset growth to business cycle in countries using
macroprudential policy instruments and in which the competition intensity measure
equals 0. Although macroprudential policy instruments are expected to reduce
procyclicality, Danielsson et al. (2016) suggest that their final effect may be
increased procyclicality. Thus, in such a case, we expect to obtain a positive
coefficient on GDPG. However, as Danielsson’s et al. (2016) focus mainly on
cyclical instruments, we envisage that the results may differ between cyclical and
resilience-oriented instruments.
The β6 on double interaction term of Competition level and GDPG informs about
the sensitivity of asset growth to business cycle under high (low) competition in the
banking industry in countries which do not use macroprudential instruments
(i.e. MPI ¼ 0). Following Leroy and Lucotte (2019) for credit growth in individual
banks in European countries, we expect that increased competition reduces
procyclicality of asset growth. Thus, this coefficient will be negative for interaction
between High competition and GDPG and positive for interaction between Low
competition and GDPG.
The β7 on triple interaction term between MPI and Competition level and GDPG
informs about the sensitivity of assets growth to business cycle in countries using
macroprudential policy instruments and differing in terms of competition intensity
(high versus low competition). Looking first at High competition, we expect a
negative coefficient if increased competition reduces procylicality of asset growth
in countries using MPI. This would be in line with hypothesis H2. In contrast, a
positive coefficient on this term would be in line with hypothesis H1, that increased
competition is associated with more procyclicality of banking growth in countries
using macroprudential policy. We expect that the role of competition level may
differ between cyclical and resilience oriented instruments.
In the estimation of baseline version of model expressed with Eq. (1) we do not
include interaction terms, and only use bank level asset growth regressed on bank
level control variables real GDP growth and Lerner. We shall use several estimation
Does Competition Matter for the Effects of Macroprudential Policy on Bank. . . 149

techniques in base estimations, i.e.: ordinary least squares panel estimation (OLS),
FE and RE as well as we use a simple dynamic identification scheme based on
Blundell and Bond’s (1998) GMM approach. The proper use of GMM models
requires robust selection of instruments (Roodman 2009). Considering the hetero-
geneity of the sample applied in our study, this requirement is difficult to fulfill.
Thus, the GMM approach will not be applied in our analysis.
As in our estimations we use time invariant variables to measure intensity of
competition (High competition and Low competition) and to identify the use of
individual macroprudential policy instruments (such as ltv_cap, dti, etc.), the fixed
effects estimator is also not adequate, because we will not be able to show the
regression coefficients on both, competition dummies and individual
macroprudential policy instruments. We need these coefficients (β1, β2) to test our
hypotheses about the role of macroprudential instruments and of competition level in
the asset growth. Therefore, in the final estimations of Eq. (1) we shall use the robust
random effects estimator.
In our analysis we use both individual macroprudential policy and aggregated
macroprudential policy indices (i.e. borrower and financial). Therefore in our anal-
ysis we will run separate models for aggregated tools. This analysis will inform us
about the role of more intensive use of macroprudential policy (in terms of the
number of instruments applied) in asset growth and in sensitivity of asset growth to
business cycle. We will also run models expressed with Eq. (1) for each individual
macroprudential instrument.

3.2 Data

We use pooled cross-section and time series data of individual banks’ balance-sheet
items and profit and loss accounts from over 90 countries and country-specific
macroeconomic indicators for these countries, over a period from 2004 to 2016.
The data on commercial banks are taken from unconsolidated financials available in
the Bankscope database, whereas the macroeconomic data were accessed from the
World Bank and the IMF web pages. We apply several filters to remove potential
data errors and outliers. We exclude outlier banks from our sample by eliminating
the extreme bank-specific observations when a given variable adopts extreme values
(e.g. negative capital ratios which may be the result of misreporting or other data
problems). Due to the fact that we are interested in the role of macroprudential policy
in reducing the growth of banking activity or in reducing the procyclicality of the
banking activity, we focus on those banks for which we have at least 6 years of
observations on asset growth (and loans growth)—to take into account the whole
business cycle. Our final sample consists of over 90,000 observations (over 9000
banks) (see Table 1, PANEL A) operating in 107 countries (see Table 6 in
Appendix).
We conduct our analysis for the whole sample of countries instead of using
country subsamples (e.g. Asia, European Union, Latin America, or high versus
150 M. Olszak and I. Kowalska

low income countries), because we need to have access to data about diversity across
countries in terms of competition intensity (high versus low competition) and of the
use of macroprudential policy instruments. In our sample we have over 30 countries
denoted as High competition countries and over 30 countries with low competition
intensity. The other countries are denoted as having medium level of competition.
The use of macroprudential instruments has evolved in the period of analysis years,
with more countries using greater range of instruments recently. Considering this we
need to use as more countries as possible to be able to identify relively large
subsamples covering more and less competitive banking industries. In previous
research it would be interesting to test the role of country clustering to find out
how this affects the links between competition and cyclicality of banking activity in
countries using macroprudential policy instruments.
Table 1 (PANEL A) provides some descriptive statistics about variables in our
estimation sample. Looking at variable of interest to our study we find that the mean
asset growth equals to 7.66%, suggesting that on average banking sectors tend to
grow. The correlations indicate at statistically-significant association between asset
growth and each of the explanatory variables. In particular, the correlation coeffi-
cient for GDPG is positive, suggesting that asset growth is procyclical. The corre-
lation between asset growth and Lerner index is (see Table 1, PANEL B) is positive,
suggesting that increases in competition are related with decreased asset growth. Of
macroprudential policy instruments, mainly restrictions targeted at bank balance-
sheets (Financial) tend to be negatively correlated with asset growth, which may
potentially imply that they work countercyclically.

4 Estimation Results

4.1 Baseline Results

Table 2 reports the base results. The coefficients on bank-specific control variables
are largely as expected when significant. Banks with higher values of loans to
deposits, which is a proxy for liquidity risk, grow faster. This means that greater
liquidity risk is related with increased asset growth. Leverage does not affect bank
asset growth in our sample. Large banks are growing faster because regression
coefficient on bank size (proxied with bank assets) is positive and statistically
significant in all estimations presented in Table 2. Bank asset growth is procyclical
because in all statistically significant specifications in Table 2, the coefficient on
GDPG is positive, and ranges between 0.11 and 0.13. Looking now at the role of
macroprudential policy in reducing asset growth we find that both, borrower targeted
macroprudential policy index which covers instruments targeted on taming the risk-
taking by borrowers (Borrower) as well as macroprudential index which covers
instruments targeted on taming the risk-taking by financial institutions, in particular
by banks (Financial) reduce asset growth, with stronger effect of Borrower than
Table 1 Descriptive statistics (PANEL A) and correlation matrix (PANEL B) for the full sample
Log
Variable Δassets Δloan Δdeposits loans/deposits Capital ratio (assets) GDPG Lerner borrower financial
Panel A: Descriptive statistics
Mean 7.66 8.08 7.66 100.99 0.54 12.53 2.30 0.22 0.04 1.84
Std. dev. 22.03 31.73 31.63 103.43 13.03 1.87 2.87 0.30 0.23 1.12
Min 661.91 777.38 830.50 0.00 946.52 3.65 14.81 8.66 0.00 0.00
Max 687.32 906.47 978.71 999.93 518.4 21.89 34.50 0.94 2.00 5.00
No of obs. 95,240 95,024 95,109 94,019 74,840 972,630 109,908 108,350 109,944 1,099,440
No of banks 9154 9150 9152 9074 7983 9155 9159 9145 9162 9162
Panel B: Correlation matrix
Δassets 1
Δloan 0.69*** 1
Δdeposits 0.83*** 0.63*** 1
loans/deposits 0.03*** 0.05*** 0.02*** 1
Capital 0.33*** 0.55*** 0.51*** 0.02*** 1
ratio
Log(assets) 0.06*** 0.03*** 0.04*** 0.02*** 0.04*** 1
GDPG 0.17*** 0.17*** 0.12*** 0.07*** 0.00 0.05*** 1
Lerner 0.03*** 0.02*** 0.02*** 0.03*** 0.01 0.06*** 0.06*** 1
Does Competition Matter for the Effects of Macroprudential Policy on Bank. . .

borrower 0.00 0.00 0.00 0.06*** 0.00 0.08*** 0.06*** 0.02*** 1

financial 0.05*** 0.04*** 0.02*** 0.17*** 0.02*** 0.21*** 0.03*** 0.25*** 0.05*** 1
*, **, *** denote an estimate significantly different from 0 at the 10%, 5% and 1% levels, respectively
151
Table 2 Baseline results
152

OLS FE RE GMM RE RE GMM GMM

1 2 3 4 5 6 7 8
Δassets(t1) 0.027*** 0.011*** 0.019*** 0.027** 0.019*** 0.019*** 0.023 0.024*
(0.001) (0.001) (0.001) (0.011) (0.001) (0.001) (0.014) (0.014)
Δloan 0.108*** 0.104*** 0.107*** 0.035 0.107*** 0.106*** 0.048 0.04
(0.001) (0.001) (0.001) (0.048) (0.001) (0.001) (0.041) (0.039)
Δdeposits 0.732*** 0.731*** 0.729*** 0.777*** 0.732*** 0.732*** 0.752*** 0.749***
(0.002) (0.002) (0.002) (0.036) (0.002) (0.002) (0.05) (0.052)
loans/deposits 0.001 0.005*** 0.002*** 0.018 0.003*** 0.002*** 0.03** 0.024*
(0) (0.001) (0.001) (0.016) (0.001) (0.001) (0.014) (0.013)
Capital ratio 0.217*** 0.205*** 0.211*** 0.064 0.214*** 0.214*** 0.029 0.01
(0.003) (0.003) (0.003) (0.067) (0.003) (0.003) (0.096) (0.066)
log_assets 0.076*** 0.804*** 0.109*** 0.262 0.118*** 0.103*** 0.085 0.077
(0.012) (0.069) (0.017) (0.28) (0.017) (0.018) (0.421) (0.381)
GDPG 0.15*** 0.115*** 0.131*** 0.11*** 0.131*** 0.133*** 0.124*** 0.12***
(0.008) (0.01) (0.009) (0.031) (0.009) (0.009) (0.032) (0.033)
borrower 0.312 1.609
(0.2) (3.155)
financial 0.144*** 0.033
(0.028) (0.115)
Lerner 0.74*** 0.844*** 0.357 0.516
(0.08) (0.082) (0.581) (0.5)
cons 0.628*** 9.791*** 1.025*** 4.102 1.384*** 0.899*** 2.654 2.102
(0.143) (0.864) (0.214) (3.14) (0.22) (0.237) (4.625) (4.4)
p-value F test 0.00 0.00
p-value Wald test 0.00 0.00 0.00
M. Olszak and I. Kowalska
p-value Sargan test 0.00 0.00 0.00
p-value AR(2) test 0.01 0.02 0.01
R-sq: within 0.87 0.87 0.87 0.87
R-sq: between 0.89 0.90 0.90 0.89
R-sq: overall 0.88 0.88 0.88 0.88 0.88
No of observations 73,650 73,650 73,650 73,650 73,066 73,066 73,066 73,066
No of banks 7942 7942 7942 7918 7918 7918 7918
No of instruments 193 193 193
t-Statistics are included in parentheses. *, **, *** denote an estimate significantly different from 0 at the 10%, 5% and 1% levels, respectively
Does Competition Matter for the Effects of Macroprudential Policy on Bank. . .
153
154 M. Olszak and I. Kowalska

Financial. Generally, our baseline results seem to be in line with previous evidence
(see Claessens et al. 2014; Cerutti et al. 2017).
Looking at the link between Lerner index and asset growth we find that it is
positive and ranges between 0.74 (see regression 5 in Table 2) and 0.84 (see
regression 6 in Table 3), suggesting that higher levels of Lerner index (i.e. lower
competition intensity) are related with higher asset growth. Therefore, we might
expect that intense competition (i.e. low Lerner index) shall be related with
decreased asset growth.

4.2 Competition Intensity and Bank Asset Growth

and Sensitivity of Asset Growth to Business Cycle
in Countries Applying Macroprudential Instruments

In Table 3 we present results of the analysis of the role of competition intensity and
aggregated instruments of macroprudential policy—borrower and financial.
Looking first at the role of macroprudential policy for asset growth we find that
countries using more macroprudential policy instruments denote reduced asset
growth, independent of the competition intensity (see columns 1–8 in Table 3).
These results are line with previous evidence for asset growth (Claessens et al. 2014)
and for bank lending activity, showing that macroprudential policy effective in
reducing credit growth (Cerutti et al. 2017; Gomez et al. 2020; Gambacorta and
Murcia 2020).
High competition intensity is related with increased asset growth in countries
applying more financial instruments than low and medium level competition in the
banking industry because the coefficient β1 on financial is negative and statistically
significant in less competitive countries (compare columns 2 and 6) than in more
competitive countries (compare columns 4 and 8). This finding is further supported
by the β3 regression coefficient on double interaction term of MPI x High competi-
tion which is positive and statistically significant—taking the value of 0.39 (columns
2) and 0.291 (column 6). This result thus seems to support expectation expressed in
hypothesis H1, that increased competition is associated with weakened effectiveness
of macroprudential policy. But the results are binding only for financial index,
because we do not find support for this view for countries using more borrower
targeted instruments (ltv caps and dti). However, competition level does not matter
significantly for the effects of borrower on asset growth.
Turning to analysis of procyclicality of asset growth we find that it is still
supported with all regressions included in Table 3, because the β4 regression
coefficients on GDPG are positive and statistically significant. In line with sugges-
tion of Danielsson et al. (2016) we find support for the view that countries using
more macroprudential instruments exhibit increased procyclicality. In our study we
find that β5 regression coefficient on double interaction term of MPI x GDPG is
positive and statistically significant in all models (see columns 5–8).
Table 3 Role of aggregated macroprudential policy indices in asset growth and in sensitivity of asset growth to business cycle under high (low) competition
intensity
High competition Low competition High competition Low competition
intensity intensity intensity intensity
MPI
instrument: borrower financial borrower financial borrower financial borrower financial
1 2 3 4 5 6 7 8
The role of MPI and compe- MPI β1 0.05 0. 0.263 0.14*** 0.125 0.329*** 0.268 0.145***
tition for assets growth 2323***
(0.285) (0.033) (0.214) (0.028) (0.285) (0.034) (0.215) (0.028)
High β2 0.541*** 1.29*** 0.236** 0.95***
competition
(0.097) (0.128) (0.103) (0.136)
MPI  High β3 0.32 0.39*** 0.419 0.291***
competition
(0.401) (0.084) (0.521) (0.102)
Low β2 0.282* 0.29 0.526*** 0.73*
competition
(0.155) (0.312) (0.189) (0.409)
MPI  Low β3 0.381 0.048 0.252 0.098
competition
(0.617) (0.175) (1.109) (0.232)
Analysis of the role of MPI GDPG β4 0.129*** 0.132*** 0.128*** 0.131*** 0.154*** 0.164*** 0.101*** 0.045***
and competition for sensi-
Does Competition Matter for the Effects of Macroprudential Policy on Bank. . .

(0.009) (0.009) (0.009) (0.009) (0.01) (0.02) (0.009) (0.017)

tivity of assets growth to MPI  β5 0.068*** 0.003 0.026* 0.027***
business cycle GDPG
(0.014) (0.007) (0.015) (0.006)
(continued)
155
Table 3 (continued)
156

High competition Low competition High competition Low competition

intensity intensity intensity intensity
MPI
instrument: borrower financial borrower financial borrower financial borrower financial
1 2 3 4 5 6 7 8
High com-  
petition  0.188*** 0.243***
GDPG
(0.022) (0.027)
MPI  High β7 20.129 0.069***
competition
 GDPG
(0.084) (0.019)
Low com- β6 0.169*** 0.199***
petition 
GDPG
(0.027) (0.05)
MPI  Low β7 0.023 0.009
competition
 GDPG
(0.182) (0.031)
Wald test 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
p-value
R-sq: 0.88 0.88 0.88 0.88 0.88 0.88 0.88 0.88
overall
No of 73,066 73,066 73,066 73,066 73,066 73,066 73,066 73,066
observation
No of banks 7918 7918 7918 7918 7918 7918 7918 7918
M. Olszak and I. Kowalska

Notes: This table presents a reduced view of model Eq. (1). Regression coefficients correspond to β1–β7 as presented in Sect. 3.1; t-statistics are included in
parentheses. *, **, *** denote an estimate significantly different from 0 at the 10%, 5% and 1% levels, respectively
Does Competition Matter for the Effects of Macroprudential Policy on Bank. . . 157

In line with findings of Leroy and Lucotte (2019), our results give support to the
view that more competitive banking industry is associated with decreased
procyclicality. As we can see, the β6 regression coefficients are negative and
statistically significant in more competitive banking industry (see the interaction
term of High competition x GDPG) and positive and statistically significant in low
competition industries (see the interaction term of Low competition x GDPG).
Intense competition is associated with decreased procyclicality of assets growth
in countries using more Borrower instruments, because the β7 regression coefficient
on triple interaction of MPI x High competition x GDPG is negative and statistically
significant. This results seems to support hypothesis H2, that intense competition
reduces the risk of bank assets, and thus increases the effectiveness of
macroprudential policy to reduce procyclicality. In contrast, for countries using
more Financial instruments we find support for hypothesis H1 that intense bank
competition is associated with increased procyclicality in countries using more such
instruments.
In Tables 4 and 5 we present the effects of macroprudential policy on cyclicality
of asset growth for each instrument one by one. We present separate analysis for
cyclical instruments in Table 4 (as defined in previous research, e.g. by Altunbas
et al. 2018) and for resilience-oriented tools in Table 5. Therefore we omit the
analysis of the role of competition intensity for asset growth (regression coefficients
from β1 to β3).
For the sake of simplicity we shall focus here on the links between
macroprudential policy instruments and the sensitivity of asset growth to business
cycle. As the results for β4–β6 confirm our analysis presented for Table 3, we shall
concentrate only on β7 to find out how competition affects the procyclicality of assets
growth in countries using individual macroprudential instruments. Intense competi-
tion in the banking industry is associated with decreased procyclicality of assets
growth in countries using ltv caps, dti, fc and rr rev, because the regression
coefficients on triple interaction terms of MPI x High competition x GDPG are
negative (columns 1, 2, 3 and 5 in Table 4)—ranging between 0.74 and 0.251,
and significant (but for ltv cap, in column 1). Such a result is in line with hypothesis
H2, that intense competition reduces procyclicality of assets of banks in countries
applying macroprudential instruments. This expectation is further supported for low
competition in the banking industry in countries using dti, because the β7 regression
coefficient is positive and significant at 10%.
Turning our attention to resilience-oriented macroprudential instruments (see
Table 5) we find support for hypothesis H1, that intense competition increases
procyclicality of asset growth in countries using dp, lev, inter and conc because
the β7 coefficients on triple interaction terms of MPI x High competition x GDPG are
positive and statistically significant. This result is additionally supported in less
competitive banking sectors in countries using conc, because the β7 on the triple
interaction term is negative and statistically significant and equals 0.511.
Two individual macroprudential policy instruments work countercyclically inde-
pendent of the competition level. They include fc (see columns 3 and 8 in Table 4)
and tax (see columns 5 and 10 in Table 5). Foreign currency limit are associated with
stronger reduction in procycality of asset growth in low competition countries than
Table 4 Role of competition intensity for the links between individual cyclical macroprudential policy instruments and asset growth and association between
158

cyclical instruments ant sensitivity of asset growth to business cycle

Role of high competition intensity Role of low competition intensity
MPI
instrument: ltv cap dti fc cg rr rev ltv cap dti fc cg rr rev
1 2 3 4 5 6 7 8 8 10
The role of MPI β1 1.246*** 1.526*** 0.968*** 0.864 0.807** 0.824*** 0.204 1.441*** 0.023 0.297
MPI and (0.27) (0.39) (0.353) (0.632) (0.35) (0.238) (0.313) (0.308) (0.506) (0.293)
competition High β2 0.293*** 0.294*** 0.239** 0.25** 0.335***
for assets competition
growth
(0.104) (0.104) (0.103) (0.104) (0.105)
MPI  High β3 1.286*** 2.383*** 0.701 0.75 1.742***
competition
(0.464) (0.537) (0.583) (0.812) (0.528)
Low β2 0.568*** 0.383* 0.696*** 0.64*** 0.424**
competition
(0.196) (0.196) (0.191) (0.194) (0.202)
MPI  Low β3 0.23 1.64*** 4.818*** 1.55* 0.14
competition
(0.615) (0.624) (0.943) (0.827) (0.602)
Analysis of GDPG β4 0.153*** 0.156*** 0.161*** 0.167*** 0.151*** 0.1*** 0.102*** 0.098*** 0.103*** 0.099***
the role of (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.009) (0.009) (0.009) (0.01)
MPI and MPI  β5 0.212*** 0.261*** 0.163*** 0.05 0.207*** 0.098** 0.008 0.127*** 0.03 0.076*
competition GDPG
for sensitivity
(0.038) (0.049) (0.046) (0.082) (0.035) (0.043) (0.058) (0.045) (0.097) (0.043)
of assets
growth to High com- β6 0.198*** 0.192*** 0.18*** 0.202*** 0.187***
business cycle petition 
GDPG
(0.022) (0.022) (0.022) (0.022) (0.023)
M. Olszak and I. Kowalska
 MPI  High 20.074 0.172 20.22***
competition
 GDPG
(0.08) (0.084) (0.093) (0.128) (0.077)
Low com- β6 0.149*** 0.161*** 0.205*** 0.199*** 0.156***
petition 
GDPG
(0.029) (0.029) (0.026) (0.026) (0.029)
MPI  Low β7 0.062 0.169* 20.559*** 20.16 0.04
competition
 GDPG
(0.08) (0.087) (0.13) (0.129) (0.068)
Wald test 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
p-value
R-sq: 0.88 0.88 0.88 0.88 0.88 0.88 0.88 0.88 0.88 0.88
overall
No of 73,066 73,066 73,066 73,066 73,066 73,066 73,066 73,066 73,066 73,066
observation
No of banks 7918 7918 7918 7918 7918 7918 7918 7918 7918 7918
Notes: This table presents a reduced view of model Eq. (1). Regression coefficients correspond to β1–β7 as presented in Sect. 3.1.; t-statistics are included in parentheses. *, **, ***
denote an estimate significantly different from 0 at the 10%, 5% and 1% levels, respectively
Does Competition Matter for the Effects of Macroprudential Policy on Bank. . .
159
Table 5 Role of competition intensity for the links between individual resilience oriented macroprudential policy instruments and asset growth and association
160

between cyclical instruments ant sensitivity of asset growth to business cycle

Role of high competition intensity Role of low competition intensity
MPI
instrument: dp lev inter conc tax dp lev inter conc tax
1 2 3 4 5 6 7 8 8 10
The role of MPI β1 1.314*** 0.629*** 0.318** 0.36 3.101*** 0.55 0.306*** 0.472*** 0.396*** 0.835***
MPI and (0.423) (0.141) (0.151) (0.224) (0.514) (0.612) (0.097) (0.107) (0.105) (0.14)
competition High β2 0.231** 0.281* 0.449** 0.045 0.057
for assets competition
growth
(0.103) (0.156) (0.182) (0.242) (0.127)
MPI  High β3 0.988 1.296** 0.897*** 0.342 2.626***
competition
(1.543) (0.625) (0.221) (0.328) (0.541)
Low β2 0.463** 0.388* 0.834*** 1.23** 0.749***
competition
(0.195) (0.201) (0.212) (0.503) (0.19)
MPI  Low β3 0.583 0.918 0.186 0.66 4.299***
competition
(0.877) (0.982) (0.607) (0.541) (1.277)
Analysis of GDPG β4 0.159*** 0.277*** 0.249*** 0.26*** 0.161*** 0.103*** 0.09*** 0.081*** 0.053*** 0.118***
the role of (0.01) (0.018) (0.019) (0.034) (0.01) (0.009) (0.015) (0.017) (0.019) (0.01)
MPI and MPI  β5 0.223*** 0.154*** 0.112*** 0.099*** 0.423*** 0.023 0.025 0.032 0.197*** 0.086**
competition GDPG
for sensitivity
(0.054) (0.021) (0.022) (0.035) (0.064) (0.138) (0.019) (0.02) (0.022) (0.042)
of assets
growth to High com- β6 0.186*** 0.312*** 0.214*** 0.392*** 0.141***
business petition 
cycle GDPG
(0.021) (0.026) (0.029) (0.041) (0.025)
M. Olszak and I. Kowalska
 MPI  High 0.309 0.595*** 0.052*** 0.439***
competition
 GDPG
(0.362) (0.151) (0.55) (0.058) (0.081)
Low com- β6 0.157*** 0.194*** 0.197*** 0.629*** 0.174***
petition 
GDPG
(0.029) (0.028) (0.03) (0.079) (0.026)
MPI  Low β7 0.193 0.141 0.149 20.511*** 20.252*
competition
 GDPG
(0.156) (0.13) (0.107) (0.084) (0.147)
Wald test 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
p-value
R-sq: 0.88 0.89 0.89 0.89 0.88 0.88 0.89 0.88 0.89 0.88
overall
No of 73,066 73,066 73,066 73,066 73,066 73,066 73,066 73,066 73,066 73,066
observation
No of banks 7918 7918 7918 7918 7918 7918 7918 7918 7918 7918
Notes: This table presents a reduced view of model Eq. (1). Regression coefficients correspond to β1–β7 as presented in Sect. 3.1.; t-statistics are included in parentheses. *, **, ***
denote an estimate significantly different from 0 at the 10%, 5% and 1% levels, respectively
Does Competition Matter for the Effects of Macroprudential Policy on Bank. . .
161
162 M. Olszak and I. Kowalska

in high competition countries, because the β7 the triple interaction term of MPI x
Low competition x GDPG is higher in absolute terms than respective coefficient on
MPI x High competition x GDPG. The opposite result is found for tax, which is
associated with stronger reduction of procyclicality in more competitive banking
industries.

5 Conclusions

This paper attempts to find out what is the role of competition in the effects of
macroprudential policy in a cross-country sample of commercial banks in the period
of 2004–2016. We focus on two areas of work of macroprudential policy that may be
potentially affected by competition. The first one is the area of banking activity
growth proxied with asset growth. The other is procyclicality proxied with the link
between bank asset growth and business cycle.
Our results show that competitive environment in the banking industry does
affect the conduct of macroprudential policy. High competition is associated with
increased asset growth in countries applying more macroprudential policy instru-
ments affecting risk taking by banks.
The association between competition and effects of macroprudential policy on
sensitivity of asset growth to business cycle differs between cyclical and resilience-
oriented tools. Intense competition in the banking industry is associated with
decreased procyclicality of assets growth in countries using cyclical instruments
including loan to value caps, debt to income ratios, credit growth limits and reserve
requirements. Such a result is line with competition-stability hypothesis, that intense
competition reduces procyclicality due to declined risk-taking incentives of banks in
more competitive industries.
In contrast, more competitive banking industry is associated with increased
procyclicality of asset growth in countries using resilience-oriented tools. Thus, for
these tools competition—fragility hypothesis is supported. In particular, this expec-
tation is verified in countries using dynamic provisions, leverage limits, interbank
transaction limits and concentration limits.
Our results therefore imply that the conduct of macroprudential policy may be
affected by the intensity of competition in the banking industry. However, the
association between competition and effects of macroprudential policy instruments
on cyclicality of banking growth depends on the type of instrument.

Acknowledgements We gratefully acknowledge the financial support provided by the National

Science Centre (NCN) in Poland, decision no. DEC-2016/23/B/HS4/02230. This paper’s findings,
interpretations, and conclusions are entirely those of the authors and do not necessarily represent the
views of the institutions with which the authors are affiliated.
Appendix

Table 6 Descriptive statistics per country

Country Obs Δassets GDPG Lerner_index borrower financial Country Obs Δassets GDPG Lerner_index borrower financial
ALBANIA 77 19 3.55 0.26 0 0.42 KENYA 260 16.84 5.33 0.37 0 2
ALGERIA 142 16.87 3.29 0.53 1.42 1 KOSOVO 10 14.17 3.85 0 0 2
ANGOLA 108 31.67 9.72 0.41 0 1.58 KUWAIT 20 10.13 3.82 0.56 0 2
ARGENTINA 532 9.21 4.23 0.27 0 5 KYRGYZSTAN 34 27.07 4.5 0.44 0 2
ARMENIA 133 27.39 5.69 0.32 0 3 LAO PEOPLE’S 12 25.17 7.71 0 0 2
DEM. REP.
AZERBAIJAN 106 33.84 11.03 0.3 0 0.69 LATVIA 171 13.04 3.06 0.31 0 2
BAHAMAS 41 1.69 0.62 0.34 0 0 LEBANON 135 14.11 4.77 0.04 0 2
BAHRAIN 43 16.23 5 0.28 0 3.07 LESOTHO 27 13.63 4.13 0 0 2
BANGLADESH 267 18.38 6.15 0.23 0 3.83 LITHUANIA 77 14.49 3.43 0.24 0 2
BELARUS 71 21.71 5.32 0.26 0 3.83 MACEDONIA 112 12.62 3.48 0.3 0 1.33
(FYROM)
BELGIUM 186 4.43 1.49 0.14 0 3.32 MALAWI 44 10.23 5.52 0.25 0 1
BELIZE 18 10.95 2.89 0.29 0 1 MALAYSIA 255 9.84 5.1 0.2 0 1
BHUTAN 22 10.01 7.37 0 0 1 MALTA 52 9.72 3.4 0.28 0 1
BOTSWANA 72 17.78 4.61 0.21 0 0 MAURITIUS 135 14.24 4.37 0.37 0 1
BRAZIL 691 15.16 3.11 0.22 0 0 MEXICO 142 20.22 2.4 0.62 0 1
BRUNEI 11 6.91 0.33 0 0 0 MONGOLIA 11 28.11 8.52 0.6 0 1
DARUSSALAM
BULGARIA 150 16.59 3.27 0.34 0 0 MONTENEGRO 70 23.31 3.11 0.02 0 1
Does Competition Matter for the Effects of Macroprudential Policy on Bank. . .

BURUNDI 28 16.89 3.48 0.32 0 0 MOROCCO 45 9.07 4.43 0.26 0 1

CAMBODIA 118 23.38 7.77 0.38 0 0 MOZAMBIQUE 99 19.17 7.43 0.25 0 1
CANADA 97 12.42 1.94 0.02 0 0.4 NEPAL 233 18.66 4.34 0.18 0 1.14
CAPE VERDE 42 11.61 4.55 0 0 0.67 NETHERLANDS 160 10.01 1.24 0.14 0 0.47
CHILE 85 17.32 4.29 0.23 0.71 1.43 NEW ZEALAND 88 9.51 2.45 0.18 0 0
163

(continued)
Table 6 (continued)
164

Country Obs Δassets GDPG Lerner_index borrower financial Country Obs Δassets GDPG Lerner_index borrower financial
CHINA 881 23.81 9.79 0.35 0.04 1.71 NORWAY 85 11.53 1.68 0.38 0.06 0.44
COSTA RICA 148 15.72 4.23 0.25 0 2 PAKISTAN 216 11.09 4.35 0.17 0 0.64
CROATIA 282 8.48 0.87 0.28 0.02 1.74 PARAGUAY 135 21.16 4.73 0.19 0 0.92
CURACAO 19 15.57 0 0.3 0.08 1.33 PERU 140 18.97 5.82 0.3 0 1.93
CYPRUS 45 6.92 1.08 0.3 0.06 1.78 PHILIPPINES 219 14.08 5.49 0.21 0 0.52
CZECH REPUBLIC 138 10.81 2.71 0.36 0 2.15 POLAND 275 13.18 3.97 0.31 0.67 0.73
DEM. REPUBLIC 71 22.02 6.62 0.14 0 3 PORTUGAL 150 0.03 0.12 0.19 0 1
OF CONGO
DOMINICAN 329 14.27 5.54 0.12 1.24 2.81 REPUBLIC OF 132 7.82 3.71 0.32 0 1
REPUBLIC KOREA
ECUADOR 186 11.69 4.41 0.23 0.5 1.33 REPUBLIC OF 111 16.9 4.29 0.3 0 1
MOLDOVA
ELSALVADOR 104 6.26 2.11 0.38 0.5 1.33 ROMANIA 147 18.11 3.4 0.25 0 1.81
ESTONIA 32 2.59 2.85 0.24 0.5 1.33 RUSSIAN 7116 10.02 3.4 0.21 0.06 1.15
FEDERATION
ETHIOPIA 84 15.53 10.91 0.54 0.5 1.33 SAINT KITTS AND 22 11.04 2.75 0 0 1
NEVIS
FINLAND 67 6.2 0.94 0.09 0.5 1.33 SERBIA 239 16.58 2.51 0.2 0 1
FRANCE 907 5.91 1.1 0.2 0.47 1.39 SINGAPORE 80 8.31 6.09 0.77 0 1
GAMBIA 20 14.23 3.41 0.24 0 0.42 SLOVAKIA 57 7.53 4.09 0.27 0 1
GEORGIA 17 10.72 5.55 0.31 0 0.42 SLOVENIA 126 3.82 1.65 0.21 0 1
GERMANY 10,775 4.02 1.39 0.11 0.01 0.4 SOUTH AFRICA 130 8.07 3.05 0.16 0 1
GHANA 148 21.34 6.88 0.37 0 0 SPAIN 254 9.96 0.96 0.33 0 1
HAITI 6 10.14 1.58 0.18 0 0 SRILANKA 130 16.7 6.21 0.22 0 1
HONDURAS 177 13.22 4.02 0.26 0 0 SWEDEN 193 9.34 2.19 0.32 0 1
HONGKONG 246 7.91 4.23 1.06 0 0 SWITZERLAND 962 8.39 2.1 0.16 0.2 0.53
HUNGARY 104 7.36 1.53 0.22 0 0 TAJIKISTAN 22 27.81 7.08 0 0 0
ICELAND 6 4.58 2.88 0.21 0 0 THAILAND 214 13.47 3.68 0.39 0.85 1.7
M. Olszak and I. Kowalska
INDIA 566 14.57 7.75 0.27 0 0 TRINIDAD AND 26 12.33 3.14 0.35 1 2
TOBAGO
INDONESIA 477 15.67 5.59 0.36 0 0 TUNISIA 128 5.54 3.46 0.43 1 2
IRELAND 71 7.69 4.6 0.25 0 0 TURKEY 185 10.8 5.92 0.21 1 2
ISRAEL 88 7.04 3.87 0.22 0 0 UGANDA 123 15.95 6.72 0.28 1 2
ITALY 555 7.58 0.14 0.14 0 1.07 UKRAINE 284 10.41 0.92 0.25 0.73 1.73
JAMAICA 45 4.21 0.26 0.34 0 2 UNITED ARAB 29 16.63 4.41 0.5 0 0
EMIRATES
JAPAN 1343 2.21 0.81 0.36 0 2 UNITED KINGDOM 857 5.2 1.47 0.31 0 0
JORDAN 12 4.78 5.13 0.36 0 2 UNITED STATES OF 59,138 6.67 1.83 0.27 0 2.41
AMERICA
KAZAKHSTAN 134 26.75 6.19 0.34 0 2
Does Competition Matter for the Effects of Macroprudential Policy on Bank. . .
165
166 M. Olszak and I. Kowalska

References

Acharya V, Richardson M (eds) (2009) Restoring financial stability: how to repair a failed system?
Wiley, New York
Aiyar S, Calorimis CW, Wieladek T (2014) Does macro-prudential regulation leak? Evidence from
a UK policy experiment. J Money Credit Bank 46:181–214. https://doi.org/10.1111/jmcb.
12086
Allen F, Gale D (2000) Comparing financial systems. MIT Press, Cambridge, MA
Allen F, Gale D (2004) Competition and financial stability. J Money Credit Bank 36(2):453–480
Altunbas Y, Binici M, Gambacorta L (2018) Macroprudential policy and bank risk. J Int Money
Financ 81(2018):203–220. https://doi.org/10.1016/j.jimonfin.2017.11.012
Bank for International Settlements (BIS), Financial Stability Board (FSB), International Monetary
Fund (IMF) (2011) Macroprudential policy tools and frameworks. Progress Report to G20,
27 October 2011
Beck T (2008) Bank competition and financial stability: friends or foes? Policy Research Working
Paper Series 4656. The World Bank
Beck T, Demirguc-Kunt A, Maksimovic V (2004) Bank competition and access to finance:
international evidence. J Money Credit Bank 36(3):627–648
Beck T, De Jonghe O, Schepens G (2013) Bank competition and stability: cross-country heteroge-
neity. J Financ Intermed 22:218–244
Berger AN, Klapper LF, Turk-Ariss R (2009) Bank competition and financial stability. J Financ
Serv Res 35:99–118
Blundell R, Bond S (1998) 1009. Initial conditions and moment restrictions in dynamic panel data
models. Journal of Econ 87(1):115–143. https://doi.org/10.1016/S0304-4076(98)00009-8
Borio C (2003) Towards a macroprudential framework for financial supervision and regulation?
BIS Working Papers No 128. Bank for International Settlements, Basel
Borio C (2014) The financial cycle and the macroeconomics: what have we learnt? J Bank Financ
45(C):182–198
Borio C, Zhu VH (2012) Capital regulation, risk-taking, and monetary policy: a missing link in the
transmission mechanism? J Financ Stab 8:236–251. https://doi.org/10.1016/j.jfs.2011.12.003
Borio C, Furfine C, Lowe P (2001) Procyclicality of the financial system and financial stability:
issues and policy options. BIS Papers No 1. Bank for International Settlements, Basel
Boyd JH, De Nicoló G (2005) The theory of bank risk taking and competition revisited. J Financ
60:1329–1343
Boyd JH, De Nicolo G, Jalal A M (2009) Bank competition, risk and asset allocations (July 2009).
IMF Working Paper No. 09/143, Available at SSRN: https://ssrn.com/abstract=1442245
Bretschger L, Kappel V, Werner T (2012) Market concentration and the likelihood of financial
crises. J Bank Financ 36(12):3336–3345
Carletti E, Hartmann P (2003) Competition and financial stability. what’s Special about Banking?
In: Mizen P (ed) Monetary history, exchange rates and financial markets: essays in honor of
Charles Goodhart, vol 2. Edward Elgar, Cheltenham
Cerutti E, Claessens S, Laeven L (2015) The use and effectiveness of macroprudential policies: new
evidence. IMF Working Paper IMF/15/61
Cerutti E, Claessens S, Laeven L (2017) The use and effectiveness of macroprudential policies: new
evidence. J Financ Stab 28:203–224. https://doi.org/10.1016/j.jfs.2015.10.004
Cipollini A, Fiordelisi F (2012) Economic value, competition and financial distress in the European
banking system, J Bank Finan 36:3101–3109
Claessens S (2014) An overview of macroprudential policy tools. IMF Working Paper No WP/14/
214
Claessens S, Ghosh SR, Mihet R (2014) Macro-Prudential Policies to Mitigate Financial System
Vulnerabilities. IMF Working Paper WP/14/155
Does Competition Matter for the Effects of Macroprudential Policy on Bank. . . 167

Committee on the Global Financial System CGFS (2012) Operationalising the selection and
application of macroprudential instruments. CGFS Papers No 48. Bank for International
Settlements
Craig B, Dinger V (2013) Deposit market competition, wholesale funding, and bank risk. J Bank
Financ 37:3605–3622
Cuestas JC, Lucotte Y, Reigl N (2020) Banking sector concentration, competition and financial
stability: the case of the Baltic countries. Post-Communist Econ 32(2):215–249
Danielsson J, Macrae R, Tsomocos D, Zigrand J-P (2016) Why macropru can end up being
procyclical. VOX EU/CEPR, 15 December 2016. https://voxeu.org/article/why-macropru-can-
end-being-procyclical
Dell’Ariccia G, Deniz I, Laeven L, Tong H, Bakker B, Vandenbussche J (2012) Policies for
macrofinancial stability: how to deal with credit booms. IMF Staff Discussion Note 12/06
Dick A (2006) Nationwide branching and its impact on market structure, quality and bank
performance. J Bus 79:567–592.
European Systemic Risk Board (ESRB) (2014a) The ESRB handbook on operationalising macro-
prudential policy in the banking sector
European Systemic Risk Board (ESRB) (2014b) Flagship report on macro-prudential policy in the
banking sector
Gaganis C, Lozano-vivas A, Papadimitri P, Pasiouras F (2020) Macroprudential policies, corporate
governance and bank risk: cross-country evidence. J Econ Behav Organ 169:126–142. https://
doi.org/10.1016/j.jebo.2019.11.004
Gambacorta L, Murcia A (2020) The impact of macroprudential policies in Latin America: an
empirical analysis using credit registry data. J Financ Intermed 42(April 2019):100828. https://
doi.org/10.1016/j.jfi.2019.04.004
Gómez E, Murcia A, Lizarazo A, Carlos J (2020) Evaluating the impact of macroprudential policies
on credit growth in Colombia. J Financ Intermed 42(August 2019):100843. https://doi.org/10.
1016/j.jfi.2019.100843
Hellmann T, Murdock K, Stiglitz J (2000) Liberalization, moral hazard in banking and prudential
regulation: are capital requirements enough? Am Econ Rev 90:147–165
International Monetary Fund (IMF) (2013) Key aspects of macroprudential policy – Background
paper. IMF Policy Paper. Washington, International Monetary Fund
Keeley MC (1990) Deposit insurance, risk and market power in banking. Am Econ Rev
80:1183–1200
Kuttner K, Shim I (2013) Can non-interest rate policies stabilise housing markets? Evidence from a
panel of 57 countries. BIS Working Paper No 433, Bank for International Settlements
Lee LH, Chiang YY (2012) Loan market competition and bank stability – a re-examination of
banking competition and risk taking. Working Paper. National Chengchi University
Leroy A, Lucotte Y (2019) Competition and credit procyclicality in European banking. J Bank
Financ 99:237–251. https://doi.org/10.1016/j.jbankfin.2018.12.004
Lim C, Columba F, Costa A, Kongsamut P, Otani A, Saiyid M, Wezel T, Wu X (2011)
Macroprudential policy: what instruments and how to use them? Lessons from country expe-
riences. IMF Working Paper 11/238
Liu H, Molyneux P, Wilson JOS (2010) Competition and stability in European Banking – a regional
analysis. Bangor Business School Working Paper BBSWP/10/019
Marcus AJ (1984) Deregulation and bank policy. J Bank Financ 8:557–565
Martinez-Miera D, Repullo R (2010) Does competition reduce the risk of bank failure? Rev Financ
Stud 23(10):3638–3664
Minsky H (1986) Stabilizing an unstable economy. Mc Graw Hill
Mirzaei A, Moore T, Liu G (2013) Does market structure matter on banks’ profitability and
stability? Emerging versus advanced economies. Journal of Bank Financ 37:2920–2937
Olszak M, Kowalska I, Roszkowska S (2018) Macroprudential policy instruments and
procyclicality of loan-loss provisions – cross-country evidence. Int J Financ Inst Markets
Money. https://doi.org/10.1016/j.intfin.2018.01.001
168 M. Olszak and I. Kowalska

Repullo R (2004) Capital requirements, market power, and risk-taking in banking. J Financ
Intermed 13:156–182
Roodman D (2009) Practitioners corner: a note on the theme of too many instruments. Oxf Bull
Econ Stat 71:135–156. https://doi.org/10.1111/j.1468-0084.2008.00542.x
Salas V, Saurina J (2003) Deregulation, market power and risk behavior in Spanish banks. Eur Econ
Rev 47:1061–1075
Schaeck K, Cihák M (2010) Competition, efficiency and soundness in banking: an industrial
organization perspective. Tilburg University European Banking Center Discussion Paper
2010–20S
Schaeck K, Cihák M, Wolfe S (2009) Are competitive banking systems more stable? J Money
Credit Bank 41:711–734
Stiglitz J, Weiss A (1981) Credit rationing with imperfect information. Am Econ Rev 71:393–410
Tabak BM, Fazio D, Cajueiro DO (2012) The relationship between banking market competition
and risk-taking: do size and capitalization matter? J Bank Financ 36(12):3366–3381
Vandenbussche J, Vogel U, Detragiache E (2012) Macroprudential policies and housing prices—a
new database and empirical evidence for Central, Eastern, and Southeastern Europe. IMF
Working Paper 12/303
Systemic Risk in Selected Countries
of Western and Central Europe

Marta Karaś and Witold Szczepaniak

1 Introduction

Monitoring and maintaining financial stability is one of the goals of central banks
around the world. The experience shows that monetary authorities are even willing
to adjust short- and medium-term inflation targets to avoid destabilization. Such an
approach was uncommon before the global financial crisis. Financial stability
authorities aim to oversee systemic risk. To be successful in this respect, they need
to employ various methods to identify, measure, and manage risks that encompass
the financial systems under their jurisdiction. To form informed views on risk, the
regulators and supervisors should combine the probability aspect with the impact
aspect. In simple terms, they need to know the likelihood, the expected magnitude,
and the expected impact of risk materialization on the financial system and the real
economy. Despite a heated dispute, no single golden standard on quantifying
systemic exists (for an overview, see Bisias et al. (2012), Hattori et al. (2014), or
Benoit et al. (2017) or Karaś (2019).

This work was supported by the NCN Grant 2018/29/N/HS4/02783. The early draft of the paper
was presented in the closed seminar of the National Bank of Poland in Warsaw, as well as in the
EuroConference 2018 Emerging Market Economies and International Risk Management
Conference in Milan 2019. The authors are thankful for all the comments that enabled to
improve the paper.

M. Karaś (*) · W. Szczepaniak

Department of Financial Investment and Risk Management, Wrocław University of Economics,
Wrocław, Poland
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 169
K. Jajuga et al. (eds.), Contemporary Trends and Challenges in Finance, Springer
Proceedings in Business and Economics,
https://doi.org/10.1007/978-3-030-73667-5_10
170 M. Karaś and W. Szczepaniak

The paper adds to the dispute by presenting the empirical results of systemic risk
measured based on the changing prices of assets of systemically important financial
institutions in the studied countries. The research is the first attempt to measure
systemic risk with quantile-based methods for this particular set of countries using
the data related to systemically important financial institutions and a methodology
that compares the results on a cross-country level.
The layout of the paper is a follows. First, we define systemic risk and outline the
methods of its quantification used in the study. Then we present and briefly discuss
the results obtained for Germany, Poland, Slovakia, Hungary, Romania, and
Bulgaria. We conclude with indications, suggestions, and further research ideas.

2 Relevant Definitions

Commonly, the literature defines financial stability by associating it with some lack
of (negative) risk materialization or with the system’s low sensitivity to adverse
events (see, e.g., Cihak 2007; Liang 2013). Additionally, from the macroprudential
viewpoint, one thinks of lowering the possible threat of the financial crisis.
Following this concept, we define systemic risk as a risk that “concerns a large
part of the financial system or a significant number of financial institutions and is
considered to disrupt the performance of the financial system and its functions, such
as financial intermediation” (Smaga 2014, p. 4). This definition allows us to model
systemic risk like the reaction of financial systems to low probability events (i.e.,
systemic triggers) related to the losses exceeding the Value at Risk at financial
institutions.
We define the financial system as a system of interconnected financial institutions
and markets (cf. Matysek-Jędrych 2007; Jajuga et al. 2017). More precisely, we treat
each financial system as a collective of (only) Systemically Important Financial
Institutions (SIFI). Each national macroprudential regulator identifies such institu-
tions for their country, and we use these institutions in the study (see Table 2 in the
Appendix). Therefore, we take the macroprudential approach, according to which
only SIFIs play a prominent role in the transmission of the systemic risk triggers:
Systemically Important Financial Institutions are institutions whose distress or disorderly
failure, because of their size, complexity, and systemic interconnectedness, would cause
significant disruption to the wider financial system and economic activity. To avoid this
outcome, authorities have all too frequently had no choice but to forestall the failure of such
institutions through public solvency support. (Financial Stability Board 2011, 1).

A crucial characteristic of the studied emerging financial systems is their struc-

ture. They are not very complex, with relatively small stock exchanges and a limited
number of financial instruments in use. In effect, the intermediation of funds in these
countries tends to concentrate in the banking sector. These characteristics are
especially true for the emerging countries studied in this paper. Consequently, we
Systemic Risk in Selected Countries of Western and Central Europe 171

focus on the condition of the banking sector, which is the most relevant systemic risk
area for the studied region.

3 Challenges of Quantile-Based Systemic Risk

Measurement in Developing Europe

For emerging Europe, the banking sector is the most central part of the financial
system and “the main source of risk for financial stability” (Karkowska 2013, 4).
Moreover, in this region, foreign owners control over 90% of the total banking assets
(Radulescu et al. 2018, 7–8). As a consequence, European emerging countries have
relatively homogeneous financial systems, where foreign-owned banks are the
primary credit suppliers (Dumicic 2018, 2).
In all analyzed countries (except Poland that has the most developed stock
exchange in the sample), almost none of these foreign-owned banks are locally
listed. This characteristic renders quantile-based systemic risk measures unsuitable
since they require a rich pole of stock-market data. We overcome this issue by
applying the methodology outlined in Karaś and Szczepaniak (2020).
Since 2015, the European Union requires national macroprudential bodies to
identify Other Systematically Important Institutions (O-SIIs). These are the “insti-
tutions that, due to their systemic importance, are more likely to create risks to
financial stability whilst maximizing private benefits through rational decisions [and]
these institutions may bring negative externalities into the system and contribute to
market distortions” (European Banking Authority 2014). Said identification is made
based on Systemic Importance Scores (SIS).
The regulators in emerging Europe do not use quantile-based systemic risk
measures that combine bank-level and market-level data. One exception is the
National Bank of Poland that uses a relatively elaborate early warning model
based on the estimated output gap (Narodowy Bank Polski 2019). Research-wise,
only a handful of studies attempt to apply quantile-based systemic risk measures in
the studied region. In all these cases, a significantly limited sample of systemically
important banks was used: they exclude many foreign-owned banks that are not
listed on the domestic stock exchanges. On the other hand, some of the studies
include the banks that are not systemically important, potentially blurring the
systemic risk view. Thus, the previous results are incomparable to the results
presented in this paper.
Table 1 presents the overview of all the studies published in the last decade that
use the described type of systemic risk measures for the region, to the best of our
knowledge.
Table 1 indicates that previous studies were ineffective in capturing the effect of
O-SIIs. In all cases except Poland, the O-SIIs fraction is marginal. The last column
indicates that our methodology significantly increases the number of the analyzed
OSII-s, making SRISK and CoVaR effectively applicable for Central and
 172

Table 1 The systemic risk measures that combine bank-level and stock market data: application to selected emerging European countries
Bulgaria Hungary Poland Romania Slovakia
Number of O-SIIs in the system 10 8 12 9 5
The number of analyzed O-SIIs per study: Period Frequency Measure
Karkowska (2013) 3 1 8 1 0 2006–2012 Quarterly CCA
Engle et al. (2019) 0 1 8 2 1 2000–2019 Monthly SRISK
Jajuga et al. (2017) 2 1 8 2 2 2005–2016 Daily SRISK, CoVaR
Andrieş et al. (2018) 3 1 9 2 2 2005–2012 Weekly
This paper 9 6 8 7 4 2006–2018 Daily
Source: own elaboration based on lists of O-SIIs 2015–2019 (European Banking Authority 2019) and sources cited in the table
M. Karaś and W. Szczepaniak
Systemic Risk in Selected Countries of Western and Central Europe 173

South-Eastern Europe. The banks that remain beyond the scope of the study are
institutions whose ownership structure changed during the sample period (due to,
e.g., mergers) or is else incompatible with the assumed methods (e.g., credit unions).

4 Selected Risk Measures and Estimation Methods

The study applies two types of well-established quantile-based risk measures: one
derived from Value at Risk and three based on Expected Shortfall. The models
selected for the empirical analysis include1:
– Conditional VaR (CoVaR) (Adrian and Brunnermeier (2016); as modified by
Karaś and Szczepaniak (2017, 2020)),
– Expected Shortfall, Marginal Expected Shortfall (Acharya et al. 2017),
– the Long Run Expected Shortfall of the system (Acharya et al. 2012),
– SRISK (Brownlees and Engle 2017).
The selection of these measures is not incidental. On the one hand, these measures
have the advantage of having been successfully used to measure systemic risk
around the time of the global financial crisis for advanced economies, such as e.g.,
the USA, the U.K., or France. Moreover, all the selected measures allow producing
currency-valued outputs, providing transparent comparative results on a cross-
country level.
Moreover, we expect these measures to produce varying results because they
measure different aspects of systemic risk. CoVaR is a contagion-focused measure
that measures risk conditionally on its co-occurrence in multiple financial institu-
tions. At the same time, SRISK is a fragility-focused measure that depends heavily
on leverage.
Finally, it is worth investigating whether the more complex measures, such as
those mentioned above, give a different overview of systemic risk than the relatively
more straightforward measures, such as the Expected Shortfall. Such property is
reported for advanced economies (Acharya et al. 2017; Adrian and Brunnermeier
2016; Benoit et al. 2017; Brownlees and Engle 2017). However, it may not neces-
sarily be valid for emerging markets.
We apply several computation modifications, inter alia, using2:

1
This paper presents a fraction of a larger-scale research project and is based on the solutions
developed and results obtained previously (c.f. Karaś and Szczepaniak 2017; Jajuga et al. 2017;
Karaś and Szczepaniak 2020).
2
The methodology has been developed in the course of the larger research project. A paralely
printed paper (Karaś and Szczepaniak 2020) presents the details regarding mentioned modifica-
tions, their justification and the analysis on how the introduced changes affect the data input and
output. It is an elaborate discussion incompatibile with the space limits and the nature of the current
empirical paper, and as such it is beyond the scope of the study presented here.
174 M. Karaś and W. Szczepaniak

– econometric GARCH models (as Engle et al. 2019),

– stock market data to obtain CoVaR values (same as Benoit et al. 2014 or Karaś
and Szczepaniak 2017),
– Systemic Importance Scores for weights of financial institutions in the financial
system model (Karaś and Szczepaniak 2020), and
– a set of proxying methods (Karaś and Szczepaniak 2020).
Marginal Expected Shortfall (MES) (Acharya et al. 2017) indicates the system’s
extreme contribution to systemic risk. Thus, MES indicates if the expected shortfall
of the system S will change once the i entity’s share in it changes in the extreme. It is
the marginal measure of the Expected Shortfall (ES):
XN
ESSt ðC Þ ¼ E t1 ðr St jr St < CÞ ¼ i¼1
wit Et1 ðr it jr St < CÞ,

where C—a quantile of the distribution of system returns rS equal to VaRqs,t , for
q ¼ 1%.
The Marginal Expected Shortfall (used to compute the Long Run MES (LRMES)
in this study (c.f. Acharya et al. 2012) is defined as a partial derivative:

∂ESSt ðC Þ
MESit ðC Þ ¼ ¼ Et1 ðr it jr St < C Þ:
∂wit

LRMES equals:

LRMESit ðCÞ  1  exp ðγ  MESit ðCÞÞ,

where γ—is the correcting factor relative to the length of the assumed horizon.
The SRISK (Brownlees and Engle 2017) determines the expected shortage of
equity in the event of a systemic crisis. It is based on LRMES that indicates the
expected decline in the equity of an institution if the equity of the financial system
falls below the assumed marginal threshold (within the next 6 months). The follow-
ing formula defines this measure:

SRISK it ¼ max ½0; kðDit þ ð1  LRMESit ðC ÞÞW it Þ  ð1  LRMESit ðCÞÞW it ,

where:
Di, t—value of debt
Wi, t—market value of equity,
k—prudential capital fraction.
The last measure used in the study is CoVaR (Adrian and Brunnermeier 2016). It
is a Conditional Value at Risk to the system, provided that there is a threat to the
financial condition in the analyzed entity (cf. Benoit et al. 2014):
Systemic Risk in Selected Countries of Western and Central Europe 175

 
P r St  CoVaRqSt jr it  VaRqit ¼ q:

We apply Delta CoVaR (c.f. Karaś and Szczepaniak 2017) for empirical com-
parisons. It is the difference between the system’s value at risk if the given financial
institution is financially at risk and the system’s value at risk if the financial position
of the given entity is normal (average, for instance, median):
   
ΔCoVaRqit ¼ CoVaRqSt jr it ¼ VaRqit  CoVaRqSt jr it ¼ VaR0:5
it :

In order to estimate the analyzed systemic risk measures, a two-dimensional

process of retaining the rates of return of the system S and institution i was adopted:
pffiffiffiffiffi
rt ¼ H t υt ,

where
rt—the vector of (rs, t, ri, t),
Ht—the conditional variance-covariance matrix of the form:

!
σ 2St σ it σ St ρit
Ht ¼ ,
σ it σ St ρit σ 2it

with a conditional standard deviation of the rate of return of the system σ St and the
institution σ it, and the conditional correlation ρit; while υt is a vector (εit, εSt) of
independent
  random variables with the same distribution, such that E(υt) ¼ 0 and
E υt υ0t ¼ I 2 is a two by two unit matrix (cf. Benoit et al. 2014). Conditional
volatility of the rates of return of the system σ st and institution σ it was estimated
based on the GJR-GARCH model. In contrast, the conditional correlation of the
institution and the system ρit is based on the GJR-GARCH DCC model (cf. Engle
et al. 2019). To obtain the individual conditional expected value for the institution i,
we use the estimator:

VaRqit ¼ σ it F i 1 ðqÞ

For the institution’s contribution to the CoVaR of the system, we adopt the
estimator:
 
ΔCoVaRqit ¼ bγ VaRqit  VaR0:5
it ,

γ ¼ bρitbσ St . The MES was estimated based on the estimator:

where: b
bσit
176 M. Karaś and W. Szczepaniak

  qffiffiffiffiffiffiffiffiffiffiffiffiffi
MESit VaRqSt ¼ b
σ it b b t1 ðεSt jεSt < κÞ þ b
ρit E σ it 1  b ρ2it Eb t1 ðεit jεSt < κÞ,

where:
PT 
κεSt

t¼1 K εSt
b t1 ðεSt jεSt < κ Þ ¼
E PT κε h
 ,
t¼1 K
St
h
PT 
κεSt

t¼1 K εit
b t1 ðεit jεSt < κ Þ ¼
E PT κε h
 ,
t¼1 K
St
h

VaRq Rx
for κ ¼ σSt St , K ðxÞ ¼ 1
h
kðuÞdu for normal distribution density function k(u) and
1
h ¼ T . In turn, the long-term marginal shortfall for the surveyed institutions was
5

determined based on the following estimator, as proposed originally by the authors

of the SRISK measure:

LRMESit ðC Þ  1  exp ð18  MESit ðC ÞÞ:

The measures put forward allow calculating the risk from the perspective of an
individual financial institution. However, they can be used to measure the risk of the
whole system (see: the applications developed by Jajuga et al. (2017, pp. 54–62) and
by Karaś (2019)). This process is a two-step calculation. Firstly, we measure the risk
of each OSII. Secondly, we aggregate the results, taking into account the relevance
of the individual institutions in each financial system, proxied with their Systemic
Importance Score. The choice of institutions follows the regulators’ suggestions who
identify OSIIs for the analyzed region (European Banking Authority 2019). The set
of such largest and most interconnected financial institutions, i.e., the banks included
in the study, is presented in Table 2 in the Appendix.

5 Empirical Results

The following section entails the discussion of the empirical results obtained in the
course of the study. The sample period includes the years 2006–2018. Interesting
sub-periods are:
– the global financial crisis,
– the public debt crisis,
– the related economic slowdown.
The raw data used for our calculations include the quotations from the stock
markets. Data were obtained from the Thomson Reuters Datastream, while all the
calculations were conducted in the MATLAB environment. In the remainder of the
paper, we discuss the results ordered per the measurement method.
Systemic Risk in Selected Countries of Western and Central Europe 177

5.1 Expected Shortfall of the Financial Systems

Expected Shortfall (ES) of the financial system depicts the average of all the losses
which are greater or equal to 1% Value at Risk, realized by financial institutions
systemically important for a given country. Plotted as a time series, the aggregate ES
shows the financial system’s exposure to a loss, conditional on the institutions’
equity being in distress relative to its average condition.
ES may be analyzed in nominal terms—we use local currencies, as the local
perspective on risk is taken. This way, we observe the size of the total potential loss
relative to the total financial institutions’ equity value. Therefore, the levels of
systemic risk measured with ES are high only when a distressful state concerns
either a very large financial institution or many smaller institutions at the same time.
As presented in Fig. 1, we observe the most significant risk peaks around the
global financial crisis for all the countries. The increased risk is also visible between
2010 and 2014. However, in this case, the moments of the risk peaking vary between
the analyzed countries. The earliest reaction is observed for Germany and the latest
for Bulgaria, suggesting that economic proximity to most indebted European coun-
tries might have worked as a catalyst for the distress. Notably, all countries recorded
a significant lowering of volatility clustering and of the absolute level of risk after
2010.
The scale of systemic risk is closely related to the total value of assets in a given
financial system. Thus, the highest risk peaks are significantly high in nominal terms
for Germany, medium for Poland, and smaller for Bulgaria, Romania, Hungary, and
Slovakia.3 Nonetheless, when we refer these levels to the size of each economy, we
see that the potential burden of systemic risk materialization in each of the countries
is similar.

5.2 Long Run Marginal Expected Shortfall of the Financial

Systems

Marginal Expected Shortfall measures the institution’s expected equity loss when
the market falls below a certain threshold over a given time horizon. If computed as
an aggregate of all systemically important financial institutions in a given country, it
indicates the total loss of the financial system conditional on a marginally probable
systemic event. When the long-run perspective is considered, only the most pessi-
mistic scenarios for the market return are considered, i.e., the market index falling by
40% over the next 6 months.

3
The size of the highest peak observed for Slovakia is driven by one outlaying observation. Given
the model specificity it cannot be excluded that it is a result of model risk materialization, and as
such is not interpreted here. The remaining peaks correspond to fundamental events as is the case
for all other observation for the remaining countries.
178 M. Karaś and W. Szczepaniak

Fig. 1 ES of selected financial systems for years 2006–2018

The empirical results show an interesting pattern of relative changes in systemic

risk. When we consider the marginal scenario, the potential losses are almost the
same for all the analyzed counties. This observation suggests a high level of
convergence between all the studied systems when facing a financial crisis (Fig. 2).
The long-run perspective significantly changes the systemic risk horizon,
pointing to the fact that simple risk measures, such as the Expected Shortfall
aggregated over a set of SIFIs, are not broad enough to effectively measure systemic
risk for the analyzed region—which is in accordance with the expectations based on
the literature review.
In all the observations (all the measures), Slovakia and Romania stand out with
the highest mean volatility but much less time-persistent peaks. For Slovakia, this
Systemic Risk in Selected Countries of Western and Central Europe 179

Fig. 2 LRMES of selected financial systems for years 2006–2018

may be explained by a smaller number of OSIIs. For Romania, the unstable flow of
foreign investment and variable currency may be responsible. In these systems, the
reactions to changes in the funding structure (as is the case in the second quarter of
2013) show a significant effect on the equity value. Thus, even if the size of
the negative peaks should be interpreted with caution, the general characteristics
of the systemic risk drivers in Slovakia and Romania seem different than elsewhere
in the geographical neighborhood.
180 M. Karaś and W. Szczepaniak

5.3 SRISK of the Financial Systems

SRISK, an extension of the measures presented above, allows us to consider the

liabilities and size of each system’s financial institutions. In other words, it enables
considering the levels of leverage, so crucial for the systemic risk perspective. Here
the expected capital shortfall of a given financial institution is conditional on the
systemic event. As such, it allows computing the capital that the financial system
(e.g., the regulator) is expected to need (for a possible bailout) if the systemic risk
materializes into a financial crisis (Fig. 3).

Fig. 3 SRISK of selected financial systems for years 2006–2018

Systemic Risk in Selected Countries of Western and Central Europe 181

It seems interesting to investigate the size of such potential losses in nominal

terms and relative to each financial system size. We present the results relative to the
capitalization of each financial system. This way, we show the scale of the potential
losses relative to the system’s total capacity to handle them (market depth).
There seem to be two types of different countries. The first one relates to Poland
that has a rather stable and safe condition in quiet times (very low base-line SRISK),
and reacts in a significant manner only to global events. Meanwhile, Hungary,
Romania, Slovakia, and Bulgaria show properties of oversized and overleveraged
financial sectors, where the loss potential remains high at all times. That is notwith-
standing also very significant reactions to global events. Germany has incomparably
higher nominal SRISK, which relates to the size of its financial system.
Another interesting observation is that SRISK truly individualizes between the
countries and monitors the risk changes over time quite smoothly. This is evident,
for instance, when we look at Germany and Hungary, where the significant trending
growth of leverage in the financial system is visible. Such an observation is not
present for other measures in the sample period.

5.4 Conditional Value at Risk of the Financial Systems

The final measure analyzed in this study, CoVaR, also corresponds to the Value at
Risk of the given financial system, conditional on its financial institutions’ distress.
However, in this case, the institution’s contribution to systemic risk is computed as
the difference between the CoVaR conditional on the institution’s equity quoted
below VaR and the CoVaR of this institution with equity prices being in the median
state.
The results point to the same periods of distress for the analyzed countries as other
measures. However, they show similar risk patterns for Germany, Poland, and
Hungary, with higher variability for Romania and Bulgaria. Also, the scale of risk
seems to be directly related to the number and the total size of internationally
connected banks in a given financial system, showing high contagion potential
among the emerging European countries, a similar size in scale as for Germany
(Fig. 4).
Importantly, we may note that the peaks in the emerging financial systems follow
the peaks of the German system. As the CoVaR is a risk-spillover measure, it is not
surprising that it allows capturing sequentiality of the systemic risk peaks. Such
sequentiality might be indicative of an existing contagion channel, which should be
investigated further in future studies.
182 M. Karaś and W. Szczepaniak

Fig. 4 CoVaR of selected financial systems for years 2006–2018

6 Conclusions

The analysis presented in the paper focuses on measuring the levels of systemic risk
for selected countries of the Central part of Europe. The study covered geographi-
cally affiliated countries that are developed differently and have very diverse finan-
cial systems in terms of complexity and innovation. Several applied measures
showed commonalities in terms of risk between various countries. The results also
indicate that these commonalities are not persistent when we use different systemic
risk measures.
Systemic Risk in Selected Countries of Western and Central Europe 183

From the policymakers’ perspective, this gives several vital conclusions. Above
all, the measures presented in the paper can be successfully used in systemic risk
monitoring in all studied countries. The second conclusion is that further studies are
required to specify what differentiates the results produced by various measures.
Such studies would conclusively indicate which of the measures would best serve
each specific goal of the regulators and monetary authorities.
What is visible, fragility and contagion relate directly to the size of leverage
relative to the size of the financial system and the presence (importance measured
with SIS) of internationally linked big banks. We also see that the scale of risk in
Central and South-Eastern Europe is as big as in Germany. At the same time, we
know that emerging European countries cannot carry the same bailout burden in case
a crisis materializes. All of this sheds light on the directions that regulators should
take in their macroprudential policies for the analyzed region.
The results show that the global financial crisis and the European debt crisis have
affected the levels of risk in all the analyzed countries, while the scale of risk is much
bigger for these countries where the financial system is more interconnected with the
global markets. Additionally, the role of the German financial system might be
significant for systemic signal transmission, especially given the pace of reaction
to adverse global events. This observation calls for further research into Germany as
a contagion channel for the analyzed countries and the whole CEE region.

Appendix

Table 2 Systemically Important Institutions for the analyzed region listed in stock markets
(directly or by proxy)
Ultimate E.U. Parent (owns or
Bank Score controls)
Bulgaria UniCredit Bulbank A.D. 1880 UniCredit S.p.A.
United Bulgarian Bank A.D. 1120 KBC Group N.V.
First Investment Bank A.D. 1100 X
DSK Bank A.D. 1040 OTP Bank Nyrt.
Societe Generale Expressbank A.D. 720 Société Générale S.A.
Raiffeisenbank (Bulgaria) A.D. 670 Raiffeisen Bank International A.G.
Eurobank Bulgaria A.D. 630 Eurobank Ergasias S.A.
Central Cooperative Bank A.D. 520 X
Piraeus Bank Bulgaria A.D. 310 Piraeus Bank S.A.
Hungary OTP Bank Nyrt. 3095 X
UniCredit Bank Hungary Zrt. 960 UniCredit S.p.A.
Kereskedelmi és Hitelbank Zrt. 830 KBC Group N.V.
ERSTE BANK HUNGARY Zrt. 655 Erste Group Bank A.G.
Raiffeisen Bank Zrt. 600 Raiffeisen Bank International A.G.
CIB Bank Zrt. 420 Intesa San Paolo S.p.A.
(continued)
184 M. Karaś and W. Szczepaniak

Table 2 (continued)
Ultimate E.U. Parent (owns or
Bank Score controls)
Poland PKO BP S.A. 1580 X
Bank Polska Kasa Opieki S.A. 1050 X
Bank Zachodni WBK S.A. 960 Banco Santander
ING Bank Śląski S.A. 950 ING Bank N.V.
mBank S.A. 930 Commerzbank A.G.
Millennium Bank S.A. 424 Banco Comercial Portugues
Bank Handlowy w Warszawie S.A. 440 X
Deutsche Bank Polska S.A. 400 Deutsche Bank A.G.
Romania Banca Transilvania S.A. 1620 X
UniCredit Bank S.A. 1525 UniCredit S.p.A.
Banca Comercială Română S.A. 1390 Erste Group Bank A.G.
BRD—Groupe Societe Generale S. 1165 Société Générale S.A.
A.
Raiffeisen Bank S.A. 1000 Raiffeisen Bank International A.G.
Alpha Bank România S.A. 445 Alpha Bank
OTP Bank Romania S.A. 305 OTP Bank Nyrt.
Garanti Bank S.A. 300 Turkiye Garanti Bankasi A.S.
Slovakia Všeobecná Úverová Banka A.S. 2070 Intesa San Paolo S.p.A.
Slovenská Sporiteľňa A.S. 1800 ERSTE Group Bank A.G.
Tatra Banka A.S 1390 Raiffeisen-Landesbanken-Holding
GmbH
Československá Obchodná Banka 1205 KBC Group N.V.
A.S.
Germany Deutsche Bank A.G. 2765 X
Commerzbank A.G. 830 X
Unicredit Bank A.G. 470 UniCredit Group
ING DiBa A.G. 145 ING Bank N.V.

References

Acharya V, Engle R, Richardson M (2012) Capital shortfall: a new approach to ranking and
regulating systemic risks. Am Econ Rev 102(3):59–64
Acharya V, Pedersen L, Philippon T, Richardson P (2017) Measuring systemic risk. Rev Financ
Stud 30(1):2–47
Adrian T, Brunnermeier MK (2016) CoVaR. Am Econ Rev 106(7):1705–1741
Andrieş AM, Nistor S, Sprincean N (2018) The impact of central bank transparency on systemic
risk—evidence from central and Eastern Europe. Res Int Bus Financ 50
Benoit S, Colletaz G, Hurlin Ch, Perignon Ch (2014) A theoretical and empirical comparison of
systemic risk measures. HEC Paris Research Papers (FIN-2014-1030), 42
Benoit S, Colliard J-E, Hurlin C, Perignon C (2017) Where the risks lie: a survey on systemic risk.
Rev Financ 21(1):109–152
Bisias D, Flood M, Lo AW, Valavanis S (2012) A survey of systemic risk analytics. Annu Rev
Financ Econ 4(1):255–296
Systemic Risk in Selected Countries of Western and Central Europe 185

Brownlees C, Engle R (2017) SRISK: a conditional capital shortfall index for systemic risk
assessment. Rev Financ Stud 30(1):48–79
Cihak M (2007) Systemic loss: a measure of financial stability. Czech J Econ Financ 57:5–26
Dumicic M (2018) Effectiveness of macroprudential policies in central and eastern European
countries. Publ Sect Econ 42(1):1–19
Engle R, Capellini R, Reis B (2019) V-Lab: systemic risk analysis summary. Retrieved July
30, 2019, from https://vlab.stern.nyu.edu/welcome/srisk
European Banking Authority (2014) Guidelines on criteria to assess other systemically important
institutions (O-SIIs). European Banking Authority, December 16. Retrieved July 31, 2019, from
https://eba.europa.eu/regulation-and-policy/own-funds/guidelines-on-criteria-to-to-assess-
other-systemically-important-institutions-o-siis-/-/regulatory-activity/press-release
European Banking Authority (2019) List of O-SIIs notified to the EBA. Retrieved August 1, 2019,
from https://eba.europa.eu/risk-analysis-and-data/other-systemically-important-institutions-o-
siis-/2017
Financial Stability Board (2011) Policy measures to address systemically important financial
institutions, Technical Report, 4th of November. http://www.fsb.org/2011/11/r_111104bb/
[10.12.2016]
Hattori A, Kikuchi K, Niwa F, Uchida Y (2014) A survey of systemic risk measures: methodology
and application to the Japanese market. IMES Discussion Paper Series, Institute for Monetary
and Economic Studies, Bank of Japan, 14(E-03)
Jajuga K, Karaś M, Kuziak K, Szczepaniak W (2017) Ryzyko systemu finansowego. Metody oceny
i ich weryfikacja w wybranych krajach [The risk of the financial system. Methods of assesment
and their verification, in Polish]. NBP Research Papers, No. 329 (PL)
Karaś M (2019) Measuring financial system stability – analysis and applications for Poland, Ph.D.
thesis, Wrocław University of Economics, February, unpublished manuscript
Karaś M, Szczepaniak W (2017) Measuring systemic risk with CoVaR using a stock market data-
based approach. In: Jajuga K, Orlowski L, Staehr K (eds) Contemporary trends and challenges
in finance. Springer proceedings in business and economics. Springer, Cham, pp 135–143
Karaś M, Szczepaniak W (2020) Three ways to improve systemic risk analysis of the CEE region
using SRISK and CoVaR. J Credit Risk (in print)
Karkowska R (2013) Systemic risk in central and eastern European banking and its determinants.
Merton option model approach. SSRN Electron J. https://doi.org/10.2139/ssrn.2349017
Liang N (2013) Systemic risk monitoring and financial stability. J Money Credit Bank 45:129–135
Matysek-Jędrych A (2007) Financial system – definition and functions. Bank & Credit 10:38–50
Narodowy Bank Polski (2019) Financial stability report, 140
Radulescu M, Banica L, Sinisi CI (2018) Developments of the CEE banking sectors after the
financial crisis. Proc Int Conf Bus Excell 12(1):851–863
Smaga P (2014) The concept of systemic risk. Systemic Risk Centre Special Paper No 5. The
London School of Economics and Political Science, August
 Part III
Corporate Finance
Industry and Size Effect in the Relation
Between Corporate Material and Financial
Decisions: Findings from the EU Countries

Julia Koralun-Bereźnicka

1 Introduction

Are companies with high operational risk more likely to be conservative in terms of
financial policies? Or does aggressive operational strategy imply similar behaviour
when it comes to financing decisions? Does this relationship depend on firm size, or
is it more likely to be industry-dependent? Interactions of real and financial decisions
have been focus of multiple studies, e.g. by Ravid (1988), Campello and Giambona
(2013), or Ortiz-Molina and Phillips (2014), who found operating inflexibility an
economically important source of risk.
However, despite the profusion of corporate finance literature dedicated to asset
flexibility as a factor affecting capital structure, some questions in the field remain at
least partly unanswered. Specifically, it seems that little attention has been paid to the
notion that the way in which the commonly recognized factors impact leverage may
vary depending on some other, indirect circumstances or features, such as the firm
size or its industrial classification. This study adds to the existing literature by
addressing the issue with yet another approach, i.e. by searching for the indirect
factors of debt. The applied method can be considered as contributive to the vast
majority of empirical research, where the significance and the direction of the asset
tangibility impact on capital structure, along with other factors, is usually verified
directly, i.e. without introducing any additional categories, which could potentially
affect the analysed relationships.

J. Koralun-Bereźnicka (*)
University of Gdańsk, Gdańsk, Poland
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 189
K. Jajuga et al. (eds.), Contemporary Trends and Challenges in Finance, Springer
Proceedings in Business and Economics,
https://doi.org/10.1007/978-3-030-73667-5_11
190 J. Koralun-Bereźnicka

2 Theoretical Background and Hypotheses Development

Asset tangibility, also known as collateral, is one of the key determinants of

corporate financing policies recognized by the two main capital structure theories,
i.e. the static trade-off theory (TOT) developed through the dispute over the capital
structure irrelevance proposition by Modigliani and Miller (1958), and the pecking-
order theory (POT) by Myers and Majluf (1984). The two theories offer different
explanations of the direction in which financial leverage is affected by asset tangi-
bility. According to the TOT, since tangible assets can be pledged as collateral, as
they are easier to value and therefore more easily convertible to cash in the case of
default, their share in total assets affects potential bankruptcy costs. As a result, a
higher proportion of tangible assets should make lenders more willing to supply
loans. The positive tangibility–leverage relation might also stem from the inclination
of firms to follow the maturity matching principle, according to which assets with a
certain degree of liquidity should be financed with liabilities of corresponding
maturity. Thence, since the fixed assets constitute long-term items of the firm
property, they should be covered by long-term liabilities, such as equity or at least
fixed capital, i.e. equity and long-term debt.
The POT predictions on the asset structure–leverage relation are more ambiguous
in terms of its sign. On the one hand, as argued by Weill (2002), asset tangibility can
mitigate the information asymmetry between firms and creditors by allowing lenders
to better evaluate the quality of businesses. Therefore, tangibility can act as a means
of controlling such problems as adverse selection or moral hazard. On the other
hand, however, an inverse relationship between firm leverage and asset tangibility
could result from the fact that large holdings of tangible assets may be used as a
source of internally generated funds and thus discourage firms from turning to
external financing (Daskalakis and Psillaki 2008). A negative relation can also be
attributed to low information asymmetry associated with tangible assets, which
should make equity issuances less costly (Sibindi 2016). Finally, the inverse relation
between the asset structure and capital structure may originate from the tendency of
firms to maintain a reasonable balance between operational and financial risk.
Therefore, the negative tangibility–leverage relation may be characteristic for com-
panies which tend to compensate high operational risk resulting from a larger share
of long-term fixed assets by lower financial risk associated with reduced
indebtedness.
Another capital structure determinant considered in this study is the firm size,
whose positive relation with debt expected by TOT is associated with the usually
lower risk of large companies (Frank and Goyal 2009; Kurshev and Strebulaev
2008). The issue of the firm size significance as an indirect determinant of capital
structure has been addressed e.g. by Daskalakis and Thanou (2010). Their study of
the Greek SMEs broken into size classes of medium, small and micro-enterprises,
indicates that, although the firm size does affect its debt level, no such influence was
found with respect to the relation between other determinants of leverage and the use
of debt. Authors conclude that companies belonging to different size groups behave
Industry and Size Effect in the Relation Between Corporate Material and. . . 191

similarly in terms of the relationship between debt and factors such as e.g. size or
asset structure. However, the opposite evidence is reported by Ramalho and Silva
(2009) in their study of different sized firms in Portugal. The authors found that the
determinants of leverage differ between micro, small, medium and large-sized
companies.
Along with the firm-level variables, financial leverage can also be determined by
external factors, such as the industrial specifics. The significance of the industrial
classification in terms of debt has been reported e.g. by Harris and Raviv (1991),
Phillips and Mackay (2005), Talberg et al. (2008), Degryse et al. (2012), or more
recently by Stancic et al. (2017) and is attributed to such industry features as the
assets flexibility (Shleifer and Vishny 1992), technological differences (Maksimovic
and Zechner 1991), or industrial competition (Leibenstein 1966). Corporate deci-
sions concerning their assets and the corresponding liabilities are taken with the
consideration of similar decisions taken by other companies in the industry. This
might result in more diversified financial structures within a given sector, rather than
in a target debt ratio common to the entire industry. As evidenced by Almazan and
Molina (2005), the leverage diversity is higher in concentrated industries, as well as
those with more freedom in terms of corporate governance practices and greater
assets liquidity. The authors also found greater diversity of capital structure in
industries, where companies use different production technology, in older industries
and in sectors with significant growth opportunities.
It appears, however, that the influence of firm size or industry on capital structure
may be dual (De Jong et al. 2008). Apart from their direct impact on debt, they may
also affect leverage indirectly, i.e. by influencing the way in which direct capital
structure determinants, such as asset tangibility, impact debt level.
Taking into account the main theories and previous empirical evidence, the three
research hypotheses are formulated: (1) the asset tangibility–capital structure relation
depends on the firm size; (2) the asset tangibility–capital structure relation depends
on the firm industrial classification; (3) the industry specifics is more important than
the firm size effect in the relation between corporate material and financial decisions.
The verification of these hypotheses, based on reliable European data, would provide
more insights into the indirect effect of firm size and industry in capital structure.

3 Data and Methods

The data is retrieved from the BACH-ESD (Banque de France 2019), which pro-
vides harmonized and aggregated corporate financial information from Austria,
Belgium, Czech Republic, Denmark, France, Germany, Italy, Luxemburg, Poland,
Portugal, Slovakia and Spain. The data is also broken by size groups (small,
medium, and large firms), by industries according to the NACE classification at
the section level, and by years (2000–2017). Several industries were excluded from
the analysis due to either data gaps or lack of cross-industry comparability. The
ratios are calculated with the use of median balance sheet data for a given category of
192 J. Koralun-Bereźnicka

Table 1 Construction of variables

Variables Definition
Debt to assets ratio (D/A) Total debt/assets
Assets tangibility (TNG) Tangible fixed assets/assets
Size dummies (D_SIZE) SM, ME, LA
Industry dummies (D_IND) A, B, C, D, E, F, G, H, I, J, L, N, P, Q, R, S
Tangibility–size interactions TNG*SM, TNG*ME, TNG*LA
Tangibility–industry interactions TNG *A, TNG *B, . . ., TNG *S
Notes: SM small, ME medium, LA large, industry symbols by NACE (Nomenclature Statistique des
Activités économiques dans la Communauté Européenne), i.e. A—Agriculture, forestry and fish-
ing, B—Mining and quarrying, C—Manufacturing, D—Electricity, gas, steam and air conditioning
supply water, E—Water supply, sewerage, waste management and remediation activities, F—
Construction, G—Wholesale and retail trade, repair of motor vehicles and motorcycles, H—
Transportation and storage, I—Accommodation and food service activities, J—Information and
communication, L—Real estate activities, N—Administrative and support service activities, P—
Education, Q—Human health and social work services, R—Arts, entertainment and recreation, S—
Other service activities

country, size, industry, and year. Due to the missing data for Q2 in the Czech
Republic and Slovakia, the means of ratios were used instead. The variables
employed in the study are defined in Table 1.
Initially, the descriptive statistics of the main variables were analysed in the three
sections, i.e. across size groups, countries and industries. This preliminary analysis
was meant to discover the basic regularities concerning the main asset structure and
liabilities structure within the analysed population. Then, in order to identify the
industry and size effect in the tangibility–capital structure relation, panel regression
models with interactions between variables were estimated, as specified by formula
(Almazan and Molina 2005):

ðD=AÞicst ¼ α þ β1 TNGicst þ β2 D SIZE icst þ β3 D INDicst

þ β4 ðD SIZE icst ∙ TNGicst Þ þ β5 ðD INDicst ∙ TNGicst Þ þ ξicst ð1Þ

where i denotes industry (i ¼ 1, . . . 16), c—country (c ¼ 1, . . . 12), s—size group

(s ¼ 1, 2, 3), and t—year (t ¼ 1, . . . 18). The estimation method was OLS with
standard errors robust for heteroscedasticity and autocorrelation of error terms
(Baltagi 2008).

4 Results

The descriptive statistics of the two main variables, i.e. debt ratio and asset tangi-
bility, are shown in Tables 2 and 3, respectively.
It can be seen from the descriptive statistics of the dependent variable shown in
Table 2 that small companies have the highest mean value of the debt ratio.
Industry and Size Effect in the Relation Between Corporate Material and. . . 193

Table 2 Descriptive statistics of the debt to asset ratio

Size group, country, Mean Minimum Maximum Standard
industry N value Median value value deviation
ALL 6451 0.637 0.653 0.051 1.748 0.142
SM 2334 0.640 0.657 0.106 0.960 0.130
ME 2249 0.634 0.645 0.120 1.108 0.134
LA 1868 0.638 0.656 0.051 1.748 0.164
AT 732 0.692 0.695 0.355 0.978 0.091
BE 659 0.595 0.618 0.106 0.919 0.136
CZ 596 0.537 0.529 0.051 1.710 0.181
DE 711 0.675 0.678 0.392 0.944 0.112
ES 682 0.589 0.588 0.199 0.971 0.106
FR 745 0.689 0.688 0.465 0.981 0.086
IT 650 0.717 0.719 0.395 0.973 0.074
NL 269 0.603 0.599 0.120 0.978 0.166
PL 499 0.494 0.503 0.079 0.973 0.143
PT 594 0.713 0.716 0.239 0.994 0.123
SK 314 0.636 0.662 0.155 1.748 0.161
A 370 0.553 0.583 0.051 1.108 0.162
B 369 0.556 0.567 0.120 0.981 0.150
C 456 0.604 0.610 0.308 0.815 0.083
D 454 0.610 0.624 0.106 0.956 0.163
E 418 0.610 0.646 0.226 0.942 0.146
F 456 0.731 0.761 0.143 1.748 0.123
G 456 0.672 0.676 0.279 0.858 0.081
H 456 0.641 0.659 0.237 0.994 0.134
I 435 0.667 0.676 0.119 0.978 0.137
J 454 0.626 0.643 0.192 0.905 0.115
L 380 0.630 0.667 0.079 1.710 0.171
N 442 0.740 0.755 0.336 1.051 0.116
P 219 0.611 0.615 0.347 0.882 0.116
Q 350 0.586 0.579 0.142 0.973 0.134
R 402 0.657 0.669 0.155 1.560 0.142
S 334 0.656 0.666 0.254 0.973 0.119
Notes: N stands for the number of companies in each category, i.e. ALL—in all countries,
industries, and size groups; SM, ME, LA—in size groups as defined in Table 1; AT-SK—in
countries defined in Table 1; A-S—in industrial sections defined in Table 1

However, the average debt level is very similar across size groups of firms. The same
regularity can be noticed when the median is compared across size groups.
As for the assets tangibility measure (TNG), shown in Table 3, a negative
correlation between firm size and assets structure can be identified; the average
relation of fixed to total assets decreases along with the firm size. Consequently,
small enterprises are characterised by the highest share of fixed assets. However, the
pattern between corporate material decisions measured by TNG and capital structure
194 J. Koralun-Bereźnicka

Table 3 Descriptive statistics of the asset tangibility ratio

Size group, country, Mean Minimum Maximum Standard
industry N value Median value value deviation
ALL 6467 0.386 0.370 0.014 0.917 0.195
SM 2342 0.414 0.407 0.021 0.856 0.175
ME 2250 0.384 0.367 0.014 0.910 0.196
LA 1875 0.353 0.310 0.018 0.917 0.210
AT 732 0.455 0.463 0.080 0.836 0.174
BE 659 0.300 0.280 0.036 0.762 0.157
CZ 596 0.479 0.490 0.014 0.917 0.204
DE 712 0.421 0.434 0.074 0.836 0.210
ES 685 0.379 0.349 0.053 0.822 0.162
FR 745 0.290 0.269 0.036 0.749 0.158
IT 650 0.318 0.316 0.051 0.666 0.151
NL 273 0.272 0.214 0.021 0.832 0.187
PL 502 0.485 0.491 0.019 0.867 0.200
PT 599 0.380 0.350 0.042 0.884 0.196
SK 314 0.478 0.478 0.082 0.875 0.189
A 370 0.438 0.456 0.080 0.917 0.164
B 369 0.385 0.393 0.021 0.856 0.160
C 456 0.281 0.275 0.059 0.497 0.098
D 454 0.518 0.553 0.028 0.910 0.172
E 422 0.505 0.472 0.111 0.881 0.191
F 456 0.195 0.159 0.037 0.581 0.118
G 456 0.185 0.186 0.038 0.336 0.066
H 456 0.494 0.494 0.091 0.875 0.145
I 435 0.485 0.504 0.042 0.804 0.186
J 454 0.214 0.180 0.032 0.810 0.112
L 380 0.609 0.640 0.018 0.909 0.181
N 447 0.353 0.329 0.019 0.832 0.137
P 219 0.355 0.353 0.036 0.884 0.150
Q 352 0.444 0.468 0.108 0.795 0.164
R 406 0.406 0.423 0.057 0.818 0.170
S 335 0.356 0.363 0.014 0.791 0.152
Notes: N stands for the number of companies in each category, i.e. ALL—in all countries,
industries, and size groups; SM, ME, LA—in size groups as defined in Table 1; AT-SK—in
countries defined in Table 1; A-S—in industrial sections defined in Table 1

is less evident. Although small companies are clearly the riskiest in terms of leverage
and assets structure, the mean and median values for medium and large-sized firms
do not reveal any clear regularities in terms of the trade-off between operational and
financial risk.
The level of debt seems much more varied across countries than across size
groups of firms. Companies in Italy and Portugal rely on debt considerably more
than in Poland, where the share of debt is clearly the lowest of all countries.
Industry and Size Effect in the Relation Between Corporate Material and. . . 195

Some clearer patterns in the relation between capital and assets structure can be
observed in the international cross-section than across size groups. The two coun-
tries with the lowest leverage, namely Poland and the Czech Republic, are at the
same time those with the highest share of fixed assets, which suggests a negative
relation between debt ratio and assets tangibility. However, the cross-industry
comparison of descriptive statistics does not provide support for the above rule.
As for the industrial cross-section, it is clear that firms from the sections of admin-
istration and construction follow the most aggressive financing strategies with the
highest mean level of debt, whereas agricultural and mining companies tend to be
more conservative in terms of leverage. The highest intra-industry variation of
capital structure is observed for the accommodation industry. The accommodation
industry is the one with distinctly highest share of fixed assets, as opposed to the
sections of construction and trade, characterized by larger flexibility of assets.
This indicates that the impact of asset structure on debt varies both across
industries and size groups of firms. Certainly, the values of R2 would be far form
satisfactory if the aim was to fully explain the capital structure variability. This
study, however, is only meant to analyze in detail the impact of just one of the many
factors affecting leverage, along with its size and industry interactions. The estima-
tion results of model (Almazan and Molina 2005) are shown in Table 4.
In order to capture the relative importance of size and industry interactions, the
model was estimated first only with size-tangibility interactions, then only with
industry-tangibility interactions, and finally with both. All three estimations reveal
significant negative relation between asset tangibility and debt ratio. Also, both types
of interactions prove significant, although only in the model where both size and
industry interactions were included.
In order to verify which category of interactions is more relevant, the AIC values
were compared. The omission of industry interactions has more considerable effect
on the AIC than the omission of size interactions, which indicates lower importance
of the size effect in the relation between asset tangibility and capital structure within
the analyzed sample.
The impact of tangibility on total debt ratio is also visualized by Fig. 1, which
shows that the direction of the relation between TNG and D/A is clearly industry-
dependent, although with only two sections (A—agriculture and L—administrative
activities) deviating from the generally negative relation.
However, in all cases the relation remains unchanged across size groups for a
given industry. Moreover, a characteristic feature noticeable in all industries with
negative tangibility-leverage relation is the weakest impact of asset structure on debt
level in small-sized companies. On the contrary, in the two industries for which the
positive impact is observed, the relation occurs the strongest for small firms.
196 J. Koralun-Bereźnicka

Table 4 Estimation results of panel regressions explaining D/A

Model (Almazan and Model (Banque de
Molina 2005) Model (Baltagi 2008) France 2019)
(size interactions) (industry interactions) (both interactions)
Variable Estimate Std. error Estimate Std. error Estimate Std. error
Const. 0.680*** 0.028 0.699*** 0.026 0.693*** 0.027
TNG 0.236*** 0.038 0.280*** 0.038 0.270*** 0.037
ME 0.055*** 0.019 0.035*** 0.011 0.081*** 0.024
LA 0.063*** 0.019 0.040*** 0.012 0.093*** 0.025
B 0.051* 0.031 0.077* 0.044 0.107** 0.043
C 0.039 0.026 0.053 0.037 0.086** 0.039
D 0.046 0.033 0.037 0.050 0.059 0.050
E 0.010 0.031 0.085* 0.051 0.057 0.053
F 0.043 0.027 0.070** 0.033 0.031 0.038
G 0.004 0.026 0.012 0.037 0.024 0.039
H 0.058** 0.029 0.048 0.060 0.009 0.061
I 0.062** 0.028 0.040 0.043 0.013 0.044
J 0.062** 0.029 0.067 0.044 0.092** 0.045
L 0.078** 0.037 0.010 0.067 0.053 0.068
N 0.045 0.029 0.107** 0.043 0.139*** 0.046
P 0.059* 0.033 0.021 0.056 0.034 0.054
Q 0.081*** 0.030 0.095 0.063 0.121* 0.062
R 0.017 0.032 0.007 0.064 0.024 0.065
S 0.003 0.029 0.061 0.041 0.093** 0.043
TNG*ME 0.046 0.049 0.117* 0.064
TNG*LA 0.060 0.055 0.146** 0.073
TNG*B 0.068 0.118 0.148 0.119
TNG*C 0.043 0.112 0.138 0.120
TNG*D 0.170* 0.089 0.231** 0.094
TNG*E 0.155* 0.092 0.075 0.104
TNG*F 0.172 0.116 0.040 0.126
TNG*G 0.079 0.138 0.046 0.145
TNG*H 0.001 0.102 0.110 0.111
TNG*I 0.040 0.090 0.107 0.094
TNG*J 0.014 0.183 0.062 0.183
TNG*L 0.143 0.122 0.244* 0.130
TNG*N 0.426*** 0.092 0.531*** 0.108
TNG*P 0.131 0.154 0.089 0.148
TNG*Q 0.023 0.125 0.098 0.125
TNG*R 0.065 0.123 0.008 0.128
TNG*S 0.144 0.100 0.232** 0.106
No. obs. 6734 6734 6734
R2 0.262 0.286 0.294
Heteroscedasticity 1484.7 [0.000] 890.8 [0.000] 936.6 [0.00]
Normality 247.7 [0.000] 281.0 [0.000] 255.4 [0.000]
(continued)
Industry and Size Effect in the Relation Between Corporate Material and. . . 197

Table 4 (continued)
Model (Almazan and Model (Banque de
Molina 2005) Model (Baltagi 2008) France 2019)
(size interactions) (industry interactions) (both interactions)
Variable Estimate Std. error Estimate Std. error Estimate Std. error
AIC 9901.8 10095.5 10162.1
Hausman test 114.6 [0.000] 142.7 [0.000] 154.8 [0.000]
Joint significance of interactions
Size 1.310 [0.191] 2.162 [0.031]
Industry 0.833 [0.405] 2.088 [0.037]
Notes: (1) ***p < 0.01, **p < 0.05, *p < 0.1, (2) White’s test for heteroscedasticity, (3) Doornik–
Hansen test for normality of residuals, (4) Interpretation of parameters in relation to section A and
small firms

-0.600 -0.400 -0.200 0.000 0.200 0.400 0.600

A
B
C
D
E
F
G
H
I
J
N
L
P
Q
R
S

S M L

Fig. 1 The impact of asset tangibility on total debt across industries and size groups. Note: the bars
represent sums of parameters from the model with industry and size interactions. Source: own study

5 Conclusions

Generally, the study highlights the importance of the relation between corporate
material and financial decisions, reflected in asset structure and capital structure,
respectively. Contrary to the TOT expectations, as well as denying the maturity
matching principle, the relation between asset tangibility and debt level proved
198 J. Koralun-Bereźnicka

mainly negative, which is in line with the findings by Daskalakis and Psillaki (2008),
or Pepur et al. (2016), but opposing the evidence reported by Degryse et al. (2012).
The main conclusion corresponds to the view that a higher share of long-term fixed
assets, meaning lower elasticity and therefore higher operational risk, tends to be
compensated by lower leverage-induced financial risk. This suggests the occurrence
of a trade-off between the activity-related risk level and the risk resulting from
indebtedness.
However, findings provide evidence that the relationship in question varies
depending on the indirect factors. The direction of the relation between asset
tangibility and capital structure depends significantly on the industrial classification
of firms, and to a lesser extent on their size. This provides support for hypothesis (2),
concerning the relevance of the industry effect for the examined asset tangibility–
capital structure relation, and only weak support for hypothesis (1), which assumes
the importance of the firm size effect in this area. Daskalakis et al. (2014) report
similar conclusions on the size effect in capital structure: while the firm size does
affect how much debt a firm will use, it does not influence the relationship between
the other factors and debt usage.
The reported prevalence of the industry effect over the firm size effect in the
assets–capital structure relation indicates the likely truthfulness of hypothesis (3),
according to which the industry specifics is more important than the firm size effect
in the relation between corporate material and financial decisions.
The impact of the above findings on theory is therefore such that the competing
capital structure theories should not be treated as universal or comprehensive
concepts. Rather, their applicability, interpreted as the ability to explain corporate
behaviour, may differ depending on a number of indirect circumstances, many of
which certainly yet to be identified.
As for the limitations of the study, the sample covering only 12 EU countries
certainly does not meet the requirement for generalization of the research results,
probably not even within the EU area, whose various regions are far from homoge-
neity in terms of economy. Therefore, extending the database with harmonized and
comparable financial information would create opportunities for a more complete
analysis and understanding of the phenomena in question. Nevertheless, the study
offers framework for further investigation of the indirect effects in capital structure,
e.g. by considering international cross-sections and (or) debt maturity. It might be
also worthwhile to determine whether and how the regularities within the area of
corporate material and (or) financial decisions are affected by the financial crisis.

References

Almazan A, Molina CA (2005) Intra-industry capital structure dispersion. J Econ Manage Strat 14
(2):263–297
Baltagi BH (2008) Econometric analysis of panel data. Wiley, Chichester
Industry and Size Effect in the Relation Between Corporate Material and. . . 199

Banque de France (2019) Bank for the accounts of companies harmonised – European sectoral
references database. http://www.bachesd.banque-france.fr. Accessed 13 Dec 2019
Campello M, Giambona E (2013) Real assets and capital structure. J Financ & Quant Anal 48
(5):1333–1370
Daskalakis N, Psillaki M (2008) Do country of firm explain capital structure? Evidence from SMEs
in France and Greece. App Financ Econ 18(2):87–97
Daskalakis N, Thanou E (2010) Capital Structure of SMEs: to what extent does size matter? http://
ssrn.com/abstract¼1683161
Daskalakis N, Eriotis N, Thanou E, Vasiliou D (2014) Capital structure and size: new evidence
across the broad spectrum of SMEs. Managerial Financ 40(12):1207–1222
de Jong A, Kabir R, Nguyen T (2008) Capital structure around the world: the roles of firm- and
country-specific determinants. J Bank & Financ 32(9):1954–1969
Degryse H, De Goeij P, Kappert P (2012) The impact of firm and industry characteristics on small
Firm’s capital structure. Small Bus Econ 38(4):431–447
Frank MZ, Goyal VK (2009) Capital structure decisions: which factors are reliably important?
Financ Manage 38(1):1–37
Harris M, Raviv A (1991) The theory of capital structure. J Financ 46(1):297–355
Kurshev A, Strebulaev IA (2008) Firm size and capital structure. AFA 2008 New Orleans Meetings
Paper. https://doi.org/10.2139/ssrn.686412
Leibenstein H (1966) Allocative efficiency vs. X-efficiency. Am Econ Rev 56(3):392–415
Maksimovic V, Zechner J (1991) Debt, agency costs, and industry equilibrium. J Financ 46
(5):1619–1643
Modigliani F, Miller MH (1958) The cost of capital, corporation finance, and the theory of
investment. Am Econ Rev 48(3):261–297
Myers SC, Majluf N (1984) Corporate financing and investment decisions when firms have
information that investors do not have. J Financ Econ 13(2):187–221
Ortiz-Molina H, Phillips G (2014) Real asset illiquidity and the cost of capital. J Financ Quant Anal
49(1):1–32
Pepur S, Ćurak M, Poposki K (2016) Corporate capital structure: the case of large Croatian
companies. Econ Res – Ekonomska Istraživanja 29(1):498–514
Phillips GM, MacKay P (2005) How does industry affect firm financial structure? Rev Financ Stud
18(4):1433–1466
Ramalho JS, Silva JV (2009) A two-part fractional regression model for the financial leverage
decisions of micro, small, medium and large firms. Quant Financ 9(5):621–636
Ravid SA (1988) On interactions of production and financial decisions. Financ Manage 17
(3):87–99
Shleifer A, Vishny RW (1992) Liquidation values and debt capacity: a market equilibrium
approach. J Financ 47(4):1343–1366
Sibindi AB (2016) Determinants of capital structure: a literature review. Risk Gov Control: Financ
Mark Instit 6(4):227–237
Stancic P, Janković M, Čupić M (2017) Testing the relevance of alternative capital structure
theories in Serbian economy. Teme: Časopis za društvene nauke 40(4):1309–1325
Talberg M, Winge C, Frydenberg S, Sjur W (2008) Capital structure across industries. Int J Econ
Bus 15(2):181–200
Weill L (2002) Determinants of leverage and access to credit: evidence on Western and Eastern
Europe countries. https://www.researchgate.net/publication/253293146_Determinants_of_
Leverage_and_Access_to_Credit_Evidence_on_Western_and_Eastern_Europe_countries
Technology Level and Financial Constraints
of Public Listed Companies

Katarzyna Prędkiewicz, Paweł Prędkiewicz, and Marek Pauka

1 Introduction

The role of corporate innovation for economic growth is undoubtful. With varying
influences depending on the country’s economic development and its phase of the
economic cycle, innovation accounts for approximately 50% of a country’s total
GDP growth (OECD 2015).
From a theoretical point of view, the asymmetry of information, problems at the
principal-agent interface, and the phenomenon of moral hazard and inverse selection
are the reason for various disturbances in the debt and equity financing market.
These phenomenons may transfer into problems with obtaining external financing
(financial constraints) at the level of individual enterprises.
That is why governments improve access to capital for innovative and high-
technology companies to support them by direct (e.g., grants) and indirect (e.g., tax
incentives) instruments and developing the financial system, among other things, the
stock exchange market.
Theoretical models support the hypothesis that the development of a capital
market should lower capital costs for innovative companies and ease access to
funds (Brown et al. 2009). However, there is no consensus in the literature about
whether the stock exchange market can conclusively solve financial constraints for
innovative companies. For example, Brown et al. (2009) proved that despite the
development of the US capital market, the companies may still have a problem with
financing R&D projects. Other authors concluded that companies in the USA, which
undertake R&D projects are not financially constrained, whereas those in Europe
still are (Cincera and Ravet 2010).

K. Prędkiewicz (*) · P. Prędkiewicz · M. Pauka

Wroclaw University of Economics and Business, Wrocław, Poland
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 201
K. Jajuga et al. (eds.), Contemporary Trends and Challenges in Finance, Springer
Proceedings in Business and Economics,
https://doi.org/10.1007/978-3-030-73667-5_12
202 K. Prędkiewicz et al.

We want to contribute to previous studies and analyze whether one of the best-
developed stock exchange markets in Central and East Europe (Poland) alleviates
financial constraints for high-technology companies. Based on the Warsaw Stock
Market (Poland) data, we will check whether high-technology companies are finan-
cially constrained compared to low and not technological firms.

2 Literature Review

The returns on innovative activities are uncertain and make innovation riskier to
finance (Hall 2002; Mazzucato 2013; Coad et al. 2016). Companies that undertake
highly risky projects have an informational advantage over external agents who may
not be able to assess a firm’s quality based on their innovative activity (Stiglitz and
Weiss 1981; Carpenter and Petersen 2002; de Rassenfosse et al. 2011). High
asymmetry of information causes investors or suppliers of external capital to have
difficulty distinguishing good projects from bad ones. To assess a project in an early
stage of technological development, skilled experts are necessary what can occur
problem for banks (Ueda 2004). Additionally, innovative projects are typically long-
term with uncertain outcomes and challenging to predict revenue (Brown et al.
2012). Moreover, the knowledge asset created during the innovation process could
not be used in an easy way as collateral because they are intangible (however, there
is a possibility of IP-backed debt). Also, most R&D investments are human expenses
(Hall 2010).
Problem with access to capital (financial constraints) may force innovative firms
to rely on internally generated funds. However, it is only possible for older compa-
nies, which have established a stream of revenue and cash flow from earlier closed
projects yet. According to Kaplan and Zingales, “Firm is considered more finan-
cially constrained as the wedge between its internal and external cost of funds
increases” (Kaplan and Zingales 1997). Tirole indicated that financial constraints are
caused by the disorders in the supply of external capital, and the mean reason is
information asymmetry between an investor and a firm (Tirole 2006). Hall provides
three main reasons for financial constraints: (1) information asymmetry (2) moral
hazard problem (3) tax reasons, which lead to changes in preferences between
internal and external capital (Hall 2010). When a company is financially constrained,
it cannot undertake part or all of its investments because of problems with capital
access. Nonoptimal investments of individual companies lead to also to nonoptimal
development of the economy.
The key queries in literature are whether companies undertaking investments
(innovative, but also ordinary) are financially (liquidity) constrained. This question
is followed by next, how to measure financial constraints. Because the supply of
capital is not easy to capture by econometric methods, measuring them is a big
empirical challenge. The indirect approach is based on the investment equation. The
companies are considered financially constrained when investments are sensitive to
cash-flow changes, assuming that cash-flow represents internal capital’s arability.
Technology Level and Financial Constraints of Public Listed Companies 203

When the company invests only in response to cash-flow fluctuations, it cannot gain
external capital at a lower cost than a marginal return from the project. The early
research focused on ordinary (non-R&D) investments, e.g., works of Fazzari et al.
(1988), which were, however, criticized by Kaplan and Zingales (1997). In regards
to investment in R&D, pioneer research was initiated by Hall (1992), Himmelberg
and Petersen (1994), and are continued contemporary inter alia in works of Cincera
and Ravet (2010), Czarnitzki and Hottenrott (2011), Lööf and Nabavi (2016).
The results of previous studies confirmed that innovative and high-technology
companies behave dissimilarly compared to non-innovative, low-technology firms.
They rely mainly on equity finance rather than debt.
For example, Brown et al. (2009) note that in high-tech companies, there is a
specific financing hierarchy—R&D projects are financed first from internal equity
(cash flow) and then from external (share issue). However, this hierarchy does not
apply to investments in fixed assets that provide the possibility of securing a debt.
Based on empirical research, the authors stated that young high-tech companies are
financially limited in terms of financing R&D projects, but mature (operating on the
market for more than 15 years) are not. They have proven that in the case of
investment in innovation, an equally important source of financing as cash flow is
external capital obtained from the issue of shares. It follows that enterprises operat-
ing on market-oriented financial markets are more likely to implement R&D projects
than those operating on banking-oriented markets.
Also, Casson et al. (2008) conclude that the “probability of issuing new equity
rises monotonically with R&D intensity, while the use of debt finance starts to
decline eventually as R&D intensity increases.” Similarly, Aghion et al. (2004)
noticed that “firms that report positive but low R&D use more debt finance than
firms that report no R&D, but the use of debt finance falls with R&D intensity among
those firms that report R&D”, moreover “firms that report R&D are more likely to
raise funds by issuing shares than firms that report no R&D, and this probability
increases with R&D intensity.” It also confirms that large and highly innovative
companies behave differently to non-innovative ones. Schäfer et al. (2004), based on
a sample of German high-tech companies, conclude that the probability that a young
high-tech company “receives equity financing is an increasing function of the
financial risk”.
The research confirmed that companies implementing R&D projects operating in
emerging markets, like Poland, are financially constrained (Nehrebecka and Białek-
Jaworska 2015). A higher self-financing capacity (internal financing) favors the
development of research and development activities. Enterprises that maintain a
higher level of savings finance R&D investments to a greater extent with their funds.
They are less dependent on external debt financing sources (credit, loans, and debt
securities).
In turn, Bah and Dumontier (2001) confirmed that a lower level of financial
leverage characterizes enterprises in the USA, Great Britain, and Japan with high
R&D spending. Similarly, Friend and Lang (1988) and Hall (2010) confirmed a clear
negative correlation between the intensity of R&D expenditure and financial
204 K. Prędkiewicz et al.

leverage in American enterprises. Hall et al. (2007) had the same conclusions
observing European enterprises.
The results of so far conducted studies confirmed that more innovative and high-
technology companies rely mainly on equity finance rather than debt compared to
non-innovative, low-technology firms. It means that the stock exchange may be
crucial for the development of innovative companies. Brown et al. (2009) confirmed
that with the development of the US capital market, the financial constraints for
investments in fixed assets have disappeared; however, they are persisted for R&D
projects. Simultaneously, the number of enterprises with negative cash flows was
growing. For them, after exhausting internal cash flow, the main source of financing
is equity obtained from the issuing of the shares. Brown et al. (2009) also observed
that American manufacturing companies significantly changed their investment
structure with time. In typical production companies, the share of investments in
fixed assets decreased and increased in R&D projects. Thus, although, according to
the authors, financial constraints in the US for projects in fixed assets have practi-
cally disappeared; however, they are still present for total corporate investment.
Opposite conclusions formulated Cincera and Ravet (2010). They analyzed
financial constraints for R&D projects, both in the USA and Europe in the period
2000–2007 and concluded that all companies undertaking R&D projects were
financially constrained. However, they analyzed subsamples separately for USA
and Europe, and it occurred that only European companies were financially
constrained. In contrast, those in the USA had no problem with capital access.
However, it should be stressed that the research sample covered huge companies
(median of employees was 6000).
However, according to most studies, a market-oriented financial system has not
entirely solved the problem of financial constraints. Innovative, high-technology
companies undertaking R&D projects may have problems with capital access, even
in countries with well-developed capital markets (Brown et al. 2009).

3 Data and Methodology

In this study, we used indirect methods of measuring financial constraints based on

the investment equation. This method assumes that financially constrained enter-
prises invest only when internal capital (cash flow) increases. Then cash flow
represents the availability of internal capital.
The general reduced-form of investment equations that we employed in our
studies is following (Fazzari et al. 1988):

ðI=K Þit ¼ f ðX=K Þit þ gðCF=K Þit þ uit

where:
Technology Level and Financial Constraints of Public Listed Companies 205

• Iit represents an investment in plant and equipment for firm i during period t or
R&D expenditure;
• X represents a vector of variables, possibly including lagged values, that have
been emphasized as determinants of investment from a variety of theoretical
perspectives;
• CF (firm’s internal cash flow) and it represents the potential sensitivity of
investment to fluctuations in available internal finance after investment opportu-
nities are controlled for through the variables in X; the function g depends on the
firm’s internal cash flow;
• K is the beginning of period capital stock;
• uit is an error term.
The statistically significant coefficient for variable CF/K will confirm that the
investment is undertaken when the operating cash flow is available, which means
that the company is financially constrained.
The above presented basic model was developed in further studies, then we took
into account improvements and suggestions formulated by other researchers.
In most studies, investments in fixed assets or operational cash flow were
calculated based on the balance sheet (Fazzari et al. 1988; Hall 2010).
Lewellen and Lewellen (2016) compared the CF determined according to the
traditional methodology (net profit plus depreciation) with the CF based on addi-
tional information obtained from the cash flow statement. They concluded that
simplified CF is a distorted measure, especially after 1990, due to the growing
importance of the profit’s non-cash position. A similar remark applies to invest-
ments, which can be determined based on fixed assets changes from the balance
sheet or all capital expenditure from the cash flow statement. According to the
authors, data from the cash flow statement allows for the correct financial constraints
measurement. Our research took into account the above suggestions—operational
cash flow and capital expenditure were directly derived from the cash flow
statement.
The access to information on R&D expenditure is limited and not available in
Polish enterprises’ financial statements. In this situation, the sensitivity of invest-
ments in tangible fixed assets to a change in cash-flow was analyzed in previous
studies (Scellato 2007; Ughetto 2008), assuming that investments in innovations are
partially included in the value of tangible fixed assets.
In the investment model, the company’s development prospects are controlled to
separate the influence on investment decisions of the company’s management
expectation regarding the future demand for products from decisions related to
increasing investments in response to changes in available internal funds (CF). In
the literature, various proposals can be found—sales revenues (Fazzari et al. 1988;
Kaplan and Zingales 1997; Harhoff 1998; Ughetto 2008), revenues growth rate
(Konings et al. 2003), as well as Tobin’s Q-index (Kaplan and Zingales 1997;
Audretsch and Elston 2002; Carpenter and Petersen 2002; Wagenvoort 2003). We
adopted the value of sales revenues for the control of investment opportunities,
206 K. Prędkiewicz et al.

referring to the principle that the demand for capital (investments) results from the
level or change in the company’s sales.
Apart from the classic variables in the investment models, the cash holdings, debt,
and the possibility of new shares are often considered.
In addition to the currently generated operational cash flow, enterprises may also
“accumulate” cash from previous periods, especially if they perceive themselves as
financially constrained. Cash holdings can provide a kind of “financial pillow” that
reduces the investment’s sensitivity to CF changes. Therefore, a positive relationship
between the measures of financial liquidity and undertaken investments should be
expected. The more financially constrained company, the stronger the relationship is
(according to Fazzari et al.). The high cost of external capital prompts the need to
raise funds to mitigate the impact of cash-flow “shocks” on the company’s invest-
ments. Brown et al. (2012) paid attention to the need to control the investment-CF
model’s savings level. They proved that without the control of smoothing out
expenditures from the kept funds and the possibility of issuing shares, it is impos-
sible to confirm predicted by theoretical models the relationship between CF sensi-
tivity and R&D expenditures. Our model assumes that the level of savings from the
previous period will affect the current investments; therefore, we included as a
control variable the last cash period holding.
In our model, we also assumed the need to control the debt level. According to
Scellato (2007), this variable should reflect the restriction in access to a new debt due
to the previously identified high debt level. Also, Brown et al. (2009) used the
variable relating to the rising of new long-term interest debt, i.e., change in debt. The
debt role turned out to be different for investment in fixed assets and investment in
R&D projects. In the first case, the debt is essential.
In contrast, in the second case (investment in innovation), debt plays a less critical
role. The possibility of obtaining capital from the issue of the new shares may be
crucial. In our research, we control the debt level in the period preceding the
investment (t-1). Firms with higher debt levels will find it more challenging to
undertake new investments. However, we do not control the possibility of issuing
new shares (this will be future research).
As a standard, the variables in investment models are scaled. However, there is no
agreement in the literature on the methodology. It is often the value of capital at the
beginning of the period (in t-1). We assumed for all variables scaling the total assets
at the beginning of the period.
To summarize, we used in our study the following model. Expectations about
future demand changes are controlled by revenues with additional control variables
in the field of cash holding and the possibility of obtaining debt:

I i,t
¼ β1 ðSi,t =K i,t1 Þ þ β2 ðCFOi,t =K i,t1 Þ þ β3 ðCASH i,t1 =K i,t1 Þ
K i,t1
þ β4 ðDi,t1 =K i,t1 Þ þ dt þ αi þ vi,t

where:
Technology Level and Financial Constraints of Public Listed Companies 207

• Ii,t—represents all investment of firm i during period t—we used capital expen-
diture from cash flow statement,
• Si,t—operating revenues (control of companies expectations about future demand
changes),
• CFOi,t—operational cash flow for firm i during period t (cash flow statement),
• CASHi,t-1—Cash and cash equivalent (control of companies’ savings),
• Di,t-1—interest debt (long-term debts and short-term loans),
• Ki,t-1—is the total assets at the beginning of period t.
We have collected a sample of companies publicity traded in Poland on the
Warsaw Stock Exchange from 2004 to 2018 (334 companies). We gathered their
financial statements from the EMIS database—the financial reports covered the time
from the initial public offering to 2018. The correctness of data was verified based on
their financial statements on companies’ websites or the National Court Register.
We used the Eurostat indicators on the High-tech industry and Knowledge-
intensive services (Annex 3—High-tech aggregation by NACE Rev. 2).1 The
manufacturing companies were qualified based on NACE code as high-technology,
medium high-technology, medium-low-technology, and low-technology and others
that have not been qualified for any groups. We aggregated them into two groups:
high-tech (we included companies-technology, medium high-technology, medium-
low-technology) and low-tech (low-technology, not qualified to any group—not-
technology). The sample structure is presented in Table 1.
Data were cleaned—the abnormal and outstanding data were excluded from the
sample. Every observation which got variable outside the range: Q1–3*(Q3-Q1);
Q3 + 3 * (Q3-Q1) was removed from the sample.
The means of standardized capital expenditure is higher for HT companies
(Table 2). When we compare the mean of standardized operating cash flow, we
noticed that the value is positive both for HT and LT firms, but it is a bit higher for
HT companies. HT firms in line with expectations and previous studies have lower
debt. Cash holding is also lower in HT compared to peers. All descriptive statistics
are presented in Table 2.
We used panel models with fixed effects and random effects, but Hausman test
showed that fixed-effects model was more efficient in all cases.

4 Results

To answer our research question, we have executed two models for panel data with
fixed-effects. The first model is calculated for HT companies and the second for LT.
Based on the first model (Table 3) we can conclude that HT companies invest in
fixed assets in response to changes in operating cash flow ( p-value < 0.0001), which

1
https://ec.europa.eu/eurostat/cache/metadata/en/htec_esms.htm.
 208

Table 1 Sample structure

Year 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018
HT 3 12 17 38 45 50 54 62 65 72 81 87 89 90 85
LT 16 29 46 79 98 107 130 150 164 177 191 208 221 229 227
Note: HT high-tech companies, LT low-tech companies
K. Prędkiewicz et al.
Technology Level and Financial Constraints of Public Listed Companies 209

Table 2 Descriptive statistics I S CFO DEBT CASH

HT mean 0.0517 1.2820 0.0637 0.1583 0.0763
LT mean 0.0451 1.0588 0.0543 0.1992 0.0894
HT S. dev. 0.0516 0.8870 0.0927 0.1513 0.0903
LT S. dev. 0.0557 0.9521 0.1092 0.1901 0.0963

Table 3 Investment models High technology Low technology

Coefficient (std. error) Coefficient (std. error)
Const 0.0197 (0.0106)* 0.0252 (0.0049)***
S 0.0118 (0.0057)** 0.0128 (0.0038)***
CFO 0.0936 (0.0316)*** 0.0144 (0.0175)
DEBT 0.0576 (0.0214)*** 0.0394 (0.0088)***
CASH 0.1423 (0.0367)*** 0.0659 (0.0171)***
Note: S—operating revenues; CFO—operational cash flow,
DEBT—interest debt (long-term debts and short-term loans),
CASH—Cash and cash equivalent; *p-value < 0.10, **p-
value < 0.05, ***p-value < 0.01

supports the hypothesis that this group of companies is financially constrained. Also,
the previous year’s cash holding is crucial for investments of HT companies.
Investments are undertaken in response to current operating cash flow, but also
prior savings impact the capital expenditure level. Cash-holdings then affect the
investment –HT firms invest when they have enough financial resources (savings).
Statistically significant is the debt level. However, there is the opposite relation
between capital expenditure and debt. A high debt level in the previous year limits
investments in fixed assets. It means that not only operational cash flow plays an
important role for investments, but also access to debt.
Based on the second model (Table 3), we concluded that LT companies do not
invest in fixed assets in response to changes in operating cash flow. The coefficient
for variable CFO (operational cash flow) is not statistically significant. It means that
this group is not financially constrained. However, similarly to HT companies, there
is a negative relationship between debt level and investment in LT firms, but weakest
compared to HT companies. The level of cash holding positively impacts the
expenditure level in LT companies; however, also the previous savings are less
critical than for HT firms. Every variable (except S) t-Welch test confirmed signif-
icant differences between HT and LT companies’ coefficients.
We run some robustness checks and divided our sample into subgroups using size
as the criterion (Table 4). If company sales were in the upper 50% of sample
distribution for the given year, it was treated as a BIG company otherwise
SMALL. We used every year subsampling (the company could move between
BIG/SMALL subgroups as time goes by) [method A] and a single subsampling
approach, where companies were divided based on the criterion of the majority of
every year subsampling [method B]. Both methods gave similar results.
210 K. Prędkiewicz et al.

Table 4 Investment model—size subsamples [method A]

BIG BIG SMALL SMALL
High technology Low technology High technology Low technology
Coeff. (std. error) Coeff. (std. error) Coeff. (std. error) Coeff. (std. error)
Const 0.02686 (0.01531) 0.02956 (0.00864) 0.00492 (0.01450) 0.02552 (0.00566)
* *** ***
S 0.01004 (0.00519) 0.01426 (0.00494) 0.01452 (0.01085) 0.00797 (0.00568)
* ***
CFO 0.12844 (0.03903) 0.00066 (0.02230) 0.04131 (0.05067) 0.02749 (0.02477)
***
DEBT 0.04493 0.05374 (0.01450) 0.05430 0.04317 (0.01241)
(0.02481)* *** (0.04551) ***
CASH 0.16782 (0.05893) 0.06792 (0.02847) 0.12604 (0.04495) 0.06818 (0.02197)
*** ** *** ***
Note: S—operating revenues, CFO—operational cash flow, DEBT—interest debt (long-term debts
and short-term loans), CASH—Cash and cash equivalent; *p-value < 0.10, **p-value < 0.05,
***p-value < 0.01

The robustness check (Table 4) confirms that only big high-technology compa-
nies are financially constrained, whereas small firms are not. We think that the result
can be linked with the size of risk equity capital available for all companies on the
capital market. The demand for equity capital for risky projects exceeds supply on
the less developed capital market, like Poland. Then only the demand of smaller
companies may be fulfilled. Investors should invest their money in a few projects
(companies) undertaking risky projects than in one big high-technology company
because they can diversify their capital. This phenomenon requires a more in-depth
analysis in future studies. However, the low technology companies, both small and
big, are not financially constrained, which is in line with our main findings.

5 Discussion and Conclusion

In this paper, we verified whether the Warsaw Stock Exchange (Poland) is an

effective means of alleviating financial constraints for high technology companies.
Based on the literature review, we supposed that the high-technology enterprises,
compared to low-technology firms, may be still financially constrained. It follows
from the literature review that for innovative companies and high technology firms
equity market is crucial. The high-technology firms have a specific financing
hierarchy—new projects (especially R&D) are financed first from internal equity
(cash flow) and then from external capital. However, not debt, but equity is firstly
preferred—opposite to classic pecking order theory. That is why the development of
the equity market should be a priority for governments. Although even well-
developed market-oriented financial systems (e.g., USA) have not entirely solved
financial constraints—even if financial constraints for investments in fixed assets
Technology Level and Financial Constraints of Public Listed Companies 211

partially disappeared, they still present for innovative investment. To answer our
research question, we used an indirect measure of financial constraints—the invest-
ment equation. The method assumes that when companies invest in reaction to
operational cash-flow changes, which represent internal capital—the company has
a problem with capital access, then is financially constrained. We executed models
for fixed assets—measuring them by capital expenditure from the cash flow state-
ment. We controlled future demand by the level of revenues, previous year level of
leverage, and past cash-holding. Two models—for high-technology and
low-technology companies were executed. Panel models with fixed effects
were used.
The statistically significant coefficient for variable CF confirmed that the HT
companies undertake investments when the operating cash flow is available. It
means that this group is financially constrained. Simultaneously, for LT coefficient
for the same variable (CF) was statistically insignificant. We have a basis for
concluding that LT public companies had no problem financing their investments
in fixed assets (then are not financially constrained). When we came to the control
variable—cash-holding—the models confirmed that previous savings are significant
for HT and LT companies but more critical for HT firms. The results may also be an
indication that the HT firms are financially constrained.
Summarizing, based on our results, we can conclude that the stock market in
emerging markets, like Poland, has not entirely solved the problem of capital access
for innovative, high-technology firms. However, it plays a role in the case of
non-technology companies. Our studies then contribute to previous research because
they extend the research area to the emerging markets. Moreover, we analyzed
financial constraints based on the newest data from IPO date to 2018 for the Warsaw
Stock Exchange data. The literature confirms that financial constraints may change
over time (Brown et al. 2009). Especially development on the financial market may
change the situation of innovative firms and measuring them, by methods we used,
help to answer the question of whether changes are going in the right direction.
However, our studies have limitations. We could not execute the model for
investment in R&D because such data are not available in Polish enterprises’
financial statements. Also, our criteria for dividing companies into high-technology
and low-technology firms is based on NACE codes. It would be better to collect
more information about companies’ activities and define the measure of technology
level and analyzed companies’ innovativeness.
Our research has implications for future studies. The models may be extended by
additional control variables—especially referring to new shares issuing. According
to previous studies, this variable should also be controlled in the investment model
(Brown et al. 2012). Moreover, it would also be interesting to analyze how specific
improvements implemented on the stock exchange market changed the financial
constraints over time.

Acknowledgments The project is financed by the Ministry of Science and Higher Education in
Poland under the programme “Regional Initiative of Excellence” 2019–2022 project number
015/RID/2018/19 total funding amount 10 721 040,00 PLN.
212 K. Prędkiewicz et al.

References

Aghion P et al (2004) Technology and financial structure: are innovative firms different? J Eur Econ
Assoc 2(2–3):277–288. https://doi.org/10.1162/154247604323067989
Audretsch DB, Elston JA (2002) Does firm size matter? Evidence on the impact of liquidity
constraints on firm investment behavior in Germany. Int J Ind Organ 20(1):1–17
Bah R, Dumontier P (2001) R&D intensity and corporate financial policy: some international
evidence. J Bus Financ Acc 28(5–6):671–692. https://doi.org/10.1111/1468-5957.00389
Brown JR, Fazzari SM, Petersen BC (2009) Financing innovation and growth: cash flow, external
equity, and the 1990s R&D boom. J Financ 64(1):151–185. https://doi.org/10.1111/j.1540-
6261.2008.01431.x
Brown JR, Martinsson G, Petersen BC (2012) Do financing constraints matter for R&D? Eur Econ
Rev 56(8):1512–1529. https://doi.org/10.1016/j.euroecorev.2012.07.007
Carpenter RE, Petersen BC (2002) Is the growth of small firms constrained by internal finance? Rev
Econ Stat 84(2):298–309. https://doi.org/10.1162/003465302317411541
Casson PD, Martin R, Nisar TM (2008) The financing decisions of innovative firms. Res Int Bus
Financ 22(2):208–221. https://doi.org/10.1016/j.ribaf.2007.05.001
Cincera M, Ravet J (2010) Financing constraints and R&D investments of large corporations in
Europe and the US. Sci Public Policy 37(6):455–466. https://doi.org/10.3152/
030234210X508642
Coad A, Segarra A, Teruel M (2016) Innovation and firm growth: does firm age play a role? Res
Policy 45(2):387–400. https://doi.org/10.1016/j.respol.2015.10.015
Czarnitzki D, Hottenrott H (2011) R&D investment and financing constraints of small and medium-
sized firms. Small Bus Econ 36(1):65–83. https://doi.org/10.1007/s11187-009-9189-3
de Rassenfosse G, Jensen P, Webster E (2011) Understanding innovation: the role of policy
intervention. A report for Victorian department of treasury and finance. https://www.
melbourneinstitute.com/downloads/industrial/Reports/Understanding%20innovation%202011.
pdf. Accessed 30 Sept 2016
Fazzari SM et al (1988) Financing constraints and corporate investment. Brook Pap Econ Act 1988
(1):141–206. https://doi.org/10.2307/2534426
Friend I, Lang LHP (1988) An empirical test of the impact of managerial self-interest on corporate
capital structure. J Financ 43(2):271–281. https://doi.org/10.1111/j.1540-6261.1988.tb03938.x
Hall B (1992) Investment and research and development at the firm level: does the source of
financing matter? NBER Working Paper 4096. National Bureau of Economic Research. http://
econpapers.repec.org/paper/nbrnberwo/4096.htm. Accessed 10 July 2016
Hall B (2002) The financing of research and development. Oxf Rev Econ Policy 18(1):35–51.
https://doi.org/10.1093/oxrep/18.1.35
Hall B (2010) The financing of innovative firms. Rev Econ Instit 1(1). http://www.rei.unipg.it/
index.php/rei/article/view/4. Accessed 1 Oct 2016
Hall B, Thoma G, Torrisi S (2007) The market value of patents and R&D: evidence from European
firms. Acad Manag Proc 2007(1):1–6. https://doi.org/10.5465/AMBPP.2007.26530853
Harhoff D (1998) Are there financing constraints for R&D and Investment in German manufactur-
ing firms? Ann Écon Stat 49(50):421–456. https://doi.org/10.2307/20076124
Himmelberg CP, Petersen BC (1994) R & D and internal finance: a panel study of small firms in
high-tech industries. Rev Econ Stat 76(1):38–51. https://doi.org/10.2307/2109824
Kaplan SN, Zingales L (1997) Do Investment-cash flow sensitivities provide useful measures of
financing constraints? Q J Econ 112(1):169–215. https://doi.org/10.1162/003355397555163
Konings J, Rizov M, Vandenbussche H (2003) Investment and financial constraints in transition
economies: micro evidence from Poland, the Czech Republic, Bulgaria and Romania. Econ Lett
78(2):253–258. https://doi.org/10.1016/S0165-1765(02)00210-0
Lewellen J, Lewellen K (2016) Investment and cash flow: new evidence. J Financ Quant Anal 51
(4):1135–1164. https://doi.org/10.1017/S002210901600065X
Technology Level and Financial Constraints of Public Listed Companies 213

Lööf H, Nabavi P (2016) Innovation and credit constraints: evidence from Swedish exporting firms.
Econ Innov New Technol 25(3):269–282. https://doi.org/10.1080/10438599.2015.1076196
Mazzucato M (2013) Financing innovation: creative destruction vs. destructive creation. Ind Corp
Chang 22(4):851–867. https://doi.org/10.1093/icc/dtt025
Nehrebecka N, Białek-Jaworska A (2015) Inwestycje polskich przedsiębiorstw w B+R a cash-flow i
dostępność kredytu bankowego [Investments of Polish enterprises in R & D and cash-flow and
availability of a bank loan]. Stud Ekon 3:366–385
OECD (2015) Entrepreneurship at a glance 2015. Organisation for Economic Co-operation and
Development, Paris. http://www.oecd-ilibrary.org/content/book/entrepreneur_aag-2015-en.
Accessed 10 Sept 2015
Scellato G (2007) Patents, firm size and financial constraints: an empirical analysis for a panel of
Italian manufacturing firms. Camb J Econ 31(1):55–76. https://doi.org/10.1093/cje/bel006
Schäfer D, Werwatz A, Zimmermann V (2004) The determinants of debt and (private) equity
financing: the case of young, innovative SMEs from Germany. Ind Innov 11(3):225–248.
https://doi.org/10.1080/1366271042000265393
Stiglitz JE, Weiss A (1981) Credit rationing in markets with imperfect information. Am Econ Rev
71(3):393–410
Tirole J (2006) The theory of corporate finance. Princeton University Press, Princeton, NJ
Ueda M (2004) Banks versus venture capital: project evaluation, screening, and expropriation. J
Financ 59(2):601–621
Ughetto E (2008) Does internal finance matter for R&D? New evidence from a panel of Italian
firms. Camb J Econ 32(6):907–925. https://doi.org/10.1093/cje/ben015
Wagenvoort R (2003) Are finance constraints hindering the growth of SMEs in Europe? EIB Pap 8
(2):23–50
 Part IV
Personal Finance
Differences in Use of Credit Products
Between the Old and New Member States
of the European Union

Agnieszka Huterska

1 Introduction

Financial behaviour of consumers are related not only to the amount of earned
income (Potocki 2018), but result primarily from financial literacy, which includes
financial knowledge, behaviour and attitude (OECD 2016: 8). Financial knowledge
alone is not sufficient to make rational financial decisions. The consumer’s ability to
use knowledge in the effective management of financial resources is necessary in
order to ensure financial well-being (OICU-IOSCO 2014: 5–6; NFCS 2015:1).
Therefore, financial literacy, i.e., a combination of knowledge, attitudes and behav-
ior related to financial decisions made, leads to the financial well-being of house-
holds (Musiał and Świecka 2016; Lusardi and Mitchell 2011; Maciejasz-
Świątkiewicz 2015). Financial literacy is crucial not only in terms of involving as
many consumers as possible in using basic banking products. It allows a rational
selection of products offered by financial institutions. Furthermore, the relationship
between the use of financial products and financial literacy is two-way. Less interest
shown in using financial products in some EU countries proves weaker
financial literacy among their inhabitants. However, the lack or very limited use of
financial products makes it difficult for such communities to practically develop
financial literacy through their personal experiences. Simpson and Buckland (2009),
Lusardi and Mitchell (2006), Lusardi et al. (2010) among others, indicated the links
between financial literacy, and financial exclusion and credit constraint in their
research. The conducted research (Musiał and Świecka 2016; OECD 2016: 8;
Frączek et al. 2017) indicates a low level of financial knowledge among young
people. This is an extremely worrying phenomenon because decisions made in

A. Huterska (*)
Nicolaus Copernicus University in Toruń, Toruń, Poland
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 217
K. Jajuga et al. (eds.), Contemporary Trends and Challenges in Finance, Springer
Proceedings in Business and Economics,
https://doi.org/10.1007/978-3-030-73667-5_13
218 A. Huterska

young adulthood at the beginning of professional activity, often affect the financial
situation in later life, including the risk of a debt spiral.
Adequate financial knowledge of young people affects not only the rational
management of current income, but also the awareness of the need to accumulate
savings throughout the entire period of professional activity, which is the subject of
research by Lusardi and Mitchell (2011). This is important when there is a need to
collect funds outside of mandatory retirement systems in order to have sufficient
funds after retirement. Particular attention should be paid to young adults and their
financial education. As indicated (Friedline and Rauktis 2014), young people ‘may
be the front lines of financial inclusion’. Financial inclusion in young adulthood is
more likely. However, reckless use of credit products may lead to excessive indebt-
edness in young adulthood and result in financial exclusion or remaining outside
mainstream banking. This threat, resulting from excessive indebtedness of young
people, is presented also in works by Williams and Oumlil (2015).
Numerous works, e.g., Demirguc-Kunt et al. (2015), Frączek (2017), Huterski
et al. (2020), Korzeniowska and Huterska (2020), indicate the existence of a strong
spatial differentiation of financial inclusion and the popularity of using specific
financial products. The results of the research conducted by Lusardi (2015) showed,
however, no correlation between the country’s GDP and the results of the test on the
knowledge and financial skills of young people. The level of financial skills depends
not only on the type and manner of transferring knowledge to young people through
formal education systems, which varies between countries. Banks, financial institu-
tions and various types of organizations, both governmental and non-governmental,
operating in individual countries also play an important role in providing informa-
tion about the benefits of using specific financial products and the related risks. The
influence of the peer group on young people also plays an extremely important role.
Taking into account the diversity of the European Union member states not only in
terms of education systems, but above all the experiences of individual societies in
taking advantage of the opportunities offered by the market economy.
The research objective of the study was to assess the specificity of disproportions
in the use of loan products by young people aged 15–24 in the old and new European
Union countries.
In the article, the following research hypothesis was formulated:
The differences in the level of financial exclusion of young people in the scope of
using credit products among new and old member states of the European Union
cannot be found.
The analyses conducted constitute an introduction to a broader study on the
indebtedness of young people in the European Union member states, and the
diversity of this phenomenon, while taking into account the level of financial
knowledge of young people.
Differences in Use of Credit Products Between the Old and New Member. . . 219

2 Data and Research Methods

The study analysed the degree of use by people aged 15–24 of banking products
such as a credit card and a loan in a financial institution. The analysis used data
obtained from the Global Findex Database (The World Bank 2017b), which contains
the results of a survey conducted among households in 2017. The data collected in
the Global Findex database are drawn from survey data covering almost 150,000
people in 144 economies—representing more than 97% of the world’s population.
The data used in the article come from a survey conducted in 2017 by Gallup Inc.
According to 2017 Global Findex Survey Methodology, they carried out surveys of
approximately 1000 people in each of the member states of European Union,1 using
randomly selected, nationally representative samples. The target population was the
entire civilian, noninstitutionalized population aged 15 and above. The interviews
were conducted by means of landline and mobile telephones*, face to face conver-
sations** and only with mobile telephones*** (The World Bank 2017a).
Statistical analysis of data on the degree of use of the financial products in
question was applied as a research method.
A preliminary analysis of the relationship between the shares of respondents
using specific types of financial services was also carried out. For the dependency
between having a credit card in and borrowing from a financial institution, Pearson
correlation coefficients were calculated and their significance was checked using
t-statistics (Sobczyk 1997, p. 253; Piłatowska 2006, p. 103; Kufel 2013):
In the European Union countries (EU_ALL) there is a moderate, positive corre-
lation (0.4320) between the share of credit card holders and the share of borrowing in
a financial institution.2 On the correlations the t-statistics prove that calculated
correlation for variables are statistically significant (i.e., the null hypothesis about
the irrelevance of the correlation coefficient at the significance level of a ¼ 0.05 is
rejected, |t| ¼ 2.442 > tα,s ¼ 2.056, where α ¼ 0.05, s ¼ 26.).
Taking into account the above dependencies, the further part of the paper presents
the differentiation of share of consumers’ own credit card and loan in a financial
institution in the old European Union member states compared to the new.

1
Austria* May 30–Jun 28, Belgium* Jul 11–Sep 18, Bulgaria** May 11–Jun 26; Croatia** May
23–Jul 9, Cyprus Apr 27–Jun 20, Czech Republic** Apr 4–Jul 11, Denmark May 5–May 30, Esto-
nia** Jun 15–Jul 15, Finland*** Apr 26–May 30, France * Apr 19–May 18, Germany* Apr 19–
May 18, Greece** May 20–Jun 16, Hungary** May 14–Jun 21, Ireland* Mar 14–Apr 10, Italy* Jan
30–Feb 23, Latvia** Jun 5–Jul 27, Lithuania** Jul 17–Aug 6, Luxembourg* Apr 19–May
18, Malta* Mar 17–Apr 15, Netherlands* Jul 11–Sep 1, Poland** Aug 12–Sep 25, Portugal*
Mar 27–May 3, Romania ** Apr 12–Jun 15, Slovak Republic** May 12–Jun 6, Slovenia* Mar 3–
Apr 5, Spain* Jan 30–Feb 23, Sweden* May 3 May 30, United Kingdom* Mar 14–Apr 10.
2
Correlation coefficients between credit card ownership and borrowed in financial institution,
calculated using the observations 1–28, 5% critical value (two-tailed) ¼ 0. 0217.
220 A. Huterska

United Kingdom 49%

Denmark 45%
Luxembourg 44%
Spain 25%
Ireland 25%
Belgium 24%
Austria 21%
Germany 21%
Finland 19%
Italy 18%
Netherlands 14%
France 13%
Portugal 13%
Sweden 12%
Greece 3%
0% 10% 20% 30% 40% 50% 60%

Fig. 1 Credit card ownership by young adults in the old European Union countries

3 Credit Card Ownership

In both groups there were countries where only about 3% of respondents declared
that they had credit cards. The highest proportion of respondents with a credit card
(48.57%) was in Great Britain (see Fig. 1), which until 2020 had belonged to the old
EU countries. The percentage of credit card holders in Denmark and Luxembourg,
belonging to the group of old member states of the European Union, and Malta (see
Fig. 2), which has been a member of the European Union since 2004, was at a similar
level. Greece was the country of the old European Union in which the number of
young adult people declaring owning a credit card was the smallest (3.03%). Also,
Sweden (12.36%), Portugal (12.65%), France (12.66%) and the Netherlands
(14.25%) were among the countries from the old EU with the number of people
using this financial product considerably below the median.
In the new member states of the European Union, only in Malta the percentage of
young adults owning a credit card (41.91%) was similar to the percentage of people
possessing this product in Denmark or Luxemburg (see Figs. 1 and 2). In Bulgaria,
Poland, Hungary and Romania the percentage did not exceed 5%.
Despite similar minimum and maximum values in both groups, the median in the
EU_old group (22.60%) was twice as high as the median in the new EU countries
group. On the one hand, this proves that the new EU countries belong to the
countries where the popularity of credit cards among young people is very low.
On the other hand, however, this group includes countries where young people use
this product to a similar degree as in the old EU countries.
Differences in Use of Credit Products Between the Old and New Member. . . 221

Malta 42%
Slovenia 26%
Cyprus 19%
Estonia 18%
Czech Republic 17%
Lithuania 14%
Slovak Republic 10%
Latvia 9%
Croatia 9%
Romania 4%
Hungary 3%
Poland 3%
Bulgaria 3%

0% 5% 10% 15% 20% 25% 30% 35% 40% 45%

Fig. 2 Credit card ownership by young adults in the new European Union countries

The much lower popularity of credit cards (The median ¼ 20.6% for EU_old and
10.0% for EU_new) when compared to accounts in financial institutions (The
median ¼ 92.6% for EU_old and 62.6% for EU_new) among people aged 15–24
results, on the one hand, from the natural income limitations for this age category,
and, on the other hand, from the existence of alternative methods of satisfying the
needs of short-term increase in available funds (see Table 1).
In the first case, one should take into account the limited creditworthiness of the
majority of young people due to the lack or limitation of personal permanent sources
of income. In the second case, the option of obtaining a current account credit facility
and using it with a debit card may affect the reduction of interest in credit cards. In
some countries, this may also be the result of the high popularity of easy, though
costly, indebtedness in non-bank loan companies, which, unlike banks, do not use
extensive creditworthiness assessment procedures (payday loans, e.g., the United
Kingdom, where the authorities recognized such debts among young people as a
serious problem—Rowlingson et al. 2016).
It is also important that the card owner pays fees and commissions to the financial
institution that provides the service.
Factors affecting the popularity of having credit cards among young people are
certainly even more complex and varied, as evidenced by the lack of a clear
geographic (north-south) or income (rich-poor) pattern in the distribution of credit
card popularity among countries, as in the case of accounts and debit cards. For
example, Slovenia, a post-socialist country from the EU_new group of countries, has
a share of credit card holders of almost 26%, while in the EU_old group in Greece
222 A. Huterska

Table 1 Statistics for credit card ownership (%)

Group n Sd 25% Me 75% min max Skew kurtosis Se
EU_new 13 13.6 4.4 10.0 17.9 2.8 41.9 1.4 2.5 11.1
EU_old 15 23.0 13.5 20.6 24.9 3.0 48.6 0.9 0.1 13.3
EU_ALL 28 18.6 9.5 17.4 24.2 2.8 48.6 1.0 0.4 13.0

Source: the author’s own calculations based on data from The World Bank (2017b)

this share is 3%, in Sweden it is just over 12%, similar to Portugal and France. Even
in the Czech Republic, another post-socialist country in the EU_new group, the share
of credit cards is over 17% despite the fact that it is a country with extremely low
popularity of having accounts in financial institutions and debit cards among young
people. At the same time, in neighbouring Poland (also a post-socialist EU_new
country), this share is the same as in Greece, i.e., 3%.

4 Taking Loans from a Bank or Other Formal Financial

Institution

While the possibility of getting into debt due to having a credit card was personalized
in the Findex survey, in the case of other forms of borrowing any money by
respondents, both indebtedness incurred individually and jointly with another person
were taken into account. Here we focus on borrowing from a bank or another type of
formal financial institution. Data on borrowing from family, relatives, or friends or
Differences in Use of Credit Products Between the Old and New Member. . . 223

Finland 25%
Luxembourg 18%
Netherlands 15%
Ireland 14%
Germany 13%
Belgium 13%
Sweden 11%
Denmark 11%
United Kingdom 8%
France 6%
Portugal 5%
Austria 4%
Spain 2%
Italy 2%
Greece 0%
0% 5% 10% 15% 20% 25% 30%

Fig. 3 Borrowing from financial institution by young adults in the old European Union countries

from an informal savings group/club as alternative sources of additional funds are

interesting material for comparison, however, it is beyond the scope of this article.
In the old EU, Finland was the country in which the most young adults had a debt
in a financial institution (%). However, in Portugal, Austria, Spain, Italy and Greece
the percentage did not exceed 5% (see Fig. 3).
The percentage of young adults with a debt in a financial institution in the group
of the new members of the European Union does not exceed 11%. The highest
percentage can be seen in Lithuania and Malta, the lowest in Latvia, Hungary,
Cyprus and Bulgaria (see Fig. 4).
Borrowing from a credit institution is even less popular among young people than
having a credit card. There were countries within the analysed groups where less
than 1% of young people used a bank loan or a loan from a financial institution. The
exceptions are the new EU countries, where the lowest share was over 2% in Latvia.
In this group, the median was also at the level of 8.8%, while in the country with the
highest use of loan products, i.e., Lithuania, this share was 10.3%, which is signif-
icantly below the highest values in the old EU (see Table 2).
A significant element that hinders inference on the basis of data on the popularity
of loans among the young is the differentiation of the percentage of students (tertiary
level education) in relation to people in the typical study age range between the
countries surveyed. According to the data of the UNESCO Institute for Statistics (for
2017–2018), tertiary school enrolment in Central Europe and the Baltics is 62%, in
the European Union 69%, in the OECD member countries 74%, and in all high
income countries it is 75%. However, within the European Union, there is a
224 A. Huterska

Lithuania 10%
Malta 10%
Slovenia 9%
Estonia 9%
Slovak Republic 5%
Czech Republic 5%
Romania 5%
Poland 5%
Croatia 5%
Bulgaria 4%
Cyprus 4%
Hungary 3%
Latvia 2%

0% 2% 4% 6% 8% 10% 12%

Fig. 4 Borrowing from financial institution by young adults in the new European Union countries

significant variation, from 47% in Slovakia, 88% in Latvia in the EU_new countries
group, and from 60% in the United Kingdom to 89% in Spain in the EU_old group.
We omit Luxembourg (19%), due to the specificity of this small country, and Greece
(137%), due to the specific use of higher education to combat unemployment in the
conditions of the economic crisis.
Data on the popularity of loans from financial institutions among adolescents
aged 15–24 are not accurate enough to distinguish having a student loan from other
types of loans, and also from ordinary consumer loans. This means that the correct
interpretation of data would require not only knowledge of the participation of
students in this age group, but also of the financial student support systems in
individual countries. The availability of preferential loans for students may in
practice be used to a different extent by young people from different countries,
due to the different structure and competitiveness of scholarship systems.
The differences in the level of life and financial independence of young people in
individual countries should also be considered, which may result from both eco-
nomic and cultural background, as well as from differences in the functioning of
banking systems in individual countries. All these factors require taking a cautious
approach when looking for certain regularities in the differences in the popularity of
borrowing from financial institutions by young people from different countries.
As mentioned above, credit facilities may take the form of an acceptable overdraft
in the current personal account or in the limits assigned to credit cards, which
naturally competes with typical loans up to a certain amount. Also, access to payday
Differences in Use of Credit Products Between the Old and New Member. . . 225

Table 2 Statistics for money borrowed from the bank or other formal financial institution (%)
Group n Sd 25% Me 75% min max Skew kurtosis Se
EU_new 13 5.9 4.2 4.9 8.8 2.1 10.3 0.6 1.0 2.7
EU_old 15 9.9 4.4 10.7 13.6 0.4 25.4 0.6 0.3 6.8
EU_ALL 28 8.0 4.3 5.8 10.9 0.4 25.4 1.2 2.0 5.6

Source: the author’s own calculations based on data from The World Bank (2017b)

loans from institutions that are not banks or other formal financial institutions, such
as pawnshops and many online loan companies, can have a similar effect.

5 Discussion and Conclusions

Several important conclusions of a different nature follow from the above consider-
ations. The new member states of the European Union (EU_new) as a group show a
lower degree of inclusion in financial services among young people aged 15–24 in
terms of credit cards and loans in a financial institution compared to the group of
countries that belonged to the EU already before May 2004 (EU_old). However, the
EU_new group turns out to be highly diversified and this is not only due to the fact
that Malta and Cyprus are the only countries in this group that have not gone through
the era of real socialism. High levels of inclusion in financial services, as evidenced
not only by the percentage of people who have an account, but also by those who use
other financial products and services, including loans, are present in Slovenia and
226 A. Huterska

Estonia and the distance between them and the weakest EU_new countries in this
area, i.e., Bulgaria, Croatia, and Romania, is significant.
Among the countries of the EU_old group, attention should be paid to the low
level of having a credit card and a loan in a financial institution in southern European
countries, i.e., Greece, Italy, and Spain, compared to northern countries such as the
Netherlands, Finland, Denmark, and Sweden. However, the disproportions among
the EU_old countries are not as strong as among the EU_new countries. The
exception is Greece, which clearly stands out from other EU_old countries and has
a banking services level comparable to that of Bulgaria or Romania.
The above comments indicate that although the differences in the wealth of the
EU_old and EU_new countries may partly explain the differences in the degree of
use of credit products by young people in these countries; however, this link is not
clear. The examples of Slovenia, Estonia, and Latvia indicate that it is possible to
promote the use of banks and other formal financial institutions among young people
effectively also in relatively less wealthy countries.
Careless usage of credit products can lead to the situation in which young people
at an early stage of the working life incur debts, which will in the future prevent them
from using any credit offer of banks. In Poland, according to the data presented by
BIG InfoMonitor and Biuro Informacji Kredytowej (BIK), the credits granted to the
18–24 age group amount to 1.3% of the value and 4% of the number of all active
credits. It is also significant that consumer credits constitute the share of almost 70%
(cash loans—40.7% and instalment loans—24.8%), overdraft limits amount to
20.9%, credit card debts amount to 12.2% and mortgage loans to as little as 1%.
Reckless expenditure, living beyond means or unexpected spending with no savings
lead to the fact that one out of every 20 people aged 18–24 experiences financial
difficulties (BIG InfoMonitor 2019; Busines Insider Polaska 2019). Moreover,
people in the 18–24 age group constitute 1.5% of unreliable debtors in Poland
(BIG InfoMonitor 2020). Having an account, which is the first step to financial
inclusion, creates an opportunity for young people to develop proper financial habits,
the ability to rationally manage their budget and to collect savings. This is extremely
important in a situation where young people show a need for credit products later in
their lives. However, only skillful use of this group of products, with an appropriate
level of financial literacy covering not only knowledge, but also financial habits, can
prevent young people from falling into a spiral of debt in the future.

Acknowledgement This work was supported by the National Science Centre, Poland under Grant
No. 2017/26/E/HS4/00858.

References

BIG InfoMonitor (2020) Info Dług, Ogólnopolski raport o zaległym zadłużeniu i niesolidnych
dłużnikach, 36 edycja, March. Retrieved from https://media.big.pl/publikacje/att/1697945
Differences in Use of Credit Products Between the Old and New Member. . . 227

BIG InfoMonitor (2019) Zaległości młodych przekroczyły 1 mld zł, Press release, 16 September.
Retrieved from https://media.bik.pl/informacje-prasowe/464473/zaleglosci-mlodych-
przekroczyly-1-mld-zl
Busines Insider Polaska (2019) Długi młodych rosną. Rekordzistą 23-latek z 3,5 mln zł zaległości,
1 October. Retrieved from: https://businessinsider.com.pl/twoje-pieniadze/budzet-domowy/
zadluzenie-mlodych-dane-o-dlugach-polakow-w-wieku-18-24-lata-big-infomonitor/te492zn
Demirguc-Kunt A, Klapper L, Singer D, Van Oudheusden P (2015) The global Findex database
2014: measuring financial inclusion around the world. World Bank policy research working
paper, 7255, Washington. Retrieved from: http://www.worldbank.org/globalfindex
Frączek B (2017) Edukacja finansowa jako determinanta wzrostu włączenia finansowego.
Podejście zintegrowane. Wydawnictwo Uniwersytetu Ekonomicznego w Katowicach,
Katowice
Frączek B, Bobenič Hintošová A, Bačová M, Siviček T (2017) Simultaneous use of the financial
literacy level and the financial inclusion degree as a result of financial education efficiency in
Visegrad Group countries. J Econ Manage 27(1):5–25. https://doi.org/10.22367/jem.2017.
27.01
Friedline T, Rauktis M (2014) Young people are the front lines of financial inclusion: a review of
45 years of research. J Consum Aff 48(3):535–602. https://doi.org/10.1111/joca.12050
Huterski R, Huterska A, Łapińska J, Zdunek-Rosa E (2020) The problem of savings exclusion and
gross savings in the new European Union member states. Enterp Sustain Iss 7(3):2470–2480.
https://doi.org/10.9770/jesi.2020.7.3(67)
Korzeniowska A, Huterska A (2020) Saving inclusion in the European Union countries – trends and
differences. In: Soliman KS (eds) Proceedings of the 35th international business information
management association conference (IBIMA), 1–2 April 2020, Seville, Spain, ISBN: 978-0-
9998551-4-1. Education Excellence and Innovation Management: A 2025 Vision to Sustain
Economic Development during Global Challenges, 12, pp 9844–9854
Kufel T (2013) Ekonometria. Rozwiązywanie problemów z wykorzystaniem programu GRETL.
Wydawnictwo Naukowe PWN, Warszawa
Lusardi A (2015) Financial literacy skills for the 21st century: evidence from PISA. J Consum Aff
49(3):639–659
Lusardi S, Mitchell OS (2006) Baby boomer retirement security: the roles of planning, financial
literacy, and housing wealth. NBER Working Paper Series. https://doi.org/10.1017/
CBO9781107415324.004
Lusardi A, Mitchell OS (2011) Financial literacy and retirement planning in the United States. J
Pension Econ Financ 10(4):509–525
Lusardi A, Mitchell OS, Curto V (2010) Financial literacy among the young: evidence and
implications for consumer policy. J Consum Aff 44(2):358–380
Maciejasz-Świątkiewicz M (2015) Financial exclusion as a result of limited financial literacy in the
context of the financialization process [Wykluczenie finansowe jako skutek ograniczonych
umiejętności finansowych w kontekście procesu finansjalizacji]. Financ Sci 4(4). https://doi.
org/10.15611/nof.2015.4.06
Musiał M, Świecka B (2016) Analiza wiedzy i umiejętności finansowych młodego pokolenia. Acta
Universitatis Lodziensis. Folia Oeconomica 6(326):203–216
NFCS (2015) National financial capability strategy. Raising Nationwide Financial Capabilities: A
Review of Recommended Strategies White Pape, November
OECD (2016) OECD/INFE international survey of adult financial literacy competencies. OECD,
Paris. www.oecd.org/finance/OECD-INFE-International-Survey-of-Adult-Financial-Literacy-
Competencies.pdf
OICU-IOSCO (2014) Strategic framework for investor education and financial literacy. Final
report. The Board of International Organization of Securities Commissions. FR 09/14, October
Piłatowska M (2006) Repetytorium ze statystyki. Wydawnictwo Naukowe PWN, arszawa
Potocki T (2018) Deternminanty wyborów finansowych gospodarstw domowych o niskich
dochodach. Wydawnictwo Uniwersytetu Rzeszowskiego, Rzeszów
228 A. Huterska

Rowlingson K, Appleyard L, Gardner J (2016) Payday lending in the UK: the regul(aris)ation of a
necessary evil? J Soc Policy 45(3):527–543
Simpson W, Buckland J (2009) Examining evidence of financial and credit exclusion in Canada
from 1999 to 2005. J Socio-Econ 38(6):966–976. https://doi.org/10.1016/j.socec.2009.06.004
Sobczyk M (1997) Statystyka. Wydawnictwo Naukowe PWN, Warszawa
The World Bank (2017a) The 2017 global Findex survey methodology. Retrieved from https://
globalfindex.worldbank.org/index.php/#GF-ReportChapters)
The World Bank (2017b) The global Findex database 2017. Retrieved from https://globalfindex.
worldbank.org/#about_focus
Williams AJ, Oumlil B (2015) College student financial capability: a framework for public policy,
research and managerial action for financial exclusion’s prevention. Int J Bank Mark 33
(5):637–653. https://doi.org/10.1108/IJBM-06-2014-0081
Diversified Risky Financial Assets
in Portfolios of Risk-Averse Households:
What Determines Their Occurrence?

Katarzyna Kochaniak and Paweł Ulman

1 Introduction

Financial risk tolerance is a behavioural finance term which can be defined as the
maximum amount of uncertainty that someone is willing to accept when making a
financial decision (Grable and Joo 2004) or the willingness to engage in a financial
behaviour in which the outcomes are uncertain with a possible identifiable loss
(Irwin 1993). It can be analysed within two approaches—subjective (based on
individuals’ self-assessments of risk attitudes) and objective (related to individuals’
actual financial risk-taking). According to the so-far literature, perceptions of own
risk tolerance determine investment choices and thus the composition of portfolios.
However, the measurement of financial risk tolerance based on the above approaches
does not always lead to consistent results. In such cases, under- or overexposure to
financial risk can be concluded (Marinelli et al. 2017).
The overexposure to financial risk can be assumed especially regarding house-
holds that declare unwillingness to take any risk (i.e., risk aversion) but participate in
risky assets. In the euro area countries, this phenomenon is not commonly observed;
however, in Belgium, Cyprus, Finland, France, Germany, Ireland, Luxembourg,
Malta, and Spain it may affect from 10% to 35% of risk-averse households
(Kochaniak and Ulman 2020). According to the microdata from the Eurosystem’s
Household Finance and Consumption Survey (HFCS), it is mostly due to their
investments in one type of risky assets. Diversified risky parts of portfolios can be
recognised more often in risk-averse households residing in Belgium, Finland,
France, Germany, and Spain.

K. Kochaniak (*) · P. Ulman

Cracow University of Economics, Kraków, Poland
e-mail: [email protected]; [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 229
K. Jajuga et al. (eds.), Contemporary Trends and Challenges in Finance, Springer
Proceedings in Business and Economics,
https://doi.org/10.1007/978-3-030-73667-5_14
230 K. Kochaniak and P. Ulman

It should be noted that financial risk tolerance in both approaches became invoked
in the EU regulations—the MiFID II (Directive 2014/65/EU) and MiFIR (Commis-
sion Delegated Regulation 2017/565), which promote the rules of ‘suitability’,
‘know your client’ and ‘know your product’. They require practitioners to provide
individuals and households only with products meeting their investment objectives
and preferences. Thus, the provisioning of retail clients who declare risk aversion
with risky financial assets became legally prohibited.
Our study is based on microdata derived from the second wave of the
Eurosystem’s Household Finance and Consumption Survey. It focuses on house-
holds declaring unwillingness to take any financial risk (that can be named subjec-
tively risk-averse) and residing in Belgium, Finland, France, Germany and Spain.
Because part of them indeed participate in risky financial assets, we aim to identify
socio-demographics and socio-economics, which are decisive for their propensity to
diversify these items in portfolios. This paper seeks to answer the following research
questions related to the households in question:
1. Which characteristics are conducive to their participation in diversified types of
risky assets?
2. Which characteristics foster their consistent behaviour on the financial market,
i.e., preventing them from investments in risky assets?
Thus, based on socio-demographics and socio-economics, we aim to describe two
separate profiles of households that declare risk aversion—with a tendency to the
diversification of risky financial asset and with a tendency to remain free from risks.
Our study extends the existing research line of inquiry regarding the subjective
and objective risk tolerance, in relation to households declaring risk aversion. It fills
the gap in the so-far literature, providing knowledge about their characteristics that
favour holdings of diversified risky financial assets, despite beliefs in own unwill-
ingness to take any risk, as well as fostering limitation of their portfolios solely to
riskless components.
The results of our study may have implications for both practice and policy. The
findings can be noteworthy for the EU financial entities obliged to fulfil the MiFID II
and MiFIR especially when they offer financial assets to households declaring risk
aversion due to their vulnerability to wrong decisions. A more profound diversifi-
cation of risky assets in such a household may signal its inability to assess own risk
attitude. Furthermore, it may signal the need to re-examine its objectives and
preferences by a professional prior to offering retail financial products or services.
Our findings may also be useful for financial entities operating cross-border since we
find the differences in domestic profiles of households affected by the conflict
between their perceived risk attitude and composition of financial asset portfolios.
The policy implications refer to consumer protection since a significant part of the
risk-averse households makes financial decisions on their own (EC 2012). The
prevalence of households with a gap between subjective and objective financial
risk tolerance in a population, which signals their overexposure to financial risks,
may lead to negative social consequences, especially under stress on the financial
market.
Diversified Risky Financial Assets in Portfolios of Risk-Averse. . . 231

The remainder of the paper is as follows: Section 2 contains an overview of the

subject literature, Sect. 3 presents the method and data applied. Section 4 contains
the results of empirical analysis. Section 5 provides conclusions.

2 Related Literature

The attitude towards risk and actual risk-taking are discussed in the so-far literature,
however, more often concerning specified sub-groups of individuals or households
(Marinelli et al. 2017; Moreschi 2005) than countries’ populations (Martin 2011;
Warneryd 1996). It should be noted that the so-called subjective and objective
financial risk tolerance is analysed with the use of various measures that have their
own strong and weak sides.
Subjective risk tolerance is often discussed with the use of the single question
self-classification method, which is valued for its simplicity and easy collection of
information required. It consists of the question about own risk attitude and proposed
answers—from the willingness to take substantial risks to the risk-aversion. This
method is often applied by researchers and practitioners (Marinelli et al. 2017;
Grable 2016; Hanna et al. 2008; Roszkowski and Grable 2005; Chang et al. 2004;
Grable and Lytton 2001; Schooley and Worden 1996) and adopted in national
surveys, such as, for instance, the Eurosystem’s Household Finance and Consump-
tion Survey or the US Survey of Consumer Finances. However, it is not free from
weaknesses such as ambiguous wording of ‘substantial’, ‘above average’ and
‘average’, which respondents may interpret differently (Kimbal et al. 2007). More-
over, the lack of ‘willingness to take less-than-average financial risk’ among pro-
posed responses may be a potential cause of too often declarations of risk aversion
(Schooley and Worden 1996). Assessments of subjective risk attitudes in economic
behaviour are also based on the choices between lotteries, but they suffer from
non-homogeneous outcomes. Due to this, verbal descriptions of situations are
implemented, with different options to choose (Warneryd 1996). However, they
may give rise to associations that obscure the measure of a respondent’s subjective
risk attitude.
Objective risk tolerance relates to actual investment behaviour and is measured on
the basis of the riskiness of financial asset portfolios. It is often described by the ratio
of a respondent’s risky assets to his or her wealth, but also by a degree of the
diversification of the risky part of a portfolio. However, both measures’ weak side is
limited control over contextual variables related to liquidity needs or financial
constraints of a respondent and market expectations, influencing behaviour beyond
risk tolerance (Nosic and Weber 2007). Warneryd (1996) finds the first measure
useful only if applied to respondents with high incomes and risky assets in portfolios,
while the latter recognises as universal and informing about the risk-taking-progres-
sion. In his opinion, the simplest behavioural measure of a portfolio’s riskiness is
that a household holds risky assets. Limitations of the ratio of risky assets to wealth
as a measure of risk tolerance are also discussed by Hanna et al. (2001) who
232 K. Kochaniak and P. Ulman

emphasise its unusefulness regarding individuals or households that do not partic-

ipate in any financial asset.
The aforementioned weaknesses of subjective and objective risk tolerance mea-
sures should be kept in mind when applied in research studies as a proxy for true risk
tolerance of individuals or households. They may lead to equivocal results obtained
within both approaches, not necessarily signifying the underexposure or overexpo-
sure to financial risks. It should be noted that the so-far literature identifies gaps
between subjective and objective financial risk tolerance; however, to a limited
extent regarding households we focus on—risk-averse but risk-taking.
Due to the aim of our study, it is essential to know which socio-demographics and
socio-economics can be concluded as decisive for subjective and objective financial
risk tolerance. The literature provides vast evidence in this regard. A respondent’s
age is found to negatively influence risk tolerance (Bucciol and Miniaci 2010; Sahm
2007; Yao et al. 2005; Coleman 2003; Bakshi and Chen 1994). It is explained by the
fact that young investors have more time to recover from losses. However, the
relationship between age and risk tolerance may not be linear. The same direction
of impact is recognised for household size (Calvet and Sodini 2014; Yao and Curl
2011). Larger households tend to be more conservative in their risk attitudes since
their size negatively influences the wealth per capita and positively the committed
expenditure-to-wealth ratio. Moreover, they are more exposed to the risk of mem-
bers’ random needs (Calvet and Sodini 2014; Yao and Curl 2011). Among the
determinants positively influencing risk tolerance, there is a level of education
completed (Gilliam et al. 2008; Chang et al. 2004; Grable 2000), as well as wealth
and incomes, including their levels and sources (Gibson et al. 2013; Bucciol and
Miniaci 2010; Deaves et al. 2007; Chang et al. 2004; Grable and Joo 2004; Hallahan
et al. 2004; Grable 2000). More formal education is concluded to make an assess-
ment of risk-return trade-offs easier (Gilliam et al. 2008; Chang et al. 2004; Grable
2000). Considerable wealth or incomes allow overcoming market requirements,
including entry costs (e.g., advising fees) and minimum value of investments
(Vissing-Jorgensen 2002; Haliassos and Bertaut 1995). The gender of a responding
person is recognised significant as well since males are found more risk-tolerant than
females (Weller et al. 2010; Chang et al. 2004; Slovic 2004; Weber et al. 2002;
Grable 2000; Jianakoplos and Bernasek 1998). The respondent’s status on the labour
market also matters, and the self-employed are recognised more willing to be
exposed to risks (Alessie et al. 2002; Stewart and Roth 2001). Hallahan et al.
(2004) and Yao and Hanna (2005) find the marital status significant and conclude
greater risk-tolerance of singles than couples. The limited willingness of couples to
take risks may be supported by similar reasons to those discussed regarding large
households. It should be noted that selected studies provide opposite findings or do
not recognise the above characteristics as statistically significant (Calvet and Sodini
2014; Bucciol and Miniaci 2010; Deaves et al. 2007; Grable 2000).
A description of risk-averse retail investors who hold risky financial assets in
portfolios is presented in brief in the study of Marinelli et al. (2017). It is based on the
characteristics selected from the above ones, such as high income and lack of
dependant children, but also on financial literacy, and less cautious economic
Diversified Risky Financial Assets in Portfolios of Risk-Averse. . . 233

behaviour. However, these outcomes refer to a non-representative group of respon-

dents in question. Our study is based on data related to the populations of selected
euro area countries. It supplements our previous research profiling households in
question and identifying the causes of their overexposure to financial risks
(Kochaniak and Ulman 2020) since it discusses the phenomenon of their holdings
of diversified risky assets. To some extent, our study recalls Warneryd’s (1996).
Both are cross-sectional and based on nationally representative data. Moreover, both
use information about the numbers of risky asset types in portfolios as a proxy of
actual risk-taking of households. According to Warneryd (1996), this measure is
correlated with households’ subjective risk attitudes. However, both studies differ in
terms of their aims and methods, as well as the targeted group of households which
in Warneryd’s research consists of respondents of all possible ranges of subjective
and objective risk tolerance.

3 Method and Data

We conduct a cross-sectional study in which we analyse the phenomenon of the

diversification of risky financial assets in the portfolios of households declaring risk
aversion.
The study applies the Poisson regression model, in which the dependent variable
Y is a count variable with a conditional Poisson distribution in relation to household
characteristics (Hilbe 2014). This means that for each household with specific
characteristics, the distribution is different. It can be seen in diverse values of the λ
distribution parameter for individual households. Regarding this, the parameter λi
(for the i-th household) should be a function of the vector of characteristics (xi).
Since in the Poisson regression model this function should take non-negative values,
we apply its exponential form described as:
0
λ i ¼ exi β , ð1Þ

Thus, the Poisson regression model with parameter estimates based on the
Maximum Likelihood Estimation (MLE) can be presented as follows:

lnE ðyi jxi Þ ¼ ln λi ¼ x0i β: ð2Þ

It should be noted that the variance and expected value in the Poisson distribution
are equal. If the value of variance is higher, the overdispersion may occur and lead to
a kind of heteroscedasticity. In such a case, an appropriate assessment of standard
errors should be ensured, or a negative binomial distribution applied for the count
variable (Winkelmann 2008).
234 K. Kochaniak and P. Ulman

In our study, the dependent variable (2) refers to the number of risky financial
asset types in the portfolio of a household that declares risk aversion. It takes values
from 0 to 4.
Based on the so-far literature, we propose a set of independent variables
recognised as statistically significant for the formation of subjective and objective
risk tolerance of a household. This set refers to the following socio-demographics
and socio-economics:
1. Quintile class of total gross income of a household, at a country level (dummies):
TGI_1Q—the first (reference variable), TGI_2Q – the second quantile,
TGI_3Q—the third quantile, TGI_4Q – the fourth quantile, TGI_5Q—the fifth
quantile
2. Type of income of a household (dummies): I_Empl—employee income,
I_SEmpl—self-employment income, I_Pens—income from pensions,
I_STrans—regular social transfers (except pensions)
3. Number of adult members of a household (discrete variable): N_Adult
4. Number of dependant children in a household (discrete variable): N_Child
5. Education level of the responding person (dummies): E_1L—primary and lower
(reference variable), E_2L1S—lower secondary, E_2L2S—upper secondary,
E_3L—tertiary
6. Marital status of the responding person (dummies): MS_S (reference variable)—
single (never married); MS_M&CU—married and in a consensual union on a
legal basis, MS_Wid—widowed, MS_Div—divorced
7. Age of the responding person (dummies): A < 25 (reference variable), A_25-39,
A_40-54, A_55+
8. Gender of the responding person (a dummy): Gender—1 if male.
We apply household-level data derived from the second wave of the
Eurosystem’s Household Finance and Consumption Survey (ECB 2016), which
provides information about the distribution of households’ socio-demographics
and socio-economics in 20 euro area countries. However, due to the aim of this
study, we focus solely on populations in which the phenomenon of risky asset
diversification can be recognised—of relatively large fractions of risk-averse house-
holds with two or more types of risky financial assets in portfolios. At the threshold
of 3%, the following countries are recognised as being in the interest of the study:
Belgium, Finland, France, Germany, and Spain. Within this subset, the fractions
mentioned vary from 3% to 14% (Table 1). It is worth adding that in 10 out of
20 euro area countries participating in the HFCS, which are Austria, Cyprus,
Estonia, Greece, Ireland, Italy, Latvia, Portugal, Slovenia, and Slovakia, the problem
analysed almost does not exist.
The total number of households considered in our study is 25,829, while regard-
ing individual countries it ranges from 1617 (in Belgium) to 9266 (in France).
In the HFCS database, the subjective risk attitude of a household is recognised on
the basis of the single question self-classification method, which boils down to
asking ‘Which of the following statements comes closest to describing the amount
of financial risk that you (and your husband/wife/partner) are willing to take when
Diversified Risky Financial Assets in Portfolios of Risk-Averse. . . 235

Table 1 Number of households declaring risk aversion and holding 0–4 types of risky financial
assets in portfolios in selected euro area countries
No. of types of risky financial
assets Belgium Finland France Germany Spain Group
0 1329 4084 7225 2253 3609 18,500
1 244 2161 1538 356 954 5253
2 43 866 470 120 295 1794
3 1 163 33 19 59 275
4 0 2 0 0 5 7
Total of households 1617 7276 9266 2748 4922 25,829

you save or make investments?’ This method limits possible responses to the
following: (1) Take substantial financial risks expecting to earn substantial returns;
(2) Take above average financial risks expecting to earn above average returns;
(3) Take average financial risks expecting to earn average returns; (4) Not willing to
take any financial risk. Since we are solely interested in households with subjective
risk aversion, we focus our study on respondents indicating the last answer.
Apart from the declared risk aversion, we are interested in households’ financial
assets. Part of them hold portfolios, in which risky assets of different types can be
identified. Thus, for each risk-averse household, we count the following types of
financial assets in the case of which the loss of the capital invested is possible:
1. Mutual fund units
2. Bonds, except state and other general government ones
3. Publicly traded shares
4. Other equities related to non-self-employment not publicly traded businesses.

4 Results

The Poisson regression provides us with useful results for discussing the propensity
of households with subjective risk aversion to hold diverse types of risky assets, both
for individual countries as well as for the entire group of these countries. The results
allow distinguishing both profiles being the subject of our interest—involved in
different types of risky assets and consistent, i.e., avoiding risky financial assets.
Most of the parameter estimates of Poisson models for the countries and the group
are statistically significant at 0.1 or less (Table 2). It can be concluded on the basis of
the standard Wald test. The lowest number of independent variables with statistically
significant parameter estimates concerns Germany. However, the Likelihood Ratio
test proves that all models applied in the study contribute significant information to
explaining the variability of the dependent variable. It is worth adding that the
problem of overdispersion is recognised solely in the model for Germany, for
which we apply negative binomial distribution. Despite this, the differences between
236 K. Kochaniak and P. Ulman

Table 2 Results obtained from the Poisson regression model (dependent variable: the number of
types of risky financial assets held by a household declaring risk aversion)
Variable Belgium Finland France Germany Spain Group
Intercept 3.8467 2.5706 4.0300 3.6429 3.0541 3.1843
TGI_2Q 0.7733 0.3389 0.6030 – 0.4574 0.4611
TGI_3Q 0.9718 0.6295 1.0198 0.4361 0.8700 0.8192
TGI_4Q 1.5212 0.8786 1.5483 0.8766 1.1261 1.1433
TGI_5Q 1.7230 1.0974 2.2804 1.5336 1.7058 1.5349
N_Adult 0.4032 0.1363 0.1296 0.3920 0.1330 0.1005
N_Child 0.1702 – 0.0455* – – –
Educ_2L1S 0.4794* 0.7318 0.4471 – 0.4639 0.9236
Educ_2L2S 0.5112 0.7300 0.4336 1.2930 0.7231 0.7886
Educ_3L 0.6626 0.9125 0.7711 1.5866 1.0721 1.0670
I_Empl – 0.0897 0.0998* – 0.4174 –
I_SEmpl – 0.2756 – – – 0.3766
I_Pens 0.5393 0.1715 – 0.2794 – 0.3026
I_STrans 0.2982 0.0932 0.1120 – 0.1891 0.0582
MS_M&CU 0.3351 – – – 0.2434 0.0990
MS_Wid – – – – 0.2298 0.1765
MS_Div 0.5446 0.1028 – 0.6669 0.4495 0.2850
A_25–39 – – 0.6795* – – –
A_40–54 0.8822 0.1517 0.8686 0.5552 0.6294 0.2276
A_55+ 1.0601 0.3527 0.8881 0.7068 1.1414 0.4905
Gender – – 0.1330 – – –
Note: () refers to a significance level above 0.1; (*) denotes significance at the level of 0.051–0.1;
in the remaining cases, the p-value did not exceed 0.05

the individual parameters estimates in the Poisson and negative binomial distribu-
tions do not differ significantly and do not affect the merit conclusions.
The outcomes obtained from the model for the group of countries allowed us to
conclude about the significance of selected characteristics of risk-averse households
for the formation of the number of risky asset types they hold. The propensity to
diversify risky parts of portfolios is found to be increasing along with increases in the
levels of their incomes, and education level and age (aged 40 and over) of
responding persons (Table 2). More risky asset types can be expected in households
receiving incomes from selected sources, mostly self-employment and pensions,
ceteris paribus. Additionally, they are recognised among relatively small house-
holds, particularly of the singles, if we measure their size by the number of adult
members.
Opposite to the above, the propensity to make investment choices adequate to
declared risk aversion can be recognised mostly among large, low-income house-
holds of relatively young (up to 39 years of age) and poorly educated responding
persons, also divorced or widowed ones.
It should be noted that the statistical significance of characteristics such as the
number of dependants or the gender of a responding person is not confirmed at the
Diversified Risky Financial Assets in Portfolios of Risk-Averse. . . 237

group level. Also, we cannot conclude about the peculiar investment behaviour of
risk-averse households living on incomes from employment.
The statistically significant characteristics mentioned above also emerged as
significant for the analysed propensity at the country level. In each of the countries,
the propensity to hold diverse types of risky assets increases along with increases in
incomes of risk-averse households, ceteris paribus. However, in Germany, more
types of risky assets in portfolios can be identified only within a higher range of this
characteristic (at least assigned to the third quintile class of total gross income).
Regarding the types of incomes, their significance for the analysed phenomenon
is not so clear cut. In Finland, more types of risky asset can be recognised in
households with incomes from self-employment and pensions, in Belgium among
living on pensions and regular social transfers, while in Germany solely among
living on pensions, ceteris paribus.
In selected countries, a greater propensity to hold diversified risky assets is
conditioned by a higher level of education completed by a responding person. In
Belgium and Spain, the strength of the impact of this characteristic increases within
its full range—from primary and lower up to tertiary. However, in all of the countries
analysed, the most diversified financial risky assets should be expected among
respondents who graduated from tertiary education, ceteris paribus.
The analysed propensity remains under the influence of the age of responding
persons as well. In France, the impact of this characteristic strengthens along with its
increase from the lowest to the highest range. In the remaining countries, its growing
influence is confirmed starting from the age of 40, ceteris paribus.
The significance of household size for the analysed phenomenon turned out to be
equivocal, and dependent on whether it is described by the number of adult members
or children. Moreover, regardless of how this characteristic is defined, it is
recognised as a stimulant in selected countries, while in the others as a destimulant.
In Spain and Finland, more types of risky assets should be expected in subjectively
risk-averse households with more adults, while in Belgium, Germany, and France
with their lower numbers. Regarding the number of dependant children, its positive
impact on the discussed propensity is recognised in France, but negative in Belgium.
In all countries, except France, selected marital status types allow us to expect
more types of risky assets, ceteris paribus. This refers to singles in Spain, couples
(married and in a consensual union) in Belgium, while in Germany and Finland
regarding singles, couples, and widowed. Moreover, in France, greater diversifica-
tion of risky assets should be observed among risk-averse households represented by
males.
Out of all considered socio-demographics and socio-economics, we can also
indicate the ones favouring the adequacy of investment behaviours of the households
that declare unwillingness to take the risk. They are in line with the conclusions
based on the model for the entire group of countries. The consistency between what
household members think and what they actually do should be expected mostly
among living on low incomes (assigned to the lowest range), as well as represented
by young, poorly educated and divorced persons. It can also be recognised among
relatively small households (regarding the number of adult members) in Finland and
238 K. Kochaniak and P. Ulman

Spain, but large ones in Belgium, France and Germany. Moreover, lower propensity
to hold diversified risky assets is recognised in French households without depen-
dants and Belgian ones with their large numbers, ceteris paribus. In Finland, France,
and Spain, we can conclude about the tendency to riskless investment behaviour of
the households living on incomes from employment and regular social transfers.
Additionally, in France, less interest in risky assets is found in households
represented by females.
The Likelihood Ratio test showed that all models contribute significant informa-
tion to explaining the variability of the dependent variable.

5 Conclusions

According to the results obtained, the propensity to hold diversified types of risky
assets can be recognised in selected euro area countries, among subjectively risk-
averse households of particular socio-demographics and socio-economics.
It characterises mostly the wealthiest ones (with incomes from the largest range),
represented by well-educated persons aged 40+. Moreover, in most countries, this
phenomenon refers to households with responding persons of marital status other
than divorced. The significance of the remaining characteristics, such as, for
instance, the size of a household, sources of incomes, and gender of the responding
person is confirmed only at the domestic level, and their direction of impact may
differ.
The opposite conclusions can be drawn about households whose behaviour on the
financial market is consistent with their declared risk aversion, i.e., avoiding risky
assets. Such an attitude is more evident among households with lower incomes and
represented by poorly educated and relatively young persons. However, in selected
countries, the significant impact of the size of a household and source of incomes
(mostly from employment and regular social transfers) can be recognised, as well as
the responding persons’ marital status of being divorced.

References

Alessie RJM, Hochguertel S, van Soest A (2002) Household portfolios in the Netherlands. In:
Guiso L, Haliassos M, Jappelli T (eds) MIT, Cambridge, pp 341–388
Bakshi GS, Chen Z (1994) Baby boom, population aging, and capital markets. J Bus 67(2):165–202
Bucciol A, Miniaci R (2010) Household portfolios and implicit risk preference. https://www.unibs.
it/sites/default/files/ricerca/allegati/1006.pdf. Accessed 15 May 2020
Calvet LE, Sodini P (2014) Twin picks: disentangling the determinants of risk-taking in household
portfolios. J Financ LXIX(2):867–906
Chang C-C, DeVaney SA, Chiremba ST (2004) Determinants of subjective and objective risk
tolerance. J Pers Financ 3:53–67
Diversified Risky Financial Assets in Portfolios of Risk-Averse. . . 239

Coleman S (2003) Risk tolerance and the investment behavior of Black and Hispanic heads of
household. J Financ Couns Plan 14:43–52
Deaves RE, Veit T, Bhandari G, Cheney J (2007) The savings and investment decisions of planners:
a cross sectional study of college employees. Financ Services Rev 16:117–133
EC (2012) Special Eurobarometer 373. Retail financial services. https://ec.europa.eu/
commfrontoffice/publicopinion/archives/ebs/ebs_373_sum_en.pdf. Accessed 26 May 2020
ECB (2016) The household finance and consumption survey. Results from the second wave. ECB
Stat Pap 18:1–138
Gibson R, Michayluk D, Van de Venter G (2013) Financial risk tolerance: an analysis of
unexplored factors. Financ Services Rev 22(1):23–50
Gilliam JE, Goetz JW, Hampton VL (2008) Spousal differences in financial risk tolerance. J Financ
Counsel Plan 19:3–11
Grable JE (2000) Financial risk tolerance and additional factors that affect risk taking in everyday
money matters. J Bus Psychol 14(4):625–630
Grable JE (2016) Financial risk tolerance. In: Xiao JJ (ed) Handbook of consumer finance research.
Springer International, Cham, pp 19–31
Grable JE, Joo S-H (2004) Environmental and biopsychosocial factors associated with financial risk
tolerance. J Financ Couns Plan 15:73–80
Grable JE, Lytton RH (2001) Assessing the concurrent validity of the SCF risk assessment item. J
Financ Couns Plan 12:43–52
Haliassos M, Bertaut CC (1995) Why do so few hold stocks? Econ J 105:1110–1129
Hallahan TA, Faff RW, McKenzie MD (2004) An empirical investigation of personal financial risk
tolerance. Financ Services Rev 13:57–78
Hanna SD, Gutter MS, Fan JX (2001) A measure of risk tolerance based on economic theory. J
Financ Couns Plan 12(2):53–60
Hanna SD, Waller W, Finke M (2008) The concept of risk tolerance in personal financial planning.
J Pers Financ 7:96–108
Hilbe J (2014) Modeling count data. Cambridge University Press, Cambridge
Irwin CE (1993) Adolescence and risk taking: how are they related? In: Bell NJ, Bell RW (eds)
Adolescent risk taking. SAGE, Newbury Park, CA, pp 7–28
Jianakoplos NA, Bernasek A (1998) Are women more risk-averse. Econ Inq 36:620–630
Kimbal MS, Sahm CR, Shapiro MD (2007) Imputing risk tolerance from survey responses. https://
www.nber.org/papers/w13337. Accessed 20 Dec 2019
Kochaniak K, Ulman P (2020) Risk-intolerant but risk-taking—towards a better understanding of
inconsistent survey responses of the euro area households. Sustainability 12(17) 6912:1–26
Marinelli N, Mazzoli C, Palmucci F (2017) Mind the gap: inconsistencies between subjective and
objective financial risk tolerance. J Behav Financ 05:1–12
Martin TK (2011) Financial planners and the risk tolerance gap. https://papers.ssrn.com/sol3/
papers.cfm?abstract_id¼2358736. Accessed 02 May 2020
Moreschi P (2005) An analysis of the ability of individuals to predict their own risk tolerance. J Bus
Econ Res 3:39–48
Nosic A, Weber M (2007) Determinants of risk taking behavior: the role of risk attitudes, risk
perceptions and beliefs. https://madoc.bib.uni-mannheim.de/1774/1/001_SSRN_ID1027453_
code846681.pdf. Accessed 02 May 2019
Roszkowski MJ, Grable J (2005) Estimating risk tolerance: the degree of accuracy and the
paramorphic representation of the estimate. J Financ Couns Plan 16:29–47
Sahm CR (2007) How much does risk tolerance change? https://www.federalreserve.gov/pubs/
feds/2007/200766/200766pap.pdf. Accessed 05 May 2020
Schooley DK, Worden DD (1996) Risk aversion measures: comparing attitudes and
asset allocation. Financ Services Rev 5(2):87–99
Slovic P (2004) The perception of risk. Earthscan, London
Stewart W, Roth P (2001) Risk propensity differences between entrepreneurs and managers: a
meta-analytic review. J Appl Psychol 86:145–153
240 K. Kochaniak and P. Ulman

Vissing-Jorgensen A (2002) Towards an explanation of household portfolio choice heterogeneity:

nonfinancial income and participation cost structures. https://www.nber.org/papers/w8884.pdf.
Accessed 15 Jan 2020
Warneryd K-E (1996) Risk attitudes and risky behavior. J Econ Psychol 17:749–770
Weber EU, Blais A-R, Betz N (2002) A domain-specific risk-attitude scale: measuring risk
perceptions and risk behaviors. J Behav Decis Mak 15:263–290
Weller JA, Levin IP, Bechara A (2010) Do individual differences in Iowa Gambling Task
performance predict adaptive decision making for risky gains and losses? J Clin Exp
Neuropsychol 32:141–150
Winkelmann R (2008) Econometric analysis of count data. Springer, New York
Yao R, Curl A (2011) Do market returns influence risk tolerance? Evidence from panel data. J
Family Econ Iss 32:532–544
Yao R, Hanna SD (2005) The effect of gender and marital status on financial risk tolerance. J Pers
Financ 4:66–85
Yao R, Gutter MS, Hanna SD (2005) The financial risk tolerance of Blacks, Hispanics and Whites. J
Financ Counsel Plan 16:51–62
Financial Behavior: Preliminary Survey
Results

Kutlu Ergün

1 Introduction

Financial literacy includes three essential components: financial knowledge, finan-

cial attitude and financial behavior. Perhaps financial behavior is the most important
factor which formulates the financial literacy of an individual. According to Xiao
(2008), to develop a behavior change focused educational program, researchers of
consumer finance need to better understand how behaviors are formed and why and
how to help consumers change undesirable financial behaviors and develop positive
financial behaviors. On the other hand, psychological economists highlight the
importance of considering psychological antecedents to explain financial behavior
(Ning and Baker 2016). Apart from keeping abreast of latest technological innova-
tions, enhancing overall financial literacy is essential factor for individuals’ well-
being. Financial literacy and technological literacy are important resources that
low-income people need to exit poverty (Servon and Kastner 2008). Therefore, in
addition to having good financial knowledge, especially level of positive financial
behaviors should be increased in order to increase the level of financial literacy.
Financial literacy includes a number of behaviors that can enhance financial well-
being. One of these important behaviors is saving. Active savers exhibit a behavior
supporting their budgeting behavior. People who build savings are also likely to be
more resilient to financial shocks and better able to meet financial goals. Also
budgeting is a valuable tool for money management and a component of financial
literacy (OECD 2016).
According to Wagner and Walstad (2019), financial behaviors are defined as
short-term if they involve a money or credit management task. The four short-term

K. Ergün (*)
Balikesir University, Balikesir, Turkey
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 241
K. Jajuga et al. (eds.), Contemporary Trends and Challenges in Finance, Springer
Proceedings in Business and Economics,
https://doi.org/10.1007/978-3-030-73667-5_15
242 K. Ergün

financial behaviors cover expenses and paying all bills each month, managing
checking account, paying off credit card balances in full each month; and, making
monthly mortgage payments on time. However, if we consider long-term financial
behaviors, we can include more behaviors which are impact of positive financial
behaviors of individuals. This study included more short-term and long-term finan-
cial behaviors: spending according to personal budget, saving regularly, saving
emergency and long-term, and paying bills on time, having monthly budget plan
and comparison shop. Showing positive behavior on these components positively
affects the financial literacy. However, overconfidence and optimism are especially
important factors negatively effecting favorable financial behaviors, and it has
negative impact on financial literacy. Behavioral studies showed that unrealistic
optimism was associated with less prudent financial behaviors, such as short plan-
ning horizons and saving less (Collins and Gjertson 2013). Therefore, financial
education, especially for young individuals, is required to enhance positive financial
behavior and overall financial literacy. Financial education is also essential for the
average family trying to decide how to balance its budget, buy a home, fund the
children’s education and ensure an income when the parents retire (OECD 2006).
Well-designed and well-executed financial education initiatives can have an effect
(Hastings et al. 2013).
This paper shows the preliminary results of a study carried out among Italian and
Turkish university students. After the introduction which is the first section, the
methodology is the second section which explains methods of the study. Third
section is the section of results which summarize the survey results. Conclusions
summarize the main findings of the research.

2 Literature Review

Consumers’ behavior is important factor in shaping their financial situation and well-
being as well as having impact on the level of financial literacy. On the other hand,
some types of behavior such as failing to actively save money, putting off bills
payment, failing to plan future expenditures or choosing financial products without
shopping around, may impact negatively on an individual’s financial situation and
well-being (OECD 2020). Desirable financial behaviors and overall financial literacy
depend on socio-demographic and socio-economic characteristics. Xiao et al. (2015)
suggest that older consumers have higher level financial capability than younger
consumers. The finding show that age is an important determinant for financial
capability. Rostamkalaei and Riding (2020) examined the financial knowledge and
financial practices among immigrants and Canadian-born individuals. They showed
that immigrants had lower financial knowledge and financial behavior score (having
retirement account and saving for long-term) than Canadian-born individuals. They
also demonstrated that individuals who are not native English and French had lower
level of financial knowledge. Individuals who speak an official language were
relatively more likely to have retirement plan and check their credit reports.
Financial Behavior: Preliminary Survey Results 243

Mahendru et al. (2020) explored the antecedents of financial well-being, such as

financial knowledge, objective and subjective financial situation and personality
traits. The results of their study revealed that the attainment of financial well-being
is characterized by the fulfillment of present and future commitments, feeling of
financial security; freedom of choice; and improved quality of life.
Financial attitudes, financial literacy and behavioral control play an essential role
in explaining responsible financial consumption behavior (Barbić et al. 2019).
Financial knowledge is known to have a significant effect on financial decisions
such as saving for retirement. Therefore financial knowledge and saving behavior
could contribute to the household saving behavior (Kim and Yuh 2018). Since
financial literacy includes financial knowledge, financial behavior and financial
attitude, in particular increasing financial knowledge can positively affect financial
behavior, leading to improving financial literacy. Woodyard, Robb, Babiarz and
Jung (2017) found that having a higher level of financial knowledge is more likely to
be associated with positive financial behavior. Both objective knowledge and sub-
jective knowledge displayed consistent associations with the reported behaviors.
Financial education can contribute to a positive change in financial behavior and
overall financial literacy. Financial education has the potential to help people make
more informed financial decisions and change financial behavior. Also it has stron-
ger and more positive impact on the long-term financial goals (Wagner and Walstad
2019). If individuals have a positive perception to achieve their long-term financial
goals, they display positive financial behaviors. This increases saving tendency to
both in the short-term and in the long-term.

3 Methodology

For the purpose of this study, It was used a financial behavior survey which was open
to all voluntary participants on social media. The survey created by the author
comprised of socio-demographic variables including gender, monthly income,
working experience and pre-knowledge on financial finance. To reach the financial
behavior score, total 10 questions were included in the survey. Important elements of
financial behaviors such as saving regularly and for emergency, budgeting, compar-
ing products when shopping and paying bills on time were included to measure the
financial behavior score of university students participated in this study.
To collect data, the link of the survey was shared on social media in Italy and
Turkey. It was reached totally 455 responses. Only 394 complete survey question-
naires from Italy and Turkey were included in the study. The study consisted of
170 Italian and 224 Turkish students. Financial behavior score calculated from
close-ended statements related to spending, saving, budgeting, shopping and con-
trolling expenditure. Maximum possible score was 10. The Independent Samples
t-test was used to analyze whether there were any statistically significant differences
between each groups’ independent variables. Following statements were included in
the survey questionnaire: “I spend according to my personal budget”, “I have
244 K. Ergün

enough saving”, I pay my bills on time”, “I have a monthly budget plan”, “I save for
emergency situations”, “I have written budget plan”, “I have enough money for
each month”, I compare product when shopping”, “I save regularly for long-term”,
and “I have enough money for unpredictability”.

4 Results

Table 1 displays sample characteristics of students from Italy and Turkey including
170 participants from Italy (43.1%) and 224 participants from Turkey (56.9%). The
majority of the students were female (Italy, 57.6%; Turkey, 54.5%). For both
countries, more than 50% of students’ monthly income was under 300 Euros. Italian
students had more working experience (58.8%) than Turkish students (36.2%). The
percentage of Italian students was higher than Turkish students in terms of having
pre-knowledge on personal finance (51.8%, 33.5% respectively).
Table 2 shows positive answers to the survey questions. Out of the maximum
behavior score of 10, on average students from Italy and Turkey scored about 5.43
across the whole sample. The highest behavior score were achieved by Turkish
students (5.72), Italy (5.04). For both countries, the vast majority of students
indicated that they paid their bills on time (Italy, 77.6%; Turkey, 87.5%). Both
Italian and Turkish students had a low percentage of having written budget plan
comparing to positive answers to other survey questions measuring financial behav-
iors. While 79.5% of Turkish students compare shops while purchasing product,
63.5% of Italian students make comparison between shops.

Table 1 Sample characteristics

Italy Turkey
Variable Frequency Percentage Frequency Percentage
Gender
Male 72 42.4 102 45.0
Female 98 57.6 122 54.5
Monthly income
Lower than €300 90 52.9 140 62.5
Higher than €300 80 47.1 84 37.5
Working experience
Yes 100 58.8 81 36.2
No 70 41.2 143 63.8
Pre-knowledge on personal finance
Yes 88 51.8 75 33.5
No 82 48.2 149 66.5
Nationality
Total 170 43.1 224 56.9
Financial Behavior: Preliminary Survey Results 245

Table 2 Descriptive statics for survey questions

Italy Turkey
Question Frequency % Frequency %
Q1 Responsible spending 112 65.9 145 64.7
Q2 Enough saving 92 54.1 142 63.4
Q3 Paying bills on time 132 77.6 196 87.5
Q4 Monthly budget plan 74 43.5 142 63.4
Q5 Saving for emergency 104 61.2 127 56.7
Q6 Written budget plan 56 32.9 81 36.2
Q7 Enough money for each month 96 56.5 171 76.3
Q8 Comparison shop 108 63.5 178 79.5
Q9 Saving regularly 90 52.9 82 36.6
Q10 Having money for unpredictability 98 57.6 145 67.7
Countries total 170 50.4 224 57.2

196
178
171

145 145
142 142
132 127
112 108 Italy
104 98
92 96 90 Turkey
81 82
74
56

Q1 Q2 Q3 Q4 Q5 Q6 Q7 Q8 Q9 Q10

Fig. 1 Positive responds to survey questions

Table 3 Independent samples t-test comparing nationalities’ financial behavior scores

Group N M SD t p
Italy 170 5.04 2.28 3.04 0.002
Turkey 224 5.72 2.10

Figure 1 shows the link between nationalities and positive responds to survey
questions. Response from Turkish students to Q3 (paying bills on time) is signifi-
cantly more positive than others. Q8 (comparison shop) is also more clear positive
response for Turkish students. Positive response to Q6 (written budget plan) ques-
tion is the lowest for both university students. Although all positive responses of
Turkish students are higher than Italian students, Italian students have more positive
answers regarding saving regularly.
Table 3 represents the t test result in terms of financial behavior scores of Italian
and Turkish participants. The independent variable, financial behavior scores ranged
from 1 to 10 with a mean of 5.72 and a standard deviation of 2.10 for Turkish
participants. Italian participants’ consumer behavior score ranged from 1 to 10 with a
246 K. Ergün

Table 4 Independent samples t-test comparing financial behavior scores for Italian students
Groups N M SD t p
Gender
Male 72 5. 27 2. 75 1. 12 0.261
Female 98 4. 87 1. 86
Monthly income
Lower than €300 90 4. 35 2. 20 4. 40 0.000
Higher than €300 80 5. 82 2. 13
Working experience
Yes 100 5. 58 2. 54 3. 77 0.000
No 70 4. 28 1. 77
Pre-knowledge on personal finance
Yes 88 5. 65 2. 15 3. 75 0.000
No 82 4. 39 2. 25

Table 5 Independent sam- Groups N M SD t p

ples t-test comparing financial
Gender
behavior scores for Turkish
students Male 102 5.60 2.14 0.75 0.454
Female 122 5.81 2.07
Monthly income
Lower than €300 140 5.49 2.02 2.13 0.034
Higher than €300 84 6.10 2.18
Working experience
Yes 81 6.14 2.09 2.29 0.023
No 143 5.48 2.07
Pre-knowledge on personal finance
Yes 75 6.05 1.99 1.67 0.096
No 149 5.55 2.14

mean of 5.04 and a standard deviation of 2.28. To make a comparison between the
differences of means among nationality group based on overall demographic con-
ditions and consumer behaviors, Independent Samples t-test was performed. It was
indicated that there was a significant difference between Italian and Turkish partic-
ipants (t ¼ 3.04, p < 0.05). Turkish participants had higher financial behavior score
than Italian participants.
Table 4 shows the t-test results of Italian students. There was no significant
relationship between male and female students (t ¼ 1.12, p > 0.05). There was a
statistically significant relationship in terms of monthly income, working experience
and pre-knowledge on personal finance (t ¼ 4.40, p < 0.01; t ¼ 3.77, p < 0.01;
t ¼ 3.75, p < 0.01 respectively).
Table 5 shows the t-test results of Italian students. There was no significant
relationship between male and female students (t ¼ 0.75, p > 0.05). There was a
statistically significant relationship in terms of monthly income, working experience
Financial Behavior: Preliminary Survey Results 247

Table 6 Independent samples t-test comparing nationalities scores each variable

Groups N M SD t p
Gender
Male Italy 72 5.27 2.75 0.88 0.376
Turkey 102 5.60 2.14
Female Italy 98 4.87 1.86 3.50 0.001
Turkey 122 5.81 2.07
Monthly income
Lower than €300 Italy 90 4.34 2.20 4.01 0.000
Turkey 140 5.49 2.02
Higher than €300 Italy 80 5.82 2.13 0.83 0.404
Turkey 84 6.10 2.18
Working experience
Yes Italy 100 5.58 2.54 1.65 0.100
Turkey 81 6.14 2.09
No Italy 70 4.28 1.77 4.13 0.000
Turkey 143 5.48 2.07
Pre-knowledge on personal finance
Yes Italy 88 5.65 2.15 1.20 0.231
Turkey 75 6.05 1.99
No Italy 82 4.39 2.25 3.88 0.000
Turkey 149 5.55 2.14

and pre-knowledge on personal finance (t ¼ 2.13, p < 0.05; t ¼ 2.99, p < 0.05;
t ¼ 1.67, p < 0.10 respectively).
Table 6 represents the comparison of the t-test results among Italian and Turkish
students. There was no significant relationship between male students among Italian
and Turkish participants (t ¼ 0.88, p > 0.05). However, there was a statistically
significant relationship between female students (t ¼ 3.50, p < 0.05). Female
students from Turkey had higher financial behavior score than female from Italy
(Turkey: M ¼ 5.81, SD ¼ 2.07; Italy: M ¼ 4.87, SD ¼ 1.86). There was no
statistically significant relationship between Turkish and Italian students in terms
of higher income category in the group of monthly income (t ¼ 0.83, p > 0.05). On
the other hand, there was a statistically significant relationship between Italian and
Turkish students among students with income level lower than €300 (t ¼ 4.01,
p < 0.05). Students with low income level from Turkey had higher financial
behavior score than Italian students (Turkey: M ¼ 5.49, SD ¼ 2.02; Italy:
M ¼ 4.34, SD ¼ 2.20). There was also statistically significant relationship between
Italian and Turkish students in terms of the variables including “working experience
and pre-knowledge on personal finance”. Turkish students who had no working
experience and no pre-knowledge on personal finance had higher financial behavior
score than Italian students who had no working experience and no pre-knowledge on
personal finance [no working experience: (Turkey: M ¼ 5.48, SD ¼ 2.07; Italy
248 K. Ergün

Male

No pre-
Female
knowledge 149*
102
122*

140* Lower than Italy

Pre-knowledge 75
300
84 Turkey

143*
No working Higher than
experince 81
300

Working
experience

Fig. 2 Significance levels of natalities’ financial behavior scores

M ¼ 4.28, SD ¼ 1.77); no pre-knowledge on personal finance: (Turkey: M ¼ 5.55,

SD ¼ 2.14, Italy: M ¼ 4.39, SD ¼ 2.25)].
Figure 2 represents the significance levels of nationalities’ financial behavior
scores. Almost all independent variables’ scores for Turkish participants are higher
than the scores for Italian participants. The scores with asterisks show the significant
variables including “being female”, “having no pre-knowledge on personal finance”,
having no working experience”, “having the income level lower than €300”. Italian
participants have higher scores only for the independent variables of “pre-knowledge
on personal finance” and “having working experience”. Both variables are not
statistically significant.
Increasing positive financial behavior does not produce the same result in both
countries. Lack of knowledge and working experience, and lower income may result
in more negative consequences for Italian individuals. On the other hand, it can be
concluded that the increase in income, working experience and financial knowledge
has more positive impact on Italian individuals. This may be due to differences in the
providing of financial knowledge in both countries. A study conducted by Henager
and Mauldin (2015) showed that perceived financial knowledge was a strong
indicator of saving behavior. They suggested that saving behavior may increase
confidence or confidence increases saving behavior even though there was no
causality. Planning also found to be significantly related to the decision to save
regularly. On the other hand, overall financial literacy can also have positive effect
on saving behavior because financial literacy because increased financial literacy
means an increase in positive financial behavior and attitude. Murendo and
Mutsonziwa (2017) analyzed the determinants of financial literacy and its effect
on individual’s savings decisions in Zimbabwe. They showed that financial literacy
has a positive impact on saving behavior of individuals living in both rural and urban
areas. Key element of increasing long-term saving is improving positive financial
behavior, leading to enhanced financial literacy.
Financial Behavior: Preliminary Survey Results 249

5 Discussion

Turkish students achieved higher financial behavior score (5. 72 out of a possible 10)
than Italian students (5.04 out of a possible 10). According to OECD/INFE 2020
International Survey of Adult Financial Literacy (OECD 2020), Italy had the lowest
financial behavior score (4.2 out of a possible 9) across the sampled 26 OECD
countries. On the other hand, according to OECD/INFE International Survey of
Adult Financial Literacy Competencies (OECD 2016), Turkey achieved 4.8 finan-
cial behavior score (out of a possible 9), which is higher than the score of Italian
participants’ score in 2020 OECD survey. Thus the result of this study is similar to
OECD/INFE 2016 and 2017 survey results. The striking result of all the studies on
financial literacy (financial knowledge, financial behavior and financial attitudes) is
that the level of financial behavior is almost always lower than the level of financial
knowledge. It takes a long time for financial knowledge to turn into positive financial
behavior.
Many studies have shown that there is a gender gap in financial literacy. The
financial literacy level of male is higher than female. Gender gap focuses on
traditional socioeconomic factors and cultural differences to explain the differences
in financial literacy between men and women. However, such factors can only
partially explain the gender gap (Arellano et al. 2018). Women are marginally
more likely to give incorrect responses than men, but they are even more likely to
give don’t know (DK) responses relative to men (Chen and Garand 2018). This study
found that there was no association between Italian and Turkish male students in
terms of positive financial behavior. However, Turkish female students had higher
level of financial behavior than Italian female students. On the other hand, there was
no statistically significant association between Turkish and Italian students in terms
of higher income category. Higher household income is consistently associated with
higher scores on financial knowledge and financial capability including attitudes and
behavior (Robson and Peetz 2020). In this study, there was a significant association
among students with lower-income category. Low-income Turkish students had
higher financial behavior than Italian students in low-income. Although studies
have found that high income level is a positive determinant of financial literacy
and financial behavior, there are also differences in terms of the effect of low income.
Personal finance skills are essential factors affecting individuals’ positive finan-
cial behaviors. Positive financial behavior increases as individual financial capabil-
ities increase. The study concluded that there was also significant association in
terms of having working experience and pre-knowledge on personal finance. Turk-
ish students with no working experience and no pre-knowledge on personal finance
had higher level of financial behavior than Italian students who had no working
experience and no pre-knowledge on personal finance.
250 K. Ergün

6 Conclusions

The most important effect of positive financial behavior is on long-term saving

behavior. Positive financial behaviors make spending behaviors more responsible.
The purpose of this study was to examine the relationship between financial behavior
and socio-demographic variables among Italian and Turkish students. It was con-
cluded that Turkish students had higher financial behavior score than Italian stu-
dents. Higher income, having working experience and pre-knowledge on personal
finance was found as determinants to have higher financial behavior scores for both
Italian and Turkish students. When Italian and Turkish students were compared,
different results were reached. Turkish students who were female, having low
income, no working experience and no pre-knowledge on personal finance had
had higher level of financial behavior scores. Increasing income, having working
experience and increasing knowledge on financial finance do not lead to positive
changes in the financial behavior of Turkish participants. Gender, high income,
having working experience and pre-knowledge on personal finance did not play a
determinative role in comparing Italian and Turkish students. The result of this
survey may provide valuable information to improve positive financial behaviors
of university students. Providing financial education can increase the level of
favorable financial behavior. If future research is included more variables and
participants it may be reached some different and fruitful results.

7 Limitations

The study included 394 students from Italy and Turkey. The results cannot be
generalized since there were small amount of participant in the survey. If the number
of the participant of this survey were, different results could be reached. It would be
better to take this into account in future studies. Adding some other socio-
demographic and socio-economic variables to the survey can produce useful results.
Although the study has some limitations, it offers useful results in terms of evalu-
ating financial behaviors.

References

Arellano A, Cámara N, Tuesta D (2018) Explaining the gender gap in financial literacy: the role of
non-cognitive skills. Econ Notes 47:495–518. https://doi.org/10.1111/ecno.12113
Barbić D, Lučić A, Chen JM (2019) Measuring responsible financial consumption behaviour. Int J
Consum Stud 43(1):102–112. https://doi.org/10.1111/ijcs.12489
Chen Z, Garand JC (2018) On the gender gap in financial knowledge: decomposing the effects of
don’t know and incorrect responses. Soc Sci Q 99:1551–1571. https://doi.org/10.1111/ssqu.
12520
Financial Behavior: Preliminary Survey Results 251

Collins JM, Gjertson L (2013) Emergency savings for low-income consumers. Focus 30(1):12–17.
Retrieved from http://www.irp.wisc.edu/publications/focus/pdfs/foc301c.pdf
Hastings JS, Madrian BC, Skimmyhorn WL (2013) Financial literacy, financial education, and
economic outcomes. Annu Rev Econ 5(1):347–373
Henager R, Mauldin T (2015) Financial literacy: the relationship to saving behavior in low- to
moderate-income households. Family Consum Sci Res J 44:73–87. https://doi.org/10.1111/fcsr.
12120
Kim KT, Yuh Y (2018). Financial knowledge and household saving: evidence from the survey of
consumer finances. Family Consum Sci Res J 47(1):5–24. https://doi.org/10.1111/fcsr.12270
Mahendru M, Sharma GD, Hawkins M (2020) Toward a new conceptualization of financial well-
being. J Public Affairs e2505. https://doi.org/10.1002/pa.2505
Murendo C, Mutsonziwa K (2017) Financial literacy and savings decisions by adult financial
consumers in Zimbabwe. Int J Consum Stud 41:95–103. https://doi.org/10.1111/ijcs.12318
Ning T, Baker A (2016) Self-esteem, financial knowledge and financial behavior. J Econ Psychol
54:164–176. https://doi.org/10.1016/j.joep.2016.04.005
OECD (2006) The importance of financial education. Policy brief. Retrieved from http://www.
oecd.org/finance/financial-education/37087833.pdf
OECD (2016) OECD/INFE international survey of adult financial literacy competencies. OECD,
Paris
OECD (2020) OECD/INFE 2020 international survey of adult financial literacy. www.oecd.org/
financial/education/launchoftheoecdinfeglobalfinancialliteracysurveyreport.htm
Robson J, Peetz J (2020) Gender differences in financial knowledge, attitudes, and behaviors:
accounting for socioeconomic disparities and psychological traits. J Consum Aff 54
(3):813–835. https://doi.org/10.1111/joca.12304
Rostamkalaei A, Riding A (2020) Immigrants, financial knowledge, and financial behavior. J
Consum Aff 54:951–977. https://doi.org/10.1111/joca.12311
Servon LJ, Kastner R (2008) Consumer financial literacy and the impact of online banking on the
financial behavior of lower-income bank customers. J Consum Aff 42:271–305. https://doi.org/
10.1111/j.1745-6606.2008.00108.x
Wagner J, Walstad WB (2019) The effects of financial education on short-term and long-term
financial behaviors. J Consum Aff 53:234–259. https://doi.org/10.1111/joca.12210
Woodyard AS, Robb C, Babiarz P, Jung J-Y (2017) Knowledge and practice: implications for cash
and credit management behaviors. Family Consum Sci Res J 45(3):300–314. https://doi.org/10.
1111/fcsr.12202
Xiao JJ (2008) Appliying behavior theories to financial behavior. In: Xiao JJ (eds) Handbook of
consumer finance research. Springer, New York. https://doi.org/10.1007/978-0-387-75734-6_5
Xiao JJ, Chen C, Sun L (2015) Age differences in consumer financial capability. Int J Consum Stud
39:387–395. https://doi.org/10.1111/ijcs.12205