University of Groningen
Essays on foreign ownership in transition bankingPoghosyan, Tigran
IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite fromit. Please check the document version below.
Document VersionPublisher's PDF, also known as Version of record
Publication date:2009
Link to publication in University of Groningen/UMCG research database
Citation for published version (APA):Poghosyan, T. (2009). Essays on foreign ownership in transition banking Enschede: PrintPartners IpskampB.V., Enschede, The Netherlands
CopyrightOther than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of theauthor(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).
Take-down policyIf you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediatelyand investigate your claim.
Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons thenumber of authors shown on this cover page is limited to 10 maximum.
Download date: 20-04-2018
Essays on Foreign Ownership in Transition Banking
Tigran Poghosyan
Publisher: PPI PublishersPostbus 3337500 AH EnschedeThe Netherlands
Printed by: PrintPartners Ipskamp
ISBN: 978-90-367-3885-9
c© 2009 Tigran Poghosyan
All rights reserved. No part of this publication may be reproduced, stored in a re-trieval system of any nature, or transmitted in any form or by any means, electronic,mechanical, now known or hereafter invented, including photocopying or recording,without prior written permission of the publisher.
Essays on Foreign Ownership in Transition Banking
Proefschrift
ter verkrijging van het doctoraat in deEconomie en Bedrijfskunde
aan de Rijksuniversiteit Groningenop gezag van de
Rector Magnificus, dr. F. Zwarts,in het openbaar te verdedigen op
donderdag 10 september 2009om 16.15 uur
door
Tigran Poghosyan
geboren op 23 oktober 1978te Yerevan, Armenië
Promotores: Prof. dr. J. de HaanProf. dr. E. Sterken
Copromotor: Dr. M. Koetter
Beoordelingscommissie: Prof. dr. B.W. (Robert) LensinkProf. dr. P. MolyneuxProf. dr. S. Ongena
Acknowledgements
This thesis has been completed with the help and cooperation of many colleagues
and individuals at the University of Groningen. During my stay in Groningen, I
have benefitted greatly from the stimulating research environment created by the
Faculty of Economics and Business and the SOM Research School.
I am particularly grateful to my promoters Prof. Jakob de Haan and Prof. Elmer
Sterken, and my co-promoter Dr. Michael Koetter for the intellectual guidance
and advice. Their liberal style of supervision and encouragement to explore my
own research topics were essential contributors to the quality of my work. I would
also like to express my gratitude to the committee members of my thesis, Prof.
Robert Lensink, Prof. Philip Molyneux, and Prof. Steven Ongena for reading the
manuscript and for their valuable comments.
I am also indebted to Tammo Bijmolt, Jan Jacobs, Gerard Kuper, Laura Spierdijk,
and other faculty members at the University of Groningen for many valuable dis-
cussions and advice during my study period. From the outside of faculty, I would
like to mention, without implication, Martin Čihák, Thomas Kick, Evzen Kocenda,
and Subal Kumbhakar for their cooperation and co-authorship of different research
papers.
This is also the place to express my gratitude to the SOM bureau members, Astrid
Beerta, Rina Koning, and Ellen Nienhuis as well as our esteemed Ph.D. coordinator
Martin Land, for their constant support and kind assistance to resolve any type of
ii
administrative issue, be it remuneration of travel expenses or problems with finding
accommodation in Groningen. Their kind cooperation has been very helpful and
has saved me a great deal of time, which I was able to devote to my research.
Perhaps I would not even have started writing this thesis if I had not met Umed
Temurshoev, my classmate from CERGE-EI and a good friend, who recommended
me to apply for a Ph.D. program in Groningen. I am very thankful to Umed for
being around during all the challenging and exciting times of our studies both in
Groningen and in Prague.
This list of other colleagues and friends can go on continuously, so I will just limit
myself by mentioning, without implication, Matilda Dorotic, Tomek Katzur, Aljar
Meesters, Ernst Osinga, Froukje Schaaf, Stanislav Stakhovich, and other members
of our dynamic Ph.D. student community. I am also thankful to Richard Jong-
A-Pin for being always ready to provide suggestions and advice when needed, and
Kees Bouwman for his help on LATEX formatting. I would also like to extend my
gratitude to all the other friends and colleagues, whose names were omitted here due
to space constraints, but who will always occupy an honorable place in my memory.
Finally, I would like to thank my family members: my father, Vladimir, my
mother, Marine, and my brother, Arsen, for their patience and moral support during
the time I was far away from home. The most important thing for me is to realize
that I was able to achieve something that my parents can be proud of.
Tigran Poghosyan
Contents
1 Introduction 1
1.1 Background and Motivation . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Outline of the Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2 Foreign Ownership and Bank Efficiency: Does Sample Selection
Matter? 9
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2 Methodology and Data . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.1 Stochastic efficiency frontier model . . . . . . . . . . . . . . . 13
2.2.2 Instrumenting foreign ownership . . . . . . . . . . . . . . . . 15
2.2.3 Data and descriptive statistics . . . . . . . . . . . . . . . . . 17
2.3 Estimation Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.3.1 Cost frontier specification . . . . . . . . . . . . . . . . . . . . 20
2.3.2 Inefficiency analysis . . . . . . . . . . . . . . . . . . . . . . . 22
2.3.3 Impact of ownership . . . . . . . . . . . . . . . . . . . . . . . 22
2.3.4 Inefficiency scores . . . . . . . . . . . . . . . . . . . . . . . . 24
2.3.5 Robustness checks . . . . . . . . . . . . . . . . . . . . . . . . 25
2.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
iv Contents
3 Determinants of Cross-Border Bank Acquisitions: The Role of In-
stitutions 37
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.2 Theoretical Background . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.3 Methodology and Data . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.3.1 Multilevel mixed-effect logistic regression . . . . . . . . . . . 42
3.3.2 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.4 Empirical Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.4.1 Do institutions matter? . . . . . . . . . . . . . . . . . . . . . 46
3.4.2 Sensitivity analysis . . . . . . . . . . . . . . . . . . . . . . . . 48
3.4.3 Analyzing the efficiency and market power hypotheses across
countries and over time . . . . . . . . . . . . . . . . . . . . . 49
3.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4 Heterogeneity of Technological Regimes and Bank Efficiency 59
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.2 Accounting for Heterogeneity of Banking Technologies: A Latent
Class Stochastic Frontier Model . . . . . . . . . . . . . . . . . . . . . 62
4.3 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.4 Estimation Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
4.4.1 Selection of the number of classes . . . . . . . . . . . . . . . . 69
4.4.2 Parameter estimates and analysis of class-specific efficiency
scores . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
4.4.3 Economic interpretation of heterogeneous technologies . . . . 71
4.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
5 Foreign Bank Entry, Bank Efficiency, and Market Power 79
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
Contents v
5.2 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5.2.1 Theoretical background . . . . . . . . . . . . . . . . . . . . . 83
5.2.2 Empirical methodology . . . . . . . . . . . . . . . . . . . . . 87
5.3 Data Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
5.4 Estimation Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
5.4.1 Foreign bank entry and cost efficiency . . . . . . . . . . . . . 93
5.4.2 Foreign bank entry and market power . . . . . . . . . . . . . 96
5.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
6 Re-examining the Impact of Foreign Bank Participation on Interest
Margins 109
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
6.2 Methodology and Data . . . . . . . . . . . . . . . . . . . . . . . . . . 114
6.2.1 Empirical model . . . . . . . . . . . . . . . . . . . . . . . . . 114
6.2.2 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
6.3 Estimation Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
6.3.1 The reference model . . . . . . . . . . . . . . . . . . . . . . . 119
6.3.2 The impact of foreign bank participation . . . . . . . . . . . 119
6.3.3 Economic significance . . . . . . . . . . . . . . . . . . . . . . 121
6.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
7 Concluding Remarks 127
7.1 Main Findings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
7.2 Policy Implications . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
Index 133
Samenvatting (Summary in Dutch) 143
List of Figures
2.1 Average inefficiency scores for individual countries. . . . . . . . . . . 30
3.1 Model 1: Average values of inefficiency (β1jt) and market power (β2jt)
coefficients across countries and over time . . . . . . . . . . . . . . . 57
3.2 Model 2: Average values of inefficiency (β1jt) and market power (β2jt)
coefficients across countries and over time . . . . . . . . . . . . . . . 58
5.1 Share of foreign-owned banks in terms of total assets (%), 1995 and
2006 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
List of Tables
2.1 Summary of results from panel data studies on bank efficiency in FSEs 31
2.2 Descriptive statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.3 Panel estimation of stochastic efficiency frontier models . . . . . . . 33
2.4 Panel estimation of stochastic efficiency frontier models, cont. . . . . 34
2.5 First-stage regression results . . . . . . . . . . . . . . . . . . . . . . . 34
2.6 Tests of instrument validity . . . . . . . . . . . . . . . . . . . . . . . 35
3.1 Cross-border bank acquisitions in FSEs, 1992-2006 . . . . . . . . . . 53
3.2 Data sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.3 Data description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.4 Descriptive statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.5 Estimates of equations (1) and (2) . . . . . . . . . . . . . . . . . . . 55
3.6 Sensitivity analysis: Time dummies and macro variables added . . . 56
4.1 Overview of the literature . . . . . . . . . . . . . . . . . . . . . . . . 74
4.2 Descriptive statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
4.3 Selection of the number of classes . . . . . . . . . . . . . . . . . . . . 76
4.4 Average efficiency scores for LCM with different number of classes . 76
4.5 LCM estimation results . . . . . . . . . . . . . . . . . . . . . . . . . 77
4.6 Comparison of efficiency scores . . . . . . . . . . . . . . . . . . . . . 78
4.7 Assigning class membership . . . . . . . . . . . . . . . . . . . . . . . 78
x List of Tables
5.1 Number of observations for domestic and foreign (acquired and green-
field) banks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
5.2 Variable definitions and sources . . . . . . . . . . . . . . . . . . . . . 104
5.3 Descriptive statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
5.4 Impact of foreign bank participation on cost efficiency: Stochastic
efficiency frontier analysis (model (5.9)) . . . . . . . . . . . . . . . . 106
5.5 Impact of foreign bank participation on market power (model (5.8)) 107
6.1 Variable definition and sources . . . . . . . . . . . . . . . . . . . . . 123
6.2 Descriptive statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
6.3 Estimation results: Does foreign bank participation affect interest
margins? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
6.4 Economic significance of interest margin determinants . . . . . . . . 126
Chapter 1
Introduction
1.1 Background and Motivation
How does foreign bank participation affect banking systems in host countries? Should
countries encourage foreign bank entry or should they instead create more favorable
business conditions for domestic banks? In the present globalized world, these ques-
tions rank high on the agenda of both academic researchers and policymakers. The
issue is particularly complicated, since the theoretical literature does not provide
unambiguous answer to these questions (Williams, 1997). Due to the absence of
a unified theoretical framework, providing the answer to these questions remains
largely an empirical issue.
The empirical assessment of the impact of foreign bank participation on domes-
tic banking systems can be performed from different (but closely interconnected)
perspectives. First of all, it is important to investigate the impact of foreign bank
participation on the overall performance of banks. It is widely recognized that for-
eign bank entry may boost the performance of banking sectors in host countries due
to better managerial expertise, application of modern technologies, and wider access
to international financial markets. However, foreign banks can also perform worse
than domestic banks due to asymmetric information problems and the low quality of
2 Chapter 1
governance in host countries (Lensink et al., 2008). Altogether, the assessment of the
impact of foreign participation on bank performance may be a daunting task given
the sample selection problems associated with the foreign entry decisions. Next,
from a social welfare perspective, it is important to examine the impact of foreign
bank participation on the market structure in the banking sectors and their com-
petitive stance. Here again, the direction of the impact of foreign bank entry is
difficult to anticipate a priori. Opening borders can boost bank competition due to
the exposure of the domestic market to competitors from abroad (Sengupta, 2007).
However, market concentration can also increase following foreign entry if better
performing foreign banks drive out less efficient domestic competitors (Detragiache
et al., 2008). The impact on the market structure might also depend on the mode of
foreign entry (Lehner and Schnitzer, 2008). While entry via establishment of a new
banking institution (greenfield investment) results in an increase in the number of
banks in the host country, entry via takeover of a domestic bank (cross-border ac-
quisition) leaves the total number of banks unchanged, affecting only the ownership
distribution. Finally, given that the banking sector remains the most important ex-
ternal source of finance for firms, it is important to investigate the impact of foreign
bank participation on the cost of bank financing. The net interest margin, defined as
the difference between the average interest rate received by banks for their lending
activities and the average interest rate paid to depositors for their funds, serves as a
good benchmark for analyzing the impact of foreign bank participation on the cost
of financing. Theoretical models of interest margin determinants do not discuss the
impact of bank ownership, but suggest that the main factors through which foreign
bank participation can influence the cost of bank financing are related to the perfor-
mance of banks and market competition (Maudos and Fernandez de Guevara, 2004).
Whether foreign bank participation can have a direct impact on bank financing af-
ter accounting for its impact on performance and competition needs to be analyzed
empirically.
Introduction 3
The aim of this book is to provide a comprehensive empirical assessment of foreign
bank entry and its implications for banking systems in former socialist economies
(FSEs). There are several reasons why FSEs provide a fertile ground for analyzing
implications of foreign bank entry on domestic banking systems. First, FSEs have
experienced the largest inflow of foreign bank participation in the world (IMF, 2000).
This provides a large amount of observations at individual bank level, which is es-
sential for conducting an empirical analysis. Second, FSEs have started from a very
low level of foreign bank participation in mid 1990s (EBRD, 2006). Transition from
a largely domestic-owned banking system to a largely foreign-owned banking system
during a relatively short period of time provides a unique opportunity to analyze
the implications of ownership change for various aspects of banking system perfor-
mance. Finally, despite a common socialistic heritage, FSEs remain heterogeneous
in terms of the progress of their economic reforms, institutional background, and
level of integration into the European Union (EU) (EBRD, 2006). Such heterogene-
ity enables analyzing the mediating effects of the macroeconomic and institutional
environments on the relationship between foreign bank participation and banking
system performance. Given these unique characteristics of the FSE banking sectors,
the outcomes of our analysis in terms of answers to the aforementioned questions
can be generalized.
The remainder of this introduction discusses the empirical methodology used
in the analysis and provides an outline of the thesis, briefly reviewing each of the
chapters and their value added to the literature.
1.2 Methodology
Given the empirical nature of the book, the main contribution to the literature is re-
lated to the application of new empirical techniques, which enable analyzing research
questions that were overlooked in previous studies. In this respect, two innovative
4 Chapter 1
approaches used in this study are worth mentioning. The first one is the applica-
tion of a two-step approach for analyzing possible endogeneity in the relationship
between foreign ownership and bank performance. Most previous studies employed
a stochastic frontier analysis, in which the efficiency of a bank is modeled as a func-
tion of its ownership, usually measured through a dummy variable distinguishing
between domestic (dummy = 0) and foreign (dummy = 1) banks. A drawback of
this approach is that the impact of foreign ownership on bank efficiency will be over-
estimated in the presence of the cream-skimming effect, according to which foreign
banks are targeting more efficient domestic banks for acquisition. To account for
possible endogeneity due to sample-selection, in the first step the propensity of a
domestic bank being acquired by foreign investors is estimated using instrumental
variables. In the second step, the dummy variable of bank ownership is replaced
by the estimated propensity score indicator. The coefficient of the propensity score
variable is free from endogeneity effects. It provides support for the existence of the
cream-skimming effect, suggesting that previous results on the relationship between
bank ownership and its efficiency should be interpreted with care and, in some cases,
reconsidered.
The second methodological innovation is the application of latent class tech-
niques. These techniques are computationally intensive and became feasible to
econometricians only recently, along with the advancement of computer technologies.
Application of the latent class stochastic frontier methodology for analyzing bank
performance allows accounting for differences in technological regimes in banking.
Empirical analysis lends statistical support for the existence of different technologi-
cal regimes in banking. These differences have not been discussed in most previous
studies, which may have led to overestimation of bank inefficiency as differences in
technological regimes were mistakenly attributed to underperformance. In addition,
application of latent class logit analysis enables testing for the importance of various
Introduction 5
institutional factors driving foreign bank entry. Previous studies based on a pooled
logistic model did not account for environmental heterogeneity, which has proven to
be a statistically significant driving factor.
1.3 Outline of the Thesis
This thesis consists of five chapters addressing the impact of foreign bank participa-
tion on banking systems in host countries. Chapter 2 assesses whether bank efficiency
endogenously determines decisions on foreign acquisition (cream-skimming effect).
Chapter 3 focuses on the institutional determinants influencing decisions of foreign
banks to go abroad. Chapter 4 analyzes how the impact of foreign ownership on
bank performance can be moderated by differences in banking technology regimes.
Chapter 5 investigates the impact of different modes of foreign entry (greenfield
investments and cross-border acquisitions) on bank competition in host countries.
Chapter 6 evaluates the impact of foreign bank participation on financial interme-
diation costs. The final chapter concludes.
Chapter 2 addresses the question of the impact of foreign bank participation on
bank performance. When policymakers in FSEs liberalized their banking markets
and encouraged foreign entry, they were largely motivated by potential efficiency
gains that foreign entry would bring to domestic banking systems. The aim of
the chapter is to test whether these expected benefits have materialized after two
decades of liberalization reforms using individual bank data. A two-stage stochastic
efficiency frontier model is applied, in which the probability that a domestic bank
will be taken over by a foreign bank, obtained in the first stage, enters the second-
stage specification among the cost efficiency determinants. The outcomes from this
model are compared to estimates obtained from the more conventional single-step
model in which a foreign ownership dummy variable enters the specification among
the cost efficiency determinants (e.g., Bonin et al., 2005, Fries and Taci, 2005). The
6 Chapter 1
comparison of the two models provides support for the cream-skimming hypothesis
(sample selection), according to which foreign banks target more efficient banks in
FSEs. This makes the interpretation of the positive impact of foreign ownership on
bank efficiency reported in previous studies less convincing.
Chapter 3 focuses on the institutional environment in host countries. Two com-
peting hypotheses explaining the decision of foreign banks to enter FSEs are distin-
guished (Lanine and Vander Vennet, 2007). According to the efficiency hypothesis,
the main motivation of foreign entry is the extraction of extra revenues resulting from
the upgrade of the efficiency of acquired banks, while the market power hypothesis
suggests that the extra revenues are expected to be obtained from possessing addi-
tional market power. The relative strength of these competing hypotheses is tested
using a multilevel mixed-effect logistic model.1 The merit of this methodological
approach in comparison to the simple logistic model applied in previous studies is
that it allows conditioning the entry decision on the heterogeneity of institutional
conditions in FSEs. The results clearly highlight the importance of the institu-
tional background and economic development of FSEs in influencing foreign entry
decisions (EBRD, 2006). Support for the efficiency hypothesis is found for foreign
entry into more developed FSEs, while the market power hypothesis is confirmed
for foreign entry into less developed FSEs with weak institutions. These findings
suggest that foreign banks find it more beneficial to upgrade efficiency of target
banks in relatively more advanced FSEs with better economic prospects, while reap-
ing monopolistic rents is more easily attainable in less developed FSEs with a weak
regulatory framework.1 Our analysis is based on a discrete choice modeling framework, in which cross-border acquisition
is defined as the acquisition of more than 50% of the outstanding equity of domestic bank by aforeign investor. An alternative approach would be to use a continuous variable measuring thepercentage of shares acquired by foreign banks and to utilize a standard regression framework.However, as shown by Lensink et al. (2008), foreign banks mostly acquire dominating shares whenentering emerging markets. Therefore, we believe that both approaches would probably lead tosimilar results.
Introduction 7
Chapter 4 provides further insights on the importance of country-specific envi-
ronmental characteristics for the performance of banks. Environmental differences
among FSEs hamper comparative analysis of bank performance, since technolog-
ical regimes of banks may be influenced by the macroeconomic and institutional
environment of the countries in which they operate. Ignoring these environmental
differences may result in biased estimates of bank performance. The implication of
environmental differences for the efficiency of banks across FSEs is explicitly tested
using the latent class stochastic frontier model, which is more general than the stan-
dard stochastic efficiency frontier model employed in previous studies. The main
advantage of the latent class framework is that it allows testing for the impact of
environmental differences on technological regimes of banks and does not impose a
priori restrictions on the sample. We show that there are three distinct technological
regimes in FSE banking, characterized by different levels of efficiency, technological
progress, and country coverage. Comparative analysis of different regimes suggests
that there exists a tradeoff between bank efficiency and technological progress. For
instance, banks located in the new members of the European Union exhibit more
technological progress and lower efficiency than banks located in CIS. Moreover,
foreign entry improves performance of banks located in the new members of the
European Union, with better progress in economic reforms, while the impact on
banks in less developed CIS countries is ambiguous. This finding is in line with
the previous result, according to which foreign bank entry is motivated by efficiency
considerations in more advanced FSEs.
Chapter 5 analyzes competition aspects of foreign bank participation. Foreign
bank participation was expected to lead to a more competitive and vibrant banking
environment in FSEs. Although the concepts of competition and performance are
intrinsically interrelated, most previous work analyzed them separately. In contrast,
this chapter tests whether foreign bank entry boosts competition in host countries by
8 Chapter 1
taking into account a possible impact of bank efficiency on market competition. To
test this hypothesis, we explicitly differentiate between two modes of foreign entry:
establishment of greenfield subsidiaries and cross-border acquisitions. While cross-
border acquisitions aim at expanding the business to the FSEs, the greenfield entry
is primarily motivated by serving the clients of the parent bank abroad. Empirical
analysis provides support for the hypothesis of increased competitive pressure follow-
ing foreign entry, but only for the case of cross-border bank acquisitions. Greenfield
entry does not result in higher competition, which may be due to the relationship
lending of greenfield banks to their clients abroad. Increased foreign participation
has not led to a fully competitive market structure in FSEs.
Chapter 6 studies the impact of foreign bank participation on financial interme-
diation costs in host countries. Foreign bank participation was expected to improve
accessibility of finance in FSEs and to decrease the cost of credit via efficiency im-
provement and greater competition. Using net interest margins as proxy for financial
intermediation costs, we analyze the impact of these channels using the dealership
model as an underlying theoretical framework (Ho and Saunders, 1981). This model
assumes that banks serve as risk-averse dealers in the deposits and loans market,
bearing the risk of refinancing due to the possible mismatch between the arrival of
deposits and demand for loans. It has become a standard benchmark in empirical
studies of interest margin determinants. We show that when all theoretically moti-
vated (e.g., market concentration, credit and market risks, bank risk aversion) and
environmental variables (e.g., liquidity) are taken into account, foreign bank partic-
ipation has no direct or indirect significant effect. This result calls for reassessment
of some of the previous studies.
The final chapter summarizes the main findings of the study and discusses its
policy implications.
Chapter 2
Foreign Ownership and BankEfficiency: Does SampleSelection Matter?
2.1 Introduction
The recently observed rapid expansion of foreign banks into former socialistic economies
(FSEs) has been largely fueled by economic reforms and special policies undertaken
by local authorities aimed at attracting foreign direct investments into the financial
sector (EBRD, 2005). One motivation for opening the borders is the expected im-
provement of bank performance in FSEs. A more efficient banking system is believed
to facilitate financial intermediation and to contribute to the optimal allocation of
financial resources in the real sector (Bonin and Wachtel, 2003).
But does foreign ownership indeed improve bank performance? Theoretical stud-
ies do not provide a straightforward answer to this question (Sengupta, 2007, Detra-
giache et al., 2008). On the one hand, foreign banks have better access to advanced
information technologies and more expertise in comparison to their domestic peers.
Foreign banks may also import better supervision and regulation practices and in-
crease competition. In addition, they may be less vulnerable to political pressure
and less inclined to lend to connected parties.
10 Chapter 2
On the other hand, domestic banks have a better know-how of the domestic
economy and understand the specifics of domestic legal systems, traditions, and
other domestic institutions. They more easily carry out lending to opaque firms,
which they can monitor better than foreign competitors.
In the absence of an unambiguous theoretical prediction on the relationship be-
tween foreign ownership and bank performance, a number of studies tried to address
this question empirically using data on FSEs. Most of these studies employed ef-
ficiency frontier methodology to analyze the impact of foreign ownership on bank
efficiency.1 The empirical evidence seems to largely support the notion that foreign
ownership has a positive impact on bank efficiency. Single-country studies report a
positive impact for Hungary (Hasan and Marton, 2003), Croatia (Jemrić and Vu-
jčić, 2002), and Poland (Nikiel and Opiela, 2002). Based on different sample periods
and country coverage, most of the cross-country studies also find a positive associ-
ation between foreign ownership and bank efficiency (see Table 2.1). Bonin et al.
(2005) report that the participation of international investors adds considerably to
cost efficiency of banks. Yildirim and Philippatos (2007) find that foreign banks
are more cost efficient but less profit efficient relative to domestic private and state-
owned banks. Fries and Taci (2005) use a unique database on banks compiled by the
EBRD and provide a detailed ownership breakdown into five categories: greenfield
foreign-owned, greenfield domestic-owned, privatized foreign, privatized domestic,
and state-owned. Their estimation results suggest that privatized banks with ma-
jority foreign ownership are the most cost efficient, followed by greenfield banks
(domestic and foreign).
There are two ways how one can reconciliate the mismatch between the ambigu-
ous theoretical predictions and consensus in the empirical literature. One is to argue
1 See Kumbhakar and Lovell (2000) and Coelli et al. (2005) for a textbook exposition of efficiencyfrontier methodology. Berger and Mester (1997) and Hughes and Mester (2008) review applicationsof the efficiency frontier methodology in the banking industry.
Foreign Ownership and Bank Efficiency: Does Sample Selection Matter? 11
that in practice the advantages related to foreign ownership outweigh its disadvan-
tages, leading to a positive overall effect of foreign ownership on bank efficiency. This
is a popular interpretation provided in most empirical studies. Another possibility is
to challenge the empirical findings on the ground of a possible endogeneity bias due
to the cream-skimming (or cherry-picking) effect (Roll, 1986). According to this hy-
pothesis, foreign investors may select the most efficient banks for acquisition, which
makes the sample from which the individual observations are drawn non-random.2
In other words, the cream-skimming effect implies that the positive impact of for-
eign ownership comes from the fact that those banks that were acquired by foreign
investors were initially more efficient (i.e., the acquired banks would perform well
even if they have had remained domestic). Surprisingly, this interpretation has been
largely neglected in the literature.
The aim of this chapter is to assess the possible endogeneity bias in the relation-
ship between foreign ownership and bank efficiency in FSEs. Our inquiry is moti-
vated by previous indirect evidence on the selection issues associated with the deci-
sion of foreign banks to enter FSEs. For instance, Lanine and Vander Vennet (2007)
show that foreign banks explicitly target large banks in FSEs in order to extract
benefits from increased market power. Similarly, Poghosyan and De Haan (2008)
document that the characteristics of target banks in terms of their size and per-
formance depend on the macroeconomic environment and institutional background
of host countries. In addition, some empirical evidence from developed economies
(Berger et al., 1999) and developing economies (Lensink et al., 2008, Detragiache
et al., 2008) suggests a negative association between foreign ownership and bank
efficiency.2 Surveying the empirical literature on FDI in developing economies, Navaretti and Venables (2004)
point out that much of the available empirical evidence “supports a statistical association betweenforeign ownership and productivity, but not a causal link”. They also report that those studiesthat examine the causal relationship more carefully conclude that the impact of foreign directinvestments is smaller and sometimes even insignificant. The reasoning is that if multinational cor-porations simply select high-performing firms in the host country for acquisition, the productivityadvantages may not be related to ownership.
12 Chapter 2
To evaluate the impact of endogeneity, we apply a two-step estimation method in
the spirit of the Heckman (1979) procedure.3 In this setup, the probability of acqui-
sition (the propensity score) is estimated in the first step, and then used to control
for the selection bias in the second step. This method has found wide-ranging ap-
plications in studies on ownership and total factor productivity of firms in many
countries, including emerging economies (Djankov and Hoekman, 2000). We are not
aware of any study that applies a two-step instrumental variable method for analyz-
ing the relationship between foreign ownership and efficiency in the banking sectors
of emerging countries. Our estimations support the cream-skimming hypothesis and
suggest that foreign banks target more efficient banks in FSEs, which makes the
empirical assessment of the relationship between foreign ownership and bank effi-
ciency complicated. After correcting for the endogeneity bias, the impact of foreign
ownership on bank efficiency becomes negative, which is in sharp contrast to most
previous evidence.
The remainder of this chapter is structured as follows. Section 2.2 describes the
two-step approach used in our empirical analysis and data. Section 2.3 discusses the
estimation results, and the last section concludes.
2.2 Methodology and Data
In this section, we describe the two-step instrumental variable approach we propose
for the investigation of the extent and significance of endogeneity bias due to the
cream-skimming effect. Following previous empirical studies on the relationship
between foreign ownership and bank efficiency in FSEs, we start by specifying a
translog cost function for the stochastic efficiency frontier analysis. The estimation
3 An alternative possibility would be to use a matching technique (non-parametric method), whichallows to control for the selection bias by examining pairs of observations with similar observablecharacteristics. Using this procedure, one is able to proxy for the unobservable counterfactual,i.e., compare the performance of the acquired bank with its performance if it had not been ac-quired. However, this method requires a large number of observations on matched bank pairs andis unsuitable in many applications (including ours).
Foreign Ownership and Bank Efficiency: Does Sample Selection Matter? 13
results from this non-instrumented specification are then compared to our two-stage
instrumental variable outcomes. Different impact of foreign ownership obtained in
these two specification indicates a bias due to the cream-skimming effect.
2.2.1 Stochastic efficiency frontier model
Cost efficiency measures the relative performance of a bank by comparing its current
level of costs to the efficiency frontier for a given technology. Since technologically
feasible cost frontiers are not observable, in practical applications the measurement
of cost efficiency is based on deviations from minimal costs observed in a sample
(Aigner et al., 1977). Following Kumbhakar and Lovell (2000), we start from a
general form of the cost function for the ith bank in country j and year t specified
as:
log TCijt = f (Yijt, Xijt, Gjt, t) + vijt + uijt, (2.1)
where TCijt is the total cost of the bank, Yijt represents the vector of outputs, Xijt
represents the vector of input prices, and Gjt is a vector of country-specific factors
driving the cost frontier. The composite disturbance term is the sum of the technical
inefficiency (uijt) and random error (vijt) components.4 The term uijt ≥ 0 captures
the deviations from the best-practice costs due to technical or allocative inefficiency
of the input usage. It is by definition nonnegative and is assumed to be drawn from
a zero-truncated normal distribution: uijt ∼ N+(µijt, σ2u), with the conditional mean
parameter µijt (i.e., the mean of the non-truncated distribution) which we explain
below. The random error term vijt captures the stochastic variability of the frontier
and is assumed to be i.i.d., vijt ∼ N(0, σ2v ). We assume an explicit dependence of the
cost function on time, which should capture the impact of technological advancement
that is otherwise unobservable in our model.4 The general specification (2.1) assumes that inefficiency and random error terms are multiplica-
tively separable from the other variables.
14 Chapter 2
Following other related papers, we apply a semi-logarithmic second-order expan-
sion of the function f (·) to obtain the well-known translog specification of the cost
function (2.1), enriched by country-specific factors. In order to reduce the number
of second-order terms in the regression equation, we assume a linear dependence
between log TC and the country-specific factors. Thus the country-specific variables
operate as linear cost frontier modifiers, and reflect changing operating conditions
within which the banks optimize their operations. This leaves us with the following
model specification:5
logTCijt
Xijt,1= β0 +
S
∑s=2
βs logXijt,s
Xijt,1+
L
∑l=1
γl log Yijt,l +
+12
S
∑s=2
S
∑l=2
δsl logXijt,s
Xijt,1log
Xijt,l
Xijt,1+
+12
L
∑s=1
L
∑l=1
ψsl log Yijt,s log Yijt,l +S
∑s=2
L
∑l=1
ωsl logXijt,s
Xijt,1log Yijt,l +
+τ1t +12
τ2t2 +S
∑s=2
τXs t log
Xijt,s
Xijt,1+
L
∑l=1
τYl t log Yijt,l +
+N
∑n=1
ξnGjt,n + vijt + uijt. (2.2)
In our model, we employ two outputs and two input prices. Variations of this
specification have been employed in other related studies to analyze different aspects
of bank efficiency in emerging countries.6
We are further interested in knowing what factors influence the inefficiency term
uijt. While the country-specific factors constitute a given economic environment for
the banks, and thus cannot form a source of individual bank’s inefficiency, uijt can
depend on bank-specific variables, like financial and ownership structure. In order
to capture these effects, we specify a linear relationship for the conditional mean µijt
5 Specification (2.2) imposes homogeneity in prices by dividing the total cost TCijt and pricesXijt,s, s ≥ 2 by price Xijt,1, i.e., by taking the first input as numeraire. The symmetry of secondpartial derivatives in input prices and in output quantities implies δsl = δls and ψsl = ψls.6 For example, Fries and Taci (2005) employ a variant of specification (2.2) with two outputs and
one input price, Yildirim and Philippatos (2007) and Rossi et al. (2004) assume three outputs andthree inputs, Lensink et al. (2008) use two outputs and two input prices.
Foreign Ownership and Bank Efficiency: Does Sample Selection Matter? 15
of the inefficiency term uijt (Battese and Coelli, 1995):
µijt = λ0 +M
∑m=1
λmZijt,m + αFDIijt, (2.3)
where Zijt is a vector of bank-specific control variables, and FDIijt is a binary variable
which is 0 if the bank is domestically-owned, and 1 if it is foreign-owned. The
control variables Zijt include indicators of the bank’s market power, diversification
of activities, and stability. The residual inefficiency is the part of inefficiency not
captured by the conditional mean described by the observable variables.
2.2.2 Instrumenting foreign ownership
In equation (2.3) it is assumed that foreign ownership is exogenous. This assumption
is not plausible in the presence of the cream-skimming effect, which suggests that
foreign investors tend to acquire the best firms (Navaretti and Venables, 2004).
This means that the decision on purchasing a bank will depend on the investor’s
assessment of the bank’s future potential in terms of cost efficiency. Mathematically
speaking, the foreign ownership dummy variable is stochastically dependent on the
residual inefficiency, which leads to an endogeneity problem. Consequently, the
estimated coefficients, including α, will be biased and inconsistent.
In order to avoid the endogeneity bias, one has to select a set of country- and
bank-specific instruments and pursue a two stage estimation approach widely used
in the treatment effect literature. The instruments are supposed to be correlated
with variable FDI and independent of the residual inefficiency term. In the first
stage, one has to estimate a probit model linking the dummy variable FDIijt and
instruments:
Pijt = Prob(FDIijt = 1|Iijt) = Φ(θ′ Iijt), (2.4)
where Φ(.) is the cumulative distribution function of a Normal distribution, and
16 Chapter 2
Iijt = (Z′ijt, W ′
ijt)′ is a vector of explanatory variables containing the bank-specific
controls Zijt from equation (2.3), and instrumental variables Wijt.
The predicted values Pijt from equation (2.4) represent the estimated probabilities
that the bank will be purchased by a foreign investor based on observed bank-specific
and other characteristics (Iijt). In the second stage, FDIijt in specification (2.3) is
substituted by Pijt. Since Pijt is a function of instruments, which are independent of
the residual inefficiency, the endogeneity bias vanishes and the estimate of parameter
α becomes more accurate.7 By using the instrumental variable method in this form,
we assume that the impact of foreign ownership does not vary with the probability of
selection and that there is no essential heterogeneity present in the data (Heckman
et al., 2006).8
In order to evaluate the expected effect of foreign ownership, we notice that the
mean of the truncated normal distribution conditional on the observables is:
E(uijt | Zijt, P(Iijt)
)= mijt + σu
φ( mijt
σu
)Φ( mijt
σu
) ,
where
mijt = λ0 +M
∑m=1
λmZijt,m + αP(Iijt).
Differentiating this expression with respect to P(Iijt), we get the (marginal) effect
of foreign ownership:
∆(Zijt, P(Iijt)) =∂E(uijt | Zijt, P(Iijt)
)∂P(Iijt)
=
7 We are aware of the fact that the predicted values Pijt contain the prediction error, and substi-tuting these into equation (2.3) without subsequently adjusting the standard errors of the resultingparameter estimates leads to an underestimation of these errors. However, the parameter estimatesthemselves remain consistent and unbiased (Pagan, 1984), which makes us confident to use thisapproach.8 Since specification (2.4) relies on distributional assumptions regarding the functional form of the
interdependence between FDI and the instruments, for the sake of robustness we also run a simpleOLS regression instead of the probit model in the first stage, and cross-check the results. Also,the OLS regression is free of the nonlinearity effects present in the probit model and offers a morerobust way of testing the validity of the instruments.
Foreign Ownership and Bank Efficiency: Does Sample Selection Matter? 17
= α
1−mijt
σu
φ( mijt
σu
)Φ( mijt
σu
) − φ
( mijtσu
)Φ( mijt
σu
)2 .
Averaging out across the sample, we get an estimate of the average unconditional
impact of foreign ownership on the studied banks. Alternatively, we could calculate
the effect of foreign ownership in discrete form as:
∆(Zijt, P(Iijt)) = E(uijt | Zijt, P(Iijt) = 1
)− E
(uijt | Zijt, P(Iijt) = 0
)=
= α + σu
φ
(mijt |P(Iijt)=1
σu
)Φ(
mijt |P(Iijt)=1
σu
) −φ
(mijt |P(Iijt)=0
σu
)Φ(
mijt |P(Iijt)=0
σu
) .
We defer the derivation of these formulas to the Appendix.
2.2.3 Data and descriptive statistics
Our data set is composed of annual bank-level data from selected European and
post-Soviet emerging economies, which is taken from the BankScope database of
Bureau van Dijk. Our sample covers the period from 1993 to 2004, and includes 20
countries: Albania (AL), Armenia (AM), Azerbaijan (AZ), Bulgaria (BG), Belarus
(BY), Czech Republic (CZ), Estonia (EE), Georgia (GE), Croatia (CR), Hungary
(HU), Kazakhstan (KZ), Latvia (LV), Lithuania (LT), Moldova (MD), Poland (PL),
Romania (RO), Slovenia (SI), Slovakia (SK) and Ukraine (UA). To make the data set
representative and mitigate the impact of temporary bank appearances, we restrict
our sample by including only those individual banks that were present in the sample
for at least 4 years. The selection process results in a sample with 1924 observations
for 305 individual banks. The composition of banks in terms of the time spell is
quite even: 55 banks were present for 4 years, 54 banks for 5 years, 42 banks for 6
years, 50 banks for 7 years, and 104 banks for 8 years, which amounts to 220, 270,
252, 350, and 832 observations, respectively.
Table 2.2 displays the distribution of banks and observations across countries
18 Chapter 2
and years. The table also summarizes the distribution of banks by ownership struc-
ture.9 Foreign banks10 are predominant (more than 50%) in 10 (mainly Central and
Eastern European countries (CEEC)) out of 20, while countries where most banks
are domestically-owned are mainly former-USSR countries.11
For the analysis of the banks’ performance, banks are modeled as firms producing
two outputs (loans and deposits) using two inputs (physical capital and labor). Loans
(Y1) are measured as the total amount of loans given out by the bank, and deposits
(Y2) as the total amount of deposits attracted. The price of physical capital (X1) is
defined as the ratio of non-interest expenses to total assets, while the price of labor
(X2) is measured as the ratio of total expenses on personnel over total assets.12
Apart from output and input prices data for individual banks and ownership
indicators, we also employ data on other important country- and bank- specific
correlates of cost efficiency. Among the country-specific correlates, we introduce
the logarithm of per capita GDP (G1), the risk-free interest rate (G2), and the
EBRD index of banking sector reform (G3).13 These variables serve as cost function
modifiers, and should represent inter-country economic and institutional differences
influencing the available cost frontier. We prefer this approach to using country
dummy variables, since the latter approach does not explain the sources of differences
between the countries.
9 The BankScope database provides data on current ownership structure only. We complementedthe database by collecting the missing historical ownership data from webpages of individual banks,public databases, and other sources, and combined them with the data provided by Hein Bogaardand Anita Taci. The cross-validation of data allows us to achieve a substantial level of ownershipdata precision.10 A bank is defined as foreign if the share of foreign stakeholders exceeds 50% of the total equityoutstanding.11 This stylized fact provides evidence that foreign investors acquired banks in relatively moreadvanced CEE economies (with a higher degree of economic development and better establishedmarket institutes) more frequently, which serves as a first empirical justification for the foreignentry based on country-specific characteristics.12 Taking a ratio over the total number of bank employees would be a better proxy for labor costs,but in the absence of data on the total number of employees this is not possible. Yildirim andPhilippatos (2007), Rossi et al. (2004), Fries and Taci (2005), and Lensink et al. (2008) also takethe ratio over total assets for measuring labor costs.13 To ensure consistency of our data set, we use country-specific variables available in variousvolumes of the EBRD “Transition Reports”.
Foreign Ownership and Bank Efficiency: Does Sample Selection Matter? 19
Further, we hypothesize that the accession to the European Union may have a
positive impact on the production opportunities in the acceding countries. Thus we
want to measure whether EU accession itself is a significant cost frontier modifier,
in addition to the indirect impacts through improvement in institutional factors and
economic conditions. Since EU accession is a gradual process, we include the EU
accession trend variable (G4) that is defined as:
Gjt,4 = min
{max
{t− EU applicationj
EU accessionj − EU applicationj, 0
}, 1
}, (2.5)
where EU application j is the year when country j submitted its application to the
European Union, and EU accession j is the year when it actually entered the EU. For
countries which filed the application but have not entered the EU by the end of our
sample, we use the expected year of entry.14 For countries which have submitted the
application, (G4) is zero for years before the submission, then gradually grows to
one at the year of accession, and is one for years after the accession.15 For countries
which have not submitted an application, we set Gjt,4 to zero for all years.
The bank-level correlates serve as explanatory variables for the conditional mean
of the inefficiency term µijt. The net interest margin (Z1) proxies the degree of
competition the bank faces (larger net interest margin indicates more market power).
The ratio of other operating assets to total assets (Z2) measures the diversification of
the individual bank’s operations. The ratio of net loans to total assets (Z3) captures
the ability to transform deposits into loans. Finally, the ratio of equity to total assets
(Z4) serves as an (inverse) indicator of the bank’s leverage. The descriptive statistics
of the data in thousands of US dollars (except for the ratios) are summarized in Table
2.2.
The instruments Wijt used in the first-stage estimation of the propensity score
14 These countries are Bulgaria and Romania, which (as expected) have entered the EU in 2007.15 These countries are Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Slovakia, andSlovenia, which have entered the EU in 2004.
20 Chapter 2
have to be linked to the foreign investor’s decision to purchase the bank but have to
be independent of the cost inefficiency of the given bank, after controlling for bank-
level correlates. We therefore exclude direct measures of the bank’s profitability
and cost structure. The included instruments are the ratio of the population to
the number of banks in the given country, the risk free interest rate, the ratios of
deposits to loans and of assets to the net interest revenue for the given bank, the
time index, and the time index squared.16
2.3 Estimation Results
The results of our empirical estimations using the parametrization of Battese and
Coelli (1995) are summarized in Tables 2.3 and 2.4. The first two columns of the
tables represent specifications with instrumented foreign ownership variables, using
a probit model and OLS, respectively. The third column contains the results of the
specification without instruments. Whenever we do not state explicitly otherwise,
we refer to the probit instrumental variable specification.
2.3.1 Cost frontier specification
Looking first at the translog time-varying cost function component of the model, we
find that most coefficients are highly significant and relatively similar in all three
specifications. This confirms the appropriateness of the time-varying cost function
model. The marginal effects evaluated at variable means are larger than one for both
outputs. This means that a one percent increase in any of the outputs is accompanied
by a more than one percent increase in costs. The sensitivity of total costs to loans
and deposits is largely comparable (the elasticities given by the marginal effects are
equal to 1.54 and 1.56, respectively). In addition, the coefficient of the cross–product
16 Naturally, the independence of the instruments of the residual inefficiency after controlling forthe bank-level correlates is at least to a certain degree a matter of faith. We subject the instrumentsto a series of validity tests that we discuss below. However, these are all based on the assumptionthat at least some of these instruments are valid.
Foreign Ownership and Bank Efficiency: Does Sample Selection Matter? 21
term between deposits and loans is negative and statistically significant. This signals
possible economies of scope in production of two types of banking services and is
consistent with findings by Fries and Taci (2005) and Lensink et al. (2008).
The negative marginal effect of time confirms a downward shift in the cost frontier
over time as a result of improvements in available production technology. However,
the coefficient is insignificant with a relatively high standard error, implying that
there may be substantial differences among the individual countries.
At the country-level, we do not find any significant association between the level
of economic development measured by per capita GDP and total costs. This finding
is consistent with results by Fries and Taci (2005), but differs from those by Lensink
et al. (2008) who report a significant negative association between per capita GDP
and banking costs.
Similarly to Fries and Taci (2005), we find that the level of nominal interest rate
has a positive and significant impact on scaled total costs: a 1 percentage point
increase in the risk-free interest rate in the economy leads to an increase in total
costs by 0.6%. The EBRD index of banking sector reform has a positive and signif-
icant impact on total costs. Fries and Taci (2005) explain the positive association
between banking sector reforms and banking costs by the fact that banks in the
studied emerging economies are moving from defensive restructuring of the banking
operations (cost cutting) to operating strategies based on service improvements and
innovation, which requires a higher level of spending.
The significant negative coefficient of the EU accession trend confirms the posi-
tive impact of EU accession on the productivity of the banking sector. Even after
controlling for the benefits linked to institutional and economic development and for
the evolution of technology over time, we still find that entering the EU shifts the
available cost frontier downward. The estimated gain in cost efficiency due to EU
accession is almost 10%.
22 Chapter 2
2.3.2 Inefficiency analysis
We find a significant negative association between banking costs and our proxy
for the market power of a bank, i.e., the level of the net interest margin (difference
between implicit rates for lending and borrowing).17 This result indicates that banks
with greater market power are able to reduce their costs, possibly due to economies
of scale and scope. It is consistent with the findings by Grigorian and Manole (2006)
and differs from those by Fries and Taci (2005) and Yildirim and Philippatos (2007).
We proxy the degree of diversification of banking activities by the ratio of other
operating income to total assets and find that it is significant and negatively as-
sociated with banking costs. This is in line with the findings of previous studies
and indicates that banks with a greater variety of banking services tend to perform
better. Similarly, banks that are more active in terms of loan provision, proxied by
the ratio of net loans to total assets, are also significantly more cost efficient, which
might be due to economies of scale.
Finally, banks that allocate a greater share of their assets to their capital for
stability reasons sacrifice in terms of cost efficiency.
2.3.3 Impact of ownership
Contrary to other cross-country panel data studies (e.g., Yildirim and Philippatos,
2007, Fries and Taci, 2005, Bonin et al., 2005, and Lensink et al., 2008), we do
not find a significant association between foreign ownership and cost efficiency in
our non-instrumented model (see the specification without instrumental variables in
Table 2.3).
In order to check for the presence of the cream-skimming effect, we first estimate
17 We believe the net interest margin is a better proxy for market power of a particular bank thanthe share of the top largest banks’ assets in the total banking assets – a popular indicator employedin related work. The net interest margin provides a qualitative measure on how banks benefit fromtheir position in the market in terms of price setting, while market share measure can be distortedby specific characteristics of banking sector regulation in a particular country.
Foreign Ownership and Bank Efficiency: Does Sample Selection Matter? 23
a probit model for the decision of foreign investors to acquire domestic banks. In
the probit specification, we use the exogenous variables from our model, and add
the instruments described above.
After instrumenting for foreign ownership, we find a substantial change in the
impact of foreign ownership on cost efficiency (see Table 2.3, columns 1 and 2). The
coefficient of the FDI variable becomes significantly positive, which implies that
there is a negative relationship between foreign ownership of a domestic bank and
cost efficiency. This suggests that foreign investors do not improve, but rather worsen
cost efficiency. The insignificant coefficient in the specification without instrumental
variables is caused by the fact that the less favorable performance in terms of cost
efficiency is partly offset by the fact that foreign investors tend primarily to acquire
banks with high residual efficiency that is not captured by our efficiency correlates.
These two effects (worse cost efficiency under foreign ownership and the endogeneity
of the foreign ownership variable) work in opposite directions, making the coefficient
of FDI in the non-instrumented model insignificant. The negative impact of foreign
ownership on cost efficiency is uncovered in the instrumental variable specification,
and confirms the cream-skimming hypothesis. Since cream-skimming is related to
the residual efficiency not captured by observable quantities, it may be partially
caused by insider information of foreign investors about the acquired domestic banks.
This finding supports the evidence by Lanine and Vander Vennet (2007) that
“large Western European banks have targeted relatively large and efficient CEEC
banks with an established presence in their local retail banking markets”. In addition,
the empirical finding has its theoretical justification in Detragiache et al. (2008), who
show that in a world with imperfect competition and informational asymmetries
foreign entry can lead to diminishing efficiency of the banking sector.
The quantitative impact of foreign ownership on the cost efficiency, averaged out
over the sample, is ∆ = 0.39 (and ∆ = 0.40 in the discrete version), which means
24 Chapter 2
that foreign-owned banks are 39% less cost-effective than their domestic counterparts
with the same observable characteristics. This seems to be a lot, and it is likely to
be a composed effect of lower cost efficiency, the pursuit of expansionary strategies,
the focus on higher-quality services, and possibly tighter accounting standards. It
also does not mean that higher cost makes the foreign-owned banks less competitive,
as long as these higher cost can be offset by higher revenues.
2.3.4 Inefficiency scores
Figure 2.1 presents estimated average inefficiency terms in both models (without
instruments and with instruments using the first-stage probit model) for the set of
countries under consideration. It can be observed that both specifications produce
comparable inefficiency scores.
The overall average inefficiency is 0.45, indicating that banks are, on average,
operating 45 % above the optimal cost frontier.18 The results among the countries
vary substantially. The worst performer is Albania, but otherwise the economically
less developed countries do not underperform. The Visegrád countries (Czech Re-
public, Hungary, Poland, and Slovakia) show above-average inefficiency, with the
Czech Republic showing the highest level of cost inefficiency in this group. This is
consistent with the findings of previous studies. Incidentally, these are the countries
which were very successful in attracting foreign direct investments into their banking
systems.
Otherwise, it is rather difficult to spot any discernible pattern. Baltic coun-
tries fare quite well, with Estonia and Lithuania being among the best performing
countries. However, Latvia shows a cost efficiency level comparable to the sample
average. Banks in the Commonwealth of Independent States (CIS) exhibit middle
range inefficiencies, with two well-performing outliers: Belarus and Georgia, the lat-
18 All levels and differences are reported in logarithmic form.
Foreign Ownership and Bank Efficiency: Does Sample Selection Matter? 25
ter being the best performing country in the sample. The three analyzed countries of
the former Yugoslavia – Slovenia, Croatia, and Macedonia – are better than average
in terms of cost efficiency. Bulgaria is among the top performers, while its neighbor
Romania lags significantly behind.
2.3.5 Robustness checks
In order to check the robustness of our results, we perform several additional esti-
mations and tests for validity of instruments.
First, as we have already mentioned, we instrumented the foreign ownership
variable with both a probit and an OLS regression. As seen in Tables 2.3 and 2.4, the
coefficients do not change substantially, and their significance remains approximately
the same. Also the average inefficiencies for individual countries (available from the
author upon request) remain almost unchanged.
Second, we want to make sure that the results are not characteristic only to our
particular selection of banks. In the original estimations, we selected only banks
that are present in the sample for at least 4 years. In order to verify the results, we
create another data set with banks that appear in the sample for at least 5 years.
The quantitative results change only slightly, and the qualitative properties remain
valid.
We estimate the model with quantities expressed in USD using the nominal
exchange rates. In order to investigate the role of possibly misaligned nominal
exchange rates, we estimate the model also with quantities denominated in USD in
terms of purchasing power parity (PPP). We find only minor changes compared to
our base model; in particular, the coefficient at the logarithm of GDP per capita
becomes significant. When expressed in PPP, GDP per capita is positively linked
with banking costs (higher GDP implies higher costs, which is consistent with the
findings of Yildirim and Philippatos, 2007).
26 Chapter 2
These findings suggest that the decision to use nominal or PPP-implied exchange
rates can play a role in studies on the relationship between GDP per capita and the
banking costs. However, our other results, most importantly the decomposition of
the cost inefficiency estimate, are virtually unchanged. The choice between nomi-
nal and PPP-implied exchange rate does not change our conclusion about the role
of cream-skimming in the evaluation of the impact of foreign ownership on cost
efficiency.
Finally, we check the validity of our instruments. First, we implement the test
procedure from Stock and Yogo (2002) and Stock et al. (2002) to determine whether
the instruments are weak or not. Using the results of the first-stage OLS regres-
sion (right column of Table 2.5), we calculate the F-statistic corresponding to the
hypothesis that the coefficients of all instruments are zero. With the value of the
F-statistic, F=24.54, we reject the null hypotheses of weak instruments outlined in
Stock et al. (2002).19
Further, since we have more instruments than endogenous variables, we can
test the overidentifying restrictions. Under the null of exogenous instruments, the
Hansen-Sargan statistic is χ2-distributed with 5 degrees of freedom. In our case, the
value of the Hansen-Sargan statistic is 8.93 (p-value is 0.112) and we thus cannot
reject the null of exogenous instruments at conventional confidence levels.
Finally, we split the instruments into two halves and use only one half of the in-
struments for instrumenting the foreign ownership dummy variable, while including
the other half as exogenous explanatory variables (see Table 2.6). The fact that the
coefficients of the instruments used as exogenous variables are insignificant strength-
ens our confidence that we are using a set of valid instruments in our estimations.
19 These null hypotheses are as follows: (i) the relative bias of the estimator in the second-stageregression is larger than 10% of the bias of the non-instrumented estimator, and (ii) the size ofthe 5% t-test for α = 0 is larger than 15%. Table 1 in Stock et al. (2002) suggests that the ruleof thumb for the rejection of the weak instruments hypothesis is the estimate of the first-stage Fstatistic that is larger than 10, which is the case in our estimations.
Foreign Ownership and Bank Efficiency: Does Sample Selection Matter? 27
2.4 Conclusions
We address the issue of foreign ownership and bank efficiency in former socialistic
emerging economies. We employ the instrumental variable approach to tackle the
sample selection problems caused by the possibility of cream-skimming. Our main
observation is that the instrumental variable approach makes the coefficient of the
impact of foreign ownership on bank efficiency positive and highly significant. This
finding indicates the presence of cream-skimming, i.e., foreign investors target the
most efficient banks for acquisition. The coefficient of the foreign ownership vari-
able becomes significant in both probit and linear regression specifications, which
implies robustness of the result with respect to the distributional assumptions and
nonlinearities present in probit model.
The quantitative evolution of the impact of foreign ownership shows that foreign-
owned banks are about 39% less cost-efficient than their comparable domestic coun-
terparts. However, this number includes both the pure cost inefficiency, as well as
possibly increased costs due to expansionary strategies, or focus on higher-quality
services.
Furthermore, our estimations suggest that emerging countries that started nego-
tiations on EU accession and eventually became (or will soon become) EU members
experienced a downward shift in the cost frontier. This result documents that im-
proved discipline resulting from the obligations related to the EU accession, together
with benefits coming from technological and market spillovers, improves the tech-
nology of the banking sector in the accession countries.
The comparison of inefficiency scores provides evidence that the most advanced
emerging countries (Czech Republic, Hungary, Poland, Slovakia) and Albania have
the most inefficient banks. This result suggests that opening the financial sector for
foreign entry does not necessarily improve the performance of banking institutions.
Drawing parallels with the previous findings on a downward shift of the cost frontier
28 Chapter 2
due to the EU accession, we interpret this result as the inability of the emerging mar-
kets that have recently entered the EU to accommodate the improved technological
possibilities and fully enjoy the gains stemming from productivity improvements.
We would like to emphasize, however, that the negative association between for-
eign ownership and cost efficiency should not be confused with the contribution
of foreign ownership to the stability of financial systems in emerging markets. The
results should be rather interpreted as evidence of inefficient use of inputs by foreign-
owned banks given the input prices and other country- and bank-specific character-
istics. In other words, foreign-owned banks in emerging economies might be more
active in terms of providing, say, more credits to local clients or extending banking
services within their local networks in emerging markets (Giannetti and Ongena,
2005, Giannetti and Ongena, 2008). As was mentioned in Detragiache et al. (2008),
a possible reason why this is not happening is that foreign-owned banks prefer sta-
bility to efficiency, and engage in activities with either top–ranked domestic clients,
or foreign firms and governmental organizations to ensure safety of their operations.
In addition, we do not want to necessarily associate the negative impact of for-
eign ownership on cost efficiency with underperformance. After entering the new
market, the foreign owner can follow strategies related to long-term success and
development, which may be costly in the short-run. These include aggressive ex-
pansion in the market, or deep modernization and restructuralization, which usually
require additional spending. However, this does not change our conclusion about
foreign banks targeting primarily more efficient domestic banks.
To conclude, the results of our estimations suggest that opening domestic finan-
cial systems for foreign entry should not be regarded as a panacea for policymakers
in emerging economies. To enjoy full benefits from foreign acquisition, the countries
should develop appropriate strategies to diminish the impact of the cream-skimming
effect.
Foreign Ownership and Bank Efficiency: Does Sample Selection Matter? 29
Appendix
Derivations of the impact of foreign ownership on cost effi-ciency
In our setup, the cost inefficiency term has a truncated normal distribution u ∼
N+ (m, σ2u). Assume that there is a random variable w ∼ N
(m, σ2
u). The mean of
this inefficiency term u is
E (u) = E (w | w > 0) = m + σuE(
w−mσu
| w > 0)
=
= m + σuE(
w−mσu
| w−mσu
> − mσu
)Since w−m
σu∼ N (0, 1), we can further write
E (u) = m + σu
φ(− m
σu
)1−Φ
(− m
σu
) = m + σu
φ(
mσu
)Φ(
mσu
)where the last fraction is the inverse Mills ratio. Since
m = λ0 +M
∑m=1
λmZm + αP (I) ,
the (marginal) impact of foreign ownership on the expected inefficiency is
∂E (u)∂P (I)
=∂m
∂P (I)∂E (u)
∂m=
= α
1 + σu
1σu
φ′(
mσu
)Φ(
mσu
)− 1
σu
(φ(
mσu
))2
(Φ(
mσu
))2
Noticing that φ′ (y) = −yφ (y), we can complete the derivation by writing
∂E (u)∂P (I)
= α
1− mσu
φ(
mσu
)Φ(
mσu
) − φ
(mσu
)Φ(
mσu
)2 .
30 Chapter 2
Figure 2.1. Average inefficiency scores for individual countries.
Notes: AL - Albania, AM - Armenia, AZ - Azerbaijan, BG - Bulgaria, BY - Belarus, CZ - Czech Republic,EE - Estonia, GE - Georgia, HR - Croatia, HU - Hungary, KZ - Kazakhstan, LT - Lithuania, LV - Latvia,MD - Moldova, MK - Macedonia, PL - Poland, RO - Romania, SI - Slovenia, SK - Slovakia, UA - Ukraine
Foreign Ownership and Bank Efficiency: Does Sample Selection Matter? 31
Tabl
e2.
1.Su
mm
ary
ofre
sults
from
pane
ldat
ast
udie
son
bank
effici
ency
inFS
EsG
rigo
rian
&M
anol
e(2
006)
Yild
irim
&P
hilip
pato
s(2
007)
Ros
si,
Schw
aige
r&
Win
kler
(200
4)B
onin
,H
asan
&W
acht
el(2
005)
Frie
s&
Tac
i(2
005)
Sam
ple
1995
-199
819
93-2
000
1995
-200
219
96-2
000
1994
-200
1N
umbe
rof
bank
s58
532
527
222
528
9N
umbe
rof
obse
rvat
ions
1074
2042
1070
856
1897
Num
ber
ofco
untr
ies
1712
911
15M
etho
dD
EA
SFA
&D
FASF
A(F
ouri
er)
SFA
SFA
Effi
cien
cyty
pes
DE
A(1
)-pr
ofit
gene
rati
onco
stan
dpr
ofit
cost
and
profi
tco
stan
dpr
ofit
cost
DE
A(2
)-se
rvic
epr
ovis
ion
Mea
neffi
cien
cyC
ost
0.39
-0.7
1D
FA-0
.72;
SFA
-0.7
60.
36-0
.87
0.41
-0.7
80.
40-0
.75
Pro
fitN
/AD
FA-0
.66;
SFA
-0.5
0.32
-0.7
10.
5-0.
82N
/AC
ount
ry-l
evel
fact
ors
GD
Pgr
owth
++
N/A
N/A
?In
flati
onra
te?
N/A
N/A
N/A
N/A
Mon
etar
yde
pth
?N
/AN
/AN
/A+
Stoc
km
arke
tca
pita
lizat
ion
+N
/AN
/AN
/AN
/AM
arke
tco
ncen
trat
ion
+−
(cos
t);
+(p
rofit
)N
/AN
/A?
Ban
king
sect
orre
form
s+
N/A
N/A
N/A
+(l
evel
);−
(squ
ared
)N
on-b
anki
ngse
ctor
refo
rms
+N
/AN
/AN
/AN
/AIn
tere
stra
teN
/AN
/AN
/AN
/A+
Ban
k-le
vel
fact
ors
Cap
ital
izat
ion
+−
(cos
t);
?(pr
ofit)
N/A
++
Fore
ign
owne
rshi
p+
+(c
ost)
;−
(pro
fit)
N/A
++
Tot
alas
sets
(in
log)
N/A
+(c
ost)
;−
(pro
fit)
N/A
N/A
N/A
Shar
eof
loan
sN
/A+
(cos
t);−
(pro
fit)
N/A
N/A
N/A
Shar
eof
non-
loan
asse
tsN
/AN
/AN
/AN
/A−
Shar
eof
non-
perf
orm
ing
loan
sN
/AN
/AN
/AN
/A−
Dep
osit
mar
ket
shar
eof
bank
N/A
N/A
N/A
N/A
+N
otes
:+
,−
and
?in
dic
ate
pos
itiv
e,n
egat
ive,
and
insi
gnifi
can
tim
pac
ton
effici
ency
,re
spec
tive
ly.
32 Chapter 2
Tabl
e2.
2.D
escr
iptiv
est
atist
ics A
LA
MA
ZB
GB
YC
ZE
EG
EH
RH
UK
ZL
TL
VM
DM
KP
LR
OS
IS
KU
AT
otal
#of
obs.
4042
584
4017
640
4223
110
810
369
112
5644
248
126
110
108
167
Tot
al#
ofb
ank
s7
69
17
276
734
1716
1117
97
4020
1718
29
Ow
ner
ship
(%)
Dom
esti
c22
.530
.987
.90.
060
.023
.942
.554
.871
.020
.476
.750
.758
.967
.981
.843
.537
.373
.628
.761
.7F
orei
gn77
.569
.012
.110
0.0
40.0
76.1
57.5
45.2
29.0
79.6
23.3
49.3
41.1
32.1
18.2
56.5
62.7
26.4
71.3
38.3
Ind
epen
den
tva
riab
leT
otal
cost
s(C
)24
.74.
57.
164
.123
0.5
247.
184
.47.
155
.624
1.8
50.1
30.6
23.1
5.0
11.3
251.
415
8.4
103.
810
9.0
38.5
St.
Dev
.40
.02.
813
.221
.751
8.9
412.
612
0.8
6.4
102.
436
6.7
77.2
35.4
31.1
3.6
19.6
389.
434
1.8
155.
414
5.0
66.7
Ou
tpu
tsT
otal
loan
s(Y
1)30
.711
.535
.555
4.9
822.
012
70.0
857.
527
.836
0.5
1236
.629
4.2
274.
516
7.8
19.7
31.1
1103
.831
1.8
669.
448
8.8
154.
2S
t.D
ev.
31.2
9.1
77.6
375.
320
36.1
2046
.415
94.7
25.8
784.
219
76.7
579.
652
2.7
344.
916
.932
.217
60.1
600.
011
17.3
606.
628
7.0
Tot
ald
epos
its
(Y2)
338.
931
.368
.587
4.1
1266
.626
35.6
1007
.635
.756
0.0
1869
.434
0.3
401.
929
9.5
26.0
59.0
1892
.067
4.9
1002
.310
89.3
214.
9S
t.D
ev.
524.
425
.914
7.8
292.
534
13.9
4397
.117
24.0
38.7
1215
.427
69.4
522.
668
5.0
465.
722
.598
.530
97.1
1252
.415
15.1
1507
.638
0.4
Inp
ut
pri
ces
(%)
Non
-in
tere
stex
pen
ses/
tota
las
sets
(X1)
2.8
7.7
9.0
4.0
7.3
3.4
4.2
7.3
4.8
3.5
5.9
5.0
4.8
6.2
11.4
3.4
5.9
2.9
5.2
6.7
St.
Dev
.2.
56.
16.
30.
12.
44.
62.
42.
24.
71.
92.
95.
45.
22.
912
.01.
84.
31.
012
.84.
2P
erso
nn
elex
pen
ses/
tota
las
sets
(X2)
1.2
2.9
2.4
1.2
4.4
0.9
2.0
3.2
2.1
1.5
3.1
3.0
2.0
3.4
2.1
1.9
3.3
1.6
1.0
2.7
St.
Dev
.0.
71.
82.
10.
12.
00.
50.
91.
31.
00.
82.
41.
41.
31.
20.
40.
92.
00.
40.
41.
7
Ban
k–s
pec
ific
corr
elat
es(%
)O
ther
oper
atin
gas
sets
/tot
alas
sets
(Z1)
1.6
6.4
7.8
3.3
7.4
2.0
3.3
5.5
2.9
2.3
6.0
3.6
3.2
7.3
6.0
2.3
4.1
2.4
3.7
6.5
St.
Dev
.1.
04.
15.
80.
45.
12.
92.
52.
31.
91.
53.
52.
23.
82.
43.
81.
53.
61.
215
.35.
3N
etlo
ans/
tota
las
sets
(Z2)
22.3
37.5
42.2
47.4
52.5
39.3
51.9
54.1
50.2
50.2
50.6
49.7
36.0
47.6
46.4
46.6
38.1
53.0
42.8
52.3
St.
Dev
.16
.319
.120
.618
.812
.919
.716
.310
.311
.018
.318
.013
.523
.012
.620
.016
.419
.210
.515
.718
.1N
etin
tere
stm
argi
n(Z
3)4.
312
.56.
96.
210
.33.
05.
015
.75.
35.
07.
75.
24.
511
.17.
35.
29.
33.
73.
39.
6S
t.D
ev.
1.9
7.6
3.8
1.0
5.2
2.3
1.7
5.3
2.7
3.9
3.2
3.5
2.7
4.1
3.9
3.3
5.3
1.8
1.4
7.7
Cos
tto
inco
me
rati
o(Z
4)88
.451
.275
.557
.875
.782
.676
.553
.774
.871
.763
.891
.683
.157
.450
.969
.279
.465
.582
.060
.8S
t.D
ev.
146.
018
.677
.86.
637
.896
.139
.518
.245
.026
.521
.934
.564
.216
.518
.725
.940
.632
.781
.624
.1E
qu
ity
/tot
alas
sets
(Z5)
11.8
12.9
20.9
15.9
15.2
8.1
10.6
26.3
17.5
9.4
17.4
12.2
12.0
26.6
31.3
12.6
20.3
10.3
9.1
18.1
St.
Dev
.13
.97.
816
.30.
78.
36.
55.
112
.713
.04.
413
.911
.013
.114
.513
.212
.611
.53.
68.
210
.7
Cou
ntr
y–s
pec
ific
corr
elat
esIn
terb
ank
rate
(G1)
11.0
18.9
20.4
2.3
48.0
6.4
8.3
22.9
7.8
12.7
9.2
7.7
5.0
17.6
12.3
14.8
47.1
6.2
10.5
20.4
St.
Dev
.5.
810
.72.
01.
130
.54.
75.
89.
65.
44.
46.
64.
75.
28.
95.
27.
239
.22.
97.
118
.2P
erca
pit
aG
DP
inU
SD
(G2)
1526
.471
8.4
740.
723
29.5
1555
.766
95.7
4882
.077
1.5
5095
.959
98.9
1694
.137
67.7
3557
.247
9.6
2116
.145
45.2
2151
.211
253.
946
75.3
892.
6S
t.D
ev.
475.
517
8.5
160.
463
1.7
428.
719
01.5
1755
.217
4.4
1125
.518
68.8
499.
914
76.3
1162
.113
7.8
346.
394
6.5
620.
423
12.5
1415
.323
2.8
Ind
exof
ban
kin
gre
form
s(G
3)2.
32.
32.
23.
31.
43.
43.
62.
43.
34.
02.
63.
03.
32.
32.
83.
22.
73.
23.
12.
1S
t.D
ev.
0.2
0.0
0.1
0.3
0.4
0.3
0.3
0.1
0.4
0.2
0.3
0.3
0.4
0.2
0.1
0.1
0.2
0.1
0.3
0.1
Ind
exof
econ
omic
free
dom
(G4)
3.4
3.1
3.9
3.2
4.1
2.3
2.0
3.6
3.4
2.6
3.8
2.6
2.6
3.3
3.2
2.9
3.5
3.1
2.9
3.8
St.
Dev
.0.
20.
40.
40.
20.
10.
10.
30.
30.
20.
30.
20.
40.
20.
20.
10.
20.
20.
20.
30.
2N
otes
:A
L-
Alb
ania
,A
M-
Arm
enia
,A
Z-
Aze
rbaij
an,
BG
-B
ulg
aria
,B
Y-
Bel
aru
s,C
Z-
Cze
chR
epu
bli
c,E
E-
Est
onia
,G
E-
Geo
rgia
,H
R-
Cro
atia
,H
U-
Hu
nga
ry,
KZ
-K
azak
hst
an,
LT
-L
ith
uan
ia,
LV
-L
atv
ia,
MD
-M
old
ova,
MK
-M
aced
onia
,P
L-
Pol
and
,R
O-
Rom
ania
,S
I-
Slo
ven
ia,
SK
-S
lova
kia
,U
A-
Uk
rain
e
Foreign Ownership and Bank Efficiency: Does Sample Selection Matter? 33
Table 2.3. Panel estimation of stochastic efficiency frontier modelsInstruments – Probit Instruments – OLS Without Instruments
Independent variablesConstant 0.8857∗∗∗ 0.8955∗∗∗ 0.7941∗∗∗
(0.1513) (0.1478) (0.1556)log(y1) 0.4713∗∗∗ 0.4692∗∗∗ 0.5494∗∗∗
(0.0561) (0.0518) (0.0566)12 (log(y1))2 0.2029∗∗∗ 0.2023∗∗∗ 0.1968∗∗∗
(0.0060) (0.0060) (0.0066)log(y2) 0.6583∗∗∗ 0.6586∗∗∗ 0.5860∗∗∗
(0.0542) (0.0488) (0.0545)12 (log(y2))2 0.2132∗∗∗ 0.2109∗∗∗ 0.2099∗∗∗
(0.0110) (0.0105) (0.0113)log( x1
x2) 0.5097∗∗∗ 0.5097∗∗∗ 0.5126∗∗∗
(0.0404) (0.0374) (0.0422)12 (log( x1
x2))2 0.1774∗∗∗ 0.1776∗∗∗ 0.1784∗∗∗
(0.0084) (0.0082) (0.0086)log(y1) log(y2) -0.2071∗∗∗ -0.2055∗∗∗ -0.2027∗∗∗
(0.0066) (0.0065) (0.0069)log(y1) log( x1
x2) 0.0306∗∗∗ 0.0288∗∗∗ -0.0307∗∗∗
(0.0112) (0.0108) (0.0112)log(y2) log( x1
x2) -0.0497∗∗∗ -0.0482∗∗∗ 0.0481∗∗∗
(0.0106) (0.0105) (0.0107)t -0.0039 -0.0040 0.0119
(0.0184) (0.0181) (0.0189)12 t2 0.0020 0.0019 0.0010
(0.0017) (0.0017) (0.0018)t · log(y1) 0.0238∗∗∗ 0.0238∗∗∗ 0.0173∗∗∗
(0.0047) (0.0045) (0.0048)t · log(y2) -0.0329∗∗∗ -0.0329∗∗∗ -0.0268∗∗∗
(0.0047) (0.0045) (0.0048)t · log( x1
x2) 0.0045 0.0045 -0.0056
(0.0035) (0.0032) (0.0037)
Country-specific variables (cost frontier modifiers)Log per capita GDP (USD) 0.0162 0.0156 0.0124
(0.0128) (0.0129) (0.0132)Risk-free interest rate 0.0062∗∗∗ 0.0062∗∗∗ 0.0069∗∗∗
(0.0005) (0.0005) (0.0005)EBRD Index of banking sector reform 0.0637∗∗∗ 0.0639∗∗∗ 0.0623∗∗∗
(0.0163) (0.0162) (0.0164)EU accession trend -0.0960∗∗∗ -0.0935∗∗∗ -0.0651∗
(0.0242) (0.0231) (0.0242)
Bank-specific variables (inefficiency correlates)Net interest margin -0.0314∗∗∗ -0.0324∗∗∗ -0.0414∗∗∗
(0.0051) (0.0051) (0.0051)Other operating income/total assets -0.0266∗∗∗ -0.0267∗∗∗ -0.0308∗∗∗
(0.0047) (0.0046) (0.0048)Net loans/total assets -0.0246∗∗∗ -0.0247∗∗∗ -0.0254∗∗∗
(0.0013) (0.0013) (0.0015)Equity/total assets 0.0049∗∗∗ 0.0049∗∗∗ 0.0057∗∗
(0.0013) (0.0013) (0.0014)FDIa 0.6534∗∗∗ 0.6410∗∗∗ 0.0319
(0.1142) (0.1149) (0.0270)Notes: the dependent variable is log( c
x2). Standard errors are given in parentheses. ∗, ∗∗, and ∗∗∗ stand for
10%, 5%, and 1% significance levels, respectively.a Predicted probabilities P(·) of foreign ownership for first and second columns.
34 Chapter 2
Table 2.4. Panel estimation of stochastic efficiency frontier models, cont.Instruments – Probit Instruments – OLS Without Instruments
Marginal effectslog(y1) 1.5408∗∗∗ 1.5405∗∗∗ 1.4698∗∗∗
(0.0427) (0.0428) (0.0448)log(y2) 1.5572∗∗∗ 1.5420∗∗∗ 1.6124∗∗∗
(0.0740) (0.0710) (0.0764)log( x1
x2) 0.7162∗∗∗ 0.7165∗∗∗ 0.8454∗∗∗
(0.0099) (0.0095) (0.0100)t -0.0223 -0.0227∗∗∗ -0.0327∗∗∗
(0.0193) (0.0194) (0.0195)
Variance parametersγ 0.8569 0.8546 0.8627σ2 0.1288 0.1288 0.1356σ2
v 0.1104 0.1100 0.1170σ2
u 0.0184 0.0187 0.0186
Number of observations 1924 1924 1924Number of banks 305 305 305Notes: marginal effects evaluated at variable means. Standard errors are given in parentheses. ∗, ∗∗, and∗∗∗ stand for 10%, 5%, and 1% significance levels, respectively. γ = σ2
vσ2 and σ2 = σ2
v + σ2u
Table 2.5. First-stage regression resultsProbit OLS
Inefficiency correlatesConstant -2.0711∗∗∗ -0.2107
(0.4415) (0.1520)Net interest margin -0.0415∗∗∗ -0.0150∗∗∗
(0.0082) (0.0028)Other operating income/total assets -0.0150∗∗∗ -0.0059∗∗∗
(0.0056) (0.0023)Net loans/total assets 0.0015 0.0005
(0.0018) (0.0006)Equity/total assets 0.0018 0.0004
(0.0027) (0.0010)InstrumentsCountry population / number of banks 1.0754∗∗∗ 0.3905∗∗∗
(0.1871) (0.0674)Country risk-free interest rate 0.0042 0.0016
(0.0029) (0.0011)Deposits / loans -0.0023 -0.0005
(0.0014) (0.0003)Assets / net interest revenue 0.0020∗∗ 0.0005∗∗
(0.0008) (0.0002)t 0.2716∗∗∗ 0.0912∗∗∗
(0.0874) (0.0304)12 t2 -0.0173∗∗ -0.0055∗
(0.0087) (0.0030)Notes: the dependent variable is f oreign ownership. Standard errors are given in parentheses. ∗, ∗∗, and ∗∗∗
stand for 10%, 5%, and 1% significance levels, respectively.
Foreign Ownership and Bank Efficiency: Does Sample Selection Matter? 35
Table 2.6. Tests of instrument validityRegression 1 Regression 2
Constant 1.0120∗∗∗ 0.9853∗∗∗
(0.0459) (0.0410)Foreign ownership 0.5299∗∗ 0.5071∗∗∗
(0.2128) (0.1026)Net interest margin -0.0116∗∗∗ -0.0120∗∗∗
(0.0035) (0.0026)Other operating income/total assets -0.0084∗∗∗ -0.0087∗∗∗
(0.0019) (0.0017)Net loans/total assets -0.0146∗∗∗ -0.0147∗∗∗
(0.0005) (0.0004)Equity/total assets 0.0019∗∗∗ 0.0018∗∗∗
(0.0007) (0.0007)InstrumentsCountry population / number of banks 0.0387
(0.1024)Country risk-free interest rate 3.7× 10−4
(8.0× 10−4)Deposits / loans 3.8× 10−4
(2.4× 10−4)Assets / net interest revenue 8.0× 10−5
(1.6× 10−4)t -0.0083
(0.0081)12 t2 −5.9× 10−4
(5.5× 10−4)Notes: the dependent variable is log(ine f f iciency). Standard errors are given in parentheses. ∗, ∗∗, and ∗∗∗
stand for 10%, 5%, and 1% significance levels, respectively.Regression 1: Excluded instruments for foreign ownership: Country risk-free interest rate, Assets / netinterest revenue, 1
2 t2.Regression 2: Excluded instruments for foreign ownership: Population / number of banks, Deposits / loans,t.
Chapter 3
Determinants ofCross-Border BankAcquisitions: The Role ofInstitutions
3.1 Introduction
During the last decade, foreign investors acquired many banks in former socialist
economies (FSEs). As a consequence, the share of foreign banks in the total assets
of the banking sector in these countries has increased substantially. In the Central
and Eastern European countries (CEEC), foreign bank presence has soared from
11% in 1995 to more than 75% in 2005 (EBRD, 2005). In contrast, cross-border
bank mergers and acquisitions in advanced economies are rare compared to domestic
takeovers (Buch and DeLong, 2004).
What makes banks in FSEs lucrative targets for foreign investors? In most of
the previous studies, cross-border bank acquisitions have been analyzed at the ag-
gregate (macro) level (see De Haan and Naaborg, 2004). Variables like geographical
distance, language, and cultural similarities with the home country, and regulatory
and supervisory structures are important determinants for the decision of foreign
38 Chapter 3
banks to enter a country (Berger et al., 2001). Also the level of economic develop-
ment of the host country seems to play a role in cross-border takeovers (Focarelli and
Pozzolo, 2001, Buch and DeLong, 2004). Banks located in countries with a stable
macroeconomic environment are more likely to be targeted by foreign investors than
those in countries with an unstable environment. For the FSEs, economic reforms
are also argued to affect the intensity of foreign bank entry (Lensink and De Haan,
2002).
More recent studies focus on the individual characteristics of target and acquir-
ing banks in FSEs. These micro-level studies show that characteristics of target
banks, including size, performance, and efficiency, are important variables predict-
ing the likelihood of a takeover (Bonin et al., 2005, Lanine and Vander Vennet,
2007, Williams and Liao, 2008). Claessens and van Horen (2008) report that banks
enter those countries where they have an institutional competitive advantage over
competitor banks.
Although it is now widely acknowledged that both country-level and bank-level
variables influence cross-border bank acquisitions, the importance of bank-level fac-
tors conditional on country-level determinants has not been treated systematically
in previous work.1 Such an analysis is especially important for the transition coun-
tries as they not only have diverse economic environments but they are also very
different with respect to institutions. Some of the transition countries have become
members of the European Union (EU) and have high economic growth rates, while
others have been less successful in their economic development. This implies that
1 Lensink et al. (2008) examine the impact of the quality of institutions on the foreign ownership-bank efficiency relationship for a broad sample of commercial banks in 105 countries. Another paperthat comes close to ours is the recent study by Claessens and van Horen (2008), who examine towhat extent institutional similarities between host and home country affect bank entry. In contrastto the present analysis, these papers do not focus on FSEs. They also do not examine whether theinfluence of bank-level factors is conditional on country-level determinants, which is the focus ofour analysis. Poghosyan and Poghosyan (2009) analyze post-entry performance of target banks andshow that foreign entry results in (delayed) efficiency improvement and decline in market power.However, they do not explore the role of target banks’ characteristics and institutional environmentof their host countries as determinants of cross-border acquisitions.
Determinants of Cross-Border Bank Acquisitions: The Role of Institutions 39
the impact of microeconomic characteristics of a domestic bank on the likelihood of
being taken over by a foreign bank may be subject to variation depending on the
characteristics of the host country.
In this chapter, we address this issue by using a multilevel mixed-effect logit
model for a sample of 2,175 observations from 11 transition countries over the pe-
riod 1992-2006. Altogether, 109 banks in our sample have been taken over. Our
estimations lend support to the view that the relative strength of microeconomic
factors determining cross-border bank takeovers varies across different groups of
countries. Hence, pooled estimates of the logistic model for all transition countries,
as used by, for instance, Lanine and Vander Vennet (2007), might provide mislead-
ing results. We find that foreign banks are targeting relatively large and efficient
banks in transition economies with weak institutions, thus providing support for the
market power hypothesis according to which banks are acquired with the objective
to increase market power of the acquiring bank. However, when entering more de-
veloped transition economies that have made progress in economic reform, foreign
banks acquire relatively less efficient banks, supporting the efficiency hypothesis ac-
cording to which banks are acquired with the objective of upgrading the efficiency
of the target bank.
The remainder of this chapter is structured as follows. Section 3.2 offers a the-
oretical background, while section 3.3 describes the empirical methodology and the
data used. Section 3.4 discusses the estimation results. The final section concludes.
3.2 Theoretical Background
The theoretical literature on the determinants cross-border bank takeovers has taken
a fairly eclectic approach (see Berger et al., 1999). A very common explanation is
that takeovers allow the consolidating banks to enhance their efficiency and prof-
itability, by exploiting economies of scale or scope and improving the efficiency of
40 Chapter 3
the consolidating banks. Alternatively, takeovers may enable the merged banks to
increase their market power. Lanine and Vander Vennet (2007) therefore distinguish
two competing hypotheses explaining cross-border bank acquisitions, namely the ef-
ficiency and the market power hypothesis. According to the efficiency hypothesis,
acquisitions are undertaken with the objective of upgrading the efficiency of the tar-
get banks. According to the market power hypothesis, acquisitions are used to gain
access to a market and build up market share without necessarily improving the
efficiency of the acquired banks. Their empirical results lend support to the market
power hypothesis. We build upon Lanine and Vander Vennet (2007) and examine
whether the impact of bank-level factors is conditional on institutional differences
between countries.2
There are various reasons to expect that a home country’s institutional setting
may affect a foreign bank’s strategy. It is widely believed that – at least at the be-
ginning of the transition – foreign banks have a competitive advantage compared to
domestic banks, as they have more advanced technologies, better corporate control,
higher educated employees, and better risk management instruments (De Haan and
Naaborg, 2004). However, domestic banks incur lower costs for providing services
at home, because they have better information about their country and customers.
Taking over a domestic bank and increasing its efficiency may therefore be a more
attractive entry strategy than a greenfield investment. However, improving the effi-
ciency of the target bank may be hampered by the institutions of the host country.
For instance, if regulations and legal frameworks are very detached from interna-
tional standards, it may be hard to introduce the risk management practices of the
foreign bank. To make the investment profitable, the foreign bank may in such
2 The study of Lanine and Vander Vennet (2007) differs in various ways from our study. Whereaswe focus on a sample of 11 CEEC countries over the period 1992-2006, Lanine and Vander Vennet’ssample covers only the period 1995-2002. Furthermore, Lanine and Vander Vennet measure cross-border deals using their announcement date, while our measure is based on the date when the dealwas completed. As it may take a while before the deal is settled and not all announced deals areeventually settled, we prefer this measure.
Determinants of Cross-Border Bank Acquisitions: The Role of Institutions 41
circumstances focus on increasing market power.
Similarly, Mian (2006) argues that a foreign bank in a distant economy faces extra
informational and agency costs in making relational loans. Likewise, Galindo et al.
(2003) point to the cost of learning that will also depend on distance. For instance,
learning how to work in a corrupt system can be costly for a banker whose lifetime
experience has been in Switzerland. Broadly speaking, distance here could reflect
a number of factors, including institutional distance between the foreign bank’s
country of origin and its subsidiary. The more the host country’s institutions are
similar to those of the home country, the lower these various costs will be and
therefore the more efficient the foreign bank can operate. Consistent with this
hypothesis, Lensink et al. (2008) report for a sample of 2095 commercial banks
in 105 countries that less institutional distance between the host and the home
country governance increases foreign bank efficiency. In case foreign banks cannot
realize efficiency gains due to a poor institutional framework in the host country
they may try to get compensation by acquiring market power.
When deciding on entering a transition country, a foreign bank arguably faces a
trade-off between expected return and its variability (Buch, 2000). As the growth
perspectives of transition countries are good and there may exist ample opportunities
for efficiency improvement of target banks, the latter may offer high rates of return.
At the same time, due to the transition process the variability of the rate of return
is likely to be higher than those of other investment opportunities. Arguably, it is
easier to achieve efficiency gains in host countries with better institutions (Berger
et al., 2001). Likewise, the more underdeveloped the host country’s institutions are,
the higher the volatility of expected returns will be, which needs to be compensated
for by higher returns. In case efficiency improvements are not sufficient, the extra
revenues needed to compensate for higher volatility may be acquired by increasing
market power.
42 Chapter 3
3.3 Methodology and Data
3.3.1 Multilevel mixed-effect logistic regression
We use a multilevel mixed-effect logit model (MMEL) to examine the impact of
bank-specific factors driving the cross-border bank takeovers in transition economies
conditional on their institutional characteristics.3 Like the logistic regression model
– used for studying cross-border bank acquisitions, among others, by Focarelli and
Pozzolo (2001), Focarelli and Pozzolo (2008), Focarelli et al. (2002), Lanine and
Vander Vennet (2007) – the multilevel mixed-effect modeling approach is based on
the principle of likelihood maximization. However, it is more general as it allows for
conditioning the impact of important acquisition determinants, such as efficiency
and market power, on institutional characteristics of host countries. In addition,
the MMEL nests simple logistic regression used in previous studies and provides a
flexible tool for testing the importance of institutional heterogeneity in host countries
for foreign bank entry by the means of the likelihood ratio test.
Our dependent variable (yit) is a dummy that takes the value of one at the time
when a cross-border bank acquisition was made. The general specification of the
MMEL model in log odd’s ratio form is:
log
(Pijt
1− Pijt
)= β0 + β1jt INEFFijt + β2jt MPijt + β3CONTROLSijt (3.1)
where Pijt = Prob(yijt = 1|INEFFijt, MPijt, CONTROLSijt) is the probability that
bank i located in country j will be acquired at time t conditional on a set of ex-
planatory variables, INEFF denotes the inefficiency of the target bank, MP denotes
the market power of the target bank, CONTROL is a vector of bank-specific and
3 A detailed description of the MMEL methodology is available in Rabe-Hesketh and Skrondal(2005). An alternative to the discrete choice modeling approach is an event-study methodologyused by Williams and Liao (2008), among others. Haselmann (2006) uses an alternative approach.He estimates a model for the lending behavior of banks to examine their strategy and concludes thatthe decision of foreign banks to enter the CEE economies seems to be driven by long-term strate-gic goals. This conclusion is based on the absence of a relationship between the macroeconomicconditions in the foreign banks’ country of origin and their loan supply.
Determinants of Cross-Border Bank Acquisitions: The Role of Institutions 43
country-specific control variables, and β’s are parameters to be estimated. In the
above specification, we relax the assumption that the impact of the target bank’s
inefficiency (β1ij) and market power (β2ij) on the likelihood of its acquisition by
foreign investors is constant across host countries and over time. More specifically,
we explicitly test for the possibility that the efficiency and market power hypothe-
ses differ depending on the institutional characteristics of host countries using the
following equations for the slope coefficients:
β1jt = β1 + β11 INSTjt + µj
β2jt = β2 + β22 INSTjt + ωj
(3.2)
where INSTjt is a variable measuring the quality of institutions in country j at
time t (increase in INST indicates better quality), and µj ∼ N(0, σµ) and ωj ∼
N(0, σω) are country-specific random effects that represent the combined effect of
all omitted country-specific determinants apart from institutional characteristics of
host countries that may influence the likelihood of foreign acquisition.
The simple logistic regression as used in previous studies is a special case of
specifications (3.1) and (3.2), when β11 = β22 = 0 and σµ = σω = 0. The latter
condition implies that the efficiency and market power hypotheses are invariant to
institutional characteristics of host countries and can be tested by the means of the
likelihood ratio test. In the presence of significant effects of quality of institutions
on the efficiency and market power hypotheses, the signs of the coefficients β11 and
β22 would indicate the direction of the impact. For example, when β11 (β22) is
positive and significant the efficiency (market power) hypothesis is more pertinent
to transition countries with better institutional quality.
3.3.2 Data
We obtained data from different sources to study cross-country bank takeovers in
transition economies. First, we obtained a list of takeovers during the 1992-2006
44 Chapter 3
period from the Securities Data Company (SDC) mergers and acquisitions database
produced by Thompson Financial. This data set contains information on the an-
nouncement and effective dates of the acquisition, the names of the bidder and
target banks, the country of their ultimate parents, and the percentage of shares
owned after the acquisition.4 From this data set, we selected completed acquisitions
that involve target banks in transition economies. In our analysis we only included
cross-border acquisitions (i.e., parents of bidder and target banks are from different
countries), which resulted in the control of ownership by the bidder bank exceeding
50% of the equity.
Second, we extracted bank level balance sheet and income statement information
from Bankscope that is maintained by Bureau van Dijk. We retrieved information
for all banks located in the 11 transition countries under research, including those
that were and those that were not engaged in a takeover (target and peer banks,
respectively). Our sample covers 388 banks and contains 2,175 observations. Al-
together, there have been 109 takeover events recorded. Table (3.1) provides the
distribution of these events across countries and over time.
Third, we used different sources to obtain information on institutional charac-
teristics of the countries in our sample. To proxy economic reform we use the first
principal component of various EBRD indicators of economic reform available for
the total sample period (referring to small- and large-scale privatization, enterprise
reforms, price liberalization, foreign exchange and trade liberalization, competition
policy, banking and non-banking sector reforms, reforms in infrastructure). This
indicator is available for our full sample period. To proxy the political regime of a
country we use the first principal component of the governance indicators of Kauf-
mann et al. (2007) that refer to different dimensions of the political system available
for the period 1996-2006 (voice and accountability, political stability and absence of
4 We thank Iman van Leyveld and Emilia Jurzyk for kindly sharing their data on bank ownership.
Determinants of Cross-Border Bank Acquisitions: The Role of Institutions 45
violence, government effectiveness, regulatory quality, rule of law, control of corrup-
tion).5
Finally, we obtained information on various macroeconomic indicators and finan-
cial market conditions as additional control variables using the World Bank’s Word
Development Indicators. Table (3.2) contains details of the data sets employed in
our analysis.
We improve upon Lanine and Vander Vennet (2007) by utilizing direct measures
of bank market power and efficiency.6 For this purpose, we use the stochastic fron-
tier methodology, according to which the efficiency of individual banks is identified
by benchmarking their performance against a common frontier determined by the
best-performing banks in the sample. We utilize the time-varying bank-specific in-
efficiency scores (INEFF) instead of the proxies employed by Lanine and Vander
Vennet to test for the efficiency hypothesis (see Appendix 1 for further details).
Unlike the cost-to-income ratio, the inefficiency score provides a direct measure of
relative performance of the particular bank in comparison to similar banks. In par-
ticular, it compares the actual level of bank cost to its optimal level (cost frontier)
given the volume of output produced and input prices. Furthermore, we calculate
Lerner’s indices using cost function estimates obtained from the stochastic frontier
model as indicators of bank market power (see Appendix 2 for further details). In
addition to efficiency and market power, we augment the specification by various
5 As we use generated regressors, the standard errors of the estimated coefficients may be affected,although the consistency of the obtained coefficients is preserved. To check whether this generatedregressor problem affects our results about the impact of institutions, we have re-estimated themodel but instead of using the first principal component of the institutional indices we used theiraverage values. The estimation results (available on request) suggest that our qualitative findingsdo not change.6 Lanine and Vander Vennet (2007) use three indicators of market power of a bank (i.e., the
logarithm of a bank’s total assets, and its share of loans and deposits of all banks) and two indicatorsof efficiency (i.e., the cost-to-income ratio, and the non-interest expense ratio). However, thesemeasures do not allow for a direct measurement of market power and efficiency and cannot becompared across countries. For instance, since the financial sector in Poland is much larger thanthe financial sector in Estonia, the market share of banks in Poland tends to be smaller than thatof banks in Estonia. Likewise, cost ratios don’t take the position of a bank in comparison to similarbanks into account.
46 Chapter 3
bank-specific (capitalization, return on equity, loans to assets, and deposits to assets
ratios) and country-specific (real GDP growth, per capita GDP, share of private
sector, and ratio of credit to GDP) control variables.
Table (3.3) provides details of the variables used in our analysis, while Table
(3.4) displays descriptive statistics.
3.4 Empirical Results
3.4.1 Do institutions matter?
The first step in our empirical investigation is to estimate the logistic regression
model of Lanine and Vander Vennet (2007) using a more general mixed-effect formu-
lation (3.1)-(3.2). As the simple logistic regression is equivalent to the mixed-effect
logistic regression with the slope coefficients restricted to be constant (β1jt = β1
and β2jt = β2), this exercise allows us to test whether by conditioning the variation
in slope coefficients on institutional developments in host countries we are able to
improve the fit of the model. Note that the regressions in which the EBRD indicator
proxies institutions refer to the period 1992-2006, while the regressions in which the
Kaufman indicator is used refer to the period 1996-2006 as this indicator is only
available for those years.
We start by estimating the model (3.1)-(3.2) separately for each measure of
institutional development using only bank-specific variables. The fit of each model
is compared to the simple logistic model (with constant slopes β1 and β2) using the
likelihood ratio test. The results as reported in Table (3.5) suggest that the MMEL
model outperforms the simple logistic regression. In economic terms, this finding
implies that the relative strength of the efficiency and market power hypotheses
varies across countries and over time, depending on the dynamics of institutional
development of host countries.
In both specifications, we obtain negative and significant coefficients β1 and posi-
Determinants of Cross-Border Bank Acquisitions: The Role of Institutions 47
tive and significant coefficients β2. This suggests that, taken institutional character-
istics as given, there is significant evidence supporting the market power hypothesis
and rejecting the efficiency hypothesis. In other words, if foreign banks can choose
between two banks located in two countries having a comparable level of institutional
development, they acquire a bank that has larger market power and is more efficient.
The former result is in line with the findings of Lanine and Vander Vennet (2007).
However, the positive and significant β11 coefficients obtained in both models suggest
that support for the market power hypothesis weakens as the level of institutional
development of the host country increases. Similarly, the negative and significant
β22 coefficients obtained in both models suggest that the efficiency hypothesis finds
greater support with the improvement of the institutional development of the host
country.
The economic effect of market power as a determinant of cross-border acquisi-
tions is, on average, less sizable compared to bank efficiency. Moreover, the impact
of institutional development on the odds of acquisition is also mostly channeled
through its impact on the likelihood of targeting inefficient banks. Comparison of
Hungary and Romania as countries with highest and lowest average level of institu-
tional development according to the EBRD index helps to illustrate this point. Our
estimations suggest that the likelihood of acquisition of an inefficient bank in Hun-
gary is 14.09% larger than in Romania, while the likelihood of acquisition of a bank
possessing large market power in Hungary is only 0.85% lower than in Romania.
Similarly, comparison of Slovenia and Romania as countries with highest and lowest
average level of institutional development according to the Kaufman index suggests
that the likelihood of acquisition of an inefficient bank in Slovenia is 15.63% larger
than in Romania, while the likelihood of acquisition of a bank possessing large mar-
ket power in Slovenia is only 2.24% lower than in Romania. These results suggest
that foreign investors put large weight on the level of host countries’ institutional
48 Chapter 3
development when acquiring banks with the purpose of upgrading their efficiency.
Among the bank-specific control variables, we find that foreign banks target
better-capitalized banks and banks with greater deposit-funding capacity, while the
impact of profitability is not significant.
To summarize, our results suggest that the quality of the institutions of the
host country matters for the acquisition strategy of foreign banks. The better the
institutions of the host country, the more (less) support there is for the efficiency
hypothesis (market power hypothesis). In the next subsection we will check the
robustness of our results by introducing time fixed effects and macroeconomic control
variables.
3.4.2 Sensitivity analysis
We estimated two additional models to check the sensitivity of our results. First,
we introduced time dummies to control for time-specific common shocks that might
have influenced foreign banks to enter transition economies. Second, we introduced
country-specific macroeconomic control variables relevant for the decision of foreign
banks to go abroad, such as per capita GDP, real GDP growth, share of private sector
in the economy, and share of private credit in GDP. Estimation results for these two
sensitivity checks are reported in Table (3.6). The estimation results in both cases
are qualitatively similar to our earlier results concerning the efficiency and market
power hypothesis testing. The coefficients β2 (β1) and β11 (β22) remain positive
(negative). The impact of bank-specific control variables is also broadly consistent
with previous results. Among the macroeconomic variables, only the private sector
share in GDP has a significant positive effect on the decision of foreign banks to
enter transition countries.
To summarize, this section shows that previous results on the importance of
institutional development for the decision of foreign banks to go abroad holds when
controlling for the impact of other macroeconomic variables and time effects.
Determinants of Cross-Border Bank Acquisitions: The Role of Institutions 49
3.4.3 Analyzing the efficiency and market power hypothesesacross countries and over time
After confirming the importance of the institutional environment for foreign bank
entry, we finally analyze the magnitude of variation of coefficients measuring the
market power and efficiency hypotheses (β1jt and β2jt) across countries and over
time. For this purpose, we use the Bayesian shrinkage estimator (Rabe-Hesketh
and Skrondal, 2005) to obtain estimates of β1jt and β2jt, and calculate their aver-
age values across countries (E[β1j] = ∑tβ1jtT and E[β2j] = ∑t
β2jtT ) and over time
(E[β1t] = ∑jβ1jt
J and E[β2t] = ∑jβ2jt
J ). Figures (3.1) and (3.2) show obtained es-
timates for models with EBRD and Kaufman indices as measures of institutional
quality, respectively.
Examination of these figures provides several useful insights. First, in all cases
we find support for the market power hypothesis, since average values of coefficients
β2jt are always positive. This is also in line with our previous discussion on the
economic significance of market power as a determinant of cross-border acquisitions.
Second, cross-country variation of average coefficients implies that the efficiency
hypothesis is largely supported (positive average values of β1jt) for relatively more
developed countries, such as the Czech Republic, Hungary and Poland for the case
of model 1 and also Estonia, Slovenia and Slovakia for the case of model 2. Third,
the time dynamics of the coefficients suggests that the relative importance of the
market power and efficiency hypotheses has been changing over time. During the
1990s, foreign banks were targeting largely efficient banks (rejection of the efficiency
hypothesis) and banks having greater market power (support for the market power
hypothesis). In more recent times, perhaps due to the fact that the cream has been
already skimmed, foreign banks started targeting inefficient banks and banks with
relatively lower market power.
50 Chapter 3
3.5 Conclusions
We analyze the microeconomic determinants of cross-border bank acquisitions in 11
transition economies over the period 1992-2006. By using a multilevel mixed-effect
logit model we explicitly incorporate the macro-economic and institutional hetero-
geneity of the transition economies into our analysis. We find that foreign banks
are targeting relatively large and efficient banks in transition economies with weak
institutions, thus providing support for the market power hypothesis according to
which banks are acquired with the objective to increase market power of the acquir-
ing bank. However, when entering transition economies that have made progress
in economic and institutional reform, foreign banks acquire relatively less efficient
banks, supporting the efficiency hypothesis according to which banks are acquired
with the objective of upgrading the efficiency of the target bank.
Our findings suggest that the concerns of Lanine and Vander Vennet (2007)
regarding the limitations with respect to the commonly accepted view that foreign
entry will contribute to the competitiveness and efficiency of banking systems in
transition are only partially justified. We show that these concerns are not valid for
a small subsample of target banks located in transition economies that have made
significant progress in terms of institutional development and the restructuring of
their economies. Foreign investors enter these countries with the aim of upgrading
the efficiency of the acquired bank and utilizing the unexploited profit opportunities.
In contrast, foreign investors seem to be hesitant in entering transition countries
lagging behind in terms of economic reforms.
Our analysis also suggests that the relative importance of the market power
and efficiency hypotheses has been changing over time. During the 1990s, foreign
banks were targeting largely efficient banks and banks having greater market power.
In more recent times, perhaps due to the fact that the cream has been already
skimmed, foreign banks started targeting inefficient banks and banks with relatively
lower market power.
Determinants of Cross-Border Bank Acquisitions: The Role of Institutions 51
Appendix 1
Obtaining individual bank cost efficiency scores using the stochas-tic efficiency frontier model
Following a recent stream of the literature (e.g., Bonin et al., 2005, Fries and Taci,
2005, Poghosyan and Borovicka, 2007), we apply frontier analysis for modeling cost
efficiency of banks in FSEs. For the stochastic cost frontier, we follow the modified
production approach (see Berger and Humphrey, 1991) and use two types of bank
outputs: total loans (y1,it) and total deposits (y2,it). The banks provide their services
using two inputs, i.e., physical capital and labor. Accordingly, the price of physical
capital is measured as a ratio of non-interest expenses to total assets (w1,it), while the
price of labor is proxied by the ratio of total personnel expenses to total assets (w2,it).
The production technology might also be influenced by the technological progress,
for which we control by using a time trend (t). The dependent variable in the frontier
is the total cost of a bank (cit), which includes both interest and operating expenses.
To account for the country-specific environmental characteristics that might have an
impact on the bank’s technology, we augment the frontier by introducing real GDP
growth (GDP_GR), real GDP per capita in US dollars (GDP_PC), and the share
of domestic credit in GDP (CRED) variables. The final translog specification for
the cost function takes the following form:
ln citwit,1
= α +S∑
s=2βs ln wit,2
wit,1+
L∑
l=2γl ln yit,l + 1
2
S∑
s=2
S∑
l=2δsl ln wit,s
wit,1ln wit,l
wit,1+
+ 12
L∑
s=1
L∑
l=1ϕsl lnyit,s ln yit,l + 1
2
S∑
s=2
L∑
l=1θsl ln wit,s
wit,1ln yit,l + ρ1t + 1
2 ρ2t2+
+S∑
s=2ρw
s t ln wit,swit,1
+L∑
l=1ρ
ys ln yit,l + ψ1GDP_GR + ψ2GDP_PC+
+ψ3CRED + νit + uit
(3.3)
where i and t are bank and time indices, respectively. The linear homogeneity
restrictions are satisfied by expressing all variables in terms of a ratio with respect
52 Chapter 3
to one of the input prices, and inefficiency is modeled as a function of time using
the specification of Battese and Coelli (1992):
uit = uη(t−T)i (3.4)
where ui is the bank-specific inefficiency term that is assumed to have a non-negative
truncated normal distribution with zero mean and variance σ2u, and T is the last pe-
riod in the sample. The overall inefficiency of each individual bank, uit, is varying
over time at the exponential rate η to be estimated. The intuition behind this param-
eterization is that the inefficiency term is assumed to be monotonically increasing
(positive and significant η), monotonically decreasing (negative and significant η)
or neutral (insignificant η) over time. To estimate the model using a maximum
likelihood method we additionally assume that the random error term, vit, follows
a normal distribution with zero mean and constant variance, σ2v .
Appendix 2
Obtaining Lerner’s indices as measures of banks’ market power
Following Angelini and Cetorelli (2003) and Maudos and Fernandez de Guevara
(2007), we estimate Lerner’s index to assess the competitive behavior of individual
banks as follows:
MPijt =ARijt − MCijt
ARijt(3.5)
where ARijt is the ratio of total operating income to total earning assets as a proxy
for average price of bank products, and MCijt is the marginal cost of banks obtained
by differentiating the cost function estimate (3.3) with respect to bank outputs. In
fully competitive markets, marginal costs of banks equal their marginal revenues and
Lerner’s index is approaching zero. Therefore, larger values of the Lerner’s index
indicate larger market power possessed by individual banks.
Determinants of Cross-Border Bank Acquisitions: The Role of Institutions 53
Table 3.1. Cross-border bank acquisitions in FSEs, 1992-20061992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 Total
BG 0 0 0 0 0 0 0 0 0 0 2 1 0 1 1 5CZ 0 2 0 0 2 2 2 2 2 1 0 1 0 0 0 14EE 0 0 0 0 0 0 1 1 1 0 0 0 1 0 0 4HR 0 0 0 0 0 3 1 1 2 1 0 0 0 1 1 10HU 0 0 0 0 4 1 1 0 1 0 2 0 0 0 0 9LT 0 0 0 0 0 0 0 0 2 1 1 0 0 0 0 4LV 0 0 1 0 0 1 1 0 1 3 0 0 0 0 1 8PL 0 1 0 1 3 2 1 6 4 2 1 0 0 0 0 21RO 0 0 0 0 0 2 3 1 2 1 3 3 0 0 0 15SI 0 0 0 1 0 0 0 0 0 2 2 0 0 0 0 5SK 1 0 0 2 0 2 1 0 3 3 2 0 0 0 0 14Total 1 3 1 4 9 13 11 11 18 14 13 5 1 2 3 109
Notes: BG=Bulgaria, CZ=The Czech Republic, EE=Estonia, HR=Croatia, HU=Hungary, LT=Latvia, LV=Lithuania, PL =Poland, RO = Romania, RS = Serbia, SI=Slovenia, SK=Slovakia.
Table 3.2. Data sourcesVariable Definition SourceCross-border bank acquisi-tion
A dummy variable changing its valuefrom 0 to 1 at the time when the acqui-sition took place.
Thompson Financial
Bank financial indicators Balance sheet items and income state-ments
Bankscope of Bureau van Dijk
Reforms Indices ranging from 1 (worst) to 4(best) and indicating the progress of re-forms in the following nine areas: small-and large-scale privatization, enterprisereforms, price liberalization, forex andtrade liberalization, competition policy,banking and non-banking sector reforms,reforms in infrastructure.
EBRD Transition Reports
Governance Indices ranging from -2.5 (worst) to 2.5(best) and indicating the progress of gov-ernance in following six areas: voice andaccountability, political stability and ab-sence of violence, government effective-ness, regulatory quality, rule of law, con-trol of corruption.
Kaufman et al. (2007)
Macro data Real GDP growth, GDP per capita (real,USD), share of private sector in GDP,and ratio of domestic credit to GDP.
World Bank World DevelopmentIndicators, EBRD
54 Chapter 3
Table 3.3. Data descriptionVariable DescriptionEfficiencyINEFF Cost inefficiency of banks obtained using the stochastic efficiency
frontier model. Larger values indicate greater inefficiency.Market powerMP Lerner’s index, calculated as difference between average revenues
and marginal costs divided over average revenues. Larger valuesindicate greater market power.
Bank-specific control variablesCAP Capital adequacy ratio, calculated as ratio of bank equity to total
assetsROE Return on equity, calculated as the ratio of pre-tax profits and
total equityLTA Intensity of loan provision, calculated as the ratio of loans to total
assetsDEP Deposit funding, calculated as the ratio of total deposits to total
assetsCountry-specific control variablesGDPGR Real GDP growthGDPPC GDP per capita (in thousands of USD)PRIV Private sector share in the economyCREDGDP Share of credit to the private sector in GDPInstitutional measuresEBRD First principal component of nine EBRD indices measuring re-
forms in various sectors in the economyKAUF First principal component of six Kaufman indices measuring gov-
ernance
Table 3.4. Descriptive statisticsMean Median St. Dev. Min Max Skewness Kurtosis
INEFF 0.6776 0.6685 0.1537 0.2142 0.9824 -0.0167 2.2100MP 30.0325 30.1135 0.6703 17.9268 30.7580 -9.9603 138.2745CAP 0.1455 0.1079 0.1244 0.0435 0.9692 3.3181 16.7740ROE 0.1062 0.1164 0.2811 -5.2137 1.1120 -6.6187 96.9797LTA 0.4519 0.4623 0.1789 0.0000 0.9724 -0.2011 2.9898DEP 0.7377 0.7873 0.1673 0.0018 0.9503 -1.9444 7.2768GDPGR 4.3063 4.5240 2.8109 -16.2270 12.2350 -1.1232 7.7949GDPPC 4392.0 4066.0 1953.9 1615.9 12340.8 1.5884 6.1060PRIV 0.6695 0.6500 0.0967 0.3000 0.8000 -0.6627 3.4584CREDGDP 0.3112 0.2920 0.1396 0.0430 0.7790 0.5151 3.0021EBRD 7.7912 7.8204 1.0267 3.9821 10.1288 -0.2397 2.9952KAUF 7.1359 7.5005 1.1366 4.5462 9.0207 -0.4662 2.1817
Determinants of Cross-Border Bank Acquisitions: The Role of Institutions 55
Tabl
e3.
5.Es
timat
esof
equa
tions
(1)
and
(2) W
ith
out
inst
itu
tion
sW
ith
EB
RD
ind
exW
ith
Kau
fman
ind
exC
oeffi
cien
tsM
argi
nal
effec
tsC
oeffi
cien
tsM
argi
nal
effec
tsC
oeffi
cien
tsM
argi
nal
effec
tsβ
0-5
.148
3***
–-5
.166
1***
–-4
.888
2***
–(1
.086
0)–
(1.0
880)
–(1
.133
0)–
β1
-0.5
617
-0.0
253
-11.
0603
**-0
.465
3**
-7.5
442*
-0.3
419
(0.6
508)
(0.0
294)
(5.1
110)
(0.2
262)
(4.5
110)
(0.2
088)
β2
0.32
24**
0.01
40**
*0.
9601
*0.
0415
*1.
4442
**0.
0659
**(0
.098
9)(0
.003
5)(0
.506
3)(0
.022
3)(0
.628
6)(0
.028
9)β
11–
–1.
3651
**0.
0575
**1.
0783
*0.
0489
*–
–(0
.654
9)(0
.029
2)(0
.627
4)(0
.029
2)β
22–
–-0
.084
4*-0
.003
6*-0
.155
0*-0
.007
0*–
–(0
.043
1)(0
.001
6)(0
.081
0)(0
.003
7)C
apit
alad
equa
cyra
tio
4.78
20**
*0.
2203
***
4.90
82**
*0.
2256
***
4.92
31**
*0.
2348
***
(1.2
040)
(0.0
553)
(1.2
190)
(0.0
559)
(1.2
900)
(0.0
616)
Ret
urn
oneq
uity
0.00
630.
0002
0.00
520.
0003
0.00
40.
0002
(0.0
336)
(0.0
013)
(0.0
304)
(0.0
016)
(0.0
263)
(0.0
013)
Inte
nsit
yof
loan
prov
isio
n-1
.014
2*-0
.049
4*-0
.956
1-0
.047
7*-1
.121
0*-0
.055
3*(0
.584
8)(0
.026
2)(0
.585
5)(0
.026
1)(0
.622
1)(0
.029
2)D
epos
itfu
ndin
g2.
4871
**0.
1179
**2.
4514
**0.
1198
**2.
2120
*-0
.108
9*(1
.153
0)(0
.051
9)(1
.157
0)(0
.052
2)(1
.190
0)(0
.005
6)S
tati
stic
sN
umbe
rof
obse
rvat
ions
2175
.021
75.0
1891
.0L
R-t
est
(p-v
alue
)–
0.06
420.
0648
Not
es:
Sta
nd
ard
erro
rsar
ere
por
ted
inb
rack
ets.
***,
**,
and
*d
enot
esi
gnifi
can
ceat
10,
5,an
d1%
con
fid
ence
leve
ls,
resp
ecti
vely
.T
he
rep
orte
dm
argi
nal
effec
tsar
eev
alu
ated
atsa
mp
lem
ean
s.
56 Chapter 3
Tabl
e3.
6.Se
nsiti
vity
anal
ysis:
Tim
edu
mm
ies
and
mac
rova
riabl
esad
ded
Wit
hm
acro
econ
omic
cont
rol
vari
able
sW
ith
tim
efi
xed
effec
tsW
ith
out
inst
itu
tion
sW
ith
EB
RD
Wit
hK
aufm
anW
ith
out
inst
itu
tion
sW
ith
EB
RD
Wit
hK
aufm
anin
dex
ind
exin
dex
ind
exβ
0-5
.617
0-7
.156
0*-6
.934
0*-4
.788
1**
-4.4
015*
*-4
.825
6***
(3.6
830)
(3.7
710)
(3.9
001)
(1.5
122)
(1.5
259)
(1.3
136)
β1
-0.5
811
-11.
3401
**-8
.630
4*-0
.615
0-8
.883
9**
-2.8
334*
(0.6
544)
(5.0
990)
(4.7
500)
(0.6
903)
(4.1
579)
(1.4
779)
β2
.318
1**
1.04
40**
1.62
70**
0.34
15**
1.07
32**
0.80
86**
(0.1
045)
(0.5
171)
(0.6
554)
(0.1
159)
(0.4
515)
(0.2
796)
β11
–1.
3850
**1.
2250
*–
1.34
89**
0.58
40*
–(0
.647
8)(0
.654
7)–
(0.6
637)
(0.3
468)
β22
–-0
.095
4*-.
1787
**–
-0.1
194*
-0.0
921*
*–
(0.0
481)
(0.0
837)
–(0
.069
2)(0
.045
3)C
apit
alad
equa
cyra
tio
4.79
81**
*4.
9683
***
4.92
12**
*4.
9622
***
5.08
90**
*4.
9624
***
(1.1
920)
(1.2
060)
(1.2
780)
(1.2
423)
(1.2
585)
(1.3
203)
Ret
urn
oneq
uity
0.00
630.
0050
0.00
390.
0039
0.00
370.
0034
(0.0
311)
(0.0
274)
(0.0
237)
(0.0
230)
(0.0
225)
(0.0
220)
Inte
nsit
yof
loan
prov
isio
n-0
.857
1-0
.853
4-0
.904
6-0
.939
3-0
.901
8-0
.865
3(0
.577
0)(0
.575
4)(0
.617
8)(0
.618
0)(0
.618
7)(0
.651
9)D
epos
itfu
ndin
g2.
3302
**2.
3542
**1.
9714
*2.
5366
**2.
4824
**2.
2729
*(1
.112
0)(1
.117
0)(1
.150
0)(1
.184
5)(1
.191
6)(1
.225
4)R
eal
GD
Pgr
owth
-0.0
3582
0.00
593
0.01
166
––
–(0
.125
8)(0
.127
0)(0
.132
1)–
––
GD
Ppe
rca
pita
-0.2
335
-0.1
147
-0.1
002
––
–(0
.369
0)(0
.372
7)(0
.385
2)–
––
Pri
vate
sect
orsh
are
4.25
23**
4.87
11**
4.99
62**
––
–(1
.936
0)(1
.959
0)(2
.061
0)–
––
Pri
vate
sect
orcr
edit
-0.8
897
-1.1
680
-1.5
270
––
–(1
.314
0)(1
.320
0)(1
.386
0)–
––
Sta
tist
ics
Num
ber
ofob
serv
atio
ns21
7521
7518
9121
7521
7518
91L
R-t
est
(p-v
alue
)0.
0507
0.03
910.
0558
0.06
61N
otes
:St
anda
rder
rors
are
repo
rted
inbr
acke
ts.
***,
**,
and
*de
note
sign
ifica
nce
at10
,5,
and
1%co
nfide
nce
leve
ls,
resp
ecti
vely
.C
oeffi
cien
tson
tim
edu
mm
ies
are
omit
ted
toco
nser
vesp
ace.
Determinants of Cross-Border Bank Acquisitions: The Role of Institutions 57
-2 -1 0 1 2Random slope for inefficiency
SK
SI
RO
PL
LV
LT
HU
HR
EE
CZ
BG
0 .1 .2 .3 .4Random slope for market power
SK
SI
RO
PL
LV
LT
HU
HR
EE
CZ
BG
-4 -3 -2 -1 0 1Random slope for inefficiency
2006
2005
2004
2003
2002
2001
2000
1999
1998
1997
1996
1995
1994
1993
1992
0 .2 .4 .6Random slope for market power
2006
2005
2004
2003
2002
2001
2000
1999
1998
1997
1996
1995
1994
1993
1992
Figure 3.1. Model 1: Average values of inefficiency (β1jt) and market power (β2jt)coefficients across countries and over time
Notes: BG - Bulgaria, CZ - Czech Republic, EE - Estonia, HR - Croatia, HU - Hungary, LT - Lithuania,LV - Latvia, PL - Poland, RO - Romania, SI - Slovenia, SK - Slovakia.
58 Chapter 3
-2 -1 0 1 2Random slope for inefficiency
SK
SI
RO
PL
LV
LT
HU
HR
EE
CZ
BG
0 .2 .4 .6 .8Random slope for market power
SK
SI
RO
PL
LV
LT
HU
HR
EE
CZ
BG
-.2 0 .2 .4 .6Random slope for inefficiency
2006
2005
2004
2003
2002
2001
2000
1999
1998
1997
1996
1995
1994
1993
1992
0 .1 .2 .3 .4Random slope for market power
2006
2005
2004
2003
2002
2001
2000
1999
1998
1997
1996
1995
1994
1993
1992
Figure 3.2. Model 2: Average values of inefficiency (β1jt) and market power (β2jt)coefficients across countries and over time
Notes: BG - Bulgaria, CZ - Czech Republic, EE - Estonia, HR - Croatia, HU - Hungary, LT - Lithuania,LV - Latvia, PL - Poland, RO - Romania, SI - Slovenia, SK - Slovakia.
Chapter 4
Heterogeneity ofTechnological Regimes andBank Efficiency
4.1 Introduction
Given the key role of banks as financial intermediaries in the process of transforma-
tion from a plan to a market economy, empirical assessment of efficiency of banking
institutions in former socialist economies (FSE) has been given considerable at-
tention in the recent empirical literature. Table 4.1 provides a brief overview of
these studies, which share several common features. First, all of them are based on
the efficiency frontier methodology, according to which each bank’s performance is
benchmarked against a frontier reflecting the characteristics of the best-performing
banks in the sample.1 Most of the studies employ stochastic frontier analysis (SFA),
a parametric method that is less sensitive to the measurement errors in the sample,
relative to the alternative non-parametric method, the data envelopment analysis
1 Coelli et al. (2005) contains a textbook exposition of the efficiency frontier methodology. Bergerand Mester (1997) and Hughes and Mester (2008) review applications of these methods in thebanking industry.
60 Chapter 4
(DEA). Next, efficiency analysis is conducted for two important measures of bank
performance: costs and profits. In both cases, the variables determining technology
of banks include the amount of outputs (such as loans, investments, other earning
assets) and the level of input prices (such as cost of capital, labor, financial funds).2
Finally, all studies assume that banks share a common production technology. In
other words, production capacity of all banks are described by an identical produc-
tion possibility frontier.
The aim of this chapter is to relax the latter restrictive assumption by allowing
for multiple technology regimes, conditional on differences in economic environments
in which banks operate. The main criticism of the homogenous technological regime
assumption adopted by all studies reviewed in Table 4.1 is the potential bias in the
frontier estimates and, thus, the obtained efficiency scores (Orea and Kumbhakar,
2004). Specifically, if the true technology is heterogenous, then the omitted techno-
logical differences might be inappropriately labeled as inefficiency in single-frontier
estimations. Consequently, the measures of the impact of inefficiency determinants
will be affected. Another drawback of the homogenous technological regime assump-
tion is that it imposes restrictions on certain important characteristics of banking
activity, such as technical progress and scale economies.
There are several approaches how one can deal with the impact of technologi-
cal differences. One approach is to include country-specific environmental variables
that are likely to influence technologies of banks, such as the level of economic de-
velopment and institutional background, as additional explanatory variables in the
frontier (Berger, 2007). In fact, most of the cross-country studies reviewed in Table
4.1 augment the frontier by country-specific variables (Fries and Taci, 2005, Bonin
2 In most studies, the theoretical foundation for the choice of frontier determinants is either theintermediation approach (Sealey and Lindley, 1977) or the modified production approach (Bergerand Humphrey, 1991).
Heterogeneity of Technological Regimes and Bank Efficiency 61
et al., 2005, Yildirim and Philippatos, 2007, Poghosyan and Borovicka, 2007, Green
et al., 2007). The main disadvantage of this approach is that the introduction of
the environmental variables only affects the intercept of the frontier specification,
leaving the slope parameters unaffected (Bos and Schmiedel, 2007). Thus, although
more flexibility in intercepts may partially alleviate the bias in inefficiency estimates
(Valverde et al., 2007), the constancy of the slope parameters will still impose re-
strictions on technical progress and scale economies of banks. Another drawback of
this approach is that technological differences are assumed to be country-specific,
which rules out the possibility that banks located within the same country may
employ different business models.
An alternative approach to alleviate the impact of technological differences is
a priori sample restriction. The sample restriction can be based, for instance, on
the organizational structure of banks (Mester, 1993, Altunbas et al., 2001), or their
geographical location (Mester, 1996, Bos and Schmiedel, 2007). The main disadvan-
tage of this approach is that a priori restriction of sample groups is to some extent
arbitrary. For instance, Koetter and Poghosyan (2009) show that even banks having
similar organizational structure can operate under different technological regimes.
In this study, we account for differences in technological regimes using a latent
class stochastic frontier analysis (LCSFA), which addresses the disadvantages asso-
ciated with the aforementioned alternative approaches (Orea and Kumbhakar, 2004,
Greene, 2005).3 Unlike the first approach, the impact of the environmental factors is
not only reflected in the magnitude of the intercepts, but also affects the slope coef-
ficients. Here, the environmental variables enter as latent class determinants rather
than as a part of the frontier and thus influence both estimates of the technological
regime of banks and their efficiency simultaneously. Unlike the second approach, the
3 To our best knowledge, this is the first application of the LCSFA for studying efficiency of banksin FSE.
62 Chapter 4
latent class method does not require a priori grouping of banks. Instead, it utilizes
all information available in the sample and identifies separate technological regimes
based on the maximum likelihood principle.
Our estimations suggest that banks in FSE operate under three distinct techno-
logical regimes. Not only do we observe technological differences between new EU
member FSE countries and the rest, but also technological regimes differ within the
new EU members. Differences in technological regimes also have implications for
the impact of foreign bank participation on bank efficiency. In line with the find-
ings in Chapter 3, we show that foreign bank participation improves efficiency of
banks located in the new members of European Union, with a relatively high level
of economic development, while the impact of foreign ownership on banks in less
developed CIS countries is ambiguous.
The remainder of this chapter is structured as follows. The next section presents
the LCSFA model and estimation details. A data description is provided in section
3, while the estimation results are reviewed in section 4. The last section concludes.
4.2 Accounting for Heterogeneity of Banking Tech-nologies: A Latent Class Stochastic FrontierModel
In our LCSFA model, we assume that the technology is represented by a cost function
in the translog form. Following Orea and Kumbhakar (2004), the translog cost
function for class j may be written as:
ln Cit = ln C(yit, wit, t; βk) + uit|k + vit|k, (4.1)
Heterogeneity of Technological Regimes and Bank Efficiency 63
where subscripts i = 1, ..., N, t = 1, ..., Ti, and k = 1, ..., K stand for bank, time,
and class, respectively; Cit is individual bank total cost; yit and wit indicate vectors
of outputs and input prices; and βk is a class-specific vector of parameters to be
estimated. The two-sided random error term vit|k is assumed to be independent of
the non-negative cost inefficiency variable uit|k for each class.
To estimate the model using maximum likelihood we assume that the random er-
ror term for class k (vit|k) follows a normal distribution with zero mean and constant
variance, σ2vk. In addition, the inefficiency term for class k (uit|k) is assumed to be a
product of a time-invariant individual bank effect, ui|k and a parametric function of
time and other explanatory variables (inefficiency determinants), λit. The ui|k term
is assumed to have a non-negative truncated normal distribution with zero mean
and variance, σ2uk.
Similarly to Orea and Kumbhakar (2004), we specify the inefficiency variable
uit|k in general form as:
uit|k = λit(ηk)ui|k = exp(z′itηk)ui|k, (4.2)
where ui|k ≥ 0; ηk = (η1k, ..., ηHk)′ is a H × 1 vector of parameters and zit =
(z1it, ..., zHit)′ is a H × 1 vector of inefficiency determinants, including the Battese
and Coelli (1992) trend specification: zit = (T − t), where T = max(Ti) is the final
time period in the panel.
The conditional likelihood function for bank i at time t can be written (see
Greene, 2005) as:
ln LFit(θk) =
Φ
(−εit|k
σuk
σvk
√σ2
vk+σ2uk
)Φ(0)
1√σ2
vk + σ2uk
φ
εit|k√σ2
vk + σ2uk
, (4.3)
where εit|k = uit|k + vit|k is the compounded disturbance term; θk = (βk, σ2vk, σ2
uk, ηk)
64 Chapter 4
are parameters describing the technology of banks belonging to class k; Φ(.) and
φ(.) are standard normal cumulative and density functions, respectively. Following
Greene (2005), we assume that bank observations are independent over time, thus
the overall contribution of bank i to the conditional likelihood can be derived using
a product of likelihood functions: LFik(θk) =Ti∏
t=1LFit(θk).4
The unconditional likelihood of bank i is obtained as a weighted sum of the k-class
likelihood functions. The weights are the class membership probabilities reflecting
the uncertainty regarding the true membership in the sample. A convenient way to
parameterize the class probabilities is to employ a multinomial logit model:
Pik(δk) =exp(δ′kqi)
K∑
k=1exp(δ′kqi)
, (4.4)
where k = 1, ..., K denote classes; δK = 0 is a parameter normalization for the refer-
ence class and qi is a vector of bank-specific and time-invariant class determinants.
Using weights Pik from equation (4.4), the unconditional likelihood for bank i can
be written as:
LFi(θ, δ) =K
∑k=1
LFik(θk)Pik(δk), (4.5)
where 0 ≤ Pik ≤ 1 andK∑
k=1Pik = 1. Combining (4.3) and (4.4) results in an overall
likelihood function of parameters θ and δ:
ln LF(θ, δ) =N
∑i=1
ln LFi(θ, δ) =N
∑i=1
ln
{K
∑k=1
LFik(θk)Pikδk
}. (4.6)
Notice that to identify the parameters of latent class probabilities, the sample has
to be generated from different technological regimes in which the banks are oper-
4 It is important to notice that the inefficiency term uit|k is a deterministic function of time, i.e.,uit|k = λit(.)ui|k.
Heterogeneity of Technological Regimes and Bank Efficiency 65
ating. Hence, the number of classes K determined by the means of information
criteria should not exceed the number of true regimes in the sample, otherwise the
parameters cannot be identified.
Unlike the standard stochastic frontier approach, where the cost frontier is the
same for each bank, in the latent class stochastic frontier model we estimate several
frontiers equal to the number of classes. How can the inefficiency term be estimated
now that there are several benchmarks? One possibility is to assign class member-
ship for an individual bank based on the highest probability and, consequently, use
the stochastic frontier estimated for that class as a benchmark against which the
inefficiency can be computed. However, this approach imposes arbitrary class mem-
bership, while the posterior probabilities of class membership are far from certain.
An alternative approach, used by Greene (2005), is based on the weighted average
of the inefficiency terms:
ln EFit =K
∑k=1
P(k|i) ln EFit(k), (4.7)
where P(k|i) is the posterior probability of class-k membership for bank i; and EFit(k)
is the bank’s efficiency using class-k technology as a reference. In this case, tech-
nologies from every class are taken into account in estimating the overall efficiency.
4.3 Data
We use bank-level data for various FSE, including both former Soviet republics and
Central and Eastern European countries, for the 1995-2005 period. The bank-level
data is extracted from financial reports (balance sheets and income statements)
available though the BankScope database of Bureau van Dijk.5 The data set is
5 To alleviate the impact of randomness in our estimation outcomes, we restrict the data set tothose banks which are present in the sample for 5 or more years in a row.
66 Chapter 4
complemented by historical ownership information collected from individual bank
web-pages and from the EBRD internal database.6 The resulting sample covers
information on banks from the following twenty countries: Albania (AL), Armenia
(AZ), Azerbaijan (AZ), Bulgaria (BG), Bosnia and Herzegovina (BY), Czech Repub-
lic (CZ), Estonia (EE), Georgia (GE), Croatia (HR), Hungary (HU), Kazakhstan
(KZ), Lithuania (LT), Latvia (LV), Moldova (MD), Poland (PL), Romania (RO),
Russia (RU), Slovenia (SI), Slovakia (SK), and Ukraine (UA).
The latent class stochastic frontier model described in the previous section re-
quires three sets of variables determining (i) the stochastic frontier (Cit,yit,t,wit), (ii)
the inefficiency term (zit), and (iii) the class membership (qit). For the stochastic
cost frontier, we follow the modified production approach (see Berger and Humphrey,
1991) and use two types of bank outputs: total loans (y1,it) and total deposits (y2,it).
The banks produce their services using two inputs, physical capital and labor. Ac-
cordingly, the price of the physical capital is measured as a ratio of non-interest
expenses to total assets (w1,it), while the price of labor is proxied by the ratio of to-
tal personnel expenses to total assets (w2,it).7 The dependent variable in the frontier
is the total cost of banks (cit), which includes both interest and operating expenses.
The inefficiency term is measured as a function of the following determinants
zit.8 The first determinant is the foreign ownership dummy variable (FOREIGN).
This variable takes a value of one if more than 50% of bank capital is owned by
foreigners. The coefficient of this variable enables testing the relative efficiency hy-
pothesis of banks depending on their ownership structure. The second determinant
6 We thank Anita Taci from the EBRD for kindly sharing her data set.7 In the absence of a reliable information on the number of bank employees, it has become cus-
tomary in the literature to proxy labor costs by deflating labor expenses over total assets (see, forinstance, Fries and Taci, 2005 or Rossi et al., 2004).8 The selection of inefficiency determinants assumes that these variables can be influenced by the
decision of bank managers. The environmental variables that are out of control of bank managersare expected to influence the technology regimes of banks.
Heterogeneity of Technological Regimes and Bank Efficiency 67
is the interest rate margin (MARGIN), which we incorporate as a measure of mar-
ket power enjoyed by a particular bank. The coefficient of this variable explains
the relationship between market structure and bank efficiency. Finally, the third
determinant is the Battese and Coelli (1992) time trend variable (TIME). This
specification assumes that the inefficiency term is either increasing, or decreasing,
or staying constant over time.
To account for possible heterogeneity due to different production technologies
we employ four country-specific variables qit as latent class determinants: progress
in financial sector reforms proxied by the index of banking sector reforms (BSRF),
progress in market liberalization reforms proxied by the index of economic freedom
(FRDM), the level of GDP expressed in US dollars (GDP), and the interbank rate
(RATE).9 All these variables are not controlled by bank managers, but can po-
tentially influence the banking technology. They have been employed in previous
studies either as variables shifting the cost frontier, or influencing the inefficiency
term. The novelty of our approach is that, instead of imposing a structural relation
between these variables and the cost frontier, we test whether banking technology
varies across countries with different characteristics using the maximum likelihood
principle.
Descriptive statistics of variables employed in our estimations are displayed in
Table 4.2. The decomposition of statistics across different countries shows that there
is a great deal of variation in terms of total costs, outputs, and input prices. In most
cases, the new EU member countries are characterized by relatively higher costs
accompanied by larger outputs and input prices. These countries are also the lead-
9 All variables are time invariant and measured as average values per country (see also equation(4.4)). The index of economic freedom is measured on a yearly basis by the Heritage Foundationand covers a wide range of economic areas, including business, trade, monetary and fiscal poli-cies, property rights, corruption etc. More detailed information about the index is available at:http://www.heritage.org/research/features/index/.
68 Chapter 4
ing performers in terms of banking sector reforms. Whether superior institutional
characteristics can influence banking technology is the question we investigate in the
next step.
The final specification of our latent class cost frontier model takes the following
form:
lncit
wit,1= αik +
S
∑s=2
βsk lnwit,s
wit,1+
L
∑l=1
γlk ln yit,l +12
S
∑s=2
S
∑l=2
δslk lnwit,s
witk,1ln
wit,l
wit,1+
+12
L
∑s=1
L
∑l=1
ψslk ln yit,s ln yit,l +S
∑s=2
L
∑l=1
θslk lnwit,s
wit,1ln yit,l +
+ρ1kt +12
ρ2kt2 +S
∑s=2
ρwskt ln
wit,s
wit,1+
L
∑l=1
ρylkt ln yit,l + vit|k + uit|k, (4.8)
where index k = 1, ..., K expresses class membership. The inefficiency term for each
class is measured using a fixed effects estimator (αik), while linear homogeneity
restrictions are satisfied by expressing all variables in terms of a ratio with respect
to one of the input prices (capital costs). Inefficiency is modeled as a function of its
determinants:
uit|k = exp[η1kFOREIGN + η2k MARGIN + η3k(T − t)]ui, (4.9)
where T is the last period in the sample. The latent class probabilities are specified
as:
Pik(δk) =exp[δ0k + δ1kGDP + δ2kBSRF + δ3k NMS]
K∑
k=1exp[δ0k + δ1kGDP + δ2kBSRF + δ3k NMS]
. (4.10)
Heterogeneity of Technological Regimes and Bank Efficiency 69
4.4 Estimation Results
4.4.1 Selection of the number of classes
In estimating equations (4.8), (4.9), and (4.10) one needs to evaluate the appropriate
number of classes K. A customary way of selecting the number of classes is to draw on
the information criteria. We have computed BIC (Schwartz’s criterion) statistic for
up to three classes.10 The statistic increases with number of classes, which suggests
that the preferred model is the one with three latent classes (see Table 4.3).11
To cross-check the class size selection from the inefficiency term point of view, we
estimate the model for one, two, and three classes and compare the average efficiency
scores for each of these models. As can be observed from Table 4.4, the average
efficiency monotonically increases with the number of classes. This relationship
implies that the country-specific heterogeneity in banking technologies, if not taken
into account, would lead to downward-biased efficiency score estimates.
The high posterior class probabilities (around 90% on average) reported in Table
4.3 suggest that the country-specific variables chosen as class determinants in our
estimations provide quite a precise group classification. Therefore, classification of
banks into three groups according to their maximum probabilities can be performed
with pretty high level of confidence.
10 The BIC statistic can be written as: BIC(K) = 2 ln LF(K) − Π(K) ln(
N∑
i=1Ti
), where K is the
number of latent classes, Π(K) is the number of parameters to estimate for specification with Klatent classes and Ti is the number of observations for bank i. The best model is the one with thehighest BIC statistic.11 Models with more than three latent classes are overspecified and could not be estimated usingthe maximum likelihood methodology.
70 Chapter 4
4.4.2 Parameter estimates and analysis of class-specific effi-ciency scores
Estimates of class-specific parameters are displayed in Table 4.5. In most cases, the
parameters representing the efficiency frontiers are significant at conventional confi-
dence levels. However, the individual coefficients do not have an economic meaning.
Instead, one has to estimate auxiliary measures based on the estimated frontier pa-
rameters to provide an economic interpretation of the estimation outcomes. The
first measure is technical progress, which in our case is assumed to be an exogenous
variable proxied as a function of time. The derivative of total costs with respect to
time (∂ ln C/∂t) calculated at sample means thus measures the change in banking
production technology following innovations not explained by outputs and income
prices. A negative sign for this measure implies technological progress (decrease
in bank costs over time). We find that banks in the second and third classes ex-
hibit technological progress, while the first class is characterized by a frontier with
increasing bank costs over time.
The second measure is the returns to scale estimated as one minus the sum of
elasticities of total costs with respect to outputs (RTS = 1− ∑k
∂ ln C/∂ ln yk). For
constant returns to scale technology, this measure should be equal to zero. A negative
measure implies that banks are operating at the decreasing returns to scale part of
the cost function. Our estimation results suggest that all three technological regimes
exhibit decreasing returns to scale technology, although with different degrees of
intensity.
Average cost efficiency estimates for different classes reported in Table 4.6 show
that the first class represents banks with the highest efficiency scores (80.3%), while
the second class represents the worst performing banks (72.8%). The majority of
banks, representing 46% of the sample, are characterized by an average efficiency
Heterogeneity of Technological Regimes and Bank Efficiency 71
level (73.3%) and clustered in the third class.
Estimates for the class determining variables reported in Table 4.5 imply that the
first class represents banks from small countries with relatively high interest rates
compared to the third class, while the second class represents banks from countries
with a high level of economic freedom and high interest rates.
4.4.3 Economic interpretation of heterogeneous technologies
The next step in our investigation is to search for a pattern between class-membership
of banks and their country of origin. We assign observations for each of the countries
under research to the three classes based on their maximum probabilities (see Table
4.7). As already mentioned before, the possible imprecision in doing this is low given
very large posterior class membership probabilities (about 90% on average).
The results suggest that five out of the eight new EU member countries are
assigned to the (average performing) third class, and the rest is classified to the
worst performing second class. Although these classes are not characterized by
high efficiency levels, they exhibit positive technological progress over time. This
result is remarkable, since it implies that banks in new EU member countries may
have benefited from spillover effects coming from core EU countries and enjoyed
technological progress. However, EU membership did not result in improvement of
the efficiency of the banking system as a whole.
On the contrary, banks from many former Soviet republics with a low level of
economic development are assigned to the best performing first class. Although
relatively more efficient, the first class is also the one that does not exhibit techno-
logical progress. Thus, our results suggest that there seems to be a tradeoff between
efficiency of the banking sector and technological progress in the banking industry.
The impact of inefficiency determinants also varies across classes. Foreign-owned
72 Chapter 4
banks are more efficient in countries assigned to the third class. However, this
variable is not significant in other classes. This finding should be interpreted with
care, since it might be biased due to sample selection (Berger, 2007, Poghosyan and
Borovicka, 2007).
Finally, banks with a higher interest margin (i.e., banks with more market power)
are more efficient than banks belonging to the third class. This finding indicates
efficiency-enhancing effect of consolidation of the banking sector in countries belong-
ing to this class. Market structure is not a significant determinant of inefficiency in
other classes.
4.5 Conclusions
This study provides evidence on the heterogeneity of technology regimes in FSE
banking. Using a novel LCSFA methodology, we show that environmental variables,
such as the level of economic development, progress in economic reforms, and institu-
tional background, have an important influence on the technology regime employed
by banks. Our analysis suggests that single-frontier methods employed in previous
studies, which do not account for technological differences, result in an upward-bias
of inefficiency estimates, since technological differences are mistakenly attributed to
inefficiency.
We identify three distinct technology regimes, characterized by different levels of
technological progress and scale economies. Further analysis of the results reveals
the existence of a tradeoff between bank efficiency and technological progress. Banks
in the new EU member countries exhibit a higher degree of technological progress,
but lower efficiency levels, while former Soviet republics are largely characterized by
efficient banking systems that do not show technological progress over time.
We also find that differences in technology regimes have implications for the
Heterogeneity of Technological Regimes and Bank Efficiency 73
impact of foreign ownership on bank efficiency. In line with the results reported in
Chapter 3, we find that foreign ownership has a positive impact on bank efficiency
in FSE with a relatively higher level of economic development, such as some of the
new EU members. On the contrary, foreign ownership does not have a significant
influence on bank efficiency in most CIS countries, which are still lagging behind in
terms of economic reform.
Overall, our results signify the importance of accounting for differences in tech-
nology regimes when analyzing bank efficiency in FSE. A failure to account for tech-
nological differences may lead to erroneous conclusions regarding various aspects of
banking, including the impact of foreign ownership on bank efficiency.
74 Chapter 4
Tabl
e4.
1.O
verv
iew
ofth
elit
erat
ure
Au
thor
sS
amp
le/C
oun
trie
sM
eth
od
olog
yO
utp
uts
Inp
uts
En
vir
onm
enta
lva
riab
les
X-i
neffi
cien
cyty
pe
Ave
rage
X-
ineffi
cien
cyS
ingl
e-co
un
try
stu
die
sH
asan
and
Mar
ton
(200
3)19
93-1
998
HU
SF
Ato
tal
loan
s,to
tal
inve
stm
ents
(oth
erea
rnin
gas
sets
),n
onin
ter-
est
orfe
ere
late
din
com
e,to
tal
in-
tere
stb
eari
ng
bor
row
edfu
nd
s
bor
row
edfu
nd
s,la
bor
–co
stp
rofi
t29
%35
%
Jem
ric
and
Vu
jcic
(200
2)19
95-2
000
HR
DE
Ato
tal
loan
s,sh
ort-
term
secu
riti
es,
inte
rest
and
non
-in
tere
stre
late
dre
ven
ues
bor
row
edfu
nd
s,la
bor
,ca
pit
al–
cost
serv
ice
pro
vi-
sion
17%
34%
Kra
ftan
dT
irti
rogl
u(1
998)
1994
-199
5H
RS
FA
tota
llo
ans,
tota
ld
epos
its
lab
or,
cap
ital
,lo
anab
lefu
nd
s–
cost
24%
Nik
iel
and
Op
iela
(200
2)19
97-2
000
PL
SF
Alo
ans,
secu
riti
esb
orro
wed
fun
ds,
lab
or–
cost
pro
fit
39%
22%
Cro
ss-c
oun
try
stu
die
sW
eil
(200
3)19
97P
L,
CZ
SF
Alo
ans,
inve
stm
ent
asse
tsb
orro
wed
fun
ds,
lab
or,
cap
ital
cou
ntr
yd
um
mie
s,eq
uit
yco
st34
%
Ros
si,
Sch
wai
ger
and
Win
-k
ler
(200
4)19
95-2
002
CZ
,E
E,
HU
,L
V,
LT
,P
L,
RO
,S
K,
SI
SF
Alo
ans,
dep
osit
s,ot
her
earn
ing
as-
sets
lab
or,
cap
ital
,d
epos
its
fou
rier
term
sco
stp
rofi
t26
%57
%
Fri
esan
dT
aci
(200
5)19
94-2
001
BG
,H
R,
CZ
,E
E,
MK
,H
U,
KZ
,L
V,
LT
,P
L,
RO
,R
U,
SK
,S
I,U
A
SF
Alo
ans,
dep
osit
sla
bor
,ca
pit
alp
erca
pit
aG
DP
,in
tere
stra
te,
den
sity
ofd
epos
its
per
squ
are
kil
omet
er,
as-
set
mar
ket
con
cen
trat
ion
,sh
are
offo
reig
nb
ank
as-
sets
,in
term
edia
tion
rati
o(l
oan
s/d
epos
its)
cost
39%
Bon
in,
Has
anan
dW
ach
tel
(200
5)19
96-2
000
CZ
,H
U,
PL
,S
K,
BG
,H
R,
RO
,S
I,E
E,
LV
,L
T
SF
Alo
ans,
dep
osit
s,li
qu
idas
sets
and
inve
stm
ents
bor
row
edfu
nd
s,ca
pit
alye
ard
um
mie
s,co
un
try
du
m-
mie
sco
stp
rofi
t27
%42
%
Gri
gori
anan
dM
anol
e(2
006)
1995
-199
8C
Z,
HU
,P
L,
SK
,S
I,B
G,
HR
,E
E,
LV
,L
T,
RO
,A
M,
BY
,K
Z,
MD
,R
U,
UA
DE
Ad
epos
its,
reve
nu
es,
net
loan
s,li
q-
uid
asse
tsla
bor
,fi
xed
asse
ts,
inte
rest
ex-
pen
dit
ure
sn
one
serv
ice
pro
vi-
sion
pro
fit
gen
era-
tion
52%
47%
Yil
dir
iman
dP
hil
ipp
atos
(200
7)19
93-2
000
CZ
,E
E,
HR
,H
U,
LV
,L
T,
MD
,P
L,
RO
,R
U,
SI,
SK
DF
A,
SF
Alo
ans,
inve
stm
ents
,d
epos
its
bor
row
edfu
nd
s,la
bor
,ca
pit
aleq
uit
yco
st-D
FA
cost
-SF
Ap
rofi
t-D
FA
pro
fit-
SF
A
28%
24%
34%
50%
Pog
hos
yan
and
Bor
ovic
ka
(200
7)19
95-2
004
AL
,A
M,
AZ
,B
G,
BY
,C
Z,
EE
,G
E,
HR
,H
U,
KZ
,L
T,
LV
,M
D,
MK
,P
L,
RO
,S
I,S
K,
UA
SF
Alo
ans,
dep
osit
sla
bor
,ca
pit
alp
erca
pit
aG
DP
,in
tere
stra
te,
ind
exof
ban
kin
gse
ctor
refo
rms,
ind
exof
econ
omic
free
dom
cost
45%
Gre
en,
Mu
rin
de
and
Nik
olov
(200
7)19
95-1
999
BG
,H
R,
CZ
,E
E,
HU
,L
V,
LT
,P
L,
RO
SF
Alo
ans,
oth
erea
rnin
gas
sets
,n
on-
inte
rest
inco
me
bor
row
edfu
nd
s,la
bor
,ca
pit
alfo
reig
nb
ank
entr
yd
um
my
cost
N/A
Not
es:
AL
-A
lban
ia,
AM
-A
rmen
ia,
AZ
-A
zerb
aijan
,B
G-
Bu
lgar
ia,
BY
-B
osn
iaan
dH
erze
gov
ina,
CZ
-C
zech
Rep
ub
lic,
EE
-E
ston
ia,
GE
-G
eorg
ia,
HR
-C
roat
ia,
HU
-H
un
gary
,K
Z-
Kaz
akh
stan
,L
T-
Lit
hu
ania
,L
V-
Lat
via
,M
D-
Mol
dov
a,M
K-
Mac
edon
ia,
PL
-P
olan
d,
RO
-R
oman
ia,
RU
-R
uss
ia,
SI
-S
love
nia
,S
K-
Slo
vak
ia,
UA
-U
kra
ine.
Heterogeneity of Technological Regimes and Bank Efficiency 75
Tabl
e4.
2.D
escr
iptiv
est
atist
ics
AL
AM
AZ
BG
BY
CZ
EE
GE
HR
HU
KZ
LT
LV
MD
PL
RO
RU
SI
SK
UA
Dep
end
ent
vari
able
Tot
alco
sts
(c)
21.7
4.3
6.8
64.1
177.
424
7.9
89.9
7.1
56.5
254.
452
.130
.723
.74.
926
215
8.7
147.
310
5.5
112.
739
.6S
t.D
ev.
35.3
2.7
13.2
21.7
401
417.
694
.36.
410
3.9
373.
679
.637
.431
.73.
640
8.3
347.
364
015
6.3
138.
567
.4
Fro
nti
erva
riab
les
Tot
allo
ans
(y1)
31.2
11.4
36.1
554.
958
7.8
1274
.067
0.6
27.8
370.
913
01.1
312.
831
0.6
180.
019
.311
73.5
313.
273
8.0
685.
952
1.4
158.
8S
t.D
ev.
31.6
9.2
79.0
375.
314
15.6
2068
.683
9.8
25.8
796.
220
15.8
599.
455
7.5
355.
117
.018
36.2
609.
734
23.6
1123
.361
2.0
290.
2T
otal
dep
osit
s(y
2)31
9.9
31.2
69.5
874.
188
0.3
2667
.887
0.1
35.7
575.
719
63.8
357.
945
2.0
317.
725
.620
08.9
685.
111
89.4
1026
.812
11.6
221.
2S
t.D
ev.
516.
426
.015
0.3
292.
524
15.2
4472
.111
01.5
38.7
1234
.228
21.7
538.
772
9.3
478.
422
.632
46.2
1271
.550
69.4
1521
.515
87.0
384.
6C
ost
ofca
pit
al(w
1)4.
810
.89.
95.
412
.83.
87.
411
.36.
65.
29.
66.
96.
010
.05.
29.
47.
94.
56.
310
.2S
t.D
ev.
4.5
7.4
5.3
0.3
4.1
4.6
3.3
3.3
4.6
2.9
4.3
3.1
3.6
3.5
2.1
5.6
4.4
1.2
15.9
6.0
Cos
tof
lab
or(w
2)0.
010.
030.
020.
010.
040.
010.
020.
030.
020.
020.
030.
030.
020.
030.
020.
030.
030.
020.
010.
03S
t.D
ev.
0.01
0.02
0.02
0.00
0.02
0.00
0.01
0.01
0.01
0.01
0.02
0.01
0.01
0.01
0.01
0.02
0.02
0.00
0.00
0.02
Ineffi
cien
cyd
eter
min
ants
For
eign
own
ersh
ip(z
1)0.
80.
70.
11
0.4
0.7
0.7
0.5
0.3
0.8
0.2
0.6
0.4
0.3
0.6
0.6
0.2
0.3
0.8
0.4
St.
Dev
.0.
40.
50.
30
0.5
0.4
0.5
0.5
0.5
0.4
0.4
0.5
0.5
0.5
0.5
0.5
0.4
0.4
0.4
0.5
Inte
rest
mar
gin
(z2)
4.4
127
6.2
10.4
2.8
5.5
15.7
5.3
57.
44.
74.
411
4.9
9.4
8.3
3.8
3.4
10.1
St.
Dev
.1.
86.
83.
81
5.3
2.4
1.4
5.3
2.6
42.
71.
92.
34.
23.
15.
35.
71.
81.
17.
3
Cla
ssd
eter
min
ants
Ban
kin
gse
ctor
refo
rms
(q1)
2.3
2.3
2.2
3.3
1.4
3.5
3.6
2.4
3.3
4.0
2.6
3.1
3.3
2.3
3.3
2.7
1.9
3.2
3.2
2.1
St.
Dev
.0.
20.
00.
10.
30.
40.
30.
20.
10.
40.
00.
30.
10.
40.
20.
10.
20.
20.
10.
30.
1In
dex
ofec
onom
icfr
eed
om(q
2)2.
62.
92.
12.
81.
93.
73.
82.
42.
63.
42.
23.
43.
42.
73.
22.
52.
32.
93.
12.
2S
t.D
ev.
0.2
0.4
0.4
0.2
0.1
0.1
0.3
0.3
0.2
0.3
0.2
0.3
0.2
0.2
0.2
0.2
0.1
0.2
0.3
0.2
GD
P(q
3)47
7922
0660
7118
280
1544
670
784
5701
3580
2326
462
430
2566
514
312
8694
2038
1818
8247
756
3881
0722
369
2607
943
098
St.
Dev
.15
0552
114
1547
9841
6818
634
716
746
5119
1891
375
0642
8124
7256
932
094
1324
911
0605
4590
7658
1047
3In
tere
stra
te(q
4)10
.718
.620
.42.
348
.25.
98.
522
.97.
312
.18.
66.
54.
017
.513
.746
.623
.25.
89.
720
.7S
t.D
ev.
5.7
10.1
2.0
1.1
30.9
4.5
6.1
9.6
4.5
3.4
6.5
2.6
1.5
8.9
6.6
39.2
12.3
1.9
6.6
18.4
Nu
mb
erof
obs.
tota
l38
4256
439
169
2342
223
102
9559
104
5722
212
229
010
896
162
fore
ign
-ow
ned
ban
ks
3129
74
1612
215
1967
8423
3444
1813
979
5128
7264
Nu
mb
erof
ban
ks
76
91
726
47
3416
169
169
3419
6417
1628
Not
es:
AL
-A
lban
ia,
AM
-A
rmen
ia,
AZ
-A
zerb
aijan
,B
G-
Bu
lgar
ia,
BY
-B
osn
iaan
dH
erze
gov
ina,
CZ
-C
zech
Rep
ub
lic,
EE
-E
ston
ia,
GE
-G
eorg
ia,
HR
-C
roat
ia,
HU
-H
un
gary
,K
Z-
Kaz
akh
stan
,L
T-
Lit
hu
ania
,L
V-
Lat
via
,M
D-
Mol
dov
a,P
L-
Pol
and
,R
O-
Rom
ania
,R
U-
Ru
ssia
,S
I-
Slo
ven
ia,
SK
-S
lova
kia
,U
A-
Uk
rain
e.
76 Chapter 4
Table 4.3. Selection of the number of classesNumber of Number of Log- BIC Posterior classclasses parameters likelihood probability1 21 -355.1 -866.3 0.8802 43 -50.4 -420.5 0.9333 65 109.0 -265.3 0.884
Notes: the table features SFA estimations for 1, 2, and 3 latent classes using 2,058 observations for the
period 1995-2005. The BIC statistic is calculated as: BIC(K) = 2 ln LF(K)− Π(K) ln(
N∑
i=1Ti
), where K is the
number of latent classes, Π(K) is the number of parameters to estimate for specification with K latentclasses and Ti is the number of observations for bank i (the best model is the one with the highest BICstatistic). The posterior class probability reflects the degree of precision with which banks were classifiedto classes (higher probability implies higher precision).
Table 4.4. Average efficiency scores for LCM with different number of classesYear SFA model with SFA model with SFA model with
3 latent classes 2 latent classes 1 latent class1995 0.763 0.720 0.6141996 0.743 0.720 0.6941997 0.732 0.720 0.6901998 0.742 0.720 0.6941999 0.749 0.725 0.7022000 0.747 0.720 0.7082001 0.757 0.730 0.7252002 0.755 0.730 0.7302003 0.754 0.728 0.7342004 0.750 0.726 0.737Total 0.750 0.726 0.718
Notes: the table features average efficiency scores obtained for SFA models with 1, 2, and 3 latent classesusing 2,058 observations for the period 1995-2005.
Heterogeneity of Technological Regimes and Bank Efficiency 77
Table 4.5. LCM estimation resultsClass 1 Class 2 Class 3
Coeff. t-ratio Coeff. t-ratio Coeff. t-ratioIntercept -1.4675 -1.6590 3.6120 1.4230 6.0129 8.1120Total loans -0.4614 -1.4420 -0.5235 -1.7140 -0.3826 -2.3950Total deposits 1.3361 4.4810 1.6195 5.6780 1.2858 8.3210Price of labor/Price of capital 0.3220 1.3430 -0.7347 -1.0480 -2.0698 -9.1420Trend 0.2079 2.6980 -0.3123 -2.0310 -0.1247 -2.7020(Total loans)2 -0.0115 -0.8030 0.1523 7.5240 0.2118 15.4680(Total loans)*(Total deposits) -0.0337 -2.0630 -0.1160 -4.3260 -0.2417 -19.6840(Total loans)*(Price of labor/Priceof capital)
0.1033 2.0350 0.0860 1.5640 0.1606 6.2750
(Total loans)*Trend 0.0041 0.4070 0.0404 3.0700 0.0098 1.7280(Total deposits)2 0.1616 7.0920 0.0787 1.8540 0.3213 23.0500(Total deposits)*(Price of la-bor/Price of capital)
-0.1360 -2.9180 -0.1113 -2.0450 -0.1705 -6.8540
(Total deposits)*Trend -0.0163 -1.7040 -0.0371 -2.4780 -0.0277 -5.4910(Price of labor/Price of capital)2 0.1186 3.7120 0.2123 2.1090 0.4608 12.4120(Price of labor/Price of capi-tal)*Trend
-0.0334 -2.8140 0.0364 1.8890 0.0409 5.7570
(Trend)2 0.0010 0.2080 0.0013 0.1490 -0.0028 -1.2340Sigma 0.8206 3.0070 0.9741 32.7780 0.9211 27.1560Lambda 0.1586 0.1130 3.3844 0.0020 0.1963 0.1140
Inefficiency determinantsIntercept -0.0706 -0.1050 -2.8224 -0.1510 0.2336 1.1020Foreign ownership 0.1489 0.4560 -1.7880 -0.3040 -0.2696 1.7410Interest margin -0.0603 -0.6940 -0.7098 -0.2390 -0.0495 -2.1370Trend 0.1349 6.3160 -0.0312 -1.6510 -0.0054 -0.6450
Class probability determinantsIntercept 0.2983 0.1210 -9.1528 -3.3340 – –Banking sector reforms 0.6579 0.8930 0.1371 0.1980 – –Index of economic freedom -0.3230 -0.3920 1.4317 1.9390 – –GDP (in USD) -0.3423 -2.4300 0.1847 1.0830 – –Interbank rate 0.1243 2.6310 0.1393 2.9160 – –
Auxiliary measures at data meansTechnological progress 0.02 -0.32 -0.04Returns to scale -1.53 -0.27 -2.66
Prior class probabilities at data means0.30 0.24 0.46
Notes: 2,053 observations for the 1995-2005 period. Dependent variable is ln Citwit,1
.
78 Chapter 4
Table 4.6. Comparison of efficiency scoresYear Class-1 Class-2 Class-3 Average1995 0.9512 0.7631 0.6691 0.76311996 0.8767 0.7941 0.6555 0.74291997 0.8429 0.6929 0.7092 0.73181998 0.8451 0.6805 0.7249 0.74201999 0.8353 0.7194 0.7201 0.74902000 0.8210 0.7107 0.7206 0.74712001 0.8096 0.7377 0.7393 0.75722002 0.7907 0.7365 0.7437 0.75482003 0.7760 0.7504 0.7435 0.75382004 0.7531 0.7498 0.7487 0.7502Total 0.8029 0.7281 0.7331 0.7499
Notes: the table features average efficiency scores obtained for the SFA model 3 latent classes using2,058 observations for the period 1995-2005. The classification of banks by classes is performed using themaximum probability principle (e.g., the bank is assigned to class 1 if the probability of being in class 1 ishigher than probabilities obtained for classes 2 and 3).
Table 4.7. Assigning class membershipNumber of obs. Frequency
Class-1 Class-2 Class-3 Total Class-1 Class-2 Class-3 Classmember-ship
EUmember
AL 14 24 38 37% 63% 3AM 26 8 8 42 62% 19% 19% 1AZ 51 5 56 91% 9% 1BG 4 4 100% 3BY 22 17 39 56% 44% 1CZ 4 84 81 169 2% 50% 48% 2 YESEE 23 23 100% 3 YESGE 21 21 42 50% 50% 1/3HR 24 14 185 223 11% 6% 83% 3HU 30 41 31 102 29% 40% 30% 2 YESKZ 17 11 67 95 18% 12% 71% 3LT 4 55 59 7% 93% 3 YESLV 23 14 67 104 22% 13% 64% 3 YESMD 33 24 57 58% 42% 1PL 39 62 121 222 18% 28% 55% 3 YESRO 63 59 122 52% 48% 1RU 41 102 147 290 14% 35% 51% 3SI 29 4 75 108 27% 4% 69% 3 YESSK 12 48 36 96 13% 50% 38% 2 YESUA 79 39 44 162 49% 24% 27% 1
Notes: AL - Albania, AM - Armenia, AZ - Azerbaijan, BG - Bulgaria, BY - Bosnia and Herzegovina,CZ - Czech Republic, EE - Estonia, GE - Georgia, HR - Croatia, HU - Hungary, KZ - Kazakhstan, LT- Lithuania, LV - Latvia, MD - Moldova, PL - Poland, RO - Romania, RU - Russia, SI - Slovenia, SK -Slovakia, UA - Ukraine.
Chapter 5
Foreign Bank Entry, BankEfficiency, and Market Power
5.1 Introduction
In Chapter 2, we provided a detailed analysis of the impact of foreign bank par-
ticipation on the efficiency of banks in former socialist economies (FSEs). Another
important consideration that has motivated local authorities to encourage foreign
bank entry was the hope that opening the borders would improve the competitive-
ness in the domestic banking industries (EBRD, 2005). The outcome of these policies
aimed at attracting foreign direct investments into domestic banking systems has
been remarkable: the average market share of foreign-owned banks in 11 CEECs has
grown from 14% in 1995 to 80% in 2006 (see Figure 5.5), which is the largest increase
of foreign bank participation in emerging markets (IMF, 2000).1 This pattern of for-
eign bank participation is in contrast to developments in industrial countries, where
cross-border bank expansion is rare (Buch and DeLong, 2004).2 In this chapter, we
1 At present, foreign banks account for a dominant share of assets in most of CEECs (except forSlovenia), in some cases reaching the staggering level of more than 90%.2 The main reason for relatively scarce worldwide evidence of cross-border bank expansion can
be the limited success of international takeovers. Major impediments that make banks reluctantto go abroad are geographical distance, language barriers, cultural aspects of home countries, anddifferences in regulatory and supervisory structures (Buch, 2000, Berger et al., 2001).
80 Chapter 5
analyze the impact of increased foreign bank participation on the competitiveness
of banking industries in CEECs, after controlling for the efficiency effects associated
with different modes of foreign entry.
Theoretically, the increased foreign bank participation can affect domestic mar-
kets via increased market competition and improved banking performance due to
spillover effects (Lehner and Schnitzer, 2008). The mode of foreign bank entry
(greenfield investments versus cross-border acquisitions) plays a crucial role in the
transmission of benefits to domestic customers (Claeys and Hainz, 2007). As op-
posed to cross-border acquisition, a greenfield entry increases the total number of
banks, inducing more competition. On the other hand, the primary motivation
for the greenfield investment is usually to follow clients of the bank abroad (Aliber,
1984), which might alleviate the effect of foreign entry on competition. Similarly, the
performance of foreign banks in emerging economies constitutes a trade-off. While
foreign banks entering the market have lower refinancing costs, host country banks
have superior information about the quality of domestic borrowers (Dell’Ariccia and
Marquez, 2004).
Empirical literature provides mixed evidence on the impact of foreign bank en-
try on the performance and competitiveness of banking systems in host countries.
Claessens et al. (2001) report that foreign bank entry leads to more competitive
pressure and higher efficiency of banks in the host country, implying positive wel-
fare effects for economies liberalizing their banking markets. However, this result
holds only for the case of emerging countries, while the conclusions are reversed
when considering foreign bank entry into developed economies.3 For the case of
the CEECs, the impact of foreign bank participation on the performance measured
3 In a related study, Lensink and Hermes (2004) show that the efficiency improvement of domesticbanks following the foreign entry is inversely associated to the level of economic development ofthe host country.
Foreign Bank Entry, Bank Efficiency, and Market Power 81
by cost efficiency is also mixed. Some single-country studies report that foreign-
owned banks are more efficient than domestic banks (see Jemrić and Vujčić, 2002
for Croatia, Hasan and Marton, 2003 for Hungary, and Nikiel and Opiela, 2002 for
Poland), while other studies do not find evidence supporting this view (see Sabi,
1996 for Hungary, Kraft and Tirtiroglu, 1998 for Croatia, and Matoušek and Taci,
2002 for Poland). Evidence from cross-country studies is also inconclusive: studies
by Bonin et al. (2005) and Fries and Taci (2005) report that foreign participation
tends to improve cost efficiency of domestic banks in CEECs, while Poghosyan and
Borovicka (2007) find that the positive effect of foreign ownership on cost efficiency
may be biased due to the cream-skimming effect (sample selection bias).4
Most of this literature, however, does not distinguish between different modes
of foreign entry. The mode of entry can be crucial in interpreting the impact of
foreign bank participation, since different entry modes are driven by different mo-
tives (Claeys and Hainz, 2007, Lehner and Schnitzer, 2008). Havrylchyk and Jurzyk
(2008) distinguish between acquired and greenfield banks and provide further evi-
dence on the existence of a selection bias. However, they conclude that the superior
performance of CEEC banks acquired by foreigners is earned rather than inherited.5
Claeys and Hainz (2007) distinguish between greenfield entry and foreign acquisition
in CEEC banking sectors and find that bank lending rates have generally declined
due to foreign entry, but the impact is mainly driven by the greenfield establish-
ments.6 A similar conclusion is drawn for the case of Latin American countries by
4 The cream-skimming effect suggests that foreign investors select the best-performing banks forthe acquisition (i.e., the domestic bank would perform well even if it was not acquired by foreigners).5 Other evidence of selection bias characterizing foreign bank entry is provided by Lanine and
Vander Vennet (2007). The authors find that foreign banks explicitly target large banks in CEECsin order to extract benefits from an increase in market power. Poghosyan and De Haan (2008)show that the characteristics of target banks in terms of their size and performance depend on themacroeconomic environment and institutional background of host countries.6 It is important to note that the authors acknowledge that greenfield banks can exhibit additional
market power by specializing in particular segments of the market, but they do not provide anempirical test of this hypothesis.
82 Chapter 5
Martinez Peria and Mody (2004). They find that interest margins of foreign green-
field banks are lower than interest margins of domestic banks, as well as interest
margins of foreign banks that have entered through cross-border acquisitions.
The aim of this chapter is to investigate the relationship between different modes
of foreign entry and both cost efficiency and market power of banks in CEECs. Unlike
previous studies, this paper explicitly acknowledges the possible interplay between
efficiency and competition when examining market power of domestic and foreign
banks. Our empirical specification is derived from a simple bank intermediation
model, which allows analyzing market power of banks after taking into account the
cost efficiency effects. The analysis is performed in two steps. First, the stochastic
frontier model (SFA) is applied to evaluate the cost efficiency of banks in CEECs.
In the SFA model, time-varying efficiency scores enable us to evaluate the possible
spillover effects from the increased foreign bank participation to the efficiency of
banks in CEECs. In addition, the efficiency scores are modeled as a function of
the bank ownership in order to distinguish between the relative performance of
domestic, foreign greenfield, and foreign acquired banks. Secondly, we evaluate the
relative market power possessed by banks having different ownership structures using
an equilibrium relationship between bank lending rates, deposit rates, and marginal
costs (free of inefficiency effects) obtained from the intermediation model.
We find that greenfield banks are characterized by a higher degree of cost ef-
ficiency relative to domestic banks and foreign banks that entered through cross-
border acquisitions. Performance of the acquired banks deteriorates during the year
of entry and improves the year thereafter, resulting in an insignificant overall effect.
The hypothesis that banking systems in CEECs are characterized by a competitive
market structure is rejected. However, the market power of foreign acquired banks is
substantially lower compared to the rest of the banks, confirming the positive impact
Foreign Bank Entry, Bank Efficiency, and Market Power 83
of foreign bank entry on competition. Our results remain unchanged when riskiness
of bank portfolio, income from non-interest banking activities, and developments in
the macroeconomic environment are taken into account.
The remainder of the chapter is structured as follows. The next section presents
a simple bank intermediation model and outlines the empirical strategy for testing
the proposed hypotheses. Section 5.3 describes the data used in our analysis, while
the estimation results are provided in Section 5.4. The last section concludes.
5.2 Methodology
5.2.1 Theoretical background
The theoretical framework is based on the new empirical industrial organization
approach of Bresnahan (1982), which has been adopted for the case of banking by
Shaffer (1989) and extended to the intermediation model in more recent studies by
Barajas et al. (1999) and Vera et al. (2007).
Consider a representative bank i producing output in the form of loans or earning
assets (Li), and using deposits or financial liabilities (Di) and non-financial factors
(labor and capital) as inputs. Apart from loans, the bank is also required to hold
reserves with the monetary authority (Ri) on the asset side. The difference be-
tween total assets and deposits constitutes a residual term called other net liabilities
(ONLi).7 The balance sheet identity for each bank i is: Li + Ri = Di + ONLi. Given
the reserve requirement ratio (ρi = RiDi
), the balance sheet identity can be rewritten
as:
Li − Di(1− ρi)−ONLi = 0. (5.1)
7 This term can be further decomposed into bank equity and the rest of other net liabilities. Wemake use of the fact that the minimal amount of equity hold by the bank given its earning assetsis restricted exogenously by the regulatory authorities and focus on competition in deposits andloans markets.
84 Chapter 5
In this simple setup, there is no uncertainty and banks strive for profit maxi-
mization. Each bank earns income by the provision of loans (rLLi) and pays interest
on acquired deposits (rDDi). In addition, each bank incurs real (non-financial) costs
from engaging into financial intermediation (Ci), that depend on the output level
(Li), prices for labor and capital (w), and other non-financial inputs (x). Conse-
quently, each bank’s profits (πi) can be expressed as the difference between financial
revenues and total (financial and non-financial) costs:
πi = rLLi − rDDi − Ci(Li, w, x), (5.2)
where rL and rD are the average lending and deposit rates. Banks maximize their
profits by choosing the optimal level of output, given interest rates rL and rD. The
first order condition for profit maximization is:8
∂πi∂Li
= rL + Li∂rL∂Li
− rD∂Di∂Li
− Di∂rD∂Li
− CLi = 0, (5.3)
where CLi = ∂Ci(Li ,w,x)∂Li
is the marginal non-financial cost of loan production. Making
use of the relationship between deposits and loans ( ∂Di∂Li
= 11−ρi
) from the balance
sheet identity (5.1) and rearranging terms in the first order condition yields the
following equation for the interest rate spread:
rL −rD
1− ρi= −Li
∂rL∂Li
+ Di∂rD∂Di
11− ρi
+ CLi . (5.4)
This equation provides several useful insights. First, the interest rate spread is
affected by the reserve requirement imposed by monetary authorities, which repre-
sents financial taxation costs incurred by a bank. Second, the size of the spread is
8 Here we follow a quantity competition approach, in line with the new empirical industrial or-ganization literature. However, it is important to note that a more realistic price competitionapproach would result in a similar equilibrium condition linking marginal revenues and marginalcosts of banks, which is used to test our main hypotheses (see Freixas and Rochet, 2008, Chapter3 for technical details).
Foreign Bank Entry, Bank Efficiency, and Market Power 85
affected by the production technology used by a bank. More cost efficient banks
use fewer resources to produce the required optimal level of output, which results in
a smaller difference between lending and deposit rates. Third, the wedge between
the lending and deposit rates is driven by the market power of a bank. In the
case of a non-perfect competition, an individual bank will be able to influence the
industry-wide interest rates, as indicated by the terms ∂rL∂Li
and ∂rD∂Di
.
Shaffer (1989) assumes that deposit markets are perfectly competitive ( ∂rD∂Di
= 0)
and estimates equation (5.4) jointly with the demand function for industry-wide
loans. In his formulation, the interest rate elasticity of demand for loans in equa-
tion (5.4) is substituted from the aggregate demand function and marginal cost is
assumed to be a linear function of input prices and output quantity. The system
estimation approach yields a market power parameter estimate for the loans market
in the form of a conjectural variation coefficient, as is customary in the new empirical
industrial organization literature.
We pursue a slightly more restrictive approach suggested by Barajas et al. (1999),
which does not require a system estimation.9 Using the definitions of the interest
rate elasticity of demand for loans (ηL = ∂L∂rL
rLL < 0) and the interest rate elasticity
of demand for deposits (ηD = ∂D∂rD
rDD > 0), equation (5.4) can be rewritten as:
rL + rL
[LiL
dLdLi
1ηL
]=
rD1− ρi
+rD
1− ρi
[DiD
dDdDi
1ηD
]+ CLi , (5.5)
where D and L denote aggregate measures of deposits and loans for all banks. Let
us further denote Lshi = Li
L and Dshi = Di
D as shares of bank i in the loan and deposit
markets, respectively. In addition, let us denote Lrespi = dL
dLi(Dresp
i = dDdDi
) as the
responsiveness of the total industry supply of loans (deposits) to the adjustment of
9 Econometric estimations of a system of equations using a full information maximum likelihoodmethod is problematic, since it produces inconsistent estimates for the whole system if one or moreof the equations are misspecified. Three-stage least squares method is an alternative estimatorwidely used in the literature, but it assumes the availability of appropriate instruments.
86 Chapter 5
loans (deposits) by bank i. Using this notation, equation (5.5) can be rewritten as:
rL
[1 +
Lshi Lresp
iηL
]=
rD1− ρi
[1 +
Dshi Dresp
iηD
]+ CLi . (5.6)
Equation (5.6) explicitly reflects the different effects influencing the market power
of banks, which are summarized by the expressions in brackets. An individual bank
possesses higher market power if the industry supply is less elastic; the size of bank
operations is larger, and the response of the industry output to the individual bank
output decisions is greater. Rearranging the equation and expressing the measure of
market power in the loan market as LMPi =
[1 + Lsh
i Lrespi
ηL
]and the measure of market
power in the deposits market as DMPi =
[1 + Dsh
i Drespi
ηD
]yields:10
rL =rD
1− ρi
[DMP
iLMP
i
]+
CLi
LMPi
. (5.7)
Given the sign restrictions on the interest rate elasticities of loan demand (ηL ≤
0) and deposit supply (ηD ≥ 0), the values for the market power indicators can be
derived as LMPi ≤ 1 and DMP
i ≥ 1, respectively.
In the case of a perfectly competitive industry, both indicators take the value
of unity and, hence, the coefficient DMPi
LMPi
is equal to unity as well. In this case, the
marginal revenue (interest rate on loans) will be equal to the financial and non-
financial marginal costs (deposit rate and derivative of the cost function).
In the presence of market power in at least one of the markets (LMPi < 1 and/or
DMPi > 1), the coefficient DMP
iLMP
iwill be greater than unity. Barajas et al. (1999)
and Vera et al. (2007) use equation (5.7) as an alternative framework for testing the
competitive market structure hypothesis ( DMPi
LMPi
= 1), which is more simplistic relative
10 In the new empirical industrial organization literature, the terms LMPi and DMP
i have been givenan interpretation of conjectural variations. However, we would refrain from this interpretationand would rather view these terms as measures of gap between the price of bank output and themarginal cost.
Foreign Bank Entry, Bank Efficiency, and Market Power 87
to the system approach used in Shaffer (1989). For this purpose, these studies assume
that the marginal cost (CLi ) in equation (5.7) is a linear function of bank output
(Li) and input prices (w). This assumption, however, is not innocuous. It disregards
the cost efficiency of banks, which was found to be an important determinant of net
interest margins in several recent studies (see, for instance, Maudos and Fernandez de
Guevara, 2004). More efficient banks have the opportunity to operate with a lower
margin due to the gains from the less expensive conduct of intermediation activities.
Therefore, the analysis in this paper improves upon previous work by explicitly
taking cost efficiency of banks into account when evaluating their marginal costs.
The next subsection provides the details of our empirical approach.
5.2.2 Empirical methodology
The empirical assessment of the market power possessed by domestic and foreign
banks in at least one of the markets (loan or deposit) is based on the estimation of
the equation (5.7), which can be represented in terms of a linear model:
rLit = β0 + β1rdDit
+ β2(rdDit
∗ DGF) + β3(rdDit
∗ DA) + β4CLit , (5.8)
where indices i and t denote bank and time, respectively, rLit is the implicit loan
rate, rdDit
=rd
Dit1−ρi
is the implicit deposit rate adjusted for the impact of financial
taxation,11 DGF and DA are dummy variables for foreign greenfield and acquired
banks, and CLit is the marginal cost of producing an extra unit of output for bank i
at time t. Abstracting from interaction terms, a value of coefficient β1 significantly
larger than one would indicate the presence of market power in at least one of the
11 The level of financial taxation ρi is an approximate measure, which serves only as a guidelinefor banks in their intermediation activities. In reality, banks often hold excess reserves in theiraccounts at the central bank for liquidity reasons. In addition, banks borrow money from thecentral bank in case their reserves are not sufficient to fulfill the reserve requirements set up by theregulators. In the empirical estimations, we use country-specific reserve requirements informationfrom the international survey on banking regulation available in Barth et al. (2008).
88 Chapter 5
markets (loans or deposits) for the whole banking industry, including both domestic
and foreign banks. Introduction of the interaction terms allows to identify whether
the extent of market power differs between domestic and foreign banks. For instance,
a significantly negative (positive) coefficient β2 would suggest that market power of
foreign greenfield banks is lower (higher) than market power of domestic banks. The
magnitude and sign of the coefficient β3 can be interpreted in a similar way.
To carry out an estimation of equation (5.8), one needs to introduce a measure
of marginal costs. Instead of pursuing the strategy of Barajas et al. (1999) and Vera
et al. (2007) and proxying the linear relationship between marginal costs and their
underlying factors in an ad hoc way, the marginal costs are obtained directly from
the data using the stochastic efficiency frontier methodology.12 The advantage of
this approach is that it explicitly takes the impact of the cost efficiency of banks on
the marginal cost of producing an additional unit of output into account. By includ-
ing the inefficiency-free measure of marginal costs, we also control for the possible
relationship between market power of banks and their efficiency.13 In addition, using
information on the timing of cross-border bank acquisitions, we are able to evaluate
whether domestic banks taken over by foreigners improve their operational efficiency
after the acquisition or not.
Consistent with the intermediation model described above, let us assume that
banks produce one unit of output (L) using labor, capital and borrowed funds as
inputs. Let w1, w2 and w3 denote the prices of labor, capital and borrowed funds.
To capture the technological progress experienced by banks in CEECs during the
12 A comprehensive textbook exposition of the stochastic efficiency frontier methodology can befound in Kumbhakar and Lovell (2000) and Coelli et al. (2005).13 Efficiency of banks can affect their pricing strategy. For example, more cost efficient banks incurlower marginal costs and can set lower prices compared to the less cost efficient banks. Applicationof the inefficiency-free measure of marginal costs makes it possible to compare the market powerparameters (measured as a relative wedge between prices and marginal costs) across banks withdifferent efficiency levels.
Foreign Bank Entry, Bank Efficiency, and Market Power 89
last decade,14 a time trend (Trend) is introduced among the determinants of the cost
frontier. In line with previous cross-country studies, we also control for possible shifts
in the cost frontiers across countries due to differences in macroeconomic conditions
and institutional backgrounds by introducing country-specific (Cn) and time-specific
(Tm) dummy variables. The final translog specification of the cost function for the
stochastic frontier analysis takes the following form:15
lnCit
wit,1= αi0 + α1 ln Lit + α2 ln
(wit,2
wit,1
)+ α3 ln
(wit,3
wit,1
)+ α4Trend +
+ δ1112
(ln Lit
)2
+ δ12 ln Lit ln(
wit,2
wit,1
)+ δ13 ln Lit ln
(wit,3
wit,1
)+ δ14 ln LitTrend +
+ γ1112
(ln(
wit,2
wit,1
))2
+ γ12 ln(
wit,2
wit,1
)ln(
wit,3
wit,1
)+ γ13 ln
(wit,2
wit,1
)Trend +
+ θ1112
(ln(
wit,3
wit,1
))2
+ θ12 ln(
wit,3
wit,1
)Trend + ρ11
12(Trend)2 +
+N
∑n=1
φnCn +M
∑m=1
φmTm + uit + vit, (5.9)
where αi0 captures individual bank random effects, vit ∼ N(0, σ2v ) is the i.i.d. error
term and uit = Btui is the positive inefficiency term varying across banks and over
time, which is composed of two parts: a non-stochastic positive time component,
Bt > 0, that is time-varying but the same for all banks and a stochastic individual
component, ui ∼ N+(µ, σ2u), which follows a truncated normal distribution with a
conditional mean parameter µ. The inefficiency term can be expressed in a general
form as:
uit = exp(η′Zit)ui, (5.10)
where Zit is a vector of factors affecting bank efficiency and η is a vector of parame-14 See Fries and Taci (2005), Bonin et al. (2005) and Poghosyan and Borovicka (2007) for the recentempirical evidence of the impact of technological progress in transition banking.15 This formulation takes into account the adding-up and symmetry restrictions imposed by theory.In addition, the linear homogeneity restriction is satisfied by deflating costs and the second inputprice by the first input price.
90 Chapter 5
ters. We use several determinants of bank efficiency. First, the efficiency is modeled
as a function of time using the specification of Kumbhakar and Wang (2005): (t− t),
where t is the beginning of the sample. A significant positive (negative) parameter
estimate of this variable would indicate that over the whole sample period, effi-
ciency of banks in CEECs has deteriorated (improved). Since the sample period
was marked by increased foreign bank participation, the coefficient of this variable
can be interpreted in terms of the overall impact of foreign bank participation on
bank efficiency in CEECs. Next, in order to discern the differences in cost effi-
ciency across domestic and foreign banks, we introduce dummy variables for foreign
greenfield (DGF) and foreign acquired banks (DA) into the inefficiency specification
(5.10). A significant positive (negative) coefficient of these dummy variables would
indicate that the post-entry efficiency of the corresponding foreign-owned banks is
on average lower (higher), in comparison to their peers. Finally, in a separate set of
estimations, we introduce current and lagged dummy variables for the year when the
domestic bank was taken over in order to evaluate the dynamic effect of cross-border
bank acquisitions on the banks’ performance.
Using results from the stochastic frontier model, the estimate of the marginal
cost term for bank i at time t (CLit ) is obtained through the partial derivative of the
translog function:
CLit =CitLit
∂ ln Cit∂ ln Lit
=CitLit
[α1 + δ11 ln Lit + δ12 ln
(wit,2
wit,1
)+ δ13 ln
(wit,3
wit,1
)+ δ14Trend
].
(5.11)
The marginal cost term CLit is adjusted for the influence of bank inefficiency and
can enter as an explanatory variable in equation (5.8). Using the generated regressor
CLit on the right hand side of (5.8) will influence the efficiency of the coefficient esti-
Foreign Bank Entry, Bank Efficiency, and Market Power 91
mates due to the biased standard errors (see Pagan, 1984). Therefore, the standard
errors of the coefficient estimates are bootstrapped using 2000 replications to ensure
the robustness of our results.16
5.3 Data Description
The main source for the bank-specific information is the BankScope database of
Bureau Van Dijk, from which the information on individual banks operating in 11
CEECs (Bulgaria, Croatia, the Czech Republic, Estonia, Hungary, Latvia, Lithua-
nia, Poland, Romania, Slovakia, and Slovenia) is retrieved for the 1992-2006 period.
The data set contains information on balance sheets and income statements of 364
commercial, cooperative and savings banks.17 Unfortunately, BankScope does not
provide historical information on bank ownership, which is crucial for our analysis.
Therefore, we utilize the information on foreign-owned banks for the years 1992-
2004 from the extended data set of De Haas and Van Lelyveld (2006) employed in
Havrylchyk and Jurzyk (2008).18 This data set categorizes foreign-owned banks into
two groups: greenfield establishments and banks taken over as a result of a cross-
border acquisition. For the remaining two years, we update the missing foreign
ownership information using a list of cross-border bank acquisitions from Securities
Data Company (SDC) mergers and acquisitions database produced by Thompson
Financial. From this source, data on completed (effective) cross-border acquisitions
are extracted (i.e. parents of bidder and target banks have different countries of
origin), which involve target banks from CEECs and that result in the control of
ownership by the bidder bank exceeding 50% of the total equity outstanding.
Table 5.1 displays the evolution of foreign bank entry into CEECs. The dominant
16 The number of bootstrap replications is chosen based on the optimal criteria suggested by An-drews and Buchinsky (2000).17 We use unconsolidated statements of banks, replacing them by consolidated statements wheneverinformation on unconsolidated statements is not available.18 We thank Emilia Jurzyk and Iman Van Lelyveld for kindly sharing their data on bank ownership.
92 Chapter 5
mode of foreign entry in the initial stage of transition has been the establishment of
greenfield subsidiaries. The number of greenfield banks has grown rapidly by the mid
1990’s, remaining at comparable level afterwards. Cross-border acquisitions became
a popular mode of entry after the mid 1990’s, growing at an accelerating pace with
EU enlargement. In the last year of the sample, the share of total banking system
assets controlled by foreign banks amounted to 65.3%.19 Decomposition of this
share by the entry modes reveals that 15.1% of banking system assets is controlled
by greenfield banks, while the remaining 50.2% is under control of foreign acquired
banks.
Table 5.2 lists and describes the variables used and their sources. Before proceed-
ing with the empirical analysis, observations with missing information in at least one
of the variables listed in Table 5.2 are dropped. Furthermore, to tackle the influence
of extreme observations and reporting errors, all variables are winsorized at the 1st
and 99th percentiles.
Descriptive statistics of the resulting data set are reported in Table 5.3. The
Table shows that foreign greenfield banks have lower scale of operations and incur
lower costs in comparison to the foreign acquired and domestic banks. This is due
to the fact that the main mission of greenfield banks is to serve their clients abroad,
rather than to engage into full scale operational activities in CEECs. There is also
high variation in terms of loan rates: domestic and foreign greenfield banks charge
on average more for their loans that foreign acquired banks. However, the variation
of deposit rates across banks is relatively modest. This observation can be explained
by the fact that depositors find it easier to switch banks when discrepancy in deposit
rates is high, while lending rates are to a large extent influenced by relationships of
19 Difference between the share of total assets controlled by foreign-owned banks in the sample andthe EBRD information reported in Figure 5.5 is due to the fact that BankScope does not coverall banks in the economy. In addition, our estimates refer to commercial, cooperative and savingsbanks only, while the EBRD data covers all banks in the country.
Foreign Bank Entry, Bank Efficiency, and Market Power 93
banks with their clients (Petersen and Rajan, 1994). Domestic and foreign banks also
differ in terms of the riskiness of their loan portfolios: domestic and foreign acquired
banks have higher loan-loss provision reserves relative to the foreign greenfield banks.
To sum up, the preliminary analysis of the descriptive statistics highlights ap-
parent differences between domestic, foreign greenfield, and foreign acquired banks
in terms of the scale of their operations, incurred costs, and riskiness. These differ-
ences may be related to different missions and strategies employed by these banks,
reflected in their portfolio mix. However, the simple comparison made using sum-
mary statistics lacks theoretical argumentation and does not allow drawing firm
conclusions regarding foreign bank entry effects on efficiency and market power. In
the remainder of the paper, these issues are addressed using a more formal frame-
work.
5.4 Estimation Results
5.4.1 Foreign bank entry and cost efficiency
The empirical approach for evaluating the impact of foreign entry on bank efficiency
is based on the stochastic efficiency frontier methodology (SFA). We follow the inter-
mediation approach widely used in the banking literature (Sealey and Lindley, 1977)
and assume that banks are minimizing their costs given the optimal amount of earn-
ing assets to be generated, prices for inputs (labor, capital and financial resources)
and technological constraints. Bank costs (C) are measured as the total operating
expenses incurred by banks. Bank output (L) is proxied by the total earning assets
in the bank’s portfolio.20 Following the literature on bank efficiency, labor prices are20 In a separate set of estimations, we subdivided bank output into two categories: total loans andtotal security holdings. We also did estimations using only total loans as an output. In both cases,the estimation results yielded qualitatively similar outcomes and are available upon request. Thepossible reason for the similar outcomes is the dominating share of total loans in total earning assets(about 90%) due to underdeveloped securities market in CEECs. Therefore, in the remainder ofthe text we refer to the total earning assets as bank output L and use terms total earning assets
94 Chapter 5
measured as the ratio of personnel expenses to total assets (w1), capital prices as
the ratio of administrative expenses (other than personnel expenses) to total assets
(w2) and prices of borrowed funds as the ratio of interest expenses to a sum of total
deposits and other funding (w3). We control for the possible influence of environ-
mental differences across countries (e.g., macroeconomic developments, institutional
background) and over time (e.g., shocks common to all CEECs), by using country
and time dummies.
The outcomes of the SFA model estimations are summarized in Table 5.4. The
main focus of this analysis is the determinants of cost inefficiency, shown in the
middle panel of the Table. Let us start by introducing time trend as inefficiency
determinant in the specification (I). The negative significant coefficient of the trend
variable suggests that efficiency of banks in CEECs has on average improved over
time, which is in line with the evidence provided by Rossi et al. (2004). Increased
foreign bank participation has possibly influenced this general efficiency improve-
ment directly (through the higher efficiency of foreign banks) or indirectly (through
the increased competition due to foreign entry and knowledge spillovers).21
In order to evaluate the direct impact of foreign bank participation, in specifi-
cations (II) and (III) dummy variables for foreign greenfield and foreign acquired
banks are introduced. The estimation results suggest that foreign greenfield banks
have higher efficiency than domestic and foreign acquired banks. Introducing both
dummy variables simultaneously as inefficiency determinants in the specification
(IV) does not alter this result. This finding has important policy implications: it
highlights the importance of the entry mode on the performance of foreign banks. It
and total loans interchangeably.21 In a separate set of regressions, we replaced the time trend by the yearly series on the marketshare of foreign bank assets from EBRD (2007). In these estimations (available upon request), asignificant negative coefficient of the foreign market share variable was obtained, suggesting thatthe efficiency improvement is correlated with the increased foreign bank participation.
Foreign Bank Entry, Bank Efficiency, and Market Power 95
also suggests that the primary motivation behind foreign entry affects the post-entry
performance of banks. While foreign greenfield banks are mainly established with
the purpose to serve the clients of their parent banks, the entry via cross-border ac-
quisitions is primarily motivated by the efficiency improvements and market power
considerations (Lanine and Vander Vennet, 2007). As argued by Detragiache et al.
(2008), bank costs after the takeover can increase due to additional expenses related
to the need to increase the quality of monitoring activities.22 In order to capture
this dynamic effect, in specifications (V) - (VIII) current and lagged dummy vari-
ables for the year when the bank was taken over are introduced.23 We find two
offsetting effects on the efficiency following the foreign acquisition: the immediate
impact is significantly positive (deterioration of bank efficiency), while the one pe-
riod lagged impact is significantly negative (improvement of bank efficiency). These
two offsetting effects together with the fact that efficiency gains disappear in the
second period, as shown in the specifications (VII) and (VIII), might explain the
insignificant overall impact of the acquisition dummy variable in the specifications
(III) and (IV).
These findings are also in line with various case studies on foreign bank acquisi-
tions in CEECs. For instance, Abarbanell and Bonin (1997) discuss the impact of
privatization of the Polish Bank Slaski (BSK) to a foreign investor in the 1990s. The
authors find that the privatization of the bank by foreign investors did not lead to
an immediate improvement of its managerial performance. One explanation is that
the top management who ran the bank prior to the privatization did not change
22 Another explanation for the insignificant relationship between the bank acquisition and its subse-quent efficiency improvement might be the additional costs incurred in the process of reorganizationand restructuring, which most of the banks undergo following the takeover. Still another possibilitymight be that target banks introduce new services, which requires installation of new equipmentand facilities causing an upsurge of costs in the short-run.23 This dummy variable captures 64 cross-border bank acquisition events. The number of feasibleobservations for cross-border acquisitions decreases to 53 (44) when the impact of the takeover isevaluated with a one period (two periods) time lag.
96 Chapter 5
following the privatization, due to the “...strength of personality, political influence,
and superior knowledge of banking...” (Abarbanell and Bonin, 1997, p. 46). Simi-
lar evidence has been documented in a case study on privatization of the Russian
Zhilsotsbank (Abarbanell and Meyendorff, 1997). However, the authors caution that
the results of privatization should not be judged only on the basis of the short-run
financial performance and that a “...critical lesson to be learned from the privatiza-
tion of BSK is the importance of a foreign financial investor taking an active role in
the development of bank strategy to bring about the fundamental changes necessary
to realize the potential franchise value.” (Abarbanell and Bonin, 1997, p. 57).
To sum up, we find that the mode of foreign entry has different implications for
bank efficiency. Foreign greenfield banks outperform domestic banks in terms of cost
efficiency, while the efficiency of foreign acquired banks is not significantly different
from that of domestic banks. The later result can be explained by offsetting effects
on efficiency following the foreign acquisition.
5.4.2 Foreign bank entry and market power
In order to evaluate the market power of banks, the following variables are used in
model (5.8): the implicit lending rate (rLit ) is defined as the ratio of total interest
income to total loans, and the implicit deposit rate (rDit) is proxied by the ratio
of total interest expenses to total deposits. The deposit rates are adjusted by the
corresponding reserve requirement ratios in each of the CEECs (see Table 5.2). To
evaluate the impact of foreign ownership on market power of banks, interaction
terms of the average deposit rate with a foreign greenfield bank dummy (rDit ∗DGF)
and with a foreign greenfield bank dummy (rDit ∗DA) are introduced. Together with
the marginal cost estimates (MC) obtained from the SFA specification (IV) in Table
5.4, these variables can be used for conducting the market power test using model
(5.8).
Foreign Bank Entry, Bank Efficiency, and Market Power 97
Table 5.5 shows the estimation results of (the augmented) equation (5.8). We ac-
count for heterogeneity across banks located in different CEECs with varying levels
of economic development and regulatory structures by applying a panel data estima-
tion technique. All estimations are done by fixed-effects method, which was found
to outperform the random-effects method based on the Hausman test. Standard
errors are estimated using residuals clustered by countries, to relax the assumption
of cross-sectional independence. Panel test for serial correlation based on the pro-
cedure of Drukker (2003) suggests that residuals in all specifications are free from
first order autocorrelation effects.
Specification (I) describes the baseline model. The coefficient of the deposit rate
variable is significant and larger than one. The Wald test indicates that the market
power coefficient is significantly larger than one, suggesting rejection of the com-
petitive market structure hypothesis for CEECs banking sector as a whole. This
finding applies to all banks in CEECs, regardless of their ownership. To evaluate
the impact of bank ownership on market power, the corresponding interaction terms
are included in specifications (II) and (III). The coefficients of interaction terms sug-
gest that foreign acquired banks have a significantly lower market power compared
to domestic and foreign greenfield banks. This finding does not alter when both
interaction terms are added to the model simultaneously in the specification (IV).
The Wald test suggests that market power coefficient of foreign acquired banks is
not significantly different from one, supporting the competitive market structure hy-
pothesis for these banks. This result contrasts the prediction of the Claeys and Hainz
(2007) model, in which competition in the domestic banking markets is stronger for
the greenfield entry, compared to the acquisition entry.24 Our results suggest that
24 Claeys and Hainz (2007) do not consider the follow clients abroad motive for foreign bank entryin their model, which might explain this contradictory result.
98 Chapter 5
cross-border bank acquisitions result in a more competitive banking environment,
which has important policy implications.
Robustness check
There are several important aspects of banking that are not captured in the theo-
retical model of market power. The first is the presence of uncertainty and credit
risk. To control for the impact of risk, we follow Barajas et al. (1999) and Vera et al.
(2007) and introduce the share of loan-loss provisions in total loans as a proxy of
quality of bank loan portfolio.25 The second aspect is the presence of non-interest
banking services, which might be considered as additional revenue for banks and
might influence their degree of riskiness and market power (Lepetit et al., 2008). To
control for the impact of fee-generating activities of banks, we follow Maudos and
Fernandez de Guevara (2004) and augment our specification by introducing the ratio
of non-interest revenues to total assets as a proxy for implicit interest revenues of
banks. Finally, macroeconomic fundamentals might influence the depth of financial
intermediation in the country (Cotarelli et al., 2005) and decision of banks to go
abroad. We control for the macroeconomic environment by introducing real GDP
growth, inflation and exchange rate changes in our specification.
The introduction of additional variables to control for banking risks (LLP), ser-
vice incomes (IMPL) and macroeconomic environment (GDP, INFL and FX) in
specifications (V), (VI), and (VII) does not change the main results. In particu-
lar, the coefficient of the interaction term with foreign greenfield dummy remains
insignificant, implying that even after accounting for credit risks, non-interest bank-
ing activities and macroeconomic variables, greenfield banks do not exhibit lower
25 A more direct measure of loan portfolio quality would be the share of non-performing loans intotal loans. However, BankScope is missing information on non-performing loans for more thanhalf of banks in the sample, for which reason we rely on loan-loss provisions as an indicator of loanportfolio quality.
Foreign Bank Entry, Bank Efficiency, and Market Power 99
market power than domestic banks. This insignificant decrease in market power
can be explained by the absence of alternative sources of bank financing for the
customers of greenfield banks, who already established relationships with their long-
term partner banks.
In line with the theoretical prediction, banks with riskier loan portfolios and
higher share of non-interest banking activities charge higher lending rates.26 The
later result supports the findings of Lepetit et al. (2008), according to which banks
expanding to non-interest income activities are riskier than banks focused on lending,
which is reflected in higher loan rates. Among macroeconomic variables, we find
positive and significant effect of exchange rate depreciation on loan rates, which
suggests that currency stability has important implications for lending decisions of
banks.
To sum up, the estimation results reject the competitive market structure hy-
pothesis in CEECs, as the estimated market power coefficients are significantly larger
than one. This indicates that banks in CEECs possess market power at least in one
of the markets (loans or deposits).27 The market power of foreign acquired banks
is significantly lower than that of domestic and foreign greenfield banks, suggesting
that increase in competition as a result of the foreign entry is mainly driven by
cross-border acquisitions.
26 Since interest income of banks can be affected by the quality of loan portfolio, using LLP amongexplanatory variables may introduce endogeneity bias in coefficient estimates. To control for pos-sible endogeneity, in a separate set of regressions we use lagged LLP among explanatory variables.The estimation results are qualitatively similar to the specification with contemporaneous LLP andare available upon request.27 Since the deposit market is likely to be more competitive than the loan market due to the negli-gible bank switching costs for depositors and prevalence of relationship-based lending, we suggestthat the main part of the market power comes from the loan markets. Relatively lower variationof deposit rates relative to the loan rates in our sample lends support for this argumentation (seealso discussion in Section 5.3).
100 Chapter 5
5.5 Conclusions
This paper has studied the implications of the recent sharp increase in foreign bank
participation in CEECs for the post-entry banking performance. The study has
highlighted the existence of a complex relationship between different modes of foreign
bank entry and both cost efficiency and market power of banks.
Foreign greenfield banks exhibit superior operational efficiency in comparison to
domestic and foreign acquired banks. This can be explained by the specialization
of greenfield banks to serve customers of their parent banks abroad and already
established banking relationships. The performance of foreign acquired banks ex-
hibits an offsetting dynamic pattern: the efficiency deteriorates in the initial year
of acquisition, slightly improving in the subsequent year. The overall impact on the
post-acquisition performance evaluated for the whole sample is insignificant, which
can be due to the poor managerial and financial characteristics of target banks in
CEECs inherited by foreign investors.
We also find evidence on differences in market power across domestic and foreign
banks. Market power of foreign greenfield banks is not significantly lower than that
of domestic banks. This result holds when the impact of credit risks, non-interest
banking activities and macroeconomic environment are taken into account, contrast-
ing the evidence from studies, which do not control for the cost efficiency of banks
when analyzing market power. Unlike greenfield entrants, foreign acquired banks ex-
hibit a substantially lower degree of market power, which can be explained by their
strategic considerations to expand activities in CEECs and subsequent increase of
the competitive pressure.
The analysis conducted in this study provides important policy implications. It
documents a significant improvement of banking performance in CEECs measured
Foreign Bank Entry, Bank Efficiency, and Market Power 101
by cost efficiency during the sample period corresponding to an increase in foreign
bank participation. CEECs banks and customers have benefited from foreign partic-
ipation both directly (superior post-entry performance of greenfield banks) and indi-
rectly (overall increase in bank efficiency due to spillover effects to domestic banks).
Opening the borders for foreign entry has also contributed to the competitiveness
of the banking industry in CEECs, but largely due to cross-border acquisitions. In
this sense, the findings in this study provide support for the conventional belief by
the policymakers that liberalization of domestic banking industry and promotion of
foreign entry would have a positive impact.
However, these conclusions should be interpreted with caution, since this study
has not addressed the issue of financial stability in CEECs. During the recent
financial crisis, banking sectors in CEECs have proven to be very vulnerable to
systemic external shocks. The impact of the increased foreign bank participation on
financial stability is an important topic, which requires the attention of policymakers
and needs to be addressed in the future research.
102 Chapter 5
Figure 5.1. Share of foreign-owned banks in terms of total assets (%), 1995 and 2006
Source: EBRD (2007).
Foreign Bank Entry, Bank Efficiency, and Market Power 103
Tabl
e5.
1.N
umbe
rof
obse
rvat
ions
for
dom
estic
and
fore
ign
(acq
uire
dan
dgr
eenfi
eld)
bank
sC
ount
ries
Ow
ner
ship
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
Tot
alB
ulga
ria
Dom
esti
c4
912
1517
2123
1714
1412
1111
109
199
Fore
ign
gree
nfiel
d0
00
12
22
34
67
77
77
55Fo
reig
nac
quir
ed0
00
00
01
47
79
109
1011
68C
roat
iaD
omes
tic
1222
2831
3745
3833
2726
2630
1918
1740
9Fo
reig
ngr
eenfi
eld
00
11
14
46
66
54
22
244
Fore
ign
acqu
ired
00
00
00
11
45
44
45
634
Cze
chR
epub
licD
omes
tic
1014
1920
2221
1916
98
76
66
518
8Fo
reig
ngr
eenfi
eld
37
810
1212
1010
1110
109
99
913
9Fo
reig
nac
quir
ed0
00
00
13
46
77
87
78
58E
ston
iaD
omes
tic
38
1116
1920
87
22
44
32
211
1Fo
reig
ngr
eenfi
eld
00
00
00
00
00
00
00
00
Fore
ign
acqu
ired
00
00
00
12
33
33
34
426
Hun
gary
Dom
esti
c18
2330
3430
2319
1810
1011
108
88
260
Fore
ign
gree
nfiel
d4
79
1112
1211
1516
1313
1312
1212
172
Fore
ign
acqu
ired
00
00
26
89
88
88
77
778
Lat
via
Dom
esti
c3
916
2021
2525
2316
1414
1514
1413
242
Fore
ign
gree
nfiel
d0
01
11
22
22
33
33
33
29Fo
reig
nac
quir
ed0
00
00
12
22
44
45
56
35L
ithu
ania
Dom
esti
c1
59
1114
1716
158
65
55
44
125
Fore
ign
gree
nfiel
d0
00
00
00
00
00
00
00
0Fo
reig
nac
quir
ed0
00
00
00
02
34
44
55
27P
olan
dD
omes
tic
1724
3537
4140
3228
1814
1214
1413
1335
2Fo
reig
ngr
eenfi
eld
24
57
1111
1311
1111
1313
1212
1214
8Fo
reig
nac
quir
ed0
00
01
34
1014
1616
1515
1616
126
Rom
ania
Dom
esti
c4
59
1211
1324
2316
1414
1110
108
184
Fore
ign
gree
nfiel
d0
00
01
38
99
910
1010
1010
89Fo
reig
nac
quir
ed0
00
00
01
13
56
99
911
54Sl
ovak
iaD
omes
tic
46
911
1316
1615
107
54
33
312
5Fo
reig
ngr
eenfi
eld
11
24
79
108
109
99
88
810
3Fo
reig
nac
quir
ed0
00
00
00
02
57
76
66
39Sl
oven
iaD
omes
tic
710
1320
2827
2324
1714
1012
1010
923
4Fo
reig
ngr
eenfi
eld
12
23
33
22
22
22
22
232
Fore
ign
acqu
ired
00
00
01
11
12
44
44
527
Tot
alD
omes
tic
8313
519
122
725
326
824
321
914
712
912
012
210
398
912,
429
For
eign
gree
nfi
eld
1121
2838
5058
6266
7169
7270
6565
6581
1F
orei
gnac
quir
ed0
00
03
1222
3452
6572
7673
7885
572
Sou
rce:
Ban
kS
cop
e,T
hom
pso
nF
inan
cial
SD
CP
lati
nu
mD
atab
ase,
De
Haa
san
dV
anL
ely
veld
(200
6)an
dH
avry
lch
yk
and
Ju
rzy
k(2
008)
.
104 Chapter 5
Table 5.2. Variable definitions and sourcesVariable Definition Measure SourceC Bank costs Total operating expenses BankScopeL Earning assets Total earning assets BankScopew1 Price of labor Ratio of personnel expenses to
total assetsBankScope
w2 Price of capital Ratio of administrative expenses(other than personnel expenses)to total assets
BankScope
w3 Price of borrowed funds Ratio of interest expenses to asum of total deposits and otherfunding
BankScope
DGF Foreign greenfield Dummy variable that takes valueof 1 for greenfield establishmentsof foreign banks
De Haas and Van Lelyveld(2006), Havrylchyk and Jurzyk(2008)
DA Foreign acquired Dummy variable that takes valueof 1 for domestic banks acquiredby a foreign bank
De Haas and Van Lelyveld(2006), Havrylchyk and Jurzyk(2008), and Thomson’s SDCPlatinum Database
DFE Foreign entry Dummy variable that takes valueof 1 in the year when a domesticbank was taken over by a foreignbank
De Haas and Van Lelyveld(2006), Havrylchyk and Jurzyk(2008), and Thomson’s SDCPlatinum Database
rL Implicit loan rate Ratio of interest expenses to to-tal loans
BankScope
rD Implicit deposit rate Ratio of interest expenses to to-tal deposits
BankScope
MC Marginal costs Derivative of the cost func-tion obtained from the stochas-tic frontier model with respect tooutput quantity
BankScope and own estimations
LLP Loan-loss provisions Ratio of loan-loss provisions tototal loans
BankScope
IMPL Implicit interest revenue Ratio of the net non-interest rev-enues to total assets
BankScope
ρ Reserve requirements ratio (%) Bulgaria=8, the Czech Republic= 2, Estonia = 16, Croatia =19, Hungary = 5, Latvia = 8,Lithuania = 6, Poland = 3.5, Ro-mania = 20, Slovakia = 2, Slove-nia = 2.
Barth et al. (2008)
GDP Economic activity Annual real GDP growth World Development Indicators(WorldBank)
INFL Inflation Annual growth in consumer priceindex (CPI)
World Development Indicators(WorldBank)
FX Currency stability Annual growth of average ex-change rate vis-a-vis US dollar
International Financial Statistics(IMF)
Foreign Bank Entry, Bank Efficiency, and Market Power 105
Tabl
e5.
3.D
escr
iptiv
est
atist
ics
Ban
kco
sts
Ear
nin
gas
sets
Pri
ceof
lab
orP
rice
ofca
pit
alP
rice
ofb
or-
row
edfu
nd
sL
oan
rate
Dep
osit
rate
Mar
gin
alco
sts
Loa
nlo
ssp
ro-
visi
ons
Imp
lici
tin
tere
stre
venu
esC
Lw
1w
2w
3r L
r DM
CLL
PIM
PL
Dom
esti
cM
ean
748.
617
335.
50.
512
0.02
30.
068
0.24
60.
079
0.07
60.
097
0.08
3ba
nks
Med
ian
231.
039
59.8
0.52
40.
020
0.05
40.
189
0.06
60.
068
0.06
00.
072
St.
Dev
.13
39.6
3163
0.5
0.20
10.
012
0.04
50.
186
0.05
20.
035
0.10
60.
040
Max
imum
9701
.819
3000
.00.
849
0.07
10.
324
1.84
70.
336
0.22
51.
000
0.31
0M
inim
um13
.714
5.0
0.04
20.
004
0.01
10.
066
0.01
20.
012
0.00
00.
014
Fore
ign
Mea
n19
9.2
6311
.40.
585
0.01
40.
053
0.26
50.
061
0.04
90.
019
0.05
4gr
eenfi
eld
Med
ian
104.
346
63.8
0.64
30.
010
0.04
60.
161
0.05
40.
041
0.01
60.
048
bank
sSt
.D
ev.
256.
265
60.5
0.19
40.
010
0.03
70.
367
0.04
50.
027
0.01
60.
027
Max
imum
1303
.130
823.
70.
838
0.04
90.
244
2.30
90.
305
0.14
80.
075
0.17
1M
inim
um14
.821
8.1
0.09
50.
004
0.01
10.
054
0.01
30.
014
0.00
00.
019
Fore
ign
Mea
n12
67.1
2864
1.7
0.55
60.
018
0.04
60.
171
0.05
30.
066
0.07
30.
064
acqu
ired
Med
ian
473.
313
231.
30.
569
0.01
50.
036
0.13
30.
040
0.05
40.
053
0.05
4ba
nks
St.
Dev
.23
88.1
4398
5.1
0.17
30.
010
0.03
50.
111
0.04
30.
034
0.07
60.
030
Max
imum
2132
4.9
1930
00.0
0.84
50.
072
0.21
40.
605
0.28
80.
191
0.36
10.
185
Min
imum
18.1
340.
00.
194
0.00
60.
012
0.05
00.
013
0.02
10.
000
0.02
4T
otal
Mea
n79
3.0
1837
1.3
0.52
40.
021
0.06
30.
235
0.07
40.
072
0.08
80.
078
(all
bank
s)M
edia
n23
5.9
4583
.60.
540
0.01
90.
050
0.17
90.
060
0.06
40.
054
0.06
7St
.D
ev.
1536
.033
298.
80.
198
0.01
20.
044
0.19
70.
051
0.03
50.
100
0.03
9M
axim
um21
324.
919
3000
.00.
849
0.07
20.
324
2.30
90.
336
0.22
51.
000
0.31
0M
inim
um13
.714
5.0
0.04
20.
004
0.01
10.
050
0.01
20.
012
0.00
00.
014
Not
es:
ban
kco
sts
and
earn
ing
asse
tsar
em
easu
red
inth
ousa
nd
sof
US
dol
lars
and
defl
ated
by
the
con
sum
erp
rice
ind
ex(e
xtr
acte
dfr
omth
eW
orld
Ban
k’s
Wor
ldD
evel
opm
ent
Ind
icat
ors
dat
abas
e),
usi
ng
1995
asa
refe
ren
ceye
ar.
To
con
fron
tth
ein
flu
ence
ofex
trem
eob
serv
atio
ns
and
rep
orti
ng
erro
rs,
all
vari
able
sh
ave
bee
nw
inso
rize
dat
the
1st
and
99th
per
cen
tile
s.
106 Chapter 5
Tabl
e5.
4.Im
pact
offo
reig
nba
nkpa
rtic
ipat
ion
onco
steffi
cien
cy:
Stoc
hast
iceffi
cien
cyfr
ontie
ran
alys
is(m
odel
(5.9
))(I
)(I
I)(I
II)
(IV
)(V
)(V
I)(V
II)
(VII
I)F
ront
ier
Ear
ning
asse
ts0.
5583
***
0.56
61**
*0.
5579
***
0.56
35**
*0.
5636
***
0.59
12**
*0.
6480
***
0.61
35**
*P
rice
ofla
bor/
Pri
ceof
capi
tal
0.54
84**
*0.
5544
***
0.54
87**
*0.
5561
***
0.54
57**
*0.
5756
***
0.57
33**
*0.
5837
***
Pri
ceof
borr
owed
fund
s/P
rice
ofca
pita
l-0
.075
1-0
.077
4-0
.075
-0.0
772
-0.0
78-0
.118
2-0
.202
8**
-0.1
598
Tim
etr
end
0.00
28-0
.000
90.
0026
-0.0
018
-0.0
222
-0.0
087
-0.0
068
0.01
48(E
arni
ngas
sets
)20.
0438
***
0.04
23**
*0.
0439
***
0.04
25**
*0.
0444
***
0.04
26**
*0.
0332
**0.
0389
***
(Ear
ning
asse
ts)*
(Pri
ceof
labo
r/P
rice
ofca
pita
l)-0
.000
9-0
.000
8-0
.001
-0.0
01-0
.001
90.
0004
0.00
20.
0059
(Ear
ning
asse
ts)*
(Pri
ceof
borr
owed
fund
s/P
rice
ofca
pita
l)0.
0095
0.01
040.
0095
0.01
030.
0078
0.01
040.
0169
0.00
76(E
arni
ngas
sets
)*(T
ime
tren
d)0.
0043
**0.
0044
**0.
0043
**0.
0045
**0.
0042
**0.
0025
0.00
350.
0012
(Pri
ceof
labo
r/P
rice
ofca
pita
l)2
-0.0
327*
**-0
.034
0***
-0.0
327*
**-0
.033
8***
-0.0
357*
**-0
.051
6***
-0.0
552*
**-0
.079
3***
(Pri
ceof
labo
r/P
rice
ofca
pita
l)*(
Pri
ceof
borr
owed
fund
s/P
rice
ofca
pita
l)0.
0210
**0.
0205
**0.
0210
**0.
0204
**0.
0273
**0.
0350
***
0.03
90**
0.04
63**
*(P
rice
ofla
bor/
Pri
ceof
capi
tal)
*(T
ime
tren
d)-0
.017
7***
-0.0
175*
**-0
.017
8***
-0.0
176*
**-0
.016
0***
-0.0
150*
**-0
.013
8***
-0.0
097*
**(P
rice
ofbo
rrow
edfu
nds/
Pri
ceof
capi
tal)
2-0
.058
3**
-0.0
602*
**-0
.058
2**
-0.0
598*
*-0
.065
5***
-0.0
742*
**-0
.072
9**
-0.0
542*
*(T
ime
tren
d)2
0.00
060.
0007
0.00
060.
0007
0.00
120.
0008
-0.0
006
-0.0
023
Con
stan
t1.
3176
***
1.29
53**
*1.
3193
***
1.30
50**
*1.
4700
***
1.50
39**
*0.
9721
**1.
2934
**In
effici
ency
det
erm
inan
tsT
ime
tren
d-0
.029
0**
-0.0
250*
*-0
.028
7**
-0.0
237*
Fore
ign
gree
nfiel
d-0
.372
7***
-0.3
786*
**Fo
reig
nac
quir
ed-0
.005
-0.0
275
Fore
ign
entr
y0.
3190
***
0.16
11*
Fore
ign
entr
y(1
year
lag)
-0.3
232*
*-0
.463
7**
Fore
ign
entr
y(2
year
sla
g)-0
.070
1-0
.056
3C
onst
ant
-0.4
256*
**-0
.342
1**
-0.4
358*
**-0
.403
6-0
.773
2***
-0.7
779*
**-0
.807
6***
-0.8
525*
**S
tati
stic
sN
umbe
rof
obse
rvat
ions
2,06
72,
067
2,06
72,
067
2,06
71,
613
1,29
01,
290
Num
ber
ofpa
ram
eter
s40
4141
4240
3938
40L
oglik
elih
ood
-174
.495
8-1
68.9
294
-174
.492
3-1
68.8
235
-165
.812
-52.
6434
-64.
6417
-2.0
239
log(
σ2 u)
-0.4
128
-0.5
215
-0.3
929
-0.4
018
-0.1
867
-0.3
858
-0.2
521
-0.2
76lo
g(σ
2 v)
-2.9
981*
**-2
.996
5***
-2.9
982*
**-2
.997
4***
-3.0
042*
**-3
.101
7***
-3.0
930*
**-3
.188
5***
Not
es:
the
dep
end
ent
vari
able
isth
era
tio
ofto
tal
oper
atin
gex
pen
ses
toth
ep
rice
ofca
pit
al.
All
vari
able
s(e
xce
pt
from
the
tim
etr
end
)ar
eex
pre
ssed
inth
elo
gari
thm
icfo
rm.
Est
imat
ion
sar
ep
erfo
rmed
usi
ng
max
imu
mli
keli
ho
od
met
ho
db
ased
onth
eB
roy
den
–Fle
tch
er–G
old
farb
–Sh
ann
o(B
FG
S)
opti
miz
atio
nal
gori
thm
.σ
2 uan
dσ
2 vst
and
for
the
stan
dar
dd
evia
tion
ofth
ein
effici
ency
and
ran
dom
erro
rte
rms,
resp
ecti
vely
.E
ach
spec
ifica
tion
also
con
tain
sd
um
my
vari
able
sfo
rco
un
trie
san
dti
me
(not
show
nin
the
tab
leto
con
serv
esp
ace)
.*,
**,
and
***
den
ote
sign
ifica
nce
atth
e10
per
cen
t,5
per
cen
tan
d1
per
cen
tle
vel,
resp
ecti
vely
.
Foreign Bank Entry, Bank Efficiency, and Market Power 107
Tabl
e5.
5.Im
pact
offo
reig
nba
nkpa
rtic
ipat
ion
onm
arke
tpo
wer
(mod
el(5
.8))
(I)
(II)
(III
)(I
V)
(V)
(VI)
(VII
)M
odel
Dep
osit
rate
2.14
62**
*2.
1571
***
2.12
70**
*2.
1364
***
2.05
72**
*1.
8219
***
1.64
51**
*M
argi
nal
cost
s0.
4856
**0.
4843
**0.
4852
**0.
4840
**0.
2339
*0.
2350
*0.
3869
*In
tera
ctio
nte
rm(d
epos
itra
te×
fore
ign
gree
nfiel
ddu
mm
y)-0
.058
7-0
.050
80.
4651
0.13
340.
0771
Inte
ract
ion
term
(dep
osit
rate
×fo
reig
nac
quir
eddu
mm
y)-0
.690
0**
-0.6
897*
*-0
.372
0**
-0.4
535*
*-0
.636
1**
Non
-per
form
ing
loan
s0.
1487
**Im
plic
itin
tere
stre
venu
e1.
2013
***
Rea
lG
DP
grow
th-0
.004
4C
PI
infla
tion
0.00
08E
xcha
nge
rate
chan
ges
0.00
18*
Con
stan
t0.
0455
*0.
0456
*0.
0531
*0.
0532
*0.
0532
*0.
0015
*0.
0968
***
Mar
ket
pow
erte
stH
0:D
epos
itra
teco
effici
ent
=1
10.2
29.
329.
548.
665.
786.
873.
31(p
-val
ue)
0.00
950.
0122
0.01
150.
0147
0.03
710.
0256
0.09
87H
0:D
epos
itra
teco
effici
ent
+In
tera
ctio
nte
rm(d
epos
itra
tean
dfo
reig
ngr
eenfi
eld
dum
my)
=1
–5.
32–
5.37
7.13
4.14
3.74
(p-v
alue
)–
0.04
38–
0.04
300.
0235
0.06
910.
0820
H0:
Dep
osit
rate
coeffi
cien
t+
Inte
ract
ion
term
(dep
osit
rate
and
fore
ign
acqu
ired
dum
my)
=1
––
1.03
1.17
2.38
1.72
0.00
(p-v
alue
)–
–0.
3341
0.30
530.
1543
0.21
870.
9796
Sta
tist
ics
Num
ber
ofob
serv
atio
ns1,
988
1,98
81,
988
1,98
81,
615
1,98
81,
966
R2
0.21
780.
2172
0.22
450.
2241
0.24
930.
2555
0.26
68L
og-l
ikel
ihoo
d13
05.8
1305
.913
14.4
1314
.412
02.8
1381
.113
47.8
Hau
sman
test
(p-v
alue
)0.
0255
0.03
710.
0153
0.02
500.
0007
0.04
470.
0088
Pan
elau
toco
rrel
atio
nte
st(p
-val
ue)
0.15
390.
1948
0.13
980.
2436
0.17
430.
2195
0.12
98N
otes
:th
ed
epen
den
tva
riab
leis
the
rati
oof
tota
lin
tere
stex
pen
ses
toto
tal
loan
s.E
stim
atio
ns
are
per
form
edu
sin
gfi
xed
effec
tsm
eth
od
wit
hb
oot
srap
ped
stan
dar
der
rors
usi
ng
2000
rep
lica
tion
s.S
tan
dar
der
rors
are
esti
mat
edu
sin
gre
sid
ual
scl
ust
ered
by
cou
ntr
yto
allo
wfo
rp
ossi
ble
inte
rdep
end
ence
bet
wee
nb
ank
slo
cate
din
the
sam
eco
un
try.
Th
eH
ausm
ante
stte
sts
the
nu
llh
yp
oth
esis
that
ran
dom
effec
tsm
od
elis
con
sist
ent
and
effici
ent
(i.e
.,n
osy
stem
atic
diff
eren
ceb
etw
een
coeffi
cien
tes
tim
ates
for
the
fix
edeff
ects
and
ran
dom
effec
tsm
od
els)
.P
anel
auto
corr
elat
ion
test
(nu
llh
yp
oth
esis
:n
ofi
rst
ord
erau
toco
rrel
atio
n)
isb
ased
onth
ep
roce
du
reof
Dru
kke
r(2
003)
.M
arke
tp
ower
hy
pot
hes
esar
ete
sted
usi
ng
Wal
dte
stst
atis
tic.
*,**
,an
d**
*d
enot
esi
gnifi
can
ceat
the
10p
erce
nt,
5p
erce
nt,
and
1p
erce
nt
leve
l,re
spec
tive
ly.
Chapter 6
Re-examining the Impact ofForeign Bank Participationon Interest Margins
6.1 Introduction
In the absence of developed bond and stock markets, banks continue to play a major
role as financial intermediaries in former socialist economies (FSEs) (Berglof and
Bolton, 2002; Bonin et al., 1998; Bonin and Wachtel, 2003). As a result, the costs
of financial intermediation services offered by banks remain crucial for the economic
development of FSEs. The observed massive increase of foreign bank participation
during the last decade inevitably raises the question to what extent foreign entry
has influenced bank interest margins, which is a commonly used measure of financial
intermediation costs offered by banks.
There is an established theoretical literature on the determinants of interest
margins initiated by the dealership model of Ho and Saunders (1981). This model
assumes that bank serves as a risk-averse dealer in the deposit and loan markets,
bearing the risk of refinancing due to the possible mismatch between the arrival of
110 Chapter 6
deposits and demand for loans. This mismatch is dealt with by the bank through
its activities in the money market, which creates a link between the optimal level of
the net interest margin set by the bank and the volatility of the money market rate
(the market risk). Some simplifying assumptions of the Ho and Saunders (1981)
model were later on relaxed by introducing heterogeneous bank products (Allen,
1988), credit risk (Angbanzo, 1997), and operating costs (Maudos and Fernandez de
Guevara, 2004) as important additional determinants of the bank interest margin.
The most recent development of the bank dealership model is provided by the model
of Maudos and Fernandez de Guevara (2004), in which the set of theoretically moti-
vated determinants of the net interest margin includes market structure, operating
costs, managerial risk aversion, credit and market risks, and the size of bank opera-
tions.
A notable feature of the dealership model is that foreign ownership is not consid-
ered to be a determinant of interest margins. This is in sharp contrast to a different
stream of theoretical literature, which underscores the problem of asymmetric in-
formation between entrant (foreign) and incumbent (domestic) banks that might
influence the margin. Foreign banks have better screening technologies to identify
good borrowers based on hard information, while domestic banks possess superior
soft information (Dell’Ariccia and Marquez, 2004). Differences in information dis-
tribution may result in a cream-skimming caused by foreign entry: in equilibrium
foreign banks would focus on providing services to less risky and large borrowers,
while domestic banks would concentrate their lending to more opaque and small
firms (Sengupta, 2007).1
Generally speaking, foreign entry can influence banks in host countries through
1 Depending on the relative strength of the two opposite effects, the host countries can evenexperience a decline in total lending following foreign bank entry, which has been empiricallydocumented in some less developed countries (Detragiache et al., 2008).
Re-examining the Impact of Foreign Bank Participation on Interest Margins 111
various direct and indirect channels (Lehner and Schnitzer, 2008). One possible
channel is spillover effects from foreign to domestic banks in terms of better screening
facilities, technology utilization, and transfer of know-how. These indirect benefits
from increased foreign bank participation should result in lower average unit costs
associated with the financial intermediation process, reflected in lower equilibrium
margins. Another possible channel is the increase in competition due to opening up
of the banking market for foreign competitors. The mode of foreign entry (acquisition
versus greenfield investment) has important implications in this respect. While
greenfield investments increase the number of banks in the economy, entry through
foreign acquisition only affects ownership distribution of existing banks and does not
influence the total number of banks. Therefore, theoretically, the entry via foreign
greenfield investments should result in more competition than the entry via foreign
acquisition.2 In addition, the advantage of acquisition over greenfield entry is that
the foreign bank acquires information about the quality of incumbent borrowers
using the credit information inherited from the target bank. The average quality
of incumbent borrowers may influence the lending rate demanded by the acquired
banks for extending new loans, giving rise to the portfolio composition effect (Claeys
and Hainz, 2007).
Surprisingly, this apparent contradiction between the predictions of the dealer-
ship model and the other stream of theoretical literature has not been examined in
previous empirical studies analyzing the impact of foreign bank participation on in-
terest margins. Most of these studies took an ad hoc approach by analyzing various
determinants that are likely to affect bank interest margins (some of which partially
overlap with the theoretically motivated determinants of the dealership model). The
2 Although in theory the number of banks and market concentration are considered to be im-portant determinants of the level of competition, empirical studies do not find support for thisargumentation (Claessens and Laeven, 2004).
112 Chapter 6
impact of foreign ownership is commonly estimated by introducing a dummy vari-
able for foreign-owned banks (direct effect due to the magnitude of margins set by
foreign banks) and/or a country-wide measure of foreign bank participation, such as
the market share of foreign-owned banks (indirect effect due to spillovers).
Based on this approach, the empirical literature provides mixed evidence on the
impact of foreign bank participation on interest margins in emerging economies.
Among cross-country studies, Demirguc-Kunt and Huizinga (2000) found that for-
eign bank participation had a positive effect on interest margins in a worldwide
sample of 80 countries during 1988-1995. Schwaiger and Liebeg (2008) came to a
similar conclusion using a sample of 11 FSEs during 2000-2005. In contrast, the im-
pact of foreign entry was found to be negative in 5 Latin American countries during
1995-2000 (Martinez Peria and Mody, 2004), in 11 FSEs during 1993-1999 (Drakos,
2003), and in 13 FSEs during 1994-2001 (Claeys and Vander Vennet, 2008).3 The
evidence is also mixed in single-county studies: Dabla-Norris and Floerkmeier (2007)
did not find any significant association between foreign ownership and interest mar-
gins in Armenia, whereas Denizer (2000) and Barajas et al. (2000) found that foreign
entry has driven down interest margins in Turkey and Colombia, respectively. All
in all, due to the absence of a unified theoretical framework and inconclusive empir-
ical evidence, the overall impact of foreign bank participation on interest margins
remains unclear.
The aim of this chapter is to fill this gap in the literature by re-examining the
empirical relationship between foreign bank participation and interest margins using
a more formal approach. Unlike most of the previous studies, we try to account for
theoretically motivated determinants of (the most advanced version of) the dealer-
3 In Martinez Peria and Mody (2004), the decrease is largely attributed to the participation ofgreenfield foreign banks, whereas indirect effects due to foreign bank participation were found toplay a crucial role in Claeys and Vander Vennet (2008).
Re-examining the Impact of Foreign Bank Participation on Interest Margins 113
ship model by Maudos and Fernandez de Guevara (2004) and the other stream of
literature theorizing on the impact of foreign bank participation on interest margins.
Careful analysis of the later literature suggests that most of the channels through
which foreign bank participation is expected to influence the margins are already ac-
counted for by the dealership model. For instance, Martinez Peria and Mody (2004)
argue that one of the channels through which increased foreign bank participation
can affect the margins is its impact on the cost of operations. However, the em-
pirical specification inspired by the dealership model already includes this variable
among interest margin determinants. Similarly, Bonin et al. (2005) and Lehner and
Schnitzer (2008) argue that foreign banks are able to charge lower margins due to
their superior efficiency. However, cost efficiency is taken into account by the deal-
ership model as determinant of the margins, too. Lastly, Claeys and Hainz (2007)
hypothesize that the possible negative impact of foreign bank participation may be
due to the portfolio effect, since foreign banks tend to be largely involved in financing
relatively safer clients. The dealership model, however, also considers the riskiness
of bank’s portfolio as an important factor influencing margins.
As a result, we conclude that there is no particular reason to expect that foreign
bank participation affects bank interest margins after the theoretically motivated
determinants of the dealership model are fully taken into account in the empirical
specification. Our empirical analysis supports this conclusion, as we find that after
controlling for the theoretically motivated determinants described in the dealership
model, various indicators of foreign bank participation (such as dummy variables
for greenfield and acquired foreign banks, a country-wide measure of foreign bank
participation) do not elicit a significant impact on interest margins. Intuitively, this
result suggests that both direct and indirect channels, through which the impact
of foreign bank participation on margins is expected to materialize (e.g., market
114 Chapter 6
structure), are fully accounted for by the dealership model. Our findings call for
re-examination of some of the previous studies, in which foreign bank participation
was found to have a significant own impact on interest margins.
The remainder of this chapter is structured as follows. Section 6.2 describes the
empirical methodology and data. Section 6.3 presents the estimation results and
their discussion. The last section concludes.
6.2 Methodology and Data
6.2.1 Empirical model
We estimate the dealership model using a fixed effect estimator to capture unob-
served heterogeneity at the individual bank level. The Maudos and Fernandez de
Guevara (2004) model is taken as a baseline specification, which we augment by in-
troducing two measures of foreign participation at the individual bank-level (foreign
greenfield banks and banks that entered through cross-border acquisitions) and one
measure at the country level (market share of foreign banks). We test the robust-
ness of our results regarding the impact of foreign participation by adding several
macroeconomic variables.
The general specification takes the following form:
Marginijt = αi +N
∑n=1
βnTheoreticalnijt−1 +M
∑m=1
γmEnvironmentalmijt−1 + (6.1)
+ λ1 ∗ DGF + λ2 ∗ DA + λ3 ∗ ForeignSharejt + Macrojt + DYEAR + εijt
where i, j, and t indices stand for bank, country, and time, respectively, Margin
is the interest margin, Theoretical and Environmental are vectors of bank-specific
(pure margin determinants) and environmental variables as defined in Maudos and
Fernandez de Guevara (2004), DGF is a dummy variable for greenfield foreign banks,
DA is a dummy variable for acquired foreign banks, ForeignShare is a percentage of
Re-examining the Impact of Foreign Bank Participation on Interest Margins 115
banking system assets in the country controlled by the foreign-owned banks, Macro
is a set of macroeconomic control variables, and εijt is an i.i.d. random error. The
individual bank heterogeneity is captured by the fixed effects intercept term αi and
the time-specific variation is captured by a vector of time dummies DYEAR.
Table 6.1 provides a description of all variables and their sources. The net interest
margin is measured as the ratio of the net interest income over total earning assets.
We use the following pure margin determinants in our estimations (see Maudos and
Fernandez de Guevara, 2004). Market structure is captured by the Herfindahl index
measured as the sum of squares of individual bank market shares for each country.4
Operating costs are measured as a ratio of operating expenses to total assets. Risk
aversion is proxied by the equity-to-total assets ratio, implying higher risk aversion
for banks having higher ratios. Market risk is captured by the standard deviation
of monthly interbank money market rates.5 Credit risk is measured by the ratio
of loan loss provisions to net loans.6 The interaction of market and credit risk is
controlled for by introducing the interaction term of the above two risk measures
into the specification. The size of operations is captured by the logarithm of net
loans.
Furthermore, we control for environmental factors influencing interest margins
using three variables. Implicit interest payments are measured by the ratio of oper-
ating expenses net of non-interest revenues to total assets. Higher implicit interest
payments should be compensated by an increase in interest margins. Opportunity
costs of bank reserves are measured by the ratio of liquid assets to total assets. More
4 Total assets are used as a measure of banking activity.5 In the absence of money market rates for some of the FSEs, the government T-Bill rates are used
as a measure of market rates.6 Due to a large amount of missing data, we cannot proxy credit risk by the ratio of non-performing
loans to total assets. Although a second best option, our measure of credit risk is still an improve-ment compared to the ratio of loans to total assets used by Maudos and Fernandez de Guevara(2004).
116 Chapter 6
liquid banks are expected to have higher margins in order to compensate for oppor-
tunity costs of holding extra liquidity. Finally, the managerial quality is proxied by
the cost-to-income ratio. Banks having a more qualified management are expected
to decrease interest margins due to lower cost-to-income ratio.
The model with the aforementioned theoretically-motivated and environmental
variables is based on the specification used in Maudos and Fernandez de Guevara
(2004), in which there is no role for the impact of the ownership structure on bank
interest margins. To test for the impact of foreign bank presence, we augment the
model by including proxies for foreign bank participation that are hypothesized to
affect the margin through a set of direct and indirect channels. By introducing the
DGF and DA dummies it is tested whether the average margins for foreign banks
(new and acquired) are significantly different from the average margin of the rest
of the banking institutions. By introducing ForeignShare variable, we test whether
there is a spillover effect arising from the presence of foreign banks in the banking
systems of host countries. That is, we test whether the overall level of foreign bank
participation in the banking system raises or lowers the margin after controlling for
individual bank ownership effects.
Given that the differences in margins across countries may be affected by the
macroeconomic environment in which banks operate, we control for the following
commonly used variables to check the robustness of our results. GDPPC is per
capita GDP in US dollars and GDPGR is the real GDP growth rate for each of the
countries capturing the influence of the level of economic development and economic
growth on interest margins, respectively. Inflation is the CPI-based inflation rate.7
7 In a separate set of regressions, we also included institutional characteristics of countries proxiedby the arithmetic average of EBRD indices covering small- and large-scale privatization, enterprisereforms, price liberalization, forex and trade liberalization, competition policy, banking and non-banking sector reforms, and reforms in infrastructure as an additional control variable. We obtainedinsignificant coefficients, probably reflecting that the institutional characteristics of the CEECs inour sample are relatively homogenous.
Re-examining the Impact of Foreign Bank Participation on Interest Margins 117
In order to avoid simultaneity problems, we take lagged values of the theoretically-
motivated and environmental variables. A bias due to simultaneity can arise when
dependent and independent variables are contemporaneously related due to an ac-
counting identity or via a functional form. Using lagged values of independent
variables rules out the possibility of a simultaneous interaction, as the independent
variables become predetermined with respect to the dependent variable.8
6.2.2 Data
We combine information from different data sources for our analysis. The main data
source is the BankScope database of Bureau van Dijk, from which we extract infor-
mation on individual bank balance sheets and profit and loss accounts. Our sample
is an unbalanced panel of 2,044 observations for 387 commercial, cooperative, and
savings banks from 11 CEECs for the period 1995-2006.9 Since BankScope provides
information only on current ownership of banks, we complement this data set by col-
lecting historical information on foreign ownership from different sources. First, we
use information on foreign-owned banks from the extended data set of De Haas and
Van Lelyveld (2006) employed in Havrylchyk and Jurzyk (2008). The data set covers
the period 1995-2004 and categorizes foreign-owned banks into two groups: green-
field establishments and banks taken over as a result of a cross-border acquisition.
Next, for the remaining two years, we obtain a list of cross-border bank takeovers
from the Securities Data Company (SDC) mergers and acquisitions database pro-
duced by Thompson Financial. We identify 8 cross-border bank acquisition events
that led to a transfer of bank control from domestic to foreign ownership (at least 50
8 We obtain qualitatively similar results with respect to the impact of foreign bank participation oninterest margins when the current values of the theoretically-motivated and environmental variablesare used in the estimations. Using the lagged variables only influences coefficient estimates oftheoretically motivated and environmental variables, while the impact of foreign bank participationremains unaffected.9 Our sample comprises Bulgaria, Croatia, the Czech Republic, Estonia, Hungary, Latvia, Lithua-
nia, Poland, Romania, Slovakia, and Slovenia.
118 Chapter 6
percent of capital) during 2005-2006. Finally, the aforementioned bank-level infor-
mation is complemented by country level information on the share of foreign-owned
banks in total banking assets from the EBRD Transition Report (EBRD, 2007). Our
macroeconomic variables - per capita GDP, GDP growth rates and consumer prices
- are taken from the World Development Indicators database (see Table 6.1).
Table 6.2 shows descriptive statistics of the net interest margin and its deter-
minants for the total sample, as well as for subsamples of domestic and foreign
banks. The average margin is about 4.2% but it has a large variation as shown by
its wide range. The magnitude of the margin is on average lower for the sample of
domestic banks, compared to foreign banks. This indicates that foreign banks are
charging a lower margin than domestic banks, suggesting a negative direct effect
of foreign bank participation on the margin. However, summary statistics of both
theoretically-motivated and environmental determinants of the margin suggest that
this variation can be explained by differences in variables influencing the margin.
For instance, foreign banks incur lower operating costs than domestic banks and the
credit portfolio of foreign banks is characterized by lower risks in comparison to the
credit portfolio of domestic banks.
6.3 Estimation Results
Table 6.3 presents estimation results for the reference and augmented dealership
models. All estimations are performed using the fixed effects estimator, which is
superior to the random effects estimator according to the Hausman test. We do
not present the coefficient estimates for time dummies to save space and keep the
discussion focused.
Re-examining the Impact of Foreign Bank Participation on Interest Margins 119
6.3.1 The reference model
We start by fitting the model of Maudos and Fernandez de Guevara (2004) as ref-
erence specification. In this model, some of the theoretically-motivated variables
determining the margin have a significant impact and the expected sign. Interest
margins are higher for banks incurring greater operational expenses and more risk,
as well as for banks characterized by greater risk aversion. Similar to the finding of
Maudos and Fernandez de Guevara (2004) for selected EU countries, we find that
interest margins increase with the size of operations, presumably reflecting compen-
sation for a possibility of larger losses per operation due to greater stakes. However,
contrary to Maudos and Fernandez de Guevara (2004), we do not find a significant
impact for market concentration. This result might imply that in CEECs, the impact
of bank-specific characteristics outweighs the importance of the market structure.
Although the individual impact of market and credit risks come out insignificant,
their interaction term has a negative significant impact on the margin. This sug-
gests that the impact of the credit risk on the margins is amplified by the level of
the market risk, and vice versa. The negative sign is in contrast to the theoretical
expectation and suggests that CEECs banks are unable to value their risks properly.
For the environmental variables, we find a negative association between implicit
interest payments and margins. Banks holding greater liquid reserves compensate
their alternative costs by setting higher margins. Likewise, the cost-to-income ratio
has a significantly positive impact, reflecting that more cost inefficient banks charge
higher margins.
6.3.2 The impact of foreign bank participation
In order to evaluate the indirect impact of foreign bank participation on interest mar-
gins, in specification (II) we include the market share of foreign banks as additional
120 Chapter 6
explanatory variable.10 Our estimations do not support the hypothesis that foreign
bank participation has significant spill-over effects, when theoretically-motivated and
environmental variables are controlled for.
Specification (III) tests for the direct impact of foreign bank participation on
interest margins. The dummy variable for foreign-owned banks is not significant,
implying no significant own effect above the theoretically-motivated and environ-
mental determinants. Since theoretical models of foreign bank entry underscore the
importance of the mode of entry, in specifications (IV) and (V) we split the foreign
ownership dummy variable into two components: a dummy variable for greenfield
foreign banks and a dummy variable for acquired foreign banks. Our estimations
suggest that different modes of entry do not significantly influence interest margins,
after controlling for the impact of the theoretically-motivated and environmental de-
terminants. The impact remains insignificant when both dummy variables enter the
specification simultaneously (column VI) and together with the measure of indirect
impact of foreign bank participation (column VII).
Finally, in specification (VIII) we control for the impact of macroeconomic vari-
ables as additional explanatory variables influencing the margin. This does not
change our conclusion regarding the insignificant direct and indirect impact of for-
eign bank participation on the interest margin. We find that the margin is lower in
relatively more developed countries (negative and significant coefficient of per capita
GDP), while the impact of economic growth is insignificant. The margins increase
with the level of inflation, probably reflecting additional price uncertainty risk. It is
also important to note that introducing the macroeconomic variables wipes out the
impact of the market and credit risks interaction dummy, while the direct impact of
the market risk variable becomes significant.
10 This variable was also used as a measure of spill-over effects from foreign bank participation tomargins in Latin American economies by Martinez Peria and Mody (2004).
Re-examining the Impact of Foreign Bank Participation on Interest Margins 121
6.3.3 Economic significance
So far, we have focused on statistical significance only. In this section, we analyze
the economic relevance of the determinants of interest margins. Table 6.4 presents
the economic impact of interest margin determinants, measured as a response of
the interest margin in percentages to a one percentage change in its determinants
based on specification (VIII). The results suggest that among the theoretically-
motivated determinants, the most substantive impact comes from the size of banking
operations (1.25 percentage points) and the size of operating costs (0.25 percentage
points). Among the environmental variables, the economic impact of implicit interest
payments (0.11 percentage points) and cost inefficiency (0.09 percentage points)
are comparable in size. Finally, among the macroeconomic variables, the strongest
impact comes from the level of economic development of the country measured by
the per capita GDP (-9.6 percentage points).
The analysis of the relative impact of these variables suggests that the insignif-
icant impact of the foreign participation may be explained by the fact that all the
channels through which foreign participation may affect margins are already ac-
counted for in the dealership model. The insignificant own impact of foreign bank
participation calls for reassessment of previous findings on the impact of foreign
bank participation on interest margins.
6.4 Conclusions
This chapter has re-examined the impact of foreign bank participation on interest
margins using the recent sharp increase of foreign bank presence in CEECs as a lab-
oratory experiment. We start by observing that the dealership model widely used
in empirical work to provide a quantitative assessment of factors driving the margin
122 Chapter 6
does not allow for the impact of foreign bank participation to be explicitly tested.
The mechanisms through which foreign bank participation may influence bank be-
havior and ultimately the margin are analyzed by other models in a framework
different from the dealership model. However, the majority of these mechanisms,
like market concentration, riskiness of bank portfolio, and operational costs, are
already taken into account by the margin determinants inspired by the dealership
model. This raises the question of whether the foreign bank participation has its
own direct and/or indirect impact on interest margins.
Previous empirical studies that addressed this question have produced mixed
results. Most of the studies report a negative effect, suggesting that foreign par-
ticipation helps to decrease the margin due to spillover effects and portfolio mix of
foreign banks (see, for example, Martinez Peria and Mody, 2004), while others did
not find any significant impact, or even reported a positive impact (see, for example,
Schwaiger and Liebeg, 2008). The mixed results in these studies can be explained
by differences in the coverage of theoretical determinants inspired by the dealership
model.
Using data on domestic and foreign-owned banks in 11 CEECs, we show that
after fully accounting for all interest margin determinants inspired by the dealership
model, foreign bank participation does not have any significant impact on interest
margins in CEECs. The impact remains insignificant when we differentiate between
proxies for indirect (foreign bank market share) and direct (dummy variables for
greenfield and acquired foreign banks) effects of foreign bank presence. We explain
this finding by the fact that the variables inspired by the dealership model already
account for the main mechanisms through which the impact of foreign bank partic-
ipation on the margins may be materialized. Our results call for a reassessment of
results reported in some of the previous studies, which suggest a direct impact of
foreign bank participation.
Re-examining the Impact of Foreign Bank Participation on Interest Margins 123
Tabl
e6.
1.Va
riabl
ede
finiti
onan
dso
urce
sV
aria
ble
Mea
sure
Sour
ceN
etin
tere
stm
argi
nR
atio
ofto
tali
nter
estr
even
uesn
etof
tota
lint
eres
tex
pens
esto
tota
lass
ets
Ban
kSco
pe
Mar
ket
conc
entr
atio
nH
erfin
dahl
inde
x(t
otal
asse
ts)
Ban
kSco
peO
pera
ting
cost
sR
atio
ofto
talo
pera
ting
expe
nses
toto
tala
sset
sB
ankS
cope
Ris
kav
ersi
onR
atio
ofto
tale
quity
toto
tala
sset
sB
ankS
cope
Mar
ket
risk
Stan
dard
devi
atio
nof
mon
thly
mon
eym
arke
tra
tes
Inte
rnat
iona
lFin
anci
alSt
atis
tics
(IM
F)
Cre
dit
risk
Rat
ioof
loan
loss
prov
isio
nsto
tota
lloa
nsB
ankS
cope
Size
ofop
erat
ions
Loga
rithm
ofto
tall
oans
Ban
kSco
peIm
plic
itin
tere
stpa
ymen
tsR
atio
ofop
erat
ing
expe
nses
net
ofno
n-in
tere
stre
venu
esto
tota
lass
ets
Ban
kSco
pe
Opp
ortu
nity
cost
sof
bank
rese
rves
Rat
ioof
liqui
dre
serv
esto
tota
lass
ets
Ban
kSco
peC
ost
ineffi
cien
cyR
atio
ofto
talc
osts
toto
tali
ncom
eB
ankS
cope
Mar
ket
shar
eof
fore
ign
bank
sR
atio
ofto
tala
sset
sco
ntro
lled
byfo
reig
n-ow
ned
bank
sto
tota
lban
king
syst
emas
sets
EB
RD
Tran
sitio
nR
epor
t
Fore
ign
bank
dum
my
Dum
my
varia
ble
that
take
sva
lue
of1
for
fore
ign
bank
s(b
oth
gree
nfiel
dan
dac
quire
d)D
eHaa
sand
Van
Lely
veld
(200
6),H
avry
lchy
kan
dJu
rzuk
(200
8)Fo
reig
ngr
eenfi
eld
bank
dum
my
Dum
my
varia
ble
that
take
sva
lue
of1
for
gree
n-fie
ldes
tabl
ishm
ents
offo
reig
nba
nks
DeH
aasa
ndVa
nLe
lyve
ld(2
006)
,Hav
rylc
hyk
and
Jurz
uk(2
008)
Fore
ign
acqu
ired
bank
dum
my
Dum
my
varia
ble
that
take
sval
ueof
1fo
rdom
estic
bank
sac
quire
dby
afo
reig
nba
nkD
eH
aas
and
Van
Lely
veld
(200
6),
Hav
rylc
hyk
and
Jurz
uk(2
008)
and
Tho
mso
n’s
SDC
Pla
tinum
Dat
abas
eE
cono
mic
deve
lopm
ent
Loga
rithm
ofG
DP
per
capi
ta(U
Sdo
llars
)W
orld
Dev
elop
men
tIn
dica
tors
(Wor
ldB
ank)
Eco
nom
icgr
owth
Rea
lGD
Pgr
owth
rate
Wor
ldD
evel
opm
ent
Indi
cato
rs(W
orld
Ban
k)In
flatio
nPe
rcen
tage
chan
gein
cons
umer
pric
ein
dex
Wor
ldD
evel
opm
ent
Indi
cato
rs(W
orld
Ban
k)
124 Chapter 6
Table 6.2. Descriptive statisticsMean Median Standard
deviationMaximum Minimum
Domestic banksNet interest margin 0.045 0.040 0.024 0.002 0.196Market concentration 0.158 0.129 0.070 0.084 0.473Operating costs 0.062 0.054 0.034 0.007 0.272Risk aversion 0.133 0.111 0.088 0.012 0.658Market risk 0.024 0.013 0.038 0.001 0.296Credit risk 0.036 0.019 0.052 0.000 0.574Size of operations 11.739 11.653 1.639 7.436 15.565Implicit interest payments -0.014 -0.013 0.026 -0.125 0.123Opportunity costs of bank re-serves
0.052 0.034 0.050 0.000 0.280
Cost inefficiency 0.851 0.804 0.372 0.160 3.999Foreign banksNet interest margin 0.037 0.031 0.026 0.003 0.185Market concentration 0.146 0.123 0.069 0.084 0.473Operating costs 0.048 0.039 0.031 0.010 0.237Risk aversion 0.123 0.101 0.082 0.021 0.612Market risk 0.016 0.009 0.028 0.001 0.296Credit risk 0.018 0.010 0.030 0.000 0.278Size of operations 12.439 12.500 1.557 7.787 15.560Implicit interest payments -0.011 -0.012 0.020 -0.112 0.100Opportunity costs of bank re-serves
0.040 0.025 0.045 0.000 0.264
Cost inefficiency 0.823 0.773 0.297 0.156 2.954Total sampleNet interest margin 0.042 0.036 0.025 0.002 0.196Market concentration 0.154 0.128 0.070 0.084 0.473Operating costs 0.057 0.049 0.034 0.007 0.272Risk aversion 0.130 0.106 0.086 0.012 0.658Market risk 0.021 0.011 0.035 0.001 0.296Credit risk 0.030 0.015 0.046 0.000 0.574Size of operations 11.987 11.968 1.644 7.436 15.565Implicit interest payments -0.013 -0.012 0.024 -0.125 0.123Opportunity costs of bank re-serves
0.048 0.031 0.049 0.000 0.280
Cost inefficiency 0.841 0.793 0.347 0.156 3.999Notes: all variables are measured in thousands of US dollars and deflated by the consumer price index, using 1995 as areference year. Each variable is winsorized at the 1st and 99th percentiles, to confront the influence of outliers and reportingmistakes.
Re-examining the Impact of Foreign Bank Participation on Interest Margins 125
Tabl
e6.
3.Es
timat
ion
resu
lts:
Doe
sfo
reig
nba
nkpa
rtic
ipat
ion
affec
tin
tere
stm
argi
ns?
(I)
(II)
(III
)(I
V)
(V)
(VI)
(VII
)(V
III)
The
oret
ical
ly-m
otiv
ated
dete
rmin
ants
Mar
ket
conc
entr
atio
n-0
.014
2**
-0.0
132*
*-0
.014
3**
-0.0
142*
*-0
.014
4**
-0.0
144*
*-0
.013
4**
-0.0
097*
Ope
ratin
gco
sts
0.35
82**
*0.
3578
***
0.35
77**
*0.
3582
***
0.35
73**
*0.
3573
***
0.35
71**
*0.
3397
***
Ris
kav
ersi
on0.
0465
***
0.04
66**
*0.
0463
***
0.04
65**
*0.
0463
***
0.04
63**
*0.
0464
***
0.04
26**
*M
arke
tris
k0.
0223
0.02
170.
0222
0.02
230.
0221
0.02
210.
0215
0.02
71*
Cre
dit
risk
0.01
300.
0132
0.01
330.
0130
0.01
360.
0136
0.01
380.
0036
Inte
ract
ion
term
(Mar
ket
risk*
Cre
dit
risk)
-0.2
868*
*-0
.288
8**
-0.2
882*
*-0
.286
8**
-0.2
888*
*-0
.288
8**
-0.2
905*
*0.
0681
Size
ofop
erat
ions
0.00
24**
*0.
0024
***
0.00
24**
*0.
0024
***
0.00
24**
*0.
0024
***
0.00
24**
*0.
0034
***
Env
iron
men
talf
acto
rsIm
plic
itin
tere
stpa
ymen
ts-0
.491
0***
-0.4
904*
**-0
.492
0***
-0.4
910*
**-0
.492
4***
-0.4
924*
**-0
.491
8***
-0.4
720*
**Li
quid
ity0.
0109
0.01
010.
0111
0.01
090.
0112
0.01
120.
0104
0.01
21C
ost
ineffi
cien
cy0.
0037
**0.
0037
**0.
0038
**0.
0037
**0.
0038
**0.
0038
**0.
0037
**0.
0034
**Fo
reig
now
ners
hip
inba
nkin
gM
arke
tsh
are
offo
reig
nba
nks
0.00
250.
0024
-0.0
025
Fore
ign
bank
dum
my
0.00
10Fo
reig
ngr
eenfi
eld
bank
dum
my
0.00
000.
0000
0.00
000.
0000
Fore
ign
acqu
ired
bank
dum
my
0.00
130.
0013
0.00
110.
0015
Mac
roec
onom
icva
riab
les
GD
Ppe
rca
pita
(US
dolla
rs)
-0.0
456*
**R
ealG
DP
grow
thra
te-0
.000
1In
flatio
n(c
onsu
mer
pric
es)
-0.0
000*
**In
terc
ept
-0.0
152*
-0.0
289*
*-0
.027
3**
-0.0
152*
-0.0
156*
*-0
.015
6**
-0.0
296*
*0.
3550
***
Stat
istic
sN
umbe
rof
obse
rvat
ions
2,03
92,
039
2,03
92,
039
2,03
92,
039
2,03
92,
039
Log-
likel
ihoo
d63
81.0
6382
.063
81.3
6381
.063
81.6
6381
.663
82.5
6439
.7R
20.
5764
0.57
600.
5751
0.57
640.
5743
0.57
430.
5741
0.36
07N
otes
:th
ed
epen
den
tva
riab
leis
the
net
inte
rest
mar
gin
.E
stim
atio
ns
are
per
form
edu
sin
gth
efi
ced
-eff
ects
OL
Ses
tim
ator
.E
ach
spec
ifica
tion
also
con
tain
sa
set
ofd
um
my
vari
able
sto
con
trol
for
tim
efi
xed
effec
ts(n
otsh
own
inth
eta
ble
toco
nse
rve
the
spac
e).
*,**
,an
d**
*d
enot
esi
gnifi
can
ceat
the
10p
erce
nt,
5p
erce
nt,
and
1p
erce
nt
leve
l,re
spec
tive
ly.
126 Chapter 6
Table 6.4. Economic significance of interest margin determinantsCoefficient P-value
Market concentration 0.0309 0.1800Operating costs 0.2537 0.0000Risk aversion 0.0751 0.0030Market risk -0.0351 0.0020Credit risk -0.0032 0.6740Interaction term (Market risk*Credit risk) -0.0042 0.1220Size of operations 1.2490 0.0000Implicit interest payments 0.1061 0.0000Liquidity 0.0176 0.1740Cost inefficiency 0.0931 0.0140Market share of foreign banks -0.0001 0.9990Foreign greenfield bank dummy 0.0000 0.9190Foreign acquired bank dummy 0.0090 0.2260GDP per capita (US dollars) -9.5853 0.0000Real GDP growth rate 0.0049 0.8480Inflation (consumer prices) 0.1090 0.0000Notes: reported are economic significance results from specification (VII) in Table 6.3. The coefficientsreflect percentage point changes in the interest margin in response to a 1 percent change in correspondingdeterminants.
Chapter 7
Concluding Remarks
7.1 Main Findings
During the last two decades, the financial landscape around the world has under-
gone dramatic changes following a wave of financial liberalization, globalization,
and removal of restrictions on cross-border banking activities. Motivated by these
developments in international banking, this thesis analyzes the impact of foreign
bank participation on banking systems in host countries. In particular, the thesis
addresses the following research questions:
• What motivates banks to expand their activities internationally?
• What is the impact of foreign bank participation on the performance and
competition of banking systems in host countries?
• Does the mode of foreign entry matter for the post-entry performance of banks?
• How does increased foreign bank participation affect the costs of financial
intermediation?
The key challenge in analyzing these research questions is that the theoretical studies
provide contrasting predictions regarding the ultimate impact of foreign bank par-
128 Chapter 7
ticipation on banking systems in host countries. Empirical investigations are also
plagued with a number of difficulties, such as scarcity of adequate data, different
macroeconomic and institutional characteristics of host countries, sample-selection
issues related to the decision of banks to go abroad. This thesis tries to tackle these
empirical challenges by: (i) using bank-level data on FSEs that have experienced a
substantial increase of foreign bank participation during the last two decades, (ii)
applying innovative empirical methodologies to confront difficulties associated with
the empirical assessment of the impact of foreign bank participation.
Chapter 2 analyzes the impact of foreign bank participation on bank performance,
focusing on the impact of sample-selection on the decision of foreign banks to go
abroad. In particular, the chapter examines whether the positive impact of foreign
ownership on the efficiency of banks in FSEs documented in previous studies (Bonin
et al., 2005, Fries and Taci, 2005, Yildirim and Philippatos, 2007) may be biased
due to the cream-skimming effect.1 Using a two-step approach (Heckman, 1979),
we come up with new evidence suggesting that foreign banks tend to acquire good
performing banks when expanding abroad. We further show that after controlling for
the sample selection, the positive impact of foreign ownership on bank performance
documented in previous studies vanishes. In addition, our results suggest that those
FSEs that have attracted more foreign direct investment into their banking sectors
are characterized by a lower level of bank efficiency. These findings underscore the
importance of exercising care in drawing conclusions regarding the impact of foreign
ownership on bank performance in the presence of sample selection problems.
Chapter 3 provides further evidence on the motives driving banks to expand their
activities internationally. We build on the previous literature that distinguishes be-
1 The cream-skimming effect suggests that foreign banks select best performing banks for ac-quisition, which complicates the empirical analysis of the impact of foreign ownership on bankperformance due to the sample selection problem.
Concluding Remarks 129
tween the efficiency versus market power hypotheses2 as motives for foreign expan-
sion (Lanine and Vander Vennet, 2007) and hypothesize that the relative strength
of these hypotheses may vary depending on the institutional environment in host
countries (EBRD, 2006; Lensink et al., 2008). Using a novel multilevel mixed-effect
logistic regression framework, we find support for the market power hypothesis in
relatively less advanced FSEs in terms of their economic development and institu-
tional background. This finding is in line with previous evidence of Lanine and
Vander Vennet (2007). However, we also show support for the efficiency hypothesis,
which holds for relatively more advanced FSEs. Our findings highlight the impor-
tance of macroeconomic heterogeneity in FSEs and its relevance for the decision of
foreign banks to go abroad.
The discussion of the implications of heterogeneous economic environments in
which banks operate for the assessment of their performance is continued in Chapter
4. We start our analysis by noticing that previous studies analyzing performance of
banks in FSEs based on the efficiency frontier framework impose a single technology
regime in banking. One of the consequences of this restrictive assumption is that
in the presence of multiple technology regimes, the obtained inefficiency estimates
will be biased (Orea and Kumbhakar, 2004). Moreover, the technology regimes in
transition banking are very likely to be affected by notable differences in macroeco-
nomic environments of these countries. Using a novel latent class stochastic frontier
methodology, we relax the single-frontier assumption of previous studies and al-
low for multiple technology regimes in transition banking. Our estimations suggest
that transition banking is characterized by three distinct technology regimes. These
technology regimes differ not only in terms of relative performance, technological
2 The efficiency hypothesis suggests that foreign banks enter host countries with the aim of ex-tracting revenues as a result of upgrading performance of target banks. In contrast, the marketpower hypothesis suggests that the main motivation for foreign entry is acquisition of large localbanks that would allow to exercise market power and extract monopolistic rents.
130 Chapter 7
progress, and returns to scale, but also in terms of the impact of foreign ownership
on bank efficiency. More specifically, we find that foreign entry improves efficiency of
banks located in FSEs characterized by better economic development prospects and
institutional background, while the impact of foreign ownership on the efficiency of
banks in less developed FSEs is ambiguous. This result confirms our previous finding
on the importance of accounting for the macroeconomic environment in evaluating
the impact of foreign bank participation.
Chapter 5 deals with another important aspect of opening the borders for for-
eign entry: its implications for the competitiveness in the domestic banking industry.
The novelty of our approach is that we take into account the impact of foreign en-
try on bank efficiency when assessing its implications for market competition. In
addition, we differentiate between two modes of foreign entry, foreign acquisitions
and greenfield establishments, when analyzing the impact of foreign entry on bank-
ing competition. This differentiation is important given different motives behind
these modes of entry: while greenfield investments are motivated by the follow the
client abroad considerations, cross-border acquisitions aim at establishing full scale
operations in FSEs. Our results suggest that foreign entry contributes to the com-
petitiveness in the banking industry only for the case of cross-border acquisitions,
while the impact of greenfield investments is insignificant. The latter finding can be
explained by the special relationships between foreign banks and their customers in
FSEs, which adds to the market power of greenfield foreign banks.
In Chapter 6 we investigate the impact of foreign bank participation on the
costs of financial intermediation in FSEs, proxied by net interest margins. The-
oretical studies on determinants of interest margins (the dealership model) do not
consider the role of bank ownership among the determinants (Ho and Saunders, 1981;
Maudos and Fernandez de Guevara, 2004), while other theoretical studies outline
Concluding Remarks 131
various direct and indirect channels through which foreign ownership may matter
(Claeys and Hainz, 2007; Dell’Ariccia and Marquez, 2004; Lehner and Schnitzer,
2008). Comparative analysis of both types of theoretical studies reveals that the
main channels through which foreign ownership may matter for the cost of financing
are taken into account by the dealership model. Our empirical analysis supports this
hypothesis and suggests that after taking into account the theoretically motivated
determinants of interest margins discussed in the dealership model, the own impact
of foreign ownership is insignificant. This finding is in contrast to previous studies,
which did not take into account all theoretically motivated determinants and found
significant impact of foreign ownership dummies, interpreting those as own effects
of foreign ownership on the cost of financing.
7.2 Policy Implications
The analysis conducted in this thesis confirms the general expectations of policymak-
ers that increased foreign bank participation will have a positive impact on FSEs,
but with some caveats. First of all, the analysis shows that the impact of foreign
bank entry is not uniform across FSEs. On average, more developed FSEs with a
better record for policy reforms seem to have gained more from foreign bank partici-
pation than the others. Related to this, the causal relationship between foreign bank
participation and performance may have gone also in the opposite direction, namely
improvement of overall economic performance and positive prospects of EU member-
ship have attracted foreign banks to the advanced FSEs. Next, the mode of foreign
entry needs to be taken into account by the policymakers when formulating policies
encouraging the foreign bank entry. Different motivations behind these modes re-
sult in different post-entry performance of foreign banks and should be weighed by
policymakers with care. Finally, further efforts need to be undertaken to improve
132 Chapter 7
the competitive stance of transition banking systems. Although foreign entry im-
proves competition on the margin, it should not be treated as panacea of solving all
problems in the domestic banking markets. A substantial degree of market power is
still present in most FSEs’ banking sectors.
Bibliography
Abarbanell, J. and J. Bonin: 1997, ‘Bank Privatization in Poland: The Case of BankSlaski’. Journal of Comparative Economics 25, 31–61.
Abarbanell, J. and A. Meyendorff: 1997, ‘Bank Privatization in Post-CommunistRussia: The Case of Zhilsotsbank’. Journal of Comparative Economics 25, 62–96.
Aigner, D., K. Lovell, and P. Schmidt: 1977, ‘Formulation and Estimation ofStochastic Frontier Production Function Models’. Journal of Econometrics 6,21–37.
Aliber, R.: 1984, ‘International Banking: A Survey’. Journal of Money, Credit, andBanking 16, 661–678.
Allen, L.: 1988, ‘The Determinants of Bank Interest Margins: A Note’. Journal ofFinancial and Quantitative Analysis 23(2), 231–235.
Altunbas, Y., L. Evans, and P. Molyneux: 2001, ‘Bank Ownership and Efficiency’.Journal of Money, Credit, and Banking 33(4), 926–954.
Andrews, D. and M. Buchinsky: 2000, ‘A Three-Step Method for Choosing theNumber of Bootstrap Repetitions’. Econometrica 68(1), 23–51.
Angbanzo, L.: 1997, ‘Commercial Bank Net Interest Margins, Default Risk, Interest-Rate Risk and Off-Balance Sheet Banking’. Journal of Banking and Finance 21,55–87.
Angelini, P. and N. Cetorelli: 2003, ‘The Effects of Regulatory Reform on Compe-tition in the Banking Industry’. Journal of Money, Credit and Banking 35(5),663–684.
Barajas, A., R. Steiner, and N. Salazar: 1999, ‘Interest Spreads in Banking in Colom-bia, 1974-96’. IMF Staff Papers 46, 196–224.
134 BIBLIOGRAPHY
Barajas, A., R. Steiner, and N. Salazar: 2000, ‘The Impact of Liberalization andForeign Investment in Colombia’s Financial Sector’. Journal of Development Eco-nomics 63, 157–196.
Barth, J., G. Caprio, and R. Levine: 2008, ‘Bank Regulations Are Changing: ForBetter or Worse?’. World Bank Policy Research Working Paper No. 4646.
Battese, G. and T. Coelli: 1992, ‘Frontier Production Function, Technical Efficiencyand Panel Data: With Application to Paddy Farmers in India’. Journal of Pro-ductivity Analysis 3, 153–169.
Battese, G. and T. Coelli: 1995, ‘A Model for Technical Inefficiency Effects in aStochastic Frontier Production Function for Panel Data’. Empirical Economics20, 325–332.
Berger, A.: 2007, ‘International Comparisons of Banking Efficiency’. Financial Mar-kets, Institutions and Instruments 16(3), 119–144.
Berger, A., R. Demsetz, and P. Strahan: 1999, ‘The Consolidation of the FinancialServices Industry: Cause, Consequences, and Implications for the Future’. Journalof Banking and Finance 23, 135–194.
Berger, A., R. DeYoung, and G. Udell: 2001, ‘Efficiency Barriers to the Consolida-tion of the European Financial Services Industry’. European Financial Manage-ment 7, 117–130.
Berger, A. and D. Humphrey: 1991, ‘The Dominance of Inefficiencies over Scaleand Product Mix Economies in Banking’. Journal of Monetary Economics 28,117–148.
Berger, A. and L. Mester: 1997, ‘Inside the Black Box: What Explains Differencesin the Efficiency of Financial Institutions?’. Journal of Banking and Finance 21,895–947.
Berglof, E. and P. Bolton: 2002, ‘The Great Divide and Beyond: Financial Archi-tecture in Transition’. Journal of Economic Perspectives 16(1), 77–100.
Bonin, J., I. Hasan, and P. Wachtel: 2005, ‘Bank Privatization and Performance:Evidence from Transition Countries’. Journal of Banking and Finance 29, 31–53.
Bonin, J., K. Miszei, I. Szekely, and P. Wachtel: 1998, Banking in TransitionEconomies: Developing Market Oriented Banking Sectors in Eastern Europe. Ed-ward Elgar Publishing: Cheltenham and Northampton.
BIBLIOGRAPHY 135
Bonin, J. and P. Wachtel: 2003, ‘Financial Sector Development in TransitionEconomies: Lessons from the First Decade’. Financial Markets, Institutions andInstruments 12, 1–63.
Bos, J. and H. Schmiedel: 2007, ‘Is There a Single Frontier in a Single EuropeanBanking Market?’. Journal of Banking and Finance 31, 2081–2102.
Bresnahan, T.: 1982, ‘The Oligopoly Solution is Identified’. Economics Letters 10(1-2), 87–92.
Buch, C.: 2000, ‘Why Do Banks Go Abroad? - Evidence from German Data’.Journal of Financial Markets 9, 33–67.
Buch, C. and G. DeLong: 2004, ‘Cross-Border Bank Mergers: What Lures the RareAnimal?’. Journal of Banking and Finance 28, 2077–2102.
Claessens, S., A. Demirguc-Kunt, and H. Huizinga: 2001, ‘How Does Foreign EntryAffect Domestic Banking Markets’. Journal of Banking and Finance 25, 891–911.
Claessens, S. and L. Laeven: 2004, ‘What Drives Bank Competition? Some Inter-national Evidence’. Journal of Money, Credit, and Banking 36(3), 563–583.
Claessens, S. and N. van Horen: 2008, ‘Location Decisions of Foreign Banks andInstitutional Competitive Advantage’. De Nederlandsche Bank Working PaperNo. 172/2008.
Claeys, S. and C. Hainz: 2007, ‘Acquisition versus Greenfield: The Impact of theMode of Foreign Bank Entry on Information and Bank Lending Rates’. ECBWorking Paper No. 653.
Claeys, S. and R. Vander Vennet: 2008, ‘Determinants of Bank Interest Margins inCentral and Eastern Europe: A Comparison With the West’. Economic Systems32, 197–216.
Coelli, T., P. Rao, C. O’Donnell, and G. Battese: 2005, An Introduction to Efficiencyand Productivity Analysis. Springer, 2 edition.
Cotarelli, C., G. Dell’Ariccia, and I. Vladkova-Hollar: 2005, ‘Early Birds, Late Ris-ers, and Sleeping Beauties: Bank Credit Growth to the Private Sector in Centraland Eastern Europe and in the Balkans’. Journal of Banking and Finance 29(1),83–104.
136 BIBLIOGRAPHY
Dabla-Norris, E. and H. Floerkmeier: 2007, ‘Bank Efficiency and Market Struc-ture: What Determines Banking Spreads in Armenia?’. IMF Working PaperWP/07/134.
De Haan, J. and I. Naaborg: 2004, ‘Financial Intermediation in Accession Coun-tries: The Role of Foreign Banks’. In: D. E. in Masciandaro (ed.): FinancialIntermediation in the New Europe: Banks, Markets and Regulation in EU Acces-sion Countries. Edward Elgar, pp. 181–207.
De Haas, R. and I. Van Lelyveld: 2006, ‘Foreign Banks and Credit Stability inCentral and Eastern Europe: A Panel Data Analysis’. Journal of Banking andFinance 30, 1927–1952.
Dell’Ariccia, G. and R. Marquez: 2004, ‘Information and Bank Credit Allocation’.Journal of Financial Economics 72, 185–214.
Demirguc-Kunt, A. and H. Huizinga: 2000, ‘Determinants of Commercial BankInterest Margins and Profitability: Some International Evidence’. World BankEconomic Review 13, 379–408.
Denizer, C.: 2000, ‘Foreign Entry in Turkey’s Banking Sector, 1980-97’. World BankPolicy Research Working Paper No. 2462.
Detragiache, E., T. Tressel, and P. Gupta: 2008, ‘Foreign Banks in Poor Countries:Theory and Evidence’. Journal of Finance 63(5), 2123–2160.
Djankov, S. and B. Hoekman: 2000, ‘Foreign Investment and Productivity Growthin Czech Enterprises’. World Bank Economic Review 14, 49–64.
Drakos, K.: 2003, ‘Assessing the Success of Reform in Transition Banking 10 YearsLater: An Interest Margin Analysis’. Journal of Policy Modeling 25, 309–317.
Drukker, D.: 2003, ‘Testing for Serial Correlation in Linear Panel-Data Models’.Stata Journal 3, 168–177.
EBRD: 2005, Transition Report. Oxford University Press: Oxford, U.K.
EBRD: 2006, Transition Report. Oxford University Press: Oxford, U.K.
EBRD: 2007, Transition Report. Oxford University Press: Oxford, U.K.
Focarelli, D., F. Panetta, and C. Saleo: 2002, ‘Why do banks merge?’. Journal ofMoney, Credit, and Banking 34(4), 1047–1066.
BIBLIOGRAPHY 137
Focarelli, D. and A. Pozzolo: 2001, ‘The Patterns of Cross-Border Bank Mergersand Shareholdings in OECD Countries’. Journal of Banking and Finance 25,2305–2337.
Focarelli, D. and A. Pozzolo: 2008, ‘Cross-Border M&A in the Financial Sector: IsBanking Different from Insurance?’. Journal of Banking and Finance 32, 15–29.
Freixas, X. and J.-C. Rochet: 2008, Microeconomics of Banking. The MIT Press,second edition.
Fries, S. and A. Taci: 2005, ‘Cost Efficiency of Banks in Transition: Evidence from289 Banks in 15 Post-Communist Countries’. Journal of Banking and Finance 29,55–81.
Galindo, A., A. Micco, and S. Cesar: 2003, ‘Better the Devil that You Know: Ev-idence on Foreign Costs Faced by Foreign Banks’. Inter-Americal DevelopmentBank Working Paper No. 477.
Giannetti, M. and S. Ongena: 2005, ‘Financial Integration and EnterpreneurialActivity: Evidence from Foreign Bank Entry in Emerging Markets’. CEPR Dis-cussion Paper No. 5151.
Giannetti, M. and S. Ongena: 2008, ‘"Lending by Example": Direct and IndirectEffects of Foreign Banks in Emerging Markets’. CEPR Discussion Paper No. 6958.
Green, C., V. Murinde, and I. Nikolov: 2007, ‘The Efficiency of Foreign and DomesticBanks in Central and Eastern Europe: Evidence on Economies of Scale and Scope’.Journal of Emerging Market Finance 3(2), 175–205.
Greene, W.: 2005, ‘Reconsidering Heterogeneity in Panel Data Estimators of theStochastic Frontier Model’. Journal of Econometrics 126, 269–303.
Grigorian, D. and V. Manole: 2006, ‘Determinants of Commercial Bank Performancein Transition: An Application of Data Envelopment Analysis’. Comparative Eco-nomic Studies 48, 497–522.
Hasan, I. and K. Marton: 2003, ‘Development and Efficiency of the Banking Sectorin a Transitional Economy’. Journal of Banking and Finance 27, 2249–2271.
Haselmann, R.: 2006, ‘Strategies of Foreign Banks in Transition Economies’. Emerg-ing Markets Review 7, 283–299.
138 BIBLIOGRAPHY
Havrylchyk, O. and E. Jurzyk: 2008, ‘Inherited or Earned? Performance of ForeignBanks in Central and Eastern Europe’. CEPII Working Paper No. 2008-16.
Heckman, J. J.: 1979, ‘Sample Selection Bias as a Specification Error’. Econometrica47, 153–161.
Heckman, J. J., S. Urzua, and E. Vytlacil: 2006, ‘Understanding Instrumental Vari-ables in Models with Essential Heterogeneity’. The Review of Economics andStatistics 88(3), 389–432.
Ho, T. and A. Saunders: 1981, ‘The Determinants of Bank Interest Margins: Theoryan Empirical Evidence’. Journal of Financial and Quantitative Analysis XVI(4),581–600.
Hughes, J. and L. Mester: 2008, ‘Efficiency in Banking: Theory, Practice, andEvidence’. Wharton School Working Paper No. 08-1.
IMF: 2000, ‘International Capital Markets: Developments, Prospects, and Key Pol-icy Issues’. Washington, DC.
Jemrić, I. and B. Vujčić: 2002, ‘Efficiency of Banks in Croatia: A DEA Approach’.Comparative Economic Studies 44, 169–193.
Kaufmann, D., A. Kraay, and M. Mastruzzi: 2007, ‘Governance Matters VI: Ag-gregate and Individual Governance Indicators 1996–2006’. World Bank PolicyResearch Working Paper No. 4280.
Koetter, M. and T. Poghosyan: 2009, ‘The Identification of Technology Regimes inBanking: Implications for the Market Power-Fragility Nexus’. Journal of Bankingand Finance 33, 1413–1422.
Kraft, E. and D. Tirtiroglu: 1998, ‘Bank Efficiency in Croatia: A Stochastic-FrontierAnalysis’. Journal of Comparative Economics 26, 282–300.
Kumbhakar, S. and K. Lovell: 2000, Stochastic Frontier Analysis. Cambridge Uni-versity Press.
Kumbhakar, S. and H.-J. Wang: 2005, ‘Estimation of Growth Convergence Using aStochastic Production Frontier Approach’. Economics Letters 88, 300–305.
Lanine, G. and R. Vander Vennet: 2007, ‘Microeconomic Determinants of Acqui-sitions of Eastern European Banks by Western European Banks’. Economics ofTransition 15, 285–308.
BIBLIOGRAPHY 139
Lehner, M. and M. Schnitzer: 2008, ‘Entry of Foreign Banks and Their Impact onHost Countries’. Journal of Comparative Economics 36(3), 430–452.
Lensink, R. and J. De Haan: 2002, ‘Do Reforms in Transition Economies AffectForeign Bank Entry?’. International Review of Finance 3, 213–232.
Lensink, R. and N. Hermes: 2004, ‘The Short-Term Effects of Foreign Bank Entryon Domestic Bank Behavior: Does Economic Development Matter?’. Journal ofBanking and Finance 28, 553–568.
Lensink, R., A. Meesters, and I. Naaborg: 2008, ‘Bank Efficiency and Foreign Own-ership: Do Good Institutions Matter?’. Journal of Banking and Finance 32,834–844.
Lepetit, L., E. Nys, P. Rous, and A. Tarazi: 2008, ‘Bank Income Structure and Risk:An Empirical Analysis of European Banks’. Journal of Banking and Finance 32,1452–1467.
Martinez Peria, M. and A. Mody: 2004, ‘How Foreign Participation and MarketConcentration Impact Bank Spreads: Evidence from Latin America’. Journal ofMoney, Credit, and Banking 36(3), 511–537.
Matoušek, R. and A. Taci: 2002, ‘Banking Efficiency in Transition Economies: Em-pirical Evidence from the Czech Republic’. Discussion Paper No: 02-3, Centre forInternational Capital Markets, London Metropolitan University.
Maudos, J. and J. Fernandez de Guevara: 2004, ‘Factors Explaining the InterestMargin in the Banking Sectors of the European Union’. Journal of Banking andFinance 28, 2259–2281.
Maudos, J. and J. Fernandez de Guevara: 2007, ‘The Cost of Market Power inBanking: Social Welfare vs. Cost Efficiency’. Journal of Banking and Finance 31,2103–2135.
Mester, L.: 1993, ‘Efficiency in the Savings and Loan Industry’. Journal of Bankingand Finance 17(2/3), 267–286.
Mester, L.: 1996, ‘A Study of Bank Efficiency Taking Into Account Risk-Preferences’.Journal of Banking and Finance 20, 1025–1045.
Mian, A.: 2006, ‘Distance Constraints: The Limits of Foreign Lending in PoorEconomies’. The Journal of Finance 61(3), 1465–1505.
140 BIBLIOGRAPHY
Navaretti, G. and A. Venables: 2004, Multinational Firms in the World Economy.Princeton University Press. with F. Barry, K. Ekholm, A. Falzoni, J. Haaland, K.Midelfart and A. Turrini.
Nikiel, E. and T. Opiela: 2002, ‘Customer Type and Bank Efficiency in Poland:Implications for Emerging Banking Market’. Contemporary Economic Policy 20,255–271.
Orea, L. and S. Kumbhakar: 2004, ‘Efficiency Measurement Using a Latent ClassStochastic Frontier Model’. Empirical Economics 29, 169–183.
Pagan, A.: 1984, ‘Econometric Issues in the Analysis of Regressions with GeneratedRegressors’. International Economic Review 25(1), 221–47.
Petersen, M. and R. Rajan: 1994, ‘The Benefits of Lending Relationships: Evidencefrom Small Business Data’. Journal of Finance 49, 3–37.
Poghosyan, T. and J. Borovicka: 2007, ‘Banking Efficiency in Emerging Economies:The Impact of Ownership Endogeneity and EU Accession’. CERGE-EI DiscussionPaper No. 189.
Poghosyan, T. and J. De Haan: 2008, ‘Determinants of Cross-Border Bank Acquisi-tions in Transition Economies: A Latent Class Analysis’. CESifo Working PaperNo. 2372.
Poghosyan, T. and A. Poghosyan: 2009, ‘Foreign Bank Entry, Bank Efficiency andMarket Power in Central and Eastern European Countries’. Economics of Tran-sition forthcoming.
Rabe-Hesketh, S. and A. Skrondal: 2005, Multilevel and Longitudinal Modeling usingStata. Stata Press.
Roll, R.: 1986, ‘The Hubris Hypothesis of Corporate Takeovers’. The Journal ofBusiness 59(2), 197–216.
Rossi, S., M. Schwaiger, and G. Winkler: 2004, ‘Banking Efficiency in Central andEastern Europe’. Financial Stability Report, OeNB 8, 77–91.
Sabi, M.: 1996, ‘Comparative Analysis of Foreign and Domestic Bank Operationsin Hungary’. Journal of Comparative Economics 22, 179–188.
BIBLIOGRAPHY 141
Schwaiger, M. and D. Liebeg: 2008, ‘Determinants of Bank Interest Margins inCentral and Eastern Europe’. OeNB Financial Stability Report, Austrian NationalBank.
Sealey, C. and J. Lindley: 1977, ‘Inputs, Outputs, and a Theory of Production andCost at Depository Financial Institutions’. Journal of Finance 32, 1251–1266.
Sengupta, R.: 2007, ‘Foreign Entry and Bank Competition’. Journal of FinancialEconomics 84, 502–528.
Shaffer, S.: 1989, ‘Competition in the U.S. Banking Industry’. Economics Letters29, 321–323.
Stock, J. and M. Yogo: 2002, ‘Testing for Weak Instruments in Linear IV Regression’.NBER Technical Working Paper 284.
Stock, J. H., J. H. Wright, and M. Yogo: 2002, ‘A Survey of Weak Instruments andWeak Identification in Generalized Method of Moments’. Journal of Business andEconomics Statistics 20(4), 518–529.
Valverde, S., D. Humphrey, and R. Lopez del Paso: 2007, ‘Opening the Black Box:Finding the Source of Cost Inefficiency’. Journal of Productivity Analysis 27,209–220.
Vera, L., L. Zambrano-Sequin, and A. Faust: 2007, ‘The Efficiency-Stability Trade-Off: The Case of High Interest Rate Spreads in Venezuela’. The DevelopingEconomies XLV(1), 1–26.
Williams, B.: 1997, ‘Positive Theories of Multinational Banking: Eclectic Theoryversus Internalisation Theory’. Journal of Economic Surveys 11(1), 71–100.
Williams, J. and A. Liao: 2008, ‘The Search for Value: Cross-Border Bank M&A inEmerging Markets’. Comparative Economic Studies 50, 274–296.
Yildirim, H. and G. Philippatos: 2007, ‘Efficiency of Banks: Recent Evidence fromthe Transition Economies of Europe, 1993-2000’. European Journal of Finance 31,123–143.
Samenvatting
Gedurende de laatste twee decennia is de financiële wereld drastisch veranderd alsgevolg van een golf van financiële liberalisaties en globalisatie in de banksector.Tegen deze achtergrond wordt in dit proefschrift de invloed van de toetreding vanbuitenlandse bank op het bancaire stelsel van gastlanden geanalyseerd. In dit proef-schrift worden in het bijzonder de volgende onderzoeksvragen behandeld:
• Wat brengt een bank ertoe om activiteiten in het buitenland op te zetten?
• Wat is de invloed van participatie van buitenlandse banken op de prestatiesvan en de concurrentie binnen het bancaire systeem van het gastland?
• Is de wijze van toetreding van invloed op de prestaties na toetreding?
• Hoe beïnvloedt toetreding van buitenlandse banken de kosten van financiëlebemiddeling?
Theoretische studies leveren tegenstrijdige voorspellingen over de invloed vantoetreding van buitenlandse banken op het bancaire systeem van de gastlanden.Empirisch onderzoek wordt bemoeilijkt door schaarsheid van data en verschillenin macro-economische en institutionele karakteristieken van de gastlanden waarmeerekening dient te worden gehouden. Bovendien kunnen selectie invloeden die gerela-teerd zijn aan de keuze van een bank om internationaal te gaan opereren de resultatenbeïnvloeden. In dit proefschrift worden deze problemen aangepakt door: (i) Datate gebruiken op bank niveau van banken uit voormalige socialistische economieën(former socialists economies FSEs). Er is voor FSEs gekozen omdat deze groepvan landen te maken heeft gehad met een grote toename van buitenlandse bankparticipatie gedurende de laatste twee decennia. (ii) Toepassing van innovatieve em-pirische methoden die kunnen omgaan met de moeilijkheden die het analyseren vanbuitenlandse bank participatie met zich mee brengt.
144
In Hoofdstuk 2 wordt onderzocht wat de invloed van toetreding van buitenlandsebanken is op de prestaties van banken. Dit hoofdstuk richt zich voornamelijk opde invloed van sampleselectie die ontstaat door de keuze van banken om naar hetbuitenland te gaan. Specifiek wordt gekeken of de positieve invloed van buitenlandseigendom op de efficiency, zoals beschreven in eerdere studies (Bonin et al., 2005,Fries and Taci, 2005, Yildirim and Philippatos, 2007), verklaard kan worden doorhet zogenoemde cream-skimming effect. Hiermee wordt bedoeld dat buitenlandsebanken alleen de best presterende binnenlandse banken overnemen.
In dit hoofdstuk maken we gebruik van een twee-staps procedure (Heckman,1979) waarmee we laten zien dat buitenlandse banken voornamelijk goed presterendebanken overnemen wanneer ze naar het buitenland gaan. Wanneer er gecontroleerdwordt voor deze sampleselectie, is niet langer sprake van een positieve invloed vanbuitenlandse banken op de prestaties van de bancaire sector zoals die in eerderestudies werd gerapporteerd. Bovendien wijzen de resultaten erop dat FSEs die meerbuitenlandse directe investeringen hebben aangetrokken een minder efficiënte banksector hebben.
In Hoofdstuk 3 wordt nieuw bewijs geleverd voor de motieven die een bank heeftom zijn activiteiten naar het buitenland uit te breiden. Wij bouwen op voorgaandeliteratuur die onderscheid maakt tussen de efficiency en de market power hypothesenals motieven om buitenlandse banken over te nemen (Lanine and Vander Vennet,2007) en veronderstellen dat de relatieve kracht van deze motieven afhankelijk kanzijn van de institutionele omgeving in het gastland (EBRD, 2006; Lensink et al.,2008). De efficiency hypothese veronderstelt dat buitenlandse banken die bankenkopen waarvan ze verwachten dat ze de efficiëntie kunnen verbeteren. De marketpower hypothese veronderstelt dat banken juist banken kopen met veel marktmacht.Met gebruikmaking van een recent ontwikkeld latente klasse logistische regressieraamwerk, laten we zien dat de market power hypothese opgaat voor FSEs die relatiefminder ontwikkeld zijn in termen van inkomen en kwaliteit van hun instituties. Voorde meer ontwikkelde FSEs vinden we echter bewijs voor de efficiency hypothese.Onze bevindingen benadrukken dat het belangrijk is om rekening te houden metheterogeniteit binnen FSEs bij het testen van de invloed toetreding van buitenlandsebanken.
De discussie over de invloed van heterogene economische omgevingen waarbinnenbanken opereren op het analyseren van hun prestaties, wordt voortgezet in Hoofdstuk4. Eerdere studies die bankprestaties meten met behulp van een efficient frontier
Samenvatting (Summary in Dutch) 145
raamwerk veronderstellen dat de te bestuderen landen beschikken over dezelfde tech-nologie. Een gevolg van deze nogal restrictieve veronderstelling is dat wanneer blijktdat er verschillende technologieën zijn, de verkregen efficiëntie scores gekleurd kun-nen zijn (Orea and Kumbhakar, 2004). Bovendien is het waarschijnlijk dat bank tech-nologieën worden beïnvloed door de verschillen in de macro economische omgevingvan FSEs. Met behulp van een recent ontwikkeld latente klasse stochastic frontiermethodologie kan de assumptie dat alle landen beschikken over dezelfde technologieworden versoepeld. Onze schattingen duiden erop dat bankieren in transitie lan-den wordt gekarakteriseerd door drie verschillende technologieën. De technologieënverschillen niet alleen in termen van relatieve prestaties, technologische vooruitgangen schaalvoordelen, maar ook met betrekking tot de invloed van buitenlands eigen-dom op de efficiëntie van een bank. Meer specifiek vinden we dat toetreding vanbuitenlandse banken de efficiëntie van banken verbetert in landen die sinds kortlid zijn van de EU. Deze landen hebben betere economische vooruitzichten en eensterkere institutionele achtergrond. De invloed van buitenlands eigendom op minderontwikkelde landen is ambigu. Deze resultaten onderbouwen de eerdere bevindingenvan het belang van de macro economische en institutionele omgeving als het gaatom het evalueren van buitenlandse bank participatie.
In Hoofdstuk 5 wordt gekeken wat de invloed van het openstellen van grenzenis op de concurrentie binnen het binnenlandse bancaire systeem. Het vernieuwendevan onze aanpak is dat we rekening houden met de invloed van het toetreden vanbuitenlandse banken bij het analyseren van de concurrentie binnen het bancaire sys-teem. Bovendien maken we onderscheid tussen overnames en greenfield investments.Dit onderscheid is belangrijk omdat er voor de verschillende manieren van toetredingmogelijk verschillende motieven zijn. Bij een greenfield is het waarschijnlijk dat debank zijn klanten achterna gaat en slechts beperkte diensten aanbiedt, terwijl bij eenovername het waarschijnlijk is dat de bank een breeds scala van diensten wil gaanaanbieden. Onze resultaten duiden erop dat toetreding van buitenlandse bankenalleen bijdraagt aan meer concurrentie in de bank sector wanneer er sprake is vaneen overname. De invloed van greenfields op de concurrentie is niet significant.
In Hoofdstuk 6 onderzoeken we de invloed van buitenlandse banken op de kostenvoor financiële bemiddeling in FSEs, door te kijken naar netto interest marges. The-oretische studies over de determinanten van interest marges (’het dealership model’)gaan ervan uit dat karakteristieken van de eigenaar van een bank hierin geen rol spe-len (Ho and Saunders, 1981; Maudos and Fernandez de Guevara, 2004). Andere stud-
146
ies geven echter aan dat buitenlands eigendom via verschillende directe en indirectekanalen wel degelijk van invloed kan zijn op interest marges (Claeys and Hainz, 2007;Dell’Ariccia and Marquez, 2004; Lehner and Schnitzer, 2008). Vergelijkende analy-ses van beide typen theoretische studies laten zien dat de hoofd kanalen waarmeebuitenlands eigendom van invloed is op de kosten van financiering meegenomen wor-den in het dealership model. Onze empirische analyse ondersteunt deze hypotheseen laat zien dat wanneer rekening wordt gehouden met theoretisch gemotiveerdedeterminanten, eigendom niet van invloed is op de interest marge. Deze bevindingwijkt af van eerdere studies die geen rekening houden met theoretisch gefundeerdedeterminanten en alleen eigendom opnemen als determinant van interest marges. Indeze studies komt naar voren dat eigendom significant is.
Onze analyses bevestigen de verwachting van beleidsmakers dat een toenamevan buitenlandse banken een positieve invloed heeft op FSEs, maar leiden ook totenige nuanceringen. Allereerst is de invloed van buitenlandse banken niet overalhetzelfde. Gemiddeld genomen profiteren FSEs waar hervormingen zijn doorgevoerdmeer van toetreding van buitenlandse banken dan FSEs waar deze hervormingennog onvoldoende zijn doorgevoerd. Het is echter ook mogelijk dat de causaliteitomgekeerd is en dat banken voornamelijk naar die landen zijn gegaan die al meereconomisch ontwikkeld waren en zich richtten op toetreding tot de EU.
Vervolgens blijkt dat beleidsmakers rekening moeten houden met de wijze waaropbuitenlandse banken toetreden. De motieven die schuilgaan achter de manier vantoetreding resulteren in verschillende in prestaties na toetreding. Ten slotte moet erextra inspanning geleverd worden om de concurrentie binnen het bancaire systeemin de transitie landen te verbeteren. Hoewel toetreding van buitenlandse bankenleidt tot meer concurrentie, blijkt in de meeste FSEs sprake te zijn van marktmacht.