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.