BANK CONSOLIDATION AND BANK EFFICIENCY IN EUROPE
Marko Košak
University of Ljubljana, Faculty of Economics, Ljubljana, Slovenia
marko.kosak@ef.uni-lj.si
Peter Zajc
European Investment Bank, Luxembourg
p.zajc@eib.org
Abstract:
Changes in the banking environment, particularly the increasing integration of the EU
financial markets, have accelerated bank consolidation in the EU over the past decade. The
focus of the consolidation process has gradually shifted from domestic to cross-border M&A
activities. Bank consolidation has decreased the number of banks, which has had two key
implications for the banking sectors: a change in bank concentration and thus market power,
and an impact on bank efficiency. In our paper we focus on bank efficiency and particularly
on the efficiency difference between the new and the old EU member countries (East-West
efficiency gap). At the beginning of the 1990s, banking sectors in transition countries of
Central and Eastern Europe had been in a relatively poor shape and significantly less efficient
than their western counterparts. However, over the past decade, banks in Central and Eastern
Europe have significantly improved their efficiency. We apply the stochastic frontier
approach to estimate the East-West banking efficiency gap and its dynamics in the 1996-2003
period. The analysis was performed on a sample of commercial banks from fifteen EU
countries. We find a narrowing efficiency gap, which might indicate that bank consolidation
in Central and Eastern Europe has contributed to increasing bank efficiency in this region.
Key words: bank consolidation, banking efficiency, stochastic frontier, European Union
JEL classification: G2; G21; P30; P34; P52

Corresponding author.
The views expressed are those of the authors and do not reflect those of the institutions with which the authors
are affiliated.

1
Povzetek:
Spremembe v bančnem okolju, še posebej pa vse večja integracija finančnega prostora EU, so
spodbudile bančno konsolidacijo v državah EU v zadnjem desetletju. Proces konsolidacje, ki
poteka predvsem preko združitev in pripojitev, se je postopoma spreminjal in je v zadnjih
letih dobil povsem mednarodni značaj, kar pomeni, da združitve in pripojitve vključujejo
bančna podjetja iz različnih držav. Posledica konsolidacije je bilo znižanje števila bank, kar je
vplivalo na tržno koncentracijo, tržno moč in učinkovitost v bančnih sektorjih. V prispevku se
osredotočava na bančno učinkovitost, s poudarkom na razliki v učinkovitosti med bankami v
starih in novih članicah EU. Na začetku 90. let so bile namreč banke v tranzicijskih državah
srednje in vzhodne Evrope občutno slabše razvite in manj učinkovite kot zahodnoevropske
banke, vendar so uspele v obdobju tranzicije v veliki meri izboljšati svojo učinkovitost. Za
ovrednotenje razlik v učinkovitosti in za merjenje časovne dinamike učinkovitosti v obdobju
1996-2003 je uporabljena metoda stohastične mejne funkcije učinkovitosti. Analiza je bila
opravljena na vzorcu poslovnih bank iz 15 držav EU, rezultati pa so pokazali, da so se razlike
v učinkovitosti bank iz starih in novih držav članic zmanjšale, iz česar bi lahko sklepali, da so
se učinki konsolidacije v državah srednje in vzhodne Evrope odrazili tudi večji učinkovitosti
bank iz teh držav.
Ključne besede: konsolidacija bank, učinkovitost bank, stohastična mejna funkcija,
Evropska unija
1. Introduction
In the recent past, the international banking industry has been undergoing far-reaching
structural changes. The processes of liberalisation, globalisation and integration have
dramatically changed the banking landscape around the world. In the European Union, the
Second Banking Directive of 1989 opened up the banking sectors of all EU member
countries to other EU banks. The introduction of the Economic and Monetary Union in 1999
contributed to a large and transparent common banking sector. This has spurred the largest
wave of mergers and acquisitions (M&As) so far. The banking sector consolidation
accelerated in the early 1990s and peaked in 2000. It resulted in a reduction of the number of
banks, and an increase in the average bank size and the concentration of the banking sector.
The main factor driving the consolidation process has been value maximisation, i.e. the
expectation that the value of the new bank will exceed the sum of the individual values of
banks to be merged. Value maximisation is generally achieved through gains in market
power and efficiency (Berger et al., 2000). Empirical evidence indicated that efficiency gains
2
are realised through economies of scale, economies of scope and improvements in
managerial efficiency, also known as X-efficiency. The literature on bank efficiency, e.g.
Berger, Hunter and Timme (1993), shows that the former two sources of efficiency gains are
relatively small compared to potential gains in managerial efficiency.
At the beginning of transition, banking sectors in Central and Eastern European countries
were in a gloomy condition with significantly lower efficiency compared to that in the EU.
However, these countries embarked on the process of radical and profound restructuring of
their banking sectors, in which bank consolidation played a key role. This resulted in
significant improvements in the banking industry in Central and Eastern Europe, and in
increased bank efficiency. In this respect, we address the question of the efficiency
discrepancy between banks in the EU. We estimate cost efficiency of banks in ten new EU
Member States (the Czech Republic, Cyprus, Estonia, Hungary, Lithuania, Latvia, Malta,
Poland, Slovakia and Slovenia) and five old EU Member States (Austria, Belgium, Germany,
Italy, Netherlands) during the 1996-2003 period1. Cost efficiency provides information on
how close (or far) a bank’s costs are from a best-practice bank’s costs. It is estimated using a
common frontier, thus allowing us to compare efficiency estimates across countries and to
examine the development of the East-West efficiency gap over the sample period.
The rest of this paper is structured as follows. First, section two gives an overview of the bank
consolidation process with a particular focus on European banking. The next section connects
areas of bank consolidation and efficiency analysis. Section four sketches out the theory of
efficiency measurement. This is followed by a discussion on the theoretical foundations and
design of the empirical estimation used in our analysis. Finally, our estimation results for cost
efficiency are presented. The chapter concludes with a comment on the results.
2. An overview of bank consolidation
The key characteristic of consolidation is change of control, which takes place through a
transfer of ownership (Ajadi and Pujals, 2004). This generally takes place through a merger or
an acquisition, whereby an M&A is defined by transfer of ownership in which one company
increases its stake in the other company’s equity to above 50%. Looking at the history of
M&A activity since the beginning of industrialisation, we observe four consolidation waves:
1987-1904, 1916-29, 1965-69 and 1984-89 (Ajadi and Pujals, 2004). The current
consolidation wave, i.e. the fifth one, has been by far the most intense in terms of the number
of transactions and their total value. Another key feature of the recent M&A wave has been its
cross-border dimension. In an increasingly global in connected world economy, cross-border
M&As started to increase in the early 1990s and peaked in 2000 with a total value of almost
USD 1.2 trillion (Chart 1). This was followed by a decline in the following two years. Broken
The expression ‘New EU member-countries’ refers to those countries which joined the EU on 1 May 2004,
while the expression ‘Old EU member-countries’ refers to all the other EU countries.
1
3
down by different categories, we observe a similar pattern for manufacturing, tertiary sectors
(e.g. energy, construction, tourism etc.) and the financial industry. As to the latter, in 2000 the
total volume of cross-border M&As in the financial industry was around USD 190 billion.
Chart 1: Cross-border M&As 1987-2002 (in USD billion)
1,200
1,000
800
600
400
200
0
1987
1989
1991
Total
1993
1995
Manufacturing
1997
Tertiary
1999
2001
Finance
Source: UNCTAD 2004
Similar to the worldwide M&A trend, bank consolidation in Europe has accelerated in the last
decade. It has been driven by the technological progress, which resulted in high fixed costs
and demanded large volume business to cover the initial outlay, a reduction of overcapacity,
and particularly by changes in the banking environment. 1990s brought a liberalisation of the
EU capital market, the banking directives, the Financial Services Action Plan (FSAP), and the
euro as a single currency, all contributing to an increasingly integrated EU financial market.
This has most notably increased competition, spurred deregulation, and reduced bank
margins. One can also see these developments as a wake-up call for the European banks to
redesign their business strategies and focus on the growth of business volumes to remain
competitive and to restore revenues. With growth opportunities becoming scarce in the
domestic markets, European banks have been increasingly looking across national borders.
It is interesting to observe the nature of cross-border M&As in the EU in terms of the origin
of the acquirers. The data show that the acquirers in the EU banking sector were throughout
the period predominantly institutions based in the countries of Euroland. Their share in the
total number of M&As increased from 53% in the 1990-1997 period to 59% in the 1998-2001
period (Table 1). The share of other EU member countries has also increased significantly –
to 38% in the 1998-2001 period. On the other hand, the share of acquirers from the USA and
4
the “rest of the world”2 has decreased from around 20% in the 1990-1997 period to 2% for the
USA and 1% for the rest of the world in the 1998-2001 period. All these figures clearly
indicate a strong “European nature” of the M&As in the most recent period, which could be
also explained by a greater degree of banking market integration within the EU. Baele et al.
(2004) suggest that the integration has been significantly stimulated by the introduction of the
single currency in 1999.
Table 1: Origin of acquirers in cross-border bank mergers in the euro area
(% of total)
Euro area
Other EU countries
USA
Rest of the world
Within-industry
Cross-industry
1990-97
1998-01
1990-97
1998-01
53
9
21
17
59
38
2
1
50
1
2
47
47
0
13
40
Source: Cabral et al., 2002
In contrast to the within-industry M&A figures, the data for cross-industry M&As unveil
more international nature of M&As in the EU countries, since the share of USA based
acquiring institutions was 13% in the 1998-2001 period. The share of institutions from “rest
of the world” 40% in 1998-2001 period.
These developments may potentially indicate the establishment of a relatively closed EU
banking market, which would consist of EU member countries’ regional banking markets and
would be relatively closed for entrant banks from non-EU countries. In addition, the
implementation of the EU Takeover Directive and the on-going withdrawal of the State from
banking industry may further intensify the (EU) cross-border merger activity and increase
concentration (ECB, 2003)
The currently achieved degree of integration in EU banking differs across banking business
lines. Therefore, in order to be able to observe the integration effects in the EU banking
markets we need to take into account distinctive characteristics of banking operations.
Typically, studies on banking integration distinguish wholesale banking activities from capital
market-related banking activities and retail banking services. According to recent studies
(Cabral et al., 2002; Baele et al., 2004), wholesale operations in euro-area banking appear to
be the most integrated, whereas in the capital market-related operations some severe obstacles
to integrations still exists (e.g. fragmented infrastructure for cross-border clearing and
settlement of securities transactions). Similarly, the markets for retail services also lack
integration, which is due to completely different nature of these markets, where the proximity
of banks to local customers and retail services diversification are very important. Thus, even
2
Switzerland is the major acquirer in the “rest of the world” category.
5
in the more distant future it is not likely that this market segment will achieve the level of
integration similar to the one of the wholesale banking.
Similar to differences among various lines of bank business, there are, despite the advanced
EU financial and banking market integration, also striking differences in the approach to bank
consolidation. Without going into too much detail and by looking merely at large EU member
countries, we can identify somewhat different developments. The number of banks remains
the highest in Germany and France (Chart 2). In Germany, this is a reflection of the historical
division of the banking sector into three pillars. The smallest share of the market is in the
hands of private banks. The second pillar consists of savings banks (Sparkassen) and
wholesale banks (Landesbanken), together accounting for more that a half of the German
banking sector. The third pillar includes the co-operative banks (Genossenschaften). When
state guarantees are discontinued in July 2005, and with a possible change in legislation,
which currently prevent private investors from buying Sparkassen, consolidation in the
German banking sector in likely to accelerate.
Chart 2: Number of banks in 2003
2500
2000
1500
1000
500
0
DE
FR
IT
UK
ES
Source: Deutsche Bank Research 2004
Differences in the number of banks are also reflected in the bank density, defined as the
number of banks per 100,000 inhabitants. Compared to Spain and the UK, Germany and
France seem to be overbanked, i.e. it seems that there are too many banks chasing too few
customers. This can be an important impediment to bank efficiency and profitability, and
could as such pose a major obstacle in the competing with other EU banks, particularly those
from the UK.
6
Chart 3: Bank density in 2002 (Banks per 100,000 inhabitants)
3.5
3.1
3.0
2.5
2.0
1.7
1.6
1.4
1.5
1.0
0.7
0.6
0.5
0.0
DE
FR
EU (ex DE)
IT
ES
UK
Source: Deutsche Bank Research 2004
Looking from the perspective of the consolidation process, one can again observe significant
differences. The UK, with the lowest number of banks, has undergone the most pronounced
and fastest consolidation over the past ten years. The number of banks has dropped by more
than a half compared to 1992 (Chart 4). The pace of bank consolidation in France and
Germany has been significantly and the number of banks declined by 43% and 40%,
respectively, over the 1992-2003 period.
Chart 4: Change in number of banks 1992-2003
%
0
-10
-16
-20
-30
-28
-40
-43
-50
-40
-56
-60
UK
FR
DE
IT
ES
Source: Deutsche Bank Research 2004
Germany and France, two countries with a relatively slow progress of bank consolidation,
demonstrate the highest cost-to-income ratios, i.e. their banks operate on average with
7
relatively higher costs than their counterparts in other EU countries. Cost-to-income ratios
can be interpredted as an indicator of bank efficiency. Hence, this empirical evidence seems
to indicate that bank consolidation has an implication for the efficiency of the banking sector.
It suggests that a higher level of bank consolidation is correlated with higher bank efficiency.
Chart 5: Cost-to-income ratio (1992-2001 average)
%
75
70
67.1
64.3
65
63.1
60.7
60.4
ES
UK
60
55
50
FR
DE
IT
Source: Deutsche Bank Research 2004
After a brief empirical overview, we turn to the several factors driving the consolidation
process3. The key motivation is value maximisation, which is based on the assumption that
the value of the new bank will exceed the sum of the individual values of banks to be merged.
M&As in the banking sector have the potential to create value through gains in market power
and/or efficiency gains (Berger et al., 2000).
Market power is a reflection of a dominant position in a banking sector and is generally
achieved by merging two competitors in the same market. A bank with significant market
power can use its market position in two ways. First, it can manipulate prices on the assets
side of its operation. By lowering prices of its products such as loans it attempts to squeeze
out of market other less competitive and less strong rivals and discourage new entry.
Alternatively, when market conditions allow it, a bank can increase prices and enhance its
revenues. Secondly, on the liabilities side of its business, a new bigger bank is likely to secure
more favourable funding conditions that the individual merged entities because of its larges
size (a larger bank is less vulnerable to economic shocks), enhanced reputation and the
diversification effect.
3
This section draws on Ajadi and Pujals (2004).
8
Efficiency gains from a merger or an acquisition materialise through economies of scale,
economies of scope, and through a transfer of knowledge and managerial skill, which are
aimed at enhancing managerial efficiency of the merged entity. Economies of scale result
from the relationship between the average production cost per unit of output and the
production volume (Kwan, 2004). Merging two banks with a similar line of activity leads to
economies of scale due to an increase in business volume and a simultaneous reduction of
fixed costs. The latter is achieved by merging different support function such as the
information system, marketing, back office, personnel management, and by streamlining the
branch network. Economies of scope result from a situation where the joint costs of producing
two complementary outputs are less the combined costs of producing the two outputs
separately (Kwan, 2004). This arises because the production processes of both outputs share
some common inputs (for instance labour and branch network). Economies of scope are
generally achieved in mergers of two entities with not completely overlapping activities, i.e. a
bank and an investment bank or a bank and an insurance company (e.g. Deutsche Bank and
Banker’s Trust; Dresdner Bank and Allianz).
Beyond this aspect, i.e. efficiency gains realised through economies of scale and scope,
M&As generally bring about at least some organisational and restructuring efforts in the new
entity. For instance, adjustments to bank management system and adoption of better business
practices contribute to improved managerial efficiency of the merged bank. Leibenstein
(1966) was the first to introduce the concept of managerial efficiency or X-efficiency, which
reflects the differences in managerial ability to control costs and/or maximise profits. The
general idea is that one of the parties in an M&A, generally the bidding bank, possesses
superior management skills and practices, which are then applied in the newly merged bank
and hence improve the managerial efficiency of this new bank. Berger, Hunter and Timme
(1993) observe that scale and scope economies have been extensively studied in the past, but
another source of differences in efficiency, X-inefficiency, has been neglected although it
accounts for 20% or more of the costs in banking, while scale and scope inefficiencies
account for less than 5%. The way banks are run is more important than their size and/or the
selection of banking products they offer (Walter, 1999). The second part of our paper is
dedicated to an X-efficiency study of banks in Europe, focusing of the efficiency wedge
between new and old EU member countries.
There are some other factors that seem to play a role in the M&A process. One such factor is
management’s desire for prestige and power. In absence of effective management control,
managers can pursue goals in their own interest rather than maximising shareholder value. For
example, a merger can give managers more power in the banking industry, which is closely
connected to higher remuneration and prestige. Another factor discussed in the literature is the
so-called mimicry effect. Managers might follow actions of other major banks, leading to
identical or very similar business strategies. In the banking industry, the common strategies in
the 1980s was to increase the size of the bank, in the 1990s to enhance ROE, and at present
bank strategy is geared toward value creation. Finally, a factor driving bank consolidation is
9
also the defensive reaction: in a dynamic M&A market, bank managers would like to defend
their banks from a hostile takeover but realise that they are too small. A friendly merger can
generate a significantly bigger entity and reduce the threat of a hostile take over. In Germany,
where domestic banks fell far behind the leading global players with respect to size, top
politicians in Germany called for domestic mergers of the largest banks to make them less
vulnerable and increase the chance of remaining independent.
3. Bank consolidation and efficiency
Bank consolidation can have positive but also negative effects for the merged bank and for the
banking sector as a whole (e.g. adverse price changes for bank customers, diversion of banks
away from locally oriented services, diseconomies of scale, increased risk due to
organisational complexity). Not only the banks involved but also the supervision and
competition authorities in respective banking markets evaluate net benefits of M&A activities
and the consolidation effects. A key aspect of this evaluation is bank efficiency, one of the
most important driving factors of bank consolidation. It has attracted bank practitioners,
supervision authorities and academics to study, analyse and developed policy
recommendations to enhance efficiency of banking.
Although the body of literature on bank efficiency is substantial, it is heavily geared towards
studies of US banks, followed by European banks in a distant second place. There are only
few studies on bank efficiency in less developed countries. Central and Eastern European
countries have not received much attention so far, even though the intensity of restructuring
and consolidation processes (including numerous M&As) has been extremely pronounced in
this region over the last 10 to 15 years. Most of the efficiency studies involving CEE banking
sectors have been conducted for specific national banking markets only, whereas the crosscountry efficiency studies which provide comparable efficiency results are still extremely
scarce. At least three such studies need to be mentioned here.
The first is the 2002 working paper Determinants of Commercial Bank Performance in
Transition: An Application of Data Envelopment Analysis by Grigorian and Manole (2002).
They estimate bank efficiency using the DEA technique. Their sample is quite heterogeneous,
i.e. it includes countries at substantially different development levels. They divide the
countries included in the study into three groups: Central Europe, Southeast Europe and the
Commonwealth of Independent States. Overall, banks from Central Europe are found to be
the most efficient.
The second study is the recent working paper Efficiency of banks: Recent evidence from the
transition economies of Europe 1993-2000 by Yildirim and Philippatos (2002). They use both
the SFA and the DFA approach to estimate bank efficiency for 12 Central and Eastern
European countries, as well as efficiency correlates for the banking sectors in this group of
10
countries. The estimated inefficiencies were found to be significant, with respective average
cost efficiency levels for 12 countries of 72 and 76 percent according to the DFA and SFA.
However, the relatively large number of countries in the sample suggests the sample is quite
heterogeneous, meaning it encompasses more advanced economies which recently obtained
EU membership as well as economies that still need to make significant progress in order to
draw near development levels common in the EU.
The third and most recent paper by Bonin et al. (2004) focuses on evaluating bank efficiency
and identifying relevant efficiency correlates in transition countries, with special attention
being paid to the efficiency-ownership relationship. The authors applied stochastic frontier
estimation procedures to banks in eleven transition countries. The results provided by Bonin
et al. (2004) indicate that private ownership is, by itself, insufficient to ensure bank efficiency
in transition countries because no statistically significant evidence of an adverse effect of
government ownership relative to private domestic ownership was found. Foreign-owned
banks turn out to be more cost efficient than other banks and they also provide better services,
particularly if they have a strategic foreign owner.
In the rest of the paper we present our bank efficiency study with which we contribute to the
field of cross-country efficiency studies for Central and Eastern European banking markets by
introducing a mixed sample of countries. The sample includes, on one hand, banks from ten
new EU member countries (the Czech Republic, Cyprus, Estonia, Hungary, Lithuania, Latvia,
Malta, Poland, Slovakia and Slovenia), and, on the other hand, banks from five old EU
member countries (Austria, Belgium, Germany, Italy, Netherlands). It covers the period 19962003. Since the banks in the new EU member countries had to adopt the legislative and
regulatory framework common in the EU well in advance their EU accession, our efficiency
analysis is based on a reasonable assumption that all banks in our sample are operating in a
very similar regulatory and legislative environment. Consequently, any efficiency differences
among the banks from different countries can be largely attributed to a greater managerial
efficiency. We suggest that a substantial part of the managerial efficiency improvements
might have resulted from the well-advanced consolidation process in these countries. In the
next section, we elaborate on some general issues related to the selection of efficiency
measurement techniques, and present the efficiency estimation model and data. Results follow
in section seven.
4. Efficiency measurement techniques in banking
The concept of efficiency measurement assumes that the production function of the fully
efficient firm or firms is known. Since this is not the case in practice, one has to estimate the
production function. A number of different techniques are used to estimate efficiency. Farrell
(1957) proposed that the production function could be estimated from sample data applying
11
either a non-parametric (mathematical programming) or a parametric (econometric)
approach.4
The three main parametric techniques are the stochastic frontier approach (SFA), the
distribution-free approach (DFA) and the thick frontier approach (TFA). These methods focus
on the difference or distance from the best-practice bank (efficient frontier), i.e. this distance
reflects the inefficiency effect ui. For example, if costs are higher than those of the bestpractice bank, then the bank is cost inefficient. The key characteristic of parametric
techniques is that they a priori impose a rule (assumption) for how random errors can be
separated from inefficiency. Thus, they make an arbitrary distinction between randomness and
inefficiency, which is the main drawback and criticism of parametric techniques (Schure and
Wagenvoort, 1999). Estimation techniques differ in the way they handle the composite error
term vi + ui, i.e. in the way they disentangle the inefficiency term ui from the random error
term vi. In the empirical part below we apply the SFA technique, which is based on the
assumption that the random error vi is symmetrically distributed (normal distribution) and that
the inefficiency term ui follows an asymmetric (one-sided) distribution (truncated normal
distribution).
There is a general distinction between deterministic and stochastic frontier production
functions (Kaparakis, Miller and Noulas, 1994). The main drawback of the deterministic
frontier is that it does not account for measurement errors and statistical noise problems, thus
all deviations from the frontier are assumed to reflect inefficiency (Coelli, Rao and Battese,
1998). This can seriously distort the measurement of efficiency. The stochastic frontier
production function avoids some of the problems associated with the deterministic frontier.
Aigner, Lovell and Schmidt (1977), and Meeusen and van den Broeck (1977) independently
proposed a stochastic frontier function with a composite error term, which allows the
production function to vary stochastically:
yi = xi β + ei
where yi
xi
β
ei
i = 1..N
(1)
is the logarithm of the maximum output obtainable from xi
is a vector of logarithms of inputs used by the i-th firm
is the unknown parameter vector to be estimated
is the error term.
The error term ei is composed of two parts:
ei = vi - ui
i = 1..N
(2)
where vi is the measurement error and other random factors
4
See Bauer et al. (1998) for a discussion of parametric and non-parametric estimation techniques.
12
ui is the inefficiency component.
The vi component captures the statistical noise, i.e. measurement error and other random or
uncontrollable factors. Aigner, Lovell and Schmidt (1977) assumed that vis are independently
and identically distributed normal random variables with mean zero and a constant variance,
i.e. vi ~ iid N(0,  v2 ). The ui component is a non-negative random variable accounting for
technical inefficiency in the production of firms. It measures technical inefficiency in the
sense that it measures the shortfall of output from its maximal possible value given by the
stochastic frontier xiβ + vi (Jondrow et al., 1982). This shortfall or, more generally, deviations
from the frontier are due to factors that are under the control of management, as opposed to
vis, which are not under management control (Chang, Hasan and Hunter, 1998). uis are
distributed either iid exponential or half-normal.
The main shortcoming of the SFA is the a priori distributional assumption of uis. This
assumption is necessary in order to use the maximum likelihood method to solve for the
parameters. In general, the stochastic frontier model can be estimated by using corrected
ordinary least squares (OLS), but maximum likelihood is asymptotically more efficient. In our
estimation, we apply the maximum likelihood method.
The mean of the distribution of the ui (the mean technical inefficiency) is easy to compute.
One simply calculates the average of ei estimates, and the statistical noise component vi
averages out. Computing technical inefficiency for individual firms is more demanding. The
decomposition of the error term into its two components, vi and ui, remained unresolved until
Jondrow et al. (1982) proposed how to calculate the observation (bank) specific estimates of
inefficiency conditional on the estimate of the error term ei.. The best predictor for ui is the
conditional expectation of ui given the value of ei = vi - ui. The predictor for efficiency is
obtained by subtracting the inefficiency from one.
Battese and Coelli (1988) showed that the best predictor of technical efficiency, exp(-ui), is
obtained by using
 exp  ui  |  i  
1  ( A 
 i
)
A

1  ( i )
A
exp( i 
where
13
 A2
2
)
(3)
 (.) is the cummulative density function of a standard normal random variable
 A   1     S2
2
  2 ,  S2   2   V2
S
 2 is the variance of ui s
 V2 is the variance of vi s.
5. Cost efficiency model
The technical efficiency concept based on a production function is easily modified and
extended to measure bank cost efficiency. Cost efficiency is derived from the cost function. It
provides information on how close (or far) a bank’s costs are from a best-practice bank’s
costs, producing the same output in the same conditions. In other words, cost efficiency
reflects the position of a particular bank relative to the cost frontier. A stochastic cost frontier
is presented below, where C(.) is a suitable functional form.
ln ci  C ( yi , wi ;  )  vi  ui
i=1,2,...,N
where ci is the observed cost of production for the i-th firm
yi is the logarithm of the output quantity
wi is a vector of logarithms of input prices

is a vector of unknown parameters to be estimated
vi is the random error
ui is the non-negative cost inefficiency effect.
Note that the inefficiency factor ui is added because the cost frontier represents minimum
costs (Coelli, Rao and Battese, 1998).5 The random error vi accounts for measurement errors
and other random factors. The inefficiency factor incorporates both technical inefficiency (i.e.
employing too many of the inputs) and allocative inefficiency (i.e. failures to react optimally
to changes in relative prices of inputs) (Berger and Mester, 1997). The random error and the
inefficiency term are assumed to be multiplicatively separable from the cost frontier.
Efficiency measurement techniques differ in how they separate the composite error term vi +
ui , i.e. how they distinguish the inefficiency term from the random error.
5
The production frontier represents maximum output and ui is subtracted from it.
14
We use panel data on banks from Central and Eastern Europe and the model by Battese and
Coelli (1992) to estimate cost efficiency. They proposed a stochastic frontier model with
time-varying inefficiency effects. The model can be written as
ln( yit )  xit   vit  uit
i  1,2,...N ;
t  1,2,..., T
(4)
where yit is the output of i-th firm in the t-th time period
x it is a K-vector of values of logarithms of inputs and other
appropriate variables associated with the suitable functional form

is a K-vector of unknown parameters to be estimated
vit are random errors assumed to be iid N(0,  v2 ) independent of uit s
u it are technical inefficiency effects.
Different distributions of uits have been assumed for this panel data model (see Coelli, Rao
and Battese, 1998, for a short overview of the evolution of this model). The model permits
unbalanced panel data and uits are assumed to be an exponential function of time, involving
only one unknown parameter,
u it  exp   (t  T )  ui
i=1,2,...,N;
t=1,2,...,T
(5)
where ui are assumed to be iid generalised truncated normal random variables
 eta is an unknown scalar parameter to be estimated.
In period T (i.e. t=T), the exponential function exp  (t  T ) has a value of one and thus the
ui is the technical inefficiency for the i-th firm in the last period of the panel. Inefficiency
effects in all previous periods of the panel are the product of the technical inefficiency for the
i-th firm in the last period of the panel and the value of the exponential function
exp  (t  T ) . The value of the exponential function is determined by the parameter eta (η)
and the number of periods in the panel. Inefficiency effects can decrease, remain constant or
increase as time increases, i.e. η > 0, η = 0 and η < 0, respectively. This specification of
inefficiency effects implies that the ranking of firms according to the magnitude of their
technical inefficiency effects is the same in all time periods. Thus, this model cannot
accommodate the situation where an initially relatively inefficient firm becomes relatively
more efficient (a change in relative ranking) in subsequent periods (Coelli, Rao and Battese,
1998).
15
6. Data
The analysis covers eight Central and Eastern European countries: Czech Republic, Estonia,
Hungary, Latvia, Lithuania, Poland, Slovakia and Slovenia – advanced transition countries –
and Malta and Cyprus, all being new EU members. We also included five old EU countries
(Austria, Belgium, Germany, Italy, Netherlands). Although there are differences between the
banking sectors of these countries, they nevertheless form a relatively homogeneous group. In
particular, preparations for EU membership and membership itself saw the installation of the
common EU legislative framework and common regulation standards. This allows us to
perform an efficiency analysis and compare estimated efficiencies across countries.
To construct the sample, we used information drawn from the financial statements of
individual banks provided by the Fitch IBCA’s BankScope database. Fitch IBCA collects data
from balance sheets, income statements and other relevant notes in audited annual reports. To
ensure consistency, only data for commercial banks in the unconsolidated format were used.
Data, expressed in euros, were collected for the 1996-2003 period and corrected for inflation
in order to ensure comparability (see the Table 3 for descriptive statistics of the data).
Mathieson and Roldos (2001) indicated three important characteristics of the BankScope
database. First, its comprehensive coverage as BankScope has data on banks accounting for
around 90% of total bank assets in each country. Second, comparability – the data-collection
process is based on separate data templates for each country to accommodate different
reporting and accounting standards. Fitch IBCA adjusts the collected data for country
specificities and presents them in a so-called global format, i.e. a globally standardised form
for presenting bank data. Thus, BankScope data is comparable across banks and across
countries, i.e. it allows cross-country comparisons (Claessens, Demirguc-Kunt and Huizinga,
2001). Third, BankScope provides balance sheet data for individual banks, which are usually
not available from other sources.
In specifying input prices and outputs of the cost function, we follow the intermediation
approach as suggested by Sealey and Lindley (1977). Three inputs (labour, funds and physical
capital) are used to produce three outputs (loans, other earning assets and deposits) (Table 2).
The three inputs reflect the three key groups of inputs in the bank production process: bank
personnel and the management expertise necessary for the provision of bank services
(labour), funds collected on the liabilities side (funds), and offices, branches and computer
hardware (physical capital).
Table 2: Input and output variables
Variable
Dependent
C
Name
Description
Total cost
Sum of labour, interest, physical capital and other costs
Variables
16
Input
w1
Price of labour
Personnel expenses over total assets
Prices
w2
Price of funds
Interest expenses over the sum of deposits, other funding
w3
Price of physical capital Depreciation over fixed assets
Output
y1
Total loans
Sum of short- and long-term loans, mortgages and other
quantities
y2
Other earning assets
y3
Total deposits
Sum of total securities, deposits with banks and equity
investments
Sum of demand and savings deposits, deposited by bank
and non-bank depositors
z
Equity capital
Other
variables
Total amount of equity capital
Source: Authors.
BankScope does not provide data on the price of labour ( w1 )directly, i.e. there is no
information on the number of employees to enable the construction of the ratio of personnel
expenses to the number of employees as the unit price of labour. Instead, we use the ratio of
personnel expenses over total assets, which is a common approach in bank efficiency studies
based on BankScope (Yildirim and Philippatos, 2002). Price of funds ( w2 ) was constructed as
the ratio of interest expenses over funding. Price of physical capital ( w3 ) also cannot be
directly taken from BankScope and was constructed as depreciation over fixed assets. The
three outputs, loans, other earning assets and deposits are proxies for banking services
provided. Total loans ( y1 ) is the total customer loans item from BankScope. Other earning
assets ( y 2 ) is the sum of total securities, deposits with banks and equity investments. Total
deposits ( y 3 ) is the sum of demand and savings deposits held by bank and non-bank
depositors. The dependent variable, total cost ( C ), is the sum of total operating expenses and
interest expenses. Equity capital ( z ) is the amount of bank equity that reflects both the size
and riskiness of banking operations.
Following Berger and Mester (1997), cost, and input prices were normalised by the price of
labour in order to impose homogeneity. Cost and output quantities were normalised by equity
to control for potential heteroscedasticity. Large banks have much larger costs (and profits)
than smaller banks, thus their random errors would have substantially larger variances if no
normalisation were performed. However, ratios of costs to equity vary much less across banks
of different sizes. As the inefficiency terms are derived from the (composite) random error,
the variance of the inefficiency term might be strongly influenced by bank size if it were not
for the normalisation by equity. Normalisation also allows the model a more economic
interpretation.
Table 3: Descriptive statistics of dependent variables, inputs and outputs for cost
17
Variable
Units
Mean
Std. Dev.
CV
Total assets
EUR mil
6,509
32,042
4.92
Total loans
EUR mil
3,303
15,837
4.80
Total other earning assets
EUR mil
2,811
14,787
5.26
Total deposits
EUR mil
4,420
22,789
5.16
Price of labour
%
1.81%
2.55%
1.41
Price of funds
%
13.05%
241.71%
18.52
Price of physical capital
%
2.64%
38.33%
14.52
Total cost
EUR mil
377
1,646
4.37
Total equity
EUR mil
355
1,460
4.11
Notes: Figures in EUR million are in 2003 prices.
Source: Authors’ calculations.
The sample of banks is not constant, i.e. we do not require a bank to have existed throughout
the sample period for it to be included in the sample. Thus, in the unbalanced panel the
number of banks across years varies for all countries. In Table 4 we summarize the number of
banks included in the sample in specific years and across sub-regions. The largest group
represent banks from the EU-5 countries and the smallest banks from Cyprus and Malta, but
we decided for this type of segmentation in order to assure the homogeneity across groups.
For the last year of the observation period (2003) the number of banks drops substantially in
all sub-regions due to data incompleteness in BankScope database. This might be also
reflected in the efficiency estimates for year 2003, so results for this year should be
interpreted cautiously.
Table 4: Number of banks across sub-regions:
Region
1996
1997
1998
1999
2000
2001
2002
2003
EU-5 countries
CEE-5 countries
Baltic countries
Cyprus and Malta
493
126
36
16
506
128
39
19
496
110
33
22
488
113
33
19
475
117
34
21
485
103
34
20
445
91
34
19
219
29
25
7
Total
671
692
661
653
647
642
589
280
Source: Authors’ calculations.
7. Results
In order to be able to make a cross-country comparison of cost efficiency, we employ a
common frontier function by pooling the data set of all banks comprising all countries
included in the analysis. The cross-country frontier function in the form of a translog function
18
was estimated for the 1996-2003 period by using an unbalanced panel data set on an annual
basis.
The translog functional form was specified as follows:
2
ln( C w3 z )   0    i ln( wi w3 ) 
i 1
3
   k ln( y k z ) 
k 1
1 2 2
 ij ln( wi w3 ) ln( w j w3 ) 
2 i 1 j 1
1 3 3
  km ln( y k z ) ln( y m z ) 
2 k 1 m 1
(6)
3
2
1
   ki ln( y k z ) ln( wi w3 )   1 ln z   2 (ln z 2 ) 
2
k 1 i 1
2
3
i 1
k 1
   i ln( wi w3 ) ln z    k ln( y k z ) ln z  ln v  ln u
where
C
yk
is total cost
is the k-th output
wi
is the i-th input price
z
is the equity capital
is measurement error term
is the inefficiency term
v
u
The duality theorem requires the cost function to be linearly homogeneous in input prices and
for the second-order parameters to be symmetric (Altunbas et al., 2001b).6 Therefore, the
following restrictions apply to the parameters of the cost function:

i
i
1

ij
 0 , for all i
i

ki
 0 , for all k
k
 km   mk , for all k, m
 ij   ji , for all i, j
The maximum likelihood method was applied for estimation. The inefficiency effects are
incorporated in the error term. The error term in a stochastic cost frontier model is assumed to
have two components. One component is assumed to have a symmetric distribution
(measurement error, v it ) and the other is assumed to have a strictly non-negative distribution
6
The duality theorem states that any concept defined in terms of the properties of the production function has a
dual definition in terms of the properties of the cost function and vice versa. See Varian (1992) for more details.
19
(inefficiency term, u it ). The estimation technique we use is based on the Battese-Coelli
(1992) parameterisation of time effects in the inefficiency term and accordingly the
inefficiency term is modelled as a truncated-normal random variable multiplied by a specific
function of time. The idiosyncratic error term is assumed to have a normal distribution. As is
always the case when implementing frontier estimation techniques, the efficiency score
acquired from the frontier function measures the efficiency of a specific bank relative to the
best-practice or most efficient bank.7
Since the aim of our work is not to investigate the reasons underlying cost efficiencies within
individual banks but to find evidence of the existence or inexistence of a cost-efficiency gap
in EU banking markets, we present the results as average efficiency scores for individual
countries and groups of countries that form sub-regions within the EU.
In the process of constructing the cost function, when we were altering the normalisation of
cost and input prices (normalisation with personnel cost vs. normalisation with other
operating costs) and when we were assessing specifications with three vs. four product
variables, we ended up with a three-product cost frontier function (loans, other earning assets,
deposits), normalised with personnel cost as a preferred cost function. The inclusion of offbalance-sheet items as a fourth product variable turned out to significantly reduce the total
number of observations, whereas the normalisations with personnel costs increased the
number of statistically significant coefficients.
We report selected summary statistics of the estimated translog function in Table 5. The
parameters  and  u2 represent the distributions parameters of the inefficiency effects,
parameter

is
the
decay
parameter
in
modelling
the
inefficiency
effects
uit  exp t  Ti ui as in Battese and Coelli (1992) and indicates the time dynamics of
measured inefficiencies. Parameter  indicates the proportion of the variance in disturbance
that is due to inefficiency,    u2 /( v2   u2 ) . The  value is high and shows that
inefficiency variation is more important than any stochastic variation in the frontier itself.
Table 5: Selected estimation results for the translog cost function specification
Coefficient
Ln(  )
2


4.980
Std. Err.
0.0020008
-451.518
-0.00128
7
0.0044023
Cost efficiency can take values between zero and one. For example, a bank with cost efficiency of 0.80 is 80%
efficient. In other words, the bank could improve its cost efficiency, i.e. reduce its costs, by 25%. The bank’s
cost inefficiency is 1-0.80=0.20.
20
-59.021
Log likelihood
u
v
12.062
0.2909844
0.198
0.0010922
0.999
7.53e-06
   /(   )
2
u
2
v
2
u
Source: Authors’ calculations.
The average efficiency scores calculated for the entire sample of countries/banks and for three
EU sub-regions (the EU-5 group, the CEE & Baltic countries group and Cyprus & Malta) are
reported in Table 6. The average efficiency score for every specific group of countries for
each year is obtained as a weighted average of individual banks’ efficiency scores, where the
relative importance of the total assets of specific banks is used as a weight for the bank. We
consider the weighting approach to be essential for the correct interpretation of the average
efficiency results for specific sub-regions.
Table 6: Average efficiency scores for the entire sample of banks, EU-5 countries,
CEE & Baltic countries, and Cyprus & Malta in the 1996-2003 period
Entire sample
EU-5
CEE & Baltic
Cyprus & Malta
Year
Mean
efficiency
score
Se(mean)
Mean
efficiency
score
Se(mean)
Mean
efficiency
score
Se(mean)
Mean
efficiency
score
Se(mean)
1996
1997
1998
1999
2000
2001
2002
2003
0.821
0.817
0.831
0.823
0.800
0.798
0.818
0.775
0.0075
0.0073
0.0082
0.0083
0.0086
0.0086
0.0091
0.0173
0.823
0.818
0.833
0.824
0.799
0.798
0.818
0.773
0.0088
0.0088
0.0099
0.0101
0.0108
0.0106
0.0111
0.0215
0.759
0.763
0.756
0.787
0.803
0.781
0.791
0.810
0.0111
0.0102
0.0112
0.0109
0.0098
0.0109
0.0103
0.0113
0.937
0.930
0.931
0.927
0.916
0.913
0.903
0.904
0.0057
0.0095
0.0084
0.0050
0.0142
0.0165
0.0251
0.0408
Total
0.815
0.0029
0.816
0.0036
0.783
0.0038
0.916
0.0062
SD
CV
SD
CV
SD
CV
SD
CV
0.169
0.207
0.171
0.209
0.112
0.143
0.064
0.070
Variability
measures
1996/2003
Source: Authors’ calculations.
The calculated average efficiency scores indicate that the average efficiency of the entire
sample of banks for the entire period was 0.815, meaning that banks were, on average, only
81.5% efficient and could reduce their costs by 23%. As one can see in Table 6 the average
efficiency score differs among regions, the highest being for Cyprus & Malta (91.6% efficient
banks) and the lowest for CEE & Baltic countries (78.3%). The differences in efficiency
scores proved to be statistically significant at p<0.05. Although the efficiency scores vary in
time the rankings between all three regions remain unchanged. Time dynamics statistics are
reported later in this section.
21
The measured variability reported in Table 6 demonstrates the differing variability of
efficiency scores in specific regions. While the coefficient of variation (CV) for the entire
sample amounts to 0.207 (standard deviation SD=0.169), the CV statistics for CEE and Baltic
countries do not exceed 0.15 (SD=0.112). The CV statistics for Cyprus & Malta even lie
below 0.10 (SD=0.064). The highest efficiency variability among all three regions is
estimated for the EU-5 group (CV=0.209 and SD=0.171), which might, at least partially, also
be a consequence of the large number and great diversity of banks included in the analysis
being from that region .
As is evident from Table 7, the variability of efficiency scores differs widely among the
countries included in the sample. So the lowest efficiency variability was recorded for Cyprus
(CV = 0.037) and the highest, somewhat surprisingly, in the case of Netherlands (CV = 0.61).
The latter is a direct result of the surprisingly low average efficiency level (Eff = 0.305),
which could also be a consequence of the weighting scheme that was deployed for calculating
the average efficiencies.
Table 7: Average efficiency scores and statistics for individual countries
Region
Baltic
CEE-5
CY&MT
EU-5
Country
Mean
Efficiency
Estonia
0.739
Latvia
0.741
Lithuania
Czech R.
SD
Se(Mean)
N
Max
Min
CV
0.0461
0.0067
47
0.895
0.498
0.0624
0.0561
0.0069
67
0.934
0.502
0.0757
0.672
0.0760
0.0063
147
0.962
0.487
0.1131
0.653
0.0703
0.0062
127
0.911
0.352
0.1076
Hungary
0.786
0.0928
0.0106
77
0.948
0.613
0.1182
Poland
0.875
0.0384
0.0025
236
0.975
0.562
0.0439
Slovenia
0.823
0.0600
0.0065
84
0.967
0.621
0.0729
Slovakia
0.759
0.1146
0.0116
98
0.906
0.035
0.1510
Cyprus
0.912
0.0340
0.0046
54
0.967
0.782
0.0373
Malta
0.919
0.0804
0.0109
54
0.954
0.410
0.0875
Austria
0.841
0.1820
0.0124
214
0.954
0.074
0.2163
Belgium
0.850
0.1461
0.0097
225
0.963
0.290
0.1719
Germany
0.844
0.1406
0.0043
1061
0.970
0.162
0.1665
Italy
0.778
0.1520
0.0057
713
0.974
0.200
0.1954
Netherlands
0.305
0.1855
0.0199
87
0.980
0.231
0.6087
Source: Authors’ calculations.
The average efficiency results for other countries do not reveal any surprising findings. In the
EU-5 group, apart from Dutch banks only Italian banks (Eff = 0.778) turned out to be less
efficient than most banks in other EU-5 peer countries. Among 15 banking markets included
in the analysis, the banking sectors of Cyprus and Malta clearly stand out by their superior
efficiency since their average efficiency reaches 91% in the case of Cyprus and 92% in the
case of Malta. Obviously, these two countries have banking sectors that are more advanced
22
than the banking sectors of CEE and Baltic countries even though they joined the EU
together. However, a more detailed analysis would be necessary in order to isolate the factors
making these two countries’ banking sectors superior in cost efficiency, not only in
comparison with CEE and Baltic countries but also relative to old EU members.
Another important aspect of bank efficiency studies that needs to be addressed is the time
dynamics of banking efficiency. The time varying decay model developed by Battese and
Coelli (1992) models inefficiency effects as: uit  exp  (t  Ti )ui . The estimated 
coefficient provides information on the time dynamics of inefficiency effects. When   0 ,
the degree of inefficiency is decreasing over time and when   0 , the degree of inefficiency
is increasing over time. For the purpose of  coefficients estimations for specific regions, we
estimated a cost frontier function for each region separately. The estimated  coefficients are
presented in Table 8. The  coefficients for the entire sample, for EU-5 countries and for
Cyprus & Malta are negative, indicating increasing average bank inefficiencies for the entire
sample and both sub-regions. However, none of the estimated coefficients turns out to be
statistically significant, meaning that any conclusions about time dynamics are unreliable. On
the contrary, the estimated coefficient for CEE & Baltic countries proved to be positive and
statistically significant at p<0.01, indicating that average banking efficiency in this group of
countries improved in the 1996-2003 period. This finding is encouraging since it shows that
the reforms conducted in CEE & Baltic countries in the 1990s and the adoption of EU
legislative and regulation standards did in fact contribute to significant improvements in the
banking sectors of these countries.
Table 8: Time dynamics of banking efficiency in the entire sample and
three sub-regions within the EU for the 1996-2003 period

coefficient
Std. Err
z
P>|z|
Log likelihood
EU-5
-0.0039
0.0046
-0.84
0.40
102.6758
CEE & Baltic
Cyprus & Malta
0.0535
-0.3132
0.0128
0.2016
4.18
-1.55
0.00
0.12
-5.5397
63.4700
Entire sample
-0.0013
0.0044
-0.29
0.77
-59.0210
Region
Source: Authors’ calculations.
8. Conclusions
The world banking industry has been undergoing an unprecedented wave of consolidation,
which reached its peak in 2000. The main driving force of the consolidation process in the
banking industry has been value maximisation of bank shareholders. Value creation in the
23
banking industry can be usually realised through gains in market power and/or efficiency
gains. Hence, efficiency gains seem to be also a reflection of the bank consolidation process.
In Europe, consolidation processes have overlapped with significant restructuring and
transformation processes in banking. In the EU banking sectors, changes and innovations (e.g.
implementation of the supranational legislation framework in the form of banking directives)
have been directed towards the establishment of a single market for financial services,
including a single banking market.
Banking sectors of most Central and Eastern European transition countries underwent a
profound restructuring process, which included a comprehensive consolidation and, after an
initial increase, a reduction in the number of banks. The question of potential efficiency gains
achieved through the consolidation processes in the enlarged EU banking market is a central
question of our analysis.
We have applied the standard efficiency measurement methodology to estimate the average
cost efficiency for selected countries and geographical regions. We have used the stochastic
frontier approach and the Battese and Coelli (1992) specification of the technical efficiency
model with a truncated normal distribution of efficiency effects and a time varying decay
model. Data was obtained from the BankScope database for ten new and five old EU
member-countries. The unbalanced panel covers the 1996-2003 period. As expected, the
results confirm the existence of an East-West efficiency gap since banks in the old EU
countries proved, on average, to be significantly more cost efficient than their counterparts in
the CEE and Baltic countries. The analysis of time dynamics showed that average bank
inefficiency in CEE and Baltic countries decreased over the entire period, whereas statistically
significant changes in the time dynamics of the average efficiency of EU-5 banks were not
confirmed. This might be an indication that the East-West efficiency gap has been gradually
narrowing. It is likely that this process will continue in the future until a complete or at least a
high degree of convergence in average bank efficiency is achieved.
The consolidation wave in the European banking seems to have spurred significant
improvements in the efficiency of the banking operations, particularly in the former transition
countries. However, the efficiency gains are not evenly distributed across banking sectors of
different countries. It will probably take some time until the efficiency levels of banking
sectors in the new EU countries reach those in the old EU countries.
24
References
Aigner, D., Lovell, C. A. K. and Schmidt, P. (1977): "Formulation and estimation of
stochastic frontier production function models", Journal of Econometrics, Vol. 6, pp.
21-37.
Amel D., Barnes C., Panetta F. and Salleo C. (2004): »Consolidation and efficiency in the
financial sector: A review of the international evidence«, Journal of Banking and Finance,
Vol. 28, Issue 10, pp. 2493-2519.
Ayadi, R. and Pujals, G. (2004): Banking Consolidation in the EU: Overview and Prospects.
CEPS Research Report in Finance and Banking No.34.
Baele L., Ferrando A., Hördahl P., Krylova E. and Monnet C. (2004): “Measuring financial
integration in the Euro area”, European Central Bank, Occasional paper series, No. 14,
April 2004, 95 p.
Battese, G. E. and Coelli, T. J. (1988): "Prediction of Firm-Level Technical Efficiencies With
a Generalised Frontier Production Function and Panel Data", Journal of Econometrics,
Vol. 38, pp. 387-399.
Battese, G. E. and Coelli, T. J. (1992): "Frontier Production Functions, Technical Efficiency
and Panel Data With Application to Paddy Farmers in India", Journal of Productivity
Analysis, Vol. 3, pp. 153-169.
Berg, S. A., Forsund, F. R., Hjalmarsson, L. and Suominen, M. (1993): "Banking efficiency in
the Nordic countries", Journal of Banking and Finance, Vol. 17, pp. 371-388.
Berger, A.M., DeYoung, R. and Udell, G.F. (2000): “Efficiency Barriers to the Consolidation
of the European Financial Services Industry”, European Financial Management, Vol. 6.
Berger, A. N. and Humphrey, D. B. (1997): "Efficiency of Financial Institutions: International
Survey and Directions for Future Research", European Journal of Operational Research,
Vol. 98, pp. 175-212.
Berger, A. N. and Mester, L. J. (1997): "Inside the black box: What explains differences in the
efficiencies of financial institutions?", Journal of Banking and Finance, Vol. 21, pp.
895-947.
Berger, A. N., Hunter, W. C. and Timme, S. G. (1993): “The efficiency of financial
institutions: A review and preview of research past, present, and future”, Journal of
Banking and Finance, Vol. 17, No. 2-3, pp. 221-249.
Bonin J. P., Hasan I. and Wachtel P. (2004): “Bank performance, efficiency and ownership in
transition countries”, Journal of Banking and Finance, forthcoming
Cabral I., Dierick F. and Vesala J. (2002): »Banking integration in the Euro area«, European
Central Bank, Occasional paper series, December 2002, 52 p.
Chang, E. C., Hasan, I. and Hunter, W. C. (1998): "Efficiency of multinational banks: an
empirical investigation", Journal Applied Financial Economics, Vol. 8, pp. 689-696.
25
Claessens, S., Demirguc-Kunt, A. and Huizinga, H. (2001): "How does foreign entry affect
domestic banking markets?", Journal of Banking and Finance, Vol. 25, pp. 891-911.
Coelli, T. J., Rao, D. S. P. and Battese, G. E. (1998): An Introduction to Efficiency and
Productivity Analysis. Boston: Kluwer Academic Publishers.
DeYoung, R. (1997): "Measuring Bank Cost Efficiency: Don't Count on Accounting Ratios ",
Financial Practice & Education, Vol. 7, No. 1, pp. 20-31.
Farrell, M. J. (1957): "The Measurement of Productive Efficiency", Journal of the Royal
Statistical Society, Vol. 120, pp. 253-281.
Grigorian, D. A. and Manole, V. (2002): "Determinants of Commercial Bank Performance in
Transition: An Application of Data Envelopment Analysis". IMF Working Paper No. 146.
World Bank.
Jondrow, J., Knox Lovell, C. A., Materov, I. S. and Schmidt, P. (1982): "On the estimation of
the technical inefficiency in the stochastic frontier production function model", Journal of
Econometrics, Vol. 19, pp. 233-238.
Kaparakis, E. I., Miller, S. M. and Noulas, A. G. (1994): "Short-Run Cost Inefficiency of
Commercial Banks: A Flexible Stochastic Frontier Approach", Journal of Money, Credit,
and Banking, Vol. 26, No. 4, pp. 875-893.
Leibenstein H. (1966): “Allocative Efficiency vs. "X-Efficiency"”. American Economic
Review, Vol. 56, No. 2, pp. 392-415.
Mathieson, D. J. and Roldos, J. (2001): "Foreign Banks in Emerging Markets". In Litan, R.
E., Masson, P. and Pomerleano, M. (editors): Open Doors - Foreign Participation in
Financial Systems in Developing Countries. Washington DC: Brookings Institution Press,
pp. 15-55.
Meeusen, W. and van den Broeck, J. (1977): "Efficiency Estimation from Cobb-Douglas
Production Functions with Composed Error", International Economic Review, Vol. 18,
No. 2, pp. 435-444.
Molyneux, P., Altunbas, Y. and Gardener, E. (1996): Efficiency in European banking.
Chichester: John Wiley & Sons Ltd.
Schure, P. and Wagenvoort, R. (1999): Economies of Scale and Efficiency in European
Banking: New Evidence. Economic and Financial Reports 99/01. Luxembourg: EIB.
Sealey Jr., C. W. and Lindley, J. T. (1977): "Inputs, Outputs, and a Theory of Production and
Cost at Depository Financial Institutions", Journal of Finance, Vol. 32, No. 4, pp. 12511266.
Yildirim, H. S. and Philippatos, G. C. (2002): "Efficiency of banks: Recent evidence from the
transition economies of Europe 1993-2000". University of Tennessee.
26