Jurnal 1

Large Bank Efficiency in Europe and the United States: Are There Economic Motivations
for Geographic Expansion in Financial Services?

Author(s): JaapW.B. Bos and JamesW. Kolari
Source: The Journal of Business, Vol. 78, No. 4 (July 2005), pp. 1555-1592
Published by: The University of Chicago Press
Stable URL: http://www.jstor.org/stable/10.1086/430869
Accessed: 19-11-2016 12:03 UTC
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted
digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about
JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms
The University of Chicago Press is collaborating with JSTOR to digitize, preserve and extend access to
The Journal of Business
This content downloaded from 36.86.56.161 on Sat, 19 Nov 2016 12:03:56 UTC
All use subject to http://about.jstor.org/terms
Jaap W. B. Bos
Utrecht University
James W. Kolari
Texas A&M University
Large Bank Efficiency in Europe

and the United States: Are There
Economic Motivations for
Geographic Expansion in
Financial Services?*
I. Introduction
Deregulation of the banking industry in Europe and

the United States in the 1980s and 1990s stimulated an unprecedented merger and consolidation
wave (see Berger and Strahan 1998). However,
most geographic expansion has been limited to continental borders, rather than crossing over the Atlantic Ocean (Berger and Humphrey 1997). In view
of the changes in European and U.S. banking laws,1
* We thank seminar participants at Maastricht University
and De Nederlandsche Bank for their helpful remarks. We are
especially thankful for valuable comments from Jaap Bikker,
Lawrence Goldberg, Clemens Kool, Iman van Lelyveld, and Erik
de Regt on earlier drafts of this paper, as well as participants at the
2002 annual meetings of the Financial Management Association
in Copenhagen, Denmark, and an anonymous referee. We also
thank Alan Montgomery for excellent research assistance. We
gratefully acknowledge financial support from the Netherlands
Organization for Scientific Research ( NWO), De Nederlandsche
Bank, and Center for International Business Studies, Mays Business School, Texas A&M University. The second author is the
Chase Professor of Finance. Contact the corresponding author,
James W. Kolari, at j-kolari@tamu.edu.
1. In the European Union, banking legal reforms include the
Second Banking Coordination Directive of 1988 and the establishment of the single market for financial services in 1993. In the
United States, the Riegle-Neal Interstate Banking and Branching
Act of 1994 and the Financial Services Modernization Act of
1999 expanded the geographic and financial services powers of
banking institutions.
This paper employs

stochastic frontier cost
and profit models to
estimate the efficiency
of multibillion dollar
European and U.S.
banks. Empirical results
suggest that both large
European and U.S.
banks have decreasing
(increasing) cost (profit)
returns to scale. Also,
large banks in Europe
and the United States
similarly exhibit
increasing returns to
scale and decreasing
(increasing) scope
economies for the cost
(profit) model.
However, large U.S.
banks have higher
average profit efficiency
than European banks on
average. We conclude
potential efficiency
gains are possible via
geographic expansion of
large European and U.S.
banks.
(Journal of Business, 2005, vol. 78, no. 4)

B 2005 by The University of Chicago. All rights reserved.
0021-9398/2005/7804-0014$10.00
1555
1556
Journal of Business
the next expansion wave can be expected to involve an international

consolidation. Over the past half-century, financial systems in Europe
and the United States have become increasingly integrated by virtue of
the international flow of money and capital in securities markets. In the
near future, the process of integration may be completed by increasing
the structural overlap among financial institutions via cross-Atlantic
consolidation.2
The prospect of joint European-U.S. consolidation of financial services
raises numerous questions. Will such intercontinental expansion result in
public gains in terms of the quality and prices of financial services? What
driving forces motivate large banks in Europe and the United States to
expand? In this regard, could mega-institutions combining geographically
dispersed operations be more efficient in terms of costs and profits than
otherwise? If increasing returns to scale are available to large banking
institutions, an economic rationale for further geographic expansion exists.
Moreover, efficiency differences between large banks in Europe and the
United States imply that there is an opportunity to enhance efficiency via
cross-Atlantic expansion.3 Of course, aside from cost and profit considerations, cross-Atlantic consolidation of large banks could be motivated
by a variety of other factors, including market power, diversification, and
management incentives (e.g., see Milbourn, Boot, and Thakor 1999).
This paper seeks to examine the question of economic motivations
for cross-Atlantic geographic expansion by conducting stochastic frontier cost and profit analyses to estimate economies of scale as well as
X-efficiency for multibillion dollar European and U.S. banks in the period
199599. Based on separate analyses of large European and U.S. banks,
we find that cost and profit functions for banks in both regions are strikingly similar, with increasing returns to scale and decreasing (increasing)
scope economies for the cost (profit) model. Importantly, X-efficiency
scores derived from these cost and profit models reveal that, on average,
European banks have lower cost and profit efficiencies than U.S. banks,
2. We refer to geographic expansion as any means by which a bank can expand geographically, both within and across borders. As such, geographic expansion is more encompassing than bank mergers. A second and related reason for our focus on geographic expansion
is that there have been very few cross-Atlantic mergers to date, which would worsen the Lucascritique in any test of these mergers success.
3. The literature on the performance benefits of bank mergers is mixed (see Berger and
Humphrey 1992; Rhoades 1994; and Peristiani 1997). However, a recent study by Rhoades
(1998) performed case studies of nine large U.S. bank mergers and found that cost efficiency
improved in each case. Seven of the bank merger cases exhibited increased profitability
relative to their peers. In general, acquiring (acquired) banks were more (less) efficient than
their peer group. Thus, while not all mergers yielded increased performance, most large bank
combinations provided economic benefits relative to peer banks. Moreover, Siems (1996)
reported evidence that shareholder wealth significantly increased in response to megamergers of U.S. banks in 1995. We infer that the results tend to support in part the synergy
hypothesis, whereby acquiring banks reap economies of scale and scope via cutting costs of
redundancies and duplication of operations.
Large Bank Efficiency
1557
and the dispersion of both profit and cost efficiency scores is considerably
smaller for U.S. banks than for European banks.
Further analyses evaluate the reasonableness of estimating a combined
cost or profit frontier for European and U.S. banks. A necessary condition
for comparable-shaped frontiers is for economies of scale to be similar
among efficient banks in Europe and the United States. We therefore test
for differences in economies of scale by moving progressively closer to
the frontier in an effort to evaluate the appropriateness of estimating a
single frontier for both regions. In brief, we find that, while no single cost
frontier exists, a single profit frontier is implied. U.S. banks tend to
exhibit higher average profit and cost efficiency than European banks in
general and compared to most individual European countries. Further
results for small banks reveal that they are less cost and profit efficient
than large banks in Europe and the United States. We conclude that,
based on profit model evidence of both increasing returns to scale and
differences in cost and profit X-efficiency, our empirical results tend to
support the notion that potential profit efficiency gains are possible in
cross-Atlantic bank mergers between European and U.S. banks.
Section II is an overview of the related literature on the efficiency
of European and U.S. banks. Section III describes our methodology,
Section IV gives details of the data, and Section V reports the empirical
results. The last section gives conclusions.
II. Related Literature
A large body of literature spanning a half-century exists on banking

efficiency in the United States (e.g., see surveys in Berger and Humphrey
1997; Berger and Strahan 1998; and Berger, Demsetz, and Strahan 1999).
Likewise, a more recent but growing literature on European banking
efficiency is developing (e.g., see Molyneux, Altunbas, Gardener 1997;
Sheldon 1999; and Altunbas, Gardener, Molyneux, and Moore 2001).
Studies prior to the 1980s tended to report U-shaped cost curves with
economies of scale exhausted by $100$500 million for the most part.
Results for economies of scope (i.e., joint production of outputs) were
mixed, with most authors concluding that banks do not gain efficiencies
from providing multiple financial services to the public. Altering the path
of efficiency research, Berger and Humphrey (1991) showed that U.S.
banks could improve their cost efficiency more by reducing frontier
inefficiencies than by reaching some optimal level of scale and scope
economies to minimize average costs. Subsequent research further investigated this issue using both parametric and nonparametric frontier
estimation methods (e.g., see Lovell 1993). Moreover, recent research
has expanded the analyses to consider both cost and profit efficiency
(e.g., see Berger and Humphrey 1997; Berger and Mester 1997; and
others), as well as risk variables (e.g., see Berg, Frsund, and Jansen 1992;
1558
Journal of Business
McAllister and McManus 1993; Mester 1996a, b; Berger and DeYoung

1997; and others). In general, studies have confirmed Berger and
Humphreys result that cost and profit frontier inefficiencies outweigh
output inefficiencies associated with scale and scope economies by a
considerable margin.
A major gap in early bank efficiency literature was the scant evidence
on large banks. This shortfall was substantial due to the pivotal role of
large institutions in shaping the structure of the banking industry. If large
bank size is closely related to the efficient production of financial services, the implication is that the post-deregulatory consolidation movement will result in a highly concentrated banking industry dominated by a
relatively small number of institutions. Conversely, if inefficiencies occur
as banks expand output beyond some size threshold, it is likely that the
organizational structure of the industry would be less concentrated, with
a larger number of banks offering services to the public. Without research
on large-bank cost and profit efficiencies, no inferences about deregulation and related policy implications to banking industry structure could
be made.
In the 1980s and 1990s, numerous studies on large U.S. banks were
published to overcome this shortfall in the literature.4 Summarizing this
literature, scale economies were found for banks between $1 billion and
$15 billion in assets with diseconomies thereafter. The existence of scope
economies was more elusive, with most studies reporting insignificant
results. One exception is that, upon resolving some econometric problems plaguing previous work in this area, Pulley and Humphrey (1993)
reported significant scope economies in the joint production of two types
of deposit services and three credit areas. Finally, consistent with Berger
and Humphrey, studies have confirmed that frontier inefficiencies are far
greater than scale and scope inefficiencies.
To our knowledge no European studies have focused on large bank
efficiency per se; instead, banks of different sizes are comparatively
examined.5 Only a few studies have been published on the subject of
large bank efficiency outside the United States. Allen and Rai (1996)
4. See studies by Hunter and Timme 1986; Shaffer and David 1986; Kolari and
Zardkoohi 1987; Elyasiani and Mehdian 1990; Evanoff and Israilevich 1990; Hunter, Timme,
and Yang 1990; Noulas, Ray, and Miller 1990; Berger, Hunter, and Timme 1993; Saunders
and Walter 1994; Hunter 1995; Jagtiani, Nathan, and Sick 1995; Jagtiani and Khanthavit
1996; Miller and Noulas 1996; Mitchell and Onvural 1996; Rhoades 1998; and Rogers
1998.
5. For example, see studies by Berg et al. 1992; Berg et al. 1993; Fecher and Pestier 1993;
Pastor, Perez, and Quesada, 1997; Vander Vennet 1994, 1999; Altunbas and Molyneux 1996;
Griffell-Tatje and Lovell 1996; Ruthenberg and Elias 1996; Pinho 1994; Economic Research
Europe 1997; Mendes and Rebelo 1997; Molyneux et al. 1997; Resti 1997; Altunbas and
Chakravarty 1998; Battese, Heshmati, and Hjalmarsson 1998; Dietsch, Ferrier, and Weill
1998; Altunbas, Goddard, and Molyneux 1999; Bikker 1999; Sheldon 1999; Berger et al.
2000; Hassan, Lozano-Vivas, and Pastor 2000; and Altunbas et al. 2001.
1559
estimated a global cost function for 194 banks in 15 countries (including

the United States). Banks were divided by asset size into two groups (i.e.,
small versus large banks below and above the median asset size of a
countrys banks, respectively) and by regulatory environment (i.e., countries with functionally integrated or universal banking versus those with
functionally separated banking). Since the median bank asset size in most
countries exceeded $40 billion U.S., most banks in the study can be considered large banks. They found that universal banking countries were
more efficient than nonuniversal banking countries, which implies that
banks offering a variety of financial services (e.g., loans, deposits, insurance, securities investment, real estate, etc.) are more competitive
than banks offering selected financial services. Also, France, Italy, the
United Kingdom, and the United States had the most inefficient banking
institutions.
Another study by Saunders and Walter (1994) measured scale and
scope economies for 133 of the largest 200 banks in the world at year-end
1988. Based on a translog cost model, they found that, while banks with
loans less than $10 billion and more than $25 billion exhibited scale
diseconomies, banks in the middle range had scale efficiencies. Also,
scope diseconomies between fee-earning and interest-earning financial
services existed. Given that their sample banks typically operated in
multiple countries, they concluded that international expansion may well
offer economies of scale opportunities for many financial institutions.
They also concluded that it was too early in the 1980s to make clear
inferences about potential scope economies, which they believed might
materialize after some initial fixed costs of expanding beyond traditional
commercial banking activities had been incurred.
The aforementioned large bank studies are unique from other European work in that consolidated data are employed in the analyses. By
consolidating the data from the entire organization, they provide insight
into bank efficiency at the firm level, rather than at divisional or branch
levels. Other efficiency studies of European banking provide some large
bank results, but they typically employ unconsolidated data. For example, Berger et al. (2000) reported efficiency results for European banks
with assets exceeding $100 million; however, they are careful to observe
that their results are relevant only for subsidiaries of banking organizations due to the use of unconsolidated data. In this regard, they noted
that transfer pricing can shift profits from one affiliate to another and
affect efficiency estimates for subsidiaries. While they did not believe
that this potential bias was significant in their study, it is not possible to
determine the extent to which transfer pricing, shared inputs, and other
intraorganizational arrangements might affect efficiency assessments.
The authors estimated separate cost and profit functions for 2,123 commercial banks (i.e., other types of banks are excluded) with data available
for the period 199398 in France, Germany, Spain, the United Kingdom,
1560
Journal of Business
and the United States. Comparative analyses revealed that domestic

banks on average had higher cost and profit efficiency than foreign banks
in these countries. However, when the results were disaggregated on
a country-by-country basis, they found that foreign banks from some
countries were equal to or more efficient than domestic banks. Interestingly, U.S. banks overseas tended to be more efficient than domestic
banks in their respective countries. They inferred that, since there is not
necessarily a home field advantage, additional global consolidation is
likely in the future. Relevant to our analyses, they recommended that
future empirical work in this area should expand the analyses to a substantial number of countries.
Another study by Sheldon (1999) used unconsolidated data for 1,783
commercial and savings banks in the European Union, Norway, and
Switzerland for the period 199397. Data envelopment analysis (DEA)
was employed to examine cost and profit efficiency. They found that
large banks, specialized banks, and retail banks are more cost and profit
efficient than small banks, diversified banks, and wholesale banks, respectively. Average frontier efficiency was fairly low, at about 45% for
costs and 65% for profit. Banks in Denmark, France, Luxembourg, and
Sweden had the highest average efficiency, and banks in Greece, Italy,
Portugal, Spain, and United Kingdom had the lowest average efficiency.
Furthermore, estimates of economies of scale in costs and profits indicated that most banks in their sample were suboptimal in size, with
optimal scales in the range of $0.5 to 1.5 billion U.S. Decreasing cost and
profit returns to scale were reported for most multi-billion dollar banks.
These and other results led them to conclude that inefficient operations,
rather than unexploited economies of scale, explain cost and profit differences across European banks. The authors inferred that diseconomies of large scale with respect to both costs and profits prohibit a high
degree of industry concentration in the European banking market.
In view of previous literature, the present study contributes new
empirical evidence on large bank efficiency. Comparative analyses of
European and U.S. large banks provide perspective in understanding
similarities and differences between institutions located in these two
regions of the world. Unlike previous European bank studies, we employ
consolidated data for independent banks, as opposed to all entities contained within the organization. While some studies do use consolidated
data, they do not exclude banking entities for which the organization does
not have a controlling interest (e.g., at least 25% ownership of common
stock).6 Instead, they include banks that are under controlling interest by
6. We develop independent bank accounting statements for European banks by including
only entities that are not controlled by other firms. Most research has used consolidated data
from IBCA/ BankScope. However, BankScope has ownership data that are kept on record for
only the last year of a sample period; moreover, as noted by the publisher of BankScope,
Bureau van Dijk, ownership data are incomplete and depend upon availability. National
1561
others as separate entities. Consequently, these studies suffer from double

counting and tend to include banks that are not independent organizations.
To a significant extent, this also runs counter to the profit-maximization
and cost-minimization assumptions behind the efficiency measures employed, where banks are assumed to choose the input-output mix that is
most efficient. Finally, we extend prior work by Allen and Rai (1996) and
Saunders and Walter (1994) by collecting larger samples of multibillion
dollar banks and updating analyses to the last half of the 1990s. Like Allen
and Rai, we give some comparative results for small, independent banks as
well. We next turn to further details of our methodology.
III. Methodology
A. Bank Efficiency7
The general concept of efficiency refers to the difference between observed and optimal values of inputs, outputs, and input-output mixes.
Efforts to measure how efficiently a firm produces outputs from its inputs
led to the development of a number of efficiency concepts, including scale
efficiency, scope efficiency, and X-efficiency. Berger et al. (1993) define
X-efficiency as the economic efficiency of any single firm minus scale
and scope efficiency effects, thereby allowing for suboptimal (beneathfrontier) operations.8 In this paper we employ stochastic frontier models
that allow us to measure scale and scope efficiency as well as X-efficiency.
According to Berger and Humphrey (1991) and Berger et al. (1993), the
significance of scale and scope inefficiencies (amounting to about 5%) is
less important in the banking industry than X-inefficiencies (in the range
of 2025%).9
We use stochastic frontier (SFA) models for two important reasons:
(1) they allow for measurement error, which is an important feature in
light of the fact that measuring bank production can be difficult due to
data availability and the choice of a set of inputs and outputs; and (2) they
generate firm-specific efficiency estimates, which allow us to test for
accounting standards require majority ownership to be published in some countries, whereas
for others this information is provided on a voluntary basis. With the kind assistance of Mark
Wessels and Patrick Oosterling of Bureau van Dijk, we were able to manually retrieve updated
ownership and control data from the 19952000 issues of BankScope. Gaps in these reports
were filled by gathering further data from Dun and Bradstreet Linkages, Reuters News, and
banks annual reports. In this way, we constructed a complete set of large, independent commercial banks for each of the years included in our sample.
7. We refer to Lovell (1993) and Coelli, Prasado Rao, and Battese (1998) for an in-depth
discussion of different efficiency measures. See also Berger et al. (1993) for an excellent
overview of the use of different efficiency concepts in banking.
8. Economic efficiency is the sum of technical and allocative efficiency. Technical efficiency is a measure of a banks distance from the frontier, minimizing inputs given outputs
or vice versa. Allocative efficiency measures the extent to which a bank is able to use inputs
and outputs in optimal proportions given prices and the production technology.
9. See also Berger and Humphrey (1997) and Molyneux et al. (1997).
1562
Journal of Business
differences in efficiency among banks from different countries as well as

measure the scale and scope economies of banks that operate close to the
frontier.10
B. Model Specification
We employ SFA cost and profit models similar to those in Berger and
Mester (1997), Humphrey and Pulley (1997), and DeYoung and Hassan
(1998). Banks are assumed to face perfectly competitive input markets
but operate in output markets where price differentiation is possible. This
framework easily accommodates our cost model with only trivial modifications. It also allows for market power in our profit model. Hence,
banks can compete via their output pricing strategies by adjusting prices
and fees according to market conditions. The extent to which they can
influence prices depends on output quantities, input prices, and other
factors, all of which are given at the time of price setting. Additional
features of the profit model are that it can account for differences in the
quality of outputs (to the extent that it is reflected in prices) as well as
correct for scale bias. Also, output prices, which are subject to severe
measurement problems according to Berger and Mester (1997) and
Vander Vennet (1999), are not required for the empirical analysis.11
For the estimation of the cost and profit frontier functions, we use the
translog functional form. This form has been widely employed and allows for the necessary flexibility when estimating frontier models. Berger
and Mester (1997) compared the translog to the Fourier flexible form
(FFF). Despite the latters added flexibility, the difference in results between these methods appears to be negligible (see also Swank 1996).
Moreover, previously cited bank efficiency studies show that the translog
cost and profit functions are locally stable in large bank applications.
We define profit before tax as PBT, outputs as y, and input prices as w.12
Also, we let control variable z reflect differences in risk-taking behavior
of banks. We also include linear and quadratic trend terms. For the specific choice of variables, we refer to Section IV. In the specification that
follows, the optimal profit level for bank k in period t is now a function
of the number of outputs, input prices, and the control variable z. In a
10. Concerning the measurement of X-efficiency, Bauer et al. (1997) imposed six consistency conditions and examined the extent to which stochastic frontier (SFA) models, thick
frontier models ( TFA), distribution-free models ( DFA), and data envelopment analysis ( DEA)
meet these consistency conditions. They found that the choice among these models did
not appear to significantly alter efficiency measures. In a study comparing DEA and SFA,
Eisenbeis, Ferrier, and Kwan (1999) reported that SFA efficiency scores were more closely
related to risk-taking behavior, managerial competence, and bank stock returns than those for
DEA.
11. For a theoretical framework for the SFA models used here, see Coelli et al. (1998) and
Bos (2002).
12. With respect to notation, we use lower case symbols in italics to denote logarithms.
Upper case symbols represent actual values of the variables.
1563
three-input, three-output translog setting, u and v are the inefficiency and

random error terms, respectively, and ai ; aij ; bi ; bij ; ci; di ; dij ; ei ; fi ; gi ; and
hi are parameters:
PBTkt y; w; z a0
3
X
ai yikt
i1
3 X
3
1X
i1 j1
3
X
3 X
3
3
X
1X
aij yikt yjkt
bi wikt
2 i1 j1
i1
3 X
3
X
1
bij wikt wjkt c1 zkt c2 zkt2
dij wikt yjkt
2
i1 j1
eij wikt zkt
i1
3
X
i1
1
fi yikt zkt g0 T g1 T 2 vkt ukt
2
1
The error term vkt is normally distributed, independently and identically

distributed (i.i.d.) with vkt N 0; s2v . The inefficiency term ukt is drawn
from a nonnegative half-normal distribution truncated at m and i.i.d. with
ukt jN m; s2u j. It carries a negative sign because all inefficient firms operate below the efficient profit frontier. The trunctation mean m results from the
maximum likelihood estimator (MLE). For the estimations involving U.S.
and European banks, we also use n1 dummies for each of the European
countries. For the cost model the left-hand side is replaced with the log of total
costs (after an identical transformation) and the inefficiency term ukt carries a
positive sign, as all inefficient firms operate above the efficient cost frontier.
Duality requires the imposition of symmetry and linear homogeneity
in input prices to estimate our cost and profit models (see Beattie and
Taylor 1985 and Lang and Welzel 1999):
aij taji for all i; j;
bij bji for all i; j;
3
X
bi 1;
i1
3
X
j1
bij 0 for all i;
3
X
i1
bij 0 for all j;
3
X
eij 0;
i1
3
X
dij 0:
i1
We impose linear homogeneity in input prices by normalizing the dependent variable and all input price variables (w) before taking logarithms (see
Coelli et al. 1998).13 Profit efficiency for bank k th time t is defined as14

PEkt E expukt "kt :
13. Each of these variables is included as a ratio to one of the input price variables, and the
coefficient for each input price is inferred ex post from the imposed restriction. This procedure
ensures homogeneity of only degree one in factor prices. Imposing constant returns to scale
would require normalization of the output variables as well.
14. The complete derivation of the maximum likelihood estimator is available upon request.
1564
Journal of Business
This measure takes on a value between 0 and 1 (fully efficient) and indicates how close a banks profits (conditional on its outputs, input prices, and
the control variable) are to the profits of a fully efficient bank under the
same conditions. Cost efficiency also takes on a value between 0 and 1
(fully efficient) and is defined as15
1

CEkt E expukt "kt
:
3
In estimating our profit and cost models, we apply the usual reparameterization by replacing s2u and s2v with s2 s2u s2v and g s2u =s2u s2v .16
We evaluate CE and PE for the whole sample as well as for four asset quartile
size classes.
C. Scale Economies for Profit and Cost Frontiers in Europe
As mentioned earlier, to evaluate whether a single profit frontier and a

single cost frontier for Europe and the U.S. can be estimated, it is necessary to determine the shape of the individual frontiers. We should
estimate a joint frontier only if estimated European and U.S. frontiers are
similar in shape. Characteristics of cost and (to a lesser extent) profit
functions are often expressed in terms of economies of scale measures.
Output-specific economies of scale are calculated by taking the derivative of the profit (or cost) model with respect to an output. For example,
based on equation (1), scale economies for output y1 can be estimated as:
BPBTkt y; w; z
1
1
a1 a11 y1kt a12 y2 a13 y3 d11 w1kt
By1
2
2
d21 w2kt d31 w3kt f11 zkt :
For the profit function, a value larger (smaller) than 1 indicates increasing
(decreasing) returns to scale, and unity indicates constant returns to scale.
For the cost function, scale estimates are oppositely interpreted. Overall
economies of scale are simply the sum of output-specific economies of scale.
When measured at the frontier, economies of scale are a good indicator of the
shape of the frontier.
15. A bank that lies above the cost frontier has cost efficiency in the range from 1
(efficient) to 1. We invert this measure to get efficiency scores comparable to those of our
profit model.
16. The parameter g represents the share of inefficiency in the overall residual variance and
ranges between 0 and 1. A value of 1 for g suggests the existence of a deterministic frontier,
whereas a value of 0 represents evidence in favor of a standard ordinary least squares estimation. Note that a deterministic frontier is by no means necessarily identical to a DEA model,
given the latters restrictions on the shape of the frontier and the distribution of the inefficiency.
See Coelli et al. (1998) for further discussion. As part of our robustness tests, we vary maximum likelihood convergence criteria (one-by-one) from 0.1 to 104 to see whether our results
are robust. In these estimations, the likelihood value, model coefficients, and efficiency measurements do not change significantly but the number of iterations does vary.
1565
Berger et al. (1993) identified four aspects of measuring economies of

scale that are relevant to our analyses. First and foremost, research has
confirmed that banks have U-shaped cost curves. Economies of scale
increase up to a relatively modest size, often estimated in the range of
$100$500 million, after which they tend to decrease (albeit slowly).
Since the large banks in our samples normally operate well beyond the
minimum average cost point, we need only match one segment of the cost
curve rather than the entire curve.
Second, risk variables are often excluded when measuring economies
of scale. Following Mester (1996a) and Berger and Mester (1997), we
attempt to overcome this problem by including an equity/total assets ratio
that enters scale measures via interaction terms.
Third, many studies base their scale measures on observations that do
not lie on or close to the efficient frontier. As such, economies of scale
(i.e., the marginal effects of outputs on profits or costs) cannot be separated from X-efficiency (i.e., the distance from the efficient frontier). In
this case, economies of scale are biased to the extent that banks do not
lie on or close to the efficient frontier. Unlike DEA, in which actual observations make up the edges of the frontier, the frontier is estimated in
SFA. For this reason, even the most efficient bank is rarely 100% efficient, and in turn, identifying the banks that are most important in shaping
the frontier is not possible. This problem is worsened because the distribution of efficiency measures, when truncated toward the frontier, is generally highly skewed, with the majority of banks lying close to the
frontier.17 For these reasons, choosing a set of efficient banks is an arbitrary decision. Interestingly, we do not need a single cutoff point,
beyond which we have a number of efficient banks; instead, our solution
to this problem derives from the fact that, if for example the cost frontiers
in Europe and the United States have the same shape, we should expect to
see a decrease in the difference in economies of scale as we get closer to
the frontier. By measuring cost and profit efficiencies as we gradually
move closer to the frontier, we decrease the bias from X-efficiency but
simultaneously decrease the number of observations in our independent
sample tests. We therefore start with the fiftieth percentile as a cutoff
point and end with the ninetieth percentile, which still provides sufficient
observations to obtain reliable estimates. At each increment, we test the
significance of the difference in average economies of scale for European
and U.S. banks, both with and without correction for the difference in
variance. If the frontiers have the same shape, this significance should
decrease as attention is narrowed to the most X-efficient banks.
Fourth, the most reliable measure of economies of scale is an overall
estimate, defined as the sum of output-specific economies of scale. The
sum of the partial derivatives of each output is dependent less on changes
17. This is reflected by a positive, significant value for m=su.
1566
Journal of Business
and differences in the output mix. Therefore, we report overall economies

of scale, rather than output-specific economies of scale, to compare the
results for Europe and the United States and evaluate the shape of the
frontiers. We report both results for the whole sample as well as for four
asset quartile size classes.
D. Scope Economies
Unfortunately, calculating scope economies is not as straightforward

as calculating scale economies. However, scope economies are not crucial to evaluating whether European and U.S. banks operate under a single frontier. In addition, Berger and Humphrey (1991) and Berger et al.
(1993) have shown that scope economies are far less significant than
X-efficiencies. Berger, Hanweck, and Humphrey (1987) observed that
for translog functions complementarities cannot exist at all levels of
output. Berger and Humphrey (1997) noted that an additional problem
with scope estimation is the possible existence of zero outputs. Another
potential pitfall is that there often is an extrapolation problem. Given
a sample containing both universal banks and other banks, only the
former banks typically offer the full range of financial services. Consequently, the economies of scope derived from the cost (or profit) function
tend to overestimate the true economies of scope among most sample
banks. A further problem is that the measurement of average economies of scope yields values that are biased due to the inclusion of
X-(in)efficiencies. In the search for a better functional form, some researchers have used a Box-Cox transformation for outputs, while others
have used a composite function with a separate fixed-costs component
of scope economies.
For cost models, Molyneux et al. (1997) proposed a comparison of the
separate cost functions for individual outputs with the joint cost of production. However, the plant-and firm-level data required for this type of
analysis are not available for our sample banks. We cannot claim to solve
all these problems. Instead, we propose a rather simple way of measuring
economies of scope that overcomes some problems and mitigates other
problems.
In our models we have three outputs, or y1 ; y2, and y3, which sum to y.
We start by taking the ratios y1 =ya; y2 =yb, and y3 =yc. If
such a ratio is high, a bank is relatively specialized. For overall scope
economies, we calculate d a2 b2 c 2 . This measure is bounded
between 1/3 (not specialized) and 1 (specialized). We define high [H ] as
referring to the upper twenty-fifth percentile and low [L] for the remainder of the observations.18 Next, for the cost model we calculate the
18. We tested for the robustness of our results by taking other cutoff points. Our results
stay qualitatively the same for a range of approximately 10%.
1567
ratio (TCH TCL)/TCL for y1 ; y2 ; y3, and y. Likewise, for the profit
model, we calculate the ratio (PBTL PBTH)/ PBTH. Total costs and
total profit are divided by total revenues to adjust for the possibility that
banks in high-and low-bank groups may have different size. In both
cases, if scope economies exist, the ratio is greater than 0. Note that
these ratios can be constructed only using averages; as such, the scope
measure itself does not have a standard deviation. This is a common
problem, as recognized in Berger and Humphrey (1997). Instead, we
report a t-value for an independent samples test for TCH TCL (similar
to our X-efficiency tests). Note that, by varying the cutoff point to
more or less than the twenty-fifth percentile, we can check for extrapolation problems. Also, by measuring scope economies for four size
classes as well as for the whole sample, we control for some of the
X-(in)efficiencies that can differ among size classes.
E. Geographic Expansion
As noted by an anonymous referee, a major factor potentially affecting

differences in efficiency among large banks is the geographic distance
between banking offices. Recent papers by Berger and DeYoung (2001,
2002) argued that efficient lead banks in a multibank organization can
export management skills, technology, and operating practices that enhance the efficiency of smaller affiliated banks. On the other hand, geographic distance could diminish lead bank control, cause organizational
diseconomies, strain relationships and related information monitoring,
and lead to agency costs in overseeing junior managers. In general, their
empirical findings reveal that, even though organizational control over
affiliates does decrease as distance from the lead bank increases, the effect is small. Also, affiliates tend to benefit from efficient organizations,
which suggests that technology overcomes barriers to efficient operation
of widely dispersed banking facilities. More benefits were found in terms
of profit than cost efficiency.
Following Berger and DeYoung, we conduct multivariate regression
tests with cost and profit efficiency scores (CE and PE) as dependent
variables and geographic distance (in miles) between banks in banking
organizations as the focal independent variable. We also include the
following as control variables: asset size (rescaled to a 0, 1 range) and
cost and profit levels (i.e., total cost or total profit divided by size defined
as total assets plus off-balance sheet activities). Since distance and size
are highly correlated with one another, we orthogonalized these variables
by regressing distance on size and using the residual as the distance
measure. Data was available for 516 European banks and 410 U.S. banks.
For both Europe and the United States, distance data have been gathered
for 1997, the middle year in the sample period. In this regard, data were
not available on European bank affiliate locations in all 5 sample years. If
1568
Journal of Business
locational data were missing in 1997, data for the closest available year
were available. For U.S. banks, this approach simplified the collection of
geographic distances for thousands of banking offices. As in the Berger
and DeYoung study, geographic distance for Europe is measured by the
sum of distances from the lead bank to affiliated banks (including affiliated
banks outside of Europe) divided by the total assets of the banking organization. For the United States, we compute the sum of distances from
the lead bank to affiliated banks (including foreign bank affiliates) as well
as branch offices divided by total assets. The reason for the latter definition
is that many U.S. banks converted their organizations to one-bank holding
companies in the 1990s by merging bank affiliates into the lead bank as
branch offices.
IV. Data
A. Sample Data
Data on European banks are collected from IBCA reports found in

BankScope. For U.S. banks data are gathered from the Call Reports for
Income and Condition provided by the Federal Reserve System. Data for
both samples are pooled for the period 1995 to 1999. The beginning year,
1995, was chosen due to the advent of interstate banking in the United
States in 1995 as well as a single market for financial services in the
European Union in 1993. Because the Financial Modernization Act became effective in 2000 and fundamentally altered the organizational
structure of U.S. banks to include greater securities and insurance powers, we ended our analyses in 1999. During our sample period, few regulatory changes occurred in European and U.S. banking markets, with
the exception of the introduction of the euro in 1999. It should also be
mentioned that economic conditions were stable in both regions during
this period.19
Since we comparatively examine both cost and profit efficiency, sample banks must be profit maximizing. Also, to test for one cost frontier and
one profit frontier for all banks, it is necessary to include banks with
similar output mixes. In this respect, we exclude brokerage firms with a
banking license due to their large securities activities and relatively small
commercial lending services. Likewise, while general banks and bank
holding companies are included in the analyses, cooperative banks, credit
unions, mutual banks, and credit card banks20 are excluded.
19. An exception is the Asian crisis in 1997. Since we do not estimate a panel data model,
and given that this crisis affected all the countries involved in our study, we do not expect it to
significantly affect the comparative results.
20. In the United States, the Comptroller of the Currency defines credit card banks as
those banks with credit card loans totaling 6.5% or more of total assets. We excluded banks
that exceeded this threshold from the analyses.
1569
To ensure that variables are computed similarly across countries and

over time, all values are expressed in constant 1995 U.S. dollars.21 We
then set a bank asset size cut-off point of $1 billion 1995 U.S. Although
this restriction substantially reduces our sample size, multibillion dollar
banks represent the dominant share of the banking assets in Europe and
the United States due to concentration of resources among larger banks
and are most likely to engage in cross-Atlantic expansion.
As recommended by an anonymous reviewer, we conducted tests
of the possible effect of the euros introduction on January 1, 1999. A
dummy variable equal to 1 in 1999 and 0 in other sample years was
included in the U.S., Europe, and combined cost and profit models. The
dummy variable was mixed in sign and rarely significant across models.
Importantly, the correlation of X-efficiency scores generated from
models with and without the euro dummy exceeded 0.90 in all cases and
was more than 0.96 in five out of the six models. Further analyses on the
correlation of predicted values of the dependent variables (i.e., the total
derivative) for the sample periods 199598 and 199599 yielded results
exceeding 0.97 in all models. Rank tests for both X-efficiency and predicted values with and without the year 1999 did not significantly affect
rank orders. Thus, we infer that the advent of the euro did not have a
significant impact on our results. Only independent banks and bank
holding companies (BHCs) that are not owned or controlled by other
firms are included in our sample. In this regard, our models presume
banks have some degree of freedom to choose their inputs and set output
prices, which would not be entirely true of component subsidiaries within
the organization. Also, and as cited earlier, Berger et al. (2000) have
noted that transfer pricing and other intraorganizational funds flows
could affect efficiency estimates. We therefore use consolidated statements and set limits on the maximum ownership and control by outside
parties. With respect to ownership, U.S. law requires banks that are 25%
or more owned by a single shareholder to be included in a BHC. For U.S.
bank data, we applied this legal criterion to the construction of our sample
banking organizations. In Europe, under the First Banking Coordination
Directive of 1977, similar rules allow us to apply the same criterion to
European banks.22 Due to numerous mergers and takeovers in Europe
21. As a robustness check, we also used German marks as a reference currency with no
change in results. The dollar exchange rate in the sample period was relatively stable for the
17 European countries with banks in our analyses. Consumer price index (CPI) and exchange
rate series were taken from Economist Intelligence Unit (EIU) Country Data. All European
and U.S. data are expressed in 1995 dollars using the U.S. CPI. Note that, since both dependent
and independent variables are expressed in the same currency, our transformation to constant
1995 U.S. dollars has an impact on only the cutoff value of $1 billion 1995 U.S. As shown in
table 1, only a fraction of our sample banks are close to this cutoff point. In this regard,
changing the cutoff point to slightly higher values did not change our results.
22. The First Banking Directive of 1977 and further requirements set forth in 1983 state
that: 25% ownership of a banks shares constitutes official participation; 50% ownership
1570
Journal of Business
and the United States during the sample period, ownership thresholds are
checked for all large banks in each year of the sample period.
We also exclude all observations with variable values missing and
values less than or equal to zero. A negative or zero value for a variable
implies that the respective banks production function is quite different
from other sample banks. Moreover, if such observations were included
in the analyses, an arbitrary transformation would be required because
zero and negative values are disallowed in the translog model. For the
combined data set only 10 observations were dropped for this reason.
Last, outliers are considered by estimating the models with all observations and checking for outliers in the efficiency scores. As found in
other studies, outliers tend to have higher scores than other sample banks.
If outliers in the independent variables are consistent with outliers in
the dependent variable, we omit the observations as long as skewness
is not substantially altered. We then reestimate the models and report
the results with all observations (with outliers omitted) if the coefficients have not changed (have changed) significantly. To avoid heteroscedasticity problems, we use weighted least squares, with weights based
on the log of total assets.23
B. Variables
Consistent with the intermediation approach used by most SFA studies,

bank outputs are defined as follows: loans ( y1), investments ( y2), and offbalance sheet activities ( y3). Loans aggregate commercial and industrial,
real estate, consumer, agriculture, and other outstanding credit. Investments include securities, equity investments, and all other investments
reported on the balance sheet. Off-balance sheet activities are credit items
and other guarantees, loan commitments, derivative securities, and other
loan and securities exposures not reflected on the balance sheet. Of course,
these activities are particularly important among large banks.
Turning to input prices, it is not possible to calculate a traditional
measure of the price of labor due to incomplete data on the number of
employees in BankScope. Even if the number of employees were available, part-time employees are not counted despite their increasing usage
in the banking industry. Also, given the dispersion in the types of jobs
at large banks (from cashier to investment banker), the price of labor
as measured by average labor cost is not likely an accurate proxy for the
marginal cost of labor. In an effort to overcome these difficulties, we express prices as the input-specific cost per unit of total output. Therefore,
requires that proportional consolidation take place; and ownership below 50% is left to the
discretion of member states concerning the consolidation procedure, with most member
states requiring proportional consolidation for ownership between 25% and 50%.
23. In the presence of heteroscedasticity, the least squares estimator is not efficient but
still unbiased, consistent, and asymptotically normally distributed.
1571
we define the price of labor (w1) as total employee expenses divided by

the sum of assets and off-balance sheet activities. As such, difficulties
related to missing employee data, part-time workers, and average labor
costs are mitigated.
Likewise, the price of financial capital (w2) equals total interest expenses divided by the sum of assets and off-balance sheet activities, and
the price of physical capital (w3) equals noninterest operating expenses
divided by the sum of assets and off-balance sheet activities. As justification for this approach, we argue that, to the extent that inputs are substitutes for one another, they individually contribute to the cost per dollar
output in the same way. Because all input prices have the same denominator, we checked for multicollinearity but found it was not important
among input prices. We also performed a robustness check by rerunning the models using observations that had data available for the number of employees but the results were unchanged for the most part.
Finally, as mentioned earlier, the control variable equity/total assets (z)
is included in the model to adjust for differences in equity capital risk
across banks.
Table 1 contains descriptive statistics for input prices, outputs, and
dependent variables. Total costs, profit before tax, outputs, and input
prices as well as total assets are in thousands of constant 1995 dollars.
Our samples consist of 519 European banks and 476 U.S. banks, or 995
banks in total. As shown in table 1, the size distributions of European
and U.S. banks in our samples differ to some extent. Compared to U.S.
banks, the European bank distribution has a larger mean size (i.e., on
average $20 billion larger), is less skewed, and has a lower maximum
size.
V. Bank Efficiency Results
The forthcoming results and discussion focuses on how the X-efficiency

of large European and U.S. banks compares in the period 199599. To
begin, we estimate separate frontiers for Europe and the United States,
respectively. Next, economies of scale and scope are calculated, and
economies of scale are compared as measured at the frontier for Europe
and the United States, respectively. In this respect, the scale estimates are
evaluated for whether they are similar enough to estimate a single profit frontier and cost frontier. Additionally, profit and cost efficiency results are compared for each of the European countries separately and for
Europe as a whole and the United States, too.
Ordinary least squares models were initially run to examine goodness
of fit. In all cases, the adjusted R2 values exceeded 95%, with most models
exceeding 99%. These results provide strong support for our model design. Since our focus is on frontier analyses, to conserve space, we report
only the SFA models and results (i.e., OLS results are available from the
1572
TABLE 1
Descriptive Information
Variable
Mean
Std. Dev.
Skewness
Kurtosis
Minimum
Maximum
A. European Banks (N = 519)
TC (total cost)
PBT (profit before taxes)
y1 (loans)
y2 (investments)
y3 (off-balance sheet)
w1 (labor price)
w2 (financial capital price)
w3 (physical capital price)
z (equity/assets)
4,791,070
831,208
32,366,000
12,396,900
19,342,800
.013
.041
.012
.083
8,124,660
1,443,620
56,840,900
26,578,300
42,105,800
.013
.039
.014
.093
2.27
2.66
2.53
3.31
3.78
3.69
4.89
4.22
3.84
8.11
11.03
9.77
15.58
22.51
21.03
32.44
26.62
18.51
16358.6
655.76
13382.7
116.78
2058.87
.0003
.0013
.0003
.01
49,246,900
9,177,220
343,000,000
191,000,000
370,000,000
.11
.38
.13
.66
Assets (total assets)
64,400,000
119,000,000
2.70
10.62
1,012,440
751,000,000
50.14
59.64
57.58
42.34
100.49
5.34
5.63
5.80
22.04
52.55
18426
3654
42092
1715
24638
8.29E06
4.65E04
5.39E08
.05
1,020,880
98,457,300
34,826,600
953,435,000
282,533,000
3.13E+10
.03
.05
.01
.26
1.60E+09
B. U.S. Banks (N = 476)
2,684,750
854,247
24,701,600
92,70,800
351,000,000
.011
.024
.003
.094
42,986,500
9,692,600
3,124,390
89,628,300
32,095,400
2.50E+09
.004
.007
.001
.027
153,938,000
6.28
6.76
6.69
5.91
9.53
.92
1.47
.07
3.99
6.37
Note.Total costs, profit before taxes, outputs, and total assets are measured in thousands of 1995 dollars.
Journal of Business
TC (total cost)
PBT (profit before taxes)
y1 (loans)
y2 (investments)
y3 (off-balance sheet)
w1 (labor price)
w2 (financial capital price)
w3 (physical capital price)
z (equity/assets)
Assets (total assets)
1573
authors upon request). Also, we should mention that the price of financial
capital (w2) is used to ensure linear homogeneity in input prices.
A. European Banks
Panel A of table 2 reports results for the estimated SFA cost and profit
models for large European banks.24 For both economies of scale/scope
and X-efficiency, we report results for four asset quartiles as well as
for the total sample. Importantly, X-efficiency estimates based on the
cost model for European banks average 0.947, which implies that frontier efficient banks could further reduce operating costs by 5.3% on average. As in other SFA studies, the distribution of efficiency scores is
highly skewed (e.g., the minimum is 0.064). Relative to the cost results,
X-efficiency estimates based on the profit model for European banks on
average are considerably lower at 0.721, which implies potentially large
profit improvements of 27.9% on average are possible for frontier banks.
With a standard deviation of 0.219 (i.e., approximately twice that of the
cost model), the distribution of profit efficiency scores is quite large. Apparently, large banks in Europe had wide disparities in profit X-efficiency
but much smaller disparities in cost X-efficiency during our sample period. These results are generally consistent with previous European studies of bank cost and profit X-efficiency (e.g., see Ruthenberg and Elias
1996; Dietsch et al. 1998; and Vander Vennet 1999). However, some
studies reported profit X-efficiencies ranging between 0.20 and 0.25
for large European banks (e.g., see Griffell-Tatje and Lovell 1996; Resti
1997; and Altunbas et al. 2001). Assuming that profit efficiency is a
function of not only internal production (as in the case of cost efficiency)
but external market forces, higher cost efficiency and lower profit efficiency could be attributable in part to differences in market power among
large banks. Alternatively, Sheldon (1999) argued that these results could
be due to high-cost outputs with service and quality features that are not
in demand by bank customers.
Panel A of table 2 also contains average scale estimates derived from
the cost and profit SFA models for large banks in Europe. As discussed
earlier, scale estimates are derived from the estimated coefficients for the
entire cost or profit equation, rather than a single estimated coefficient.
24. Since estimated coefficients for l (i.e., the ratio of the variance of the truncated
normal inefficiency term to the variance of random noise) are significantly greater than zero
(at the 0.01 level) in both models, we infer support for our frontier approach relative to OLS.
The total variance of the error term (or s) is low. Similar results hold for all our estimations.
Results with respect to individual variables are difficult to interpret due to second-order and
interaction variable effects. For example, in the cost model the estimated coefficient for loans
( y1) was negative and significant, which implies lower operating costs as loans are increased,
all else the same. However, this interpretation does not take into account the nonlinear cost
implications of loans captured in squared loans ( y1 y1) and multiple interactions of loans with
other variables in the model. For these reasons, we do not report estimated coefficients for
the cost and profit models.
1574
TABLE 2
Journal of Business
Comparison of Independent Large Banks in Europe

A. Efficiency
Cost Models
Europe
Quartile
Scale
Scope
X-efficiency
Q1
Q2
Q3
Q4
Total
Q1
Q2
Q3
Q4
Total
Q1
Q2
Q3
Q4
Total
Mean
Profit Models
United States
t
Mean
1.143 4.767
1.055 17.108
1.126 4.699
1.071 17.376
1.153 5.661
1.049 13.572
1.095 7.494
.938 8.376
1.127 5.991
1.042 9.294
.231
.327 .947 3.357
.009
.680 .975 1.038
.166
.986 .989 11.006
.631
.561 .867 15.932
.340
.083 1.024 10.484
.924 5.755
.973 12.281
.924 5.573
.976 21.321
.962 15.571
.978 35.126
.957 12.235
.978 190.13
.947
8.454
.976
18.098
Europe
United States
Mean
Mean
1.208
1.159
1.169
1.100
1.151
.455
.245
.248
.534
.367
.636
.600
.714
.831
3.646
3.497
4.387
5.986
4.564
1.658
1.516
.264
.714
.590
2.169
1.977
3.074
7.353
1.134
1.148
1.098
0.907
1.099
.895
.934
.982
.865
.950
.676
.754
.778
.866
11.186
11.322
8.856
5.237
5.808
.404
1.135
4.808
15.955
1.362
3.711
5.381
6.542
15.232
.721
2.964
.749
4.790
B. Frontier Test
Independent Samples Test
Cost
Profit
Europe
United States
Percentile
NEurope
NUS
Mean
SD
Mean
SD
50th
60th
70th
80th
90th
50th
60th
70th
80th
26.3
12.9
7.2
3.6
8.2
40.7
27.8
28.0
22.3
8.934
8.957
7.986
6.595
5.282
2.246
1.766
1.899
.876
259
207
155
104
51
260
208
155
104
238
191
143
95
47
238
190
143
95
1.156
1.163
1.167
1.176
1.227
1.167
1.164
1.186
1.173
.186
.186
.193
.213
.249
.260
.262
.285
.311
1.032
1.023
1.016
1.007
1.009
1.124
1.126
1.137
1.144
.113
.122
.129
.145
.153
.151
.153
.145
.142
90th
13.6
1.027
52
47
1.201
.383
1.140
.185
.976
.749
.054
.156
C. X-Efficiency Test
Cost
Profit
59.8 5.005
96.3 2.181
519
519
476
476
.947
.721
.112
.243
Note.Asset quartiles (in millions of U.S. dollars) are distributed as follows: Q1 = [1012345,
1982082}, NEurope = 93, NUS = 155; Q2 = [1982083, 4420897], NEurope = 84, NUS = 166; Q3 =
[4420898, 30193469], NEurope = 165, NUS = 84; Q4 = [30193470, 160096490], NEurope = 177, NUS =
71. All independent samples tests use Levenes test (at the 5% level) for the equality of variances. We
report only the absolute t-values for mean differences that are not rejected by this F test. Scope t-values
correspond to an independent samples test that expected cost (or profit) is equal for specialized and
nonspecialized banks.
1575
Consistent with prior studies of large banks (e.g., Allen and Rai 1996 and
Altunbas et al. 2001), cost model estimates suggest decreasing economies
of scale for outputs. An increase of 1 U.S. dollar in total output results in
an increased cost of almost 1.13 U.S. dollars for European banks. Since
overall economies of scale are negative and significant, the results imply
cost diseconomies of scale. Contrary to the cost model findings, overall
scale economies for the profit model are increasing and significant at 1.15
U.S. dollars for large European banks. This result suggests that large
banks can increase profits by expanding their size.25 The scale economies
results were similar across quartile size ranges.
Finally, scope economies are negative but insignificant for the cost
model. European banks that produce a disproportionate amount of a particular output have total costs approximately 34% lower (but statistically
insignificant) than banks that have a more balanced output mix. These
results are in line with the large international bank results of Allen and
Rai (1996), as well as work by Vander Vennet (1999). For the profit model,
scope economies are positive (with total profits approximately 37% higher
among banks with a more balanced output mix) but again insignificant. This result is consistent with the finding of higher revenues and
profitability among universal banks compared to specialized banks by
Vander Vennett (1999). The insignificance of the cost and profit scope
estimates suggests that scope economies are small in general, which agrees
with the consensus in the empirical banking literature (e.g., see Berger
2003).
B. U.S. Banks
Panel A of table 2 further reports the economies of scale/scope and

X-efficiency results from estimating the cost and profit SFA models for
large U.S. banks. For the cost model, the average X-efficiency score is
0.976 (or higher than European banks) with a standard deviation of 0.047
(or lower than European banks).26 When evaluated at their respective cost
frontiers, we infer that large U.S. banks have higher cost efficiency on average than European banks. For the profit model, the average X-efficiency
score is also relatively higher than European banks at 0.749 with a standard
25. The negative relationship between cost and profit scale economies as bank output
expands is consistent with work by Berger and Mester (1997), who found that cost and revenue
inefficiencies can be negatively correlated (i.e., cost inefficiencies do not necessarily imply
profit inefficiencies). On the other hand, our results differ from Sheldon, who reported decreasing returns to scale for both costs and profits.
26. For the U.S. data, we observe very high kurtosis for the dependent variables and
outputs. The variable assets used in weighted least squares (WLS) also has high kurtosis,
potentially aggravating estimation problems arising from heteroscedasticity. This latter condition appears to be the case for our U.S. cost model. We checked for robustness of our results
by using different WLS estimations; in general, this problem does not significantly affect our
efficiency estimates and our coefficients, which are still unbiased and consistent. Also, the
mean efficiency and efficiency rankings were not significantly affected.
1576
Journal of Business
deviation of 0.092.27 This result contrasts with that of Miller and Noulas
(1996), who conducted a DEA analysis oflarge U.S. banks and found profit
X-efficiency of 0.97 with almost half the banks 100% technically efficient.
Not surprising, the U.S. market is more homogeneous than the European
market, as confirmed by the lower standard deviation for both cost and
profit X-efficiency. Also, as implied by the descriptive data in table 1,
whereas average return on assets is higher in the United States, the standard
deviation of the return on assets is lower than in Europe. The average cost
ratio of U.S. banks is lower than for European banks also.
Cost model results in panel A of table 2 show that overall economies
of scale for U.S. banks significantly decrease but are smaller in magnitude than those for the European cost model (i.e., 1.127 and 1.042 for
European and U.S. banks, respectively).28 Average scale economies for
U.S. banks generated from the profit model are positive but somewhat
smaller than those for European banks; for example, a 1 U.S. dollar
increase in total outputs results in an almost 1.10 U.S. dollar increase in
profits (compared to 1.15 U.S. dollars for European banks). As in Europe,
it appears that large U.S. banks expansion tends to boost profits. One
exception is the profit economies of scale result of 0.907 for U.S. banks in
the largest size quartile. For these banks, profit scale diseconomies appear
to exist.
Scope economies for large U.S. banks are negative and significant (at
the 0.01 level) for the cost model, whereas they are positive but insignificant for the profit model. Cost scope diseconomies for U.S. banks are
about three times larger than for European banks. This difference in
results could be due to greater number of specialized banks in the United
States and universal banks in Europe. Previous U.S. studies on large U.S.
banks and scope economies are generally mixed, with relatively small
economies or diseconomies, as mentioned earlier (e.g., see Pulley and
Humphrey 1993 and Mitchell and Onvural 1996 and citations therein).
Our results are likewise mixed, with scope economies in profits but scope
diseconomies in costs for both Europe and the United States. Also, like
most previous work, most scope economies are not significant.
In general, our efficiency results for large European and U.S. banks
reflect more similarities than differences. Estimates of scale economies
for costs and profits reveal functional relationships between output level
and efficiency that are strikingly comparable to one another. However, on
average, European banks have lower cost and profit X-efficiencies than
U.S. banks, and the dispersion of both profit and cost efficiency scores is
considerably smaller for U.S. banks than for European banks. Consistent
27. Comparing the values for m=su , truncation is very similar for the U.S. and Europe
bank samples, with a large proportion of efficient banks.
28. Our finding of decreasing returns to scale is consistent with numerous prior studies of
large U.S. banks (e.g., see Hunter et al. 1990; Noulas et al. 1990; Jagtiani and Khanthavit
1996; and Miller and Noulas 1996).
1577
with previous literature, we do not normally find scope economies, with

the exception of significant cost diseconomies of scope for U.S. banks.
We should mention that our results do not necessarily imply that U.S.
banks are more efficient than European banks, as each sample of banks is
evaluated against its own efficient frontier. It is possible that the most
efficient U.S. banks are only average when compared to the European
banks. To compare efficiency results for European and U.S. banks, we
next turn to a combined analysis of both samples.
C. European and U.S. Banks Combined
Panel B of table 2 reports the results for a series of independent sample

t-tests for mean differences in economies of scale estimates among
European and U.S. banks. As discussed earlier, these tests are based on
bank samples that progressively move closer to the frontier (i.e., from the
fiftieth to the ninetieth percentile). The t-statistics test for the mean differences in average scale estimates for European and U.S. banks that are
not rejected by Levenes F-test for equality of variances. The cost
models t-tests for all percentile cutoff points are significant, which is
evidence against a single cost frontier. For the profit models, the evidence
is mixed, with significant t-tests for the fiftieth, sixtieth, and seventieth
percentile samples but insignificant t-tests for higher percentile samples.
Placing greater weight on the ninetieth percentile banks that are closest to
the profit frontier, due to insignificant differences in mean scale economies, we infer that the evidence favors estimating a single profit frontier
for European and U.S. banks.
We next estimate the cost model for the combined sample with the
addition of dummy variables for 17 countries and a dummy variable for
Europe (versus the United States), respectively. These tests seek to capture geographic differences in cost scale estimates. The left-hand side of
table 3 reports the results. While some of the country-specific dummy
variables are significant, it is noteworthy that the Europe dummy variable
is not significant. Since average X-efficiency is 0.961 for the country
dummy cost model and 0.958 for the Europe dummy cost model, overall
cost efficiency results are not affected by the inclusion of country-specific
dummy variables.29 In both estimations m=su is significant and negative,
as a relatively large number of bank observations can be found in the tail
of the efficiency distribution.
In panel C of table 2, we employ the results from the country dummy
cost model to test whether the European banks jointly are on average as
29. As a further test, we also estimated the European model with country dummies where
Germany is the reference country. Efficiency scores and rankings were not significantly
different using the model without dummies as presented in table 2. Thus, whereas Europes
banking market as a whole may be quite heterogeneous, the estimation of a single European
frontier poses no problems.
1578
TABLE 3
Combined Cost and Profit Frontiers

Profit Models
Cost Models
Country Dummy
Variable
Country Dummy
Europe Dummy
4.767
12.948
2.334
.031
5.934
1.598
11.074
16.822
.048
.225
.068
4.497
.197
.017
.099
.110
.167
.173
.084
.270
.042
.116
.094
.063
.156
.067
2.336
.472
4.878
4.694
5.263
9.552
1.466
9.942
2.179
4.480
3.161
2.827
2.847
2.189
.160
.149
.016
.295
.237
.102
.078
.488
.187
.676
.452
.164
.144
.204
.910
1.428
.320
5.582
3.662
2.097
1.265
9.374
4.298
9.010
4.130
3.715
.903
2.908
Journal of Business
Intercept
Europe
Austria
Belgium
Switzerland
Germany
Denmark
Spain
Finland
France
Great Britain
Greece
Ireland
Italy
Luxembourg
Netherlands
Europe Dummy
.092
.176
.023
.663
.303
.106
.374
.384
.703
.020
.005
2.423
2.936
.249
743.85
.0092
.07573
.961
.090
995
1.718
2.739
.709
8.060
3.895
1.912
10.694
11.739
7.830
1.470
1.251
4.476
22.914
23.422
.582
.270
.110
.561
.243
.765
.014
.004
1.754
2.362
.239
611.66
.00866
.04829
.958
.081
995
7.551
3.712
1.701
17.510
6.554
7.380
.893
.750
3.824
21.074
22.660
.440
.330
0.218
.885
.090
.875
.179
.628
.166
.094
.012
2.139
4.208
.625
90.43
.02087
.36947
.735
.207
995
2.027
2.656
3.287
4.319
.517
7.762
2.304
8.512
.913
2.924
1.127
3.437
13.345
13.881
1.555
.847
.384
.425
.106
1.563
.129
0.027
.500
1.203
0.364
129.17
.05409
.07823
.853
.118
995
8.639
5.085
3.202
6.002
1.354
9.038
3.537
2.117
.841
12.694
24.387
Norway
Portugal
Sweden
w1
w3
y1
y2
y3
z
T
0.5T2
m=su
l
s
LLF
ss(v)
ss(u)
Mean
SD
Observations
Note.Variables are defined as follows: y1 = loans, y2 = investments, y3 = off-balance sheet activities, w1 = price of labor, w2 = price of financial capital, w3 = price of physical
capital, z = equity/total assets ratio, and T = time. Other terms are LLF= the log likelihood ratio test, ss(v) = the variance of random noise, or s2v ; ssu = the variance of the
truncated efficiency term, or s2u ; m=su = the truncation point for u divided by the standard deviation of the truncated efficiency term, l= the ratio of standard deviations of the truncated normal efficiency term and random noise, or su =sv , and s = the total variance of the error term equal to the sum of the variance of random noise plus the variance of the
truncated efficiency term, or s2 s2v s2u. The full estimation results are available from the authors upon request.
1579
1580
TABLE 4
Journal of Business
Independent Country Sample Tests of X-Efficiencies Compared

to the United States
Cost Model
Profit Model
Country
Mean
Mean
Austria
Belgium
Switzerland
Germany
Denmark
Spain
Finland
France
Great Britain
Greece
Ireland
Italy
Luxembourg
Netherlands
Norway
Portugal
Sweden
5
16
46
64
47
52
10
44
63
22
16
54
13
24
16
11
18
.902
.877
.974
.940
.956
.973
.971
.969
.947
.973
.976
.976
.937
.981
.971
.971
.974
777.18
813.72
1874.98
1363.42
5848.73
626.19
674.15
240.28
714.38
215.42
1222.1
.06
1905.46
205.55
205.55
606.82
244.06
1.23
2.79
.23
4.20
5.42
1.60
.76
6.09
4.28
1.32
.66
13.20
4.11
2.67
2.67
.93
.31
.518
.791
.824
.805
.796
.849
.772
.783
.831
.739
.865
.803
.683
.845
.751
.855
.823
2313.37
316.61
26.14
386.44
47.91
10.66
176.69
325.63
60.53
41.74
8.68
194.26
843.00
17.94
327.08
.18
218.79
1.69
1.73
4.94
6.09
9.86
3.54
1.89
5.37
6.83
12.96
.03
7.47
2.96
2.38
4.45
1.22
1.35
Note.The F-statistic is Levenes test (at the 5% level) for the equality of variances. We report the
t-values for mean differences only in average scale economies that are not rejected by the F test.
Efficiency scores are weighted by assets per country in each year to avoid small banks in small countries biasing results. The reference country is the United States, with an average weighted cost efficiency of 0.974 and an average weighted profit efficiency of 0.865.
cost efficient as U.S. banks. Mean efficiency for European banks is 0.947,
which is significantly less than U.S. banks at 0.976, albeit by only 2.9%.
We infer that European banks are slightly less cost efficient than U.S.
banks on average but the difference is not economically meaningful.
The left-hand side of table 4 compares each European countrys bank
cost efficiency to U.S. bank cost efficiency. Given the relatively low number of observations for some countries and to take into account the different market structures, we weighted the efficiency scores for each bank in
a particular country by its share of the countrys total banking assets. For
the United States, the weighted cost efficiency is 0.974. The F-statistic
tests for the equality of variances fail to accept the null hypothesis (except
for Italy); as such, type II t-tests under the assumption of unequal variances are most appropriate. A negative and significant t-statistic means a
countrys banks are less cost efficient than U.S. banks. Excluding some
countries due to relatively low sample sizes (i.e., Austria, Finland,
Luxembourg, and Portugal), the type II t-test results are mixed, with five
(eight) countries banks having cost efficiency similar to (different from)
U.S. banks. Only banks in Italy and the Netherlands had significantly
higher average cost efficiency scores than U.S. banks.30 Banks in Belgium
30. As shown in table 8, these findings are not related to the individual coefficients for the
country dummies, as these dummies capture country-specific effects but not X-efficiency.
1581
stand out as the least efficient among the European countries. Finally,
while banks in Germany, Denmark, France, Great Britain, and Norway
were less efficient than U.S. banks from a statistical standpoint, the magnitude of the difference appears to be fairly modest in economic terms.
Turning to the estimated profit models for the combined sample, recall
that earlier tests moving progressively closer to the profit frontier implied
differences between European and U.S. banks that became insignificant.
The right-hand side of table 3 shows the estimated models with the addition of dummy variables for 17 countries and a Europe dummy (versus the United States), respectively. Profit X-efficiency scores averaged
0.735 for the country dummy model and 0.853 for the Europe dummy
model. Hence, relative to the cost model results, country-specific dummy
variables (which are generally statistically significant) provide incremental information that is averaged out upon lumping together the European
countries. Focusing on the Europe dummy variable t-test, X-efficient
frontier U.S. banks scale economies are not significantly higher than those
for European banks. Also, m=su is positive and significant, implying that
most banks are efficient. This evidence tends to confirm our earlier
finding of a single profit frontier for European and U.S. banks. As mentioned earlier, panel C of table 2 shows that, for U.S. banks, the average profit efficiency score is 0.749 compared to 0.721 for European
banks, which are significantly different at the 5% level.
We next consider the independent sample tests for mean differences in weighted profit efficiency scores between individual European
countries banks and U.S. banks, where the latter have a profit efficiency
weighted by assets of 0.865. Again excluding selected countries due to
small sample sizes (i.e., Austria, Finland, Luxembourg, and Portugal),
the type II t-test results in table 4 clearly indicate that U.S. banks are significantly more profit efficient than banks in 11 countries and that only
banks in Ireland and Sweden had profit efficiency similar to U.S. banks.
In sum, the profit efficiency results for banks in Europe and the
United States are consistent with the cost efficiency results. As a robustness check, we also estimated the cost and profit models using traditional labor prices for U.S. banks as well as using flow outputs (e.g.,
substituting interest earnings on loans for the stock of loans, securities
earnings for the stock of investments, and noninterest income for the
stock of off-balance sheet activities), but the cost and profit efficiency
results remained qualitatively the same. Figure 1 graphically summarizes
the mean cost and profit efficiency scores for large banks by country.
Casual inspection of this graph suggests that cost efficiency and profit
efficiency are correlated with one another to some extent. Viewing countries in the upper-right quadrant of the graph as operating within a relevant range of cost and profit efficiency, there is considerable dispersion
among sample countries cost and profit efficiency. Banks in the United
States, Ireland, and the Netherlands tend to dominate other countries
1582
Journal of Business
Fig. 1.Mean weighted (by total assets) cost and profit efficiency for large
banks.
banks in cost/profit efficiency space. The relative strength of U.S. banks

is consistent with Berger et al. (2000), who found that U.S. banks tended
to be relatively efficient compared to banks in France, Germany, Spain,
and the United Kingdom. Belgian banks have relatively low cost efficiency but their profit efficiency is well within the range of profit
efficiency of other countries banks. Only banks in Luxembourg and
Austria, for which the number of observations is small, appear to be
outside relevant ranges for cost and profit efficiency. While differential
regulatory and economic environments in the past no doubt influenced
the relative efficiency of banks in these and other countries, harmonization of banking services as European countries move toward a single market will likely lead to increasing convergence of cost and profit
efficiency among large banks in the future. In the meantime, differential competitive advantages exist for large banks located in different
European countries; moreover, large banks in the United States are competitive relative to the most efficient banks in Europe.
D. Geographic Dispersion and Bank Efficiency
Here we report the results for empirical tests on the relationship between
geographic dispersion as measured by distance between banks within a
banking organization and efficiency measures. Table 5 gives the descriptive statistics for the six independent variables discussed previously.
U.S. banks have higher mean distances than European banks; nonetheless,
U.S. banks average cost and profit efficiency scores are higher, average cost
ratios are lower, and average profit ratios are higher than European banks.
TABLE 5
1583
Descriptive Statistics: Geographic Distance Analyses

Mean
SD
Minimum
Maximum
A. European Banks (516 obseravations)
Distance
Size
CE
PE
Cost ratio
14,356
84,235,200
.9490
.7212
.0652
30,041
151,781,000
.1051
.2438
.0558
0
1,071,080
.1059
.0153
.0051
165,215
956,470,000
.9953
.9895
.5028
Profit ratio
.0136
.0158
.0001
.1322
B. U.S. Banks (410 observations)
Distance
Size
CE
PE
Cost ratio
Profit ratio
53,662
454,657,000
.9744
.7660
.0448
.0140
226,378
2,799,080,000
.0580
.1459
.0134
.0066
0
1,045,520
.2009
.0583
.0009
.0003
1,454,040
31,869,400,000
.9902
.9874
.1188
.0840
Note.Distance is the sum of geographic distances in miles between banks within the banking
organization. Size is total assets and off-balance sheet items in thousands of dollars. CE and PE are cost
and profit efficiency scores, respectively. Total costs and total profits divided by total assets and offbalance sheet items are the cost and profit ratios.
Table 6 shows the correlation coefficients between these variables. As may

be expected, a higher cost efficiency is negatively and significantly correlated with the cost ratio. Likewise, a higher profit efficiency is positively and
significantly correlated with the profit ratio. Interestingly, the cost and profit
ratios are positively and significantly correlated, with a value of 0.645.
Distance was not significantly correlated with any of the other variables. As
defined previously, distance is orthogonalized with respect to size.
Table 7 reports the results for the multivariate regression models using cost efficiency and profit efficiency as dependent variables. We ran
three models for each efficiency measure due to the inclusion of the
following three dummy variables: (1) outside country (i.e., 1 for banks
with affiliates outside their home country in Europe or the United States
TABLE 6
Correlations between Independent Variables

Distance
Distance
Size
CE
PE
Cost ratio
Profit ratio
1.000
.000
.018
.026
.001
.055
Size
CE
PE
Cost Ratio
Profit Ratio
.000
1.000
.013
.076*
.121**
.108**
.018
.013
1.000
.059
.262**
.245**
.026
.076*
.059
1.000
.114**
.236**
.001
.121**
.262**
.114**
1.000
.645**
0.055
.108**
.245**
.236**
.645**
1.000
Note.Distance is the sum of geographic distances in miles between banks within the banking
organization. Size is total assets and off-balance sheet items in thousands of dollars. CE and PE are cost
and profit efficiency scores, respectively. Total costs and total profits divided by total assets and offbalance sheet items are the cost and profit ratios. The symbols * and ** denote 5% and 1% significance
(two-tailed), respectively. Distance is orthogonalized with respect to size (see also table 7).
1584
TABLE 7
Geographic Expansion and Bank Efficiency: Estimated Regression Coefficients (and t Values)
Profit Efficiency Dependent Variable
Cost Efficiency Dependent Variable

Model (1)
b
Intercept
.102
Distance
.826E08
Size
.568E03
Total Cost Ratio 0.362E01
Total Profit Ratio .338
Outside Country .178
Outside EU
Outside Europe
F value
Adjusted R2
Model (2)
t
22.80 2.92
0.66 .217E06
2.47 0.147E01
1.97 1.172
4.75 9.725
6.33
.452
19.33
.090
Model (3)
t
Model (1)
t
Model (2)
t
Model (3)
t
30.13
7.919
25.11 0.388
31.94
0.389
33.17
0.389
33.18
.57
.448E06 .41 .157E06 2.29 .150E06 2.19 .154E06 2.23
2.19
.311E01
0.47
.185E01 2.43
0.161E01 2.27
.165E01
2.27
2.15 3.239
2.06 1.883
11.72 1.960
12.25 1.945
11.85
4.75 26.45
4.43 9.360
13.21
9.603
13.59
9.446
12.38
.030
2.54
6.49
.048
4.00
1.231
5.61
.048
3.87
18.61
18.17
43.27
45.35
45.57
.087
.085
.186
.193
.194
Journal of Business
Note.We estimate a truncated regression model, with truncation at 1 from above. Mean cost efficiency is 0.960, and mean profit efficiency is 0.741. Distance is the sum of
geographic distances in miles between banks within the banking organization. It is computed orthogonal to size (i.e., the residual of a regression of distance on size and a constant)
to avoid multicollinearity problems. Size is total assets plus off-balance sheet items. CE and PE are cost and profit efficiency scores, respectively. Total costs and total profits
divided by total assets plus off-balance sheet items are the cost and profit ratios. Three dummy variables are defined as follows: (1) outside country (i.e., 1 for banks with affiliates
outside their home country in Europe or the United States, 0 otherwise), (2) outside EU (i.e., 1 for European banks with affiliates outside the European Union or U.S. banks with
offices outside the United States, 0 otherwise), and (3) outside Europe (i.e., 1 for European banks with affiliates outside of Western and Eastern Europe or U.S. banks with offices
outside the United States, 0 otherwise). We use a White estimator to correct for possible heteroscedasticity problems.
1585
0 otherwise), (2) outside EU (i.e., 1 for European banks with affiliates

outside the European Union or U.S. banks with offices outside the United
States, 0 otherwise), and (3) outside Europe (i.e., 1 for European banks
with affiliates outside of Western and Eastern Europe or U.S. banks with
offices outside the United States, 0 otherwise). The estimated distance
coefficients have negative signs for all cost and profit efficiency models,
with significant coefficients (at the 0.01 level) for all profit models.
Hence, as banks become more geographically diversified, profit efficiency tends to significantly diminish. For example, an increase of total
distance within the banking organization of 1,000 miles would decrease
profit efficiency by about 0.016%, and vice versa for a decrease in distance. The results for the dummy variables reflecting foreign versus domestic banking operations are negative (positive) and significant in the
cost (profit) efficiency models. These results suggest that cost (profit)
efficiency diminishes (increases) as banks move beyond their domestic
geographic area. The positive effect on profit efficiency may reflect the
net positive diversification benefits banks reap from their geographical
expansion. Regarding the control variables, all three control variables
(i.e., profit ratios, cost ratios, and asset size) are significant in both the cost
and profit models, except for cost model 3, in which asset size is insignificant. The latter finding as well as the relatively low spread in cost
efficiency explain the higher OLS adjusted R2 values in the profit models
compared to the cost models.
In sum, based on the distance variable results, we infer that geographic
dispersion tends to negatively affect banks profits. The dummy variable results further reveal that expansion outside of a banks domestic
area can have negative effects on cost efficiency but positive effects on
profit efficiency. These findings are generally consistent with Berger and
DeYoung, who found more profit than cost efficiency benefits of geographic expansion.
E. Small-Bank Results
In this section, we broaden our analyses to include small banks. International mergers and acquisitions are most likely to occur between large
institutions and smaller banks, as the former seek a foothold position
in new markets. Historically, small, independent banks are much more
common in the United States than in Europe. Our European sample includes all small banks not affiliated with a bank holding company or other
banking organization in the 199599 period for which data were available.
A total of 383 observations met these criteria.31 For the United States,
we randomly sampled 500 small, independent banks. Compared to our
large bank samples, the size distribution is much less dispersed.
31. Most small, independent European banks are located are in Germany, Switzerland,
and Denmark. No observations were obtained for Finland, Greece, Ireland, and Sweden.
1586
Journal of Business
Repeating the empirical methodology applied to large banks, table 8

contains the results for small, independent commercial banks with
less than $1 billion in constant 1995 dollars.32 The estimated cost and
profit frontier X-efficiency results for small banks in Europe, the
United States, and combined reveal relatively higher cost efficiency
but lower profit efficiency compared to large banks, especially among
European banks. Cost X-efficiency is on average 0.743 and 0.871 for
small European and U.S. banks, respectively. Cost economies of scale
are negative, especially for small U.S. banks. While our sample tests
show significant differences between the European and U.S. cost frontiers, these differences become less significant as we approach the
frontier. Cost scope economies are also more negative in the United
States, with mixed significance in Europe and the United States. Consistent with our large bank results, cost-efficient small banks in Europe
and the United States are comparable. However, based on joint frontier results, there are more inefficient banks in Europe than in the
United States, as reflected by the difference between the average cost
X-efficiency of U.S. (0.871) and European (0.743) banks.
Regarding the small-bank profit efficiency results, average profit
X-efficiency is 0.607 in Europe and 0.644 in the United States As in
the large-bank analyses, European small banks have lower profit
X-efficiencies than U.S. small banks. However, for all percentile sample
cutoff points, we find significant differences between the two frontiers.
For the separate profit frontier results, small U.S. banks gain more than
small European banks on a dollar of additional output (i.e., the estimated
coefficients are 1.31 and 0.49, respectively). This higher profit efficiency
among small U.S. banks may well be due to the fact that independent,
small U.S. banks tend to operate in rural areas, in which there is less
competition than experienced by their European counterparts. When
these two groups of banks are compared using a joint frontier, we find that
small U.S. banks are also significantly more profit efficient than small
European banks, albeit by a small absolute margin (i.e., 0.644 for U.S.
banks versus 0.607 for European banks). Scope economies are also
higher for U.S. banks but are not significant.
Figure 2 shows the mean cost and profit efficiencies of each countrys
small banks, which are weighted by assets as before. Similar to the largebank results, there appears to be a positive correlation between cost and
profit efficiency scores. Also, like the results for large banks, small banks
in the United States and the Netherlands dominate other banks in the cost/
profit efficiency space. Luxemburg was relatively inefficient among large
banks but is highly efficient in small bank comparisons. Small banks
in Spain and France are relatively cost/profit inefficient. Unlike the
32. Complete details of the empirical results are available upon request from the authors.
TABLE 8
1587
Comparison of Independent, Small Banks in Europe

A. Efficiency
Cost Models
Europe
Scale
Scope
X-efficiency
Profit Models
United States
Quartile
Mean
Mean
Q1
Q2
Q3
Q4
Total
Q1
Q2
Q3
Q4
Total
Q1
Q2
Q3
Q4
.895
.996
1.026
1.111
1.059
.113
.739
.699
1.501
.816
.814
.817
.710
.740
3.814
4.243
5.096
6.850
5.068
1.286
2.383
3.297
.872
3.037
6.292
4.165
3.273
3.416
1.149
1.142
1.144
1.142
1.145
.437
.573
.596
.578
.599
.877
.854
.882
.892
Total
.743
3.495
.871
Europe
Mean
United States
t
6.387
.359 .836
6.343
.252 .589
9.927
.413 .976
6.938
.601 1.324
8.007
.490 1.079
3.312
.091 .519
.585 1.608 .102
.297 .429 1.714
.370 1.201 .299
2.369 0.507 1.777
9.699
.527 2.448
7.361
.564 2.561
9.514
.586 2.753
8.351
.638 3.554
8.519
.607 3.042
Mean
1.324 6.337
1.295 6.198
1.308 11.077
1.339 8.586
1.311 7.701
.485
.115
.550 1.097
.554 1.016
.663
.884
.608
.805
.615 3.933
.650 4.278
.692 6.362
.625 3.156
.644
4.249
B. Frontier Test
Independent Samples Test
Cost
Profit
50th
60th
70th
80th
90th
50th
60th
70th
80th
25.6
12.3
6.4
10.6
6.2
163.8
180.6
162.4
122.4
4.561
3.274
2.772
3.763
2.526
26.459
22.026
18.441
14.940
191
153
115
76
38
191
153
115
76
260
208
156
104
52
259
207
156
104
57.7 12.250
38
51
90th
NEurope
NUS
Percentile
Europe
Mean
United States
SD
Mean
SD
1.061
1.073
1.071
1.043
1.034
.496
.497
.487
.436
.211
.199
.206
.213
.250
.440
.441
.465
.490
1.136
1.136
1.136
1.145
1.147
1.304
1.312
1.319
1.308
.141
.152
.170
.119
.132
.189
.144
.154
.161
.344
.469
1.326
.180
.743
.607
.213
.200
C. X-Efficiency Test
Cost
Profit
261.3 11.950 383

47.2 2.999 383
519
519
.871 .102
.644 .151
Note.Asset quartiles (in millions of U.S. dollars) are distributed as follows: Q1 = [4069, 45720],
NEurope = 24, NUS = 201; Q2 = [45721, 102338], NEurope = 39, NUS = 187; Q3 = [102339, 264629],
NEurope = 122, NUS = 103; Q4 = [264630, 990942], NEurope = 198, NUS = 28. All independent samples
tests use Levenes test (at the 5% level) for the equality of variances. We report the absolute t-values
only for mean differences that are not rejected by this F test. Scope t-values correspond to an independent samples test that expects cost (or profits) is equal for specialized and nonspecialized banks.
1588
Journal of Business
Fig. 2.Mean weighted (by total assets) cost and profit efficiency for small
banks.
large-bank results, there is much higher dispersion of countries cost/

profit efficiency scores for small banks.
In general, the small-bank efficiency results are similar in some ways
to those for large banks in Europe and the United States but different
in other respects. Small U.S. banks tend to have greater cost and profit
X-efficiencies than small European banks, which is similar to our findings for large banks. One difference in results is the normally lower average
cost and profit X-efficiency scores for small banks relative to large banks.
We infer that small banks tend to be less efficient than large banks. One possible implication of these findings is that small banks face less competitive
pressure to be cost and profit efficient than large banks. If this is indeed true,
an efficiency motive for large banks to merge or acquire small banks exists.
VI. Conclusion
The last two decades of the twentieth century witnessed rapid bank consolidation around the world, which has taken place for the most part within
countries and on an intracontinental basis. At some point, when regional
concentration of bank resources becomes satiated, it is reasonable to believe
that intercontinental mergers will be the next phase in the merger movement. In this paper, we examine the question of economic motivations
for further geographic expansion by conducting stochastic frontier cost
and profit analyses to estimate economies of scale as well as X-efficiency
for multibillion dollar European and U.S. banks. Comparable data for input prices and outputs are gathered for large banks for the period 199599.
1589
Our empirical results with respect to separate analyses for large

European and U.S. banks cost and profit curves are strikingly similar.
For the cost model in both regions, we found small, increasing, and significant returns to scale. Returns to scale for the profit models are positive
and significant for the most part. Generally speaking, these results suggest that large banks in Europe and the United States can increase profits
via output expansion. Based on these cost and profit models, X-efficiency
scores revealed that, on average, European banks have lower cost and
profit efficiencies than U.S. banks. In absolute terms, U.S. banks have
lower (higher) cost (profit) ratios. Also, the dispersion of both profit
and cost efficiency scores is considerably smaller for U.S. banks than
European banks. This result is not surprising in view of the greater
homogeneity of banking markets in the United States than Europe. Scope
estimates were generally not significant, with the exception of cost diseconomies of scope among U.S. banks.
Further analyses evaluated the reasonableness of estimating a combined
cost or profit frontier for European and U.S. banks. For this purpose we
measured economies of scale moving progressively closer to the cost or
profit frontier, as banks closest to the frontier are most important in terms
of assessing the feasibility of a single cost or profit frontier. We found that
mean scale economies for European and U.S. banks measured very close
to the profit frontier are not significantly different and, therefore, imply
a single profit frontier. Profit X-efficiency scores revealed lower average efficiency among different European countries banks compared to
U.S. banks. By contrast, the results for cost scale economies comparisons
failed to accept the null hypothesis of a single cost frontier. Results of
X-efficiency comparisons revealed that U.S. banks cost efficiency tended to be higher than that of European banks.
We conclude that our empirical results tend to support the notion
that potential profit efficiency gains are possible in cross-Atlantic bank
mergers between European and U.S. banks. Thus, an economic motivation appears to exist in favor of geographic expansion in the years ahead.
Likewise, in view of considerable heterogeneity in cost and profit efficiency, our evidence suggests that intra-European mergers among large
banks are possible.
Direct tests for how geographic distance affects cost and profit efficiencies of large banks in the United States and Europe indicated that efficiency tends to decline as banking organizations spread out their offices.
However, profit efficiency tends to increase with respect to expansion beyond their home country. Since further analyses showed that small banks
tend to have lower average cost and profit X-efficiency scores than large
banks, large banks may well be expected to expand via small bank mergers.
Todays global banks developed gradually over time by expanding
abroad to meet their domestic business clients international financial
needs. As such, branch office acquisitions and mergers and de novo entry
1590
Journal of Business
have traditionally been utilized to achieve a foothold in foreign locations.

While other motivations may well exist for cross-Atlantic mergers among
large banks, potential profit efficiency gains appear to be a plausible
factor in decisions to merge beyond continental shores and form global
organizations composed of large European/U.S. bank combinations.
References
Allen, L., and A. Rai. 1996. Operational efficiency in banking: An international comparison.
Journal of Banking and Finance 20:65572.
Altunbas, Y., and S. P. Chakravarty. 1998. Efficiency measures and the banking structure in
Europe. Economics Letters 60:2058.
Altunbas, Y., E. P. M. Gardener, P. Molyneux, and B. Moore. 2001. Efficieny in European
banking. European Economic Review 45:193155.
Altunbas, Y., J. Goddard, and P. Molyneux. 1999 Technical change in banking. Economics
Letters 64:21521.
Altunbas, Y., and P. Molyneux. 1996. Economies of scale and scope in European banking.
Applied Financial Economics 6:36775.
Battese, G. E., A. Heshmati, and L. Hjalmarsson. 1998. Efficiency of labour use in the
Swedish banking industry: A stochastic frontier approach. Working paper, Goteborg
University, Goteborg, Sweden.
Bauer, P. W., A. N. Berger, G. D. Ferrier, and D. B. Humphrey. 1997. Consistency conditions
for regulatory analysis of financial institutions: A comparison of frontier efficiency methods. Working paper 0297, Federal Reserve Financial Services, Washington, DC.
Beattie, B. R., and C. R. Taylor. 1985. The economics of production. New York: John Wiley.
Berg, S., F. Frsund, L. Hjalmarsson, and M. Suominen. 1993. Banking efficiency in the
Nordic countries. Journal of Banking and Finance 17:37188.
Berg, S., F. Frsund, and E. Jansen. 1992. Malmquist indices of productivity growth during
the deregulation of Norwegian banking 198089. Scandinavian Journal of Economics 94:
21128.
Berger, A. N. 2003. The efficiency effects of a single market for financial services in
Europe. European Journal of Operational Research 150:46681.
Berger, A. N., R. S. Demsetz, and P. E. Strahan. 1999. The consolidation of the financial
services industry: Causes, consequences, and implications for the future. Journal of
Banking and Finance 23:13594.
Berger, A. N., and R. DeYoung. 1997. Problem loans and cost efficiency in commercial
banks. Journal of Banking and Finance 21:84970.
. 2001. The effects of geographic expansion on bank efficiency. Journal of Financial
Services Research 19:16384.
. 2002. Technological progress and the geographic expansion of the banking industry. Working paper 200207, Federal Reserve Bank of Chicago.
Berger, A. N., R. DeYoung, H. Genay, and G. F. Udell. 2000. Globalization of financial
institutions: Evidence from cross-border banking performance. In Brookings-Wharton
papers on financial services, 3, ed. R. E. Litan and A. Santomero. Washington, DC:
Brookings Institution.
Berger, A. N., G. A. Hanweck, and D. B. Humphrey. 1987. Competitive viability in banking:
Scale, scope and product mix economies. Journal of Monetary Economics 20: 50120.
Berger, A. N., and D. B. Humphrey. 1991. The dominance of inefficiencies over scale and
product mix economies in banking. Journal of Monetary Economics 28:11748.
. 1992. Megamergers in banking and the use of cost efficiency as an antitrust defense.
Antitrust Bulletin 37:541600.
Berger, A. N. and Humphrey, D. B. 1997. Efficiency of financial institutions: International
survey and directions for future research. European Journal of Operational Research
98:175212.
Berger, A. N., W. C. Hunter, and S. G. Timme. 1993. Efficiency of financial institutions: A
review and preview of research past, present and future. Journal of Banking and Finance
17:22149.
1591
Berger, A. N., and L. J. Mester. 1997. Inside the black box: What explains differences in the
efficiencies of financial institutions. Journal of Banking and Finance 21:895947.
Berger, A. N., and P. E. Strahan. 1998. The consolidation of the financial services industry:
Causes, consequences, and the implications for the future. Working paper, Federal Reserve Bank of New York.
Bikker, J. A. 1999. Efficiency in the European banking industry: An exploratory analysis to
rank countries. Working paper. Section Banking and Supervision Strategies, Directorate
Supervision, De Nederlandsche Bank, Amsterdam.
Bos, J. W. B. 2002. European banking: Market power and efficiency. PhD thesis,
Maastricht University, Maastricht.
Coelli, T., D. S. Prasado Rao, and G. E. Battese. 1998. An introduction to efficiency and
productivity analysis. Boston: Kluwer Academic Publishers.
DeYoung, R. and Hassan, I. 1998. The performance of de novo commercial banks: A profit
efficiency approach. Journal of Banking and Finance 22:56587.
Dietsch, M., G. Ferrier, and L. Weill. 1998. Integration and banking performance in the
European Union: Productivity, technical efficiency, scale efficiency, cost efficiency, and
profit efficiency.Mimeo, Universite Robert Schuman, Strasbourg.
Economic Research Europe Ltd. 1997. Single market integration and X-inefficiency. In The
single market review: Credit institutions and banking, ed. European Commission. Luxembourg: Office for Official Publications of the European Communities.
Eisenbeis, R. A., G. D. Ferrier, and S. H. Kwan. 1999. The informativeness of stochastic
frontier and programming frontier efficiency scores: Cost efficiency and other measures of
bank holding company performance. Working paper no. 9923, Federal Reserve Bank of
Atlanta.
Elyasiani, E., and S. M. Mehdian. 1990. A nonparametric approach to measurement of
efficiency and technological change: The case of large U.S. commercial banks. Journal of
Financial Services Research 4:15768.
Evanoff, D. D., and P. R. Israilevich. 1990. Cost economies and allocative efficiency of
large U.S. commercial banks. Proceedings of a Conference on Bank Structure and
Competition, Federal Reserve Bank of Chicago:15269.
Fecher, F., and P. Pestier. 1993. Efficiency and competition in OECD financial services. In
The measurement of productive efficiency: Techniques and applications, ed. H. O. Fried,
C. A. K. Lovell, and S. S. Schmidt, 37485. New York: Oxford University Press.
Griffell-Tatje, E., and C. A. K. Lovell. 1996. Deregulation and productivity decline: The
case of Spanish savings banks. European Economic Review 40:12811303.
Hassan, I., A. Lozano-Vivas, and J. T. Pastor. 2000. Cross-border performance in European
banking, Bank of Finland Discussion Papers, 24/2000.
Humphrey, D. B., and L. B. Pulley. 1997. Banks response to deregulation: Profits, technology and efficiency. Journal of Money, Credit and Banking 29:7393.
Hunter, W. C. 1995. Internal organization and economic performance: The case of large
U.S. commercial banks. Economic Perspectives. Federal Reserve Bank of Chicago
19:1020.
Hunter, W. C., and S. G. Timme. 1986. Technical change, organizational form, and the
structure of bank production. Journal of Money, Credit and Banking 18:15266.
Hunter, W. C., S. G. Timme, and W. K. Yang. 1990. An examination of cost subadditivity
and multiproduct production in large U.S. banks. Journal of Money, Credit and Banking
22:50425.
Jagtiani, J., and A. Khanthavit. 1996. Scale and scope economies at large banks: Including
off-balance sheet products and regulatory effects (19841991). Journal of Banking and
Finance 20:127187.
Jagtiani, J., A. Nathan, and G. Sick. 1995. Scale economies and cost complementarities in
commercial banks: On-and off-balance-sheet activities. Journal of Banking and Finance
19:1175189.
Kolari, J. W., and A. Zardkoohi. 1987. Bank costs, structure and performance. Lexington,
MA: D.C. Heath and Company.
Lang, G., and P. Welzel. 1997. Mergers among German cooperative banks: A panel-based
stochastic frontier analysis. Working paper, University of Augsburg.
Lovell, C. A. K. 1993. Production frontiers and productive efficiency. In Fried, H. O.; Lovell,
C. A. K.; and Schmidt, S. S. (eds.), The measurement of productive efficiency: Techniques
1592
Journal of Business
and applications, ed. H. O. Fried, C. A. K. Lovell, and S. S. Schmidt, 367. New York:
Oxford University Press.
McAllister, P. H., and D. McManns. 1993. Resolving the scale efficiency puzzle in banking.
Journal of Banking & Finance 17:389405.
Mendes, V., and J. Rebelo. 1997. Productive efficiency in Portuguese banking: Agricultural
banks vs. universal banks. Mimeo, University of Porto.
Mester, L. J. 1996a. A study of bank efficiency taking into account risk-preferences. Journal
of Banking and Finance 20:102545.
. 1996b. Measuring efficiency at U.S. banks: Accounting for heterogeneity is important. Working paper no. 9611/R, Federal Reserve Bank of Philadelphia.
Milbourn, T. T., A. W. A. Boot, and A. V. Thakor. 1999. Megamergers and expanded scope:
Theories of bank size and activity diversity. Journal of Banking and Finance 23:195214.
Miller, S. M., and A. G. Noulas. 1996. The technical efficiency of large bank production.
Journal of Banking and Finance 20:495509.
Mitchell, K., and N. M. Onvural. 1996. Economies of scale and scope at large commercial
banks: Evidence form the Fourier flexible functional form. Journal of Money, Credit and
Banking 28:17899.
Molyneux, P., Y. Altunbas, and E. Gardener. 1997. Efficiency in European banking. West
Sussex, UK: John Wiley and Sons.
Noulas, A. G., S. C. Ray, and S. M. Miller. 1990. Returns to scale and input substitution for
large banks. Journal of Money, Credit and Banking 22:94108.
Pastor, J. M., F. Perez, and J. Quesada. 1997. Efficiency analysis in banking firms: An
international comparison. European Journal of Operational Research 98:395407.
Peristiani, S. 1997. Do mergers improve the X-efficiency and scale efficiency of U.S.
banks? Evidence from the 1980s. Journal of Money, Credit and Banking 29:32637.
Pinho, P. S. 1994. Economies of scale and scope and productive efficiency in Portuguese banking: A stochastic frontier approach. Ph.D. thesis, City University Business School, London.
Pulley, L. B. and D. B. Humphrey. 1993. The role of fixed costs and cost complementarities in determining scope economies and the cost of narrow banking proposals.
Journal of Business 66:43762.
Resti, A. 1997. Evaluating the cost-efficiency of the Italian banking system: What can be
learned from the joint application of parametric and non-parametric techniques. Journal
of Banking and Finance 21:22150.
Rhoades, S. A. 1994. A summary of merger performance studies in banking, 198093, and
assessment of the operating performance and event study methodologies. System
staff study no. 167, Board of Governors of the Federal Reserve System.
. 1998. The efficiency effects of bank mergers: An overview of case studies of nine
mergers. Journal of Banking and Finance 22:27391.
Rogers, K. 1998. Product mix, bank powers, and complementarities at U.S. commercial
banks. Journal of Economics and Business 50:20518.
Ruthenberg, D. and R. Elias. 1996. Cost efficiencies and interest margins in a unified
European banking market. Journal of Economics and Business 48:23149.
Saunders, A., and I. Walter. 1994. Universal banking in the United States: What could we
gain? What could we lose? New York: Oxford University Press.
Shaffer, S. 1984. Scale economies in multiproduct firms. Bulletin of Economic Research
36:5158.
Shaffer, S., and E. David. 1986. Economies of superscale and interstate expansion. Research
paper no. 8612, Federal Reserve Bank of New York.
Sheldon, G. 1999. Costs, competitiveness and the changing structure of European banking.
Working paper, Fondation Banque de France pour la Recherche.
Siems, T. F. 1996. Bank mergers and shareholder wealth: Evidence from 1995s megamerger
deals. Financial Industry Studies, Federal Reserve Bank of Dallas (August).
Swank, J. 1996. How stable is the multiproduct translog cost function? Evidence from the
Dutch banking industry. Kredit und Kapital 29:15372.
Vander Vennet, R. 1994. Concentration, efficiency and entry barriers as determinants for EC
bank profitability. Journal of International Financial Markets, Institutions, and Money
4:2146.
. 1999. Cost and profit dynamics in financial conglomerates and universal banking in
Europe. Mimeo, University of Ghent.

Jurnal 1

Загружено:

Сведения о документе

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

Jurnal 1

Загружено:

Авторское право:

Доступные форматы

Large Bank Efficiency in Europe and the United States: Are There Economic Motivations

for Geographic Expansion in Financial Services?

Large Bank Efficiency in Europe

Deregulation of the banking industry in Europe and

This paper employs

(Journal of Business, 2005, vol. 78, no. 4)

the next expansion wave can be expected to involve an international

Large Bank Efficiency

A large body of literature spanning a half-century exists on banking

McAllister and McManus 1993; Mester 1996a, b; Berger and DeYoung

Large Bank Efficiency

estimated a global cost function for 194 banks in 15 countries (including

and the United States. Comparative analyses revealed that domestic

Large Bank Efficiency

others as separate entities. Consequently, these studies suffer from double

differences in efficiency among banks from different countries as well as

Large Bank Efficiency

three-input, three-output translog setting, u and v are the inefficiency and

eij wikt zkt

The error term vkt is normally distributed, independently and identically

bij bji for all i; j;

bij 0 for all i;

bij 0 for all j;

As mentioned earlier, to evaluate whether a single profit frontier and a

Large Bank Efficiency

Berger et al. (1993) identified four aspects of measuring economies of

and differences in the output mix. Therefore, we report overall economies

Unfortunately, calculating scope economies is not as straightforward

Large Bank Efficiency

As noted by an anonymous referee, a major factor potentially affecting

Data on European banks are collected from IBCA reports found in

Large Bank Efficiency

To ensure that variables are computed similarly across countries and

Consistent with the intermediation approach used by most SFA studies,

Large Bank Efficiency

we define the price of labor (w1) as total employee expenses divided by

The forthcoming results and discussion focuses on how the X-efficiency

A. European Banks (N = 519)

Assets (total assets)

B. U.S. Banks (N = 476)

Large Bank Efficiency

Comparison of Independent Large Banks in Europe

Large Bank Efficiency

Panel A of table 2 further reports the economies of scale/scope and

Large Bank Efficiency

with previous literature, we do not normally find scope economies, with

Panel B of table 2 reports the results for a series of independent sample

Combined Cost and Profit Frontiers

Large Bank Efficiency

Independent Country Sample Tests of X-Efficiencies Compared

Large Bank Efficiency

banks in cost/profit efficiency space. The relative strength of U.S. banks

Large Bank Efficiency

Descriptive Statistics: Geographic Distance Analyses

A. European Banks (516 obseravations)

B. U.S. Banks (410 observations)

Table 6 shows the correlation coefficients between these variables. As may

Correlations between Independent Variables

Cost Efficiency Dependent Variable

Large Bank Efficiency

0 otherwise), (2) outside EU (i.e., 1 for European banks with affiliates

Repeating the empirical methodology applied to large banks, table 8

Large Bank Efficiency

Comparison of Independent, Small Banks in Europe