2002 JBF Isik - Technical, Scale and Allocative Efficiencies of Turkish Banking Industry

Journal of Banking & Finance 26 (2002) 719–766
www.elsevier.com/locate/econbase
Technical, scale and allocative efficiencies of

Turkish banking industry
a,b,1 c,*
Ihsan Isik , M. Kabir Hassan
a
School of Management, New Jersey Institute of Technology, Newark, NJ 07102-1982, USA
b
Finans Bolumu, Isletme Fakultesi, Gebze Yuksek Teknoloji Enstitusu, 41400 Kocaeli, Turkey
c
Department of Economics and Finance, University of New Orleans, New Orleans, LA 70148, USA
Received 6 April 1999; accepted 20 December 2000
Abstract
This paper investigates input and output efficiency in the Turkish banking industry to
understand the impact of size, international variables, ownership, control and gover-
nance on profit, cost, allocative, technical, pure technical and scale efficiency measures.
Employing a non-parametric approach along with a parametric approach, we estimate
the efficiency of Turkish banks over the 1988–1996 period. This period allows us to
account for the changes in the macro economy and regulatory treatment of the Turkish
banking industry over time. Our results suggest that the heterogeneous characteristics of
banks have significant impact on their efficiency. Moreover, cost and profit efficiencies
of the Turkish banks have exacerbated over time. Results also indicate that the domi-
nant source of inefficiency in Turkish banking is due to technical inefficiency rather than
allocative inefficiency, which is mainly attributed to diseconomies in scale. To the extent
that they chose an inefficient level of production, bank management is responsible for
scale inefficient operations. However, increasing demand for banking services in the
nineties, fueled by the state’s increasing demand for funds to finance chronic budget
deficits and high growth policies, and the oligopolistic nature of the Turkish banking
market do not justify scale adjustments. Our policy suggestions are that the government
implement financial reform packages that foster competition in the banking market, and
*
Corresponding author. Tel.: +1-504-280-6163; fax: +1-504-280-6397.
E-mail addresses: isik@adm.njit.edu (I. Isik), mhassan@uno.edu (M. Kabir Hassan).
1
Tel.: +973-596-6426; fax: +973-596-3074
0378-4266/02/$ - see front matter Ó 2002 Elsevier Science B.V. All rights reserved.
PII: S 0 3 7 8 - 4 2 6 6 ( 0 1 ) 0 0 1 6 7 - 4
720 I. Isik, M.K. Hassan / Journal of Banking & Finance 26 (2002) 719–766
that the industry devise incentive schemes to improve managerial efficiency. Ó 2002
Elsevier Science B.V. All rights reserved.
JEL classification: G2; D2; G21; G28

Keywords: Turkish banks; Cost efficiency; Ownership and control
1. Introduction
In a rapidly changing financial market worldwide, bank regulators, man-

agers, and investors are concerned about how efficiently banks transform their
expensive inputs into various financial products and services. According to
Berger et al. (1993), although rapid changes in the financial services industry
have been taking place all around the globe, the efficiency research has not kept
pace with these changes, in terms of both maturity and breadth. In their recent
excellent international survey paper, Berger and Humphrey (1997) also un-
derscored the imbalance of the focus in the literature after reviewing 130
frontier (X -) efficiency studies from 21 countries and various types of financial
institutions. They reported that the vast majority of the studies on banking
efficiency focus on the banks of developed countries in general (about 95%) and
of the US in particular (about 70%). While giving possible directions for future
research, both survey studies suggest that more research is needed in measuring
and comparing the efficiency of banks and other financial institutions from
different countries. The economic and political environments surrounding
financial institutions differ substantially across countries. For research and
policy purposes, results from banking markets that are more national in scope
with much higher levels of concentrations may shed some light on the efficiency
impact of various regulatory policies.
During the last two decades, there were continuous legal, structural and
institutional changes in the Turkish financial sector. A series of economic
reforms were introduced in the 1980s to establish a free market economy and
promote financial market development. With these reforms, interest and
foreign exchange rates were freed, and new financial products and institutions
were permitted. The Interbank Money Market and Istanbul Stock Exchange
(ISE) were established in 1986 to provide liquidity in the financial system. As a
reflection of internationalization policies, foreign banks, joint ventures and
partnerships between domestic and foreign banks were also welcomed to the
system. Also, Turkish banks took an interest in doing business abroad and
began operating in international markets. Turkey’s determination for a per-
manent membership in the European Union motivated its banking authori-
ties to ensure that their regulations are in harmony with those in the Union.
Turkey has already accepted EU practices on capital adequacy and ad-
I. Isik, M.K. Hassan / Journal of Banking & Finance 26 (2002) 719–766 721
justed its customs and tariffs according to a mutual agreement with the
EU.
A changing and distinct market structure, with a rich variety of organiza-
tional and ownership types, presents an opportunity to examine efficiency
differences among different forms of banks in Turkey. In comparison with their
North American and Western European peers, Turkish banks do not face any
serious competition from non-bank companies because most of the insurance,
factoring, leasing, and brokerage companies are affiliated with banks through a
holding company structure encompassing financial and non-financial entities.
Since the universal-banking system is in effect in Turkey, banks undertake most
of the activities in both capital and money markets. The banking industry that
dominates the entire financial system has recorded phenomenal earnings in
recent years. At times in the 1990s, the return on asset ratio in Turkish banking
sector has been as high as five times the OECD average (Denizer, 1997).
Moreover, Turkish banks have been getting much higher returns than indus-
trial firms (Zaim, 1995). 2 Traditional banking activities tended to disappear as
a result of increasingly more profitable arbitrage activities, much of which
revolves around the management and funding of large portfolios of govern-
ment securities, which have been the highest yielding asset in Turkey in recent
years. More strikingly, as a result of their extraordinary activities in govern-
ment papers, an increasing number of industrial firms’ non-operating income
has surpassed the income from their core operations.
In this still mixed economic system, publicly and privately owned banks
operate side by side. That is, aside from its regulatory power and interference
in the system, the state also occupies a prominent place in banking. As of 1996,
the state controls about 43% of the total banking assets. Although most of the
banks are young and small, one foreign bank, the Ottoman Bank, is as old as
144 years and a domestic public bank, the T.C. Ziraat Bank, accounts for
about 25% of all commercial banking assets. 3 There are no local banks in
Turkey and all banks are multi branched. All types of individual deposits (TL
or foreign exchange denominated) are fully insured. 4
In spite of continuous efforts to foster competition among economic units
in the recent decades, the Turkish banking industry still seems to suffer from
2
Evidently, profitability of the financial sector was five times higher than that of manufacturing
sector in 1989 (Zaim, 1995). The number one corporate income tax payer in 1999 fiscal year was a
commercial bank.
3
The T.C. Ziraat Bankasi was founded in 1863 whereas the Osmanli (Ottoman) Bankasi was
founded in 1856. Ottoman Bank has a British origin and history of serving as a central bank of
Turkey until 1930 (BAT, 1996).
4
It is claimed that this insurance scheme put into effect following the 1994 crisis signifies fragility
of the system and creates moral hazard incentives as evidenced by the recent speculative
replacement of eight banks into state custody like an ‘‘empty shell’’.
oligopolistic competition. It would probably take over 1000 banks to account

for 70% of the US bank assets, but it takes only the 10 largest banks to account
for that share of the Turkish bank assets. 5 However, especially since the fi-
nancial reforms, the Turkish banks have been operating in an increasingly
competitive market. Existing banks have faced stiff competition not only from
new domestic banks but also from foreign banks. Banks began to shut down
their unprofitable branches and/or reduce branch sizes by laying-off redundant
employees. In a perfectly competitive and contestable financial market, ineffi-
cient banks will be driven out or acquired by efficient banks. Thus, to augment
competitive viability of banks in the new regulatory environment, it is imper-
ative for bank managers and regulators to determine the level and causes of
inefficiency in the banking industry.
There have been a few studies on Turkish banking efficiency. Oral and
Yolalan’s (1990) DEA study measures operating efficiency and profitability of
branch offices of a major domestic commercial bank. Fields et al.’s (1993) non-
frontier translog cost function study focuses on scale economies of national
commercial banks between 1987 and 1988. They showed that a bank does not
have to be large in order to be competitive from a cost perspective, a finding
that is similar to those of the earlier studies from developed countries. Altunbas
et al.’s (1994c) stochastic cost frontier study focuses on inefficiency differences
mainly between public and private banks during the period 1991–1993, and
reported no strong inefficiency differences between public and private banks.
Zaim’s (1995) DEA frontier study investigates the impact of the financial re-
forms introduced in the early 1980s on Turkish bank efficiency. His results
suggest that Turkish banks experienced improved efficiency in a more liber-
alized banking environment, which resulted in improved technical and allo-
cative efficiencies in 1990 with respect to those in 1981.
Our paper differs from other Turkish banking papers in several respects.
First, unlike other studies, we estimate five different measures of non-stochastic
efficiency scores such as allocative efficiency, technical efficiency, pure technical
efficiency, scale efficiency and overall cost efficiency. Second, we also calculate
stochastic cost and alternative profit efficiencies of the Turkish banks, and
compare the results across methods and other studies on Turkish banking. This
study is the first that measures and analyzes the profit efficiency of Turkish
banks. Third, we estimate and report returns to scale for all Turkish banks
according to various size measures during the post liberalization period.
Fourth, unlike other studies, we correlate four measures of financial perfor-
mance with five measures of cost efficiency to investigate whether higher fi-
5
The 5- and 10-bank concentration ratios in the Turkish bank assets in 1988, 1992, and 1996 are
62%, 53%, 49% and 84%, 76%, 72%, respectively. Although the degree of concentration in the
market tends to decline, it is still considerably high.
nancial performance impact bank cost efficiency. Fifth, we also investigate the
impact of size, international presence, control and governance, holding affili-
ation as well as ownership variables on the Turkish bank efficiency scores.
Finally, unlike other studies, we include off-balance sheet activities, inter-bank
funds, directed lending, and investment security portfolios, which are signifi-
cant in volume, in estimation of the efficiency measures.
This paper is divided into five parts. Following the introduction, Section 2
presents the methodology and data. The point estimates of various efficiency
measures and sample statistics of the inputs, outputs and input prices are
discussed in this section. Section 3 provides the empirical results and analyzes
the inefficiency of the banks by tracing their major sources. We also discuss the
returns to scale in Turkish commercial banking, and test whether efficiency
measures are statistically correlated with simple traditional accounting ratios
of financial performance. Section 4 examines the relationship between effi-
ciency and bank characteristics. Section 5 concludes the paper.
2. Methodology and data
Since the production technology of the fully efficient firm in a banking in-
dustry is not known, it must be estimated from observations in practice. Of the
available methods in the efficiency literature, in this paper, we prefer ‘‘the op-
timization of the performance of each production unit’’ (non-parametric ap-
proach) to ‘‘the single optimized regression equation, or a mythical ‘average’ for
each production unit’’ (parametric approach). Hence, to measure the efficiency
of the Turkish banks, we employ Data Envelopment Analysis (DEA). Our
choice of DEA is also driven by the expressed interest in the Turkish banking
industry of reducing costs in the recent years. However, in order to control the
impact of the method choice on the qualitative inferences, we also compare our
DEA results with the results from a parametric frontier method called Eco-
nomic Frontier Approach (see Section 4.1 for the discussion of EFA).
DEA is a linear programming technique that forms a piecewise-linear
convex isoquant over the data points such that no observed point lies to the left
or below it. Thus, DEA frontier represents the set of efficient observations for
which no other production unit or linear combination of units employs as little
or less of every input without changing the output quantities generated or
produces as much or more of every output without altering the input quantities
used. DEA basically allows us to focus on the input saving (cost) efficiency
(CE) of banks, which can be detailed into allocative (AE) and technical (TE)
efficiency components. Thus, the CE measures the proportional reduction in
costs that can be obtained if the bank is both allocatively and technically ef-
ficient. The AE by itself measures the proportional reduction in costs if the
bank chooses the right mix of inputs given the prices (the point of tangency
between the isoquant and budget line). TE measures the proportional reduc-
tion in input usage that can be attained if the bank operates on the efficient
frontier. The DEA also permits us to further decompose the TE into its
components, ‘‘pure’’ technical efficiency (PTE), proportional reduction in input
usage if inputs are not wasted, and scale efficiency (SE), proportional reduction
if the bank achieves constant returns to scale (CRS). Obviously, the decom-
position lets one better trace the sources of inefficiency in production units. All
these point estimates attain values between 0 and 1 for the least and the most
efficient units in the sample, respectively. A more detailed description of DEA
can be found in the literature. 6
2.1. Empirical design for efficiency estimations
Rather than estimating a common frontier across time, we preferred to es-

timate three separate annual efficiency frontiers, specifically for the years 1988,
1992, and 1996. Constructing an annual frontier specific to each year is more
flexible and thus more appropriate than estimating a single multiyear frontier
for the banks in the sample (Bauer et al., 1993; DeYoung and Hasan, 1998).
We also believe that the principal advantage of having panel data is the ability
to observe each bank more than once over a period of time. This is a critical
issue in a continuously changing business environment because the technology
or bank that is most efficient in one period may not be the most efficient in
another. Furthermore, by doing so, we alleviate, at least to an extent, the
problems related to the lack of random error in DEA by allowing an efficient
firm in one period to be inefficient in another, assuming that the errors owing
to luck or data problems are not consistent over time. Given the ongoing re-
structuring of the bank market in Turkey, separate frontiers for different years
would better reflect the changes. We chose the period 1988–1996 in order to
study the performance of the Turkish banks in the post liberalization envi-
ronment. According to Denizer (1997), who studied the impact of the financial
liberalization in Turkey on the competition in the banking market, the interest
rates liberalization process, and elimination of restrictions on foreign exchange
and other financial operations were not complete until 1988, although financial
reforms had began in the early 1980s. Therefore, we choose 1988 as the base
year of the post liberalization period and study the bank performance after
then periodically.
Adoption of periodic rather than continuous assessment of banks in this
study is, however, based on two important reasons. First, unless there is a
fundamental innovation or shock to the banking technology, efficiency mea-
sures are expected to be stable over time; that is, it is unlikely that efficiencies
6
As a reference, see Fried et al. (1993) and Charnes et al. (1994).
will fluctuate markedly over short periods of time (Berger and Humphrey,
1997). Although a couple of banks may marginally improve or worsen their
efficiency over short periods of time, it is unlikely that a very efficient bank in
one year would become very inefficient the next year, only to regain its high
efficiency in the following year (Bauer et al., 1998). Indeed, several studies have
found that efficiency is reasonably persistent over time (Berger and Humphrey,
1991; Berg, 1992; Kwan and Eisenbeis, 1994; Eisenbeis et al., 1996). A relatively
long period is needed for developments in the macro-economy, marketplace and
regulatory environment to exert their influence upon the banking technology.
We assume that a 4-year time interval is adequate to capture any structural
change in the performance of banks. Second, Turkey experienced a funda-
mental economic crisis in 1994 whose impact was devastating for banks: about
30% shrinkage in banking assets, skyrocketing interest rates (700% at times in
the interbank market), and about a 50% devaluation of the Turkish Lira in the
first quarter of the year (Celasun, 1998). Isik and Hassan (2000) reported that
Turkish banks have incurred a 17% productivity loss, a 10% technical regress,
and a 7% efficiency decrease during the crisis. Thus, it would be more appro-
priate to avoid not only 1994 but also 1993 and 1995 because the crisis was an
exogenous phenomenon, whose roots implicitly or explicitly evolved gradually
before its inception and whose impacts lasted for a while after its realization.
2.2. Data
In a performance analysis, production units are expected to be relatively

homogenous, providing similar services and using similar resources. Com-
mercial banks operating in Turkey are depository institutions that cannot take
part in the leasing and trading real goods for commercial purposes. In contrast,
development and investment banks can engage in such activities, but they
cannot accept deposits. These non-depository institutions also do not extend
small commercial and individual loans, which require a substantial amount of
investment in a brick-and-mortar branching network, work force, and red tape,
etc. In fact, they are mostly single branch banks that finance large long term
projects, such as those financed by funds borrowed from the World Bank, IMF
or other international organizations, which provides substantial savings on
overhead, monitoring and control costs. Because of their small market share in
the sector (6% in 1996) as well as quite different technology, structure and goal,
this study excludes development and investment banks and instead concen-
trates on commercial banks. 7 We obtained data from various issues of the
7
Not also included here are special finance houses (profit sharing institutions) that operate
according to Islamic banking principles and other financial institutions that are engaged in
insurance, leasing or factoring business.
Istanbul Stock Exchange (ISE), and Banks Association of Turkey (BAT). Our
sample includes the universe of the commercial banks that have operated in
Turkey between 1988 and 1996, i.e., a raw total of 149 annual observations
from three different years: 39 banks from 1988, 54 banks from 1992 and 56
banks from 1996.
By definition, non-parametric best-practice frontiers are determined by ex-
treme values in the dimensional space created by the choice of inputs and
outputs. Unlike parametric approaches that are non-deterministic, a single
outlier can have a much greater effect on measured efficiency. That is, single
outliers can significantly influence the calculated efficiency measures for each
firm using the non-parametric DEA approach. In order to avoid bias in the
construction of efficiency frontier against which the efficiency scores of all banks
in the sample are relatively measured, we eliminated a few bank observations
because of their either suspicious values or outlier nature. For some banks,
input prices could not be constructed because either the required relevant stock
value for the input or flow value for the expense was reported zero. At times,
even though input prices could be constructed, they were unrealistically large or
small. Therefore, despite the expense of losing some information, we excluded
those observations whose input prices could not be obtained and/or are more
than 2.5 standard deviations away from the mean value of the respective year.
In addition, one bank, Derbank (Caybank), did not report to the BAT in 1988,
for an unknown reason. As a result of this data filtering process, we lost 3
observations in 1988, 4 in 1992 and 3 in 1996, which leaves us with a net total
of 139 observations: 36 from 1988, 50 from 1992 and 53 from 1996.
2.3. Definition of variables: Inputs, input prices, and outputs
‘Reliable’ efficiency prediction requires appropriate definitions and certain

assumptions regarding the measurement of input, output and input price
variables. The exclusion of certain important bank inputs and/or outputs might
bias the final efficiency measures by distorting construction of the frontier (the
locus of the efficient combination of inputs and outputs). To determine what
constitutes inputs and outputs of banks, one should first decide on the nature
of banking technology. In literature on the theory of banking, there are two
main approaches competing with each other in this regard: the production and
the intermediation approaches (Sealey and Lindley, 1977). Like many studies
on banking efficiency (e.g., Aly et al., 1990; Zaim, 1995; DeYoung and Nolle,
1998; Berger and Mester, 1997; Resti, 1997; DeYoung and Hasan, 1998), we
adopt the intermediation approach in this paper. 8 Accordingly, we model
8
Humphrey (1985) presents an extended discussion of the alternative approaches over what a
bank produces.
commercial banks as multi-product firms, producing 4 outputs and employing

3 inputs. All variables except for the input factor labor are measured in millions
of US dollars. 9
The input vector includes (1) labor [LABOR], the number of full-time em-
ployees; (2) capital [CAPITAL], the book value of premises and fixed assets;
and (3) loanable funds [FUNDS], the sum of deposit (demand and time) and
non-deposit funds. 10 Hence the total costs include both interest expense and
operating costs and are proxied by the sum of labor, capital and loanable funds
expenditures. Obviously, all input prices are calculated as flows over the year
divided by these stocks: (1) price of labor [P(LABOR)], total expenditures on
employees such as salaries, employee benefits and reserves for retirement pay
divided by the total number of employees; (2) price of capital [P(CAPITAL)],
total expenditures on premises and fixed assets divided by book value of pre-
mises and fixed assets; and (3) interest rate on loanable funds [P(FUNDS)],
total interest expenses on deposit and non-deposit funds divided by loanable
funds. Expenditures on these inputs account for the vast majority of all
banking costs in Turkey.
The output vector includes: (1) Short-term loans [ST_LOANS], (2) long-term
loans [LT_LOANS]: the loans with less than and more than a year maturity,
respectively; (3) risk-adjusted off-balance sheet items [RA_OFF_B/S]: guar-
antees and warranties (letters of guarantee, bank acceptance, letters of credit,
guaranteed pre-financing, endorsements and others), commitments, foreign
exchange and interest rate derivatives (e.g., forwards, swaps) as well as other
off-balance sheet activities; 11 and (4) other earning assets [OTHER_EA]: loans
to special sectors, inter-bank funds sold, and investment securities (treasury
bills, government bonds and other securities). 12 Off-balance sheet items are
risk-adjusted using Basle Accord risk weights to provide conformity with di-
rectly issued loans in terms of credit risk.
While our definition of bank inputs and outputs is not free from short-
comings, we believe that it might be a reasonable challenge to improve the way
to represent bank production in Turkey. Previous studies on Turkish bank
efficiency did not account for directed lending (i.e., loans to special sectors such
as agriculture, housing, small business, marine or tourism) as they drew no
9
The denomination of the variables in US$ is expected, to an extent, to eliminate the adverse
impact of the inflation on the real magnitudes.
10
Non-deposit funds include borrowed funds from Interbank, central bank, domestic banks,
abroad and others as well as funds raised by issuing securities.
11
Other off-balance sheet activities contain various types of guarantees, mostly regarding
foreign trade, foreign currency and interest rate transactions in addition to loan commitments.
12
Special loans were peculiar to a number of banks and significant in volume. Exclusion of
special loans would probably lead to underestimation of the technical and cost efficiencies for these
banks.
distinction between directed lending and other lending. All state banks and
some private banks make such subsidized loans and ignoring these activities
may lead to understatement of efficiency and productivity scores for such
banks. Another point concerns the investment security portfolios of banks.
Omission of such assets does not reflect the development in the banking
business in Turkey. In recent years, traditional banking operations have been
displaced by the management and funding of large portfolios of government
securities. For example, security portfolios, which are predominantly skewed
towards public sector securities, constituted 13% of total bank assets in 1996,
and have become the most lucrative asset in Turkey in recent years. 13 Thus,
the impact of ignoring security investments as bank outputs will not be uni-
form across banks because some banks, especially small ones are more active in
such investments than others. 14 Another challenge is the inclusion of the very
short-term inter-bank loans. Following the foundation of Interbank Money
Market for Turkish Lira in 1986, banks began to lend and borrow from each
other extensively to meet their short-term liquidity needs. Some banks are more
involved in inter-bank activities than other banks. Since parsimony is desirable
in non-parametric models, we specified the fourth output, other-earning assets
(OTHER_EA), as a ‘‘control bucket’’, which comprises (1) loans to special
sectors (directed lending), (2) inter-bank funds and (3) various investment se-
curities including government papers. 15
Due to tax, regulatory and inflationary distortions, the volume of repo-
transactions as well as other off-balance sheet activities exploded in the more
liberal environment. As a result, repo transactions have reached levels as high
as one quarter of deposits and 16% of total assets, and exceeded the volume of
securities portfolio recorded on the balance sheet. As Berger and Mester (1997)
reported, no earlier frontier study (parametric or non-parametric) accounted
13
On the US$ basis, the realized average period real interest rate in the 3–6 and 6–9 months T-
bills and government bonds were 9% and 27% and 43% in 1995, respectively (BAT).
14
Small banks’ most operations revolve around the management of such portfolios of
government securities (IMF Staff Report).
15
Screening and monitoring costs could be lower for large directed loans than for non-directed
business loans. Such transaction costs could be higher for small directed loans due to more
intensive screening and collection. Similarly, transaction costs for inter-bank loans and investment
activities may involve lower transaction costs. In order to clarify the relative roles of directed loans
and non-directed loans on efficiency scores and organizational forms, we ran sensitivity tests by
using a variety of output mixes. We used a combination of short-term and long-term loans, directed
loans, other loans devoid of directed loans, and risk-adjusted off-balance sheet loans to re-estimate
efficiency scores by using both parametric and non-parametric methods. Our results indicate that
overall efficiency scores have marginally declined, but the trends, sources of inefficiency and relative
relationships between organizational forms remain intact. We, therefore, report the results using
the specification of short-term, long-term, risk-adjusted off-balance sheet loans and other loans as
output variables. We thank the referee for bringing this to our attention.
for off-balance sheet items although they were comparable to loans in terms of
risk and revenue. 16 Thus, in a sense, this study will be the first non-parametric
efficiency study that takes the off-balance sheet items fully and directly into
account and the first frontier efficiency study that treats these items as variable
output. It is critical to include such activities in the efficiency studies because
they are often four or five times greater than the on-balance sheet items when
measured by notional values (Saunders, 1993). These items are often effective
substitutes for loans that demand similar information gathering, origination,
monitoring and control costs. Thus, they might require presumably similar
revenues if they are competitive substitutes for direct loans. Moreover, the risk
weights suggested by the Basle Accord imply that off-balance sheet items have
approximately the same perceived credit risk as loans (Berger and Mester,
1997).
2.4. Sample statistics for the variables and size
Table 1 displays summary statistics for outputs, inputs and input prices of
the Turkish banks for the fiscal years 1988, 1992 and 1996. Panel A reports the
average outputs, inputs and input prices for each year along with their stan-
dard deviations. Panel B details the input prices and size according to the four
groups of banks: (1) national state banks (NAT_STATE), (2) national private
banks (NAT_PRIVT), (3) foreign banks founded in Turkey (FOR_FOU_TR),
and (4) foreign banks having branch offices in Turkey (FOR_BRN_TR).
The first important observation regarding inputs and outputs is the concen-
tration of short-term loans. In every year, short-term loans dominate long-term
loans by at least a factor of six. Instead of long-term risky loans, banks seem to
prefer to extend short term (and floating rate) loans. This is mainly because of
the expected high inflation in the country. The short-term nature of loans can
be considered a rational method of managing interest rate risk. Most of the
deposits in this inflationary environment are short terms as well. For instance,
in 1994, 90% of the total deposits were in 3-month or shorter time maturity
class (BAT, 1994). The second important point is that risk-adjusted off-balance
sheet activities account for a large part of the bank output. Since we weighted
these items according to the Basle risk weights, they are directly comparable to
loans. When compared with loans, RA_OFF_B/S items outweigh the sum of
the short-term and long-term loans in each year. Further analysis of the off-
balance sheet activities indicates that they mainly exist in the national private
and foreign bank groups. Since some of these activities, such as currency and
16
Only a prior non-frontier study, Jagtiani et al. (1995) took a number of off-balance sheet
activities into account and found that cost scale and product mix economies were affected
insignificantly by these activities (Berger and Mester, 1997).
Table 1
Sample statistics of variables: Outputs, inputs and input prices (millions of US dollars)a
Variables Years
1988 1992 1996
Mean S.E. Mean S.E Mean S.E.
Panel Ab
Outputs
ST_LOANS 187.707 306.901 296.970 441.760 410.072 580.706
LT_LOANS 35.412 86.895 29.289 83.315 44.834 109.379
RA_OFF_B/S 337.186 454.883 389.357 551.884 863.020 1114.739
OTHER_EA 220.090 625.692 273.235 881.615 386.456 1263.475
Inputs
LABOR 4012 7905 2797 6564 2677 5630
CAPITAL 37.407 67.781 62.136 180.332 78.774 263.879
FUNDS 682.922 1213.197 873.511 1678.419 1232.219 2222.385
Input prices
P(LABOR) 0.008 0.004 0.019 0.008 0.017 0.009
P(CAPITAL) 0.194 0.174 0.270 0.237 0.271 0.270
P(FUNDS) 0.161 0.052 0.133 0.084 0.174 0.104
Panel Bc
P(LABOR)
NAT_STATE 0.004 0.002 0.010 0.002 0.008 0.003
NAT_PRIVT 0.007 0.003 0.016 0.006 0.014 0.006
FOR_FOU_TR 0.010 0.003 0.025 0.007 0.021 0.008
FOR_BRN_TR 0.012 0.005 0.025 0.008 0.028 0.010
P(CAPITAL)
NAT_STATE 0.052 0.028 0.065 0.043 0.048 0.047
NAT_PRIVT 0.173 0.145 0.228 0.170 0.272 0.237
FOR_FOU_TR 0.375 0.267 0.320 0.260 0.320 0.436
FOR_BRN_TR 0.264 0.174 0.429 0.333 0.354 0.316
P(FUNDS)
NAT_STATE 0.194 0.038 0.207 0.051 0.249 0.147
NAT_PRIVT 0.158 0.050 0.138 0.086 0.165 0.093
FOR_FOU_TR 0.136 0.074 0.133 0.106 0.143 0.083
FOR_BRN_TR 0.158 0.056 0.090 0.410 0.185 0.121
ASSETS
NAT_STATE 2925.242 2476.829 6146.119 3917.786 6379.214 5470.071
NAT_PRIVT 770.306 1189.30 1031.521 1442.258 1330.377 1356.315
FOR_FOU_TR 183.205 175.331 168.325 142.438 283.377 343.189
FOR_BRN_TR 43.561 25.534 90.107 96.837 86.067 86.077
a
Labor is measured in terms of number of employees by the end of the respective year.
b
Panel A summarizes outputs, inputs and input prices of the Turkish commercial banks. Out-
puts: (1) short-term loans, ST_LOANS, and (2) long-term loans, LT_LOANS, comprise the loans
with less than and more than a year maturity, respectively; (3) risk-adjusted off-balance sheet items,
RA_OFF_B/S, include guarantees and warranties (letters of guarantee, bank acceptance, letters of
Table 1 (continued)
credit, guaranteed prefinancing, endorsements and others), commitments, foreign exchange and
interest rate transactions as well as other off-balance sheet activities; (4) other earning assets,
OTHER_EA, consist of loans to special sectors, interbank funds, and investment securities
(treasury bills, government bonds and other securities). Inputs: (1) labor, LABOR, is the quantity
of labor by the number of full-time employees on the payroll; (2) capital, CAPITAL, is the book
value of premises and fixed assets; (3) loanable funds, FUNDS, is the sum of deposit (demand and
time) and non-deposit funds as of the end of the respective year. Input prices: (1) price of labor,
P(LABOR), total expenditures on employees such as salaries, employee benefits and reserves for
retirement pay divided by the total number of employees; (2) price of capital, P(CAPITAL), total
expenditures on premises and fixed assets divided by book value of premises and fixed assets; (3)
price of loanable funds, P(FUNDS), total interest expenses in deposit and non-deposit funds
divided by loanable funds.
c
Panel B details the input prices and size according to four banking groups: (1) NAT_STATE,
national public banks that are owned predominantly by the Turkish taxpayers and voters; (2)
NAT_PRIVT, national private banks whose more than 50% of shares is owned by the Turkish
residents; (3) FOR_FOU_TR, foreign banks founded in Turkey whose more than 50% of shares is
owned by the residents of foreign countries; (4) FOR_BRN_TR, foreign banks’ branches operating
in Turkey.
interest rate swaps and forwards, are complicated financial instruments, they
may need professional management. Such know-how is more likely to exist in
wholesale banks like foreign banks. In notional values, the ratio of the off-
balance sheet items to the on-balance sheet items for the entire banking system
is 1.95, 1.82 and 2.36 for 1988, 1992 and 1996, respectively. 17 This implies that
exclusion of off-balance sheet items in cost efficiency analysis might cause
considerable bias in the results against the banks that are actively involved in
such activities. Finally, the value of both outputs and inputs increases through
time except for labor. The number of employee per bank decreases sharply
from 4012 in 1988 to 2797 in 1992, and then slightly to 2677 in 1996. This fall
might be related to (a) an increase in the number of banks in the system, (b) the
privatization of some state banks and large lay-offs in these banks after priv-
atization, and (c) the spin-off of some state banks’ (e.g., Etibank and Sumer-
bank) non-banking activities. Another plausible reason is the automation in
the bank business and increased consciousness of using resources productively
in the more competitive business environment of the 1990s. However, the
downward trend in the number of bank personnel reversed recently owing
mainly to the growth of small and medium size private banks.
Panel B in Table 1 further elaborates the input prices by groups of banks. As
we look at the cost of inputs, we see that the most expensive bank production
factor in each year is capital, followed by funds, and the least expensive factor
is labor. The investigation of the labor price across groups in each year shows
17
However, this ratio takes a value as high as 11.0 for some banks in the sample.
that foreign banks pay higher wages and benefits than the domestic banks. Per
employee in 1996, FOR_FOU_TR and FOR_BRN_TR paid $21 000 and
$28 000, respectively, whereas, NAT_STATE and NAT_PRIVT paid $8000
and $14 000, respectively. Foreign banks may pay higher wages and benefits to
entice senior managers to live abroad, or to offset the perception among the
foreign staff that the opportunity for advancement within the organization is
limited (DeYoung and Nolle, 1998). Since FOR_FOU_TR banks are usually
partnerships with Turkish investors, they lie in between and have the features
of both domestic and foreign banks. FOR_FOU_TR pay higher wages than
domestic banks because they are foreign but they pay less than ‘‘pure’’ foreign
banks (FOR_BRN_TR) because they are domestic. It also seems that the state
employees receive similar salaries regardless of the state bank they work for, as
evidenced by the relatively low standard deviations of the labor price for state
banks. The highest salary dispersion is in foreign banks perhaps because of the
substantial differences in their origins. While a number of them come from
more developed countries, such as Citibank, Chase Manhattan (US), ABN
Amro (Netherlands), Banca Di Roma (Italy), a number of them come from less
developed countries, such as Bank Mellat (Iran) and Habib Bank (Pakistan).
The analysis of the price of capital shows that foreign banks pay much higher
prices for capital than domestic banks. Foreign banks generally use more state-
of-art technology and newer equipment and buildings than their domestic
counterparts. National banks are relatively old banks and they might still be
working with their old buildings and equipment. The domestic traditional
banks might also be slower to adopt new technology and make investments in
the automation. The cost of capital also includes rent, and foreign banks might
be paying more rents for the offices and equipment than the domestic banks,
which generally own their fixed assets. The price of funds is higher for state
banks than other banks. This might be partly due to the well-known and
chronic cash shortage of state banks. The state often makes use of its banks to
pay public worker salaries and pensions. In case of cash shortages, public
banks borrow from the inter-bank market, usually paying too high rates of
returns, so their liquidity difficulties periodically disrupt the money market and
bid up the cost of funds to other banks (IMF Staff Report). In general, foreign
banks pay less than national banks on funds. Foreign banks might be using
external funding sources by borrowing from international markets or from
their parent companies. The increased demand of the domestic banks for funds
from abroad implies that external funding costs are cheaper than internal
funds, mainly due to a lack of adequate domestic savings.
As for size, on average, state banks are much larger than both domestic and
foreign private banks. The largest commercial bank in the country is a state
bank (T.C. Ziraat Bank). This is also clear from the market share of the state
banks. State banks account for about 41% of the entire commercial banking
assets. Foreign banks generally used to engage in wholesale banking with a
limited number of branch offices. However, they have recently adopted growth
policies to reach individual and corporate customers by extending their pres-
ence into the local deposit market. Consequently, they have increased their
personnel by 35% in 1994 and 7% in 1996.
3. The empirical results pertaining to efficiency estimates
Like earlier studies, we have so far assumed that all types of banks are
coming from the same legal and business environment. However, it may be
questionable to pool domestic and foreign commercial banks into one sample.
Foreign banks have entered Turkish market by establishing either (1) a branch,
(2) a subsidiary or (3) a joint venture with foreign or domestic banks. Foreign
subsidiaries and partnerships have their own sets of books, i.e., they are distinct
organizations from their parents. Both foreign bank branches and subsidiaries
operate much like domestic banks and share the same legal framework. They
can open new branches, accept deposits, make loan, trade securities, and
provide fee-based services (Banks Act, BAT, 1998). 18 However, foreign banks
might have a distinct disadvantage in terms of input efficiency, which is pri-
marily driven by excess expenditures on personnel or dependence on expensive
and unstable purchased funds. Foreign banks could have quite different goals
from domestic banks, as they may be inclined to tradeoff between efficiency and
market share in order to penetrate a local market. Alternatively, foreign banks
might possess some distinct advantages, stemming mainly from the composi-
tion of their asset portfolios. Relative to domestic banks, foreign banks’ asset
portfolios are more skewed to investment securities, whose administrative and
transactional costs are much lower than loans. Also, most foreign banks make
fewer but larger loans, which decreases monitoring and control costs. Also,
lack of exposure in a lesser-known market may manifest itself in the form of
extra information gathering costs for clients. As a consequence, whether banks
of different nationalities share identical production technology (i.e., frontier)
stands as a testable hypothesis. 19 This is a critical issue because if technologies
18
However, a reciprocity provision is in force with respect to the operations of foreign banks.
This provision allows the Council of Ministers to take counter measures if the conditions in any of
the countries in which Turkish banks operate are changed unfavorably.
19
Because pooling increases the reliability of the estimates, we combined state and private
domestic banks into national banks, and foreign bank branches and subsidiaries into foreign
banks, assuming that the differences within each group are a lesser concern. Although the number
of observations in state bank group is limited, we also tested the null hypothesis that state and
private banks share common technology using the same tests, and we generally failed to reject the
null hypothesis. The results are available from authors upon request.
are distinct, the efficiency measures for each group must be estimated relative
to a separate frontier constructed for that specific group.
Assessing domestic and foreign banks first relative to their own group-
specific (separate) frontiers, and then to the pooled (common) frontier, their
efficiency measures are summarized in Panel 1 of Table 2. The mean efficiency
measures calculated relative to separate frontiers are always greater than those
computed relative to common frontier regardless of the year and group, which
implies that separate frontiers always either coincide with or lie inside of the
common frontier. 20 Following the procedures outlined in Byrnes (1985), Aly
et al. (1990) and Elyasiani and Mehdian (1990), we tested the hypothesis of
identical frontiers for each year under study. Overall, both the parametric and
non-parametric test statistics given in Table 3 fail to reject the null hypothesis
of identical technology between the two groups, implying that it is not inap-
propriate to construct production frontiers by pooling data on domestic and
foreign banks. Therefore, the rest of the study continues with the results de-
rived from the common frontier.
Panel 2 of Table 2 presents sample statistics of the various efficiency esti-
mates of Turkish banks calculated relative to the common frontier for the fiscal
years 1988 (Panel 2.A), 1992 (Panel 2.B), 1996 (Panel 2.C), and overall (Panel
2.D). There appears to be a downward trend in the cost efficiency of Turkish
banking. The cost efficiency (inefficiency) was 78% (28%) in 1988, 71% (41%) in
1992, and 68% (47%) in 1996. 21 This means that the average Turkish bank
could have used only 78%, 71% and 68% of the resources actually employed in
1988, 1992 and 1996, respectively, to produce the same level of output in these
years. More plainly, while the average input waste was 28% in 1988, it climbed
to 41% in 1992, and then to 47% in 1996. Apparently, there is substantial room
for significant cost savings if Turkish banks utilize their productive inputs more
efficiently.
Altunbas et al. (1994c) reported that average scores of the cost inefficiency
for the entire Turkish banking industry were 46% in 1991, 32% in 1992 and
49% 1993. As discussed above, their study includes public and private devel-
opment and investment banks in addition to commercial banks. Thus, to make
the results comparable, we calculated the mean inefficiency scores pertinent to
commercial banks from their study (Table 3, p. 7). Accordingly, the inefficiency
20
Byrnes (1985) and Elyasiani and Mehdian (1992) provide a geometric proof for this.
21
The relationship between efficiency (E) and inefficiency (IE) is E ¼ 1=ð1 þ IEÞ or IE ¼
ð1 EÞ=E. Thus, the 78% efficiency implies 28% inefficiency, not 22%. The 78% efficiency figure
means that the average bank in the sample could have produced the same level of output using only
78% of the resources actually employed, if it were producing on the frontier rather than at its
current location. On the other hand, the 28% inefficiency figure means that the average bank needs
28% more resources to produce the same output as the average efficient bank.
Table 2
Efficiency measures of Turkish banks calculated relative to common and separate frontiers
Panel 1. Mean efficiency measures of national and foreign banks in Turkey (1988–1996)
Seperate frontier Common frontier
National banks Foreign banks National banks Foreign banks
1988 1992 1996 1988 1992 1996 1988 1992 1996 1988 1992 1996
CE 0.845 0.789 0.844 0.868 0.778 0.660 0.758 0.690 0.704 0.844 0.747 0.638
AE 0.944 0.902 0.924 0.938 0.864 0.803 0.925 0.897 0.841 0.920 0.841 0.796
TE 0.892 0.875 0.906 0.925 0.873 0.821 0.817 0.764 0.832 0.898 0.855 0.778
PTE 0.949 0.931 0.956 1.000 0.955 0.916 0.920 0.909 0.914 0.983 0.938 0.893
SE 0.939 0.937 0.948 0.925 0.901 0.896 0.889 0.832 0.911 0.913 0.898 0.874
Panel 2. Summary statistics of common frontier efficiency measures of Turkish banks (1988–1996)
Efficiency Obs. Mean Std error Minimum Maximum
measure
Panel 2.A: 1988
(CE)a 36 0.782 0.209 0.238 1.000
(AE)a 36 0.924 0.079 0.651 1.000
(TE)a 36 0.840 0.194 0.366 1.000
(PTE)a 36 0.937 0.116 0.437 1.000
(SE)a 36 0.896 0.167 0.366 1.000
Panel 2.B: 1992
(CE) 50 0.710 0.252 0.120 1.000
(AE) 50 0.877 0.151 0.232 1.000
(TE) 50 0.797 0.218 0.225 1.000
(PTE) 50 0.920 0.131 0.546 1.000
(SE) 50 0.856 0.174 0.381 1.000
Panel 2.C: 1996
(CE) 53 0.685 0.245 0.067 1.000
(AE) 53 0.828 0.191 0.067 1.000
(TE) 53 0.817 0.207 0.281 1.000
(PTE) 53 0.908 0.167 0.409 1.000
(SE) 53 0.901 0.150 0.281 1.000
Panel 2.D: All b
(CE) 139 0.719 0.240 0.067 1.000
(AE) 139 0.871 0.157 0.067 1.000
(TE) 139 0.816 0.207 0.225 1.000
(PTE) 139 0.920 0.142 0.409 1.000
(SE) 139 0.883 0.164 0.281 1.000
a
CE: Cost efficiency, AE: Allocative efficiency, TE: Technical efficiency, PTE: Pure technical
efficiency, SE: Scale efficiency.
b
Panel 2.D: All gives overall summary statistics for the entire industry over 1988–1996.
scores for the commercial banks were 51% in 1991, 34% in 1992, and 49% in
1993. About 30% to 50% of resource waste found in Turkish banking is rela-
tively too high when compared to the inefficiencies in the banks of more
Table 3
Summary of parametric and nonparametric tests for the null hypothesis that national (n) and
foreign (f) banks possess identical technologies (frontiers)
Eff. mea- Test groupsa
suresb Parametric test Nonparametric tests
Individual Analysis of Mann–Whitney Kruskal–Wallis Kolmogorov–
tests Variance [Wilcoxon equality of Smirnov [K–S]
(ANOVA) test Rank-Sum] test populations test test
Hypotheses Meann ¼ Meanf Mediann ¼ Distributionn ¼ Distributionf
Medianf
Test statistics F ðPrb > F Þ z ðPrb > zÞ X 2 ðPrb > X 2 Þ K–S
ðPrb > K SÞ
Panel A: 1988
(CE) 0.110 0.420 0.180 0.238
(0.744) (0.672) (0.672) (0.806)
(AE) 0.050 )0.070 0.005 0.169
(0.825) (0.944) (0.944) (0.986)
(TE) 0.060 0.600 0.361 0.161
(0.803) (0.548) (0.548) (0.992)
(PTE) 2.660 1.590 2.526 0.346
(0.112) (0.112) (0.112) (0.352)
(SE) 0.370 0.260 0.070 0.123
(0.545) (0.791) (0.791) (1.000)
Panel B: 1992
(CE) 0.030 0.570 0.320 0.194
(0.874) (0.572) (0.572) (0.776)
(AE) 0.650 0.080 0.007 0.194
(0.425) (0.936) (0.936) (0.776)
(TE) 0.000 )0.060 0.004 0.170
(0.967) (0.952) (0.952) (0.893)
(PTE) 0.420 0.320 0.105 0.139
(0.521) (0.746) (0.746) (0.979)
(SE) 0.810 )0.360 0.132 0.097
(0.373) (0.716) (0.716) (0.991)
Panel C: 1996
(CE) 6.530 )1.390 1.938 0.374
(0.01) (0.164) (0.164) (0.099)
(AE) 6.530 )1.550 2.402 0.388
(0.01) (0.121) (0.121) (0.079)
(TE) 2.620 )0.680 0.464 0.254
(0.112) (0.496) (0.496) (0.490)
(PTE) 1.120 0.220 0.047 0.174
(0.295) (0.828) (0.828) (0.902)
(SE) 2.170 )0.360 0.126 0.189
(0.147) (0.722) (0.722) (0.835)
a
Test methodology follows Byrnes (1985); Aly et al. (1990) and Elyasiani and Mehdian (1992).
Parametric (ANOVA) and Nonparametric (Mann–Whitney, Kruskal–Wallis and K–S) tests test
Table 3 (continued)
the null hypothesis that national and foreign banks are drawn from the same efficiency population
(environment). The numbers in parentheses are the p-values associated with the relative test.
b
(CE) ¼ Cost efficiency, (AE) ¼ Allocative efficiency, (TE) ¼ Technical efficiency, (PTE) ¼ Pure
technical efficiency, (SE) ¼ Scale efficiency.
developed countries. For instance, the estimated inefficiencies are 24% for the
German universal banks (Altunbas et al., 1994a), 13–17% for the Italian credit
cooperative banks (Altunbas et al., 1994b), 5–10% for the English banks (Al-
tunbas, Molyneux, 1995a), and around 20% of total banking industry costs in
the US (Berger et al., 1993). Zaim (1995) showed that the financial reforms of
the early 1980s have succeeded in fostering competition among commercial
banks and eventually in stimulating them to increase their efficiencies. How-
ever, our results indicate that the efficiency of Turkish commercial banks had
deteriorated in the 1990s. There might be several reasons for this adverse de-
velopment. In contrast to the 80s, the 90s are characterized by high rates of
interest on deposits due to increased competition for funds, and a greater
emphasis on short-term deposits by the public due to high rates of inflation
(BAT, 1996). A more fundamental reason might be that competitive pressures
were not yet strong enough in Turkey to force banks to sustain their efficiency
for survival.
Demand for financial services tends to grow as economies expand and so-
cieties become wealthier. The Turkish economy has grown handsomely in re-
cent years (e.g., 8.1% in 1995 and 7.9% in 1996). Economic expansion seems to
have fueled the demand for banking services and products, as evidenced by the
impressive growth rate in Turkish bank assets (e.g., 32% in 1995 and 22% in
1996). De novo banks enter the market to absorb such heightened demand for
banking services. However, the continuous growth trend of the average
Turkish bank implies that de novo banks were inadequate to supply all of the
new services demanded, which in turn forced the existent banks to enlarge their
scales to meet the unsatisfied demand. However, banks’ operating problems are
accentuated when they grow too much and too fast. Generally, as banks grow,
managers usually do not have sufficient time and expertise to handle all op-
erational details. Thus, many Turkish banks might simply have outgrown their
existing management skills. Also, banks that emphasize growth do so at the
expense of credit quality. Returns, non-performing loans, allowances, and
eventually loan losses rise so that net margins decline.
The greater demand for banking services might have also permitted less cost
efficient production in the banks operating in a concentrated market like
Turkey, at least in the short run. Turkey has an underdeveloped capital mar-
ket. Thus, disintermediation in financial services has not threatened banks yet
in Turkey unlike in the USA or EU, where disintermediation from corpora-
tions along with heightened competition from non-banks have prompted a
frantic search for ways for banks to cut costs, make better use of technology
and squeeze greater returns out of their branches. As mentioned earlier,
competition from non-banks is minimal in Turkey because most of them are in
one way or another affiliated with banks. The liberalization policies adopted in
the 1980s removed the barriers to new entries, which resulted in substantial
increases in the number of commercial banks (e.g., there were 31 de novo
entries into the system, of which 19 were foreign and 12 domestic during the
1980–1990 period). On the other hand, the majority of these new entrants have
been small and in specialized areas such as trade financing and wholesale
corporate financing. Therefore, de novo banks’ marginal contribution to
competition especially in retail banking was limited (Denizer, 1997). 22
3.1. Sources of inefficiency in general
As the results in Panel 2.D of Table 2 indicate, over the years under study, the
average allocative inefficiency is about 15%, whereas the average technical in-
efficiency is about 23%. In each year, allocative inefficiency is always smaller
than technical inefficiency, which suggests that the dominant source of the cost
inefficiency is technical (managerial) rather than allocative (regulatory). Turkish
banks appear to use too much input per unit of output. The higher technical
inefficiency relative to allocative inefficiency also implies that the managers of
Turkish banks were relatively good at choosing the proper input mix given the
prices, but they were not that good at utilizing all factor inputs. Hence, overall
inefficiency in Turkish banks may be attributed, to a great extent, to under-
utilization or wasting of resources rather than choosing the incorrect input mix.
Nevertheless, their ability to chose the optimum input mix seems to be declining,
as evidenced from the fall of allocative efficiency over time (92% in 1988, 88%
in 1992 and 83% in 1996). The reason for increasing allocative inefficiency over
time might be high fluctuation and instability in factor prices due to chronic
inflation in the country in the recent years. If bank managers are uncertain
about prices, they are likely to make inefficient decisions.
The reason for relatively higher technical inefficiency, however, might be
attributed partly to the recent abnormal growth of the small and medium
banks, and partly to heavy investment in expensive computerization and au-
tomation projects and consequently idle capacity. While computers were used
only for back office services in the beginning of 1980s, they were being used
later as on-line real time systems. The 1990s saw the beginning of electronic
banking in Turkey. Turkish banks have invested heavily in computer pro-
cessing and data transmission systems. Several banks have established coun-
22
Using the structure–conduct–performance (SCP) paradigm in his empirical analysis, Denizer
(1997) demonstrated that new entries did not positively influence competition in banking in Turkey.
trywide electronic networks and a number of them began to offer direct access
terminals to their major customers. There were about 4 million credit cards in
the market by the end of 1994, although the Turkish banks had just started
issuing credit cards in August 1988. At the same time, there was a rapid spread
of Automatic Telling Machines (ATMs), Point of Scale (POS) terminals and
bankcards. By the end of 1995, about 5000 ATMs, and 25 000 thousand POS
machines were installed; and nearly 15 million bankcards have been issued
(IGEME). Almost all banks participated in this technological transformation
and established their own ‘‘communication networks’’, but these projects were
expensive, and created a substantial idle capacity. This, in turn, forced some
banks to share the systems they created with other banks. The introduction of
the ‘‘Golden Points’’ by some private banks to share their ATMs is a good
example of the idle capacity created in banking.
Furthermore, the decomposition of total technical (in)efficiency into its
components signals that the main source of inefficiency in Turkish banking is
managerial inefficiency. ‘‘Pure’’ technical efficiency (PTE) is simply technical
efficiency (TE) devoid of scale effects, i.e., the difference between technical ef-
ficiency and ‘‘pure’’ technical efficiency represents the cost of operating at an
incorrect scale. Overall, average scale inefficiency is about 13%, and average
pure technical inefficiency is about 9% (Panel 2.D, Table 2), suggesting that the
major source of the total technical inefficiency for Turkish banks is scale in-
efficiency (output related) and not pure technical inefficiency (input related). 23
During the pre-liberalization era, there was no price (interest rate) competition
among Turkish banks to attract deposits due to the interest rate ceiling im-
posed by the state. When the interest rates were regulated, competition could
only take place through convenience to clients. Accordingly, Turkish banks
competed for scarce deposits by trying to make their services more accessible to
customers even in remote regions. Another incentive for extensive branching
was the increasingly negative interest rates effective in the inflationary envi-
ronment of this era. These remarkable growth efforts provided increases in
deposits but at the expense of investing in costly new brick-and-mortar branch
23
This result contradicts the finding of Zaim (1995), who reported that much of the technical
inefficiency for Turkish banks was from the pure technical component, not scale component. This
variation might be because of the difference in the period studied by Zaim and us. He used 1981 and
1990 years to compare the impact of deregulation on banking efficiency in Turkey. These years
correspond to the period in which there has been a downward trend in both branching and staffing
in commercial banking due to widely spread productivity consciousness after the deregulation.
However, our period, as specified earlier, corresponds to the period of revival of growth in
commercial banking, which brought about again typical extensive employment and branching
issues of the pre-liberalization. The finding of the dominant effect of pure technical efficiency over
scale efficiency was also reported in some studies for other countries (see Aly et al. (1990) for the US
banking and Fukuyama (1993) for the Japanese banking).
offices. On one hand, such growth policies resulted in extreme bank sizes and
eventually overwhelming overhead costs whose adverse impacts on banking
have continued to the present day. On the other hand, the extensive branching
network and heavy technology build-up of the existent banks, together with
consumer inertia and the goodwill for established institutions, have served as
formidable barriers for potential entries and de novo banks’ expansion into the
retail banking.
It seems that in the case of Turkish banks, technical inefficiencies have much
more to do with the scale of production rather than the inefficient utilization of
resources. If the development and growth trend of the relatively young and
small Turkish banks as compared to the established larger banks of more
developed markets are taken into account, our results make sense. The dom-
inant effect of the scale inefficiency indicates that most of the Turkish banks
operate at this ‘incorrect’ scale. They either experience economies of scale (i.e.,
increasing return to scale (IRS)) due to being at less than optimum size or
diseconomies of scale (i.e., decreasing returns to scale (DRS)) due to being at
more than the optimum size. While decreasing or increasing the scale of pro-
duction could result in cost savings or efficiencies, the nature of market demand
may not allow it. The scale inefficiency due to IRS might be attributed to new
and small banks, which desire to grow to reach the ‘right’ scale but currently
fail to do so. Whereas, the scale inefficiency due to DRS might be related to
established large banks, which transgressed the ‘right’ scale perhaps to meet the
excess market demand for financial services and products induced by a growing
Turkish economy. Our results suggest that of the scale inefficient banks, the
majority are non-small banks (80%) and suffering from diseconomies of scale
(77%). This implies that scale inefficiency in Turkish banking is related pri-
marily to an excess level of output production, which seems to be both market
and regulatory driven.
3.2. Returns to scale in the Turkish commercial banking
Since the dominant source of the total technical inefficiency in the Turkish
banking seems to be scale related, it is worth further examining the trend in the
returns to scale of the Turkish banks. As Panel 1 in Table 4 shows, the majority
of Turkish banks have increasingly experienced diseconomies of scale (DRS),
47% in 1988, 48% in 1992 and 53% in 1996, confirming the extra production
costs faced by rapidly growing banks. The share of scale efficient (CRS) banks
has declined from 39% in 1988, to 34% in 1992 and 33% in 1996, signaling
worsening scale efficiency over time. The share of the banks experiencing
economies of scale (IRS) rose from 14% in 1988 to 18% in 1992, and then fell
back to 14% in 1996.
Panel 2 displays the returns to scale by size measured in millions of US
dollars. While Panel 2.A presents overall summary results from the sample of
139 observations over three years, Panel 2.B details them by smaller size
classes. Examination of the panel reveals that while, on average, 35% of all
banks are CRS, the majority, 65%, is scale inefficient (DRS or IRS). Of the
scale inefficient banks, 77% are experiencing DRS, and 23% are experiencing
IRS. Of the scale efficient banks, 39% are small banks, 45% are medium banks,
and 16% are large banks. Of the banks experiencing DRS, only 4% are small
banks and the majority, 96% are medium and large banks (50% due to medium
banks, and 46% due to large banks). Whereas, of the banks experiencing IRS,
the majority (85%) is small, 10% are medium, and only 5% are large. As ob-
served here, the convexity of the frontier assures that banks experiencing IRS
are more frequently smaller banks.
As Panel 2.B shows, except for one bank, none of the banks with total assets
in excess of $400 million has experienced increasing returns to scale (IRS) over
the period under study. 24 Lack of scale economies beyond a certain size is a
common finding in conventional average cost studies. Clark (1988) reports in
his survey that only two studies found significant overall economies of scale
above $100 million of deposits. However, some studies based on samples of
large banks found economies of scale at much larger banks (e.g. Shaffer and
David, 1986; Shaffer, 1988; Noulas et al., 1990). Based on these different
findings, Berg et al. (1991) state that the critical size apparently depends on
the size of distribution of the sample. This could be explained by the fact
that larger banks are involved in more activities. The special activities [input
and output specifications] thus may be less representative of their true activi-
ties.
3.3. Correlation of efficiency measures with financial performance
In order to complement the results of efficiency measures, we correlate

various accounting measures of bank performance with various efficiency
scores. As ‘‘proxy measures of performance’’, we selected the following ratios:
COSTTA (total cost/total assets), TAPE (total asset/number of employee),
ROA (net income/total assets), and ROE (net income/total equity). We cal-
culated both rank-order Spearman and regular Pearson correlation coefficients
to examine the possible relationship among the X -efficiency measures and
‘‘proxy measures of performance’’. The Spearman [s] and Pearson [p] corre-
lation coefficients are presented in Table 5. The null hypothesis is that the
correlation coefficient between two variables is zero.
24
97% of the banks experiencing decreasing returns to scale are the banks with total assets in
excess of $400 million. This and the lack of IRS beyond $400 million imply that the optimal scale
for the Turkish banks lies between 100 and 400 million dollars of assets.
742
I. Isik, M.K. Hassan / Journal of Banking & Finance 26 (2002) 719–766
Table 4
Returns to scale (RTS) in Turkish commercial banking
Panel 1. Developments in RTSa
RTS Years
1988 1992 1996
# of % share # of % share # of % share
banks banks banks
CRS 14 39 17 34 18 33
DRS 17 47 24 48 28 53
IRS 5 14 9 18 7 14
Total 36 100 50 100 53 100
b
Panel 2. RTS by size
Size CRS DRS IRS Total
# % col. % row # % col. % row # % col. % row # % col. % row
Panel 2.A
SML_BNKS 19 39 48 3 4 7 18 85 45 40 29 100
MED_BNKS 22 45 38 34 50 59 2 10 3 58 42 100
LAR_BNKS 8 16 20 32 46 78 1 5 2 41 29 100
Total 49 100 35 69 100 50 21 100 15 139 100 100
Panel 2.B
<40 8 16 36 2 3 9 12 59 55 22 16 100
40–100 11 22 61 1 1 6 6 29 33 18 13 100
100–400 13 29 43 15 22 50 2 8 7 30 22 100
400–1000 9 17 32 19 28 68 0 0 0 28 20 100
1000–2500 4 8 20 14 20 75 1 4 5 19 14 100
>2500 4 8 18 18 26 82 0 0 0 22 15 100
Total 49 100 35 69 100 50 21 100 15 139 100 100

a
Panel 1 presents the trend in the RTS of the Turkish commercial banks by year; where RTS are the increases in output that result from increasing

all inputs by the same percentage. There are three possible cases: (1) Constant returns to scale (CRS), which arise when D% in outputs ¼ D% in inputs,
i.e., if a firm doubles all its inputs, its output exactly doubles; (2) Decreasing returns to scale (DRS), which occur when D% in outputs < D% in inputs,
i.e., if a firm doubles all its inputs, its output less than doubles; (3) Increasing returns to scale (IRS), which happen when D% in outputs > D% in
inputs, i.e., if a firm doubles all its inputs, its output more than doubles.
b
Panel 2 provides the summary of overall RTS according to various size groups, first by larger (Panel A), then smaller (Panel B) size ranges, over the
years 1988, 1992 and 1996. For instance, over the years studied, 40 observations (29% of total 139 obs. ¼ % col.) belonged to SML_BNKS, of which 19
(48% of 40 SML_BNKS obs. ¼ % row) experienced CRS, 3 (7%) experienced DRS, 18 (45%) experienced IRS. Over these years, 49 observations (35%
of total 139 obs. ¼ % row) belonged to the banks that experienced CRS, of which 19 (39% of 49 CRS obs. ¼ % col.) were SML_BNKS, 22 (45%) were
MED_BNKS, and 8 (16%) were LAR_BNKS; where SML_BNKS (small banks): assets < 100; MED_BNKS (medium banks): 100 6 assets < 1000;
LAR_BNKS (large banks): assets P 1000 [$ million].
743
Table 5
Spearman rank order [s] and regular Pearson [p] correlation coefficients among efficiency estimates
and proxy-measures of performancea
(CE)b (AE)b (TE)b (PTE)b (SE)b COSTTAb TAPEb OAb
(AE)
s 0.698
p 0.695
(TE)
s 0.884 0.395
p 0.868 0.283
(PTE)
s 0.611 0.194 0.695
p 0.576 0.093 0.692
(SE)
s 0.773 0.420 0.884 0.396
p 0.702 0.296 0.793 0.123
COSTTA
s )0.389 )0.251 )0.364 )0.270 )0.310
p )0.366 )0.225 )0.316 )0.245 )0.223
TAPE
s 0.416 0.225 0.405 0.267 0.333 )0.465
p 0.325 0.135 0.332 0.191 0.290 )0.394
ROA
s 0.360 0.162 0.346 0.274 0.252 )0.329 0.197
p 0.277 0.062 0.296 0.254 0.178 )0.272 0.125
ROE
s 0.422 0.214 0.374 0.307 0.235 )0.329 0.390 0.791
p 0.374 0.194 0.365 0.343 0.203 )0.222 0.324 0.700
a
Spearman [s] correlation coefficient – first rows of each cell. Parametric ordinary Pearson [p]
correlation coefficient – second rows of each cell.
b
(CE) ¼ Cost efficiency; (AE) ¼ Allocative efficiency; (TE) ¼ Technical efficiency; (PTE) ¼ Pure
technical efficiency; (SE) ¼ Scale efficiency; COSTTA, sum of total interest and non-interest ex-
penses divided by total assets, is average cost and proxies cost efficiency; TAPE, total assets divided
by total number of employees, proxies productivity level per employee; ROA, return on assets is net
income divided by total assets and proxies productivity of asset base; ROE, return on equity is net
income divided by total equity and proxies productivity of equity.

10% significance levels respectively for the tests of zero correlation for the non-parametric rank
order.

order.

order.
The results show that the Spearman [s] correlation coefficients are all sig-
nificantly different from zero, indicating that there is a strong association
among the X -efficiency measures and proxy measures of performance. Gener-

ally, the Pearson [p] coefficient results confirm all the relationships found with
the Spearman in the direction (positive or negative) and significance. 25 Cost
efficiency (CE) is highly positively and statistically significantly associated with
other X -efficiency measures, namely, AE, TE, PTE, and SE (qCE;AE ¼ 0:698,
qCETE ¼ 0:884, qCEPTE ¼ 0:611, qCESE ¼ 0:773, respectively). TE is more
related to SE than to PTE (qTESE ¼ 0:884 versus qTEPTE ¼ 0:695), confirming
the dominant effect of scale efficiency in determining the technical efficiency of
the Turkish commercial banks. As expected, COSTTA, a proxy for cost
(in)efficiency, is significantly negatively correlated with all efficiency scores and
proxy measures of performance. COSTTA is also associated more with the X -
efficiency measure it proxies, CE, than with the other efficiency measures, AE,
TE, PTE and SE. Total asset per employee, TAPE, a proxy for the average
productivity or asset generating ability per employee, is significantly and posi-
tively correlated with all X -efficiency measures as well as ROA and ROE, and
significantly negatively correlated with COSTTA. There is also statistically
strong correlation between profitability ratios (ROA and ROE) and efficiency
measures. This finding supports Miller and Noulas (1996) and Hasan and
Marton (2000) in that most profitable banks are also the most efficient.
In sum, the statistically and significantly different from zero correlation
coefficients suggest that our X -efficiency measures are strongly associated with
conventional proxy measures of performance, i.e., they are robust and are not
‘meaningless’ artifacts of our techniques. Elyasiani et al. (1994) have also
shown that information contained in efficiency measures closely corresponds to
that contained in standard financial ratios. 26
4. Organizational structure and efficiency in the Turkish banking industry
Banks are not homogenous in terms of ownership structures and organi-

zational forms. Relying on different theories and hypotheses such as the theory
of the firm, theory of international trade, agency theory and market discipline
hypothesis or their combinations, studies have examined whether ownership
structure or organizational form is correlated with differences in frontier effi-
ciency (e.g. Aly et al., 1990; Elyasiani and Mehdian, 1992; Grabowski et al.,
25
In a few cases, the relationship found significant with Spearman rank-correlation is not
supported by the ordinary Pearson (AE-PTE, AE-ROA, PTE-SE, TAPE-ROA). Because of the less
stringent assumptions required with Spearman, the results obtained with Spearman are more
credible. Thus, we attribute the difference found in a few cases to the question of satisfying
assumptions underlying each method.
26
Berger and Mester (1997) also report a strong correlation between efficiency scores and
commonly used financial ratios such as C/GTA (Total Cost/Total Assets), ROA and ROE.
1993; Pi and Timme, 1993; Mester, 1993; Berger and Mester, 1997; DeYoung
and Hasan, 1998; Hasan and Merton, 2000). Most of these studies hypothesize
that the type of ownership and organizational form that generates stronger
incentives to control costs and/or boost profits is expected to be more efficient.
Furthermore, investigation of the differences in efficiency measures with
respect to ownership and organizational structures might be helpful in guiding
government policy and private practice in the banking business. If the gov-
ernment considers privatizing state enterprises (the Turkish government cer-
tainly does), it should know whether privatization is beneficial and how this
should be done. If the privatization of state banks can be justified on efficiency
grounds, then it will be both economically and politically easy to convince the
public. Knowledge of the level and types of inefficiency in banking can help
the government devise programs suitable to successful privatization and im-
provement of the sector performance.
4.1. Comparison of the efficiency across organizational forms
Evidence to date suggests that inefficiencies on the output side may be as

large or larger than those on the input side (Berger et al., 1993). Unlike cost
efficiency measures, profit efficiency accounts for errors on the output side as
well as on the input side. That is, profit efficiency is concerned with both cost
and revenue efficiency but only under certain conditions would it be likely that
the former will equal the sum of the latter. For example, high-quality producers
may appear to be cost inefficient because of the extra expenses associated with
producing the higher quality output. The studies on the cost efficiency aspect of
banking firms are voluminous. However, although profit efficiency has become
more popular among researchers in recent years because of its more accepted
economic goal of profit maximization, there are only a handful of studies on
profit efficiency in the banking literature (Berger and Humphrey, 1997). In this
line of research, our study will be the first to investigate the profit efficiency of
the Turkish commercial banks.
To measure the stochastic cost (CostE1) and ‘alternative’ profit (ProfitE)
efficiencies of Turkish banks, using the same input and output vectors, we
employed the Economic Frontier Approach (EFA), which was first introduced
by Aigner et al. (1977). Our definitions of CostE and ProfitE follow the studies
of Berger and Mester (1997) and DeYoung and Hasan (1998). Stochastic cost
efficiency is defined as the estimated cost needed to produce the bank’s outputs
if the bank was as efficient as the best practice bank in the sample facing the
same exogenous variables such as input prices and variable outputs, divided by
the actual cost of the bank, adjusted for random error. Alternative (nonstan-
dard) profit efficiency is defined as how close a bank comes to earning maxi-
mum profits given its output levels rather than its output prices as in standard
profit efficiency. According to both studies, alternative profit efficiency is su-
perior to standard profit efficiency in many respects, one of which is the

problems in measuring output prices, a real concern for our RA_OFF_B/S
output because it is hard, if not impossible, to get accurate price information
for off-balance sheet activities. 27
Table 6 reports the mean efficiency measures of the three studies on Turkish
banking according to banking groups. It is important to note that these three
studies use different sets of inputs and outputs, utilize different approaches
(parametric versus non-parametric), employ distinct efficiency concepts (cost/
input versus profit) and have different time periods. Our non-parametric (CE,
AE, TE, PTE, and SE) and parametric (CostE1 and ProfitE) results are from
1988, 1992 and 1996. While Zaim’s (1995), Zaim (‘95), non-parametric results
(CE, AE, TE, PTE, and SE) are from 1981 and 1990, Altunbas et al.’s (1994c),
AMM (‘94), parametric (CostE2) results are from 1991, 1992, and 1993. Since
we are interested in relative relationships across banking groups as well as a
general level comparison rather than year specific, we averaged the efficiency
measures obtained in each study over the relevant years. Apparent differences
between ‘raw’ efficiency measures of different groups may not be statistically
significant. In this study, we compare and contrast bank performance not only
according to nationality (domestic versus foreign) and ownership (private
versus public), but also according to size (small versus medium versus large),
holding affiliation (in-holding versus independent banks), international pres-
ence (multinational versus ‘pure’ domestic banks), corporate control and
governance (publicly versus privately held banks; board- and management-
same-banks versus otherwise). We also provide Kolomogorov–Smirnov non-
parametric significance pair-wise test results in Table 6. The purpose is to
investigate the potential correlates and sources of efficiency as well as to
compare the average efficiencies across the studies on Turkish banking. 28
The overall cost efficiency for all commercial banks in the system is 72% in
current paper, 65% in Zaim (‘95), and 56% in AMM (‘94). In order to receive
non-biased efficiency measures, appropriate set of inputs and outputs that
would represent the ‘true’ nature of the production process should be defined.
As Berg et al. (1991) stated, interpretation of the efficiency measures varies for
different sets of inputs and outputs. Unlike other studies, current paper takes
into account the off-balance sheet items and directed lending as well as in-
vestment securities and inter-bank funds (both bought and sold). Off balance
sheet items account for the vast majority of the banking business in Turkey.
27
A detailed description of the efficiency concepts (stochastic cost efficiency, standard and
alternative profit efficiencies) could be found in Berger and Mester (1997).
28 base
indicates that the variable is the base variable for the pair-wise significance tests of the
means in each class in Table 6, e.g.; in the size class, SMALL is the base against which the
significance of the means of MEDIUM and LARGE banks is tested pair-wise (SMALL
MEDIUM and SMALLLARGE).
Table 6
748
Mean efficiency scores according to various banking groups (in comparison to other studies on Turkish banking)a
Studies CURRENT PAPER [CP(’00)] CP(‘00) AMM(’94) Zaim (’95)
Techniques Nonparametric frontier (DEA) Parametric frontier Parametric Nonparametric frontier (DEA)

frontier
Variables/Efficiency CE AE TE PTE SE CostE1 ProfitE CostE2 CE AE TE PTE SE
measures
A. Size variables
SMALLbase 0.720 0.843 0.834 0.901 0.923 0.882 0.894 – – – – – –
MEDIUM 0.783 0.903 0.861 0.939 0.908 0.843 0.829 – – – – – –
LARGE 0.674 0.896 0.755 0.939 0.799 0.817 0.793 – – – – – –
B. Corporate control variables

NON_CEO_CHAbase 0.731 0.886 0.826 0.925 0.892 0.891 0.828 – – – – – –
CEO_CHA 0.655 0.881 0.736 0.878 0.836 0.901 0.843 – – – – – –
C. Corporate governance variables
PRIVTLY_HELDbase 0.704 0.873 0.802 0.897 0.891 0.893 0.839 – – – – – –
PUBLCLY_HELD 0.742 0.909 0.817 0.970 0.839 0.905 0.829 – – – – – –
D. Holding affiliation variables
INDEPENDENTbase 0.642 0.827 0.776 0.866 0.896 0.825 0.704 – – – – – –
IN_HOLDING 0.803 0.864 0.930 0.998 0.932 0.871 0.845 – – – – – –
E. International presence variables
PURE_ 0.663 0.804 0.813 0.884 0.925 0.772 0.814 – – – – – –
DOM_BNKSbase
MN_DOM_BNKS 0.736 0.871 0.848 0.939 0.901 0.899 0.719 – – – – – –
F. nationality variables
NAT_BNKSbase 0.714 0.882 0.806 0.914 0.879 0.880 0.824 0.583 0.622 0.797 0.842 0.868 0.970
FOR_BNKS 0.731 0.844 0.838 0.933 0.893 0.901 0.868 0.520 0.760 0.795 0.947 0.960 0.986
G. Ownership variables
STATE_BNKSbase 0.708 0.911 0.778 0.914 0.846 0.844 0.769 0.397 0.653 0.703 0.919 0.952 0.966
Variables/Efficiency CE AE TE PTE SE CostE1 ProfitE CostE2 CE AE TE PTE SE
measures
PRIVT_BNKS 0.720 0.866 0.820 0.921 0.887 0.898 0.845 0.581 0.665 0.820 0.861 0.880 0.977
H. Nationality and ownership variables

STATE_BNKS
NAT_STATE 0.708 0.911 0.778 0.914 0.846 0.844 0.769 0.397 0.653 0.703 0.919 0.952 0.966
FOR_STATE – – – – – – – – – – – – –
PRIVT_BNKS
NAT_PRIVTbase 0.715 0.877 0.810 0.914 0.884 0.898 0.833 0.623 0.609 0.835 0.810 0.833 0.972
FOR_PRIVT 0.731 0.844 0.838 0.933 0.893 0.886 0.868 0.520 0.760 0.795 0.947 0.960 0.986
FOR_FOU_TR 0.798 0.908 0.870 0.905 0.960 0.855 0.872 0.523 – – – – –
FOR_BRN_TR 0.688 0.802 0.818 0.951 0.850 0.907 0.866 0.517 – – – – –
I. All commercial banks 0.719 0.871 0.816 0.920 0.883 0.895 0.837 0.560 0.647 0.790 0.863 0.885 0.974
a
CP (‘00) refers to the current paper (2000) whose efficiency measures summary is based on 1988 þ 1992 þ 1996 for all variables, except for holding
affiliation and international presence variables that were measured only in 1996 due to data limitation; Zaim (’95) refers to Zaim (1995)’s DEA work
whose efficiency measures summary is based on 1981 þ 1990; and AMM (’94) refers to Altunbas, Altunbas et al. (1994c)’ s stochastic cost study whose
cost efficiency measures summary is based on 1991 þ 1992 þ 1993.
*
Significance at 10% confidence levels, respectively, in the difference between the means of the base and comparison group according to non-
parametric K–S test.
**
***
749
These activities are also concentrated mainly in domestic and foreign private
banks. Thus, a great portion of the banking activity was ignored in these earlier
studies. Consequently, the mean efficiency measures that we found for all
Turkish banks in general and for private banks in particular are generally
higher than those of the other two studies. 29
4.1.1. Size variables

The results show that there exists a negative relationship between bank
size and efficiency. Both small (SML_BNKSbase ) and medium banks
(MED_BNKS) have significantly higher costs (both non-parametric, CE, and
parametric, CostE1), technical and scale efficiencies than large banks
(LAR_BNKS). It appears that small banks are also more profit efficient than
large banks. These findings support the studies of Hermalin and Wallace
(1994), Kaparakis et al. (1994), DeYoung and Nolle (1998), and Berger and
Mester (1997), who reported that efficiency measures demonstrate no potential
scale biases favoring larger banks. In Turkey, smaller banks are typically
newer. They generally specialize in trade finance and wholesale corporate
banking and employ more professional and astute management teams. Al-
though there are no restrictions on the scope and scale of their operations,
these banks have not extended their presence beyond the three largest cities of
the country, Istanbul, Ankara and Izmir (Denizer, 1997). Hence, competition
in rural markets, where there are only a few large banks, is not as strong as the
competition in metropolitan markets, where banks of all sizes fight head-on.
Hence, small banks face relatively stronger competition in all of their markets
than large banks do. This might have induced more market discipline on small
banks, which needed to be more cost efficient to survive. However, it may be
that higher inefficiency in large banks stems simply from the nature of their
business in rural markets, where the population density is low, making the
delivery systems more costly. Our results also suggest that there is no signifi-
cant relationship between allocative efficiency and bank size, a finding con-
sistent with that of Aly et al. (1990), and not surprising in the sense that
inefficiencies might arise even though banks do their best in response to shadow
market prices. Small banks are significantly more scale efficient than both
medium banks (by about 2%) and large banks (by about 12%). This is con-
sistent with what we observe in Table 4: of the scale inefficient banks, only 20%
are small and the rest are non-small (medium and large) banks.
All banks, whether small, medium or large, seem to be suffering more from
technical inefficiency rather than allocative inefficiency. But, unlike small
29
For instance, our CostE1, 89%, is overall substantially higher than the CostE2 of AMM (’94),
56%. Holding the differences in the sample and time period constant, this could be a product
of difference in representation of the bank production in Turkey between the two studies.
banks, the technical inefficiency in medium and large banks is driven mostly by
scale inefficiency, not pure technical inefficiency. In addition, scale problems
appear to be more significant with large level of output production.
4.1.2. Corporate control and governance variables

The agency cost theory implies that the objectives of managers and stock-
holders may differ from each other, resulting in non-optimal decisions by
management and inefficient firm behavior. When the CEO is also chairman of
the board, decision management and decision control reside in one person, and
managers might take actions that may lead to poor performance (Pi and
Timme, 1993). Consistent with expectations, whether parametric or non-
parametric, all ‘‘raw’’ efficiency scores of the Turkish banks whose CEO is also
the chairman of the board (CEO_CHA), are lower than those of banks whose
CEO is not the chairman of the board (NON_CEO_CHAbase ). Efficiency dif-
ferential is also statistically material in cost (CE), technical (TE), and scale (SE)
efficiencies.
According to market discipline hypothesis, to the extent that the stockholders
of the bank can exert discipline over bank management, banks whose shares
are publicly traded are expected to be more efficient (Berger and Mester, 1997).
Our results suggest that the Turkish banks whose shares are publicly traded in
the ISE (PUBLCLY_HELD) dominate the banks whose shares are not traded
(PRIVTLY_HELDbase ) in terms of every ‘‘raw’’ efficiency measure except for
scale (SE) and profit (ProfitE) efficiency. However, the superiority of the
publicly traded banks to the privately owned banks is more stressed in allo-
cative (AE) and ‘‘pure’’ technical efficiency (PTE) scores. For each group, the
dominant source of the cost inefficiency is again not allocative but technical
inefficiency, which is driven predominantly by scale inefficiency.
4.1.3. Holding affiliation variables

An important feature of the banking industry in Turkey is the interwoven
ownership structure of banking and non-banking businesses. Holding com-
panies control the ownership and management of some banks. Likewise, some
banks have a significant ownership stake and thus control the management of
industrial corporations. This kind of ownership structure has a historical basis
that dates back to early 1970s, an era marked as the beginning of bank holding
companies. During this period, Turkey adopted high growth policies led by
both public and private sector investments. While the state had access to
capital using several means to finance its large investments, the private sector
did not. Due to lack of capital markets, industrial firms had to depend on bank
loans to fund their projects (Fry, 1979). Because public banks were the major
commercial banks in the industry and committed primarily to financing public
projects, the private sector had no choice but to establish or acquire banks to
meet their growing funding needs (Denizer, 1997).
Holding company structure could result in complicated organizational

forms or internal management structure, centralization of major decisions,
different goals, connected lending, and funding of negative NPV investments
of the affiliates, which in turn might lead to cost and/or profit inefficiency. On
the other hand, such a structure might also bring about some opportunities
as well as discipline for banks. Different parts of the holding company can
derive benefits from one another. For instance, the holding company struc-
ture enables firms to tap resources of each other like labor, capital, informa-
tion, know-how, and funds and becomes a savior resort for each other’s
idle assets. Retail banking, with its branch structure, also creates customer
relationships that help the bank to market the products of its affiliates beside
its own services. Apart from financing its affiliates, banks also enjoy fund-
ing the purchases and leases of its affiliates’ customers. Thus, there could be
some synergy opportunities both on output and input side for banks affili-
ated with a holding company, contributing to the efficient utilization of re-
sources.
In order to examine the impact of holding company affiliation on efficiency,
we divided domestic banks into two groups in 1996: (1) those operating in a
holding company structure encompassing financial and/or non-financial enti-
ties (IN_HOLDING), and (2) those operating independently (INDEPEN-
DENTbase ). 30 All point efficiency scores indicate that banks in holding
companies tend to outperform independent banks. As can be seen from Table
6, their profit (ProfitE) as well as technical (TE) and cost (both non-parametric,
CE, and parametric, CostE1) efficiencies are significantly greater than those
of independent banks. Some earlier efficiency studies on US banks also found
that being in a holding company confers some cost advantages compared to
staying independent (Newman and Shrieves, 1993 and Berger and Mester,
1997). 31 It seems that there are some synergies resulting from working under
the umbrella of a holding company where firms are connected but have sep-
arate entities.
As for sources of inefficiency, INDEPENDENT banks suffer less from
allocative inefficiency than technical inefficiency, which is mostly related to
poor utilization of resources rather than operating at incorrect scale. In
IN_HOLDING banking firms, problems appear to be more allocative than
technical. This could be a result of a multi-layered management structure in
these institutions, which hampers optimal decision-making and denies the in-
dependence needed to choose the correct mix of inputs.
30
Lack of data for earlier years forces us to conduct our analysis for only 1996.
31
Elyasiani and Mehdian (1990) found no difference between one-bank and multibank holding
company affiliation and Grabowski et al. (1993) found that branching organizational forms are
more efficient than multibank holding companies.
4.1.4. International presence variables

There exist a number of theories seeking to explain the multi-nationalization
of banks. Among them are the theory of comparative advantage (e.g., Lees, 1974;
Aliber, 1976), the theory of industrial organization (e.g., Hymer, 1976; Kindle-
berger, 1969; Caves, 1971), and the eclectic theory of international production
(e.g., Dunning, 1977; Dunning, 1980; Gray and Gray, 1981). 32 The cost ad-
vantages due to the presence of scale and scope efficiency resulting from foreign
expansion are the major reason implicit in these theories. Therefore, a finding of
greater efficiency for banks with an international presence versus purely do-
mestic one might suggest that multi-nationalization enhances the efficiency due
to the exploitation of scope or scale economies or new techniques and concepts
‘exported’ from abroad. In order to investigate this case, we excluded the for-
eign banks operating in Turkey and focused on only domestic (national) banks
with and without presence in international markets as of 1996. 33
It appears that multinational domestic banks (MN_DOM_BNKS) are su-
perior to purely domestic banks (PURE_DOM_BNKSbase ) in terms of all
‘‘raw’’ efficiency measures except for scale efficiency. The reason for the lower
scale inefficiency in MN_DOM_BNKS might be the fact that they are mainly
medium and large banks that already suffer from diseconomies of scale. Thus,
exploitation of scale economies cannot be the motivation for these already large
banks’ foreign expansion. This finding is consistent with the assumption com-
monly shared in the literature that multi-national banks would have exhausted
their scale economies when they were small, pure domestic banks (e.g., Gray and
Gray, 1981). While the major source of cost inefficiency in MN_DOM_BNKS is
technical inefficiency, it is allocative inefficiency in PURE_DOM_BNKS. The
observation that MN_DOM_BNKS suffer relatively less from allocative inef-
ficiency might imply that an international presence might allow those banks
more price choice to optimally determine their input mix. MN_DOM_BNKS
seem to suffer more from scale inefficiency than pure technical inefficiency
whereas PURE_DOM_BNKS seem to experience the opposite. Purely do-
mestic banks are relatively small and usually are not exposed to the problems
stemming from an increase in scale. Their problem looks more related to the
under-utilization of resources rather than to operating at an incorrect scale.
However, these results should be regarded as exploratory because the efficiency
32
For more discussion of the theories about multinationalization of banks, see Mahajan et al.
(1996).
33
As of 1996, Turkish banks have individual or joint equity participation in over fifty banks
and/or other financial institutions in the USA, Germany, France, the Netherlands, Switzerland, the
UK, Hungary, the Russian Federation, Kazakhstan, Azerbaijan, Uzbekistan, Turkmenistan,
Ireland, the Channel Islands, Bahrain, and Luxembourg, and the Turkish Republic of Northern
Cyprus, a reflection of the progressive multinationalization of the financial institutions in Turkey
(BAT, 1996).
differentials between the two groups do not seem to be statistically strong. Thus,
a more focused study is needed to draw a conclusion about the impact of multi-
nationalization on Turkish banking efficiency.
4.1.5. Nationality variables

In contrast to scale and scope efficiency studies, X -efficiency studies have found
that foreign-owned banks in the US were significantly less efficient than US-
owned banks (DeYoung and Nolle, 1998; Mahajan et al., 1996; Chang et al.,
1998). These studies suggest that foreign-owned banks usually had to trade effi-
ciency (both profit and cost) for rapid expansion of market share. The proposed
reason for lower efficiency is that foreign banks in the US finance their rapid
growth by relying on purchased funds, which are more expensive than core
deposits raised through a network of branches, which takes time to establish.
Like Zaim (1995), we find that foreign banks (FOR_BNKS) operating in
Turkey dominate national banks (NAT_BNKS) with respect to ‘‘raw’’ effi-
ciency scores CE, TE, PTE, and SE. 34 However, both his and our results
suggest that the national banks are more allocatively efficient (AE) than foreign
banks. The observation that national banks are superior in terms of allocative
efficiency might be due to the fact that national banks face less expensive input
prices, particularly in labor. Unlike Zaim (1995), we test the significance of
efficiency differential between the two groups, and find that efficiency differ-
ences between national and foreign banks in CE, TE and SE are also statis-
tically significant. In both types of ownership, it seems that the source of the
cost inefficiency is mainly technical, and the source of the technical inefficiency
is mainly scale inefficiency.
The results found in the non-US X -efficiency studies seem to contradict the
results found in the US studies. Foreign banks in India were found to be
somewhat more efficient than privately owned domestic banks (Bhattacharya
et al., 1997). Hungarian banks with foreign involvement were also found to be
more efficient than their domestic counter-parts (Hasan and Marton, 2000).
Specific market conditions in these countries may have presented opportunities
for foreign banks to exploit their comparative advantages, resulting in higher
efficiencies. One implication is that the context in which foreign banks operate
has an important role in determining their efficiency. Foreign banks operating
in Turkey predominantly come from more developed countries. They found a
banking system that is newly opening to the world and awakening after a long
34
Altunbas et al. (1994a,b,c) reported that the foreign banks (subsidiaries and branches) are less
cost efficient (CE) than the national banks in terms of ‘‘raw efficiency scores’’. However, the
efficiency difference found was statistically insignificant. Because of their non-homogenous
business, Zaim (1995) and we, unlike Altunbas et al. (1994a,b,c), excluded foreign and domestic
development and investment banks in the construction of the benchmark frontier. Thus, our results
are more comparable with Zaim’s.
period of protection by the state against competition both from inside and
outside. In spite of their small market share, foreign banks have a significant
place in the Turkish banking industry due to the new concepts and practices
they have introduced. After fortifying their position, they began to compete
with their local counterparts in a wide spectrum of services from consumer
banking to corporate financing. As previously mentioned, in order to reduce
their dependence on expensive brokered deposits, foreign banks recently in-
creased their presence in the local deposit markets, as evidenced by their
substantial employment of new labor and opening of new branches. It seems
that foreign banks have begun to overcome the lack of exposure in a lesser-
known market and improve their operating efficiency over their domestic peers.
4.1.6. Ownership variables

The period 1983–1993 corresponds to the period of major structural changes
in the financial system triggered by the reforms of the late Turgut Ozal’s suc-
cessive governments. During this time, Turkish public banks have been oper-
ating pretty much like their private counterparts under very professional
managers (nicknamed as ‘princes’ of Ozal), most of who transferred from
overseas. They have competed intensively with private banks for deposits,
money market funds, and loan originations. In 1993, the quasi-fiscal opera-
tions of the state banks (Sumerbank, and Etibank) have been disentangled
from their banking operations. Reorganization, downsizing, the introduction
of new products and services, internal training programs to enhance the quality
of the human capital (like the Banking and Insurance School), and operating
increasingly in international markets were common attributes of the state
banks during this era. As a reflection of these policies, the state banks’ average
profitability has exceeded those of the private banks in the early 1990s. Their
performance was also remarkable internationally. For instance, one public
bank’s CEO has been given the ‘‘Bank Chairman of the Year’’ award by the
Trade Finance in August 1991, one of the leading financial periodicals in the
world. Moreover, while the state-owned T.C. Ziraat Bank ranked 452nd in
the Euromoney’s 1988 top 500 banks list, its ranking improved to 202nd in
1993. In the same list, it was the number one bank according to ROE, the 41st
according to net profits, and the 12th according to growth in 1993 worldwide.
However, under political influences, the scene began to change for the state
banks, especially upon displacement of the so-called ‘‘princes’’ by the mid-
1990s. State banks became over-employed and registered substantial loan
write-offs and considerable operating losses driven particularly by ‘‘duty los-
ses’’ incurred as a result of subsidized lending to favored sectors. Thus, their
profitability, liquidity, and capitalization had been undermined until a capital
injection in 1997 (IMF Staff Report). Many private banks extend loans to
special sectors, too, but state banks are dominant in this area. It can be argued
that the cost of making and collecting loans to favored sectors could be
significantly lower than in the private sector due to less intensive screening and
collection. Alternatively, the costs could be higher if the banks are obligated to
make more small loans. In order to alleviate the potential biases against the
state banks, we accounted for directed lending (loans to special sectors) by
including them in the construction of efficiency scores as mentioned earlier.
In the literature, there are a number of theories trying to explain the be-
haviors of private and public enterprises such as property rights theory, agency
cost theory, transaction costs theory, contract theory, among others (Wei et al.,
2000). All these theories suggest that publicly owned firms should perform less
efficiently and less profitably than private firms. There are two major reasons
behind the efficiency difference between public and private firms. The first is the
difference in objectives: while all private firms are profit maximizing, public
firms do not necessarily pursue profits. Public firms would pursue whatever
objectives the government demands or the controlling politician wants, such as
hiring a large number of redundant workers in the public sector in order to
promote politicians’ welfare (Boycko et al., 1996). The second is the difference
in budget constraints: private firms are subject to a relatively ‘‘hard’’ budget
constraint while public firms are subject to ‘‘soft’’ budget constraints. This
would necessarily lead private investment to more efficient places. In contrast,
the public sector would direct investment funds to the most politically desirable
projects. In light of these propositions, we tested the significance of the effi-
ciency differentials between the two ownership types, including and excluding
foreign private banks over the years under study (1988–1996).
Our empirical results suggest that national private banks (NAT_PRIVT) are
more cost and profit efficient than state banks (NAT_STATE), consistent with
expectations. When the group of private banks is defined in a way that it would
contain both domestic and foreign private banks (PRIVT_BNKS), all three
studies on Turkish banking efficiency imply that the banks in private hands are
more cost efficient than banks in public sector. 35 In addition, our results
suggest that private banks (PRIVT_BNKS) seem to have higher TE, PTE and
SE but lower AE than the state banks (STATE_BNKS). Of the efficiency
differentials between the two types, those in AE and SE are statistically more
significant. Labor intensive policies in public banks appears to be technically
cost inefficient but allocatively efficient, given that public workers cost about
three times less than private workers. 36
35
Unlike Zaim (1995) but like us Altunbas et al. (1994a,b,c) also reported that public banks are
less cost efficient than domestic private banks, but Altunbas et al. (1994a,b,c) found that the
efficiency difference between the two groups are not statistically significant. In India state banks
were found to be more efficient than private banks (Bhattacharya et al., 1997).
36
Allocative efficiency refers to optimal decision to choose the right mix of inputs given the
prices. Employing more labor given its very low price and less capital given its high price is
allocatively efficient decision, at least in the short run.
Private banks suffer from under-utilization of resources, too, but not as

severely as the state banks. The main source of technical inefficiency in both
state and private banks appears to be scale inefficiency rather than pure
technical inefficiency. However, the scale inefficiency is more acute in the case
of state banks. Public banks include some quasi-fiscal operations. They act as
trustees for the financial assets of local provinces and municipalities, a ‘free’
source of funds, and as a collection and payment agency of the state. This is a
costly duty that requires extra branches and labor. Overall our results suggest
that such activities create distortions for state banks that contribute more to
their size, thus higher scale inefficiencies, and more to a high cost structure,
thus lower profit and input efficiencies.
4.2. Marginal impact of organizational variables on banking efficiency
Qualitative conclusions on pair-wise significance tests of efficiency differ-

ences between banking groups or solely on their point estimates are common in
the literature (e.g., Berg et al., 1991; Elyasiani and Mehdian, 1992; Grabowski
et al., 1993; Altunbas et al., 1994a; Altunbas et al., 1994b; Altunbas et al.,
1994c; Zaim, 1995; Bhattacharya et al., 1997; among others). Some of the key
differences observed, however, could be correlated with other factors. For in-
stance, the efficiency differences we find between private and public banks
might be a result of size, not ownership per se, because all state banks are either
medium or large sized.
Additionally, cost (input) efficiencies of banks whose asset portfolios are
tilted towards loans might be lower simply because of the fact that operating
costs related to originating, maintaining and monitoring loans are much
greater than those of investment securities. Foreign banks, which depend
heavily on brokered deposits, could be found less efficient than domestic banks
because purchased deposits are more expensive and unstable than core de-
posits. Alternatively, foreign banks could be found to be more efficient than
domestic banks because reliance on purchased funds does not require invest-
ment in expensive branch networking. Private banks could be more profit ef-
ficient than public banks because private banks can involve in highly risky, but
also highly profitable, activities but public banks by law cannot do so. Like-
wise, small banks could be less efficient than large banks if small banks are new
and large banks are old and there are benefits resulting from ‘‘learning by
doing’’ and goodwill, which comes with experience and age. It is important to
measure efficiency without controlling for certain factors because the ultimate
goal in the second-stage regression analysis is to test how efficiency is related to
these factors by treating them as potential correlates of efficiency.
‘Net impact’ of each factor (organizational form) on the average efficiency
could be detected if the impact of other possible factors could be controlled as
in a regression analysis (Mester, 1993; Berger and Mester, 1997; Chang et al.,
1998; Hasan and Marton, 2000), which we endeavor to address next. Following
Saxonhouse (1976) and Mester (1993), we employ generalized least squares
(GLS) regressions to control for heteroscedasticity. Table 7 presents the results
from seven regressions, one for each efficiency estimate, along with model
statistics and definitions of variables. We have seven dependent variables, of
which five are non-parametric [CE, AE, TE, PTE, and SE] and two are
parametric [CostE1, ProfitE] efficiency scores, and thirteen independent vari-
ables (excluding the intercept). Our focus is on the correlation of efficiency with
organizational forms (core variables) [CEO_CHA, PUBLCLY_HELD,
NAT_PRIVT (excluded from the regressions as the base case), NAT_STATE,
FOR_FOU_TR, and FOR_BRN_TR] on which we have complete data over
the respective years under study. For the reasons cited above, we use size
[SMALL (excluded from regressions as the base), MEDIUM, LARGE], time
[YEAR_88 (excluded from regressions as the base), YEAR_92, YEAR_96],
concentration of total assets on loans [LOAN/TA], the degree to which total
assets are financed by brokered (purchased) funds [BROKERED FUND/TA],
variability of bank’s return on equity over time [RISK: r (ROE)] and bank’s
age (AGE) as control (peripheral) variables.
As the table shows, even after controlling for bank-specific attributes, the
results from regression analyses support, to a great extent, our earlier results
from the pair-wise significance tests although a few relationships lost their
significance. For example, although the coefficient of publicly traded banks
(PUBLCLY_HELD) dummy is positive in all regressions, it is significant at the
10% level only in ‘‘pure’’ technical efficiency. However, consistent with earlier
results, the banks that separate decision making from decision control by
having different CEOs and chairmen of board (NON_CEO_CHA) are found
to be significantly more input efficient than the banks that fail to separate
(CEO_CHA), regardless of the method and efficiency concept. Results also
show that level of input efficiency in foreign private banks (both subsidiaries
and branches) like in public banks is greater than the level of efficiency in
domestic private banks, which is again in line with our earlier findings. Another
indication from the results is that all forms of banks are equally profit efficient,
perhaps implying no potential earnings biases favoring any organizational
form. Banking is a very profitable business in Turkey and all types of banks
show this result equally.
As for size, the results show that all sizes of banks could be equally cost and
profit efficient. However, small banks are much more scale efficient than both
large and medium size banks, implying that there are no economies to be ex-
ploited by expanding scale of operations. This supports our earlier finding that
Turkish banks are oversized on average and suffers from diseconomies of scale.
The deterioration of efficiency (input and profit) over time is also evident from
significantly negative coefficients of the years 1992 and 1996. Contrary to ex-
pectations, concentration of bank assets on loans (LOAN/TA), reliance on
Table 7
Correlates of the nonparametric and parametric efficiency scoresa
Techniques Nonparametric frontier (DEA) Parametric frontier
Independent/dependent CE AE TE PTE SE CostE1 Profit E

variables
INTERCEPT 0.383 0.904 0.533 0.770 0.780 0.925 0.930
(5.287) (15.997) (8.056) (17.648) (17.196) (56.685) (23.234)
Organizational forms:b
CEO_CHA )0.097 )0.043 )0.077 )0.098 )0.057 )0.046 )0.043
()1.934) (1.824) ()1.532) ()2.372) ()1.854) ()1.684) ()0.820)
PUBLCLY_HELD 0.056 0.036 0.035 0.038 0.004 0.006 0.022
(1.173) (1.292) (0.895) (1.628) (0.128) (0.259) (0.456)
NAT_STATE 0.155 0.833 0.122 0.074 0.107 0.032 )0.004
(2.356) (0.031) (1.816) (1.356) (2.242) (0.903) ()0.050)
FOR_FOU_TR 0.121 0.049 0.512 0.053 0.015 0.013 0.020
(2.550) (1.826) (1.079) (1.603) (0.570) (0.688) (0.543)
FOR_BRN_TR 0.119 0.080 0.092 0.188 0.020 0.049 0.073
(1.643) (1.734) (1.347) (3.772) (0.425) (1.648) (1.257)
Control variables:c
Size variables:
MEDIUM 0.030 0.013 )0.012 0.042 )0.085 )0.013 )0.004
(0.660) (0.461) ()0.314) (1.462) ()3.080) ()0.651) ()0.097)
LARGE )0.023 )0.0003 )0.074 0.073 )0.182 )0.200 )0.039
()0.397) ()0.010) ()1.411) (2.135) ()5.131) ()0.877) ()0.741)
Stability variables:
YEAR_92 )0.186 )0.054 )0.140 )0.080 )0.091 )0.006 0.018
()5.311) ()2.471) ()4.311) ()3.482) ()3.454) ()0.775) (1.001)
YEAR_96 )0.136 )0.102 )0.045 )0.055 )0.036 )0.225 )0.073
()3.444) ()3.862) ()1.226) ()2.306) ()1.397) ()13.716) ()1.950)
Other bank variables:
LOAN/TA 0.918 0.242 0.772 0.324 0.380 0.093 )0.112
(5.615) (2.156) (5.583) (3.700) (4.297) (1.950) ()0.946)
759
(continued on next page)
Table 7 (continued )
760
Techniques Nonparametric frontier (DEA) Parametric frontier
Independent/dependent CE AE TE PTE SE CostE1 Profit E
variables

BROKERED FUND/TA 0.292 0.014 0.284 0.115 0.217 0.055 )0.101
(2.681) (0.172) (3.212) (2.116) (3.226) (1.906) ()1.378)
RISK: r(ROE) 0.186 0.039 0.106 0.073 0.069 0.034 0.022
(2.306) (1.148) (1.608) (1.795) (1.547) (1.688) (0.428)
AGE )0.0007 )0.0004 )0.0005 )0.0002 )0. 0001 )0.00007 )0.0002
()1.317) ()1.003) ()0.955) ()0.821) ()0.275) ()0.254) ()0.291)
Model statistics
# of observations 139 139 139 139 139 139 139
Adjusted R-squared 0.427 0.273 0.392 0.327 0.467 0.736 0.124
F-statistics 7.179 4.185 6.198 3.817 6.845 20.234 1.176
DW-statistics 1.833 1.921 2.138 2.145 2.050 2.179 1.781
a
CE: Cost efficiency; AE: Allocative efficiency; TE: Technical efficiency; PTE: Pure technical efficiency; SE: Scale efficiency, CostE1: Stochastic cost
efficiency; ProfitE: Stochastic alternative profit efficiency. The columns report multivariate Generalized least squares (GLS) regression coefficients (with
t-statistics in parentheses).
b
Organizational forms.: CEO_CHA, the banks whose chief executive officer is also the chairman of the board; PUBLCLY_HELD, the banks whose
shares are publicly traded in the Istanbul Stock Exchange (ISE); NAT_STATE, public banks that are owned predominantly by the domestic taxpayers
and voters; NAT_PRIVT, private banks whose more than 50% of shares is owned by the Turkish residents (base group); FOR_FOU_TR, foreign
banks founded in Turkey whose more than 50% of shares is owned by the residents of foreign countries; FOR_BRN_TR, foreign banks’ branches
operating in Turkey.
c
Control variables: Size variables: SMALL banks: assets <100; MEDIUM banks: 100 6 assets < 1000; LARGE banks: assets P 1000 ($ million);
Stability variables: YEAR_88: 1 if the efficiency score belongs to 1988 (base) YEAR_92: 1 if the efficiency score belongs to 1992; YEAR_96: 1 if the
efficiency score belongs to 1996; Other bank variables: LOAN/TA: total loans (short term, long term, and specialized loans) divided by total assets;
BROKERED FUND/TA: brokered (non-deposit) funds divided by total assets; RISK: standard deviation over time of the bank’s annual return on
equity; AGE: the time that has passed since the foundation before 1988, or 1992, or 1996.

Significance at 10% confidence level.


purchased funds (BROKERED FUND/TA), and variable returns (RISK: r

(ROE)) are significantly positively related to input efficiency. DeYoung (1998)
demonstrated that there is a strong association between management quality
and efficiency. Thus, it may be that poor managers are also poor in operations
and passive in assuming higher risks and higher profits. Higher loan and
brokered fund involvement can be actually considered proxies for risky oper-
ations. If there is a trade-off between risk and return, pursuit of higher profits
may often dictate investing in riskier assets like loans and funding them with
riskier funds like brokered deposits. After all, the cost of attaining riskier
ventures is substantially ‘subsidized’ in Turkey by full protection of individual
deposits introduced after the 1994 crisis, resulting in moral hazard incentives
for bank owners. Finally, bank age does not seem to be correlated with effi-
ciency.
The results from the two rival methods of estimating efficiency, parametric
and non-parametric frontiers, are generally in agreement in direction and
generally statistically significant. Parametric input efficiency (CostE1) generally
supports all the major relationships suggested by the non-parametric input
efficiencies (CE, AE, TE, PTE, and SE). However, we cannot state the same for
ProfitE. 37 Coefficients of determination (R2 s) for all models except for ProfitE
are relatively high and also statistically significant, indicating that our variables
are jointly able to explain a significant portion of the variations in input effi-
ciencies but not in profit efficiency. Due to potential multicollinearity between
organizational structure and size, we ran the same regression without having
both variables simultaneously in the regression equation. The basic results
prevailed in this robustness check and these results are available upon request.
5. Concluding remarks
Employing both parametric and non-parametric methods, we estimate the

cost and profit efficiencies of Turkish banks over the 1988–1996 period to study
the impact of different ownership and organizational structures on efficiency of
Turkish banking sector. Over the years under study, we find that the overall
cost and profit efficiencies for the Turkish banks are 72% and 83%, respectively,
implying that on average, about 40% of the bank resources and about 20% of
the potential bank profits are wasted during the production of banking ser-
vices. The results also indicate that the production efficiencies of the industry
consistently fell over time, which seems to have resulted from increases in the
37
It is also evident from the fact that CostE1 is significantly correlated with nonstochastic cost
efficiencies (qCostE1;CE ¼ 0:55, qCostE1AE ¼ 0:32, qCostE1TE ¼ 0:57, qCostE1PTE ¼ 0:46, qCostE1SE ¼
0:39) but insignificantly so with ProfitE (qCostE1;ProfitE ¼ 0:19).
cost of funding and growth of the banks in the recent years. The mean effi-
ciency measures that we find for the industry in general, and for the private
banks in particular, are higher than those reported in earlier studies on Turkish
bank efficiency. We attribute this finding mainly to the inclusion of the omitted
factors in other studies such as off-balance sheet activities, repo transactions,
inter-bank funds, and lending to special sectors.
We also decompose the overall cost efficiency into its allocative and tech-
nical efficiency components. The results suggest that the dominant source of
the cost inefficiency in the Turkish commercial banking sector is due to tech-
nical inefficiency rather than allocative inefficiency. While regulation is typi-
cally given as a major source of allocative inefficiency, poor management is
usually associated with technical inefficiency. Relatively higher technical effi-
ciency implies that the competitive viability of the Turkish commercial banks
depends, to a great extent, upon improved managerial efficiency. However,
further analysis suggests that most of the observed technical inefficiencies are
due to operating at incorrect scales rather than operating off the efficient
frontier. To the extent that it involves the choice of an inefficient level, scale
inefficiency can be considered a form of managerial inefficiency.
The period of our study corresponds to the period of revival of growth in
commercial banking that seems to have encouraged the over-branching and
over-staffing problems of the pre-1980s. Accordingly, the results show that the
Turkish banks have increasingly experienced diseconomies of scale and most of
the scale inefficient banks suffer from ‘‘transgressing’’ the optimum size. Con-
sequently, the banks in our sample tend to incur nontrivial production costs
due to scale inefficiency, which seem to be associated more with being at a size
greater than the optimum. Although efficiency gains could be realized by re-
ducing production levels, the excess demand for banking services, induced by
the state’s high growth polices in recent years, does not justify it. The oli-
gopolistic structure of the Turkish banking industry and the increased demand
for banking services have sustained managerial inefficiency, despite the reform
agenda undertaken in the 1980s to achieve greater competition in the Turkish
economy.
A second-stage correlate of efficiency analysis suggests that the relationship
between bank size and efficiency is strongly negative. While small banks
compete with large banks primarily in metropolitan markets, they do not
compete in rural markets. Thus, strong competition might have induced more
market discipline on small banks, leading to greater cost efficiency. Unlike their
counterparts in the US, foreign banks operating in Turkey seem to be signif-
icantly more efficient than their domestic peers. The local market conditions in
Turkey provide some opportunities for foreign banks to utilize their compar-
ative advantages in addition to their relatively minor scale problems, which
appears to be a major problem for the domestic banks. In general, private
banks are found to be more efficient than public banks in terms of all types of
efficiency. We also find a strong association between X -efficiency and man-

agement-team structure. Our results indicate that banks where the board is
independent of management are significantly more efficient than those banks
with the same management and board. In addition, publicly traded banks,
which are exposed to more market discipline, are found to be more technically
efficient than privately held banks. Banks operating under a holding company
structure outperform the independent banks in terms of efficiency. Although
multinational domestic banks seem to be more efficient than pure domestic
banks, their efficiency differential does not appear to be significant.
Acknowledgements
We would like to thank our anonymous referee for comments and sugges-
tions that have substantially improved the quality of this paper. Special thanks
go to Iftekhar Hassan for his valuable technical and moral support. We also
wish to thank Ebru Isik and Ugur Meleke for their laborious efforts during
collecting and processing data. The remaining errors are ours.
References
Aigner, D.J., Lovell, C.A.K., Schmidt, P., 1977. Formulation and estimation of stochastic frontier
production function models. Journal of Econometrics 6, 21–37.
Aliber, R.Z., 1976. Toward theory of international banking. Economic Review, Federal Reserve
Bank of San Francisco, pp. 5–8.
Altunbas, Y., Evans, L., Molyneux, P., 1994a. Universal banks, ownership and efficiency – A
stochastic frontier analysis of the German banking market, Unpublished working paper,
Institute of European Finance.
Altunbas, Y., Molyneux, P., Di Salvo, R., 1994b. Inefficiency in the Italian credit cooperative bank
sector – A stochastic frontier analysis, Unpublished working paper, Institute of European
Finance.
Altunbas, Y., Molyneux, P., Murphy, N., 1994c. Privatization, efficiency and public ownership in
Turkey – An analysis of the banking industry 1991–1993, Unpublished working paper, Institute
of European Finance.
Aly, H.Y., Grabowski, R., Pasurka, C., Rangan, N., 1990. Technical, scale, and allocative
efficiencies in US banking: An empirical investigation. Review of Economics and Statistics
72, 211–218.
Bauer, P.W., Berger, A.N., Ferrier, G.D., Humphrey, D.B., 1998. Consistency conditions for
regulatory analysis of financial institutions: A comparison of frontier efficiency methods.
Journal of Economics and Business 50, 85–114.
Bauer, P.W., Berger, A.N., Humphrey, D.B., 1993. Efficiency and productivity growth in US
banking. In: Fried, H.O., Lovell, C.A.K., Schmidt, S.S. (Eds.), The Measurement of Productive
Efficiency: Techniques and Applications. Oxford University Press, Oxford, pp. 386–413.
Berg, S.A., Forsund, F., Jansen, E., 1991. Technical efficiency of Norwegian banks: A
nonparametric approach to efficiency measurement. Journal of Productivity Analysis 2,
127–142.
Berger, A.N., Humphrey, D.B., 1991. The dominance of inefficiencies over scale, product mix
economies in banking. Journal of Monetary Economics 28, 117–148.
Berger, A.N., Hancock, D., Humphrey, D.B., 1993. Bank efficiency derived from the profit
function. Journal of Banking and Finance 17, 317–347.
Berger, A.N., Hunter, W.C., Timme, S.G., 1993. The efficiency of financial institutions: A
review and preview of research past, present, and future. Journal of Banking and Finance
17, 221–249.
Berger, A.N., Humphrey, D.B., 1997. Efficiency of financial institutions: International survey and
directions for future research. European Journal of Operational Research 98, 175–212.
Berger, A.N., Mester, L.J., 1997. Inside the black box: What explains differences in the efficiencies
of financial institutions. Journal of Banking and Finance 21, 895–947.
Bhattacharya, A., Lovell, C.A.K., Sahay, P., 1997. The impact of liberalization on the productive
efficiency of Indian commercial banks. European Journal of Operational Research 98, 332–345.
Byrnes, P., 1985. Ownership and efficiency in the water supply industry: An application of the
non-parametric programming approach to efficiency measurement. Unpublished dissertation.
Southern Illinois University.
Boycko, M., Shleifer, A., Vishny, R.W., 1996. A theory of privatization. The Economic Journal
106, 309–319.
Celasun, O., 1998. The 1994 currency crisis in Turkey. Working paper no: 1913, World Bank
Development Research Group.
Caves, R., 1971. International corporations: The industrial economics of foreign investment.
Econometrica 38, 1–27.
Chang, C.E., Hasan, I., Hunter, W.C., 1998. Efficiency of multinational banks: An empirical
investigation. Applied Financial Economics 8, 1–8.
Charnes, A., Cooper, W.W., Lewin, A.Y., Seiford, L.M., 1994. Data Envelopment Analysis:
Theory, Methodology and Applications. Kluwer, Dordrecht.
Clark, J.A., 1988. Economies of scale and scope at depository financial institutions: A review of the
literature. Federal Reserve Bank of Kansas City, Economic Review, pp. 16–33.
Denizer, C., 1997. The effects of financial liberalization and new bank entry on market structure
and competition in Turkey. Working paper no. 1839, World Bank Development Research
Group, November.
DeYoung, R., 1998. Management quality and X-inefficiency in national banks. Journal of
Financial Services Research 13, 5–22.
DeYoung, R., Hasan, I., 1998. The performance of de novo commercial banks: A profit efficiency
approach. Journal of Banking and Finance 22, 565–587.
DeYoung, R., Nolle, D., 1998. Foreign-owned banks in the US: Earning market share or buying it?
Journal of Money, Credit, and Banking 28, 622–636.
Dunning, J., 1977. Trade, location of economic activity and the MNE: A search for an eclectic
approach. In: Ohlin, Heselborn, Wijkman (Eds.), The International Allocation of Economic
Activity. Holmes and Meier, New York, pp. 395–418.
Dunning, J., 1980. Toward an eclectic theory of international production: Some empirical tests.
Journal of International Business Studies 11, 9–31.
Eisenbeis, R.A., G.D. Ferrier, S.H. Kwan, April 1996. An empirical analysis of the informativeness
of programming and SFA efficiency scores: Efficiency and bank performance. Working paper,
University of North Carolina, Chapel Hill, NC.
Elyasiani, E., Mehdian, S., 1990. Efficiency in the commercial banking industry: A production
frontier approach. Applied Economics 22, 539–551.
Elyasiani, E., Mehdian, S., 1992. Productive efficiency performance of minority and nonminority-
owned banks: A non-parametric approach. Journal of Banking and Finance 16, 933–948.
Elyasiani, E., Mehdian, S.M., Rezvanian, R., 1994. An empirical test of association between
production and financial performance. Applied Financial Economics 4, 55–59.
Fields, J.A., Murphy, N.B., Tirtiroglu, D., 1993. An international comparison of scale economies
in banking: Evidence from Turkey. Journal of Financial Services Research 17, 111–125.
Fried, H.O., Lovell, C.A.K., Schmidt, S.S., 1993. The Measurement of Productive Efficiency:
Techniques and Applications. Oxford University Press, Oxford.
Fry, M., 1979. Money and Banking in Turkey. Bogazici University Publications, Istanbul, Turkey.
Fukuyama, H., 1993. Technical and scale efficiency of Japanese commercial banks: A non-
parametric approach. Applied Economics 25, 1101–1112.
Grabowski, R., Rangan, N., Rezvanian, R., 1993. Organizational forms in banking: An empirical
investigation of cost efficiency. Journal of Banking and Finance 17, 531–538.
Gray, J.M., Gray, H.P., 1981. The multinational bank: A financial MNC. Journal of Banking and
Finance 5, 33–63.
Hasan, I., Marton, K., 2000. Development and efficiency of the banking sector in a transitional
economy: Hungarian experience, Working paper, Stern School of Business, New York
University.
Hermalin, B.E., Wallace, N.E., 1994. The determinants of efficiency and solvency in savings and
loans. Rand Journal of Economics 25, 361–381.
Humphrey, D.B., 1985. Cost and scale economies in bank intermediation. In: Aspinwall, R.,
Eisenbeis, R. (Eds.), Handbook for Banking Strategy. Wiley & Sons, New York.
Hymer, S., 1976. The international operations of national firms: A study of direct foreign
investment. MIT Press, Cambridge, MA.
Isik, I., Hassan, M.K., 2000. The effect of the 1994 economic crisis on the productivity and
efficiency of Turkish banks, Working paper, School of Management, New Jersey Institute of
Technology.
Jagtiani, J., Nathan, A., Sick, G., 1995. Scale economies and cost complementarities in commercial
banks: On- and off-balance sheet activities. Journal of Banking and Finance 19, 1175–1189.
Kaparakis, E., Miller, S., Noulos, A., 1994. Short-run cost inefficiency of commercial banks: A
flexible stochastic frontier approach. Journal of Money, Credit, and Banking 26, 875–893.
Kindleberger, C.P., 1969. American Business Abroad: Six Lectures on Direct Investment. Yale
University Press, New Haven, CT.
Kwan, S.H., Eisenbeis, R., 1994. An analysis of inefficiencies in banking: A stochastic cost frontier
approach. Working paper.
Lees, F.A., 1974. International Banking and Finance. MacMillan, New York.
Mahajan, A., Rangan, N., Zardkoohi, A., 1996. Cost structures in multinational and domestic
banking. Journal of Banking and Finance 20, 238–306.
Mester, L.J., 1993. Efficiency in the savings and loan industry. Journal of Banking and Finance
17, 267–286.
Miller, S.M., Noulas, A.G., 1996. The technical efficiency of large bank production. Journal of
Banking and Finance 20, 495–509.
Newman, J., Shrieves, R., 1993. Multibank holding company effect on cost efficiency in banking.
Journal of Banking and Finance 17, 709–732.
Noulas, A.G., Ray, S.C., Miller, S.M., 1990. Returns to scale and input substitution for large US
banks. Journal of Money, Credit and Banking 22, 94–108.
Oral, M., Yolalan, R., 1990. An empirical study on measuring operating efficiency and profitability
of bank branches. European Journal of Operational Research 46, 282–294.
Pi, L., Timme, S.G., 1993. Corporate control and bank efficiency. Journal of Banking and Finance
17, 515–530.
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, 221–250.
Saunders, A., 1993. Panel discussion of the efficiency of financial institutions around the globe:
Comments on efficiency studies. Journal of Banking and Finance 17, 551–552.
Saxonhouse, G.R., 1976. Estimated parameters as dependent variables. The American Economic
Review 66, 178–183.
Sealey, C., Lindley, J., 1977. Inputs, outputs, and a theory of production and cost at depository
financial institutions. Journal of Finance 32, 1251–1266.
Shaffer, S., 1988. A revenue-restricted cost study of 100 large banks. Research paper no. 8806.
Federal Reserve Bank of New York.
Shaffer, S., David, E., 1986. Economies of superscale and interstate expansion. Research paper no.
8612. Federal Reserve Bank of New York.
Wei, Z., Varela, O., Hassan, M.K., 2000. Ownership and performance in chinese manufacturing
industry. Journal of Multinational Financial Management, forthcoming.
Zaim, O., 1995. The effect of financial liberalization on the efficiency of Turkish commercial banks.
Applied Financial Economics 5, 257–264.

2002 JBF Isik - Technical, Scale and Allocative Efficiencies of Turkish Banking Industry

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

2002 JBF Isik - Technical, Scale and Allocative Efficiencies of Turkish Banking Industry

Uploaded by

Copyright:

Available Formats

Journal of Banking & Finance 26 (2002) 719–766

Technical, scale and allocative eﬃciencies of

JEL classiﬁcation: G2; D2; G21; G28

In a rapidly changing ﬁnancial market worldwide, bank regulators, man-

oligopolistic competition. It would probably take over 1000 banks to account

2. Methodology and data

2.1. Empirical design for eﬃciency estimations

Rather than estimating a common frontier across time, we preferred to es-

In a performance analysis, production units are expected to be relatively

2.3. Deﬁnition of variables: Inputs, input prices, and outputs

‘Reliable’ eﬃciency prediction requires appropriate deﬁnitions and certain

commercial banks as multi-product ﬁrms, producing 4 outputs and employing

2.4. Sample statistics for the variables and size

3. The empirical results pertaining to eﬃciency estimates

3.1. Sources of ineﬃciency in general

3.2. Returns to scale in the Turkish commercial banking

3.3. Correlation of eﬃciency measures with ﬁnancial performance

In order to complement the results of eﬃciency measures, we correlate

Total 49 100 35 69 100 50 21 100 15 139 100 100

I. Isik, M.K. Hassan / Journal of Banking & Finance 26 (2002) 719–766

among the X -eﬃciency measures and proxy measures of performance. Gener-

4. Organizational structure and eﬃciency in the Turkish banking industry

Banks are not homogenous in terms of ownership structures and organi-

4.1. Comparison of the eﬃciency across organizational forms

Evidence to date suggests that ineﬃciencies on the output side may be as

perior to standard proﬁt eﬃciency in many respects, one of which is the

I. Isik, M.K. Hassan / Journal of Banking & Finance 26 (2002) 719–766

B. Corporate control variables

I. Isik, M.K. Hassan / Journal of Banking & Finance 26 (2002) 719–766

4.1.1. Size variables

4.1.2. Corporate control and governance variables

4.1.3. Holding aﬃliation variables

Holding company structure could result in complicated organizational

4.1.4. International presence variables

4.1.5. Nationality variables

4.1.6. Ownership variables

Private banks suﬀer from under-utilization of resources, too, but not as

4.2. Marginal impact of organizational variables on banking eﬃciency

Qualitative conclusions on pair-wise significance tests of efficiency differ-

I. Isik, M.K. Hassan / Journal of Banking & Finance 26 (2002) 719–766

I. Isik, M.K. Hassan / Journal of Banking & Finance 26 (2002) 719–766

purchased funds (BROKERED FUND/TA), and variable returns (RISK: r

Employing both parametric and non-parametric methods, we estimate the

efficiency. We also find a strong association between X -efficiency and man-

You might also like