This article was downloaded by: [National Chiao Tung University 國立交通 大學]

On: 27 April 2014, At: 17:05 Publisher: Routledge

Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

**The Service Industries**

**Journal**

Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/fsij20

**Cost efficiency and the effect**

**of mergers on the Taiwanese**

**banking industry**

Ya-Hui Peng assistant professor a & Kehluh Wang associate professor b

a

Institute of Business Administration , Hsuan Chuang University , 48 Hsuan Chuang Road, Hsinchu, Taiwan , 300

b

Graduate Institute of Finance , National Chiao Tung University , 1001 Ta Hsueh Road, Hsinchu, Taiwan , 30010

Published online: 25 Jan 2007.

**To cite this article: Ya-Hui Peng assistant professor & Kehluh Wang**

associate professor (2004) Cost efficiency and the effect of mergers on the Taiwanese banking industry, The Service Industries Journal, 24:4, 21-39, DOI: 10.1080/0264206042000275172

**To link to this article: http://dx.doi.org/10.1080/0264206042000275172**

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy,

completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions

## Cost Efficiency and the Effect of Mergers

## on the Taiwanese Banking Industry

### Y A - H U I P E N G a n d K E H L U H W A N G

This study addresses the cost efficiency, economies of scale and scope of the Taiwanese banking industry, specifically focusing on how bank mergers affect cost efficiency. Adopting stochastic frontier analysis, we employ a translog cost function for efficiency estimation. Composite error terms are used to account for managerial inefficiency and environmental effects. Empirical results suggest that economies of scale and scope exist at small and medium-sized banks. Meanwhile, government-owned or -controlled banks are the most cost efficient. Non-performing loans increase the inefficiency of the banking sector by just

under 10 per cent. Further analysis reveals that bank merger

activity is positively related to cost efficiency. Mergers can enhance cost efficiency, even though the number of bank employees does not decline. The banks involved in mergers are generally small and were established after the banking sector was deregulated.

INTRODUCTION

The banking industry in Taiwan is highly regulated, and new entrants were prohibited until the Commercial Bank Establishment Promotion Decree was implemented in 1991. In 1990, Taiwan had 24 banks with 953 branches, some government owned and operated. Over 16 commercial banks were established in 1991 and 1992. By 1996, the total number of banks had reached 42, with 1,936 branch offices. The entry of new competitors,

Ya-Hui Peng is assistant professor at the Institute of Business Administration at Hsuan Chuang University, 48 Hsuan Chuang Road, Hsinchu, Taiwan 300. Kehluh Wang is associate professor at the Graduate Institute of Finance of National Chiao Tung University, 1001 Ta Hsueh Road, Hsinchu, Taiwan 30010.

The Service Industries Journal, Vol.24, No.4, July 2004, pp.21–39 ISSN 0264-2069 print=1743-9507 online

*DOI: 10.1080=0264206042000275172 # 2004 Taylor & Francis Ltd.*

combined with internationalisation and market liberalisation, has revolutio-nised the banking industry. One result of the changes has been increasing competition that reduces the quality of loan portfolios. According to a report by the Ministry of Finance, the interest spread declined from 3.05 per cent in 1991 to 2.76 per cent in 1996. Over the same period, the non-performing loan ratio increased from 0.93 to 3.68 per cent. Furthermore, the mean return on equity and return on assets in the banking industry drastically declined from 28.89 and 1.2 per cent in 1990 to 9.7 and 0.7 per cent, respectively, in 1996.

The Taiwanese government is encouraging mergers and acquisitions to solve the problems in the banking industry. Particularly, larger commercial banks have been persuaded to take over small credit institutions. From 1997 to 1999, 16 mergers occurred. The Bank Merger Act and the Bank Holding Company Act were announced in 2000 and 2001 to further facilitate transactions.

This study investigates the cost efficiency of various types of banks in Taiwan, and seeks to determine whether mergers and acquisitions among banking firms can improve productivity. Meanwhile, the impact of non-performing loans is considered as well. The operating efficiency of banks is crucial in a sound economic system, and mergers and acquisitions are believed to be one way to improve it. In addition to examining whether merged banks are more cost efficient as expected by the government, this study also seeks to determine whether newly established banks are more efficient than older banks, or vice versa, and discusses its implementation to merging activity. Results of this study are important for bank managers, investors, policy makers and multinational banks interested in acquiring local banks.

Many bank mergers have occurred over recent years, stimulating consider-able academic interest. Prager and Hannan [1999] show that deposit rates fall at banks involved in mergers that increase market concentration. However, the results are inconclusive for mergers that do not significantly change market concentration [Simons and Stavins, 1998]. For Taiwanese mergers, while market share usually increases after mergers, no effect on pricing has been

observed [Chen and Chen, 2002]. Cornett and

Tehranian [1992] and Rhoades [1998] report an improvement in both bank profitability and market value, although other investigators do not [Berger and Humphrey, 1992; Akhavein, Berger and Humphrey, 1997; Hannan and Wolken, 1989; Pilloff, 1996].

Meanwhile, the impact of mergers on bank efficiency has also been dis-cussed substantially in the literature. The empirical results reveal little or no efficiency improvement for US mergers in the 1980s. Berger and Humphrey [1992] examine 60 large mergers in the 1980s and find no efficiency improve-ment. Pilloff [1994] studies 48 mergers from 1982 to 1991 and finds that the

value-weighted abnormal return and efficiency change are small. DeYoung [1997] finds that mergers between equally sized banks yield smaller-than-average cost efficiency improvements.

However, the results for mergers in the 1990s are mixed. Rhoades [1998] studied mergers of large US institutions and found efficiency gains in most cases. Resti [1998] analyzed 67 Italian bank mergers and found that mergers of equally sized banks yield substantial efficiency gains. Berger [1998] reported that if the participating banks are less efficient than their peers prior to consolidation, then substantial efficiency gains are pre-dicted. His result holds for both large and small banks. Lang and Welzel [1999] considered the cost effects of 283 small-scale mergers among German cooperative banks. Positive economies of scale and scope are realised only when merged banks close some branches. German cooperative bank mergers show no evidence of efficiency gains. Vennet [1996] studied 500 takeovers among European financial institutions and found that merger gains depend on the characteristics of the deal. Cross-border acquisitions and domestic mergers of equally sized banks generate significant cost effi-ciency improvement. Evidence from mergers of Australian trading banks between 1986 and 1995 proves that acquiring banks are more efficient than target banks [Avkiran, 1999].

In addition to cost efficiency, some studies have addressed the influence of loan quality on bank efficiency measurement. Bernstein [1996] considered the loan quality effect while estimating the translog cost function. He found that banks with poorer loan quality have higher costs, but the direct influence is small. Berger and DeYoung [1997] also review the loan quality problem, and consider the intertemporal relationship between loan quality and cost effi-ciency. Their results are ambiguous on the question of whether problem loans should be considered in estimates of efficiency.

The Commercial Bank Establishment Promotion Decree of 1991 dramati-cally altered the market structure of the Taiwanese banking industry. Chen and Yeh [2000] employed a non-parametric approach to measure the relative oper-ating efficiency of 34 Taiwanese commercial banks. Notably, they found that government-owned banks are less efficient than other banks, with a slightly higher Malmquist Index. Huang and Huang [2002] formulated a behavioural model under uncertainty to estimate total factor productivity in the Taiwanese banking industry. They showed no significant improvement in either total factor productivity or cost efficiency. Interestingly though, their result reveals that government-owned banks are more cost efficient than other banks. Since the problem of low quality loans is exacerbated by market com-petition, Li, Hu and Liu [2002] used the input distance function approach to elucidate the effect of non-performing loans on efficiency. Ou, Lee and

Young [2002] also tried to determine the relationship between bank asset quality and operating performance.

This study first examines the empirical measurement of cost efficiency in the Taiwanese banking industry. The stochastic frontier model and a translog cost function are used to estimate cost structure and cost efficiency. The model also takes into account the impact of low-quality loans on output measurement. Unlike the estimates from the model of Bernstein [1996], this approach allows the direct impact of low-quality loans on costs to be esti-mated. Second, this study seeks to clarify the efficiency, overall economies of scale and scope of different bank sizes and organisational types. Analytical results indicate that the size of a bank affects economies of scale and scope and that government-owned or -controlled banks enjoy greatest cost efficiency. Finally, the relationship between cost efficiency and merger activity is exam-ined. Results of this study reveal that bank mergers significantly improve cost efficiency. This finding is consistent with the findings of Shaffer [1993], Vennet [1996] and Akhavein, Berger and Humphrey [1997].

The remainder of this article is organised as follows. The next section formulates the shadow cost frontier that applies a translog function for efficiency estimation. We then present and analyse empirical results. A final section draws conclusions.

METHODOLOGY

Model Specification

There are several methods to study the efficiency and performance of com-mercial banks. The ratio approach uses financial indicators of the banking

industry to evaluate production efficiency via factor analysis,

one-way ANOVA, correlation analysis and cluster analysis. The nonpara-metric programming approach employs a mathematical programming model to measure the technical efficiency frontier. The parametric approach is based on the production or cost function. The advantage of this approach is that it can include a stochastic error term to account for environmental uncer-tainties. However, it needs to choose an explicit production or cost function with strong distributional assumptions on the error term. Many studies have focused on estimating the cost frontier based on various assumptions concern-ing the error term. [Cebenoyan, Cooperman and Register, 1993; Kaparakis, Miller and Noulas, 1994].

This study employs the parametric method with the shadow cost frontier model to measure the operating efficiency of Taiwanese banking firms. The intermediation approach is used for the bank production process. Banking firms are assumed to transform deposits, raw materials or intermediate pro-ducts into loans and investments as the outputs of the production system.

Since low-quality loans incur increased labour and administrative costs, loan output is quality adjusted. Total production cost comprises interest expenses on deposits and other operating costs of labour and capital. Consequently, the bank production process is assumed to involve a transformation of inputs (capital, labour and deposits) into outputs (loans and investments).

The dual cost function can be represented as C ¼ C(Yi, Pj) where Yi is the

ith output, i ¼ 1, 2, and Pjis the price of input j, j ¼ 1, 2, 3.

In the model, outputs are measured in terms of the dollar value of the

earning assets at the end of the fiscal year. Y1represents the loans and Y2

the investments. Moreover, P1 is the price of capital measured by the

rentals on building, equipment and maintenance [Murray and White, 1983],

and P2denotes the price of the labour calculated as the total salaries and

benef-its of each employee hour. Finally, P3 represents the total annual interest

expenses divided by average deposits and other borrowings.

Estimation of Cost Models

The shadow cost frontier approach assumes that all banks have the same underlying production frontier, which measures loans in terms of quality-adjusted units. Suppose a commercial bank produces an output vector from an input vector. The shadow cost approach postulates that various firm-specific production possibility frontiers can be pooled and represented by a

single common frontier that applies to the quality-adjusted outputs Y. The

unobserved quality-adjusted outputs Y _{are related to the observed outputs}

Y and a quality indicator ZQ. That is, Y ¼ Y(Y, ZQ). Here, the quality

indi-cator ZQdenotes the non-performing loan ratio. The shadow cost frontier is

defined as C_{¼ C}_{(Y}

i, Pj).

Following the stochastic frontier approach, we include a composite error term in the model. Consequently, the shadow cost frontier is represented as

C ¼ C(Y_{i}, Pj) þ 1 and 1 ¼ U þ V (1)

where C denotes the observed cost. 1, the composite error term, has two com-ponents, U and V. U is the neutral cost-augmenting inefficiency. Since the managerial or controllable inefficiency only increases costs above the cost frontier, U is assumed to be a one-sided error term. The three commonly assumed distributions of U are the normal truncated at zero, the half-normal truncated at a non-zero point and the exponential [Stevenson, 1980]. However, the estimates based on these various distributions are not very different [Cowing, Reifschneider and Stevenson, 1983; Greene, 1990; Mester, 1996]. Most studies have assumed U to be half-normal and truncated at zero [Mester, 1996; Huang, Fu and Huang, 1999; Hao, Hunter and Yang,

2001; Huang and Huang, 2002]. This study follows, i.e., U is from a normal

distribution with mean 0 and variance s2

U, but is truncated from below

at zero. V represents a two-sided random error, representing the fluctuations or uncontrollable factors that can either increase or decrease costs. Therefore,

V is assumed normally distributed with mean 0 and variances2

V. U and V are

distributed independently of each other [Huang, Fu and Huang, 1999]. Recent studies have suggested that the cost function of banking firms can be represented by a translog function [Hunter and Timme, 1986]. Moreover, the empirical translog model can be expressed as follows.

ln C ¼aþX 2 i¼1 biln Yiþ X3 j¼1 gjln Pjþ 1 2 X2 i¼1 X2 k¼1 bikln Yiln Y k þ1 2 X3 j¼1 X3 l¼1 gjlln Pjln Plþ X2 i¼1 X3 j¼1 rijln Yiln PjþU þ V (2)

The share equations are obtained from the partial derivatives of the above equation: Sj¼ @ ln C @ ln Pj ¼gjþ X3 l¼1 gjlln Plþ X2 i¼1 rijln Y i þWj j ¼ 1, 2, 3 (3)

where Wjare random error terms. The quality-adjusted loan output is defined as

ln Y_{1}¼(1 þd1ZQ) ln Y1 (4)

As non-performing loans are related with the loan outputs (Y1) only, the

invest-ment outputs (Y2) need not adjust. So ln Y2 ¼ln Y2. Since high quality loans

are less costly to produce than low quality loans, the coefficientd1is expected

to be positive. Homogeneity and symmetry restrictions are imposed on the estimate of the cost function parameters.

Measures of Cost Efficiency

The residuals 1i¼ UiþVican be estimated from the parameters of the

trans-log cost function. The variancess2

VandsU2can be calculated by the method of

moments [Olson, Schmidt and Waldman, 1980]:
^
s_{U}2 ¼ ﬃﬃﬃﬃﬃﬃﬃﬃm3
2=p
p
(4=p1)
2=3
(5)
^
s_{V}2 ¼m2 1
2
p
^
s_{U}2 (6)

where m2 and m3 represent the second and third central moments of the

residuals.

Jondrow et al. [1982] propose a method for estimating individual firm-specific inefficiency. This value can be defined as the conditional mean of

Uigiven the composite error 1i¼ UiþVI:

EkUij1il ¼miþs

f(mi/s)

F(m_{i}_{/}_{s}_{}_{)} (7)

wheremi¼1isU2/s2, ands2¼sU2þsV2.

f() and F() are the standard normal density function and the distribution

function, respectively. According to Jondrow et al. [1982], if the logarithmic

cost function is estimated, then the exponential of Uirepresents the cost

inef-ficiency. Battese and Coelli [1988] propose a method for estimating individual firm-specific efficiency, which can be expressed as follows.

EkeUi_{j1}
il ¼
F(mi/ss)
F(mi/s)
exp miþ
1
2s
2
(8)

The model herein follows the approach of Battese and Coelli [1988].

Furthermore, the 100(1 2a) per cent confidence interval for the individual

efficiency is further computed following the method of Bera and Sharma [1999]. The estimates are

Lower ¼miþF
1 a
2þ 1
a
2
F mi
s
s (9)
Upper ¼miþF
1 _{1 }a
2 1 F
mi
s
s (10)

The lower bound (LB) and upper bound (UB) of the confidence interval are,

LB ¼ exp (Upper) (11)

UB ¼ exp (Lower) (12)

Economies of Scale

The overall economy of scale measures the elasticity of the total cost with respect to an output vector. An overall economy of scale exists when the average or marginal costs associated with increasing output are progressively decreasing. It is measured as the inverse of the sum of the cost elasticities:

SE ¼ X @ ln C @ ln Y i 1 ¼ XEYi 1 (13)

where EYi denotes the cost elasticity of the ith output. Overall economies

(diseconomies) of scale exist if SE is greater (less) than one. Meanwhile, if SE equals one, constant returns to scale exist.

Economies of Scope

If a bank can produce two outputs together more cheaply than producing the same two outputs separately, then economies of scope exist. The relationship can be expressed as

C(Y1, 0) þ C(0, Y2) . C(Y1, Y2): (14)

Following Panzar and Willig [1981], economies of scope can be measured by,

SC ¼½C(Y1, 0) þ C(0, Y2) C(Y1, Y2)

C(Y1, Y2)

: (15)

Since the translog cost function cannot be used to estimate the cost when one or more outputs are zero, Huang, Fu and Huang [1999] present an alternative method for defining the economies of scope:

SC ¼½C(Y1Y m 1,Y2m) þ C(Y1m,Y2Y2m) C(Y1,Y2) C(Y1,Y2) (16) where Ym

1 and Y2mare the minimum values of Y1and Y2in the sample. The zero

value problem still exists for banks with minimum outputs Y1and Y2, so only

outputs that exceed the minimum values are considered here. Meanwhile, if SC is greater than zero, then overall economies of scope exist.

Cost Efficiency and Merger Activities

This study further examines the relationship between cost efficiency and

merger activity. Many variables impact the efficiency of

a bank [Mester, 1993; Kaparakis, Miller and Noulas, 1994; Hao, Hunter and

Yang, 2001]. This study employs a second-stage regression to identify the sources of cost efficiency:

eff ¼ f (MERGE, TA, TA2, GROWTH, BTD, ETA, TDTD, NINTOP) þ 1 (17) where

. _{eff is the cost efficiency obtained from Equation 8.}

. _{MERGE ¼ 1 for banks involved in merger activity, otherwise 0}

. _{TA ¼ total assets}

. _{TA2 ¼ square of TA}

. _{GROWTH ¼ growth rate of bank assets over the preceding year}

. _{BTD ¼ ratio of number of branches to total deposits}

. _{ETA ¼ ratio of number of employees to total assets}

. _{TDTD ¼ ratio of time deposits to total deposits}

. _{NINTOP ¼ ratio of non-interest income to operating profits}

Since the efficiency measure is bounded between 0 and 1, censored (Tobit) regression is used to estimate the parameters. The variable MERGE specifies the impact of merger activity on bank cost efficiency. Bank size may influence cost efficiency so the variable TA is included as a control variable for scale bias on efficiency. To clarify whether an optimal bank size exists for cost efficiency in banks, the square of TA, TA2, is also considered. GROWTH is a measure of the operating performance, and BTD represents the expense behaviour. ETA captures the impact of the size of the labour force on cost efficiency. All these variables may affect cost efficiency. Moreover, if a bank has a high percentage of time deposits, its funds are at lower costs. Therefore, the vari-able TDTD is used to measure the effect of this deposit mix on cost efficiency, and its parameter is expected to be positive. The variable NINTOP is a proxy for the output mix effect. Its impact on cost efficiency can be either positive or negative, depending on whether the bank generates more service-based revenues or more lending revenues as input costs increase.

DATA DESCRIPTION

The study sample comprises 44 banks with a range of sizes and organisational types. Panel data from 1997 to 1999 were obtained from the financial reports of these sample banks and from the Financial Statistics Abstract published by the Ministry of Finance. Sixteen bank mergers occurred during the sample period. Appendix 1 lists the sample banks.

Table 1 provides the descriptive statistics of the related variables and shows significant variation between the merged and non-merged banks, and

D E S C R I P T I V E S T A T I S T I C S O F S A M P L E B A N K S

Total assets Total costs Loans Investments

Price of capital Price of labour Price of funds Non-performing loans ratio Non-merged banks Mean 405,592 25,030 287,265 51,485 0.01913 0.00042 0.01732 0.05071 SD 414,760 24,067 293,281 53,498 0.01081 0.00011 0.00976 0.04000 Merged banks Mean 276,042 15,938 188,379 45,075 0.01501 0.00033 0.01419 0.05310 SD 501,407 26,981 319,946 109,985 0.00669 0.00011 0.00609 0.03393 Government-owned or -controlled banks Mean 946,685 55,807 672,544 119,184 0.01627 0.00051 0.01256 0.05001 SD 503,373 28,099 348,360 83,424 0.00866 0.00004 0.00403 0.02375

Old privately-owned banks

Mean 282,678 18,313 188,845 42,737 0.02152 0.00044 0.02050 0.07916

SD 189,206 11,424 111,121 39,085 0.01227 0.00013 0.01174 0.05879

New privately-owned banks

Mean 172,086 11,254 124,295 21,517 0.01856 0.00035 0.01752 0.03919

SD 101,657 7,538 65,785 16,911 0.01029 0.00008 0.00951 0.02712

Total banking firms

Mean 390,870 23,997 276,028 50,757 0.01866 0.00041 0.01696 0.05098

SD 425,323 24,476 296,812 61,897 0.01049 0.00011 0.00945 0.03924

Note: Total assets, costs, loans and investments are measured in millions of NT dollars.

across different organisational types. The merged banks are smaller, with average factor prices 20 per cent lower than those of the other banks. The majority of the merged banks are privately owned.

The government-owned or -controlled banks are relatively large in terms of total assets, while the new privately owned banks are much smaller. Con-sequently, the government-owned or -controlled banks dominate the banking industry in terms of loans and investments. The old privately owned banks have higher-than-average input prices. Notably, the government-owned or controlled banks have the highest labour costs, averaging 510 NT dollars per employee hour, compared to the industry average of just 410 NT dollars. However, the new privately owned banks face higher-than-average capital costs. Finally, the old privately owned banks have the poorest non-performing loan ratio of 7.9 per cent.

EMPIRICAL RESULTS

Parameter Estimates of the Cost Model

The cost system consists of the translog cost function and share equations. The seemingly unrelated regression method proposed by Zellner [1962] is used herein to estimate the parameters of the cost model. Appendix 2 lists the esti-mates of parameters in Equations 2 and 3. Most of the estimated parameters are

positive and significantly different from zero. The adjusted R2is 98 per cent.

Estimation of Cost Efficiency

The coefficient of the quality index,d1, is the focus of the stochastic shadow

cost frontier approach. This approach derives the distortion of the output cost

associated with output quality.d1is positive as expected. From Equation 4, the

relationship between cost distortion and the quality index can be further explored: Y 1 Y1 ¼Yd1ZQ 1 ¼ICD (18)

ICDrepresents the cost distortion as indicated by the quality index.

As shown in Table 2, the overall cost inefficiency due to non-performing loans is approximately 9.9 per cent of the total outstanding loans. The merged banks have a cost distortion 0.2 per cent lower than that of the non-merged banks. Mean-while, the cost inefficiency is greater for old privately owned banks, at about

15 per cent, significantly higher than the industry average. Sinced1ZQis less

than one, @ICD=@Y1 . 0 and @2ICD=@Y12, 0. Therefore, the cost of lower quality

loans increases at a decreasing rate with respect to the total amount of loans. Table 2 also summarises the cost efficiencies and confidence intervals across various types of banks. The merged banks are more cost-efficient,

M E A N C O S T D I S T O R T I O N A N D C O S T E F F I C I E N C Y B Y T Y P E S O F B A N K S

Bank types ICD Cost efficiency Upper bound Lower bound

Non-merged banks 1.0992 (0.0812) 0.9437 (0.0215) 0.9967 (0.0043) 0.8562 (0.0334) Merged banks 1.0977 (0.0642) 0.9475 (0.0168) 0.9975 (0.0016) 0.8615 (0.0295) Government-owned

or -controlled banks

1.1068 (0.0541) 0.9487 (0.0167) 0.9976 (0.0016) 0.8636 (0.0294) Old privately-owned banks 1.1537 (0.1186) 0.9344 (0.0216) 0.9957 (0.0042) 0.8410 (0.0321) New privately-owned banks 1.0716 (0.0509) 0.9462 (0.0216) 0.9969 (0.0048) 0.8604 (0.0331) Pooled sample 1.0991 (0.0792) 0.9441 (0.0210) 0.9968 (0.0041) 0.8568 (0.0329) Note: The sample standard deviations are in parentheses.

implying that merging affects cost efficiency. This relationship is further elu-cidated by the regression analysis. Furthermore, the old privately owned banks perform worst, while the government-owned or -controlled banks enjoy high cost efficiency. The differences in cost efficiency and cost distortion across organisational types are also examined using the Kruskal-Wallis test. The results are statistically significant, as shown in Table 3.

Economies of Scale and Scope

As shown in Table 4, most Taiwanese banks exhibit economies of scale and scope, regardless of the organisational types. This study further decomposes the samples into three size categories – small, medium and large. Table 4

indi-T A B L E 3

R E S U L T S O F T H E K R U S K A L - W A L L I S T E S T

Organisational types ICD Cost efficiency Government-owned

or -controlled banks

W1¼ 2,414, N1¼ 33 W1¼ 2,550, N1¼ 33 Old privately-owned banks W2¼ 1,402, N2¼ 30 W2¼ 2,627, N2¼ 30 New privately-owned banks W3¼ 4,962, N3¼ 69 W3¼ 3,601, N3¼ 69 H statistic 10.397 21.378 x0.012 (2) ¼ 9.210 Significant at 1% Significant at 1% Note: H ¼ 12 N(N þ 1) PW2 i Ni 3(N þ 1):

N: total sample number, Ni: sample number of the ith set, Wi: rank sum of the ith set.

T A B L E 4 E C O N O M I E S O F S C A L E A N D S C O P E B Y T Y P E S O F B A N K S Bank types Sample number Sample mean of SE Sample mean of SC Sample no. and percentage with SE . 1 Government-owned or -controlled banks 33 1.0406 (0.0696) 0.8618 (0.4188) 23 (69.7%) Old privately-owned banks 30 1.1676 (0.0707) 0.2591 (0.1174) 30 (100%) New privately-owned banks 69 1.2151 (0.0782) 0.1821 (0.0655) 69 (100%) Asset size ,250 83 1.2194 (0.0729) 0.1724 (0.0395) 83 (100%) Asset size 250-1,000 33 1.0950 (0.0458) 0.4531 (0.2159) 33 (100%) Asset size .1,000 16 0.9911 (0.0181) 1.2141 (0.2033) 6 (37.5%) Total 132 1.1606 (0.1033) 0.3718 (0.3633)

Notes: The sample standard deviations are in parentheses. Asset size is measured in billion NT dollars.

SE represents the economies of scale and SC represents the economies of scope.

cates that increasing returns to scale exist for small and medium banks, while decreasing returns to scale exist for large banks. Thus, economies of scale are larger for smaller banks. This finding implies that size expansion can yield greater cost advantages for small banks than for large banks. Specifically, banks with assets of under 1,000 billion NT dollars may improve their cost effi-ciency by size expansion, possibly through mergers and acquisitions. The sample mean of SE (economies of scale) for merged banks is 1.2211, larger than 1.1529 for non-merged banks. The percentage of banks that operate with economies of scale is also larger for merged banks (93.3 per cent). Merged banks benefit more from the economies of scale than the non-merged banks. Since all banks have SC (economies of scope) values larger than zero, cost savings can be achieved from the joint production of loans and investments. However, large banks benefit more than small banks from economies of scope.

Relationship Between Cost Efficiency and Merger Activity

Table 5 summarises the descriptive statistics of variables used in the censored regression model Equation 17. The nonparametric Kruskal-Wallis test is used to check intertemporal improvement in cost efficiency during the sample period. Notably, the H statistic (2.6468) is below the critical value, implying that there is no significant difference in cost efficiency from 1997 to 1999.

Table 6 presents the estimates of the parameters in the regression model. All the variables except TA, TA2, GROWTH and NINTOP markedly affect cost efficiency. The estimated coefficient of ETA (ratio of number of employees to total assets) is significantly negative. Mergers in Taiwan generally do not lead to large-scale layoffs. Such action can provoke employee protests and create political problems, which in turn may impede the approvals of mergers by the authorities. Consequently, cost efficiency decreases as the size of the labour force increases. The positivity of the influence of TDTD (ratio of time deposits to total deposits) shows that banks with high proportions of time deposits

T A B L E 5

D E S C R I P T I V E S T A T I S T I C S O F V A R I A B L E S U S E D I N T H E C E N S O R E D R E G R E S S I O N M O D E L

Variable Mean Std Dev. Maximum Minimum eff 0.944122 0.021031 0.976674 0.850260 TA 390,870 425,323 2,074,455 43,569 GROWTH 0.123932 0.121097 0.533819 20.110819 ETA 0.007274 0.002282 0.014247 0.002834 BTD 0.000260 0.000135 0.000768 0.000068 TDTD 0.750232 0.069555 0.865700 0.524200 NINTOP 2.379456 5.943318 65.473680 21.735294 Note: TA is measured in million NT dollars.

enjoy higher cost efficiency because such funds are stable, manageable and much cheaper than other funds. The positive BTD (ratio of number of branches to total deposits) implies that this variable affects outputs more strongly than inputs. While branching can increase input expenses, it also expands the revenue base from the outputs.

With the effects of other variables controlled, a statistically significant relationship clearly exists between bank mergers and cost efficiency. Mergers can enhance cost efficiency, even though the number of employees does not decline. The banks involved are generally small and were established after the banking sector was deregulated. Since the banking industry remains highly regulated even after its deregulation, branching barriers persisted after 1991. New branches require special approval by the Ministry of Finance and normally no more than two new branches are permitted for each bank in a given year. This is an important constraint for banks that are considering aggressive expansion, especially for new banks with insufficient market coverage. Banks that take over other financial institutions may transfer newly acquired branches to other locations. Through mergers, these banks can quickly penetrate other market areas and thus make better use of their combined resources. This argument is also supported by the positive BTD parameter.

Since cost efficiency is derived not from closing branches or laying off personnel, merging obtains operational synergies relying on economies of scale and scope. As discussed in the earlier sections, smaller banks exhibit better economies of scale than larger banks, while larger banks enjoy better economies of scope than smaller banks. Consequently, size has a mixed effect on cost efficiency. However, branching privileges show that Taiwanese

T A B L E 6

E S T I M A T E S O F P A R A M E T E R S I N C E N S O R E D R E G R E S S I O N M O D E L

Variables Coefficient z-statistic
Intercept 0.942353 35.939070
MERGE 0.010230 1.836319
TA 21.93E 2 08 21.157650
TA2 5.83E 2 15 0.681308
GROWTH 20.007842 20.867365
BTD 63.930260 3.060575
ETA 28.329762 26.618873
TDTD 0.069533 2.438017
NINTOP 20.000305 21.133296
Note: Adjusted R2_{¼ 0.307836.}
_{Significant at 10%.}
_{Significant at 5%.}
_{Significant at 1%.}

banks can probably enjoy greater economies of scope through mergers. There-fore, bank mergers are positively related to cost efficiency.

CONCLUSION

This work studies cost efficiency, economies of scale and economies of scope of the Taiwanese banking industry, and further elucidates the potential impact of bank mergers on cost efficiency. Adopting stochastic frontier analysis, this study employs a translog cost function with composite errors to explain managerial inefficiency and environmental effects. Furthermore, loan outputs are adjusted to account for non-performing loans. The sample period is from 1997 to 1999, which covers the main wave of bank mergers in Taiwan. The empirical results suggest that economies of scale and scope do exist, but depend on bank size. Further regression analysis reveals that merger activity significantly affects cost efficiency. The evidence also demonstrates variations in cost efficiency among different organisational types. Government-owned or -controlled banks enjoy the highest cost efficien-cies, while old privately owned banks have the lowest cost efficiencies.

The Taiwanese government has always encouraged bank merger activity to promote economic stability. This study supports this policy. Although entry barriers were lifted following the deregulation of the banking sector, expansion via branching remains restricted. The opening of new branches requires special approval by the authorities. However, banks that take over other financial institutions are allowed to transfer the new branches to other locations. The branching privileges associated with mergers and acquisitions in Taiwan may explain the positive effect of merging on efficiency, despite the fact that the workforce is generally not reduced.

Though merged and non-merged banks in Taiwan have different sizes and organisational types, how these factors affect cost efficiency remains unsolved. This study only considers the cost side of mergers. A complete evaluation of the effects of mergers would have to consider also the revenue side (profit efficiency). All these are left for future studies.

ACKNOWLEDGEMENTS

The authors would like to thank Cliff J. Huang and two anonymous referees for their helpful com-ments and suggestions. Any remaining errors are the responsibility of our own. This research was supported in part by the National Science Council under Grant NSC 92-2416-H-009-017.

REFERENCES

Akhavein, J.D., A.N. Berger and D.B. Humphrey, 1997, ‘The Effect of Megamergers on Effi-ciency and Prices: Evidence from a Bank Profit Function’, Review of Industrial Organization, Vol.12, pp.95 – 139.

Avkiran, N.K., 1999, ‘The Evidence on Efficiency Gains: The Role of Mergers and the Benefits to the Public’, Journal of Banking and Finance, Vol.23, pp.991 – 1013.

Battese, G.E. and T.J. Coelli, 1988, ‘Prediction of Firm-Level Technical Efficiencies with a Generalized Frontier Production Function and Panel Data’, Journal of Econometrics, Vol.38, pp.387–99.

Bera, A.K. and S.C. Sharma, 1999, ‘Estimating Prediction Uncertainty in Stochastic Frontier Production Function Models’, Journal of Productivity Analysis, Vol.12, pp.187 – 210. Berger, A.N., 1998, ‘The Efficiency Effects of Bank Mergers and Acquisitions: A Preliminary

Look at the 1990s Data’, in Y. Amihud and G. Miller (eds), Bank Mergers and Acquisitions, Dordrecht: Kluwer Academic, pp.79 – 111.

Berger, A.N. and R. DeYoung, 1997, ‘Problem Loans and Cost Efficiency in Commercial Banks’, Journal of Banking and Finance, Vol.21, pp.849– 70.

Berger, A.N. and D.B. Humphrey, 1992, ‘Megamergers in Banking and the Use of Cost Efficiency as an Antitrust Defense’, Antitrust Bulletin, Vol.37, pp.541 – 600.

Bernstein, D., 1996, ‘Asset Quality and Scale Economies in Banking’, Journal of Economics and Business, Vol.48, pp.157 – 66.

Cebenoyan, A.S., E.S. Cooperman and C.A. Register, 1993, ‘Firm Efficiency and the Regular Closure of S&Ls: An Empirical Investigation’, The Review of Economics and Statistics, Vol.75, No.3, pp.540– 44.

Chen, B.S. and R.Y. Chen, 2002, ‘The Mergers and Acquisitions of Taiwan’s Domestic Banking Industry: Impacts on Efficiency and Market Power’, Conference Paper, Asia Conference on Efficiency and Productivity Growth, Taiwan, 2002.

Chen, T.Y. and T.L. Yeh, 2000, ‘A Measurement of Bank Efficiency, Ownership and Productivity Changes in Taiwan’, The Service Industries Journal, Vol.20, No.1, pp.95 – 109.

Cornett, M.M. and H. Tehranian, 1992, ‘Changes in Coporate Performance Associated with Bank Acquisitions’, Journal of Financial Economics, Vol.31, pp.211– 34.

Cowing, T.G., D. Reifschneider and R.E. Stevenson, 1983, ‘A Comparision of Alternative Frontier Cost Function Specifications’, in Ali Dogramaci (ed.), Developments in Econometric Analyses of Productivity, Nijhoff: Kluwer Academic, ch.4.

DeYoung, R., 1997, ‘Bank Mergers, X-Efficiency, and the Market for Corporate Control’, Mana-gerial Finance, Vol.23, pp.32 – 47.

Greene, W.H., 1990, ‘A Gamma-Distributed Stochastic Frontier Model’, Journal of Econometrics, Vol.46, pp.141 – 64.

Hannan, T.H. and J.D. Wolken, 1989, ‘Returns to Bidders and Targets in the Acquisition Process: Evidence from the Banking Industry’, Journal of Financial Services Research, Vol.3, pp.5–16. Hao, J., W.C. Hunter and W.K. Yang, 2001, ‘Deregulation and Efficiency: The Case of Private

Korean Banks’, Journal of Economics and Business, Vol.53, pp.237 – 54.

Huang, M.H. and C.J. Huang, 2002, ‘The Impact of Banking Regulatory Reform on the Industry’s Productivity and Efficiency in Taiwan’, conference paper, Asia Conference on Efficiency and Productivity Growth, Taiwan, 2002.

Huang, C.J., T.T. Fu and M.Y. Huang, 1999, ‘Cost Efficiency of the Farmers’ Credit Unions in Taiwan’, in Tsu-Tan Fu, Cliff J. Huang and C.A. Knox Lovell (eds), Economic Efficiency and Productivity Growth in the Asia-Pacific Region, Northampton: Edward Elgar, pp.309– 26.

Hunter, C.W. and S.G. Timme, 1986, ‘Technical Change, Organizational Forms, and the Structure of Banking Production’, Journal of Money, Credit and Banking, Vol.18, pp.153– 60. Jondrow, J., C.A.K. Lovell, I.S. Materov and P. Schmidt, 1982, ‘On the Estimation of Technical

Inefficiency in the Stochastic Frontier Production Function Model’, Journal of Econometrics, Vol.19, pp.233 – 8.

Kaparakis, E.I., S.M. Miller and A.G. Noulas, 1994, ‘Short-Run Cost Inefficiency of Commercial Banks: A Flexible Stochastic Frontier Approach’, Journal of Money, Credit and Banking, Vol.5, pp.875– 93.

Lang, G. and P. Welzel, 1999, ‘Mergers among German Cooperative Banks: A Panel-Based Stochastic Frontier Analysis’, Small Business Economics, Vol.13, pp.273– 86.

Li, Y., J.L. Hu and H.W. Liu, 2002, ‘Non-Performing Loans and Bank Efficiencies: An Appli-cation of the Distance Function Approach’, conference paper, Asia Conference on Efficiency and Productivity Growth, Taiwan, 2002.

Mester, L.J., 1993, ‘Efficiency in the Savings and Loan Industry’, Journal of Banking and Finance, Vol.17, pp.267 – 86.

Mester, L.J., 1996, ‘A Study of Bank Efficiency Taking into Account Risk-Preferences’, Journal of Banking and Finance, Vol.20, pp.1025 – 45.

Murry, J. and R.W. White, 1983, ‘Economies of Scale and Economies of Scope in Multiproduct Financial Institutions: A Study of British Columbia Credit Unions’, The Journal of Finance, Vol.38, No.3, pp.887– 902.

Olson, J.A., P. Schmidt and D.M. Waldman, 1980, ‘A Monte Carlo Study of Estimators of Stochastic Frontier Production Functions’, Journal of Econometrics, Vol.13, pp.67 – 82. Ou, C.S., C.L. Lee and C.S. Young, 2002, ‘Using DEA to Examine the Association between Bank

Asset Quality and Operating Performance: the Case of Taiwan’s Banking Industry’, confer-ence paper, Asia Conferconfer-ence on Efficiency and Productivity Growth, Taiwan, 2002. Panzar, J.C. and R.D. Willig, 1981, ‘Economies of Scope’, American Economic Review, Vol.71,

pp.268– 73.

Pilloff, S.J., 1994, ‘Performance Changes and Shareholder Wealth Creation Associated with Mergers of Publicly Traded Banking Institutions’, Working Paper 94 – 10, Financial Insti-tutions Center, The Wharton School, University of Pennsylvania.

Pilloff, S.J., 1996, ‘Performance Changes and Shareholder Wealth Creation Associated with Mergers of Publicly Traded Banking Institutions’, Journal of Money, Credit and Banking, Vol.28, pp.294– 310.

Prager, R.A. and T.H. Hannan, 1998, ‘Do Substantial Horizontal Mergers Generate Significant Price Effects? Evidence from Banking Industry’, Journal of Industrial Economics, Vol.46, No.4, pp.433 – 52.

Resti, A., 1998, ‘Regulation Can Foster Mergers, Can Mergers Foster Efficiency? The Italian Case’, Journal of Economics and Business, Vol.50, pp.157– 69.

Rhoades, S.A., 1998, ‘The Efficiency Effects of Bank Mergers: An Overview of Case Studies of Nine Mergers’, Journal of Banking and Finance, Vol.22, pp.273– 91.

Shaffer, S., 1993, ‘Can Mergers Improve Bank Efficiency?’, Journal of Banking and Finance, Vol.17, pp.423– 36.

Simons, K. and J. Stavins, 1998, ‘Has Antitrust Policy in Banking Become Obsolete?’, New England Economic Review, Mar/Apr, pp.13 – 26.

Stevenson, R.E., 1980, ‘Likelihood Functions for Generalized Stochastic Frontier Estimation’, Journal of Econometrics, Vol.36, pp.57 – 66.

Vennet, R.V., 1996, ‘The Effect of Mergers and Acquisitions on the Efficiency and Profitability of EC Credit Institutions’, Journal of Banking and Finance, Vol.20, pp.1531 – 58.

Zellner, A., 1962, ‘An Efficient Method of Estimating Seemingly Unrelated Regressions and Tests of Aggregation Bias’, Journal of the American Statistical Association, Vol.57, pp.348– 68.

APPENDIX 1

S A M P L E B A N K S L I S T E D A C C O R D I N G T O O R G A N I S A T I O N A L T Y P E

Organisational type Name of banks Total

Government-owned or controlled banks

The Farmers Bank of China, Chiao Tung Bank, Bank of Taiwan, Land Bank of Taiwan, Taiwan Cooperative Bank, First Commercial Bank, Hua Nan Bank, Chang Hwa Bank, Bank of Kaohsiung, Taipeibank, Bank of Taiwan Province

11

Old privately owned banks (established before 1991)

The International Commercial Bank of China, International Bank of Taipei, Hsinchu International Bank, Taichung Bank, Tainan Business Bank, Kaohsiung Bank, Taitung Bank, Bank of Overseas Chinese, The Shanghai Commercial and Savings Bank, United World Chinese Bank

10

New privately owned banks (established after 1991)

Makoto Bank, Sunny Bank, Bank of Pan Shin, Lucky Bank, Kao Shin Bank, Grand Bank, Dah An Bank, Union Bank of Taiwan, The Chinese Bank, Bank Sinopac, Asia Pacific Bank, E. Sun Bank, Cosmos Bank Taiwan, Pan Asia Bank, Chung Shing Bank, Taishin Bank, Far Eastern Bank, Fubon Bank, Ta Chong Bank, Baodao Bank, Chinatrust Bank, En Tie Bank, Chinfon Bank

23

APPENDIX 2

T R A N S L O G C O S T F U N C T I O N E S T I M A T E S

Variable Coefficient Estimate t-Statistic
Intercept a 10.7192 7.7456
ln Y1 b1 20.7924 21.9346
ln Y2 b2 0.5841 1.7111
ln P1 g1 0.2934 5.9599
ln P2 g2 0.5788 9.6287
ln P3 g3 0.1278 1.7898
(ln Y1)2 b11 0.2470 3.0814
(ln Y2)2 b22 0.1289 2.8901
(ln Y1)(ln Y2) b12 20.1448 22.3629
(ln P1)2 g11 0.1449 26.9089
(ln P2)2 g22 0.0650 8.5648
(ln P3)
2
g33 0.1580 13.4347
(ln P1)(ln P2) g12 20.0211 25.7934
(ln P1)(ln P3) g13 20.1223 219.4509
(ln P2)(ln P3) g23 20.0388 25.7169
(ln Y1)(ln P1) r11 0.0052 0.9659
(ln Y2)(ln P1) r21 20.0082 21.7202
(ln Y1)(ln P2) r12 20.0093 22.6646
(ln Y2)(ln P2) r22 20.0050 21.4347
(ln Y1)(ln P3) r13 0.0076 1.1490
(ln Y2)(ln P3) r23 0.0124 2.1336
ZQ d1 0.1509 5.7479
sU
2
0.0093
sV2 0.0087
Note: Adjusted R2¼ 0.9845.
_{Significant at 10%.}