Banks and Bank Systems International Research

0 downloads 0 Views 1MB Size Report
The journal examines the processes of banking systems development in the countries ...... Twenty bank branches from all banking institutions ...... Empo – Emporiki Bank, gpsb – Greek Postal Savings Bank, geb – Geniki bank, egna ...... Banking // 4th Workshop on Business Process Intelligence (BPI 08), Milano, 01.09.2008.
Banks and Bank Systems International Research Journal Volume 3, Issue 4, 2008 Issued from 2006 Published quarterly ISSN 1816-7403 ISSN online 1991-7074 Head of the Editorial Board: Prof. Anatoliy Yepifanov The journal examines the processes of banking systems development in the countries of the world and individual banks; their transformation influenced by globalization and liberalization processes. It also addresses the problems of maintaining a competitive edge. Address: Publishing Company “Business Perspectives” Dzerzhynsky lane 10, Sumy 40022 Ukraine E-mail: [email protected] URL: http://www.businessperspectives.org The authors are responsible for the reliability of information which materials published contain. Reprinting and reproduction of published materials are possible in case of referring to an author and an edition. © Publishing Company “Business Perspectives”, 2008 Copyright: All rights reserved. No part of this publication may be reproduced, stored, transferred, advertised, demonstrated, adaptated, rearranged, translated in any form or bought by any means. This also concerns the distribution, disposition, property renting, commercial renting, or any other kind of renting, reprinting, siting, importing or public demonstration. The prior written permission of the Publisher is required. The above-named requirements should be also referred to non-profit basis as well as any free access to the previous, current and future issues of the publication.

Editorial Board Anatoliy Yepifanov, Dr., Prof., Honored Economist of Ukraine, Rector of Ukrainian Academy of Banking of the National Bank of Ukraine (Ukraine) – Head of the Editorial Board E. Altman, Professor of Finance, Stern School of Business, New York University, Director of the Credit and Fixed Income Research Program, NYU Salomon Center (USA) M.I. Blejer, Director, Centre for Central Banking Studies, Bank of England (UK) K. Bröker, Dr., Prof., Attorney at Law, Attys. Jena & Prof., Dr. Bröker (Germany) S.G. Cecchetti, Professor of International Economics and Finance, International Business School, Brandeis University, Director of Research, Rosenberg Institute for Global Finance (USA) J.A. Consiglio, Dr., Department of Banking and Finance, Faculty of Economics, Management & Accountancy, The University of Malta, and Chairman, The Income Tax Tribunal of Special Commissioners (Malta) B. Danylyshyn, Member-correspondent of NAS of Ukraine, Interim Chairman of Council of Investigating Productive Forces of Ukraine of NAS of Ukraine (Ukraine) V. Geyets, the Academician of NAS of Ukraine (Department of Economics), Director of Institute of Economics and Forecasting of NAS of Ukraine, Head of the Board of the National Bank of Ukraine (Ukraine) S. Heffernan, Professor of Banking and Finance, Faculty of Finance, Cass Business School, City of London (UK)

L. Pawlowicz, Prof., Gdańsk University, Director, Gdańsk Academy of Banking, Deputy President of the Management Board, Gdańsk Institute of Market Economics (Poland) S. Poloucek, Professor of Finance and Banking, Head of the Department of Finance and Head of the Institute of Doctoral Studies, School of Business Administration in Karviná, Silesian University (Czech Republic) I. Salo, Dr., Prof., Ukrainian Academy of Banking of the National Bank of Ukraine (Ukraine) D. Schönwitz, Dr., Prof., Principal of the University of Applied Sciences of the Deutsche Bundesbank (Germany) M.H. Sharma, Ph.D., Professor of Banking, School of Accounting and Finance, University of the South Pacific (Fiji Islands) M. Sõrg, Professor of Money and Banking, Head of the Institute of Finance and Accounting, University of Tartu (Estonia) P. Steiner, Head of the Institute of Banking and Finance, Graz University (Austria) M. Theobald, Professor of Finance and Investment, Head of Accounting & Finance Subject Group, Birmingham Business School, University of Birmingham (UK)

E. Hüpkes, Head of Regulation, Swiss Federal Banking Commission (Switzerland)

J.B. Thomson, Ph.D., Vice President and Economist, Research Department, Federal Reserve Bank of Cleveland; Lecturer, Weatherhead School of Management, Case Western Reserve University (USA)

A. Janc, Head of Department of Banking, Dean of the Faculty of Economics, The Poznań University of Economics (Poland)

D. Tripe, Senior Lecturer at the Department of Finance, Banking & Property; Director of the Centre for Banking Studies, Massey University (New Zealand)

T.W. Koch, SCBA Chair of Banking & Professor of Finance, Moore School of Business, University of South Carolina (USA)

T. Vasyl’yeva, Candidate of Sciences in Economics, Assoc. Prof., Ukrainian Academy of Banking of the National Bank of Ukraine (Ukraine)

O. Kozmenko, Candidate of Sciences in Economics, Assoc. Prof., Ukrainian Academy of Banking of the National Bank of Ukraine (Ukraine) M.K. Lewis, Ph.D., Professor of Banking and Finance, School of Commerce, University of South Australia (Australia) D.T. Llewellyn, Professor of Money and Banking, Loughborough University (UK) M. Makarenko, Dr., Prof., Ukrainian Academy of Banking of the National Bank of Ukraine (Ukraine) A. Mullineux, Professor of Global Finance, Department of Accounting and Finance, The Birmingham Business School, The University of Birmingham (UK) E. Orhaner, Head of Department of Banking and Insurance Education, Faculty of Education of Commerce and Tourism, Gazi University (Turkey)

O. Vasyurenko, Dr., Prof., Vice-Chancellor and Director of Kharkiv Affiliate of Ukrainian Academy of Banking (Ukraine) I. Walter, Professor of Finance, Stern School of Business, New York University; Director of the Stern Global Business Institute, NYU Stern School of Business (USA) L. White, Ph.D., CTP, Director, Banking Excellence Program and Interim Warren Chair of Real Estate, Department of Finance and Economics, College of Business and Industry, Mississippi State University (USA) L. Wu, Associate Professor of Finance, Zicklin School of Business (USA); H.-C. Yu, Chung Yuan Christian University Business School (Taiwan) S.A. Zenios, Professor of Finance and Management Science, University of Cyprus (Cyprus) M. Zineldin, Ph.D., Professor in Economics, School of Management and Economics, Växjö University (Sweden)

Contents Hai-Chin Yu, Ken H. Johnson, Der-Tzon Hsieh Public debt, bank debt, and non-bank private debt in emerging and developed financial markets

4

Kuan-Min Wang, Thanh-Binh Nguyen Thi, Shu-Hui Wu Dynamic hetero-risk: the major determinant of loan rate pass-through mechanism in Taiwan

12

Alexandra Lai, Raphael Solomon Ownership concentration and competition in banking markets

16

Chung-Hua Shen, Meng-Fen Hsieh, Chien-Chiang Lee Bank provisioning, business cycles and bank regulations: a comprehensive analysis using panel data

29

John Mylonakis The influence of banking advertising on bank customers: an examination of Greek bank customers’ choices

44

Seok Weon Lee Asset size, risk-taking and profitability in Korean banking industry

50

Christos Floros, Georgia Giordani ATM and banking efficiency: the case of Greece

55

Izah Mohd Tahir, Nor Mazlina Abu Bakar, Sudin Haron Technical efficiency of the Malaysian commercial banks: a stochastic frontier approach

65

Wolfgang Benner, Lyudmil Zyapkov A multifactoral Cross-Currency LIBOR Market Model with an FX volatility skew

73

Andreas Burger, Juergen Moormann Productivity in banks: myths & truths of the Cost Income Ratio

85

Jean Perrien, Raoul Graf, Fabien Durif, Lionel Colombel The role of norms in the evolution of a relationship: the case of an asymmetrical process in the banking industry

95

Authors of the issue

102

Submission guidelines for authors

104

A joint subscription form 2008

105

Banks and Bank Systems, Volume 3, Issue 4, 2008

Hai-Chin Yu (Taiwan), Ken H. Johnson (USA), Der-Tzon Hsieh (Taiwan)

Public debt, bank debt, and non-bank private debt in emerging and developed financial markets Abstract Using an effective sample of 3,453 observations selected from the Taiwanese stock exchange, this study documents and attempts to reconcile divergent outcomes from the extant literature on debt structure (public, bank, and non-bank private debt). Sampled firms from this emerging market generally acquire debt from both public and private sources, with a strong preference for bank debt, suggesting, among other things, that bank debt and public debt complement each other rather than acting as substitutes. Four interesting alternative explanations are provided in an attempt to reconcile the contra indicative results that arise when modeling the use of public debt. Keywords: bank debt, public debt, private debt, debt structure. JEL Classification: G32, G20, G21.

Introduction © Traditionally, the world’s financial markets are viewed as either developed or emerging ones. Today, however, this description is better restated as a process of emerging and merging markets. To continue in an orderly process, reconciling the practices and conventions across the growing world financial market is necessary. This paper contributes to this process by analyzing debt structure for publicly traded firms in one of the world’s recognized emerging markets and attempts to reconcile differential outcomes with the extant literature. In the U.S., debt financing has been the predominant source of external funds over the past two decades (Denis and Mihov, 2003). In the last decade, the same is true in the emerging market of Taiwan with roughly 60% of needed funds being raised by way of external debt based on estimates from the Central Bank of Taiwan. About 40% of this corporate debt is raised from financial institutions (namely, commercial banks), suggesting the Taiwanese financial system is significantly bank-based. Additionally, the openness of the market over the past decade has allowed a number of Taiwanese firms to readily issue bonds. Thus, most Taiwanese, as well as U.S., firms face a mixture of debt sources. In both markets, the use of bank debt as opposed to public debt is casually perceived as an issue of firms being at opposite ends of the reputation-credit quality spectrum. That is, bank debt is assumed to be more expensive than public debt and hence less desirable because of monitoring and agency costs. However, certain lending practices are significantly different in the two markets. For example, the common employment of convertible debt in the Taiwa© Hai-Chi n Y u , K e n H. Johnson, Der-Tzon Hsieh, 2008. This paper benefits greatly from helpful comments provided by Larry H.P. Lang, Shane Johnson, Raj Aggarwal, Geoffrey Friesen, Ingyu Chiou, Jim Ligon, William Welch, and numerous participants at the Financial Management Association’s Annual Meeting in Denver.

4

nese public debt market is almost unheard of in the corresponding U.S. market. Consequently, investigating firms from an emerging market with access to public debt and their subsequent debt structure is an interesting topic. What drives the makeup of this mixture of public and private debt in developed markets is the subject of much debate in the extant literature. Numerous works attempt to explain corporate preferences in debt mixture. Fama (1985), Berlin and Loeys (1988), Diamond (1991), and Berlin and Mester (1992) provide a representative sample of these works. Generally, these works conclude that banks help mitigate problems stemming from information asymmetries between firms and debt holders. Specifically, firms can use bank monitoring to reduce these asymmetric information related problems, thus increasing optimal leverage and firm value. Additionally, some studies hypothesize that private debt financing has a significant advantage over public debt in terms of monitoring efficiency (e.g., Diamond, 1984; and Boyd and Prescott, 1986), while other stress access to private information and the efficiency of liquidation and renegotiation in financial distress (e.g., Fama, 1985; Chemmanur and Fulghieri, 1994; Gertner and Scharfstein, 1991). Interestingly, however, Rajan (1992) argues private lenders can negatively affect the borrower by extracting rents and distorting management incentives. Other works including, but not limited to, Smith and Warner (1979), Blackwell and Kidwell (1988), Diamond (1984, 1991), and Berlin and Loeys (1988) argue that the difference between public and private debt is that the former has higher agency costs relative to the latter. In particular, and with regards to monitoring bond issuers, public debt is associated with lower incentives of individual bondholders. Monitoring on the part of numerous bondholders, as is the case with public debt, is also inefficient, since monitoring involves wastage through the duplication of monitoring costs. By contrast, private debt is asso-

Banks and Bank Systems, Volume 3, Issue 4, 2008

ciated with fewer free rider problems and results in an increase in monitoring efficiency. Arguments exist in favor of the use of public debt. For example, Diamond (1993) argues that excluding short-term private debt may be costly because borrowers become unable to determine whether a loan can be rolled over or whether liquidation may be enforced. On the other hand, using a mixture of both public and private debt allows borrowers to reduce the control exercised by the private lender and thereby avoid costly liquidations. Diamond (1991) and Hoshi, Kashyap, and Scharfstein (1993) suggest that reliance on private lending is related to the credit quality and the reputation of the borrower with high net worth firms tending to rely more on public rather than private debt providers. Of particular note, Blackwell and Kidwell (1988) hypothesize that firms with higher credit-worthiness rely more on public debt than firms with lower credit-worthiness, because of the lower transaction costs associated with public debt which more than offset the higher agency costs. Houston and James (1996) further examine the determinants of the mix of private and public debt using detailed information on debt structure. Their findings suggest that the relationship between bank borrowing and the importance of growth opportunities depends on the number of banks used by the firm, and on whether or not the firm has public debt outstanding. Specifically, when firms borrow from many different banks, a positive relationship exists between bank debt and growth opportunities. By contrast, when firms borrow from a single bank, the relationship is negative. The availability of collateral and its use as security also affect debt mixture. Rajan and Winton (1995) indicate a positive correlation between the use of private debt and financial distress. Therefore, banks only require collateral in the bad state of the world; thereby, signaling their asymmetric information to the public. Furthermore, as firms draw nearer to financial distress, the ratio of secured claims increases. Besanko and Thakor (1987) and Boot, Thakor, and Udell (1991), among others, however, suggest that collateral can be viewed as a signal of quality. Obviously, at some point, collateral can be sufficient enough so as to eliminate worries on the part of creditors over financial distress. All private debt is not the same however. Hooks and Opler (1993) illustrate this point when they note that the vast majority of theoretical models on the choice of debt structure assume that bank and non-bank private debts are equivalent. Johnson (1997) seems to follow this argument and expands the mixture of debt financing to include non-bank private financ-

ing. Johnson finds a difference between bank and non-bank private debts. Specifically, bank debt use is negatively correlated with the market-to-book ratio, and positively correlated with the fixed assets ratio and leverage, while non-bank private debt is positively and statistically significantly correlated with the market-to-book ratio, and negatively correlated with the fixed assets ratio and leverage. The only similarity that Johnson finds is that bank debt and non-bank private debt are both negatively correlated with age. Denis and Mihov (2003) follow from finding firms with the highest credit quality borrow from public sources, while those with medium credit quality borrow from banks, and finally those with the lowest credit quality borrow from non-bank private lenders. Thus, non-bank private debt plays a unique role in accommodating the financing needs of firms with low credit quality. This study continues in the vein of Hooks and Opler (1993), Johnson (1997), and Denis and Mihov (2003) in that a distinction is made between public, bank, and non-bank private debt in an effort to learn more about the overall debt mixture of publicly traded firms. The remainder of this study is organized as follows: Section 1 describes the data sources, Section 2 discusses modeling specifications, Section 3 presents the empirical results, and the last section provides concluding remarks. 1. Sample data This study empirically examines the choice of public and private debt among Taiwanese companies using a sample of 3,453 observations selected from 579 firms listed on the Taiwanese stock exchange over the period of 1991-2000. Data were obtained from three sources: (i) the Taiwan Economic Journal (TEJ) – financial statements, (ii) the TEJ annual lending database, and (iii) the Taiwan Security and Exchange Council (TSEC). In this study, bank debt is defined as firm borrowings from commercial banks, while non-bank private debt is defined as borrowing from insurance, finance, or leasing companies. Finally, public debt includes corporate bonds, commercial paper and bankers’ acceptances. Consistent with Johnson (1997), among others, all of the above forms of debt are further reclassified according to whether they are long-term (maturity of three years or more) or short-term (maturity less than three years). 2. Empirical methodology The relationship between debt structure and firm characteristics is examined by way of Tobit regression analysis with limits at zero and unity. Maintaining consistency with many prior works, this work employs the ratio of long-term sources of debt (pub5

Banks and Bank Systems, Volume 3, Issue 4, 2008

lic, bank, and non-bank private) to total debt as its dependent variables. In other words, the tendencies to employ particular debt sources are the primary dependent variables for this study. With respect to the cited works above, explanatory proxies for firm characteristics are briefly discussed next to provide the reader with a general understanding of the models specifications. A firm’s access to debt sources, both public and private, is deemed to increase with reputation. Accordingly, firm age, defined as the number of years since the firm was first incorporated, is specified as a proxy for reputation and should be positively associated with the tendency to employ public debt. Additionally, as in most prior studies, the natural log of the book value of total assets is used as a proxy for firm size and size should favor the use of debt in general. The fixed assets ratio, including net property, plant, and equipment divided by total assets, is used as a proxy for asset collateral value. Collateral value depends on liquidation value, and so asset collateral value can be regarded as a proxy for project liquidation values. Hence, firms with higher fixed assets ratio are viewed as highly collateralized firms and should positively influence the use of debt irrespective of its source. Tobin’s Q is used when calculating the market to book value ratio as a proxy for firm performance. The literature as a whole generally agrees that firms with higher Tobin’s Q (Q>1) have better performance than those with lower Q (Q|η 2 | denotes the adjustment with upward rigidity and | η1 |≤| η 2 | denotes the adjustment with downward rigidity. 2. Data and empirical results This study uses monthly data ranging from February 1986 to July 2005. The interbank overnight call-loan rate ( mit ) represents the market rate, and loan rate ( lit ) is the loan rate mean of 36 major banks in Taiwan. The data are provided by Taiwan Economic Journal Data Bank (TEJ). First, we used the Augmented Dickey Fuller (ADF) unit root test to check for stationarity in rate series. Table 1 reports the ADF test results of the level and first differential. Under 1% significant level, the level terms of both interbank rate and loan rate are unable to reject the null hypothesis of nonstationarity. However, the differential terms significantly reject this null hypothesis. Table 1. ADF unit root test Variable

Level

First-order differential

lit

-2.336 [1]

-9.960*** [0]

mit

-0.791 [8]

-8.711*** [7]

Note:

The

regression

∆y t = a 0 + γy t −1 +



p

φ ∆y t −i +1 i =2 i

of

ADF

is

+ ε t . The maximum lag is

14 periods. The values in [.] are the most fitting lags determined by the least AIC value. *** denotes the 1% significance level and the 1% critical value is -3.463 (see MacKinnon, 1996).

The estimated values of β 0 , β1 and results of Engle-Granger cointegration test are provided in Table 2. The estimated coefficient on the market rate vari-

able in the equation (1) is 0.259 and it is statistically significant at 1% level, showing the non-complete pass-through of market rate to loan rate. The estimated coefficient of β 0 is significantly greater than zero, indicating that the markup on loan rate is commonly applied to offset the additional charge. Besides, the results of Engle-Granger cointegration test indicate the inexistence of symmetric long-run relationship; it means that the market rate is unable to be used for forecasting the tendency of loan rate under the symmetric information. Table 2. Engle-Granger cointegration test

β1

7.043***

0.259***

-2.144

(63.50)

(13.17)

[8]

equation

lit = β 0 + β1m

H 0 : no

β0

Long-run

cointegration

Note: In Engle-Granger cointegration test, the regression of

∆y t = γy t −1 +



p φ ∆y t −i +1 i=2 i

+εt ADF is . The maximum lag is 14. The value in [.] is the most fitting lag determined by the least AIC value. The values in (.) are t-statistics. *** denotes the 1% significance level and the critical value is -3.98 (see MacKinnon, 1996).

Table 3 shows the results of TAR and MTAR tests on model using et −1 as an indicator variable (referred to as model 1) and on model using σ t −1 as an indicator variable (referred to as model 2). Under 5% significance level, it is discovered that model using et −1 ( ∆et −1 ) as an indicator variable shows the inexistence of cointegration relationship, whereas model using σ t −1 ( ∆σ t −1 ) as an indicator variable provides a robust evidence of asymmetric cointegration relationship. This result indicates that the factor determining the existence of pass-through mechanism from market rate to loan rate in Taiwan is the dynamic hetero-risk.

Table 3. TAR and MTAR cointegration test Indicator

H 0 : ρ1 = ρ 2 = 0

τ

H 0 : ρ1 = ρ 2

Lags

ρ1

ρ2

et −1

7

-0.058

-0.057

3.066

-0.061

0.000

0.983

∆et −1

7

-0.049

-0.061

3.130

-0.111

0.056

0.812

TAR

σ t −1

6

-0.161

-0.040

7.255**

0.141

5.747

0.017**

MTAR

∆σ t −1

5

-0.168

-0.037

6.479**

0.116

5.823

0.016**

Model

TAR MTAR

variable

Φ -statistic

F-statistic

p-value

Note: The maximum lag is 14. The most fitting lag determined by the least AIC value, which ensures that the residual of model is a white noise. ** denotes the 5% significance level. The critical values of Φ test are consulted, Table 2 of Wane et al. (2004).

We, then, apply the error correction model to check for the rigid property of loan rate. In order to avoid sample deficiency caused by estimated coefficients, 14

the insignificant coefficients are omitted before reestimation. The estimation of error correction model is illustrated with equation (12):

Banks and Bank Systems, Volume 3, Issue 4, 2008

) ) ∆lit = −0.046 + 0.012M it eit −1 − 0.051(1 − M it )eit −1 + 0.364∆lit −1 + 0.022∆mit −1,

(0.117) ( 0.817)

(0.048)

(0.000)

(0.058)

(12)

QLB (24) = 3.950 (1.000) ARCH (24) = 0.096 (1.000 ) ⎧1 if ∆σ t −1 ≥ 0.116 Mt = ⎨ ⎩0 if ∆σ t −1 < 0.116

The values in brackets are p-values, Q LB (24) and ARCH (24) are statistics of Ljung-Box autocorrelation test and ARCH-LM hetero-variance test with 24-period lag, respectively. Under the significance level of 1%, the residuals of the model do not have auto-correlation and heteroskedasticity variance, which shows that the model meets the stability requirement and that the adequate model is fitted. Comparing η1 and η 2 , |η€1 |=0.012 50%

Largest shareholder

Fig. 1. Bank shareholding restrictions

Tables 1 through 3 provide a full list of countries with this type of banking regulation as of 2003, created from the database of the World Bank Survey of Bank Regulation and Supervision2. Our paper examines whether this type of restriction restrains competition in an oligopolistic model of the loan market.

© Alexandra Lai, Raphael Solomon, 2008. 1 Although most OECD countries, apart from Australia, Canada, Luxembourg, and Norway, do not have formal restrictions on bank shareholding, some countries (including the United States, the United Kingdom, and Japan) exhibit widely held shareholding patterns for their largest publicly traded banks. This may suggest that norms in these countries constrain bank ownership concentration. 2 See Barth, Caprio, and Levine (2001) for more details.

16

Country

Turks 6



Table 1. Bank shareholding restrictions: high income countries

Notes: a – must have at least two shareholders; CB – Central bank; Comm. – commission; Exceed – the organization that may permit exceeding the limits; MF – Ministry of Finance; No – if none exists; Pres. – President; TR – Treasury.

Table 2. Bank shareholding restrictions: medium income countries Country

No. of

Share

Share

Exceed

banks Dec-01

Limits

Limits

People

Firms

Columbia

29

95%

95%

No

Costa Rica

21

a

a

No

China*

105

10%

10%

No

Egypt

53

10%

10%

CB

Guyana

7

20%

20%

No

Malaysia

25

10%

20%

No

Malta

15

5%

5%

CB

Mauritius

10

15%

15%

No

Mexico

32

20%

20%

No

Banks and Bank Systems, Volume 3, Issue 4, 2008

Table 2 (cont.). Bank shareholding restrictions: medium income countries Country

No. of

Share

Share

Limits

Limits

Exceed

banks Dec-01

People

Firms

2

20%

20%

MF/CB

Nicaragua

6

20%

20%

No

Oman

15

15%

25-35%

No

Philippines

42

40%

40%

Pres.

Montserrat

Puerto Rico

17

5%

5%

Comm.

Sri Lanka

25

10%

10%

CB

St. Kitt's/Nevis

6

20%

20%

No

St. Lucia

7

20%

20%

MF/CB

St. Vincent

5

20%

20%

MF/CB

Thailand

31

5%

5%

No

Turkmenistan

13

35%

35%

No

Ukraine

152

a

a

No

3

20%

20%

No

W. Samoa

Notes: a – must have at least two shareholders; CB – Central bank; Comm.– commission; Exceed – the organization that may permit exceeding the limits; MF – Ministry of Finance; No – if none exists; Pres. – President; TR – Treasury.

Table 3. Bank shareholding restrictions: low income countries Country

No. of

Share

Share

Limits

Limits

Exceed

banks Dec-01

People

Firms

Bangladesh*

50

10%

10%

No

Bhutan

2

20%

20%

No

Burundi

7

20%

20%

No

Fiji

6

15%

15%

No

Gambia

6

10%

10%

No

Georgia*

29

25%

25%

No

Grenada

5

20%

20%

MF

India

97

60%

60%

No

Kenya

46

25%

25%

No

Kyrgyzstan

20

15%

15%

No

Nepal*

13

49%

49%

No

Serbia/Montenegro

49

a

a

No

Sudan

25

10%

10%

No

Swaziland

4

25%

25%

No

Turkmenistan

13

35%

35%

No

Vietnam*

48

5%

5%

No

Zambia*

16

25%

25%

No

Zimbabwe

24

10%

25%

No

Notes: a – must have at least two shareholders; CB – Central bank; Comm.– commission; Exceed – the organization that may permit exceeding the limits; MF – Ministry of Finance; No – if none exists; Pres. – President; TR – Treasury.

Rules requiring dispersed shareholdings can cause several problems. They may deter foreign entry. They may also act as a poison pill, a mechanism to prevent hostile takeovers, without which banks

might have access to cheaper capital1. Finally, they may increase agency costs. This paper focuses on the latter problem. In an environment without shareholding restrictions, large shareholders may obtain control of banks to discipline management and to minimize agency costs. In so doing, they make the banking system more competitive (lower prices, higher output) and thus more efficient. In our model, large shareholders achieve this goal by issuing bank debt (taking uninsured deposits). In our game-theoretic model of two competing banks, managers make daily operating decisions (represented by the choice of loan output), but also divert a fraction of the bank’s residual cash flow. Either the manager or the controlling blockholder may choose the bank’s capital structure. To obtain control, the blockholder must engage in costly monitoring. Monitoring does not guarantee control; rather, it yields the blockholder control with probability (less than one) increasing in the number of shares held. The timing of the game is as follows: (i) potential blockholders simultaneously decide whether to acquire a controlling share of a bank and monitor management, (ii) the manager or the controlling blockholder chooses the capital structure of the bank, and (iii) managers compete in the market for bank loans. From a blockholder’s perspective, debt has two consequences. First, it “disciplines” a manager by reducing the amount of free cash flow from which the manager can divert funds. Second, it has a strategic effect vis-à-vis the other bank. Specifically, holding fixed the amount of debt at the rival bank, a unilateral increase in one bank’s debt increases its own output while reducing that of the other bank2. This raises the more indebted bank’s market share and profits at the expense of the other bank, since industry profits decline. In a symmetric Nash equilibrium, where both banks issue debt, each bank incurs lower profits from making loans than they would under coordinated actions. In our model, however, an increase in debt at both banks may increase bank value, even as profits fall, because debt transfers payoffs from the manager to the shareholders. Moreover, industry output is higher. We show that, since managers issue less debt than blockholders, the presence of controlling blockholders increases both firm value and competition in the loans market. Hence, if shareholding 1 See Gouvin (2001), Malatesta and Walkling (1988), and Ryngaert (1988). 2 Brander and Lewis (1986) demonstrate that an increase in debt causes profit-maximizing managers to compete more aggressively in the output market relative to the pure (debt-free) Cournot outcome.

17

Banks and Bank Systems, Volume 3, Issue 4, 2008

restrictions prevent the existence of blockholders, both firm value and competition in the loans market decrease, binding rule restricting ownership concentration creates two possibilities1. In the first, blockholdings never exist; in the second, blockholders exist but do not monitor and never gain control. The latter case is quite interesting, since it may prove challenging for banking regulators to determine that it is occurring, yet it may have the same effects on competition as if the blockholders did not exist! Our model is related to three distinct strands of literature. One strand relates capital structure to output. The key paper is by Brander and Lewis (1986), on which we draw extensively. Maksimovic (1988) models a repeated game in an oligopoly setting, in which collusive outcomes can occur. Debt holding can destroy the sustainability of collusion, leading to more competitive outcomes. Bolton and Scharfstein (1990) also relate debt financing to the aggressiveness of competition in a theory of predation. Dasgupta and Titman (1998) link pricing (and hence market share) decisions to capital structure through the effect of capital structure on the rate by which a firm discounts future profits. Campello (2003) finds capital structure empirically significant for explaining product market outcomes. A second strand of literature models ownership structure and/or capital structure as responses to agency problems. Within this strand, only Zhang (1998) links capital structure and ownership structure in a model with managerial risk aversion and inside ownership. Agency problems between management and shareholders can also take the form of empire building and diversion of perquisites2. Jensen (1986) notes that debt is a good antidote to managerial empire building and diversion of perquisites. In these models where capital structure is seen as alleviating agency problems, it also has an impact on firm value. A third strand relates ownership structure to firm value. Our results are consistent with the main message of this literature: a large blockholder increases firm value. Burkart and Panunzi’s (2001) model has a manager, a large shareholder, and some dispersed shareholders, where shareholders need to monitor to prevent managerial diversion of resources. Despite the conflict of interest between dispersed shareholders and the blockholder, it is always valueenhancing for the blockholder to win effective control of the firm, because this aligns the interests of shareholders and the manager. Burkart, Gromb, and Panunzi (1997) and Bolton and Von Thadden 1 As Figure 1 shows, 20 per cent is both the median and modal restriction in the sample of countries that have formal bank shareholding restrictions. 2 See Jensen and Meckling (1976) on agency problems.

18

(1998) consider the optimal ownership concentration as a response to agency problems between management and shareholders. Barclay and Holderness (1990) show empirically that the value of the firm increases if there is a blockholder, but that the increase is limited if the blockholder does not exercise control. In particular, actions that Barclay and Holderness (1990) interpret as monitoring, such as changing the composition of the board or replacing the management, yield the highest benefits in terms of firm value. Caprio, Laeven, and Levine (2004) provide empirical evidence of a positive relationship between ownership concentration and value for a sample of 244 publicly traded banks across 44 countries3. Our model links ownership structure to output through the choice of an optimal capital structure under agency problems, uniting these three strands of literature. A related paper examines the link between ownership structure and the incentive to acquire other firms: Allen and Cebenoyan (1991) allow for interaction between concentrated insider ownership and concentrated outsider ownership (the blockholder). In their sample of 58 American bank holding companies, they find that banks with a blockholder and no large inside shareholdings tend to be less acquisitive. To the extent that more competition results from fewer acquisitions, the blockholder without the inside shareholder may be said to be the most competitive ownership structure. We abstract from private benefits of control and focus exclusively on the existence of shared benefits of control. Barclay and Holderness (1992) and others find evidence of both shared and private benefits of control, and these generally depend on the size of the blockholding4. Holderness and Sheehan (1998), however, report evidence from the United States that large blockholders are constrained from expropriating cash flows and from other actions inimical to the interests of minority shareholders5. Pedersen and Thomsen (2003) also find that the type of blockholder – institutional investor, corporation, financial firm, or private individual – matters for control. They find that financial firms are most likely to assert control. In our model, we can interpret the idea of different propensities to assert control in terms of different (opportunity) costs of monitoring.

3 A related literature examines the effects of insider ownership, as opposed to the outsider ownership that we consider. See Stulz (1990) for a theoretical model and DeYoung, Spong, and Sullivan (2001), Morck, Shleifer, and Vishny (1988), and McConnell and Servaes (1990) for empirical evidence. 4 See also Barclay and Holderness (1989) and Mikkelson and Regassa (1991). 5 In particular, by citing the case of Turner Broadcasting (p. 8), the authors demonstrate that ownership does not always entail control, even when the blockholder owns a majority of the shares.

Banks and Bank Systems, Volume 3, Issue 4, 2008

This paper is organized as follows. Section 1 develops the model and summarizes the multi-stage game with a timeline of the model (Figure 2). In section 2, we solve for the subgame-perfect Nash equilibrium at all stages of the game. We draw out the policy implications of our analysis in section 3. The las section offers some conclusions. 1. The model There are two banks in the economy, indexed by i = 1, 21. Each bank i is run by a manager who holds no equity. Each bank has a potential blockholder who purchases a fraction, αi ≥ 0, of the bank’s shares. Atomistic shareholders own the remaining shares. If the potential blockholder declines to purchase shares (αi = 0), all shares are bought by dispersed owners. All economic agents are risk-neutral and maximize wealth. Managers choose output levels (of loans supplied) in a non-co-operative (Cournot) game. The manager also controls the choice of debt level if there is no blockholder. A blockholder, however, can influence the capital structure choice by monitoring, at cost c. A controlling blockholder chooses the level of debt issued by the bank. We assume that the proceeds of the risky debt issue are immediately distributed to shareholders as dividends. We assume that there are no conflicts of interest between the blockholder and dispersed shareholders, and we therefore focus on conflicts of interest between owners and managers. Managers are paid a fixed salary and also divert an exogenous fraction, φ , of banks' cash flow net of payments to debt holders (hereafter, the residual cash flow) for their own consumption Since the proceeds of the debt issue are distributed to shareholders, debt reduces the residual cash flow from which the manager diverts. A blockholder can thus “discipline” the manager through the choice of capital structure. The bank’s debt level also affects the manager’s output decisions. We model the output decisions of managers, capital structure decisions, and the shareholding and monitoring decisions of potential blockholders in a oneshot multi-stage game. The various stages of the model are described below. Stage 1: Blockholding and monitoring decisions At this stage, a potential blockholder acquires a share, α i ∈ [0, α max ] , of the bank and simultaneously decides whether to monitor. We represent legal restrictions on ownership concentration by αmax < 1. The decision to monitor depends on how likely monitoring leads to effective control and the bene1

The results are generalized to the case of a small number of banks.

fits of control2. Blockholders face uncertainty over whether their monitoring is successful. If the blockholder monitors, the blockholder wins control from the manager with a probability p(αi), a nondecreasing function of the size of blockholding, αi3. A controlling blockholder determines the capital structure of the bank by choosing the face value, Di, of debt to issue. With probability 1 - p(αi), the manager retains control and chooses the debt level. Stage 2: Capital structure decision At this stage, the controlling blockholder of bank i chooses the debt level, Di, to maximize the expected value of the firm, which can be decomposed into the value of the firm at stage 3 and the value of debt: V i ( Di , D j ) = V iE ( Di , D j ) + V iD ( Di , D j ).

Bank i’s equity and debt values depend not only on its own debt level but also on the level of debt issued by its competitor bank, Dj. The manager who retains control chooses debt to maximize the expected residual cash flow, which is equivalent to maximizing the value of the firm at stage 3. We interpret debt, D, as uninsured (wholesale) deposits. We also abstract from competition for deposits and the sequential-service nature of bank deposits. Whereas banks typically issue debt both to cover operational funding requirements and for strategic reasons, bank debt is purely strategic in this model. Banks issue debt to obtain an advantage in the loans market and to discipline their management. There are four different control structures: BB: Blockholders determine debt levels in both banks, D = (DBB, DBB). This occurs with probability p(αi)p(αj). MM: Managers determine debt levels in both banks, D = (DMM, DMM). This occurs with probability [1 - p(αi)][1 - p(αj)]. BM: Bank i’s blockholder and bank j’s manager choose debt levels, D = (DBM, DMB). This occurs with probability p(αi)[1 - p(αj)]. MB: Bank i’s manager and bank j’s blockholder choose debt levels, D = (DMB, DBM). This occurs with probability [1 - p(αi)]p(αj). 2

The benefits of control are endogenously determined in our model. Following Burkart and Panunzi (2001), we differentiate the rights of control from effective control. That is, blockholders have rights of control conferred by ownership, but may or may not choose to exercise control. Upon choosing to exercise control, there is some uncertainty whether they obtain effective control. This makes sense when α max < 0.5, so that blockholders need to obtain the right to vote the

proxies of dispersed shareholders. How easily they can do this depends on the size of their shareholdings, their monitoring effort, voting rules, and (potentially) luck. 3 We assume that p is non-decreasing in α but needs not be differentiable; we require p(0) = 0 for p(α) to be reasonable.

19

Banks and Bank Systems, Volume 3, Issue 4, 2008

We denote the debt choice by a controlling blockholder facing a blockholder in the other bank as DBB, and one facing a manager as DBM. Likewise, the debt choice by a manager facing another manager is denoted as DMM, while that of a manager facing a controlling blockholder in its competitor bank is DMB. The likelihood of any of these four control structures depends jointly on αi and αj. Stage 3: Output decision

bank is in default if and only if zi < z€i , implicitly defined by R i (q, z€i ) − Di = 0 . Then, bank i’s expected residual cash flow (net of debt repayment) is



1 z€i

[ R i (q i , q j , z i ) − Di ] dz i , the value of equity is

V iE = (1 − φ )



1 z€i

[ R i (q i , q j , z i − Di )]dz i ,

debt is V iD =

The managers play a Cournot game, taking Di and Dj as given. We denote bank i’s profit, net of the manager’s salary and payments to insured retail depositors, as Ri(qi, qj, zi), where q = (qi, qj) is a vector of loan quantities and zi, an independent and identically distributed state variable, is uniformly distributed over the unit interval. We let Rii denote the derivative of Ri with respect to qi, R ij the derivative of Ri



z€i 0

and the value of

R i (q i , q j , z i )dz i + (1 − z€i ) Di . It is

noteworthy that z€i is a function of qi, qj, and Di. Brander and Lewis (1986) show that z€i is increasing in Di and qi, and decreasing in qj, j ≠ i. Stage 4: Payoffs

i

Uncertainty is resolved (zi and zj are realized), profits are realized, and debt is repaid, if possible. If no default has taken place, the manager diverts a fraction, φ , of the bank’s residual cash flow, and shareholders obtain their share of the firm’s public value,

(concavity), R ij < 0 , Riji < 0 (loans are substitutes as

(1 − φ ) max R i − Di , 0 . If default occurs, debt hold-

with respect to qj, and Rz the derivative of Ri with respect to zi. Following Brander and Lewis (1986), we impose the following restrictions on Ri: Riii < 0 well as strategic substitutes, which means that a bank’s incentive to increase loans increases when the other bank reduces its loans), Rzi > 0, and Rizi > 0 (so that a higher realization of zi is beneficial for bank i). The manager of bank i chooses loan levels, qi, to maximize expected residual cash flow, which is equivalent to maximizing the value of equity at this stage. Let z€i be the critical value of zi for which the

{

}

ers get all of the bank’s profits, while the manager and shareholders get nothing (limited liability). The value of debt at time of issue (stage 2) is V iD , as defined above, and it accrues to shareholders. Hence, shareholders internalize the value of debt when choosing the optimal debt level at stage 2. Bank value and the price of bank shares at stage 1 are then E [V iE + V iD ].

2 banks, i and j Blockholding acquisition decision

Debt decision Debt face value, D

α ∈ [0, α max ]

Manager :

1

max ∫ [ R i (qi , q j , z ) − Di ] dz Di

z€i

1

1

0

z€i

Blockholder : max ∫ R i (qi , q j , z ) dz − φ ∫ [ R i (qi , q j , z ) − Di ] dz Di

Stage 1a

Stage 2

Stage 1b

Stage 3

Monitoring decision

Output decision

m ∈ {0, 1}, cost cm

1

Manager : max ∫ [ R i (qi , q j , z ) − Di ] dz

If monitor, control w. p. p (α )

Di

z€i

Fig. 2. Timeline of the model

2. Equilibrium In this section, we solve the model for subgameperfect Nash equilibria in pure strategies. That is, 20

we start with the last stage of the game (stage 3) to solve for equilibrium Cournot outputs, given debt levels. We then proceed to solve for equilibrium debt levels (stage 2) under the four possible control

Banks and Bank Systems, Volume 3, Issue 4, 2008

structures. Finally, we solve the game between potential blockholders, who simultaneously choose the size of shareholding to acquire and decide whether to monitor. Our analytical results are supplemented by results from numerical examples1

MB), the debt level is chosen to maximize the expected residual cash flow. The manager’s problem is as follows:

2.1. Equilibrium Cournot output. The manager at bank i faces the following maximization problem:

Taking the derivative with respect to Di yields



max qi ≥0

1 z€i

[ R i (q i , q j , z i ) − Di ] dz i .

(1)

Assuming an interior solution, the first-order condition is



1 z€i

[ Rii (q i , q j , z i ) − Di ] dz i = 0 .

(2)

In our proposition 1, we restate propositions 1 and 2 of Brander and Lewis (1986). Proposition 1. (i) For identical banks, equilibrium quantities are higher the higher are debt levels. That is, dq*/dD* > 0. (ii) For non-identical banks, Di ≠ Dj, a unilateral increase in bank i’s debt increases bank i’s equilibrium quantity and reduces bank j’s equilibrium quantity. That is, dq i* / dDi* > 0 and dq *j / dDi* < 0 .

2.2. Equilibrium debt levels. Whenever a manager is in control at bank i (control structures MM and max(1 − φ ) Di ≥ 0

+



z€i 0

max Di ≥ 0

R



i

1 0



1 z€i

(q i* ,

max Di ≥0



1 z€i

[ R i (qi* , q *j , z i ) − Di ] dz i .

− [1 − F ( z€i )] +



1 z€i

R ij (q i* , q *j , z i ) dz i

dq *j dDi

(3)

.

(4)

The first term is negative and reflects the decline in residual cash flow for every dollar of debt that is repaid. The second term is positive, because both R ij and dq *j / dDi* are negative. This represents the strategic effect of debt. A higher level of debt at bank i induces a decrease in the equilibrium quantity of loans at bank j. This increases bank i’s profits. The optimal debt level reflects these conflicting effects. A manager will choose a positive debt level only if the strategic effect of debt is sufficiently strong. Whenever a blockholder is in control at bank i (control structures BB and BM), debt level is chosen to maximize the value of the firm, allowing for the diversion that will happen after cash flows are realized. The blockholder’s problem is given in (5), or, equivalently, in (6):

[ R i (q i* , q *j , z i ) − Di ] dz i

(5) q *j ,

z i ) dz i + Di [1 − F ( z€i )],

R i (q i* , q *j , z i ) dz i − φ

1



z€i

[ R i (q i* , q *j , z i ) − Di ] dz i .

(6)

Taking the derivative with respect to Di while substituting the manager’s first-order condition yields



z€i 0

Rii (q i* , q *j , z i ) dz i

⎧⎪ −φ ⎨ ⎪⎩



1 z€i

R ij (q i* ,

q *j ,

dq i* + dDi



1 0

R ij (q i* , q *j , z i ) dz i

dq *j dDi

⎫⎪ − [1 − F ( z€i )]⎬ . z i ) dz i dDi ⎪⎭ dq *j

Since1 Rizi > 0, the first-order condition from the output stage, equation (2), implies that Rii < 0 for

zi ∈ [0, z€i ]. Hence, the first term of (7) is negative, because a higher level of debt, and the resulting higher output, reduce profits for low realizations of 1

While we use many parameterizations to determine the sensitivity of our results, we do not attempt to parameterize our model with actual data. Therefore, we note where a result is the outcome of those simulations. The appendix details our numerical solution procedure.

(7)

zi. The second term is positive, representing the strategic effect of debt. This strategic effect is larger for shareholders than for the manager, since it increases both the value of equity and the value of debt. In our model, shareholders internalize the value of debt, whereas managers disregard it. The expression inside the brackets in the final term of (7) is simply the derivative of the manager’s objective function. We conjecture that, evaluated at the optimal debt level for the blockholder’s problem, this final term is positive. This term represents the mar21

Banks and Bank Systems, Volume 3, Issue 4, 2008

ginal reduction in the amount the manager can divert from cash flow, the disciplining effect of debt1. That is, while the manager increases debt up to the point where the marginal contribution of debt to cash flow is zero, a blockholder increases debt beyond that point to discipline the manager. To compare optimal debt levels arising from the managers' and the blockholder’s problems, we introduce the idea of debt capacity in the context of this model. Debt capacity of bank i is some maximum debt level, denoted as D . We assume that both the manager’s and the blockholder’s problems have unique interior solutions: both the expected residual cash flow as well as the value of the firm are concave in their own debt levels. We also assume that the value of a bank’s debt is concave in the level of that debt. Our assumptions above guarantee that each bank has a unique debt capacity. Proposition 2. Given that the bank is not at its debt capacity, blockholders always prefer to issue a higher level of debt than managers would. That is, DBB > DMB and DBM > DMM. Furthermore, if debt levels across banks are strategic substitutes, then DMB ≤ DMM ≤ DBB ≤ DBM. Proof. See the appendix. The fact that blockholders always choose a higher debt level indicates that debt’s disciplining effect is present. While we are unable to show analytically that debt levels are strategic substitutes in our model, our numerical examples indicate that debt levels are indeed strategic substitutes, that is, whenever the debt level of bank i rises, that of bank j falls. We are interested in the effects of ownership regime on competition, or industry output. The next proposition deals with this relationship. Denote a bank’s output and debt choice as q x and D x , where x ∈ {BB, BM , MB, MM } indicates the control struc-

Proposition 3 follows because Cournot reaction functions have negative slopes greater than -1, a condition that is required for a stable Nash equilibrium. We also care about how the ownership regime affects the bank’s performance, as measured by the bank’s value. Since we are unable to obtain analytical results, we defer this discussion until section 3. It is not obvious a priori whether a control structure with higher debt levels results in a higher bank value. This is because a higher debt level, while “disciplining” managers, also leads to more competition in the loans market and the latter reduces profits, all things equal. 2.3. The blockholding and monitoring decision. To obtain the Nash equilibrium for this stage of the game, we first derive the conditions under which a blockholder will monitor. Returns to monitoring, measured in terms of firm value, depend not only on whether the blockholder wins control, but also on whether the other bank has a controlling blockholder. Hence, if the blockholder at bank i decides to monitor, expected payoffs are p(α i ) p(α j )V BB + [1 − p(α i )][1 − p(α j )]V MM +

p (α i )[1 − p(α j )]V BM +

[1 − p(α i )] p(α j )V MB .

wise, expected MB p(α j )V + [1 − p (α j )]V MM .

Other-

payoffs are The difference be-

tween the two has to be greater than c to induce monitoring by the blockholder. That is, a blockholder at bank i monitors if and only if ⎧⎪ p (α j )(V BB − V MB ) + ⎫⎪ α i p(α i ) ⎨ ⎬ ≥ c BM − V MM )⎪⎭ ⎪⎩+ [1 − p (α j )](V

(8)

The derivative of the left side of (8) with respect to αi is positive; thus, the condition for monitoring can be expressed in terms of a critical blockholding size (concentration):

ture. Also, denote industry output as Q x .

α i ≥ α (α j ).

Proposition 3. Industry output is highest when both banks have controlling blockholders, and lowest when both banks have managers in control of the capital-structure decision. That is, QMM < QBM < QBB, where QMM is industry output when both banks are manager controlled (M-controlled), QBM is industry output when one bank is blockholder controlled (B-controlled) and one bank is M-controlled, and QBB is industry output when both banks are Bcontrolled.

This critical shareholding size is a function of the ownership concentration at the other bank, αj.

(9)

Lemma 1. (i) Given that monitoring occurs, α i ≥ α (α j ), the value of the bank is increasing in

α i . (ii) If V BB + V MM − V MB − V BM < 0, the critical concentration necessary to induce monitoring is increasing in the other bank’s concentration: α ′(α j ) > 0. Otherwise, the reverse is true. Proof. See the appendix.

1

This is true whenever the manager has an incentive to choose a positive debt level. In all our numerical simulations, we obtain interior solutions for the manager’s debt choice.

22

Whether V BB + V MM − V B − V M is positive or negative corresponds to whether monitoring decisions are strategic substitutes or complements. Our nu-

Banks and Bank Systems, Volume 3, Issue 4, 2008

merical simulations suggest that this expression is often negative. Lemma 1 implies that, if monitoring takes place, the blockholder prefers the highest possible concentration (up to the point where p′(α ) = 0 ), in order to maximize the chances of winning control. This is because firm value increases whenever the blockholder wins control and is able to determine the capital structure of the bank. The first part of Lemma 1 leads us to the following conclusion. Corollary 1. A blockholder who monitors acquires a shareholding of size αmax.

held in equilibrium. Industry output is the same in cases II(a) and III. A relaxation of ownership restrictions, by increasing α max , increases industry output only in cases II(b) and I. However, it has no effect on industry output in cases II(a) and III unless the change results in α max ≥ α (α max ). Proof. See the appendix. 3. Numerical results

For our numerical examples, bank profits are given by R i (qi , q j , z i ) = z i (1 − qi − γ q j )qi − e − zi qi2 , (10)

Let V (α i , α j ) be the value of bank i as a function of

where zi (1 − qi − γ q j ) is the (linear) inverse de-

both banks' shareholding concentrations. By the above corollary, V takes four values:

mand function for bank i’s loans,

V (α max , α max ), V (α max , 0), V (0, α max ), and

V (0, 0).

The block-acquisition game has the following normal-form representation: B NB max max V (α ,α ), V (α max , 0), B V (α max , α max ). V (0, α max ). NB

V (0, α max ),

V (0, 0),

max

V (0, 0).

V (α

, 0).

where B denotes the action to acquire a blockholding and NB denotes the action not to acquire a blockholding. The Nash equilibrium outcome of this game depends on whether α (α ) is increasing or decreasing in α , and on how binding are legal restrictions on ownership. The following cases must be considered: I.

Non-restrictive ownership max α (0), α (α max ) ≤ α max .

{

}

constraints:

II. Moderately restrictive ownership constraints: (a) α (0) ≤ α max < α (α max ). This case is relevant only if V

BB

+V

MM

−V

MB

−V

BM

< 0.

(b) α (α max ) ≤ α max < α (0). This case is relevant only if V BB + V MM − V MB − V BM > 0. III. Highly restrictive ownership α max < min α (0), α (α max ) .

{

}

constraints:

Proposition 4. The Nash equilibrium in cases I and II(b) yields the outcome where both banks have blockholders who monitor their managers and produce a higher expected output relative to the other two cases. Case II(a) produces a Nash equilibrium in which blockholders exist at both banks but neither monitors. In case III, the legal constraints on ownership are so binding that both banks are widely

C i (qi , zi ) = e − z i qi2 is bank i’s (quadratic) cost func-

tion, and γ ∈ (0,1] is a measure of substitutability between bank loans1. Loan levels are restricted to those combinations that ensure a non-negative price. This functional form satisfies the restrictions from section 12. 3.1. Equilibrium values. In this numerical example, debt levels are strategic substitutes. In accordance with Proposition 2, blockholders always issue more debt than managers. In section 2, we show that a blockholder facing a competitor bank with a manager in control issues more debt than one facing another controlling blockholder. In this specification, managers' debt choices do not differ greatly whether they are facing another manager or a blockholder. Result 1. In equilibrium, 0 < DMM ≈ DMB < DBB < < DBM.

Furthermore, equilibrium debt issue is lower the more substitutable are bank loans. On average, a manager’s debt issue is most responsive to changes in γ, while the debt issue by a blockholder facing a manager is the least responsive. Result 2. In equilibrium, dD MM dD MB dD BB dD BM ≈ < < < 0. dγ dγ dγ dγ

We obtain the above result from regressions with data generated from the numerical simulations. Managers' debt choices are sensitive to changes in the substitutability between bank loans because this directly impacts profitability, and managers care only about profits. Blockholders are concerned with 1

Given this specification, one interpretation of z is as a shock common both to bank revenues and to bank costs, such as an exchange rate shock or a macroeconomic shock. Since the shock affects revenues linearly and costs non-linearly, revenues and costs are imperfectly correlated.

2

Riii

Rzi

= − 2( zi + e − z i ) < 0 , R ij = (1 − qi − γ q j ) qi + e − z i qi2 >

= − ziγ qi < 0 , Riji

= − zi γ < 0 ,

0 , Rizi = (1 − 2qi − γ q j ) + 2e − z i qi > 0

23

Banks and Bank Systems, Volume 3, Issue 4, 2008

mitigating the agency problem between managers and shareholders as well as with profitability. This second concern reduces the blockholder’s sensitivity towards changes in γ. Figure 3 plots equilibrium debt levels against γ; DMM and DMB do not coincide exactly, due to random deviations arising from the numerical solution procedure. However, we find the difference between the two values to be statistically insignificant1. Figure 4, which plots equilibrium industry output against γ, demonstrates Proposition 3. Debt levels (phi =0.2) 0,12

0,1

value of an M-controlled bank facing an Mcontrolled bank as VMM. Result 3. In equilibrium, VMB < VMM < VBB < VBM.

The result is demonstrated in Figure 5, which plots equilibrium firm values against γ. The value under a controlling blockholder is always higher due to the benefits of control; that is, the blockholder mitigates agency problems, and this is value-increasing. The blockholder who faces a manager-controlled bank benefits even more from the fact that the manager issues less debt, enabling the blockholder-controlled bank to gain market share. Likewise, the manager who faces a blockholder-controlled bank loses market share, and thus firm value is lower than it would be if the rival firm was manager-controlled. 0.07

0,08 0.06

DBM DMB

0,06

DMM

0.05

DBB VBM

0,04

VMB

0.04

VMM VBB

0,02

0.03

0.02

0 0

0,2

0,4

0,6

0,8

1

1,2

Gamma 0.01

Fig. 3. Debt levels for different control structures, degrees of substitution 0 0

0,6

0.2

0.4

0.6

0.8

1

1.2

Gamma

Fig. 5. Firm values for different control structures, degrees of substitution

0,5

0,4

QBM QMM

0,3

QBB

0,2

0,1

0 0

0,2

0,4

0,6

0,8

1

1,2

Gamma

Fig. 4. Output levels for different control structures, degrees of substitution

Our numerical examples lead us to the following conclusion regarding the relationship between ownership regime and firm value. Denote the value of a B-controlled bank facing a B-controlled bank as VBB, the value of a B-controlled bank facing an Mcontrolled bank as VBM, the value of an M-controlled bank facing a B-controlled bank as VMB, and the 1

Our statistical inference assumes that the deviations due to the numerical procedure are normally distributed around a mean of zero.

24

3.2. Policy implications. Our analysis demonstrates that legal restrictions on the concentration of ownership can affect bank value and competition in the loans market. Marginally relaxing this restriction will have an effect only in cases where the restriction has not prevented blockholding and monitoring to occur in the first place (case I). If ownership restrictions are binding, so that they either prevent blockholding or they prevent monitoring even in the presence of blockholdings, then a marginal increase in the maximum shareholding will generally not have any effect on bank value or competition in the loans market. For a relaxation of bank shareholding restrictions to be beneficial, the increase in maximum shareholding may need to be substantial.

It is worth highlighting Case IIb: the shareholding restriction does not prevent the existence of blockholders, but these blockholders exist only to prevent the other blockholder from monitoring. No monitoring occurs in these situations. That is, society does not derive any benefits from ownership concentration. Hence, if there are (unmodelled) costs to ownership concentration, case IIb is associated with lower social welfare relative to case III,

Banks and Bank Systems, Volume 3, Issue 4, 2008

in which no blockholders exist. A relaxation of shareholding restrictions that induces a shift from case III to case IIa results in a net decrease in social welfare. However, a complete abolition of shareholding restrictions, or an increase in maximum allowable shareholding that is sizable enough to ensure we obtain case I, can be socially beneficial1. Our model abstracts from other conflicts of interest between equity holders and debt holders (riskshifting) and between blockholders and minority shareholders (self-dealing). While the problem of risk-shifting is particularly relevant to highly leveraged institutions such as banks, capital requirements and positive franchise values mitigate the problem. Moreover, this problem is associated with leverage and not concentration ownership per se. Although the original economic justification for bank shareholding restrictions might have been to prevent self-dealing, these regulations are a relatively old phenomenon, dating back to the 1960s in some countries. Since that time, two important developments bear mention. First, in the 1980s and 1990s, there was a revolution in corporate governance in the banking sector, as well as more generally. This revolution included changes such as an increased emphasis on outside directors, new rules for electing boards, and more internal oversight. Second, in the post-Basel era, there is increased supervision of banks, particularly large, multinational banks. Taken together, these phenomena vastly reduce the scope for self-dealing by blockholders. The justification for these restrictions, while fairly universal in the 1960s, largely does not exist in most industrialized countries today. Conclusion

This paper examines whether a restriction on ownership concentration affects competition in the bank

loan market. Our analysis demonstrates that legal restrictions on the concentration of ownership can affect bank value and competition in the loans market. Marginally relaxing this restriction will have an impact only in cases where the restriction has not prevented blockholding and monitoring from occurring in the first place. If ownership restrictions are severe enough to prevent blockholding or monitoring (even if blockholders exist), then a marginal increase in the maximum shareholding will generally not have any impact on bank value or competition in the loans market. For a relaxation of bank shareholding restrictions to be beneficial, the increase in maximum shareholding may need to be substantial. Ownership concentration matters in our model because it provides the incentives to the blockholder to engage in costly monitoring, a necessary step to obtaining the right to set the level of unsecured debt. We show that blockholders always issue more debt than managers. Debt is socially beneficial in two ways: it alleviates the agency problem and it is procompetitive, because a higher level of debt induces the manager to compete more aggressively in the loans market. Hence, increased ownership concentration with monitoring creates a more competitive banking industry. We do not, however, model a cost to ownership concentration. In particular, we assume that there is no conflict of interest between blockholders and atomistic shareholders. We also do not model a mechanism by which control by the blockholder may adversely affect managerial incentives. A more balanced analysis might introduce a trade-off to ownership concentration or to debt issue. Hence, we might extend our analysis in two ways. In one approach, we can introduce some selfdealing by blockholders. In the other, we can model debt as impairing managerial incentives to exert effort that may raise firm profitability. Further research on these issues may prove fruitful.

References1

1. 2. 3. 4. 5. 6. 7.

1

Allen, L. and Cebenoyan, A.S. (1991). “Bank Acquisitions and Ownership Structure: Theory and Evidence”. Journal of Banking and Finance, Vol. 15, pp. 425-48. Barclay, M.J. and Holderness, C.G. (1989). “Private Benefits from Control of Public Corporations”. Journal of Financial Economics, Vol. 25, pp. 317-95. Barclay, M.J. and Holderness, C.G. (1990). “Negotiated Block Trades and Corporate Control”. Journal of Finance, Vol. 46, No. 3, pp. 861-78. Barclay, M.J. and Holderness, C.G. (1992). “The Law and Large-Block Trades”. Journal of Law and Economics, Vol. 35, No. 2, pp. 265-94. Barth, J., Caprio, G. and Levine, R. (2001), The Regulation and Supervision of Banks around the World: A New Database. Policy Research Working Paper No. 2588, The World Bank. Bebchuk, L.A., Fried, J.M. and Walker, D.I. (2002). “Managerial Power and Rent Extraction in the Design of Executive Compensation”. University of Chicago Law Review, Vol. 16, No. 3, pp. 751-846. Bebchuk, L.A. and Fried, J.M. (2003). “Executive Compensation as an Agency Problem”. Journal of Economic Perspectives Vol. 17, No. Summer, pp. 71-92.

This causes an increase in industry output (from QMM to QBB) and an increase in bank value (from VMM to VBB).

25

Banks and Bank Systems, Volume 3, Issue 4, 2008

8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32.

Bolton, P. and Scharfstein, D.S. (1990). “A Theory of Predation Based on Agency Problems in Financial Contracting.” American Economic Review, Vol. 80, No. 1, pp. 93-106. Bolton, P. and Thadden, E.-L.V. (1998). “Blocks, Liquidity and Corporate Control.” Journal of Finance, Vol. 53, No. 1, pp. 1-25. Brander, J. and Lewis, T. (1986). “Oligopoly and Financial Structure”. American Economic Review, Vol. 76, pp. 956-70. Burkart, M., Gromb, D. and Panunzi, F. (1997). “Large Shareholders, Monitoring, and the Value of the Firm”. Quarterly Journal of Economics, Vol. 12, No. 3, pp. 693-728. Burkart, M. and Panunzi, F. (2001). Agency Conflicts, Ownership Concentration, and Legal Shareholder Protection. Discussion Paper No. 2708, Centre for Economic Policy Research. Campello, M. (2003). “Capital Structure and Product Market Interactions: Evidence from Business Cycles”. Journal of Financial Economics, Vol. 68, pp. 353-78. Caprio, G. , Laeven, L. and Levine, R. (2004). Governance and Bank Valuation. Policy Research Working Paper No. 3202, World Bank. Chaplinsky, S. and Niehaus, G. (1993). “Do Inside Ownership and Leverage Share Common Determinants?” Quarterly Journal of Economics and Business, Vol. 32, No. 4, pp. 51.65. Dasgupta, S. and Titman, S. (1998). “Pricing Strategy and Financial Policy”. Review of Financial Studies, Vol. 11, pp. 705-37. Denis, D. J. and Sarin, A. (1999). “Ownership and Board Structures in Publicly Traded Corporations”. Journal of Financial Economics, Vol. 52, pp. 187-223. DeYoung, R., Spong, K. and Sullivan, R.J. (2001). “Who’s Minding the Store? Motivating and Monitoring Hired Managers at Small, Closely Held Commercial Banks”. Journal of Banking and Finance, Vol. 25, pp. 1209-43. Gouvin, E.J. (2001). “The Political Economy of Canada’s Widely Held Rule for Large Banks”. Law and Policy in International Business, Vol. 32, pp. 391-426. Holderness, C.G. and Sheehan, D.P. (1998). Constraints on Large-Block Shareholders. Working Paper No. 6765, National Bureau of Economic Research. Jensen, M.C. (1986). “Agency Costs of Free Cash Flow, Corporate Finance and Takeovers”. American Economic Review, Vol. 76, No. 2, pp. 323-9. Jensen, M.C. and Meckling, W.H. (1976). “Theory of the Firm: Managerial Behavior, Agency Costs and Ownership Structure”. Journal of Financial Economics, Vol. 3, pp. 305-60. Maksimovic, V. (1988). “Capital Structure in Repeated Oligopolies”. Rand Journal of Economics, Vol. 19, No. 3, pp. 389-407. Malatesta, P.H. and Walkling., R.A. (1988). “Poison Pill Securities: Stockholder Wealth, Profitability and Ownership Structure”. Journal of Financial Economics, Vol. 20, pp. 347-76. McConnell, J.J. and Servaes, H. (1990). “Additional Evidence on Equity Ownership and Corporate Value”. Journal of Financial Economics, Vol. 27, pp. 595-612. Mikkelson, W.H. and Regassa, H. (1991). “Premiums Paid in Block Transactions”. Managerial and Decision Economics, Vol. 12, No. 6, pp. 511-7. Morck, R., Shleifer, A. and Vishny, R.W. (1988). “Management Ownership and Market Valuations: An Empirical Analysis”. Journal of Financial Economics, Vol. 20, pp. 293-315. Pedersen, T. and Thomsen, S. (2003). “Ownership Structure and Value of the Largest European Firms: The Importance of Owner Identity”. Journal of Management and Governance, Vol. 7, pp. 27-55. Phillips, A. (1964). “Competition, Confusion, and Commercial Banking”. Journal of Finance, Vol. 19, No. 1, pp. 32-45. Ryngaert, M. (1988). “The Effect of Poison Pill Securities on Shareholder Wealth”. Journal of Financial Economics, Vol. 20, pp. 377-417. Stulz, R. (1990). “Managerial Discretion and Optimal Financing Policies”. Journal of Financial Economics, Vol. 26, pp. 3-27. Zhang, G. (1998). “Ownership Concentration, Risk Aversion and the Effect of Financial Structure on Investment Decisions”. European Economic Review, Vol. 42, pp. 1751-1778.

Appendix 1. Proofs of Lemma and Propositions Proof of Proposition 2. We denote a bank’s debt capacity as Di = arg max Di V iD ( Di , D j ), where V

iD

is the value of

debt for bank i, given Cournot equilibrium quantities, as functions of debt levels. That is, V iD =



z€i

0

R i (qi* , q *j ) dz + (1 − z€i ) Di . iB

Furthermore, note the following relationship between firm value (V ), debt value (V

((1 − φ )V V

iB

26

=V

iD

iM

, where (V

+ (1 − φ )V

iM

.

iM

) is the manager’s objective function:

iD

), and equity value

Banks and Bank Systems, Volume 3, Issue 4, 2008

To assume that the bank remains within its debt capacity at all optimal debt levels (for the blockholder and for the manager) is equivalent to assuming that we obtain interior solutions for both the blockholder’s and the manager’s problems. This implies that ∂V iB ∂Di

∂V iD ∂Di

= DiM ( D j )

≥ 0, and DiM ( D j )

∂V iM ∂Di

=− DiB ( D j )

∂V iD ∂Di

≤ 0, DiB ( D j )

where DiM ( D j ) is the manager’s debt reaction function and DiB ( D j ) is the blockholder’s debt reaction function. Hence, DiM ( D j ) ≤ DiB ( D j ).

(11)

If debt levels are strategic substitutes and Nash solutions are stable, then ∂DiM ( D j ) < 0, −1 < ∂D j −1
0.

Proof of Proposition 4. In the highly restrictive situation (case III), no blockholder can be given the incentives to monitor, even by the largest block possible. Hence, the unique Nash equilibrium is (NB, NB), with both banks being widely held and industry output at QMM, the lowest level of all the cases we consider. In the non-restrictive situation (case I), the unique Nash equilibrium is (B, B), where each bank has a blockholder who

has a share,

α max ,

of the bank and monitors. Industry output is a function of who (blockholder or manager) wins con-

trol. Expected industry output is [ p (α max )] 2 Q BB + 2 p(α max )[1 − p(α max )]Q BM + [1 − p(α max )] 2 Q MM . The derivais positive for all p (α ) ≤ 1 and p′(α ) > 0. Hence, an increase in the legally tive of this with respect to α allowable ownership concentration increases competition and industry output. In the moderately restrictive situation (case II), we first consider the case of VBB + VMM – VMB – VBM < 0, since this is max the more common case (for all but one) in our numerical simulations. In this case, a blockholder of size α will not monitor if faced by another blockholder of the same size. That blockholder will monitor, however, whenever the other max bank is widely held. Thus, the outcome (B, B) involves blockholders in both banks who hold α of their banks but do not monitor because this shareholding does not provide the critical ownership level to induce monitoring, max

max

max

α max < α (α max ). The bank’s value is therefore the same as for widely held banks: V (α

max

, α max ) = V (0, 0) =

V MM . If one bank, i, is widely held but the other bank has a blockholder of size α max , that blockholder will monitor, max < α (0). Hence, V (α max , 0) = p(α max )V BM + [1 − p(α max )]V MM and V (0, α max ) = p(α max )V MB + since α [1 − p(α max )]V MM . Therefore, if V (0, 0) – V (0, α max ) = p(α max )(V MM − V MB ) > 0, the Nash equilibrium is (B,B). This is indeed the case (from Proposition 5): industry output is QMM and an increase in industry as long as α BB

max MM

< α (α MB

max

). An increase in α

BM

α max

has no effect on

α (α ) puts us into the non-restrictive case (I). max > 0. Then, a blockholder of size α will monitor if the other bank has a blockmax

beyond

max

Suppose that V + V – V – V holder of the same size, but will not monitor if the other bank is widely held. Hence, V (α max , α max ) =

[ p (α max )] 2 V BB + p(α max )[1 − p(α max )](V BM + V MB ) + [1 − p (α max )] 2 V MM , while V (0, 0) = V (α max , 0) = V (0, α max ) = V MM . Since V (α , α ) – V (0, 0) > 0 for all p(α ) < 1, the Nash equilibrium is (B, B), in which both banks have blockholders who monitor. Expected industry output is the same as in the non-restrictive situation (case I). max

max

max

27

Banks and Bank Systems, Volume 3, Issue 4, 2008

2. Algorithm to solve for the Nash equilibria of stages 2 and 3 of the game The solution is computed for fixed values of φ and γ. The solutions of type BB, MM, and BM/MB are each computed separately, in order to take advantage of the symmetries in the first two cases. The first step is to specify a parameter space for the quantities of loans, qi and qj. Since quantities cannot be negative, we use the closed interval [0, 10] and partition it into 4,000 points. For each point (q i , q j ) ∈ [0, 10] × [0, 10], we compute the following quantity: z s = min z (α − qi − γ q j ) qi − e − z qi2 .

Let

z ∈ [ 0, 1]

{

Q = (q i , q j ) ∈ [0, 10] 2

z€s = z s (α − q i − γ q j ) q i − e − zs q i2 .

We

define

the

space

z s < 0.99 and z€s > 0.01 } . This process of trimming the endpoints of the unit interval makes

the solution more efficient. We specify a space for debt, DS, which is usually [0, 0.15] at the first stage. Next we fix a debt level, Di ∈ DS , and compute zi for each (qi, qj) pair. For each qj, we find qi , which maximizes the objective *

function



1 z€i

[ R − D ] dz i . The value

qi* depends on Di and qj. We repeat this process for all qj in [0, 10], and then restart

the process for another Di searching over a large space for debts issued. We eliminate those triples (qi, qj, Di) that fail to satisfy the second-order conditions of the output-stage maximization problem. We then examine pairs ( Di , D j ) ∈ DS . For each pair, we look at the Nash quantities q i* (q j , Di ) and search for 2

matched pairs. For example, suppose we are looking for the Nash equilibrium in quantities associated with debt levels (0.05, 0.03). We look at the sets of Nash quantities from Di = 0.05, calling them qi* (i ) , q *j (i ) . We label the sets of Nash quantities from Dj = 0.03 as

(

qi*

( j) ,

q *j

)

( j ) . We select pairs

{(

(

qi* (i ) ,

q *j (i )

) ), ( q ( j ) , q * i

* j(

)}

j ) , where

q*j (i ) = qi* ( j ), and look for ones where qi* (i ) = q*j ( j ). Any pair such that qi* (i) = q *j ( j ) is a Nash equilibrium in quantities. This may seem an unusual computational approach to determine a Nash equilibrium, but it relies on the fact that both banks have the same objective function (with the variables qi and qj switched). We then have, for all ( Di , D j ) ∈ DS 2 , q i* ( Di , D j ) and q *j ( Di , D j ) . The next step is to compute the Nash equilibrium debt strategy. The computational approach depends on whether the solution is for case BB, MM, or BM/MB. Since the BB and MM cases rely on symmetries, we focus on the asymmetric equilibrium for debt choice. We compute the values of the managers' objective functions and the blockholders' objective functions for banks i and j. For each level of debt, Dj, we determine the best response, Di* (i, D j ), based on the objective function of bank i. We repeat the process and compute D *j ( j , Di ), using the other objective function. The two reaction functions may have a solution, which is a Nash equilibrium. It is likely that there is no solution in the first iteration. Since it takes such a long time to compute optimal quantities given debt levels, we do not search over many debt levels at a time. In the event that no solution appears, we fit linear regressions to the reaction functions, solve for the intersection, and continue our search for the optimal debt level in the neighborhood of that intersection. Even if we do find a solution to the two reaction functions, the solution is imprecise. We 2 typically “zoom in” on the part of the space DS in which the “equilibrium” appears, in order to gain more precision. 3. Algorithm to solve for the Nash equilibria of stage 1 of the game The first step is to partition the Nash equilibria of the debt stage into two sets. Let F = VBB + VMM – VMB – VBM, where V is the value of the firm. We determine the boundaries of the three cases for F > 0 and F < 0 separately. We need to specify a parameter space for c, the cost of monitoring. Based on the values of the firm computed above, we use [0.0001, 0.0020], partitioned into 20 subintervals. We consider three functions for p (α ) : ⎧⎪0.996 α 2 , 0 ≤ α ≤ 0.3932 p1 (α ) = α , p 2 (α ) = ln (1 + α ), and p 3 = ⎨ . ⎪⎩ln (α + 0.774), 0.3932 ≤ α ≤ 1

Each of these three functions has a behavioral interpretation. p1 represents the idea that only sole proprietors can do what they want with their firms; even a majority shareholder may face legal opposition from determined minority shareholders. Both p2 and p3 have the property that p(1) < 1; a sole proprietor may not obtain full control if there are legislative restrictions or other random barriers. But the critical difference between p2 and p3 is that p2 is globally concave, whereas p3 is convex on the subset [0, 0.3932] and concave on the subset [0.3932, 1]. The convex-concave function captures the idea that, if a blockholder has a small block of shares, it may be almost infeasible to win control, but there is a threshold after which control becomes likely. Note that p3 is a continuous function on [0, 1], although not continuously differentiable. For each function, each cost, and each (φ , γ ) pair, there is a partition of α-space that defines the boundaries of the three cases. We compute α (0) and α (α ) for all α ∈ [0, 1] partitioned into 1,000 subintervals. From these calculations, the boundaries are apparent.

28

Banks and Bank Systems, Volume 3, Issue 4, 2008

Meng-Fen Hsieh (Taiwan), Chung-Hua Shen (Taiwan), Chien-Chiang Lee (Taiwan)

Bank provisioning, business cycles and bank regulations: a comprehensive analysis using panel data Abstract

The empirical analysis is performed through the system of Generalized Method of Moments (GMM) dynamic panel data estimations on a panel of 49 countries observed in the period of 1991-2002. The evidence shows that with steady growth in both the economy and bank earnings, bank management will tend to increase loan loss provisions (LLP), whereas with a buoyant economy but negative growth in bank earnings, management will exhibit an inclination to reduce LLP. Regarding the influence of bank regulation on provisions, the evidence shows that under certain circumstances banks make more provision based on regulatory considerations. This explains why bank regulations regarding LLP across countries do have an effect on the banks’ provisioning behavior. Keywords: loan loss provisions, pro-cyclicality, income-smoothing, dynamic panel data. JEL Classification: G21, E58.

Introduction♦

The relationship between bank loan loss provisions (LLP) and business cycles has been focus of a great deal of empirical study of late, in large part because of the promulgation of the Basel II Accord at the end of the year 2006. Under Basel II, banks are required to provide sufficient capital or reserve in accordance with their clients’ probability of default. The default risk of an enterprise increases when the external economy experiences a decline; this is marked by an upsurge in banks’ LLP. Against this, a sudden leap in the external economy decreases the probability of defaults, which then induces banks to decrease their LLP. The general expectation, therefore, is that bank loan loss provisioning becomes pro-cyclical. The concept of pro-cyclicality1, when applied to the new capital requirements, is that in a downturn, for instance, when risks are more likely to materialize, capital requirements might increase. Thus, capital requirements and output growth will move in opposite directions (Ayuso et al., 2004). The implication here is that when an economy faces a downturn, banks are apt to increase their loan loss reserves, meaning that they probably do not provide sufficient LLP in good years to save for bad years. In fact, numerous researchers provide evidence that many banks intend to increase their loan loss provisioning when the economy is in a downward trend. Among these are Ayuso et al. (2004) with a Spanish data who find the negative relationship between capital buffers and business cycle. Other country studies are, among others, Estrella (2004), Lindquist (2004), and Rime (2001) who analyze the capital pro♦

© Chung-Hua Shen, Meng-Fen Hsieh, Chien-Chiang Lee, 2008. To avoid confusion, in what follows the movement in a financial indicator is said to be “pro-cyclical” if it tends to amplify business cycle fluctuations. According to this definition, for instance, provisions behave pro-cyclically if they fall in economic upswings and rise in downswings (Borio et al., 2001). 1

cyclicality in a European context, respectively. And Laeven and Majnoni (2003) with a 45-country sample2, on the weight of this evidence find that the negative relationship between LLP and economic growth seems to be confirmed. Business cycles aside, bank earnings positively affect LLP. The consensus is that there is a positive relationship between LLP and earnings, so called income-smoothing effect. Greenwalt and Sinkey (1988) investigate whether large bank-holding companies employed their loan-loss provision to smooth accounting earnings. The discretionary nature of the estimation process and its use over successive periods provide managers with the opportunity to smooth income. Such behavior might be exhibited by charging additional amounts to loan-loss expenses in years of peak earnings while decreasing the loan-loss provision or delaying recognition of write-offs when earnings are down3. This has been substantiated more by Collins et al. (1995) and Kanagaretnam et al. (2004), among others, who find evidence in suppert of the income-smoothing effect. By contrast, Moyer (1990), Wetmore and Brick (1994), Beatty et al. (1995) and Ahmed et al. (1999) do not support the income-smoothing effect. They examine LLP as a mean to manipulate a bank’s capital adequacy. A manager can increase the primary capital adequacy by increasing the LLP4. More spe2 In addition to the research studies mentioned here that have empirically verified the pro-cyclicality of banks’ provisioning behavior, Borio et al. (2001) use a simulation approach to investigate the important links between financial development and business cycles; loan loss provisioning is another one of the factors. Lowe (2003) also investigates the issue during periods when the economy is in good shape, especially under circumstances marked by a dramatic growth in loans and a rapid increase in the credit risk of bank loans. 3 See page 304 for more details. 4 The other reason is based on the credibility of a bank/company, as suggested by Greenawalt and Sinky (1988), Fudenberg and Tirole (1995) and Defond and Park (1997), among others. Furthermore, rather than just serving as window dressing for a financial report, Kim and Santomero (1993) argued that a positive correlation between earnings and provisions may well be the result of optimal statistical forecasting

29

Banks and Bank Systems, Volume 3, Issue 4, 2008

cifically, managers will increase LLP when premanaged earnings (EBTP) are high, and decrease LLP when pre-managed earnings (EBTP) are low (Ahmed et al., 1999; Kim and Kross, 1998; Kanagaretman et al., 2004)1. Furthermore, by separating earnings on the basis of whether they are positive or negative, Laeven and Majnoni (2003) also find that positive earnings have a positive influence on LLP, whereas negative earnings have a negative effect. However, the interaction relationship between earnings and business cycles on provisioning has not been documented. The objective of this article is twofold. The first purpose is to explore the relationships among business cycles, earnings and bank LLP. In order to identify and test for different geographical regions and different regulations on provisioning, we use recent two-step system Generalized Method of Moments (GMM) dynamic panel data techniques proposed by Arellano and Bover (1995) and Blundell and Bond (1998), which can deal with the possible simultaneity among LLP, economic growth and bank earnings, so as to concentrate on the causal effect of the exogenous component of economic growth and bank earnings on loan loss provisioning. Using panel data also allows us to control for country-specific effects and to incorporate information from individual countries over time. This study departs from previous studies in the literature in that we assume that business cycles and earnings have an interactive effect on banks’ loan loss provisioning behavior. We attribute this to a negative relationship between LLP and business cycles and a positive relationship between LLP and bank earnings. Thus, there are four possible scenarios. In the first scenario, when both the economy and bank earnings are in good shape, in consideration of the income-smoothing effect, we expect banks to increase their LLP by virtue of the good economy and banks’ confidence in the future. In sharp contrast, in consideration of the pro-cyclicality effect, we do not expect banks to increase their LLP. Thus, these two different effects offset each other, leaving us uncertain about the sign of the coefficient. In the second scenario, when the external economy is in with respect to loan losses and hence is not necessarily due to misleading provisioning behavior as supposed in income-smoothing theory. Defond and Park (1997) used discretionary accruals as the basis to predict next period earnings. Evidence suggests that in considering job security when current earnings are ‘poor’ and expected future earnings are ‘good’, managers ‘borrow’ ‘future earnings’ for use in the current period. Conversely, when current earnings are ‘good’ and expected future earnings are ‘poor’, managers ‘save’ current earnings for possible use in the future. 1 Kanagaretman et al. (2004) employ book-to-price ratio as an alternative under/over valuation measure.

30

bad shape but bank earnings are in good shape, both forces are expected to lead to a positive coefficient. Thus, we expect the income-smoothing intense effect. Third, when the economy is in good shape but bank earnings are in negative, we expect that these conditions likely give banks the incentive to reduce their provisions. Hence, the reverse incomesmoothing intense effect should be at work. Finally, in the fourth scenario, when the economy and bank earnings are both in bad shape, it is difficult to predict the coefficient. We also contend that provisioning must be influenced not only by business cycles and bank earnings, but also by the regulatory system. Cavallo and Majnoni (2002) indicate that in countries with less governance, or with common law origin, or with relatively more external investor rights, banks tend to decrease provisions. Nevertheless, empirical research on this issue has been scarce. Is the behavior of bank provisioning in a particular country affected by that country’s banking regulations on provisions? To explain the current trends vis-à-vis provisioning, the existing literature most commonly classifies countries on the basis of either their membership in international organizations, such as the G10, OECD and the EU, or their geographic location, like Europe, Latin America and Asia. But, by any measure, even when countries are in the same international organization or located in roughly the same geographical region, the regulations on provisioning in each country are simply not the same. The second purpose of this article, therefore, is to explore whether bank regulations in different countries have an influence on bank LLP. Generally speaking, LLP can be divided into two categories: specific provisions and general provisions. A specific provision is a reserve that covers a specific loan loss and is fully or partially tax deductible, while a general provision covers a potentially uncertain loan loss, based on economic and earnings forecasts. It is worth noting that banks can take a deduction for general provisions up to a predefined percentage of eligible loans, for instance 0.3% in Japan and 2% in Taiwan. The tax deductibility of loan losses is undeniably a compelling force for banks to set aside adequate LLP. Thus, for individual countries, we explore three specific issues which may impact the income-smoothing effect. First, can general provisions be included in Tier II capital? Despite differences in the tax deductibility of loan losses, a specific provision cannot usually be included in Tier II capital. In the case of general provisions, however, there are different treatments in different countries. For instance, in France, Germany, the UK, the U.S.A. and many non-G-10

Banks and Bank Systems, Volume 3, Issue 4, 2008

countries1, including Taiwan, general provisions can be included in Tier II capital. By contrast, in Brazil, the Netherlands and Spain, general provisions cannot be counted as part of Tier II capital. We expect that if countries allow general provisions to be included in Tier II capital and bank earnings are rising, then banks probably intend to set aside greater provisions. On these grounds, the intense incomesmoothing effect should be in force. Secondly, do countries set minimum or benchmark provisioning requirements for standard loans? France, Germany, the UK, the USA, Singapore, Brazil and Chile, for instance, do not have minimum requirements. On the other hand, in several countries, general provisions are set at compulsory levels; in Italy, Argentina and China, to name a few, banks are required to provision 1 percent of their loans outstanding. In Taiwan, loans are classified into 5 categories based on the quality of loan, and banks are required to provision at least 2%, 10% and 50% for category 3, 4 and 5 type of loan, respectively2. We expect that if countries set minimum or benchmark provisioning requirements for standard loans and if bank earnings are rising, then banks likely provision more. Thus, the intense incomesmoothing effect should be in force. Thirdly, we investigate whether any legal penalties have been imposed on banks for inaccurately classifying a loan or underestimating provisions in at least the past five years. Even the most sophisticated legislative code must be tracked back to evaluate how well it is enforced. World Bank (2002) and Laeven and Majnoni (2003) also focus on the enforcement of regulations in their cross-country research. Corporate law gives directors and auditors certain rights and obligations to ensure that financial statements provide a fair and accurate statement of a bank’s financial position and that banks comply with adequate provisioning practices. Banking and financial legislation often provides specific penalties for violations of prudential regulations, in general, and for contraventions of the banking and financial services act, in particular. For instance, in Hong Kong, as in most other jurisdictions, the penalty for violating any provision of the Banking Ordinance could be a 1

Countries allowing general provisions to be counted as part of Tier II capital include some G-10 countries – France, Germany, Italy, Japan, the UK and the USA – and some non-G-10 countries, such as Argentina, Australia, Chile, China, the Czech Republic, Hong Kong, India, Mexico, Saudi Arabia, Singapore, South Africa, South Korea, the Russian Federation and Taiwan. 2 In Taiwan, according to Article 5 of the “Regulations Governing the Procedures for Banking Institutions to Evaluate Assets and Deal with Non-performing/Non-accrual Loans”, amended on Jan. 6, 2004, the minimum standard for loan loss provision shall be the sum of 2% of the balance of Category Two credit assets, 10% of the balance of Category Three credit assets, 50% of the balance of Category Four credit assets and the full balance of Category Five credit assets.

fine, imprisonment, or both. In France, underestimating provisions constitutes an offense to the extent that it affects the fairness and accuracy of the information that is provided to the public, as defined in the 1966 Commercial Company Law. Similar interpretations of commercial and banking laws are used in Mexico, Russia, Saudi Arabia, Spain and the West African Monetary Union (WAMU). In some countries, penalties are applicable to bank directors and managers, and these include fines, temporary disqualification, demotion, dismissal and even imprisonment. Not only that, when a violation affects the preparation of a final financial statement, it infringes on the auditor’s obligations (as in Germany)3. Thus, we expect that if any legal penalties have been imposed on banks because of an inaccurate classification of loans or an underestimation of provisions in at least the past five years and bank earnings are rising, then the banks implicated intend to provision more. This would mean that the intense income-smoothing effect is at play. The remainder of this paper is organized as follows. Section 1 describes the econometric model we employ. Section 2 provides the data and the descriptive statistics. Section 3 discusses the empirical results, and the last section reviews the conclusions we draw and presents some important policy implications. 1. Econometric framework 1.1. Basic model. In regard to the development of our econometric model, Degeorge et al. (1999) provided a two-period model in which managers manage reported earnings to maximize their own compensation. In their model, the firm’s latent earnings may reflect one of three situations: (1) The firm may be so far below the threshold that trying to reach it via managing earnings would be too costly. In this case, the firm seeks to report earnings that are less than its latent earnings, an approach referred to as “saving for a better tomorrow”. (2) If the firm is below its target earnings but reaching the target is not too costly, the managers may use their influence to boost reported earnings and achieve the target, a process described as “borrowing for a better today.” (3) Firms that exceed the target may reduce their current reported earnings to be able to report higher earnings in the next period, a process referred to as “reining in”. The authors noted that the three thresholds may be relevant to reported earnings: zero earnings, the prior year’s earnings per share, and stock analysts’ earnings expectations.

3 See World Bank (2002), pp. 30-31 for details. In the case of Taiwan, the relevant regulations stipulate the terms of punishment in the event that banks or managers violate applicable laws, regulations, or bank rules. However, no records of actual punishment are made available.

31

Banks and Bank Systems, Volume 3, Issue 4, 2008

In order to carry this two-period model out, this study extends the model proposed by Laeven and Majnoni (2003) by using loan loss provisions divided by total assets as the dependent variable (LLPTA) and explore the influence of business cycles and bank earnings on bank provisioning1. In a departure from the Laeven and Majnoni (2003) model, we classify business cycles into growth and decline. We also distinguish between positive and negative earnings that may influence banks’ provisioning behavior. Regarding the estimation methodology, we follow the dynamic panel data approach suggested by Arellano and Bover (1995) and Blundell and Bond (1998), and we use recently developed two-step dynamic panel system generalized GMM techniques to address potential endogeneity in the data. This method is also helpful to amend the bias induced by omitted variables in cross-sectional estimates, and the inconsistency caused by endogeneity both in cross-sectional and traditional static panel regressions. Our dynamic panel models are written as

major explanatory variable of loan loss provisioning. Thus, real GDP growth referring to as business cycle factor is included in the model. GDP + indicates that the economy is in good shape (i.e., the GDP real growth rate is higher than the 1991-2002 average value), while GDP − shows it is in a state of decline (i.e., the GDP real growth rate is less than the average value for that period). When equations (2) and (3) are inserted into (1), then GDP+ EBPT + denotes conditions in which both the economy and bank earnings are in good shape; GDP+ EBPT − denotes conditions in which the economy is in good shape but bank earnings are falling.

+ + α EBPT − + LLPTAijt = α 0 + α1 EBPTijt ijt 2 , + α 3 ∆LOAN ijt + α 4 Z ijt + α 5Tt + vi + ε ijt

termined variables Z. The instruments for the regression are levels of the right-hand side variables and the country-specific effect in equation (1); there is no correlation between the differences of these variables and the country-specific effect. We can validate the estimated model through a Sargan test of over-identifying restrictions.

α 1 = θ 11 GDPit+ + θ 12 GDPit− α 2 = θ 21 GDPit+ + θ 22 GDPit− EBPT + = max ( EBPT ,0)

,

(1)

(2)

The dependent variables include Loan growth. Z is the set of other control variables, such as Equity, NPL growth, Net Charge-Off etc., and Tt is the year dummy. We estimate the system by dynamic GMM with moment conditions E[∆LLPTAijt − s (vi + ε ijt )] = 0 and E[∆Z ijt − s (vi + ε ijt )] = 0 for s = 1 on the prede-

1.2. Sensitivity tests. Besides considering interactions between business cycles and bank earnings in bank provisioning, this study also includes legal Here, i = 1,…,N; t = 1,…, T; i is the i-th country; regulations pertaining to provisioning. For countries that allow banks to include general provisions in j stands for the j-th bank in country i; N = 49; and t Tier II capital, the dummy variable D tier2 is set as ranges from year 1991 to year 2002. vi , Tt and ε ijt one. And for those that set minimum or benchmark are, respectively, the unobservable country- and provisioning requirements for standard loans, the time-specific effects, and the error term. EBPT dummy variable Dminires is set as one. If countries (Earnings before Provision and Tax) represents net have legal penalties that have been imposed on earnings and is measured by each bank’s total as- banks for the inaccurate classification of loans or the sets. EBPT+ denotes that EBPT is positive in a spe- underestimation of provisions in at least the past cific year, but EBPT– denotes that it is negative. five years, the dummy variable Dpenalty is set as one. Since the business cycle variable appears in the Accordingly, the above model can not reflect the provisioning literature as a proxy for credit risk, a regulatory practices. Thus, our modified equation is: LLPTAijt = [ β 0 + β 1 EBPTijt+ + β 2 EBPTijt− + β 3 ∆LOAN ijt + β 4 Z ijt + β 5Tt ] × D regulation EBPT − = min ( EBPT ,0)

.

(3)

+ [γ 0 + γ 1 EBPTijt+ + γ 2 EBPTijt− + γ 3 ∆LOAN ijt + γ 4 Z ijt + γ 5Tt ] × (1 − D regulation ) ,

(4)

+ v i + ε ijt D regulation = ( Dtier 2 , D Minires , D penalty ) .

1

(5)

1 Further research may take the loan loss provisions ratio in terms of total loans, that would avoid ponderous explanations. However, Cavallo and Majnoni (2002), Laeven and Majnoni (2003) use total assets as the denominator of the dependent variable.

32

Banks and Bank Systems, Volume 3, Issue 4, 2008

The impact of three regulations on incomesmoothing effect can be analyzed as follows. First, when a country allows its banks’ general provisions to be included in Tier II capital, its banks have one more motivation to provision when earnings increase. Thus, permitting general provision to be included in Tier II strengthens the incomesmoothing effect, suggesting coefficient of β 1 Dtier 2 > γ 1 (1 − Dtier 2 ) > 0 . Next, when a country requires minimum provision, banks are prone to provision more in order to fulfill this requirement. Thus, permitting the minimum requirement also uphold the income-smoothing effect, suggesting that β 1 D min ires > γ 1 (1 − D min ires ) > 0 . Last, banks are more prudent with LLP if they are operating in a jurisdiction that imposes penalties on not provisions. Thus, once they have earnings, they tend to provision more, enhancing the income-smoothing effect. and β 1 D penalty > γ 1 (1 − D penalty ) > 0 β 2 D penalty < γ 2 (1 − D penalty ) < 0 .

2. Data and descriptive statistics

This study analyzes the commercial banks of 49 countries, and to do so, it takes banks’ financial statements from BankScope database. The empirical analysis covers the 1991-2002 sample period. Since most studies on cyclicality have been longer than 10 years and each version of BankScope only provides 8-year data, we combine 1999 and 2002 data (data can be traced back to 1991, the earliest data set provided by BankScope). As the number of banks in each edition is not constant, we compile bank names one by one from two different editions and delete the repeated years. The definitions of the variables and sources of the data are given in Table 1. In this study, there are two possible treatments for the data. In one treatment, we use the equity ratio as the benchmark since, from a comparative standpoint, it is the most complete variable. If a bank lacks 5 years or more of data, then that bank is deleted from the sample. We use the other treatment to deal with overlapping data if a bank provides consolidated and unconsolidated financial reports at the same time; normally the latter is more complete, and thus, it is kept. Table 2 presents the compilation of our data for the number of banks and the descriptive statistics of the main economic variables for each country. As shown in the first column of Table 2, of the total number of banks (4,024) in the sample countries, the USA (446), Germany (372) and France (356) have the largest number. The second column shows banks’ total assets (TA) by country, and the top three countries are Sweden (49,468 million USD),

Japan (49,258) and the Netherlands (30,370), values far higher than the whole sample average (9,540 million USD). The third column shows the loan ratio, which is total loan divided by total assets (Loan/TA), where the full sample average is 53.89%. The fourth column is the loan loss provision ratio, which is loan loss provision divided by total assets (Loan loss provision/TA), with the average value of 1.09% for all countries. It should be pointed out that the loan ratio for New Zealand is the highest three, thus, in theory, it might be expected that New Zealand’s loan loss provision ratio should be relatively high. But it is the lowest, revealing that it violates the Matching Principle between loan (revenue) and provision (expense). Particularly interesting too is the case of Taiwan. Its loan ratio is 68.07%, which is significantly higher than the 53.89% average, but its loan loss provision ratio is only 0.72%, which is significantly lower than the average 1.09%. As shown in column 5, the loan loss reserve ratio for Taiwan is only 0.87%, ever so much lower than the full sample average of 2.90%; in no way, therefore, is provisioning in the Taiwan banking industry sufficient. Turning to the GDP growth rate in column 6 of Table 2, the average value is 3.16%. Table 3 provides the statistical information for other economic variables. The first column is Equity/TA, where the sample average is 9.80%, the highest value is in Brazil (17.19%), and the lowest in Ecuador (-0.17%). The second column is EBPT/TA, where the average value is 1.06%; the highest value is in Turkey (3.37%), while the lowest is in Uruguay (-4.87%). Column 3 lists the data for NCO/TA, and the average is only 0.87%; Argentina has the highest (6.14%), and the Netherlands has the lowest (0.01%). The fourth column lists the data for NPL/TA; the highest ratios are in Thailand (22.98%), Indonesia (16.97%) and Kenya (12.90%), all well above the average (4.59%). The lowest is in the USA (0.47%). Table 4 reports the mean results for the independent and dependent variables. On the basis of GDP and earnings, when the economy and banks’ earnings are both in a good shape, the LLP/TA ratio is 0.72% which is the lowest level. When the economy remains good, but earnings become negative, the LLP/TA ratio is higher (1.85%). The third scenario is where the ratio is 0.95%. However, when both the economy and banks’ earnings exhibit a downward trend, the LLP/TA ratio reaches the highest level (3.63%) and has the highest standard error (13.49%), indicating that banks on average provision more and are much more volatile. 33

Banks and Bank Systems, Volume 3, Issue 4, 2008

Loan growth exhibits only one positive figure when the economy and earnings are both good, i.e., 0.23%, showing that a good external and internal environment will induce banks to lend more. However, when the economy and earnings are in a bad state, the banks’ loan growth will have the highest negative growth (-3.52%), meaning that banks will take more precautions in their lending. It is interesting to find that when banks’ earnings are positive, regardless of whether the economy is good or bad, the equity ratios are ranked as the lowest two ones, indicating that positive earnings make banks feel safe by keeping their equity structure at a lower level. As regards the Non-Performing Loan growth and Net Charge-Off ratios, these two ratios reach their lowest level (and highest volatility also) when the economy and earnings are in good condition, but they also reach their highest level when the economy, both externally and internally, is facing a downward trend, meaning that banks are not forward looking. If variables are classified based on geographical location, it is found that banks located in Latin America set aside the highest provision (1.98%) and that highest equity ratio (0.02%) that are accompanied by the highest NPL growth and Net Charge-Off of 0.98% and 2.55%, respectively. It is learnt from this case that these variables cannot be overlooked. Besides Latin America, banks in Asia have the second high loan loss provisions (1.59%), and the rest are ordered as follows: Japan (0.66%), the USA (0.60%) and Europe (0.59%). It is particularly worth mentioning that the equity ratio in Japan has a small value (0.00003%) compared with the others, and this may not only reflect the fact that Japanese banks are extremely large in size, but that Japan has a fragile financial system. Table 4 also presents data pertaining to relevant regulations on provisioning. First, as concerns countries that allow general provisions to be counted as part of Tier II capital, all five variables are smaller than those of other countries, thus contravening our intuition. Secondly, with respect to countries that set minimum or benchmark provisioning requirements for standard loans, banks on average set more provisions than the other countries (0.94% vs. 0.74%), which coincides with our expectations. The same pattern occurs with countries that have imposed legal penalties on banks for having inaccurately classified loans or for having underestimated provisions in at least the past five years. This shows that legal enforcements will push banks to provision more (0.78% vs. 0.61%). Table 5 presents the correlation coefficients between the independent and dependent variables. It is clear 34

that the pro-cyclical effect holds. For instance, the correlation between GDP growth rate and Loan loss provision is negative, which means that when the economy is growing, banks tend to decrease their provisions. 3. Empirical results 3.1. Basic model. Columns A and B in Table 6 show the dynamic interaction effect of GDP growth, Earnings and Loan growth. We find that the coefficient of GDP growth is positive, whereas that of Earnings is positive, and the latter is consistent with Laeven and Majnoni (2003). And, important to note, only the income-smoothing effects is supported. A clear trend is noted in these results: with a higher growth rate for Non-performing loans or Net charge-offs, or a higher equity ratio, banks increase their provisions.

Columns C and D in Table 6 report our modified dynamic model which considers whether business cycles and earnings have an interactive effect on provisioning. First, in column C, the coefficient of GDP+ EBPT + is 0.023, which shows that banks provision more when earnings are positive without being affected by the good economic condition into account. The coefficient of GDP+ EBPT − is -0.143 and significant, suggesting that good economy upholds the negative-earnings incomesmoothing effect. That is, banks provision less when earnings are negative and this effect is enhanced by the good economy. In the third scenario, i.e., when the economy is in a downward trend but banks enjoy positive earnings, the coefficient of GDP− EBPT + turns out to be a desirable positive (0.021), thus bust of the economy strengthens the positive-earnings income-smoothing effect. Finally, the coefficient of GDP − EBPT − is -0.046 and significant, indicating that when the economy is in a downturn and bank earnings are not satisfactory, banks tend to decrease their provisions. In column D, the control variable is added. All the coefficients of the variables remain the same except for GDP − EBPT − as the coefficient changes from negative to desirable positive, and this, in a significant manner. This shows that bust of the economy also strengthens the negative-earnings income-smoothing effect and this irrefutably confirms that policy in the financial system vis-à-vis provisioning is not forward looking1. Since the regressions in Table 6 pass the Sargan tests, this two-step system GMM estimator seems to offer a particularly useful assessment of GDP growth, Earnings and Loan growth. 1

See Borio et al. (2001) and Beattie et al. (1995) for more details.

Banks and Bank Systems, Volume 3, Issue 4, 2008

3.2. Impact of geographical location on provisioning. Table 7 groups countries by location: Europe, the USA, Japan, Latin America and Asia. Except for Asia, all the coefficients of

GDP + × EBPT + are significantly positive, imply-

ing that when both the economy and earnings are growing, banks raise their provisions. This shows that the income-smoothing effect is stand and less influenced by business cycle. By contrast, the coefficient for Asia is significantly negative; this attests to the dominance of the pro-cyclical effect in Asia. The GDP + × EBPT − coefficients are all negative. This implies that when the economy is on an upward trend but earnings are poor, banks tend to decrease their provisions. Again, the incomesmoothing effect is stand and less influenced by business cycle. The GDP − × EBPT + coefficients are not consistent across all five zones. Significantly negative as they are in Europe, the USA and Latin America, they uphold neither the pro-cyclical effect nor the income-smoothing effect. In Japan and Asia, the coefficients are desirable positive ones, meaning both the pro-cyclical effect and the incomesmoothing effect hold. Finally, the coefficients for GDP − × EBPT − also differ by geographical location. In Europe and Latin America, they are positive and significant for the latter. This signals that when the economy is on a downturn trend of the business cycle and earnings are not good, banks tend to raise their provisions, showing that the income-smoothing effect is not stand and much influenced by business cycle. In the USA, Japan and Asia, on the other hand, the coefficients are negative showing the income-smoothing effect is dominant. The major difference between Asia and the other four zones is that banks do not provide sufficient (higher) provisions during periods when economic growth and earnings are on the increase. Implicit here is that prior to the 1997 Asian crisis, banks had evidently not provisioned enough to be able to confront the serious loan losses that were about to occur. Instead, banks provisioned more during the very period in which the economy was suffering when bank earnings were good. It is suggested that this is an important lesson that should have been learned from the crisis. Table 8 has the same geographical classifications as Table 7, but the control variables are added and led to a significant drop of observations. Thus, there are some exceptions compared with Table 7. As for the control variables, the results do not change much and are basically consistent with those in Table 6. There are exceptions in Europe, Japan and Asia. The coefficient for the equity ratio shifts from signifi-

cantly positive to significantly negative. In Japan, the absolute value increases, which shows that Japanese banks with a higher ratio of equity capital tend to decrease provisions proportionately. 3.3. Impact of regulatory systems on provisioning. Table 9 reports on the impact of regulatory systems pertaining to provisioning. That is, can general provisions be included in Tier II capital? Is there any minimum provision required? Have any legal penalties been imposed on banks in at least the past five years? First of all, compared with benchmark model (Column C of Table 6) countries with penalties records on provisioning, the GDP + × EBPT + coefficients intend to be greater than the bench model ones (0.0754 versus 0.023). This result is consistent with our hypothesis, which is the intense income-smoothing effect is at play under taking regulations into consideration.

Secondly, the coefficients of GDP + × EBPT − found that countries do not allow general provisions as part of Tier II and countries without any minimum requirements, their banks intend to provision less than benchmark model. Thirdly, the coefficient of GDP − × EBPT − for countries which have no minimum requirements is significantly negative. This illustrates that when regulatory systems are taken into account, the income-smoothing effect is still existed and less impacted by business cycles. Table 10 follows Table 9, and the control variables are included. Sample size also drops significantly, though, there are still some findings. First, for countries which allow general provisions to be considered as Tier II capital, the coefficient of GDP + × EBPT − is significantly negative and smaller than the benchmark model matching our expectation (-0.463 versus -0.095). Second, the situation with GDP − × EBPT + is also different; in countries which have penalty records available on provisioning, the coefficient becomes significantly negative and smaller than that shown for the basic model (-0.717 versus 0.015). This means the income-smoothing intense effect is upheld. It is important to note, therefore, that the enforcement of bank regulations on loan loss provisioning does indeed have an impact on banks’ provisioning behavior. Conclusions

The first objective of this article is to explore the relationships among business cycles, earnings and bank loan loss provisions by employing recent twostep system GMM techniques developed for dynamic panels on a panel of 49 countries observed in the period of 1991-2002. This study differs from previous studies in the literature because we assume 35

Banks and Bank Systems, Volume 3, Issue 4, 2008

that business cycles and earnings have an interactive impact on the behavior of bank loan loss provisioning. We attribute this to a negative relationship between business cycles and LLP and a positive relationship between bank earnings and LLP. Next, unlike pervious studying, the GMM panel estimator exploits the time-series variation in the data, accounts for unobserved country-specific effects, allows for the inclusion of lagged dependent variables as regressors, and controls for endogeneity of all the explanatory variables. The evidence shows that with steady growth in both the economy and bank earnings, bank management tends to increase LLP, whereas with a buoyant economy but negative growth in bank earnings, management shows a tendency to reduce LLP. In these scenarios, the income-smoothing effect appears to be held and less affected by business cycles. By contrast, when the economy is in a downward trend and banks suffer losses, management evidently increases LLP. In this case, the reversed incomesmoothing effect is stand and strongly influenced by business cycles. When geographical location is taken into account, the income smoothing effect has a dominant power and less affected by business cycles. The implication that emerges from this part of the empirical results is that prior to the 1997 Asian cri-

sis, banks had obviously not been provisioning enough to be able to meet the challenge they faced when the severe loan losses later occurred. Instead, banks provisioned more during the very period in which the economy was suffering but bank earnings were good. It is suggested that the relevant authority and banks take serious note of this given that they had been provisioning more, the severity of the crisis may have been mitigated. As to whether country-wide bank regulations influence bank loan loss provisioning, it is found that when the economy and bank earnings are both showing steady or negative growth, the intense income-smoothing effect is at work. In other scenarios, neither income-smoothing nor pro-cyclicality holds. This accounts for the fact that bank regulations on loan loss provisioning across 49 countries do have an impact on banks’ provisioning behavior. The implication here is that even when countries are in the same international organization or are located in roughly the same geographical region, the regulations with respect to provisioning in each country are just not the same. Policymakers and researchers need to pay considerably more attention to the enactment and enforcement of relevant regulations on provisioning.

References

1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14.

36

Ahmed, A.S., Takeda, C., Thomas, S. (1999). Bank loan loss provisions: a re-examination of capital management, earnings management and signaling effects. Journal of Accounting and Economics 28, 1-25. Arellano, M., Bover, O. (1995). Another look at the instrumental variable estimation of error-components models. Journal of Econometrics 68, 29-51. Ayuso, J., Perez, D., Saurina J. (2004). Are capital buffers pro-cyclical? Evidence from Spanish panel data. Journal of Financial Intermediation 13 (2), 249-264. Beattie, V.A., Casson, P.D., Dale, R.S., McKenzie, G.W., Sutcliffe, C.M.S., Turner, M.J. (1995). Banks and Bad Debts: Accounting for loan losses in international banking. New York USA, John Wiley and Sons. Beatty, A., Chamberlain, S.L., Magliolo, J. (1995). Managing financial reports of commercial banks: The influence of taxes, regulatory capital, and earnings. Journal of Accounting Research 33, 231-261. Blundell, R., Bond, S. (1998). Initial conditions and moment restrictions in dynamic panel data models. Journal of Econometrics 87, 115-143. Borio, C., Furfine, C., Lowe, P. (2001). Procyclicality of the financial system and financial stability: Issues and policy options. BIS Paper No. 1, 1-57. Cavallo, M., Majnoni, G. (2002). Do banks provision for bad loans in good times? Empirical evidence and policy implications. In: Levich, R., Majnoni, G., Reinhart C. (Eds), Ratings, Rating Agencies and the Global Financial System, 319-342. Boston, Kluwer Academic Publishers, Dordrecht and London. Collins, J.H., Shackelford, D.A., Wahlen, J.M. (1995). Bank differences in the coordination of regulatory capital, earnings, and taxes. Journal of Accounting Research 33, 263-291. Defond, M.L., Park, C.W. (1997). Smoothing income in anticipation of future earnings. Journal of Accounting and Economics 23, 115-139. Degeorge, F., Patel, J., Zeckhauser, R. (1999). Earnings management to exceed thresholds. Journal of Business 72, 1-33. Estrella, A. (2004). The cyclical behavior of optimal bank capital. Journal of Banking and Finance 28, 1469-1498. Fudenberg, D., Tirole, J. (1995). A theory of income and dividend smoothing based on incumbency rents. Journal of Political Economy 103, 75-93. Greenawalt, M.B., Sinky, J.F. (1988). Bank loan-loss provisions and the income smoothing hypothesis: An empirical analysis, 1976-1984. Journal of Financial Services Research 1, 301-318.

Banks and Bank Systems, Volume 3, Issue 4, 2008

15. Kanagaretnam, K., Lobo, G.J., Yang, D.H. (2004). Joint tests of signalling and income smoothing through bank loan loss provisions. Contemporary Accounting Research 21, 843-884. 16. Kim, D., Santomero, A.M. (1993). Forecasting required loan loss reserves. Journal of Economics and Business 45, 315-329. 17. Kim, M.S., Kross, W. (1998). The impact of the 1989 change in bank capital standards on loan loss provisions and loan write-offs. Journal of Accounting and Economics 25, 69-99. 18. Laeven, L., Majnoni, G. (2003). Loan loss provisioning and economic slowdowns: Too much, too late? Journal of Financial Intermediation 12, 178-197. 19. Lindquist, K.G. (2004). Banks’ buffer capital: How important is risk? Journal of International Money and Finance 23, 493-513. 20. Lowe, P. (2003). Credit Risk Measurement and Pro-cyclicality. Monetary and Economic Department, Bank for International Settlements, Banking supervision at the crossroads, 136-163. 21. Moyer, S.E. (1990). Capital adequacy ratio regulations and accounting choices in commercial banks. Journal of Accounting and Economics 13, 123-154. 22. Rime, B. (2001). Capital requirements and bank behaviour: Empirical evidence for Switzerland. Journal of Banking and Finance 25, 789-805. 23. Wetmore, J.L., Brick, J.R. (1994). Loan loss provisions of commercial banks and adequate disclosure: A note. Journal of Economics and Business 46, 299-305. 24. World Bank (2002). Bank Loan Classification and Provisioning Practices in Selected Developed and Emerging Countries. Washington D.C., Finance Forum. Appendix

Table 1. Definitions and sources of the variables from BankScope -- Bureau van Dijk

Micro LLP/TA

Loan loss provision / Total assets

LLR/TA

Loan loss reserve / Total assets

EBPT

Earnings before provision and tax / Total assets

EBPT +

EBPT is positive in the specific year, and it is a true value rather than a dummy variable.

EBPT + = max(EBPT ,0) EBPT −

EBPT is negative in the specific year, and it is a true value rather than a dummy variable.

EBPT − = min(EBPT ,0) Loan ratio

Total Loan / Total Assets

Loan growth

(LOAN/TA) t – (LOAN/TA) t-1

Equity

(Equity / Total Assets)

NPL growth

(NPL/TA) t – (NPL/TA) t-1

Net charge-off

Net charge-off / Total Assets from World Bank Development Indicator

Macro

GDP

GDP growth is real growth in GDP per capita (annual %).

GDP

+

GDP real growth rate is greater than the 1991-2002 average value.

GDP



The GDP real growth rate is less than the 1991-2002 average value.

GDP per capita Bank regulation

GDP per capita (constant 1995 U.S.$) From Laeven and Majnoni (2003), for the case of Taiwan is collected by the authors. In countries which allow banks to include their general provisions in Tier II capital, the dummy variable is set as 1; other-

Dtier 2

wise as 0.

D Minires

In countries which set minimum or benchmark provisioning requirements for standard loans, the dummy variable is set as

D penalty

In countries which have imposed legal penalties on banks for the inaccurate classification of loans or the underestimation

1; otherwise as 0. of provisions in at least the past five years, then the dummy variable is set as 1; otherwise as 0.

37

Banks and Bank Systems, Volume 3, Issue 4, 2008

Table 2. Number of banks and the descriptive statistics of the main economic variables Total assets No.

Country

Number of banks

Million USD

Loan / TA (%)

Loan loss provision /TA (%)

Loan loss reserve/TA (%)

GDP growth rate (%)

1

Argentina

64

1,582

48.21

2.51

5.65

2.63

2

Australia

27

15,318

73.35

0.65

1.30

3.52

3

Austria

31

4,828

44.98

0.49

1.07

2.15

4

Belgium

36

16,833

34.94

0.31

0.18

1.93

5

Brazil

90

4,859

33.84

2.09

2.82

2.5

6

Canada

35

23,764

65.77

0.80

1.41

2.78

7

Chile

15

2,837

62.66

0.57

1.48

5.88

8

Colombia

21

935

60.32

1.79

2.23

2.37

9

Denmark

55

4,661

55.80

1.06

3.13

2.24

10

Ecuador

22

242

48.52

4.82

12.24

2.25

11

Egypt

27

2,450

45.83

1.17

6.44

4.18

12

Finland

5

17,427

49.35

0.51

1.06

1.95

13

France

206

10,776

48.64

0.80

4.14

1.84

14

Germany

162

7,922

47.56

0.52

2.01

1.67

15

Greece

19

7,216

41.74

0.52

1.44

2.62

16

Hong Kong

34

11,211

48.86

0.47

1.36

4.05

17

India

56

3,742

43.12

0.67

0.94

5.4

18

Indonesia

42

1,727

57.84

4.57

8.85

4.28

19

Ireland

12

14,404

57.20

0.23

1.01

7.09

20

Israel

15

8,700

63.83

0.64

2.71

4.41

21

Italy

54

18,821

48.09

0.50

2.02

1.52

22

Japan

140

49,258

70.25

0.66

1.46

1.27

23

Jordan

10

4,068

43.36

0.60

5.20

5.1

24

Kenya

17

248

52.10

1.40

5.80

1.6

25

Malaysia

23

5,165

57.66

0.90

2.96

6.37

26

Mexico

25

7,427

53.96

1.19

2.81

2.98

27

Netherlands

29

30,370

46.73

0.26

0.92

2.56

28

New Zealand

6

7,910

75.58

0.12

0.69

3.04

29

Nigeria

14

794

27.74

1.22

5.32

2.59

30

Norway

9

9,552

79.85

0.68

2.41

3.35

31

Pakistan

19

1,375

43.01

0.62

3.11

3.75

32

Peru

14

1,094

56.01

1.94

4.12

3.79

33

Philippines

13

928

54.23

0.61

2.52

3.18

34

Portugal

27

7,980

42.16

0.42

1.48

2.53

35

Singapore

10

14,550

64.32

0.74

4.53

6.47

36

South Africa

13

7,057

74.76

1.18

2.68

1.99

37

South Korea

16

20,924

57.29

1.09

1.60

6.03

38

Spain

84

8,719

44.15

0.37

1.53

2.62

39

Sri Lanka

6

811

56.76

0.53

2.59

4.55

40

Sweden

5

49,468

51.19

0.91

3.68

1.91

41

Switzerland

143

1,453

53.68

0.36

2.29

0.81

42

Taiwan

35

12,198

68.07

0.72

0.87

5.47

43

Thailand

13

10,207

74.70

1.49

5.51

4.49

44

Turkey

26

3,248

35.23

1.18

1.43

3.12

45

U.K.

93

19,830

37.87

0.66

2.15

2.27

46

United States

324

10,217

60.19

0.60

1.17

2.92

47

Uruguay

5

627

77.43

4.83

1.42

1.37

48

Venezuela

10

1,292

43.54

1.29

2.86

1.25

49

Zimbabwe

Average Total

2,163

6

433

58.31

0.98

5.62

0.28

44

9,540

53.89

1.09

2.90

3.16

467,460

Note: Values reported in this table are the average values after those banks which lack at least 5 years of data have been deleted. The rest of this study adopts this sample.

38

Banks and Bank Systems, Volume 3, Issue 4, 2008

Table 3. Descriptive statistics of the micro and macro economic variables No.

Country

Equity/TA

EBPT/TA

NCO/TA

(%)

(%)

(%)

NPL/TA (%)

-0.91

6.14

7.89 1.41

1

Argentina

18.15

2

Australia

7.37

0.46

0.62

3

Austria

8.81

0.79

Na

Na

4

Belgium

9.30

0.95

Na

0.52

5

Brazil

17.19

2.57

2.27

2.86

6

Canada

8.39

0.93

0.57

2.54

7

Chile

12.31

0.97

0.51

0.64

8

Colombia

16.26

1.43

0.89

4.04

9

Denmark

10.68

1.20

Na

1.12

10

Ecuador

-0.17

-1.60

3.93

3.35

11

Egypt

9.00

1.21

0.20

Na

12

Finland

5.45

0.47

0.18

1.29

13

France

9.96

0.85

0.76

6.06

14

Germany

9.32

0.79

Na

Na

15

Greece

7.89

1.09

0.39

2.68

16

Hong Kong

14.26

1.78

0.42

2.71

17

India

4.95

0.60

0.58

3.10

18

Indonesia

6.81

-1.76

2.55

16.97

19

Ireland

8.16

0.97

0.24

0.61

20

Israel

9.03

0.53

0.22

5.45

21

Italy

8.29

0.84

1.02

3.78

22

Japan

3.99

-0.04

0.31

3.46

23

Jordan

7.45

1.02

0.20

8.21

24

Kenya

12.22

3.12

1.17

12.90

25

Malaysia

9.80

1.53

Na

0.90

26

Mexico

16.16

1.07

0.87

3.20

27

Netherlands

9.15

1.01

0.01

0.77

28

New Zealand

4.95

1.31

0.11

0.55

29

Nigeria

9.49

3.31

0.28

5.71

30

Norway

6.13

0.71

0.28

3.22

31

Pakistan

6.92

1.25

0.10

6.17

32

Peru

10.59

1.34

0.96

6.12

33

Philippines

16.46

1.34

0.30

7.64

34

Portugal

7.26

0.82

0.33

2.36

35

Singapore

14.10

1.59

0.61

8.09

36

South Africa

14.44

1.29

0.87

3.35

37

South Korea

5.76

-0.18

1.40

4.41

38

Spain

17.80

1.56

0.40

Na

39

Sri Lanka

8.00

1.78

0.03

8.76

40

Sweden

6.81

0.49

Na

7.28

41

Switzerland

16.93

2.04

Na

1.44

42

Taiwan

13.59

0.87

2.05

1.28

43

Thailand

6.07

-0.91

0.80

22.98

44

Turkey

10.06

3.37

0.03

2.63

45

U.K.

12.56

1.74

0.58

2.21

46

United States

9.86

2.28

0.55

0.47

47

Uruguay

2.07

-4.87

1.39

7.16

48

Venezuela

12.87

4.71

0.79

2.66

49

Zimbabwe

7.34

4.34

0.51

5.52

Average

9.80

1.06

0.87

4.59

39

Banks and Bank Systems, Volume 3, Issue 4, 2008

Table 4. Means of variables (%) LLP/TA

Loan growth

Equity

NPL growth

Net charge-off

Based on GDP and earnings

+ + GDP × EBPT

0.7231 (2.389)

0.2253 (8.146)

0.0065 (0.040)

0.0745 (3.024)

0.5381 (2.999)

+ − GDP × EBPT

1.851 (5.956)

-0.4781 (13.957)

0.0154 (0.348)

1.701 (9.353)

1.663 (3.738)

− + GDP × EBPT

0.9527 (3.806) 3.629 (13.487)

-0.2615 (7.987) -3.520 (14.578)

0.0063 (0.045) 0.0476 (0.311)

0.1441 (4.534) 2.643 (20.315)

0.7392 (2.839) 5.896 (16.608)

0.5893 (1.231) 0.5977 (3.198) 0.6628 (1.389) 1.9839 (5.218) 1.5922 (7.739)

0.0999 (8.354) -0.0027 (8.137) -0.0760 (2.561) -0.5682 (12.718) -0.3707 (8.590)

0.0145 (0.110) 0.0046 (0.042) 0.00003 (0.00007) 0.0204 (0.086) 0.0043 (0.056)

-0.1341 (4.426) 0.0124 (0.419) 0.5501 (2.493) 0.9765 (4.438) 0.3259 (12.369)

0.4190 (1.395) 0.5504 (3.232) 0.3162 (0.669) 2.545 (6.996) 1.4785 (7.902)

0.0070 (0.056) 0.0371 (0.217)

0.1411 (3.025) 0.4789 (2.679)

0.6168 (3.049) 1.5699 (5.722)

0.0067 (0.052) 0.0134 (0.113)

0.3728 (3.255) 0.0093 (2.835)

0.7066 (3.007) 7.2923 (3.684)

0.0138 (0.113) 0.0062 (0.060)

0.0379 (2.871) 0.3007 (3.073)

0.7778 (3.909) 0.4004 (1.649)

− − GDP × EBPT Based on geographical location Europe USA Japan Latin America Asia

Based on regulation Can general provisions be included in Tier II capital? Yes 0.7721 -0.1461 (2.731) (8.469) No 1.0387 0.0250 (3.288) (10.872) Is there any minimum provision required? Yes 0.9421 -0.2827 (3.049) (8.224) No 0.7406 -0.0387 (2.688) (9.140) Have any legal penalties been imposed in at least the past five years? Yes 0. 7836 0.0133 (2.819) (9.983) No 0.6100 -0.1518 (2.127) (6.804)

Note: Values in parentheses are standard errors.

Table 5. Correlation coefficients of the microeconomic and economic development variables GDP LLP/TA LLP/TA LLR/TA Equity/TA NCO/TA Loan/TA NPL/TA EBPT/TA GDP growth rate

40

1

LLR/TA

Equity/TA

NCO/TA

Loan/TA

NPL/TA

EBPT/TA

growth rate

0.49

-0.24

0.40

0.03

0.38

-0.40

-0.18

1

-0.40

0.39

-0.03

0.81

-0.26

-0.08

1

0.16

-0.19

-0.10

0.33

0.01

1

-0.04

0.09

-0.01

-0.08

1

0.05

-0.05

0.05

1

-0.19

-0.20

1

0.11 1

Banks and Bank Systems, Volume 3, Issue 4, 2008

Table 6. Test results of the effects of pro-cyclicality and income-smoothing GDP growth

(A)

(B)

0.001** (2.354)

-0.001 (-1.041)

(C)

+

GDP × EBPT

+

+ − GDP × EBPT − + GDP × EBPT EBPT

0.104** (2.978)

Loan growth

0.036* (1.674)

-0.020** (-3.646)

− − GDP × EBPT

(D)

0.023 (1.115)

0.075** (4.098)

-0.143** (-2.603)

-0.095** (-3.401)

0.021 (1.247)

0.028** (3.092)

-0.046** (-2.464)

0.015** (2.882)

-0.019** (-3.840)

-0.010** (-2.056)

-0.008* (-1.778)

Loan growth

Equity

32.818** (2.513)

Equity

24.622* (1.856)

NPL growth

0.281** (11.125)

NPL growth

0.263** (11.280)

Net charge-off

0.678** (14.352)

Net Charge-off

0.689** (15.918)

Sargan test (p-value) No. of observations No. of banks

0.086 14,046 2,163

Sargan test (p-value) No. of bank-years No. of banks

0.179 3,855 2,163

0.115 14,046 2,163

0.181 3,643 2,163

Notes: 1. The dynamic panel model is adopted. Due to space constraints, the constant term is not reported. The independent variable is the ratio of loan loss provision divided by total assets (LLPTA). GDP growth is real growth in per capita GDP in annual percent. EBPT equals earnings before provision and tax divided by total assets. Loan growth is equal to the loan growth rate. Equity is the ratio of equity capital divided by total assets. NPL growth is the non-performing loan growth rate.

+



Net charge-off is the ratio of net charge-off against total assets. 2. GDP ( GDP ) means that the GDP real growth rate is greater (less) than the 1991-2002 average value. EBPT

+

( EBPT

+



) denotes that EBPT is positive (negative) in the specific

+

year, and it is a true value rather than a dummy variable. GDP × EBPT shows that both the economy and bank earnings are in good shape. 3. Values in parentheses are t-values; ***, ** and * indicate the 1%, 5% and 10% level of significance, respectively. 4. Sargan test: the null hypothesis is that the instruments used are not correlated with the residuals. All equations include time dummies as regressors and instruments. ** and * indicate the significance at the 5% and 10% levels, respectively. Instruments: lagged levels for differences, lagged differences for levels. Two-step estimates.

Table 7. Test results of the effects of pro-cyclicality and income-smoothing — geographical location added Europe

USA

Japan

Latin America

Asia

0.007** (4.216)

0.142** (25.628)

0.352** (7.104)

0.045** (8.548)

-0.499** (-7.483)

-0.033 (-1.337)

-0.031 (-0.631)

-0.380** (-2.434)

-0.209** (-5.954)

-0.119 (-0.400)

-0.032** (-3.053)

-0.033** (-4.223)

0.090** (2.900)

-0.019** (-3.524)

0.063** (10.852)

− − GDP × EBPT

0.048 (0.042)

-0.550** (-21.486)

-0.183** (-3.456)

0.104** (9.229)

-0.081** (-3.656)

Loan growth

0.001 (0.376)

-0.022** (-5.612)

0.028** (3.178)

-0.013** (-2.218)

-0.082** (-3.054)

Sargan test (p-value) No. of observations No. of banks

0.185 4,612 938

0.574 2,391 324

0.388 1,930 244

0.280 1,099 217

GDP GDP

+ +

× EBPT × EBPT

+ −

− + GDP × EBPT

0.165 992 140

Notes: 1. The same as Table 6. 2. “Europe” includes Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, the Netherlands, Norway, Portugal, Sweden, Spain, Switzerland and the United Kingdom. “Latin America” includes Argentina, Brazil, Chile, Colombia, Mexico, Peru, Uruguay and Venezuela. “Asia” includes India, Indonesia, Korea, Malaysia, Pakistan, the Philippines, Taiwan and Thailand. “USA” indicates the United States of America.

41

Banks and Bank Systems, Volume 3, Issue 4, 2008

Table 8. Test results of the effects of pro-cyclicality and income-smoothing – geographical location and control variables added Europe

USA

0.006** (3.303)

0.038** (16.023)

0.058* (1.867)

0.023** (1.964)

-0.014 (-0.489)

+ − GDP × EBPT

-1.096** (-1175.294)

0.015 (0.338)

-0.109** (-2.117)

0.042 (0.133)

-0.471** (-2.781)

− + GDP × EBPT

-0.081** (-3.151)

0.008 (1.380)

0.202** (13.524)

-0.019** (-2.357)

0.044** (13.181)

-1.025 (-0.568)

-0.755** (-12.043)

0.064 (1.076)

0.012** (4.642)

0.042** (4.976)

Loan growth

-0.009** (-28.147)

0.005** (3.094)

-0.023** (-5.555)

-0.005 (-0.781)

-0.029** (-2.703)

Equity

-20.766** (-3.891)

13.794** (4.962)

-19039.54** (-21.061)

41.484** (3.750)

-40.604 (-1.053)

NPL growth

0.243** (126.341)

0.150** (6.623)

0.093** (16.608)

0.205** (10.023)

0.309** (24.533)

Net charge-off

0.784** (120.977)

0.869** (89.308)

0.186** (7.258)

0.288** (5.046)

0.450** (16.245)

0.699 149 939

0.687 1,899 324

0.991 126 244

0.922 219 217

GDP

+

× EBPT

+

− − GDP × EBPT

Sargan test (p-value) No. of observations No. of banks

Japan

0.160 819 140

Latin America

Asia

Note: The same as Table 6 and Table 7. Table 9. Test results of the impact of regulations on provisioning Can general provisions be included in Tier II capital?

Is there any minimum provision required?

Have any legal penalties been imposed in at least the past five years?

Yes 0.069** (9.312)

No 0.091** (14.958)

Yes 0.015** (2.900)

No 0.071** (9.530)

Yes 0.0754** (11.257)

No 0.005** (8.417)

-0.117** (-2.586)

-0.770** (-13.496)

-0.095* (-1.758)

-0.489** (-7.133)

-0.140** (-2.574)

-0.163** (-6.244)

-0.026** (-4.823)

-0.035** (-4.736)

-0.021** (-6.336)

-0.033** (-5.053)

-0.017** (-2.698)

-0.019** (-3.915)

− − GDP × EBPT

0.074** (5.609)

0.068** (2.498)

0.069** (8.846)

-0.357** (-5.374)

-0.0171 (-0.597)

0.0.96 (0.860)

Loan growth

-0.009** (-2.837)

-0.004 (-1.342)

-0.008** (-2.497)

-0.010** (-3.113)

-0.013** (-4.362)

-0.001 (-0.258)

0.083 1,435 1,274

0.094 8,598 1,274

0.411 4,165 548

0.091 5,868 929

0.144 5,051 699

0.094 3,899 607

+

GDP × EBPT GDP

+

× EBPT

+ −

− + GDP × EBPT

Sargan test (p-value) No. of observations No. of banks

Notes: 1. The same as Table 6. 2. Countries that allow general provisions to be counted as part of Tier II capital include some G-10 countries, such as France, Germany, Italy, Japan, the UK and the USA, and some non-G-10 countries, such as Argentina, Australia, Chile, China, the Czech Republic, Hong Kong, India, Mexico, Saudi Arabia, Singapore, South Africa, South Korea, the Russian Federation and Taiwan. 3. Countries which set minimum or benchmark provisioning requirements for standard loans include some G-10 countries, such as Italy and Japan, and some non-G-10 countries, such as Argentina, Australia, China, Hong Kong, India, South Korea, Mexico, South Africa, Spain, the Russian Federation and Taiwan. 4. Countries which have penalty records on provisioning available include some G-10 countries, such as France, Italy and the USA, and some non-G-10 countries, such as Brazil, China, the Czech Republic, Hong Kong, Mexico, Saudi Arabia, Singapore and the Russian Federation.

42

Banks and Bank Systems, Volume 3, Issue 4, 2008

Table 10. Test results of the impact of regulations on provisioning – control variables added Can general provisions be included in

Is there any minimum provision re-

Tier II capital?

quired?

Have any legal penalties been imposed in at least the past five years?

+

Yes 0.070** (11.767)

No 0.029** (2.379)

Yes 0.005 (1.138)

No 0.049** (12.812)

Yes 0.047** (18.322)

No 0.004 (0.811)

+ − GDP × EBPT

-0.463** (-16.392)

-0.023 (-0.118)

0.086 (0.861)

-0.522** (-23.486)

-0.056 (-1.279)

-0.953** (-8.472)

− + GDP × EBPT

-0.020** (-4.617)

-0.082** (-4.635)

-0.008 (-1.469)

0.014* (1.756)

0.007 (0.833)

-0.011* (-1.804)

0.022** (4.431)

2.361 (0.383)

0.018** (6.861)

-0.859** (-11.723)

-0.717** (-0.855)

0.080** (3.763)

Loan growth

-0.003 (-1.020)

-0.004 (-0.601)

-0.021** (-6.400)

0.006** (3.514)

0.004** (2.402)

0.002 (0.732)

Equity

7.989** (2.478)

48.260** (4.797)

48.729** (2.975)

18.911** (4.947)

15.103** (3.634)

-173.312** (-8.366)

NPL growth

0.092** (5.752)

0.127** (3.469)

0.166** (9.095)

-0.005 (-0.891)

0.188** (9.053)

0.071** (17.788)

Net charge-off

0.791** (38.307)

0.076 (1.450)

0.467** (14.804)

0.827** (70.484)

0.845** (75.609)

0.478** (14.669)

0.841 3,133 1,274

0.647 96 203

0.817 1,127 548

0.577 2,102 929

0.872 1,991 699

0.792 1,160 607

+

GDP × EBPT

− − GDP × EBPT

Sargan test (p-value) No. of observations No. of banks

Note: The same as Table 6 and Table 9.

43

Banks and Bank Systems, Volume 3, Issue 4, 2008

John Mylonakis (Greece)

The influence of banking advertising on bank customers: an examination of Greek bank customers’ choices Abstract

The selection of banking advertising methods and means depends on a bank’s target group. The scope of this paper is to examine the relationship between bank advertising and the needs of a bank customer in Greece and its possible influence on potential customers to select their banks. The survey collected 260 questionnaires to provide the empirical dataset for technical inquiry based on descriptive statistics and correlation analysis. The research demonstrated the issue of customers indifference to advertising in their decision to cooperate with a bank. Advertising is not the determinant factor in their final choice. Selecting a banking institution is based on the traditional products and services it offers. However, its existence is a prerequisite, as it verifies a bank's critical presence in the market and plays an important role in customers’ choices. The examination of a banking institution is made based on price and product-related criteria and not promotion. Keywords: bank marketing, retail banking, banking advertisement, banks competition, banking products and services, bank selection. JEL Classification: M31.

Introduction♦

Banking advertising includes advertising performed by banking institutions. Apart from advertising addressed to bank customers (transformational), this category may also include business reports, information brochures of announcements on payment of new shares, reports on investment program results and several other financial announcements (informational). The first type contains dominant psychological messages, while the latter involves factual or logically verifiable messages, in order to impart banks’ products and services to current and potential customers (Punto & Wells, 1984; Rossiter & Percy, 1987). Retail banking has been undergoing a revolution since the abolition of controls and apart from size there is little to differentiate banking institutions from one another (Richardson & Robinson, 1986). The adoption and implementation of the marketing concept by banking institutions have been slow but profitable in many countries while in many other ones they are still product-oriented. Nevertheless, the traditional product-oriented banks are becoming increasingly customer-oriented focusing more and more on customer loyalty. The selection of banking advertising means depends on a bank’s target group. Most banking institutions address their advertising to holders of small accounts and thus advertise their products and services through the mass media. The press and television are the preferred means for larger banks that have branches across the country. Advertisements on investment programs usually appear in the trade press, while investment programs appear in almost all wide circulation newspapers as they are addressed to small investors. Many banking institutions rent space in ♦

© John Mylonakis, 2008.

44

several exhibitions and print information brochures to describe their activities and services; this is something that also insurance companies do even though their brochures have the form of participation requests. The scope of this paper is to examine the relationship of bank advertising with the needs of a bank customer in Greece and its possible influence on potential customers to select their banks. More specifically, it is researched the impact of communication messages provided by banks on various banking offers and packages (products & services) through their advertising strategies. 1. Past literature

Cheese et al. (1988) first indicated the importance of effective bank communications strategies. Despite the popularity of advertising in the academic field and the relating huge number of studies published on advertisement-related matters, no any previous study on the specific topic of financial advertising was found in the literature. This is not surprising as the same finding was stated by Laskey et al. (1992) in the past, where they emphasized the missing literature and the paucity of published research dealing with the effectiveness of bank advertising. Since then, many studies were published, especially for the Greek bank market (Koutouvalas et al., 2005; Mylonakis, 2004; Mylonakis et al., 1998), which covered spontaneously bank advertising issues, mainly as relatively small portion of overall bank customer researches, examining consumer satisfaction rates and more specifically whether or not bank customers were influenced by advertising to select their main or secondary bank. The only pioneering research paper found was that of Laskey et al. (1992) which dealt with strategy and structure in bank advertising. They examined the effectiveness of advertising on bank customers and found that

Banks and Bank Systems, Volume 3, Issue 4, 2008

respondents’ overall attitude and aesthetic/emotional evaluations varied significantly and that picturebased advertising elicits a higher intention to patronize a bank. They also emphasized the distinction between information and transformational advertising, the first found to be the most effective. 2. Research methodology 2.1. Bank customer profile. The research sample consists of 260 bank customers of all spectrums of banks in the Hellenic Bank market. 59.1% of the sample consists of men and 40.9% of women, with an average age of 34 years. The monthly family income of the sample can be classified in the following 4 categories: bank consumers with net income up to €20,000 represent 24% of the sample, those with net income between €20,001-30,001 represent 61%, those with net income between €30,001-40,000 represent 10%, and only 5% of the sample declares €40,001-50,000. As to marital status 75% were married and 25% divorced or nonmarried, and as to education 40% held a University degree, 34% had obtained a high school diploma and 16% held a post-graduate degree. All participants reside in various prefectures of Attica (Major Athens area). Bank customers who only cooperate with one banking institution represent 33.2%, while a slightly higher percentage of 36.9% represents those customers who cooperate with two banking institutions and, 29.9% declared that they cooperate with more than two banking institutions. 2.2. Methodology. For the purposes of this study questionnaires have been provided to collect information. At first, a pilot questionnaire was delivered to test whether it meets research needs. The distribution period of this questionnaire lasted a week and the results led to changes in its initial form, by ruling out some of the questions that were considered as irrelevant to the purpose of the study. The first page of the questionnaire was also added at this stage, along with the definition of bank advertising and modifications were made in certain questions for which the reliability analysis produced a Cronbach’s alpha coefficient below 0.6.

Reliability analysis was calculated for the scale in total, as well as, for each factor. Baker (1999) and Malhotra (1999) state that in the cource of the development of a measurement scale researchers emphasize the need for this scale to be reliable; that is the observations to be stable and coherent. The calculation of a scale’s reliability is based upon Cronbach alpha, which is the most widely used measurement method for this purpose (Peterson, 1994). Malhotra (1999) and Spector (1992) report that Cronbach’s Alpha indicator has to be at least 0.70 for a scale to be considered as a reliable one.

The Cronbach's Alpha statistic shown describes the consistency of each research factor. Consistency refers to the degree with which questions provide qualitative data for each factor. Values close to 1 indicate reliability of a factor, while usually only Cronbach’s Alpha values of 0.6 or above are acceptable for research use. Table 1. Reliability analysis Research factors

Cronbach's Alpha

No. of items

1. Receipt of information

.808*

8

2. Satisfaction from banking advertising

.798

2

3. Effectiveness of banking advertising

.748

4

4. Attitude towards banking advertising

.699

4

5. Satisfaction from banking institutions

.811

12

The sampling method used is random sampling. Twenty bank branches from all banking institutions were randomly selected and questionnaires were distributed to the customers of each bank separately. Each participant was given a brochure upon delivering their completed questionnaire; the brochure explained the purpose of the study and its expected results (debriefing). The testing stage of the pilot questionnaire aimed at assessing its degree of comprehension, acceptance and interpretation. More specifically: the potential comprehension of the questionnaire, i.e. whether the terms used can be easily conceived and whether the way in which the questions are expressed allows for the collection of the desired data. The selected type of questions used is closed-ended questions, except for 2 ones that refer to the age and place of residence and are open questions. The questionnaire consisted of 19 questions: the first five questions involved demographic data (gender, age, education, income, residence), three questions referred to closed questions, in which respondents were requested to choose between two statements (e.g., yes-no) and eleven questions were Likert-type ones. Data processing was carried out using the SPSS software, version 11.5. The research topics of all questions were concentrated into the following 5 factors: 1. Offer updates, relate to the way participants are informed of available banking services. 2. Satisfaction from advertising and banks' advertising strategies. 3. Factors affecting satisfaction from a bank (accounts, interest loans, credits, bonuses). 4. Advertisement effectiveness in bank selection. 5. Customers’ attitude towards advertisement, promotion techniques and packages offered. 45

Banks and Bank Systems, Volume 3, Issue 4, 2008

The above group of research topics is hypothesized as follows: H1. Providing information on banking offers through advertising strategies is associated with the selection of banking packages by customers. H2. The money spent by banking institutions on advertisement is associated with the degree of customer information on banking products. H3. The money spent by banking institutions on advertisement is associated with the degree of customer investment in promoted banking services. H4. The number of banking institutions with which customers cooperate is associated with the degree of customer information on banking products. 3. Research findings 3.1. Descriptive statistics. Factor 1: Offers information and updates. Results showed that 47.3% of bank customers are informed through television spots and 52.7% through other means. Regarding information through brochures, 40.9% receive information through brochures, as opposed to the remaining 59.1. Information through telephone communication represents 35.5%. Obtaining information through friends or family represents 35.5%.

Results on obtaining information through press advertising showed that 25.5% of bank customers stated that they have received information on banking services through the press, while 74.5% responded that they have not. Some of the questions that form the "Information" research factor are shown below (the number in the parentheses that follows each question refers to its coding): Have you previously been informed of this particular interest rate through television spots, press advertising, information brochures or telephone communication? Have you previously been informed of this ebanking service through television spots, press advertising, information brochures or telephone communication? Have you previously been informed of these loans through television spots, press advertising, information brochures or telephone communication? Have you previously been informed of this balance transfer program through television spots, press advertising, information brochures or telephone communication? Have you previously been informed of bonus programs through television spots, press advertising, information brochures or telephone communication? 46

Have you previously been informed of friendly and direct customer service through television spots, press advertising, information brochures or telephone communication? Do you follow banking advertisements? Factor 2: Bank customer satisfaction from advertising. The second factor refers to customer satisfaction from banking advertising. Customer satisfaction from advertising relates to the following questions: To what extent do you believe that advertisements meet what they claim to offer? and To what extent are you satisfied by banks' advertising strategies?

The study showed that 87.3% of bank customers are not happy with banks' promotional activities. On the other hand, only 12.7% are very satisfied by banking advertising. Factor 3: Factors affecting bank satisfaction. The questions that relate to customer satisfaction are:

It offers a higher bank account interest rate. 67.9% of the sample has responded positively about choosing a bank based on its interest rates, while 32.1 responded negatively. It offers the possibility to carry out transactions through the Internet (e-banking). 51.9% of the respondents pay attention to the option of receiving services through e-banking, while 48.1% do not. It offers lower-interest loans. Low interest rate of loans forms a criterion for 63% of the respondents. 7% are indifferent to this statement. It offers a balance transfer program. 57% have responded positively to this service and 43% are indifferent. It offers bonus programs to reward its customers. Customer rewards form a criterion for choosing a banking institution for 44.3% of the examined sample. 55.7% of the respondents feel indifferent about this statement. It offers more friendly and direct services (9Fa). Friendly and direct customer service is an important factor in selecting a banking institution for 78% of the respondents. For another 22% it is not important.

Results on customer satisfaction from banks' offered products regardless of their means of promotion showed that 74.5% are very satisfied from banking products and 25.5% are not very satisfied. Results, also, revealed that out of a total of 87 bank customers, who declared cooperation with one banking institution, 51 are very satisfied and 36 are not very satisfied. From those who declared cooperation with two banks, 82 out of 95 respondents stated that they are very satisfied and only 13 re-

Banks and Bank Systems, Volume 3, Issue 4, 2008

sponded negatively. Finally, 67 out of 78 who cooperate with more than two banks responded that they are satisfied with their banks, while 11 expressed dissatisfaction.

Hypothesis 1: Providing information on banking offers using marketing strategies is associated with the selection of banking packages by customers.

Data analysis led to the result that there is no correlation between the degree of obtaining information on banking services and the number of banking products purchased by customers (r = 0.501, p = 0.089). Due to this special finding, the relationship between customer information and selecting specific banking products will be examined separately. In this new analysis, banking institution products have been examined, such as interest rate, balance transfer and loans. Analysis showed that there is correlation between customer information on banking interest rates and the selection of related packages (r = 0.498, n = 110, p < 0.0005), that there is correlation between customer information on balance transfer and the selection of related packages (r = 0.650, n = 110, p < 0.0005), and that there is also correlation between information on banking loans and the purchase of loans (r = 0.559, n = 110, p < 0.0005).

Factor 4: Advertisement effectiveness. The following questions are related to the advertisement effectiveness on the selection of banking institutions: Has advertising helped you in selecting your bank? How many times have you been motivated by advertising in order to obtain more information on the services offered by a bank?, Would you ever choose a bank with low promotional coverage in the media (television, newspapers, magazines, radio) but with very good offered packages? Would you ever choose a new bank that has recently began operating in this sector, based only on its advertising campaign?

Results showed that 70.9% of bank customers believe that advertisement is very effective for a banking institution, while 29.1% thinks otherwise. 3.2. Correlation analysis. Bivariate correlation analysis was used to examine the following four hypotheses.

The following table shows all correlations among the 6 variables examined.

Table 2. Bivariate correlations among separate variables related to information and selection of banking packages

Higher interest rate Lower-interest loans

Higher interest rate

Lower-interest loans

Balance transfer

Information on interest rates

Information on loans

Information on balance transfer

1

.143

.002

.498(**)

.079

.077

1

.650(**)

.373(**)

.559(**)

.506(**)

1

.240(*)

.313(**)

.606(**)

1

.515(**)

.456(**)

1

.538(**)

Balance transfer Information on interest rates Information on loans Information on balance transfer

1

Note: * Correlation is important at a = 0.05 ** Correlation is important at a = 0.01.

Based on the above results, there is correlation between obtaining information on a specific banking product and selling it to consumers. Banking institutions, within the environment of competition in which they operate, provide consumers with informationon the variety and characteristics of their products in order to attract them. This load of information on all products is not of interest to consumers, who focus only on the products they want to obtain. They seek to get more details on the characteristics on certain products offered by banking institutions and, based on these products, they compare all banks and select the one that offers a competitive advantage against the other, i.e. the one whose products entail the characteristics that meet their needs better compared with the other.

Hypothesis 2: The money spent by banking institutions on advertising is associated with the degree of customer information on banking products.

Data analysis was performed using the variance method (ANOVA) and reached that there are statistically significant differences between the degree of customer information by three different banking institutions in which random sampling was applied (F = 4.76, df = 2, p = 0.01). A daring assumption following the above analysis results is that every banking institution follows its own advertising policy and applies its own strategies. Part of an advertising strategy is also the decision on the importance paid to information campaigns on people and, therefore, the frequency and the quality of the provision of information. Given the inability to obtain infor47

Banks and Bank Systems, Volume 3, Issue 4, 2008

mation on operational costs spent by banks on advertising and on informing customers, an explanatory approach that can be put forth with reservation is that the degree of providing information by each bank also represents the amount of money available by each bank for marketing purposes. Hypothesis 3: The money spent by banking institutions on advertisement is associated with the degree of customer investment in promoted banking services.

Continuing with the data analysis and the ANOVA variance method, the third hypothesis is confirmed: advertising costs relate to the degree of customer investments in promoted banking services. Advertisement, being a tool within a competitive environment, forces banks to invest more money in promotion in order to become competitive, i.e. to ensure a top place among customer choices and increase their profits at the same time. Bank expenses for advertising produce the desired result: clients actually invest in the promoted products and services. Advertising is the only sector that offers money and profits instead of spending it. The relationship between advertising and attracting customers is proportional. Therefore, those banking institutions that spend less on advertising will fall behind. Therefore, considering the daring assumption of the second hypothesis, each one of the three banking institutions perceives the provision of information and marketing in different ways. Hypothesis 4: The number of banking institutions with which customers cooperate is associated with the degree of customer information on banking products.

The fourth hypothesis relates to whether the number of banking institutions with which customers cooperate relates to the amount of customer information on banking products. The ANOVA variance method was also applied in this scenario and the results indicated that there is no correlation between the provision of information and the number of banks customers cooperate with (F = 0.016, df = 2, p = 0.984). Informing customers on banking packages and offers is not associated with the number of banks they cooperate with. Cooperating with many banking institutions originates from competition in the banking sector and is not an indication of the level of customer information. Banks, in their effort to be competitive, launch tempting offers in some of their products; unlike their competitors, being unable to do the same thing for the same product, they focus on a different product or service, thus enabling customers to benefit from these opportunities. This fact does not prove that the consumers themselves obtain maximum information to select their bank. Banks make 48

sure that their offers and products are made known to the public through promotion and advertising. Conclusions

The scope of this paper is to examine the relationship between bank advertising and the needs of a bank customer in Greece and its possible influence on potential customers to select their banks. By considering all the results drawn by this study, one may conclude certain things about banking advertising and the role it plays within the competitive environment in which banks operate. The most important conclusion is subversive for the concept and meaning of the term of advertising but not for everything associated with it. This study has demonstrated the issue of customers’ indifference to advertising in their decision to cooperate with a bank. Advertising is not the determinant factor in their final choice. Selecting a banking institution is based on the traditional products and services it offers. The examination of a banking institution is made based on price and product-related criteria and not promotion. Still, the fact that consumers refuse to cooperate with banks that do not have any promotional activity confirms that the role of marketing is very important for banking competition. Bank customers may not be interested in advertising at first while choosing their bank, but this is the initial reaction of all those who are interested in achieving the most cost-beneficial and favorable terms. In reality, the absence of advertising acts as a suspending factor in choosing a bank because, even though people are interested in getting good terms, they do wish that these terms come from a bank that has made its presence clear within this competitive environment, follows sector developments, competes with other banks and offers the same or improved products as its competitors. Such a banking institution inspires reliability and security, which are essential elements for consumer investments. Therefore, the conclusion is that advertising is not the main criterion for consumers in choosing their bank. However, its existence is a prerequisite, as it verifies a bank's critical presence in the market and plays an important role in their choices. Managerial implications. Banking institutions in today's world should develop multiple effective and supplementary activities to achieve benefits both for the state and for citizens. Banking institutions must spend money on technology and equipment in order to be able to promptly follow developments and secure their place among their competition. It is important that they adopt electronic banking systems, in order to lower their operating costs, remain competitive, maintain their old customers and continue to attract new devoted customers.

Banks and Bank Systems, Volume 3, Issue 4, 2008

Moreover, due to the increasing competition brought about by the entry of non-banking institutions in electronic banking, it is important to strengthen their marketing campaigns. Finally, electronic banking services ought to be more personalized and adjust to the particular needs of their customers. Equal importance should be paid to developing e-

commerce, which is expected to form the primary transaction channel in companies. Security systems relating to the operation of computerized and electronic systems, as well as safeguarding confidentiality and data processing will become a critical factor for the sustainability and growth of banking institutions.

References

Baker T.L. (1999), “Doing Social Research”, 3rd Edition. McGraw Hill, Boston. Cheese J., Day A., Wills G. (1988), “Handbook of Marketing and Selling Bank Services”, International Journal of Bank Marketing, Vol. 6, No. 3. 3. Laskey H.A., Seaton B., Nicholls J.A.F. (1992), “Strategy and Structure in Bank Advertising: An Empirical Test”, International Journal of Bank Marketing, Vol. 10, No. 3, pp. 3-9. 4. Malhotra N.K. (1999), “Marketing Research: An Applied Orientation”, 3rd ed., Prentice-Hall International, London. 5. Mylonakis J., Malliaris P., Siomkos G. (1998), “Marketing-driven factors influencing savers in the Hellenic Bank Market”, Journal of Applied Business Research, Vol. 14, No. 2, pp. 109-116. 6. Mylonakis J. (2004), “Bank market positioning maps: customer perceptions of Hellenic financial Services”, International Journal of Services Technology and Management, Vol. 5, No. 2, pp. 140-150. 7. Koutouvalas D., Siomkos G., Mylonakis J. (2005), “Perceived Service Quality Management and Loyalty in Public versus Private Banks’ Operations: Some Empirical Evidence”, International Journal of Services and Operations Management, Vol. 1, No. 2, pp. 101-122. 8. Peterson, R.A. (1994), “A Meta-analysis of Cronbach’s Coefficient Alpha”, Journal of Consumer Research, 21 (21): 381-391. 9. Puto C.P., Wells W.D. (1984), “Informational and Transformational Advertising: The Differential Effects of Time”, in Kinnear T.C. “Advances in Consumer Research XI, Association for Consumer Research, Provo, UT, pp. 638-643. 10. Richardson B. and Robinson C.G. (1986), “The impact of internal marketing on customer service in a retail bank”, International Journal of Bank Marketing, Vol. 4, No 5. 11. Rossiter J.R., Percy L. (1987), “Advertising and Promotion Management”, McGraw-Hill, New York. 12. Spector P.E. (1992), “Summated Rating Scale Construction: An Introduction”, Sage University Paper Series No. 82, On Quantitative Applications in the Social Sciences, Beverly Hills CA: Sage. 1. 2.

49

Banks and Bank Systems, Volume 3, Issue 4, 2008

Seok Weon Lee (Korea)

Asset size, risk-taking and profitability in Korean banking industry Abstract

From the cross-sectional data of Korean banks before and after the regulation of banking industry was tightened in the late 1997, we find significant evidences that larger banks have very perverse, unprofitable risk-taking incentives, but only when the regulations are loose. Thus, the typical moral hazard of larger banks does exist only in the period of deregulation in our sample. After regulations are tightened, the perverse risk-taking behavior disappears and risk-taking behavior becomes more profitable. Based on our findings, the following policy implication could be derived. If consolation and merger are believed to be one of the easiest ways to capture greater market share and make higher profit, and many banks follow this wave and trend, regulators that are required to maintain the safe and sound environment for banking should understand that the resulting net effect of this pattern on the bank’s risk and profitability, and therefore, on the safety and soundness of the entire banking industry will depend on the extent to which such activities are accompanied by proper monitoring of regulators. Keywords: asset size, risk taking, moral hazard, profitability, bank regulation. JEL Classification: G29.

Introduction•

Many researchers in the banking literature tend to presume that size and safety do not go hand in hand, especially when risk is measured as ex-ante risktaking incentives. They provide many results that large banks, while better diversified than small banks in asset-portfolio composition, are not less risky than small banks. Rather, large banks tend to use their diversification advantage to increase the riskiness of their activities such as by increasing risky lending and operating with less capital but not to operate at lower levels of overall risk. Generally, large banks realize a cost advantage over small banks because of their ability to operate with less capital. It is generally agreed that the less the capital (the higher the leverage), the riskier the firm is, because of both leverage effect and the moral-hazardincentives of stockholders associated with limited liability. Furthermore, investors would not have great incentives to monitor the risk-taking behavior of large banks because they may believe that regulators will not allow the failures of large banks due to the potential damage to the economy. Therefore, large banks would have some moral hazard incentives to try to take advantage of this less monitoring motivation to increase their riskiness. Liang and Rhoades (1991) find that large banks operate with lower capital ratios to pursue riskier activities. McAllister and McManus (1993) also show that large banks operate with lower capital ratios. Akhavein, Berger and Humphrey (1997) show that the profit efficiency associated with large-bank mergers is, at least in part, attributable to a shift in outputs from low-risk securities to higher-risk loans. However, whether the riskier activities pursued by large banks enhance profit of the banks is questionable. If the riskier activities do not contribute to •

© Seok Weon Lee, 2008.

50

higher profits, or result in worse performance, this may have to be interpreted as a typical moral hazard effect suggested by finance literature. Empirical results with respect to the relationship between large banks’ risk-taking and profit are rather mixed up. Benston, Hunter, and Wall (1955) show some results supporting that size and the levels of cash flow are positively correlated. Boyd and Runkle (1993) find different results. In this paper, we examine whether larger banks in Korean banking industry pursue riskier activities than smaller banks during the period of 1994-2005, employing both the measures for ex-ante risk-taking incentives and those for ex-post risk-taking. Especially, we compare the risk-taking behavior of large banks for two different regulatory regimes: the period of deregulation (pre-1998) vs. the period of tight and strict regulation (post-1998). Following the predictions and findings of many researchers, we believe that the ability of stockholders to maximize their profit by greater risk-taking would be enhanced in periods of deregulation and regulatory forbearance. Consequently, we presume that if larger banks had greater risk-taking incentives than smaller banks, this pattern would be stronger and more transparent when regulations are loose. To overcome the financial crisis in the late 1997, the regulations of the Korean banking industry became very tight and strict after 1997. Thus it would be a good sample period to examine the above issue. Furthermore, we examine whether the riskier activities pursued by large banks enhance profit. By this profitability test, ex-post we evaluate whether the risk-taking behavior of banks was driven by perverse moral hazard incentives or by deliberate and careful decision making. From the cross-sectional data of Korean banks before and after the regulation of banking industry was tightened in the late 1997, we find significant

Banks and Bank Systems, Volume 3, Issue 4, 2008

evidences that larger banks have very perverse, unprofitable risk-taking incentives, but only when the regulations are loose. Thus, the typical moral hazard of larger banks does exist only in the period of deregulation in our sample. After regulations are tightened, the perverse risk-taking behavior disappears and risk-taking behavior becomes more profitable. Based on our findings, the following policy implication could be derived. If consolation and merger are believed to be one of the easiest ways to capture greater market share and make higher profit, and many banks follow this wave and trend, regulators that are required to maintain the safe and sound environment for banking should understand that the resulting net effect of this pattern on the bank’s risk and profitability, and therefore, on the safety and soundness of the entire banking industry will depend on the extent to which such activities are accompanied by proper monitoring of regulators. In the next section, we describe the sample of banks. In section 2, we describe the hypotheses to be tested and the regression model used to test them. In section 3, we present the empirical results and in the lest section offer concluding remarks. 1. Sample and data

We collect the balance sheet data of banks such as asset size, equity capital, loans, fixed assets, nonperforming loans, and return on asset from the Statistics of Bank Management for each year, from 1994 to 2005, published by the Korean Financial Supervisory Service. 2. Testable hypotheses, testing models and correlation test

To examine how the risk-taking incentives of banks are associated with asset size, we estimate the following cross-sectional, univariate regression equation for each year during the period of 1994-2005 to eliminate serial-correlation problem. We omit 1997 because 1997 is a transitional year for the introduction of new regulations. (Risk-taking)i = β0 + β1(Asset size)i + εi .

(1)

Risk-taking for each individual bank i is proxied by alternative balance sheet measures. We employ capital-to-asset ratio, loan-to-asset ratio, and fixed asset-toasset ratio as the measures for the bank’s ex-ante risk-

taking incentives, and non-performing loans-to-loans ratio as the measure for the bank’s ex-post risk-taking. The first one is the capital-to-asset ratio. As discussed in this paper’s introduction, lower capital-toasset ratio is believed to represent greater risktaking incentives. The second one is the ratio of loans to total asset. Generally, loans are considered to be risky assets and are given high risk weight at the calculation of BIS (Bank for International Settlement) capital ratio. It is generally agreed that the greater the loan ratio of a bank, the more vulnerable the performance of the bank is to future economic conditions. Thus, other things being equal, higher loan-to-asset ratio is believed to represent higher risk-taking incentives. The third one is the fixed asset-to-asset ratio or operational leverage. It is very well known that operational leverage acts in a similar way to financial leverage (capital-to-asset ratio) in increasing firm risk. Thus, other things being equal, higher fixed asset-to-asset ratio is believed to represent higher risk-taking. We test our hypotheses for two interpretations of risk-taking measures, the ex-ante risk-taking incentives and the ex-post risktaking. We believe that ex-ante risk-taking incentives are more germane and include the ex-post risktaking for completeness of our test. As the measure for ex-post risk-taking, we employ the bank’s nonperforming loans-to-loans ratio. 3. Empirical results for regression analysis 3.1. Measures for ex-ante risk-taking incentives. Tables 1-3 present the results for the change in moral hazard of large banks associated with three different measures for ex-ante risk-taking incentives over the period of 1994-2005. Table 1 presents the results for the case where the risk of a bank is measured by its financial leverage, capital-to-asset ratio. As shown in the table, the slope coefficient is significantly negative for the period of 1994-1996, indicating that the larger banks have moral hazard associated with low capital when the regulations of the banking industry are loose. However, for the period of 1998-2005 after the regulations are tightened, the significant negative sign does not exist. All the coefficients are statistically insignificant. Indeed, most of the coefficients have positive sign. This may indicate that the regulations introduced in this period have been effective in moderating risk-taking in large banks to have low capital ratio.

Table 1. Regression results for capital-to-asset ratio Year

Intercept

Slope coefficient

t-value of the slope coefficient

p-value of the slope coefficient

R2

Standard error of regression

F-statistic

1994

0.0863***

-1.3×10-7 ***

-3.07

0.0055

0.30

0.0235

9.4359 ***

1995

0.0715***

-8.4×10-8 ***

-2.91

0.0078

0.27

0.0202

8.4846 ***

1996

0.0618***

-5.8×10-8 ***

-2.82

0.0096

0.26

0.0169

7.9795 ***

51

Banks and Bank Systems, Volume 3, Issue 4, 2008

Table 1 (cont.). Regression results for capital-to-asset ratio Year

Intercept

Slope coefficient

t-value of the slope coefficient

p-value of the slope coefficient

R2

Standard error of regression

F-statistic

1998

0.0085***

3.7×10-8

1.61

0.1236

1999

0.0328***

1.08×10-8

1.13

0.2740

0.12

0.0273

2.6096 ***

0.08

0.0099

2000

0.0335***

7.56×10-9

1.08

0.2966

0.07

0.0080

1.2893 ** 1.1692 **

2001

0.0346***

6.88×10-9

1.55

0.1439

0.16

0.0078

2.4175 ***

2002

0.0409***

-8.3×10-10

-0.20

0.8381

0.08

0.0081

0.0436

2003

0.0445***

-4.1×10-9

-1.13

0.2805

0.10

0.0077

1.2772 **

2004

0.0479***

6.56×10-10

0.17

0.8649

0.10

0.0075

0.0302

2005

0.0501***

8.33×10-9

1.66

0.1227

0.18

0.0103

2.7557 ***

Note: (Capital-to-asset)i = β0 + β1(Asset size)i + εi. This table shows the cross-sectional univariate regression results for capital-to-asset ratio. *, **, or *** indicate statistical significance at the 10, 5, or 1% significance level, respectively.

Table 2 presents the results for the change in moral hazard of large banks associated with loan-to-asset ratio. As shown in the table, the slope coefficient is significantly positive for the period of 1994-1996, indicating that the larger banks have moral hazard associated with high loan ratio when the regulations of the banking industry are loose. However, for the

period of 1998-2005 after the regulations are tightened, the significant positive sign does not exist. All the coefficients are statistically insignificant. The regulations introduced in this period have been effective in moderating risk-taking in large banks to have high loan ratio.

Table 2. Regression results for loan-to-asset ratio Year

Intercept

Slope coefficient 1.81×10-7 *

t-value of the slope coefficient 1.97

p-value of the slope coefficient 0.0605

1994

0.4203***

1995

0.4027***

6.43×10-8 **

2.09

0.0477

1996

0.3968***

1.29×10-7 **

2.21

1998

0.3698***

6.75×10

-8

1999

0.3946***

7.49×10-8

2000

0.4350***

2001 2002

0.15

Standard error of regression 0.0520

3.9129 ***

0.16

0.0451

4.3748 ***

0.0370

0.18

0.0483

4.8986 ***

1.45

0.1639

0.10

0.0544

2.1058 ***

1.31

0.2081

0.10

0.0598

1.7297 **

5.99×10-8

1.05

0.3076

0.07

0.0655

1.1152 **

0.4489***

1.02×10

-8

0.31

0.7570

0.01

0.0578

0.0998

0.5219***

3.54×10-9

0.12

0.9002

0.00

0.0566

0.0163

2003

0.5568***

-9.1×10

-0.29

0.7698

0.01

0.0643

0.0895

2004

0.5555***

5.68×10-9

0.17

0.8614

0.00

0.0631

0.0317

2005

0.5563***

2.43×10-10

0.01

0.9933

0.00

0.0585

0.0000

-9

R2

F-statistic

Note: (Loan-to-asset)i = β0 + β1(Asset size)i + εi. This table shows the cross-sectional univariate regression results for loan-to-asset ratio. *, **, or *** indicate statistical significance at the 10, 5, or 1% significance level, respectively.

Table 3 presents the results for the change in moral hazard of large banks associated with operational leverage, fixed asset-to-asset ratio. As shown in the table, the slope coefficient is positive for the period of 1994-1996 as we expected. However, none is statistically significant, and therefore, we do not have any strong evidence that the larger banks have

moral hazard associated with operational leverage when the regulations of the banking industry are loose. However, for the period 1999-2005 after the regulations are tightened, the coefficient is significantly negative indicating that the larger banks have significantly reduced the risk-taking incentives to have high fixed-asset ratio.

Table 3. Regression results for fixed asset-to-asset ratio

52

Year

Intercept

Slope coefficient

t-value of the slope coefficient

p-value of the slope coefficient

R2

Standard error of regression

F-statistic

1994

19.7712***

8.09×10-6

0.44

0.6622

0.01

10.3972

0.1960

1995

21.3821***

1.66×10-5

0.97

0.3403

0.04

11.9628

0.9479 **

1996

22.7294***

1.05×10-5

0.76

0.4535

0.02

11.3816

0.5813 *

1998

155.80***

0.02×10-4

0.46

0.6514

0.01

572.03

0.2109

Banks and Bank Systems, Volume 3, Issue 4, 2008

Table 3 (cont.). Regression results for fixed asset-to-asset ratio Year

Intercept

Slope coefficient

t-value of the slope coefficient

p-value of the slope coefficient

R2

Standard error of regression

F-statistic

1999 2000

67.8810***

-4.1×10-5 **

-2.21

0.0426

0.25

19.6014

4.9274 ***

69.8130***

-5.1×10-5 ***

-3.09

0.0073

0.39

19.1003

9.5902 ***

2001

55.3019***

-2.3×10 **

-2.50

0.0265

0.32

16.2713

6.2552 ***

2002

43.0117***

-1.2×10-5 *

-2.01

0.0677

0.25

11.8677

4.0321 ***

2003

39.1866***

-1.0×10-5 **

-2.31

0.0392

0.31

9.4422

5.3567 ***

2004

36.3469***

-1.1×10 **

-2.76

0.0172

0.38

7.8418

7.6257 ***

2005

30.2745***

-8.1×10-6 **

-2.72

0.0422

0.30

7.3210

5.1639 ***

-5

-5

Note: (Fixed asset-to-asset)i = β0 + β1(Asset size)i + εi. This table shows the cross-sectional univariate regression results for fixed asset-to-asset ratio. *, **, or *** indicate statistical significance at the 10, 5, or 1% significance level, respectively.

3.2. Ex-post risk-taking measure and profitability. Tables 4 and 5 present the results for the ex-post evaluation of large banks risk-taking behavior employing non-performing loans’ ratio (ex-post risktaking measure) and return on asset (profitability measure), respectively. Table 4 shows that the slope coefficient on asset size with respect to nonperforming loan ratio is significantly positive during the period of 1994-1995. This result, combined with the results found in the previous sections, indicates that the greater risk-taking of larger banks associated with low capital and high loan ratio turns out to be very unprofitable, and therefore, could be a strong evidence that the larger banks have perverse moral hazard incentives when the regulations are loose. Overall, the greater the asset size is, the

greater the risk-taking incentives are, but the more problem assets larger banks have. This is the typical moral hazard effect suggested by the literature. This conclusion is supported by the insignificant coefficient with respect to return on asset in Table 5 as well. However, after the regulations are tightened, the perverse moral hazard of larger banks disappears as indicated mainly by the insignificant coefficients with respect to non-performing loan ratio. Tables 4 and 5, respectively, show that the non-performing loan ratio of larger banks is significantly decreased and their return on asset is significantly increased in 1998 right after the regulation is tightened, indicating larger banks try to improve their risk status toward safer ones and pursue more profitable and deliberate strategies.

Table 4. Regression results for non-performing loans-to-loans ratio Year

Intercept

1994 1995 1996 1998 1999 2000 2001 2002 2003 2004 2005

2.1050*** 3.0524*** 3.5318*** 12.68*** 9.3103*** 9.2209*** 3.0272*** 1.7302*** 1.5853*** 1.5222*** 0.9853***

Slope coefficient 1.65×10-5 *** 7.33×10-6 ** 1.52×10-6 -1.1×10-5 * -1.8×10-6 -4.6×10-6 -1.7×10-7 1.86×10-7 4.75×10-7 1.6×10-7 4.7×10-8

t-value of the slope coefficient

p-value of the slope coefficient

R2

Standard error of regression

F-statistic

4.81 2.58 0.63 -1.86 -0.49 -1.26 -0.18 0.46 1.12 0.62 0.41

0.0000 0.0164 0.5347 0.0799 0.6276 0.2279 0.8583 0.6526 0.2840 0.5440 0.6838

0.51 0.22 0.02 0.16 0.02 0.09 0.00 0.02 0.09 0.03 0.01

1.9460 1.9874 2.0016 6.6672 3.7226 4.2413 1.6948 0.8258 0.8930 0.5095 0.2323

23.23 *** 6.6969*** 0.3972 3.4432 *** 0.2452 1.5804 *** 0.0331 0.2130 1.2577 ** 0.3899 0.1741

Note: (Non-performing loans-to-loans)i = β0 + β1(Asset size)i + εi This table shows the cross-sectional univariate regression results for non-performing loans-to-loans ratio. *, **, or *** indicate statistical significance at the 10, 5, or 1% significance level, respectively.

Table 5. Regression results for return on asset Year

Intercept

Slope coefficient

t-value of the slope coefficient

p-value of the slope coefficient

R2

Standard error of regression

F-statistic

1994

0.4845***

-2.8×10-7

-0.56

0.5801

0.01

0.2849

0.3153

1995

0.3166***

1.64×10-8

0.03

0.9746

0.00

0.3579

0.0010

1996

0.3121***

-1.1×10-7

-0.27

0.7862

0.00

0.3466

0.0752

1998

-6.6010***

7.17×10-6 ***

2.81

0.0113

0.30

2.9802

7.9458 ***

1999

-0.9201***

-5.2×10-7

-0.21

0.8358

0.00

2.5915

0.0444

53

Banks and Bank Systems, Volume 3, Issue 4, 2008

Table 5 (cont.). Regression results for return on asset Year

Intercept

Slope coefficient

t-value of the slope coefficient

p-value of the slope coefficient

R2

Standard error of regression

F-statistic

2000

-1.3227***

1.45×10-6

0.98

0.3420

0.06

1.7113

0.9624 *

2001

0.3894***

4.0×10-7

1.32

0.2094

0.12

0.5414

1.7433 ***

2002

0.6423***

-3.5×10-8

-0.12

0.9006

0.00

0.5691

0.0162

2003

0.5619***

-3.7×10-7

-1.04

0.3156

0.09

0.7533

1.0966 **

2004

0.7482***

1.35×10-7

0.54

0.5974

0.02

0.4937

0.2942

2005

0.9023***

2.89×10-7

0.91

0.3777

0.06

0.6491

0.8387 *

Note: (Return on asset)i = β0 + β1(Asset size)i + εi. This table shows the cross-sectional univariate regression results for return on asset. *, **, or *** indicate statistical significance at the 10, 5, or 1% significance level, respectively.

Concluding comments

From the cross-sectional data of Korean banks before and after the regulation of banking industry was tightened in the late 1997, we find significant evidences that larger banks have very perverse, unprofitable risk-taking incentives, but only when the regulations are loose. Thus, the typical moral hazard of larger banks does exist only in the period of deregulation in our sample. After regulations are tightened, the perverse risk-taking behavior disappears and risk-taking behavior becomes more profitable. Based on our findings, the following

policy implication could be derived. If consolation and merger are believed to be one of the easiest ways to capture greater market share and make higher profit, and many banks follow this wave and trend, regulators that are required to maintain the safe and sound environment for banking should understand that the resulting net effect of this pattern on the bank’s risk and profitability, and therefore, on the safety and soundness of the entire banking industry will depend on the extent to which such activities are accompanied by proper monitoring of regulators.

References

1. 2. 3. 4. 5. 6. 7. 8. 9.

54

Akhavein, J.D., Berger, A.N., and Humphrey, D.B. (1997). “The Effects of Bank Megamergers on Efficiency and Prices”, Review of industrial Organization 12. Benston, G., R. Eisenbeis, P. Horvitz, E. Kane, and G. Kaufman. (1986). “Perspectives on Safe and Sound Banking: Past, Present, and Future”, Cambridge, Mass.:MIT Press. Chen, A.H., M. Cornett, S. Mazumdar, and H. Teheranian. (1999). “An Empirical Analysis of the Effects of the FDICIA of 1991 on Commercial Banks,” Research in Finance 17, pp. 41-64. Demsetz, R.S., and Strahan, P.E. (1997). “Diversification, Size, and Risk at Bank Holding Companies,” Journal of Money, Credit, and Banking 29, pp. 300-313. Galloway, T.M., W.B. Lee, and D.M. Roden. (1997). “Banks’ changing incentives and opportunities for risk taking”, Journal of Banking and Finance 21, pp. 509-527. Gunther, J.W., and Robinson, K.J. (1990). “Empirically assessing the role of moral hazard in increasing the risk exposure of Texas banks”, Federal Reserve Bank of Dallas Economic Review. Mckenzie, J.A., Cole, R.A., and Brown, R.A. (1992). “Moral hazard, portfolio allocation, and asset returns for thrift institutions”, Journal of Financial Research, pp. 315-339. O’Hara, M. and W. Shaw. (1990). “Deposit Insurance and wealth Effects: The Value of Being “Too Big to Fail”, Journal of Finance 5, pp. 1587-1600. Saunders A., E. Strock, and N.G. Travlos. (1990). “Ownership Structure, Deregulation, and Bank Risk Taking”, Journal of Finance 2, pp. 643-654.

Banks and Bank Systems, Volume 3, Issue 4, 2008

Christos Floros (UK), Georgia Giordani (UK)

ATM and banking efficiency: the case of Greece Abstract

This paper shows how useful the number of ATMs is for modelling and estimating banking efficiency. We examine banking efficiency for Greece using data from top 10 commercial banks. To estimate banking efficiency we employ DEA and FDH using three inputs (number of employees, number of branches and number of ATMs) and one output (loans). We find that large banks are more efficient than medium and small sized banks. Furthermore, we report that banks with a large number of ATMs are more efficient than those with a less number of ATMs. Finally, we conclude that the provision of e-banking services by banks does not influence their efficiency scores. Keywords: banking efficiency, e-banking, ATM, Greece. JEL Classification: E58, G21.

Introduction•

The performance of the banking sector has become a very popular topic, especially after the provision of electronic banking services by banks, such as ATMs, Internet banking, Telephone banking etc. The usage of these technologies has increased significantly during the past 15 years mainly due to the fact that they are offering various advantages to banking customers. The use of ATMs is also very widespread as in the last years they have been used for other purposes apart from cash withdrawals. In the recent years, ATMs are able to offer their customers a wide range of services. ATMs accept deposits of money and checks, print mini statements, check customers’ balances, proceed payments of utility bills and transfer funds to other bank accounts. In addition, customers can purchase credit for their ‘pay and go’ mobile phones, as well as they can purchase other goods and services, for example train and concert tickets. Banks gain more revenues (and increase their efficiency) by offering e-banking services such as ATMs, in addition to the reduction of their costs, as fewer physical branches are needed and consequently fewer employees. A large number of studies examine banking efficiency. Duncan and Elliott (2004) explain that the concept of efficiency can be regarded as the relationship between outputs of a system and the corresponding inputs used in their production. Efficiency is treated as being a relative measure that reflects the deviations from maximum attainable output for a given level of input (English et al., 1993). Farrell (1957) draws upon the studies of Debreu (1951) and Koopmans (1951) in order to define a simple efficiency measure which would be able to account for multiple inputs. He argues that the efficiency of a firm consists of technical and allocative efficiency. Technical efficiency is the ability of a firm to obtain •

© Christos Floros, Georgia Giordani, 2008.

maximum output from a given set of inputs, and allocative efficiency is the ability of a firm to use the inputs in optimal proportions given their respective prices. When these two measures are combined they provide a measure of total economic efficiency. In this paper, we examine banking efficiency for Greece using three inputs (number of employees, number of branches and number of ATMs) and one output (loans). A limited number of papers include the number of ATMs as an input to estimate banking efficiency. This paper shows how useful ATM is for modelling and estimating banking efficiency using data from 10 commercial banks from Greece. 1. Methodology

There are a number of different approaches that can be followed in order to examine the efficiency of banks. These include the Stochastic Frontier Analysis (SFA), Thick Frontier Approach (TFA), Distribution Free Approach (DFA), Free Disposal Hull (FDH) and the Data Envelopment Analysis (DEA). This paper uses both the DEA and FDH to estimate the efficiency of Greek banks. We describe the DEA input and output orientated models by using Constant, Variable, Increasing and Decreasing Returns to scale approaches. 1.1. Input-orientated model. Farrell (1957) uses a simple example which involves the use of two inputs (x1 and x2) to produce a single output (y) under the constant returns to scale assumption. It should be noted that this assumption allows the presentment of the technology by using a unit isoquant.

Figure 1 shows the unit isoquant if a fully efficient firm which is shown as SS’ in Figure 1 and measures the technical efficiency. In point P a firm can use quantities of inputs to produce a unit of output, and the technical inefficiency is represented by the distance QP which is the amount by which inputs could be proportionally reduced without reducing the amount of output. This is also expressed in percentage by the ratio QP/OP which represents the 55

Banks and Bank Systems, Volume 3, Issue 4, 2008

percentage by which all inputs could be reduced. The technical efficiency (TE) of a firm is measured by the ratio TE =

OQ QP which is equal to 1. OP OP

Technical efficiency takes the value between one and zero and therefore indicates the degree of a firm’s inefficiency. The value one indicates that the firm is fully technically efficient. Point Q is considered technically efficient due to the fact that it lies on the efficient isoquant.

Source: Coelli (1996). Fig. 2. Output-orientated model

We also define the overall economic efficiency as the product of technical and allocative efficiency: EE =

Fig. 1. Input-orientated model

AA' line represents the input price rate, also known as allocative efficiency. The allocative efficiency (AE) of a firm can be calculated by the ratio AE = OR . RQ represents the reduction in production OQ

costs that would occur if production were to occur at the allocative (and technically) efficient point Q' instead of occurring at the technically efficient but allocative inefficient point Q. The combination of technical and allocative efficiency provides the total economic efficiency. Total economic efficiency is defined by the ratio EE OR . RP is the distance where there is a reduction OP OQ OR OR * = = EE. All in the cost. TE * AE = OQ OP OP

=

the three measures take values between one and zero. 1.2. Output-orientated model. Furthermore, Figure 2 presents the output orientated model. AB is the distance which represents technical inefficiency; the amount by which outputs could be increased without the need of extra inputs. The technical efficiency of the output orientated model is given by the ratio

TE =

OA . OB

We also define the allocative efficiency if we draw the isorevenue line DD' in the case that we have OB . available the price information AE = OC

56

OA OB OA = x = TE × AE. OC OB OC

1.3. Data envelopment analysis (DEA). DEA is a non-parametric method which was first introduced by Charnes et al. (1978) and has been used ever since to measure the empirically derived relative efficiency (Molyneux et al., 1996 ). It computes a comparative ratio of outputs to inputs for each decision making unit (DMU), which is reported as being the relative efficiency score (Avrikan, 2005). This efficiency score is usually expressed as a number between zero and one or zero and 100 per cent. A decision making unit with a score less than one is considered to be an inefficient unit relatively with other DMUs with score equal to one which are considered to be efficient. Traditionally, DEA measures the technical efficiency of DMUs opposed to their allocative efficiency1. According to Pasiouras (2006), the best practice production frontier for a sample of DMUs is constructed through a linear combination of input-output sets that envelops the input-output correspondence of all DMUs in our sample.

Charnes et al. (1978) propose a model which was input orientated and assumed constant returns to scale, while Pasiouras (2006) reports that the use of constant returns to scale is appropriate when all firms are operating at an optimal scale. In the case where firms are not operating at an optimal scale due to various constraints in finance or imperfect competition then the use of variable returns to scale was suggested by Banker et al. (1984) so as to calculate technical efficiency without containing any scale efficiency effects. Damar (2005) argues that 1

Technical efficiency examines how well the production process converts inputs into outputs whereas allocative efficiency is the effective choice of inputs vis a vis prices with the objective of minimising the costs of production.

Banks and Bank Systems, Volume 3, Issue 4, 2008

inefficiency is measured as the distance between the firms’ observed input-output combination and the frontier. Halkos and Salamouris (2004) argue that the main advantage of the DEA technique is that it can deal with the case of multiple inputs and outputs, as well as factors, which are not controlled by individual management. Another advantage is that the method skips some usual problems which arise from the use of parametric methods in the analysis of financial ratios. The usual problems are considered as being the need to determine the functional form or to determine the statistical distribution of the ratios. In addition, problems arise when the numerator or the denominator of the financial ratios take negative values. According to Gutierrez-Nieto et al. (2007), DEA can be used when the conventional cost and profit functions can not be justified. Damar (2005) explains that the DEA method allows for zero output values and handles zero input values. According to Sufian (2006), DEA is less data demanding as it handles small sample sizes. Additionally, DEA does not require a preconceived structure or a specific functional form to be imposed on the data in the process of identifying and determining the efficient frontier. Wu et al. (2005) argue that the DEA technique allows efficiency to change over time and it does not require prior specification of the best production frontier. However, Farc et al. (2000) identify that one of the most important objections to the DEA model is its non stochastic nature. When the DEA model is used any deviations from the frontier are attributed to inefficiency. They report that DEA does not adequately address the underlying economics (i.e., DEA accommodates economic behavior only by using cost and revenue specifications). Wu et al. (2005) also explain that DEA is sensitive to outliers and statistical noise, so the outcome of the analysis can be warped in the case that the data are contaminated by statistical noise. According to Pasiouras (2006), the DEA method assumes the data to be free of measurement errors and is very sensitive to outliers. Also, GutierrezNieto (2007) argues that there is no statistical framework on which significance tests can be based regarding the selection of inputs and outputs in a DEA model1. There is also a possibility that inputs and outputs might be highly correlated. 1.4. Constant returns to scale model. Here we define the constant returns to scale (CRS) model 1 When employing the DEA method we don’t apply any statistical tests and in the case that we need to solve large problems there might not be feasible results.

which was proposed by Charnes et al. (1978). We assume that we have K inputs and M outputs on each of the N firms or Decision making units (DMU). For the i-th DMU these are represented by the vectors xi and yj respectively. The best method to introduce the DEA CRS model is via the ratio form. For each DMU we obtain a measure of the ratio of all outputs over all inputs, such as u'yi / v' xi, where u is an M x 1 vector of output weights and v is a K x 1 vector of inputs weights. In order to select the optimal weights we define the following mathematical programming problem: max u,v ( u'yi / v' xi), St u'yj / v' xj ≤ 1, j= 1, 2…, N

(1)

u, v ≥ 0.

In the above model we find values for u and v, such that the efficiency measure of the i-th DMU is maximized, subject to the constraint that all the efficiency measures have to be less than or equal to one. Because this ratio formulation has an infinite number of solutions, we impose the constraint v' xi = 1, which provides: maxµ,ν ( µ' yi), st ν'xi =1,

(2)

µ' yj - ν'xj ≤ 0, j = 1, 2,…., N µ ,ν ≥ 0.

The notation has changed from u and v to µ and ν as to reflect the transformation. This is known as the multiplier form of a linear programming model. By using duality in linear programming we can derive an equivalent envelopment form of this problem: min θ,λ θ, st - yi + Yλ ≥ 0,

(3)

θ xi – Xλ ≥ 0, λ ≥ 0.

Where θ is a scalar and λ is an N x 1 vector of constants. Notice that the envelopment form involves fewer constraints than the multiplier form and therefore it is more preferable to solve. The value of θ which we obtained is the efficiency score for the i-th DMU. It will satisfy the θ ≤ 1, with the value 1 indicating a point on the frontier and consequently a technically efficient DMU according to the definition by Farrell (1957). The above linear programming model can be solved N times, once for every DMU and obtain a value of θ for each DMU. 1.5. Variable returns to scale model. The variable returns to scale (VRS) model was suggested by Banker, Charnes and Cooper (1984) as an extension to the constant returns to scale model because the 57

Banks and Bank Systems, Volume 3, Issue 4, 2008

latter model is only appropriate in the case where all the DMUs are operating at an optimal scale. The use of the CRS model when not all DMUs are operating at an optimal scale will result in measures of Technical Efficiency which are confounded by scale efficiencies. With the use of the VRS model we can calculate technical efficiency without taking into consideration the scale efficiency effects. We modify the CRS linear programming problem to account for VRS by adding the convexity constraint: N 1'λ =1 to equation (3) in order to provide: min θ,λ θ st - yi + Yλ ≥ 0, θ xi – Xλ ≥ 0,

(4)

Ν 1' λ = 1, λ ≥ 0,

where N1 is an N x 1 vector of ones. By following this approach we form a convex hull of intersecting planes which envelope the data points more tightly than the CRS conical hull and thus provide technical efficiency scores which are greater than or equal to the ones that we obtained when using the CRS model. 1.6. Super efficiency. Andersen and Petersen (1993) suggest a criterion that permits the ranking of firms which are all found to be 100 % efficient by the DEA method. We consider a single input, single output case and suppose that a firm with inputoutput (xo, yo) has been found to be technically efficient in an output-orientated problem. It is clear that if the firm’s output has been any larger than yo then it would have remained efficient. Furthermore, a small deterioration in the firm’s performance may be allowed without becoming inefficient. In other words, the firm’s observed output exceeds what is necessary for the firm to be considered efficient relative to other firms in the sample and this firm is considered to be super efficient. The Super efficiency model can be formulated after the reformulation of equation (4) and is given by:

min θ,λ θ st - yi + Yλ ≥ 0,

(5)

θ xi – Xλ ≥ 0, λ =0, Ν 1' λ =1. 1.7. Free disposal hull. The free disposal hull model (FDH) was first introduced by Deprins et al. (1984) as an alternative method to the data envelopment analysis model, where only the strong (free) disposability of inputs and outputs is assumed. This model was initially presented as a VRS DEA 58

model excluding the linear combination of observed production plans. There are two methods to solve FDH problems. The first method was introduced by Tulkens (1993) and later by Cherchye et al. (2001) and is based on enumeration algorithms, and the second method is the use of mathematical programming. The computation of the technical efficiency measures using the FDH method requires solving non-linear mixed integer programs. For a set of observed production plans (xk , yk ), k ∈ K, where K is an index set, producing R outputs with I inputs. Then ( xk, yk) ∈ RR+I, ∀ k ∈ K. Let a technology T be defined by T = {( x, y )} : y can be produced by x.

Following Leleu (2006), the technical inefficiency of an observed production plan (xo, yo) is defined by: E (xo, yo) = min (θo : (θo xo, yo) ∈ T. The FDH technology exhibits a strong free disposability assumption on T but does not impose any convexity assumption. The traditional FDH technology is under variable returns to scale and is labelled TFDH – VRS. This FDH technology is represented by its production possibility set: TFDH – VRS = { (x,y): K

∑z k =1

K

k

y k ≥ y,

∑z k =1

K

k

y k ≤ x,

∑z

k

=1,

zk



{0,1},

k =1

k ∈ K}. For more information about the FDH approach, see Leleu (2006). 1.8. Selection of inputs-outputs. In recent years research efforts have been devoted to measuring the efficiency of the banking industry. More specifically the attention has been focused on estimating an efficient frontier and measuring the average difference between banks. In order to estimate the efficiency of the banking industry the inputs and outputs that will be used have to be defined. According to Boulding (1961), the concept of economic activity as an input-output process is perhaps the most basic concept of economics. Nevertheless it is vague, and curious difficulties emerge when an effort is made to specify the inputs and outputs involved and to define the nature of transformation implied.

Frisch (1965) identifies the production process as a transformation method which is controlled by human beings and it is desirable by a number of individuals. With the term transformation it is implied that goods or services (inputs) enter a process where they lose their original form while at the same time other goods or services are generated and these are the outputs. The concept of inputs and outputs is used widely in a number of sectors such as the manufacturing and the agricultural sector. In the banking industry the production process includes the use of deposits and other assets, and the output of banks is measured in terms of quantity.

Banks and Bank Systems, Volume 3, Issue 4, 2008

Fixler and Zieschang (1992) and Beger and Humphrey (1992) argue that there is a confusion in the definition of output measurement, because of the integrated nature of the production process in the banking industry (see also Mlima and Hjalmarsson, 2002). This confusion exists mainly due to the theoretical gap in the banking literature on multi-input multi-output production structure as well as due to the non-tangible nature of the outputs. For this reason there have been made attempts by researchers to overcome this problem by introducing two different approaches which can be used in measuring the banking efficiency; the production approach or the intermediation approach. It is stated though by Wykoff (1992) that problems exist even after the adoption of one of the approaches due to the nonavailability of data on certain physical quantities such as the number of checks cashed or the number of loans issued, etc. 1.9. The production or service provision approach. The production approach is also known as service provision approach or value-added approach and it is used for analyzing the technological efficiency of the banking systems. According to this approach, banks provide their services to customers by keeping customers deposits, issuing loans, cashing checks and by administering various financial transactions that are made by customers (Berg et al., 1991, 1993). The analysis of productivity and efficiency is made by comparing the quantity of services given with the quantity of the resources used. The five activities that are performed by banks are identified by Berg at al. (1991) and these are the following: 1) supplying demand, facilitating deposit services, 2) short and long term loan services, 3) brokerage and other services, 4) property management, and 5) provision of safe deposit boxes. Apart from these five activities Berg et al. (1991) also add that a bank can incur positive cost in terms of labor, machines, buildings and materials. 1.10. The asset or intermediation approach. According to the asset or intermediation approach, banks accept deposits from customers and then transform them into loans to their clients. Mester (1997) argues that the usual inputs which are included in this approach are labor, material and deposits and the outputs used are loans and other income generating activities. This approach is mainly used by researchers in estimating the economic efficiency of the banking sectors. It must be added that the asset approach is sub divided into two groups: 1) the profit approach, and 2) the risk-management approach. The profit approach or the cost approach can be used to estimate the economic efficiency, and according to this approach, the goal of the bank’s manager is to maximize the profit function of the

bank. In the process of production bank managers must take into account all the types of costs that incur as well as the income that is being generated. The profit approach can measure simultaneously the inefficiency occurred in the inputs and the outputs side. On the other hand, the risk-management approach is used to evaluate risks that may be attached to various forms of assets in a bank, where banks take risks in order to produce acceptable returns. In the risk management approach the management decision making process is used as inputs and the shareholders’ value and bank profits are used as outputs. 2. Data description

The data for this study were obtained from the Bankscope database and the official webpage of the Hellenic Bank Association (www.hba.gr). The sample considers only the years 2004 and 2005 due to lack of data (mainly number of ATMs), for the ten largest commercial Greek banks ranked according to their total assets. These banks are the following: the National Bank of Greece, Eurobank, Piraeus bank, Alpha Bank, the Agricultural bank of Greece, Emporiki Bank, Greek Postal Savings Bank, Geniki bank, Egnatia Bank and Attica Bank. We calculate the efficiency scores using three inputs (number of employees, number of branches and number of ATMs) and one output (loans)1. The research objective of this paper is the use of the number of ATMs as an input in the modelling of banking efficiency. Table 1 shows the descriptive statistics for all variables. Table 1. Descriptive statistics Year Name

2004 Minimum

Maximum

59

455

234.4

131.8485

52

1315

491.2

386.0152

1118

12702

4888.5

3497.1503

1726.8

26052.7

11509.41

8863.8953

Number of branches ATMs Number of employees Loans Year Name Number of branches ATMs Number of employees Loans

Mean

Std. dev.

2005 Minimum

Maximum

Mean

Std. dev.

59

567

275.1

164.7674

61

1352

525.7

389.2768

1118

13175

4848.3

3662.4819

1787.6

29528.2

13691.56

10582.8563

Source: Bankscope and Hellenic Bank Association. 1 Loans are money lent by banks to borrowers, and borrowers agree to return the borrowed amount of money along with an interest rate.

59

Banks and Bank Systems, Volume 3, Issue 4, 2008

In particular, National Bank of Greece has higher efficiency scores for the year 2004 and for the super efficiency input and output models. We report a significant decrease in the efficiency in 2005 for all the methods that are employed. Eurobank follows the pattern that National Bank of Greece follows. There is a significant decrease in the super efficiency both for the input and output orientation but there is no change observed for the other methods. Alpha bank, on the other hand, exhibits an increase in the super efficiency scores for both input and output orientation for 2005 in addition to an increase in the FDH efficiency scores. Piraeus bank shows a decrease in 2005 in all methods but it should be mentioned that the FDH method generates the higher efficiency scores in comparison to the other methods. Emporiki Bank exhibits a small decrease in 2005 in all the methods except for the FDH method where a small increase since the year 2004 is observed. Agricultural bank of Greece exhibits a small decrease in all the efficiency scores obtained from all the types of method for 2005. Greek Postal Savings Bank’s efficiency scores are decreased in 2005 with the major decrease being observed for super efficiency input and output oriented methods. Egnatia Bank is also showing a small decrease in the efficiency for 2005 with the larger difference for super efficiency scores both for input and output orientated methods. On the other hand, Geniki bank is showing a significant increase in its efficiency in 2005 for all the methods employed but the larger increase is observed for super efficiency input and

Table 2. Summary of empirical results (efficiency scores) PART A. Year 2004

All banks

The results show that there is an obvious decrease in the efficiency scores in 2005 compared to 2004. More specifically, the greatest decrease is observed for the super efficiency method for both the input and the output orientation. The highest scores are observed for the all the methods employing the VRS.

output oriented methods. Lastly, Attica bank is also showing a decrease in the efficiency in 2005 for all methods with the larger increase being observed for the super efficiency input and output models. Table 2 shows a summary of the results by year (Part A shows the results for 2004, while part B shows the results for year 2005) and methods. In addition, Table 3 presents the results for each bank by year and methods.

Basic input irs

Basic input drs

89.02% Basic output irs

83.13% Basic output drs

86.55%

83.13%

Fdh vrs

Fdh irs

Fdh drs

100.00%

92.29%

84.35%

Basic input crs

Basic input vrs

Basic input irs

Basic input drs

74.71% Basic output crs

91.47% Basic output vrs

87.96% Basic output irs

78.21% Basic output drs

74.71% Super eff. input crs

89.38% Super eff. input irs

85.87%

78.21%

75.77% Super eff output crs

75.77% Super eff output drs

75.77%

80.07%

Fdh crs

Fdh vrs

Fdh irs

Fdh drs

74.71%

98.50%

93.02%

78.21%

Basic input crs

Basic input vrs

83.06%

89.10%

Basic output crs

Basic output vrs

83.06% Super eff. input crs

86.62% Super eff. input irs

93.18% Super eff output crs

93.18% Super eff output drs

93.18%

93.63%

Fdh crs 84.29% PART B. Year 2005

All banks

3. Empirical results

Table 3. Empirical results: bank vs. method (efficiency scores) Nbg Input oriented

2004 100.00%

2005 82.47%

agric Input oriented

2004 66.65%

2005 58.11%

Output oriented

100.00%

82.47%

Output oriented

64.81%

56.18%

Super effic input orient

163.25%

64.94%

Super effic input orient

64.34%

55.70%

Super effic. output orient

163.25%

86.44%

Super effic output orient

64.34%

55.70%

Fdh

100.00%

82.47%

Fdh

75.99%

70.35%

2004

2005

100.00%

100.00%

Input oriented

Efg Input oriented

gpsb

2004 100.00%

2005 82.19%

Output oriented

100.00%

100.00%

Output oriented

100.00%

82.19%

Super effic input orient

108.36%

106.41%

Super effic input orient

129.57%

64.37%

Super effic output orient

108.36%

106.41%

Super effic output orient

129.57%

64.37%

Fdh

100.00%

100.00%

Fdh

100.00%

82.19%

60

Banks and Bank Systems, Volume 3, Issue 4, 2008

Table 3 (cont.). Empirical results: bank vs. method (efficiency scores) Alpha Input oriented

2004

2005

99.62%

100.00%

Input oriented

egna

2004 87.37%

2005 85.18%

Output oriented

99.62%

100.00%

Output oriented

86.49%

81.96%

Super effic input orient

99.23%

104.26%

Super effic input orient

82.09%

75.20%

Super effic output orient

101.51%

104.26%

Super effic output orient

82.09%

75.20%

99.72%

100.00%

Fdh

93.50%

87.60%

2004

2005

95.48%

93.43%

Input oriented

Fdh Pir Input oriented

geb

2004 51.96%

2005 75.21%

Output oriented

95.39%

93.16%

Output oriented

44.07%

71.81%

Super effic input orient

92.00%

90.85%

Super effic input orient

41.41%

53.51%

Super effic output orient

92.00%

90.85%

Super effic output orient

41.41%

53.51%

Fdh

96.00%

95.43%

Fdh

64.76%

76.76%

Empo Input oriented

2004

2005

attica

64.24%

Input oriented

64.53%

62.59%

Output oriented

Super effic input orient

64.48%

62.36%

Super effic output orient

64.48%

62.36%

Fdh

75.72%

76.25%

Output oriented

66.20%

2004 93.51%

2005 90.07%

93.51%

90.07%

Super effic input orient

87.02%

80.13%

Super effic output orient

87.02%

80.13%

Fdh

96.67%

90.07%

Notes: nbg – National Bank of Greece, efg – Eurobank, Pir – Piraeus bank, Alpha – Alpha Bank, agric – Agricultural Bank of Greece, Empo – Emporiki Bank, gpsb – Greek Postal Savings Bank, geb – Geniki bank, egna – Egnatia Bank, and attica – Attica Bank.

Summary and conclusion

This paper examines the banking efficiency of top 10 largest commercial Greek banks. We employ data from 2004 and 2005, while we use both the DEA and FDH models with Constant, Variable, Increasing and Decreasing Returns to scale. The objective of this paper is twofold: (i) to describe and apply the most recent methods of efficiency in Greek banking, and (ii) to calculate the efficiency scores using three inputs (number of employees, number of branches and number of ATMs) and one output (loans). To the best of our knowledge, only a limited number of studies include the number of ATMs as an input when calculating the efficiency scores. The overall efficiency scores range between 71% (for 2004) and 73.6% (for 2005). The empirical results show that the average level of overall technical efficiency is 72%, suggesting that Greek banks could have increased their outputs by 28% with the existing level of inputs. The variation of efficiency scores is plotted in Figure 1 (Appendix). The high overall technical efficiency scores are in line with studies by Pasiouras (2006), Tsionas (2001) and Spathis (2001). Similarly the low overall efficiency scores are in line with Halkos and Salamouris (2004); they report that the average efficiency score of the Greek banking system is around 60%. As far as the various methods employed are concerned, only the FDH with VRS for case 1 and the year 2004 shows an efficiency score equal to unity. Casu and Molyneux (2003), note that when an efficiency score of unity is achieved then this

combination of inputs/outputs is the ‘best practice’ units and therefore the efficient frontier is generated. The results for the average efficiency scores show that National Bank of Greece, Eurobank, Alpha and the Greek Postal Savings Bank exhibit the higher efficiency score with a percentage almost 100 % for every case, and this indicates that these banks’ relative efficiencies are located on the efficient frontier. Those banks which have efficiency values from 0.6 to 0.9, representing fairly performance, include Piraeus Bank, Agricultural Bank and Attica Bank. Other banks (Geniki Bank) are ranking below 0.6, representing relatively poor efficiency. The competitiveness of Geniki Bank is clearly lagging behind the other 9 Greek banks. Figure 2 (Appendix) shows the range of efficiency scores per method employed for all banks in 2004 and 2005. From the DEA analysis, we find that large banks (NBG, Eurobank, Alpha Bank) are more efficient than medium and small sized ones. However, it appears that small banks without offering e-banking services exhibit very high efficiency scores (Greek Postal Savings Bank), while the results of Geniki Bank show that Geniki Bank is less efficient in terms of technical efficiency. We also find that banks with a large number of ATMs (National Bank of Greece) are more efficient than those with less ATM, and this is in line with the study of Pasiouras et al. (2007). They suggest that the banks with broader ATM networks are more technical and cost efficient. However, they notice that the influence of ATMs on the efficiency of banks disappears when a 61

Banks and Bank Systems, Volume 3, Issue 4, 2008

control on market conditions was imposed. Further, we conclude that the provision of e-banking services by banks does not influence their efficiency scores.

Future research should examine the e-banking efficiency of Greek and other European banks using recent data and methods.

References

1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22.

Avrikan, N. (2005). Developing foreign bank efficiency models for DEA grounded in finance theory. SocioEconomic planning sciences 40 (2006), pp. 75-296. Berg,S., Forsund, F. and Jansen, E. (1991). Bank output measurement and the construction of best practice frontiers. Journal of productivity analysis, 2, pp. 127-142. Berger, A. and Humphrey, D. (1993). Measurement and efficiency issues in commercial banking. University if Chicago Press, pp. 245-279. Boulding, K. (1961). Some difficulties in the concept of economic input. National Bureau of Economic Research, 25, pp. 331-334. Casu, B. & Molyneux, P. (2003). A comparative study of efficiency in European Banking. Applied Economics, 35 (17), pp. 1865-1876. Charnes, A., Cooper, W. and Rhodes, E. (1978). Measuring the efficiency of decision making units.European Journal of Operations Research, 2, pp. 429-444. Coelli T.J. (1996). A Guide to DEAP version 2.1: A Data envelopment analysis (Computer) Program. Cepa Working Papers, 8 (96), pp. 1-50. Damar, H. (2006). The effects of shared ATM networks on the efficiency of Turkish banks. Applied Economics, 38, pp. 683-697. Farc, R., Grosskopf, S., Kirkley, J. and Squires, D. (2000). Data envelopment analysis (DEA): A framework for assesing capacity in fisheries when data are limited. International Institute of Fisheries Economics and Trade. Farrell, M. (1957). The measurement of productive efficiency. Journal of Royal statistical society, 3, pp. 253-290. Frisch, R. (1965). Theory of production. Dordrecht: D. Reidel. Gutierrez-Nieto B., Serrano-Cinca C. And Molinero (2007). Microfinance institutions and efficiency. The international journal of management science, 35, pp. 131-142. Halkos, G. & Salamouris D. (2004). Efficiency measurement of the Greek commercial banks with the use of financial ratios: a data envelopment analysis approach. Management Accounting Research, 15, pp. 201-224. Mester, L. (1996). A study of bank efficiency taking into account risk preferences. Journal of banking and finance, 20, pp. 1025-1045. Mlima, A. and Hjalmarsson, L. (2002). Measurement of Inputs and Outputs in the Banking industry. Tanzaret Journal, 3 (1), pp. 12-22. Molyneux, P., Altunbas, Y.& Gardener, E. (1996). Efficiency In European Banking. Chichester: Wiley. Pasiouras, F. (2006). Estimating the technical and scale efficiency of Greek commercial banks: the impact of credit risk, off-balance sheet activities, and international operations. University of Bath. School of Management. 2007, 16. Spathis, Ch., Kosmidou, K. and Doumpos, M. (2002). Assessing profitability factors in the Greek banking system: A multicriteria methodology. International Transactions in operational research, 9, pp. 517-530. Sufian, F. (2006). Trends in the efficiency of Singapore’s commercial banking groups: A non stochastic frontier DEA window analysis approach. International Journal of Productivity and Performance Management, 56, (2), pp. 99-136. Tsionas, E., Lolos, S.& Christopoulos D. (2001). The performance of the Greek banking system in view of the EMU: results from a non-parametric approach. Economic Modelling, 20, pp. 571-592. Wu, D., Yang Z. and Liang L. (2006). Using DEA-neural network approach to evaluate branch effieciency of a large Canadian bank.Expert Systems with applications, 31, pp. 108-115. Wykoff, F. (1992). Comments on measurements and efficiency in banking. University of Chicago Press, pp. 279-287.

Appendix Basic input m ethod for all banks 2004-2005 100,00%

Basic output m ethod for all banks 2004-2005 95,00%

60,00%

2004

40,00%

2005

20,00%

90,00% Efficiency

Efficiency

80,00%

85,00%

2004

80,00%

2005

75,00% 70,00%

0,00% Basic input crs basic input vrs basic input irs basic input drs

65,00% basic output crs basic output vrs basic output irs basic output drs

Methods

62

Methods

Banks and Bank Systems, Volume 3, Issue 4, 2008 Super efficiency output m ethod for all banks 2004-2005

Super efficiency input method for all banks 2004-2005 100,00%

100.00%

60.00%

2004 2005

40.00%

80,00%

Efficiency

Efficiency

80.00%

60,00%

2004

40,00%

2005

20,00%

20.00%

0,00% super eff output crs

0.00% super eff input crs

super eff input irs

super eff output drs Method

Method

FDH m ethod for all banks 2004-2005 120.00%

Efficiency

100.00% 80.00% 2004

60.00%

2005

40.00% 20.00% 0.00% fdh crs

fdh vrs

fdh irs

f dh drs

M et ho d

Fig. 1. Variation of efficiency scores (method vs. year) EFG

110,00% 108,00%

input oriented

150,00%

output oriented

100,00% 50,00% 0,00% 2004

106,00%

input oriented

104,00%

output oriented

102,00%

super effic input orient

100,00%

super effic output orient

super effic input orient

98,00%

super effic output orient

96,00%

fdh

94,00%

fdh

2004

2005

2005 Year

Year

Alpha

Efficiency

106,00% input oriented

104,00%

output oriented

102,00%

super effic input orient 100,00%

super effic output orient

98,00%

fdh

96,00% 2004

2005 Year

Agricultural

Efficiency

80,00% 70,00% 60,00%

input oriented

50,00%

output oriented

40,00%

super effic input orient

30,00%

super effic output orient

20,00%

fdh

10,00% 0,00% 2004

2005

Year Emporiki 90,00% 80,00%

Efficiency

Efficiency

200,00%

E fficien cy

NBG

70,00%

input oriented

60,00%

output oriented

50,00%

super effic input orient

40,00% 30,00%

super effic output orient

20,00%

fdh

10,00% 0,00% 2004

2005 year

63

Banks and Bank Systems, Volume 3, Issue 4, 2008 Greek Postal Savings 140.00%

Efficiency

120.00% input orient ed

100.00%

out put orient ed

80.00%

super ef f ic input orient

60.00%

super ef f ic out put orient 40.00%

f dh

20.00% 0.00% 2004

2005 year

Efficiency

Egnatia Bank 100.00% 90.00% 80.00% 70.00% 60.00% 50.00% 40.00% 30.00% 20.00% 10.00% 0.00%

input orient ed out put orient ed super ef fic input orient super ef fic out put orient fdh

2004

2005 year

Geniki 100,00% input oriented

Efficiency

80,00%

output oriented

60,00%

super effic input orient

40,00%

super effic output orient

20,00%

fdh

0,00% 2004

2005 Year

Piraeus 98,00% input oriented

Efficiency

96,00%

output oriented

94,00%

super effic input orient

92,00%

super effic output orient

90,00%

fdh

88,00% 2004

2005 Year

Fig. 2. Range of efficiency scores (method vs. bank/year)

64

Banks and Bank Systems, Volume 3, Issue 4, 2008

Izah Mohd Tahir (Malaysia), Nor Mazlina Abu Bakar (Malaysia), Sudin Haron (Malaysia)

Technical efficiency of the Malaysian commercial banks: a stochastic frontier approach Abstract

The purpose of this study is to examine the technical efficiency of the Malaysian commercial banks over the period of 2000-2006, using the stochastic frontier approach (SFA). The findings show that Malaysian commercial banks have exhibited an average overall efficiency of 81 percent implying an input waste of 19 percent. The results also found that the level of efficiency has increased during the period of study. Finally, domestic banks are found to be more efficient relative to foreign banks. Keywords: technical efficiency, commercial banks, SFA, domestic banks, foreign banks. JEL Classification: C21, H21, E59.

Introduction•

The structure of the Malaysian financial institutions has changed dramatically for the last twenty years. In addition, global trend towards liberalization in banking has led to the blurring of demarcation lines separating activities of the different groups of financial institutions and the removal of artificial barrier to competition. Similarly, deposit taking, credit granting, investment, insurance and financial advisory services are being bundled into one financial conglomerate of financial supermarkets. The integration of financial markets within and across borders as well as mergers among banks, reflect attempts to increase the financial industry efficiency. The Malaysian experience on the merger exercise is a good example. From 58 financial institutions, the number has to reduce to 10 anchor banks and this is to be completed by 31 December 2000. This was the result of the financial crisis which has weakened the domestic banking sector and the move towards consolidation is hoped to improve the efficiency of the banking sector. The commercial banks have undergone a tremendous development with the merger exercise. Theoretically, bank merger could broaden the product mix and reduce cost, and definitely large size capital and asset are crucial for a bank to become efficient, competitive and powerful one. These elements with good quality service will enable banks to compete with foreign institutions at local as well as at international levels. The objective of this paper is to investigate the level of technical efficiency among commercial banks in Malaysia for the period of 2000-2006. The paper is structured as follows: the next section will discuss efficiency measurement in banking followed by model development and data. Empirical findings are discussed in section 3 followed by conclusion. 1. Efficiency measurement in banking

Generally, efficiency means the maximum output that can be produced from any given total of inputs. •

© Izah Mohd Tahir, Nor Mazlina Abu Bakar, Sudin Haron, 2008.

This refers to the efficiency of a firm which allocates resources in such a way as to produce the maximum quantity of output. In the context of resource allocation efficiency, Shepherd (1997) pinpoints two categories: internal efficiency and allocative efficiency. Internal efficiency refers to effective management within the firm itself; for example, the ways in which management inspires the staff, controls costs and keeps operations lean. However, when a company is increased in size, profit flows are expected to increase too. Hence, management tends to become less effective. Such shortcomings in management are known as X-inefficiencies and can be attributed to the excess of actual costs over the minimum possible costs. In other words, Xinefficiency may be measured as excess costs divided by actual costs. Early research in banking industry was mainly concerned with estimating the average productivity, using some sort of indices and with cost comparison (Farrell, 1957). Subsequently, researchers tended to proxy efficiency by market share. They assumed that banks with large market shares are expected to earn higher profits because of lower unit costs (Smirlock, 1985; Evanoff and Fortier, 1988). In other words, banks with lower cost structures could maximize their profits either by maintaining the current level of prices and size or reducing the price levels and expanding a positive relationship between firms’ profits and market structures being attributed to the gains made by more efficient firms. The financial indicators of bank’s operating performance, such as operating costs divided by total assets or the return on equity or assets, have also been used to compare efficiencies; for example, Rhoades (1986), Cornett and Tehranian (1992) and Srinivisan and Wall (1992) studied the effect of mergers among banks on efficiency. However, the use of financial ratios has its limitations. According to Berger et al. (1993), the first problem is that financial ratios are regarded as misleading indicators of efficiency because they do not control for product 65

Banks and Bank Systems, Volume 3, Issue 4, 2008

mix or input prices. Secondly, using the cost-toasset ratio assumes that all assets are equally costly to produce and all locations have equal costs of doing business. Finally, the use of simple ratios cannot distinguish between X-efficiency gains and scale and scope efficiency gains. Recent approaches to measure bank efficiency include the parametric and non-parametric approaches. These efficiency measurements differ primarily in how much shape is imposed on the frontier and the distributional assumptions imposed on the random error and inefficiency (Berger and Humphrey, 1997). In the research literature, both parametric and non-parametric approaches have been widely used but there is no consensus which of these major approaches is superior (Berger and Humphrey, 1997). There are three main parametric approaches used to estimate relative efficiency: the stochastic frontier approach, the thick frontier approach, the thick frontier and distribution-free estimates approach. The stochastic frontier approach (SFA) sometimes also referred to as the econometric frontier approach (EFA), was developed by Aigner, Lovell and Schmidt (1977), and Meeusen and Van den Broeck (1977). In this approach, the SFA specifies a functional form for the cost, profit or the production frontier and allows for random error. The SFA modifies a standard cost (production) function to allow inefficiencies to be included in the error term. The predicted standard cost function is assumed to characterize the frontier while any inefficiency is captured in the error term, which is construction orthogonal to the predicted frontier. This assumption forces any measured inefficiencies be uncorrelated with the regressors and any scale or product mix economies derived linearly from these explanatory variables (Ferrier and Lovell, 1990). Another assumption needed in the SFA is to distinguish the inefficiencies from random components of the error terms. The random components include short-term luck which places individual banks in relatively high or low cost positions and measurement error from excluded explanatory variables, misspecification etc. These two components are separated by assuming that inefficiencies are drawn from asymmetric half-normal distribution, and that random errors are drawn from a symmetric normal distribution. However, it is not possible to decompose individuals’ residuals into inefficiency or random variation; therefore, estimating technical inefficiency by observation is impossible. Okuda et al. (2003) use SFA to estimate the cost function of the Malaysian commercial banks from 1991-1997 and its impact on bank restructuring. The study observed economies of scale but not economies of scope and 66

suggested that Malaysian domestic banks were making unproductive capital investments. The thick frontier approach (TFA) has been applied to banking by Berger and Humphrey (1991, 1992). This approach, instead of estimating a frontier edge, compares the average efficiencies of groups of banks. A cost function for the lowest average cost quartile of banks is estimated and banks in this quartile are assumed to have greater than average efficiency and form a ‘thick frontier’. Similarly, a cost function is also estimated for the highest average cost quartile and banks in this quartile presumably have less than average efficiency. Differences in error terms within the highest and lowest performance quartile of observations (stratified by size class) are assumed to represent random error, while the predicted cost differences between the highest and lowest quartile are assumed to reflect inefficiencies. This inefficiency residual is then decomposed into several types of inefficiencies. The TFA thus imposes no distributional assumptions on either inefficiency or random error except to assume that inefficiencies differ between the highest and lowest cost quartile and that random error exists within these quartiles. In the distribution-free approach (DFA), a functional form for the frontier is also specified but inefficiencies are separated from random error in a different way. Unlike the SFA, the DFA makes no strong assumptions regarding the specific distributions of the inefficiencies or the random errors. The identifying assumption is that efficiency of each bank is stable over time, while random errors tend to average out over time. The estimate of inefficiency for each bank in a panel data set is then determined as the difference between its average residual and the average of the bank on the frontier with some truncated measure performed to account for the failure of the random error to fully average out. The truncation procedure is similar to the TFA treatment of outliers1. Therefore, the truncation procedure is used to remove some of the effects of the extreme observations by treating all the most efficient firms alike and, similarly, all the most inefficient firms alike2. Berger (1993) has applied the DFA to banking in the study of the US banking industry. He finds that the frequency distribution of inefficiencies appears to be closer to the shape of symmetric normal distribution than an asymmetric half-normal distribution. 1  In the TFA approach, data are averages within the very highest and lowest average cost quartile. 2 Lang and Welzel (1996) used a fixed effects model where a dummy variable is specified for each bank in a panel data set. Differences in the fixed effects estimated across banks represent bank inefficiencies. Berger (1993) finds that the fixed effects approaches (under Method 2) were confounded by large differences in scale.

Banks and Bank Systems, Volume 3, Issue 4, 2008

Yildrim and Philippatos (2007) use both SFA and DFA to examine the cost and profit efficiency of banking sectors in twelve countries in Europe and find that the average cost efficiency level was 72 percent by DFA and 77 percent by SFA. Unlike the parametric approach, the non-parametric approach assumes that random error is zero so that all unexplained variations are treated as reflecting inefficiencies. Non-parametric approaches such as Data Envelopment Analysis and Free Disposal Hull, put relatively little structure on the specification of the best-practice frontier. Data Envelopment Analysis (DEA) is rooted in the work of Farell (1957), who used the economic concept of the production frontier and the production possibility set to define technical and allocative efficiencies and later proposed measures of relative inefficiencies. DEA was first introduced by Charnes, Cooper and Rhoades (1978) to describe an application of mathematical programming to observe data to locate frontier which can then be used to evaluate the efficiency of each of the organizations responsible for the observed output and input quantities. The concept of DEA is similar to that of technical efficiency in the microeconomic theory of production. However, the main difference is that the DEA production frontier is not determined by some specific equation; instead it is generated from the actual data for the evaluated firms (DMUs). Therefore, the DEA efficiency score for a specific firm is defined not by an absolute standard but relative to the other firms under consideration. DEA also assumes that all firms face the same unspecified technology, which defines their production possibility set. The main objective of DEA is to determine which firms are operating on their efficient frontier and which firms are not. If the firm’s input-output combination lies on the DEA frontier, the firm is considered efficient; and the firm is considered inefficient if the firm’s inputoutput combination lies inside the frontier. The basic DEA model (CCR model) implied the assumption of constant returns to scale. This assumption was later relaxed to allow for the evaluation of variable returns to scale and scale economies. Specifically, the efficient frontier may be derived using four alternative returns to scale assumptions; constant returns to scale (CR); variable returns to scale (VR), non-increasing returns to scale (NI); and non-decreasing returns to scale (ND). Yue (1992) defines the following assumptions. A bank exhibits increasing returns to scale if a proportionate increase in inputs and outputs places it inside the production frontier; and constant returns to scale if a proportionate increase or decrease in inputs or outputs move the firm either along or above the frontier. A bank which is not on the frontier is defined as

experiencing non-increasing returns to scale if the hypothetical bank with which it is compared, exhibits either constant or decreasing returns to scale. A similar definition applies for non-decreasing returns to scale. A firm which is efficient under the assumption of variables return to scale (VRS) is considered technologically efficient; the VRS score represents pure technical efficiency (PT), whereas a firm which is efficient under the assumption of constant returns to scale (CRS) is technologically efficient and also uses the most efficient scale of operation. There are a number of studies examining relative efficiency using DEA (Sufian and Abdul Majid, 2007; Li, 2006; Sufian, 2006; Sufian, 2004; Katib and Mathews, 2000). Sufian and Abdul Majid (2007) analyze efficiency change of Singapore commercial banks during the period of 1993-2003. They find that commercial banks in Singapore exhibited an average overall efficiency of 95.4 percent. Li (2006) investigates the scale-efficiency and technology-efficiency of 14 Chinese commercial banks. She concludes that most banks have low comparative efficiency. She also finds that inefficient banks generally have input surplus. Sufian (2006) investigates the efficiency of non-bank financial institutions in Malaysia for the period of 2000-2004. The study finds that finance companies were more efficient than merchant banks and that the inefficiency was the result of pure technical inefficiency rather than scale inefficiency. Using DEA to examine the efficiency effects of bank mergers and acquisition in Malaysia, Sufian (2004) finds that Malaysian banks have exhibited a commendable overall efficiency level of 95.9 percent during 1998-2003 which indicates that merger program was successful. Katib and Mathews (2000) also use DEA to estimate the efficiency of 20 Malaysian commercial banks from 1989 to 1995. The results suggest that whilst efficiency ranges between 68 percent and 80 percent, the trend in efficiency is downwards. Free Disposal Hull is a special case of the DEA model where the points on lines connecting the DEA vertices are not included in the frontier. Instead, the FDH production possibility set is composed of only the DEA vertices and the Free Disposal Hull point interior to these vertices. Because the FDH frontier is either congruent with or interior to the DEA frontier, FDH will typically generate larger estimates of average efficiency than DEA. The FDH approach therefore allows for a better approximation or ‘envelopment’ of the observed data. 2. Model specification and data

This study will use the intermediation approach. Under the intermediation approach, banks are treated as financial intermediaries that combine 67

Banks and Bank Systems, Volume 3, Issue 4, 2008

deposits, labor and capital to produce loans and investments. The values of loans and investments are treated as output measures; labor, deposits and capital are inputs; and operating costs and financial expenses comprise total cost. Technical efficiency (TE) has two types of measure: output-oriented and input-oriented measures. If it is an output-oriented measure, TE is a bank’s ability to achieve maximum output given its sets of inputs. Whilst, an input-oriented TE measure reflects the degree to which a bank could minimize its inputs used in the production of given outputs. Our study adopts an output-oriented measure. A value of 1 indicates full efficiency and operations on the production frontier. A value of less than 1 reflects operations below the frontier. The wedge between 1 and the value observed measures the technical efficiency. The technical efficiency of the bank can be written in a natural logarithm form as follows: InQ = f ( x) + InU t − InVt ,         (1) where InQ is the observed outputs in natural log, f denotes some functional form, x is the vector of inputs, Ut is the inefficiency error term, Vt is random error term which accounts for measurement of error on the value of output. To put it simply, the production function describes the relationship between the output variables with quantities of input variables plus the inefficiency and random error. InQ = α 0 +

n

∑ α Inx 1

i

+ E t ,  

 

 

  (2)

i =1

Where InQ is the natural log of output variable for production function, Inxi is the vector of quantities of variable inputs in natural log, Et is the stochastic error term where Et = Ut – Vt .

Following Aigner, Lovell and Schmidt (1977), this study assumes the distribution of the error term or statistical noise, Vi , to be two-sided normal distribution while the inefficiency term, U i , is assumed to be one sided (half normal distributed). The full model thus lnQit = β0 +α1 lnX1 +α2 lnX2 +α3 lnX1 lnX1 +

(4)

+α4 lnX2 lnX2 +α5 lnX1 lnX2 +υit +νit,

where Qit is outputs: total earning assets (financing, dealing securities, investment securities and placements with other banks), Xi s are inputs: total deposits (deposits from customers and deposits from other banks) and total overhead expenses (personnel expenses and other operating expenses). Our sample is an unbalanced panel of 22 commercial banks (9 domestic banks and 13 foreign banks) during the period from 2000 to 2006, totalling 147 observations. The basic data source is BANKSCOPE – Fitch’s International Bank Database. The computer program, FRONTIER Version 4.1 developed by Coelli (1996), has been used to obtain the maximum likelihood estimates of parameters in estimating the technical efficiency. The program can accommodate cross sectional and panel data; cost and production function; half-normal and truncated normal distributions; time-varying and invariant efficiency; and functional forms which have a dependent variable in logged or original units. These features of what Frontier 4.1 can and cannot do are not exhaustive, but provide an indication of program’s capabilities. Table 1 presents the descriptive statistics of banks’ inputs and outputs used in this study.

(3) Table 1. Commercial bank’s input and output variables 2000-2006 (in RM million) Variable

N

Mean

Median

Minimum

Maximum

Std. dev.

147

28300.14

19669.00

508.90

189518.10

34256.54

147

24477.63

17172.50

190.10

164392.60

29819.88

147

1073.91

825.20

6.60

2784.00

1212.98

59

53196.17

38644.60

8826.00

189518.10

40747.25

59

46037.12

33733.30

6955.90

164392.60

35478.75

59

761.70

571.90

124.20

2784.00

572.60

88

11608.48

3124.30

508.90

39324.00

12660.97

88

10022.98

2614.20

190.10

35417.30

11249.28

88

191.09

63.25

6.60

875.10

231.24

All Q X1 X2 Domestic banks Q X1 X2 Foreign banks Q X1 X2

Notes: Q = Total earning assets, X1 = Total deposits, X2 = Total overhead expenses. 

68

Banks and Bank Systems, Volume 3, Issue 4, 2008

Source: Authors’ estimation.

3. Empirical findings

A firm is regarded as technically efficient if it is able to obtain maximum outputs from given inputs or minimize inputs used in producing given outputs. Therefore firms on the production frontier are labelled as ‘best practice’ and they demonstrate optimum efficiency in the utilization of their resources. A value of 1.0 indicates that a firm lies on the bestpractice frontier or full efficiency. A value of less than 1.0 indicates operations below the frontier or inefficient utilization of resources. In Table 2, the average technical efficiency score of Malaysian banks for the 147 observations over the years 20002006 ranges between 77 percent to 84 percent and increases over the years. Katib and Mathews (2000) find the score ranges between 68 percent and 80 percent but on a decreasing trend whilst Sufian

(2004) finds Malaysian banks exhibited 95.9 percent. As an overall, the efficiency score is 81 percent. In other words, the sample banks have wasted on average 19 percent of their inputs. Looking at the efficiency scores in Table 3, both domestic banks and foreign banks average efficiency is on increasing trend. The scores for domestic banks on average ranged between 88.8 percent and 92.8 percent whilst that of foreign banks ranged between 69.7 percent and 78.2 percent. The overall efficiency level for domestic banks was higher (90.9 percent) compared to that of foreign banks (74.4 percent) suggesting that domestic banks are on average more efficient than foreign banks. The results also suggest that there is significant mean difference between technical efficiency of domestic and foreign banks.

Table 2. Technical efficiency: summary of SFA results 2000

2001

2002

2003

2004

2005

2006

All

Mean

0.771

0.790

0.791

0.811

0.822

0.832

0.842

0.810

Median

0.840

0.849

0.858

0.867

0.876

0.884

0.891

0.871

Maximum

0.942

0.946

0.949

0.99

0.992

0.992

0.993

0.993

Minimum

0.432

0.457

0.481

0.506

0.529

0.552

0.575

0.432

S. D.

0.159

0.145

0.143

0.139

0.132

0.126

0.119

0.137

Skewness

-0.811

-0.974

-0.801

-0.801

-0.813

-0.825

-0.836

-0.844

18

20

21

22

22

22

22

147

N

Notes: N = Number of banks, S.D. denotes standard deviation. Source: Authors’ own estimates.  

This is perhaps the results of the merger waves of the 1990’s that has completed its exercise in 2000, leaving domestic banks to only 9 banks. In addition, foreign banks have been prohibited to open new

branches since 1971 while that of domestic banks were given competitive advantage and support from the government.

Table 3. Technical efficiency scores by ownership, 2000-2006 Domestic banks Mean Median Maximum Minimum Std. dev. Skewness N Foreign banks Mean Median Maximum Minimum S. D. Skewness N

2000

2001

2002

2003

2004

2005

2006

All

0.888 0.893 0.942 0.840 0.038 -0.126 7

0.889 0.894 0.946 0.845 0.038 0.126 8

0.896 0.900 0.950 0.854 0.036 0.123 8

0.913 0.912 0.992 0.863 0.043 0.567 9

0.918 0.917 0.993 0.872 0.041 0.562 9

0.923 0.923 0.993 0.880 0.038 0.557 9

0.928 0.928 0.994 0.888 0.036 0.552 9

0.909 0.909 0.994 0.840 0.040 0.267 59

0.697 0.646 0.901 0.432 0.164 0.005 11

0.724 0.722 0.907 0.457 0.154 -0.023 12

0.727 0.683 0.913 0.482 0.148 -0.024 13

0.742 0.701 0.919 0.506 0.141 -0.038 13

0.756 0.718 0.924 0.530 0.135 -0.051 13

0.769 0.734 0.929 0.553 0.128 -0.064 13

0.782 0.749 0.934 0.575 0.122 -0.075 13

0.744 0.713 0.934 0.432 0.139 -0.161 88

Notes: N = Number of banks, S. D. denotes standard deviation. Source: Authors’ own estimates.

69

Technical efficiency

Banks and Bank Systems, Volume 3, Issue 4, 2008

1.00 0.95 0.90 0.85 0.80 0.75 0.70 0.65 0.60

Domestic banks Foreign banks

2000 2001 2002 2003 2004 2005 2006 Year

Note: Figure 1 depicts the graphical presentation of the mean efficiency scores for both domestic and foreign banks. Fig. 1. Technical effciency for domestic and foreign banks

Analyzing the technical efficiency at bank level throughout 2000 to 2006, the impact of merger upon domestic banks can be seen from the increase of efficiency scores of domestic banks (Table 4). By comparing the performance of domestic banks based on the mean efficiency scores, the results show that AmBank Malaysia Berhad, the leading service provider in Malaysia, appears to perform well with 99.3

percent, outperforming the largest bank, Malayan Banking Berhad, which is ranked third (93.4 percent). RHB bank is ranked second (95.3 percent). The disparity between the highest (99.3 percent) and the lowest (86.7 percent) was quite small. Both Hong Leong Bank and Public bank have improved their efficiency levels significantly during the period. Looking at the performance of foreign banks, banks which experience high levels of technical efficiency include United Overseas Bank (91.8 percent), OCBC bank (91.7 percent) and Citibank (90.3 percent). However, the disparity between the highest (91.8 percent) and the lowest (50.5 percent) was quite large. The Royal Bank of Scotland and JP Morgan Chase Bank have both improved their efficiency levels significantly throughout the period observed.

Table 4. Mean technical efficiency of individual banks, 2000-2006 Banks

2000

2001

2002

2003

2004

2005

2006

Average

0.841

0.850

0.860

0.868

0.877

0.884

0.892

0.867

Alliance Bank Malaysia Bhd

NA

0.845

0.854

0.863

0.872

0.880

0.888

0.867

AmBank Malaysia Berhad

NA

NA

NA

0.992

0.993

0.993

0.994

0.993

CIMB Bank Berhad

0.903

0.909

0.915

0.921

0.926

0.930

0.935

0.920

EON Bank

0.840

0.850

0.860

0.868

0.877

0.884

0.892

0.867

Hong Leong Bank Berhad

0.881

0.888

0.895

0.902

0.908

0.914

0.920

0.901

Domestic banks Affin Bank Berhad

Malayan Banking Berhad

0.920

0.925

0.930

0.934

0.938

0.942

0.946

0.934

RHB Bank

0.942

0.946

0.950

0.953

0.956

0.959

0.962

0.953

Public Bank

0.893

0.899

0.906

0.912

0.917

0.923

0.928

0.911

The Royal Bank of Scotland

0.593

0.614

0.635

0.654

0.673

0.691

0.709

0.653

Bangkok Bank Berhad

0.432

0.457

0.482

0.506

0.530

0.553

0.575

0.505

Bank of America

0.560

0.582

0.604

0.625

0.645

0.664

0.682

0.623

The Bank of Nova Scotia

0.579

0.600

0.621

0.641

0.661

0.679

0.697

0.640

NA

NA

0.579

0.601

0.622

0.642

0.661

0.621

0.765

0.779

0.792

0.805

0.817

0.828

0.838

0.803

Foreign banks

Bank of China Bank of Tokyo-Mitsubishi UFJ Citibank Berhad

0.883

0.890

0.897

0.904

0.910

0.916

0.921

0.903

HSBC Bank

0.826

0.837

0.847

0.856

0.865

0.874

0.882

0.855

United Overseas Bank

0.901

0.907

0.913

0.919

0.924

0.929

0.934

0.918

NA

0.849

0.859

0.868

0.876

0.884

0.891

0.871

JP Morgan Chase Bank

0.581

0.603

0.624

0.644

0.663

0.682

0.699

0.642

OCBC Bank

0.899

0.906

0.912

0.917

0.923

0.928

0.932

0.917

Deutsch Bank

0.646

0.665

0.683

0.701

0.718

0.734

0.749

0.699

Standard Chartered Bank

Note: NA denotes data not available. Source: Authors’ own estimates.

70

Banks and Bank Systems, Volume 3, Issue 4, 2008

Conclusions

As in most previous studies on bank efficiency, we find that on average, bank deviates substantially from the best-practice frontier. The technical efficiency for the whole sample on average was 81 percent suggesting an input waste of 19 percent. Overall, the level of efficiency has slightly increased over the period of study. Our results also suggest that domestic banks on average were found to be relatively more efficient compared to foreign banks, 90.1 percent and 74.4 percent respectively. According to our results, AmBank, RHB and Malayan Banking appear to be the

most efficient domestic banks while Affin, Alliance and EON Bank were the least efficient banks. As for foreign banks, United Overseas, OCBC and Citibank were the most efficient banks while Bangkok Bank, Bank of America and Bank of China were the least efficient. As a caveat, the results should be interpreted with great caution since previous researches differ substantially across different estimation procedures. Further study should use other estimation approaches and look at the cost and profit efficiency and results thus can be compared.

References

1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22.

Aigner, D.A., Lovell, A.K., and Schmidt, P. (1977). Formulation and Estimation of Stochastic Frontier Production Models. Journal of Econometrics 6: 21-37. Berger, A.N. (1993). “Distribution-Free” Estimates of Efficiency in the US Banking Industry and Tests of the Standard Distributional Assumptions. The Journal of Productivity Analysis 4: 261-92. Berger, A.N. and Humphrey, D.B. (1997). Efficiency of Financial Institutions: International survey and Directions for Future Research. European Journal of Operational Research 98: 175-212. Berger, A.N., Hunter, W.C. and Timme, S.G. (1993). The Efficiency of Financial Institutions: A Review and Preview of Past, Present, and Future. Journal of Banking and Finance 17: 221-49. Charnes, A., Cooper, W.W. and Rhoades, E. (1978). Measuring the Efficiency of Decision Making Units, European Journal of Operational Research 6: 429-444. Coelli, T.J. (1996). A Guide to FRONTIER Version 4.1: A Computer Programme for Stochastic Frontier Production and Cost Function Estimation, CEPA Working Papers no. 7/96. Cornett, M.M. and Tehranian, H. (1992). Changes in Corporate Performance Associated with Bank Acquisition. Journal of Financial Economics. 31: 211-234. Elyasiani, E. and Mehdian, S.M. (1990). A Nonparametric Approach to Measurement of Efficiency and Technological Change: The Case of Large U.S. Commercial Banks. Journal of Financial Services Research 4 no. 2, July: 157-68. Evanoff, D.D. and Fortier, D.L. (1988). Re-evaluation of the Structure-Conduct-Performance Paradigm in Banking. Journal of Financial Services Research 1: 277-94. Farell, M.J. (1957). The Measurement of Productive Efficiency. Journal of the Royal Statistical Society 120, Part 111, Series A: 253-81. Ferrier, G.D., and Lovell, C.K. (1990). Measuring Cost Efficiency in Banking: Econometrics and Linear Programming Evidence. Journal of Econometrics 46: 229-245. Humphrey, D.B. (1990). Why Do Estimates of bank Scale Economies Differ? Economic Review, Federal Reserve Bank of Richmond, September/October: 38-50. Katib, M. Nasser and Mathews, K. (2000). A Non-Parametric Approach to Efficiency Measurement in the Malaysian Banking Sector. The Singapore Economic Review, 44: 89-114. Lang, G., and Welzel, P. (1996). Efficiency and Technical Progress in Banking: Empirical Results for a Panel of German Cooperative Banks. Journal of Banking and Finance 20: 1003-23. Li, Z. (2006). The Assessment Analysis of Efficiency of Commercial Banks Based on DEA Model, International Management Review, vol. 2 no. 3: 60-66. Meesun, W. and Broeck, J.V.D. (1977). Efficiency Estimation from Cobb-Douglas Production Functions with Composed Error. International Economic Review 18 no. 2, June: 435-44. Okuda, H., Hashimoto, H. and Murakami, M. 2003, The Estimation of Stochastic Cost Functions of Malaysian Commercial Banks and Its Policy Implications to Bank Restructuring, Centre for Economic Institutions Working Paper Series No. 2003-2. Shepherd, W.G. (1997). The Economics of Industrial Organization. 4th Edition, Prentice-Hall International, New Jersey. Smirlock, M. (1985). Evidence on the (Non) Relationship Between Concentration and Profitability in Banking. Journal of Money, Credit and Banking 17: 69-83. Srinivasan, A., and Wall, L.D. (1992). Cost Savings Associated With Bank Mergers. Working Paper 92-2, Federal Reserve Bank of Atlanta. Vol. 32: 1251-1266. Sufian, F. (2004), The Efficiency Effects of Bank Mergers and Acquisitions in Developing Economy: Evidence from Malaysia, International Journal of Applied Econometrics and Quantitative Studies Vol. 1-4: 53-74. Sufian, F. (2006). The Efficiency of Non-Bank Financial Institutions: Empirical Evidence from Malaysia, International Journal of Finance and Economics Issue 6 retrieved from www.eurojournals.com/finance.htm

71

Banks and Bank Systems, Volume 3, Issue 4, 2008

23. Sufian, F. and Abdul Majid, M.Z. (2007). Singapore Banking Efficiency and Its Relation to Stock Returns: A DEA Window Analysis Approach, International Journal of Business Studies, 15, 1: 83-106. 24. Yildrim, H.S. and Philippatos, G.C. (2007). Efficiency of Banks: Recent Evidence from the Transition Economies of Europe, 1993-2000. The European Journal of Finance, Vol. 13: No. 2: 123-143. 25. Yue, P. (1992). Data Envelopment Analysis and Commercial Bank Performance: A Primer with Applications to Missouri banks. Federal Reserve Bank of St. Louis January/February: 31-45. Appendix. Stochastic technical frontier maximum likelihood parameter estimates Parameter

Variable

Coefficient

Standard error

T-ratio

β0

Constant

4.9481

0.4401

11.2434

β1

lnX1

-0.0095

0.1614

-0.0588

β2

lnX2

0.2307

0.1447

1.5947

β3

lnX1 lnX1

0.0582

0.0202

2.8840

β4

lnX2 lnX2

0.0165

0.0184

0.8927

β5

lnX1 lnX2

-0.0405

0.0374

-1.0834

Sigma-square

σ

0.0914

0.1173

0.7797

Gamma

γ = σ

0.0209

47.1492

Log likelihood function

2



2 v

2 u



/ (σ

2 u

2 v

+ σ

2 u

)

0.9838

218.8504

Notes: X1 = Total deposits (deposits from customers and deposits from other financial institutions), X2 = Total Overhead Expenses (personnel expenses and other operating expenses). Dependent variable is Q, Total Earning Assets (financing, dealing securities, investment securities and placements with other banks).

72

Banks and Bank Systems, Volume 3, Issue 4, 2008

Wolfgang Benner (Germany), Lyudmil Zyapkov (Germany)

A multifactoral Cross-Currency LIBOR Market Model with an FX volatility skew Abstract

Based on LIBOR Market Models, we develop a rigorous pricing framework for cross-currency exotic interest rate instruments under a uniform probability measure and in a multifactoral environment that accounts for the empirically observed foreign exchange skew. The model resorts to a stochastic volatility approach with volatility dynamics following a square-root process and is designed to be flexible enough to allow for the incorporation of as much market information as possible. Using the Fourier transform, we produce closed-form valuation formulas for FX options by obtaining an explicit expression for the characteristic function, though in a mildly approximate fashion for the sake of analytical tractability. The main focus is placed on FX markets, in terms of which the calibration of model parameters can be performed on a wide range of FX options expiries and strikes. Keywords: Cross-Currency LIBOR Market Model, stochastic volatility, Fourier transform, foreign exchange skew, forward probability measure. JEL Classification: G13, E43, F31.

Introduction♦

The origins of the proposed Cross-Currency LIBOR Market Model (CCLMM) can be traced back to the need of developing a unified pricing framework for a number of cross-currency exotics. Initially confronted with a hybrid structure that required the simultaneous description of highly correlated interest rate markets, foreign exchange (forex) rate and hazard rate dynamics, the present work gained impetus from the necessity to determine the value of a cross-currency swap, which was to serve as an underlying of various derivative products, at an arbitrary future date. Typically, FX options exhibit a significant volatility skew that manifests itself in the at-the-money (ATM) implied volatility’s underestimation of in-the-money (ITM) option prices and overestimation of out-of-the-money (OTM) ones, whereby the ATM implied volatility has been obtained by inversion of an ATM option pricing formula based on a lognormal stochastic evolution of the forward forex rate. Moreover, it seems impossible for the most cross-currency derivatives to choose a particular strike, or a specific maturity of an FX option since they usually represent longdated exotic structures that either cannot be decomposed into plain-vanilla FX options, or at best depend on FX options for a wide range of strikes and maturities. Aggravating matters even further, exotic cross-currency interest rate derivatives are rarely structured to depend on ATM volatilities. They are usually designed with strikes far away from at-themoney. Hence, the volatility function needs to be calibrated to prices of FX options across all available maturities and strikes as suggested by Piterbarg (2006). He asserts that a model similar to that of Schloegl (2002) based on LIBOR Market Models, yet accounting for forex smiles in a proper manner, ♦

© Wolfgang Benner, Lyudmil Zyapkov, 2008.

and a good FX option calibration algorithm still awaits development. For this purpose, it appears natural to resort to an extension of the lognormaltype dynamics of the forward forex rate that is based on stochastic volatility. This paper proposes an integrated CCLMM under a uniform pricing measure in a multifactoral environment that allows for as much flexibility as possible in calibrating model parameters to market data. The pricing measure will be uniform as it will be applicable to (i) simple financial instruments that are affected only by the domestic interest rate market or the foreign interest rate market but not both, as well as to (ii) complex financial instruments that are affected by both the domestic and foreign interest rate markets linked by the forex market. With the intention to derive valuation formulas, we deflate all stochastic price processes using a single numeraire regardless of the market the price process belongs to or is affected by, thus ensuring pricing consistency between the markets and allowing the evaluation of complex financial structures within a LIBOR Market Model setup. The model design must be capable of reflecting market implied volatilities and exogenously assigned correlation structures between the interest rates and FX dynamics. However, the main focus will be placed on the calibration to FX options for various maturities and strikes simultaneously, while retaining one-factor assumptions for both interest rate markets. Though somewhat restrictive at first glance, this choice keeps the number of model parameters to be calibrated low affecting high speed of calibration without sacrificing accuracy of valuation. In addition, the model developed here can easily be used as a stepping stone to incorporating interest rate volatility smiles on a multi-currency basis, which remains a subject of future research. The various extensions of the forward LIBOR models could serve as a starting point of this effort. One 73

Banks and Bank Systems, Volume 3, Issue 4, 2008

possibility would be the postulation of alternate interest rate dynamics such as local volatility type of extensions based on constant elasticity of variance (CEV) processes pioneered by Andersen et al. (2000), or the adoption of a displaced-diffusion approach as elaborated, for example, by Benner et al. (2007). Jump-diffusions are treated in Glasserman et al. (2003a, 2003b), but have not gained much acceptance due to their producing of non-timehomogeneous volatility term structures and some other calibration complications. Finally as the modelling technique with probably the greatest explanatory power, the inclusion of stochastic volatility in the LMM is considered by three main research streams: Andersen et al. (2005) and Andersen et al. (2002), on which Piterbarg (2003) builds using the method of calibration by parameter averaging as described in Piterbarg (2005a, 2005b) and providing formulas that relate market and model skews and volatilities directly without the need to develop closed-form solutions of European option valuation problems. Joshi et al. (2003) choose a distinctly different way of analyzing the evolution of the swaption volatility matrix over time by assuming a specific time-homogeneous instantaneous volatility function whose parameters are allowed to vary stochastically. The paper is organized as follows. Section 1 develops a unified pricing framework under a uniform domestic forward measure. It determines both the dynamics of the domestic/foreign LIBORs and the forward forex rate with stochastic volatility. The reason why forward forex rates are being modelled directly is that, by definition, they represent price processes of tradable securities as opposed to spot forex rates. In fact, each forward forex rate follows a martingale under its natural forward measure, so that its dynamics are, under such a measure, fully specified by its volatility process. Section 2 derives an FX option pricing formula with stochastic forex volatilities based on mildly approximate assumptions in order to preserve the analytical tractability of the model. It also offers an elaborate overview of the implemented calibration procedure. The last section concludes with a brief summary of the main results and some suggestions for future work. For the sake of lucid presentation, all purely technical details are reserved to the appendices at the end of the paper.

1. Cross-Currency LMM under uniform probability measure 1.1. Definitions. Given a filtered probability space

( Ω,{F }

t t∈[ t0 ,t N ]

, P tN

)

satisfying the usual condi-

{W

(d ) tN

}

(t )

d =1,2,3,4

t∈[ t0 ,t N ] tions, let the tuple denote a ddimensional Brownian motion that introduces the source of uncertainty to the correlated dynamics of the foreign exchange market, both the domestic and foreign interest rate markets and to the meanreverting square root process of the common volatil-

ity V (t ) shared equally by forward forex rates of any maturity. In addition, we assume that the filtration

{F t }t∈[t ,t 0

N

]

t

N is the usual P -augmentation of

{W the natural filtration generated by

(d ) tN

d =1,2,3

t∈[ t0 ,t N ]

, so that it is right-continuous and complete. A com-

0 = t < t < ... < t

0 1 N is mon set of LIBOR maturities defined for both the domestic and the foreign currency markets. May the following symbols

L( s, t , T ) and Lf ( s, t , T ) indicate domestic and foreign (forward) LIBORs as of time s , starting at time t and maturing at time T respectively with s ≤ t ≤ T . For the sake of simplicity, Lti (t ) = L(t , ti , ti +1 )

Lf (t ) = Lf (t , t , t )

i i +1 and ti are designated to represent one-year forward rates start-

ing

at

ti

and

ending

at

ti +1 = ti + 1

with

f

i = 0,..., N − 1 . Moreover, L(t , t , ti ) and L (t , t , ti ) specify spot rates, whereas Q(t ) stands for the spot forex rate as units of domestic currency per unit of foreign currency. The forward forex rate, at which investors can buy or sell foreign currency for settlement at a future date, is determined by

FX i (t ) = B f (t , ti )Q(t ) B(t , ti ) . In terms of zero-coupon bonds,

B(t , ti ) and

f

B (t , ti ) denote the domestic and foreign bond respectively, while the LIBORs are given by:

⎛ B (t , ti ) ⎞ ⎛ B f (t , ti ) ⎞ − 1⎟ , Ltfi (t ) = α i −1 ⎜ f − 1⎟ with α i = ti +1 − ti = 1. Lti (t ) = α i −1 ⎜ ⎝ B (t , ti +1 ) ⎠ ⎝ B (t , ti +1 ) ⎠ In particular:

⎛ 1 ⎞ ⎛ ⎞ 1 L(t , t , t1 ) = α 0 −1 ⎜ − 1⎟ , Lf (t , t , t1 ) = α 0 −1 ⎜ f − 1⎟ with α 0 = t1 − t ≤ 1. ⎝ B (t , t1 ) ⎠ ⎝ B (t , t1 ) ⎠ 74

}

(t )

Banks and Bank Systems, Volume 3, Issue 4, 2008

Their stochastic evolution is characterized solely by the respective volatility functions γ i (t ) and γ i f (t ) , which are assumed deterministic within the main framework and can be calibrated independently for the domestic and foreign LIBORs to single-currency caps and swaptions using the well-known techniques for the single-currency LMM suggested by Rebonato (2002) and Brigo et al. (2002). To capture the implied volatility’s functional dependence on the corresponding interest rate option’s strike, a displaced-diffusion approach according to Rubinstein (1983) can be adopted, as shown by Benner et al. (2007). However, it will not be further pursued in

this place since we primarily concentrate on retaining sufficient control over forex smiles and developing a practicable FX option calibration algorithm for a wide range of maturities and across a variety of strikes. For convenience, the terminal forward probability measure PtN associated with the domestic bond maturing at the terminal date t N is chosen to be the uniform martingale measure throughout the paper. Though of marginal importance to the analysis, the structure of both bond volatilities σ ( f ) (t , t N ) is nonetheless needed and is determined according to Benner et al. (2007):

α i L( f ) (t , ti , ti +1 ) ( f ) α 0 L( f ) (t , t , t1 ) ( f ) t t t ( , , ) γ γ (t , t , t1 ). − i i +1 (f) (t , ti , ti +1 ) 1 + α 0 L( f ) (t , t , t1 ) i =1 1 + α i L

N −1

σ ( f ) (t , t N ) = −∑

1.2. Modeling forward forex rates with stochastic volatility. We begin by assuming the following general dynamics for the CCLMM:

dLi (t ) dLi f (t ) P P = µi (t )dt + γ i (t )dW1 tN (t ), = µi f (t )dt + γ i f (t )dW2 tN (t ), f Li (t ) Li (t ) dQ(t ) P P = µ q (t )dt + σ q (t )dWQ tN (t ), W1,2,tNQ → PtN − BROWNian motions. Q(t ) dW1 (t )dW2 (t ) = ρ12 dt , dW1 (t ) dWQ (t ) = ρ1Q dt , dW2 (t )dWQ (t ) = ρ 2Q dt

(1)

The last forward forex rate within the exemplary tenor structure represents a martingale under Pt N :

dFX N (t ) P P P P = σ f (t , t N )dW2 tN (t ) + σ q (t )dWQ tN (t ) − σ (t , t N )dW1 tN (t ) = σ Nfx (t )dW3 tN (t ), where FX N (t )

σ Nfx (t )2 = σ Nf (t )2 + σ q (t )2 + σ N (t )2 − 2σ Nf (t )σ N (t ) ρ1,2 − 2σ q (t )σ N (t ) ρ1,Q + 2σ q (t )σ Nf (t ) ρ 2,Q

(2)

dW1 (t )dW3 (t ) = ρ13 dt , dW2 (t )dW3 (t ) = ρ 23 dt Strictly speaking, σ Nfx (t ) is a stochastic quantity through its dependence on the realization of the LIBOR rates1. Though, it could be made conditionally deterministic to a high degree of accuracy by some quite sophisticated and extremely precise approximation methods, or simply by classical “driftfreezing” techniques2. Yet another approach of directly calibrating σ fx (t ) as a (deterministically) variable model parameter is being pursued henceforth since the forward forex rate represents the price process of a tradable asset denominated by the corresponding numeraire, hence an observable secu1

Benner, Zyapkov and Jortzik (2007) show how the spot forex volatility

σ q (t ) 2

can be calibrated.

For a brief review of existing approximations and the introduction of a new one based on BROWNian bridges refer to Benner et al. (2007), especially Sect. 3. The “drift-freezing” technique goes back to Daniluk and Gatarek (2005).

rity, as it has already been mentioned above. The decomposition of the volatility in (2) serves solely for the purpose of enabling us to determine the drift of any forward forex rate prior to the terminal one by switching from the natural to the terminal measure, as will be shown shortly. Such a model of the forward forex rate can safely be used to price FX options with different maturities but the same strike. When simultaneously pricing options with various strikes, however, the natural question arises as to how to account for the usually observed smile effect. For this reason, we resort to an extension of the lognormal-type dynamics of the forward forex rate beyond the geometric BROWNian motion and postulate a stochastic volatility evolution in conformity with Heston (1993) based on a common volatility V (t ) that follows a meanreverting square-root process under the physical probability measure: 75

Banks and Bank Systems, Volume 3, Issue 4, 2008

dFX N (t ) P = σ Nfx (t ) V (t ) dW3 tN (t ), dV (t ) = α (θ − V (t ) ) dt + ξ V (t )dW4 (t ) FX N (t )

(3)

dW1 (t )dW4 (t ) = ρ14 dt , dW2 (t )dW4 (t ) = ρ 24 dt , dW3 (t )dW4 (t ) = ρ34 dt Any previous forward forex rate no longer follows a martingale, but its dynamics under Pt N can be determined according to:

With FX i (t ) =

B f (t , ti )Q(t ) B (t , t N )

dFX i (t ) B (t , ti ) P ⇒ = µi fx (t ) dt + σ i fx (t ) V (t ) dW3 tN (t ), B (t , t N ) FX i (t )

(

P

P

where µi fx (t )dt = − σ i (t )dW1 tN (t ) − σ N (t )dW1 tN (t )



f i

P

P

)

P

P

P

(t )dW2 tN (t ) + σ q (t ) dWQ tN (t ) − σ N (t )dW1 tN (t ) − σ i (t )dW1 tN (t ) + σ N (t )dW1 tN (t )

)

α j L(t , t j , t j +1 ) P γ t (t )dW1 (t ) ⇒ j =i 1 + α j L (t , t j , t j +1 )

N −1

= −σ i fx (t ) V (t )dW3 tN (t )∑ P

tN

j

α j L(t , t j , t j +1 ) γ t (t )σ i fx (t ) V (t ) ρ13 , + α 1 L ( t , t , t ) j =i j j j +1

N −1

µi fx (t ) = −∑

j

while the volatility process evolves under the uniform martingale measure as follows: P

With dW4 (t ) = dW4 tN (t ) +

dB(t , t N ) P dW4 (t ) = dW4 tN (t ) + σ (t , t N ) ρ14 dt ⇒ B(t , t N )

P dV (t ) = ⎡⎣α (θ − V (t ) ) + ξ V (t )σ (t , t N ) ρ14 ⎤⎦ dt + ξ V (t )dW4 tN (t ).

Finally, the drift functions of both term structures of interest rates under Pt N remain to be computed. It is well-understood that the drift of the domestic LIBOR takes on the expression: N −1

µi (t ) = − ∑

α j L(t , t j , t j +1 )γ t (t )γ t (t )

j =i +1

j

i

1 + α j L(t , t j , t j +1 )

, i < N − 1.

A sequential procedure starting with the terminal foreign LIBOR, and moving backwards until the spot LIBOR rate is reached, renders the evolution of

d

LtfN −1 (t ) B f (t , t N )Q(t ) B(t , t N )

LtfN −1 (t ) B f (t , t N )Q(t ) B(t , t N )

γ

= ⎡⎣ µ

f N −1

f N −1

(t ) + γ Pt N 2

(t )dW

f N −1

(t ) +

the foreign term structure at last attainable. Since any traded asset scaled by the respective numeraire will be a martingale under the corresponding measure, we infer that LNf −1 (t ) B f (t , t N )Q(t ) , which represents a portfolio of foreign bonds in domestic currency, will be driftless when divided by the numeraire B (t , t N ) :

(t )σ (t ) V (t ) ρ 23 ⎤⎦ fx N

LtfN −1 (t ) B f (t , t N )Q(t ) B(t , t N )

LtfN −1 (t ) B f (t , t N )Q(t ) B(t , t N )

dt +

Pt N

σ Nfx (t ) V (t )dW3 (t ) ⇒

µ Nf −1 (t ) = −γ Nf −1 (t )σ Nfx (t ) V (t ) ρ 23 . Applying the same reasoning to an arbitrary foreign LIBOR prior to the terminal one, we obtain: Ltf (t ) B f (t , ti +1 )Q(t ) B (t , ti +1 ) With i and i < N − 1 ⇒ = Ltfi (t ) FX i +1 (t ) B(t , t N ) B (t , t N )

d ⎡⎣ Ltfi (t ) Bi +f 1 (t )Q (t ) / BN (t ) ⎤⎦

N −1 α L (t )γ (t ) ⎛ f ⎞ j tj tj f fx f ⎜ ( t ) ( t ) V ( t ) ( t ) µ γ σ ρ γ ρ = + + ∑ 12 ⎟ dt ti i +1 ti 23 ⎜ i ⎟ LtfN −1 (t ) Bi +f 1 (t )Q(t ) / BN (t ) j = i +1 1 + α j Lt j (t ) ⎝ ⎠ N −1 α L (t )γ (t ) j tj tj P P P γ tif (t )dW2 tN (t ) + σ i fx+1 (t ) V (t )dW3 tN (t ) + ∑ dW1 tN (t ) ⇒ j = i +1 1 + α j Lt j (t ) N −1

µi f = −γ t f (t )σ i fx+1 (t ) V (t ) ρ 23 − γ t f (t ) ∑ i

76

i

j =i +1

α j Lt (t )γ t (t ) j

j

1 + α j Lt j (t )

ρ12 .

Banks and Bank Systems, Volume 3, Issue 4, 2008

Summarizing the results so far, the proposed Cross-Currency LIBOR Market Model with an incorporated forex smile is fully described by the following system of stochastic differential equations:

dLti (t ) Lti (t )

N −1

= −γ ti (t ) ∑

j = i +1

α j L(t , t j , t j +1 )γ t (t ) j

1 + α j L(t , t j , t j +1 )

P

dt + γ ti (t )dW1 tN (t )

N −1 α L (t )γ (t ) ⎛ fx ⎞ j tj tj Pt N f f ⎜ γ ( t ) σ ( t ) V ( t ) ρ ρ = − + ∑ ti i +1 23 12 ⎟ dt + γ ti (t ) dW2 (t ) f ⎜ ⎟ Lti (t ) j =i +1 1 + α j Lt j (t ) ⎝ ⎠ N −1 α L (t , t , t dFX i (t ) j j j +1 )γ t j (t ) fx P σ i (t ) V (t ) ρ13dt + σ i fx (t ) V (t )dW3 tN (t ) = −∑ FX i (t ) j =i 1 + α j L (t , t j , t j +1 )

dLtfi (t )

(4)

P dV (t ) = ⎡⎣α (θ − V (t ) ) + ξ V (t )σ (t , t N ) ρ14 ⎤⎦ dt + ξ V (t )dW4 tN (t )

dWm (t )dWl (t ) = ρ ml dt , where m, l = {1, 2,3, 4} and m ≠ l. The correlation coefficients ρ ml , m, l = 1, 2,3 , can be chosen either by historical estimation, or by parsimonious parameterization of the correlation function1 and subsequent calibration to the information extracted from occasionally observed prices of quanto interest rate contracts. Assuming that both the domestic and the foreign LIBOR volatilities have previously been calibrated independently to single-currency caps and swaptions as indicated above, the model still needs to be calibrated to available FX options if it is intended to be used as a pricing tool for cross-currency exotics. Therefore, the main purpose of this article, aside from deriving

the CCLMM with a forex smile in (4), consists in the development of an effective and fast calibration algorithm, at the core of which a closed-form FX option valuation formula stands. 2. Option pricing formulas and calibration routines 2.1. FX option valuation by Fourier transform. Regardless of the model chosen for the evolution of

the forward forex rate, the price at option under the natural measure equivalent measure

PtN

t0 of the i th FX Pti

and under the

is given by:

⎤ + + B (t , ti ) P P ⎡ FXopti (t0 ) = B(t0 , ti ) E ti ⎡( FX i (t ) − K ) | F t0 ⎤ = B (t0 , t N ) E tN ⎢( FX i (t ) − K ) | F t0 ⎥ . ⎣ ⎦ B(t , t N ) ⎣ ⎦ 1 The driftless dynamics of FX i (t ) under their own model for the stochastic evolution of the underlynatural probability measure will produce the correct option price only if the discounting of the payoff is carried out using the appropriate numeraire – in this case the bond maturing at ti . The use of any other measure will introduce a covariance between the discounting and the payoff itself, for example when the pricing of plain-vanilla options on the whole spectrum of forward forex rates FX i (t ) ,

i = 1,..., N , is accomplished under a single measure like the terminal one. In order to recover the same option value, this fact has to be compensated for by altering the drift of the forward forex rate as shown previously. However, facing a complex pricing problem, which entails simultaneously several FX rates, a particular measure has to be specified and once it has been chosen, the presence of non-zero drifts is unavoidable and the need to formulate a 1 For the development of a semi-parametric full-rank model correlation structure consistent with historically estimated data refer to Schoenmakers and Coffey (2003).

ing(s) like (4) is inevitable. This has been the main purpose of our work – the development of a viable pricing model for exotic cross-currency interest rate instruments, which is at the same time flexible enough to allow for the incorporation of the entire market information, as of FX markets essentially meaning calibration to the whole range of FX options prices across all available maturities and strikes. As a step prior to the actual pricing of derivative structures, the calibration of the model can be carried out under any probability measure since we calibrate to plain-vanilla FX options, whose payoffs involve only a single forward forex rate at a particular point of time. It is acceptable to use a model separately calibrated for each option expiry since vanillas depend on the terminal distribution of the underlying only as opposed to exotics whose values usually depend on the full dynamics through time of a whole range of FX rates. Consequently, we revert every time to the natural measure Pti , effectively using for calibration always the same model though under different probability measure, to circumvent 77

Banks and Bank Systems, Volume 3, Issue 4, 2008

only unknown parameter is the conditional characteristic function φti0 (⋅) of ln ( FX i (t ) ) = Yi (t )

the unpleasant dependence alluded to above, which will admittedly complicate the producing of a closedform solution to the option pricing problem.

φti (u ) = E

The value of the FX option can be rewritten in terms of the real part of its Fourier transform, where the

0

Pti

(e

iuYi ( t )

)

| F t0 , eventually arriving at:

⎛ 1 1 ∞ ⎡ e −iu ln K φti (u − i ) ⎤ ⎞ ⎛ 1 1 ∞ ⎡ e− iu ln K φti (u ) ⎤ ⎞ FXopti (t0 ) 0 0 = FX i (t0 ) ⎜ + ∫ Re ⎢ ⎥ du ⎟ − K ⎜ + ∫ Re ⎢ ⎥ du ⎟ . ⎜2 π 0 ⎢ ⎟ ⎜ 2 π 0 ⎢ iuφti (−i ) ⎥ ⎟ B(t0 , ti ) iu ⎣ ⎦⎥ ⎠ 0 ⎣ ⎦ ⎠ ⎝ ⎝

(5)

Similar descriptions of the option price in a different for the Fourier transform designed to use the FFT to form have already been derived by numerous au- price options efficiently. In the appendix we propose thors, e.g., Bakshi et al. (2000) and Scott (1997), a different approach, which draws directly upon and numerically determined on the assumption that Lévy’s inversion theorem, and avail ourselves of the the characteristic function is known analytically. Gauss-Laguerre Quadrature to obtain the best nuOne disadvantage of the formula above is the singu- merical estimate of the Fourier integrals in (5). larity of the integrand at the required evaluation Therefore, the option pricing reduces to the calculapoint u = 0 , which ultimately precludes the applica- tion of the unknown conditional characteristic function of the Fast Fourier Transform (FFT). Therefore, tion. The dynamics directly relevant to valuing the Carr et al. (1999) develop a new analytic expression FX option are: i −1 α L (t ) α L (t ) j tj With σ (t , ti ) = −∑ γ t j (t ) − 0 t γ t (t ); dWm (t )dWl (t ) = ρ ml dt; m, l = {1,3, 4} 1 + α 0 Lt (t ) j =1 1 + α j Lt j (t )

1 P dYi (t ) = − σ i fx (t ) 2 V (t )dt + σ i fx (t ) V (t ) dW3 ti (t ) 2 P dV (t ) = ⎡⎣α (θ − V (t ) ) + ξ V (t )σ (t , ti ) ρ14 ⎤⎦ dt + ξ V (t )dW4 ti (t ) i −1 α L (t , t , t dLt j (t ) P k k k +1 )γ tk (t ) dt + γ t j (t )dW1 ti (t ). = −γ t j (t ) ∑ 1 α L ( t , t , t ) + Lt j (t ) k = j +1 k k k +1 It is well-known that according to the Markov property the characteristic function:

φti (u, y, v, l0 ,..., li −1 ) = E 0

Pti

(e

iuYi ( t )

| Yi (t0 ) = y, V (t0 ) = v, Lt0 (t0 ) = l0 ,..., Lti−1 (t0 ) = li −1

(6)

)

is determined as the solution of a partial differential function of V (t ) and by eliminating the stochastic equation (PDE) that can be found through Feynman- dependence on the LIBORs via the bond volatilities Kac’s theorem. To provide a closed-form solution in σ (t , t ) since the asymptotic form of the drift in the i the spirit of Heston (1993), however, we need to ensure the linearity of the coefficients in the related dynamics of the LIBORs rules out linearity with PDE. This property is obviously destroyed by the respect to lk , k = 0,..., i − 1 . The classical approach presence of the drift correction term to handling the LIBORs is by freezing them at their ξ V (t )σ (t , ti ) ρ14 in the dynamics of the volatility initial value. In addition, we need to approximate the square root of the volatility process within the process due to the change of measure. It becomes immediately apparent that the only way to explicitly drift function. Consequently, the dynamics of V (t ) calculate the wanted characteristic function is by become approximately of a square-root type and making the drift of the volatility process an affine after redefining the system of SDEs (6): 1⎛ V (t ) ⎞ P P With V (t ) ≈ ⎜ V (t0 ) + ⎟ , dW3 (t )dW4 (t ) = ρ34 dt and ⎜ ⎟ 2⎝ V (t0 ) ⎠ i −1 α L (t ) α 0 Lt (t0 ) j t 0 σ (t , ti ) ≈ σ (t0 , ti ) = −∑ γ t (t0 ) − γ t (t 0 ) ⇒ (7) 1 + α 0 Lt (t0 ) j =1 1 + α j Lt (t0 ) ti

j

ti

0

0

j

j

0

1 P dYi (t ) = − σ i fx (t ) 2 V (t )dt + σ i fx (t ) V (t )dW3 ti (t ) 2 ⎡ ⎤ 1⎛ V (t ) ⎞ P dV (t ) = ⎢α (θ − V (t ) ) + ξ ⎜ V (t0 ) + ⎟ σ (t0 , ti ) ρ14 ⎥ dt + ξ V (t )dW4 ti (t ), ⎜ ⎟ 2 V (t0 ) ⎠ ⎢⎣ ⎥⎦ ⎝

78

Banks and Bank Systems, Volume 3, Issue 4, 2008

we come by the following PDE in the backward variables and the respective boundary condition:

⎤ ∂φ 1 ∂φ fx 1 1 σ (t0 , ti ) ∂φ ⎡ ρ14 v ⎥ − σ i (t0 ) 2 v + + ⎢α (θ − v ) + ξ V (t0 )σ (t0 , ti ) ρ14 + ξ 2 2 ∂t0 ⎢⎣ V (t0 ) ⎦⎥ ∂v 2 ∂y 1 ∂ 2φ fx 1 ∂ 2φ 2 ∂ 2φ fx 2 σ ( t ) v ξ v σ i (t0 )ξρ34 v = 0 + + i 0 2 ∂y 2 2 ∂v 2 ∂v∂y

(8)

φti (u ) = eiuy . Suggested by the linearity of the PDE’s coefficients in v , we propose a solution like:

φti (u, y, v) = eC (t −t )+ D (t −t ) v +iuy , where C (0) = 0 and D(0) = 0. 0

(9)

0

0

By plugging this ansatz into (8), we obtain two ordinary differential equations (ODEs):

⎤ ∂D 1 2 2 ⎡ 1 σ (t0 , ti ) 1 1 = ξ D +⎢ ξ ρ14 + iuσ i fx (t0 )ξρ34 − α ⎥ D − u 2σ i fx (t0 ) 2 − iuσ i fx (t0 )2 ∂t0 2 2 2 V (t0 ) ⎢⎣ 2 ⎥⎦

(10)

∂C 1 ⎡ ⎤ = D ⎢αθ + ξ V (t0 )σ (t0 , ti ) ρ14 ⎥ . ∂t0 2 ⎣ ⎦ As shown in the appendix, the first one is a Riccati equation (see Oksendal (2000) Chapter 6) which can be solved by reducing it to a second-order linear ODE, whereas the second one is solved by direct integration. By means of their explicit solutions, as given by (B16) and (B18) respectively, we obtain an analytical expression for the characteristic function (9) that enables us to numerically determine the option price (5). 2.2. The calibration algorithm. For illustration purposes, we now consider a fictitious example of fitting the model to FX options across five different expiries and strikes. It is unnecessary to accentuate that the procedure can theoretically be extended to an arbitrarily wide range of maturities and strikes at the expense of rising computational time since the number of model parameters to match increases accordingly. Market prices of FX options in basis points, as displayed in Table 1, for strikes generated by K j (i ) = FX i (t 0 )e 0.1×

i ×δ ( j )

,

where i,j = 1,...5 and δ ( j ) = −1,−0.5,0,0.5,1

(11)

have been taken as a basis for the ensuing calibration routine. Table 1. Market prices of FX calls (in bp) for different strikes and maturities Expiry

FXopt 1

FXopt 2

FXopt 3

FXopt 4

FXopt 5

1y

74.31572

47.68933

25.64710

11.21897

3.99027

2y

90.20941

53.70243

23.56308

6.77430

1.26618

3y

98.96450

56.78179

21.40343

4.11798

0.38920

4y

103.73044

58.84490

20.92330

3.29027

0.20156

5y

105.93154

60.10471

21.14829

3.11624

0.15548

In the first place, we can check whether the FX option market is flat-smiled or not. The standard procedure would be to take the ATM strike, which happens to be that of the third FX option in the table δ ( j ) = 0 and consequently above since

K j (i ) = FX i (t0 ) , and to compute the ATM implied volatility by inversion of the lognormal option valuation formula. The same volatility is then utilized to price options with different strikes given by (11). The results, presented in Table 2, underline once again the fact that FX options exhibit a pronounced volatility skew with ITM options being underestimated whereas OTM ones are being systematically overestimated. From there the need stems to go beyond the standard geometric BROWNian motion and to resort to an extension of the lognormal-type dynamics of the forward forex rate that is based on stochastic volatility. Table 2. FX call prices (in bp) computed with the ATM implied volatility Expiry

FXopt 2

FXopt 3

FXopt 4

FXopt 5

71.97134

46.15283

25.64710

11.96459

4.55733

88.39820

52.03995

23.56308

7.48320

1.54316

97.65900

55.09986

21.40343

4.78883

0.53401

4y

102.85492

57.42792

20.92330

3.94375

0.31927

5y

105.22879

58.83967

21.14829

3.80896

0.28176

1y 2y 3y

FXopt 1

The calibration is performed on the FX call prices from Table 1, where the model prices are determined by (5). To obtain the best numerical estimate of both Fourier integrals, we employ the Gauss79

Banks and Bank Systems, Volume 3, Issue 4, 2008

Laguerre Quadrature, which is a Gaussian Quadrature over the interval [0, ∞) with a weighting function w( x) = e − x (see Abramowitz et al. (1972)). The model parameters to be calibrated are: (a) the parameters of the volatility process, initially set to ξ = 0.01 , α = 0.02 , θ = 0.001 and V (t0 ) = 0.01 ; (b) the FX volatility coefficients, initially set to σ i fx [i] = 1 − exp(−0.05(i − 0.2)) , i = 1,...,5 ; (c) the correlation coefficients, set to ρ14 = ρ34 = −0.2 . Aiming to reproduce FX option values for a wide range of strikes and expiries, we solve the calibration problem by simultaneously varying the model parameters (a), (b) and (c) until the sum of squared basis point differences between model and market prices has been minimized, which, for the sake of completeness, is reported here to have been achieved at 8.56059 bp. We use a fast unconstrained non-linear minimization algorithm, the Davidon-Fletcher-Powell (DFP) conjugate gradient method as described in Press et al. (1996), to make the distance ∆ 2 between the model and market matrices as small as possible: 2

5

∆2 =

∑ ωij ( FXoptijmarket − FXoptijmod el ) → Min ! (12)

i , j =1

The calibration has been carried out with identical constant weights ω , which essentially corresponds to an attempt to obtain the best global fit to the “complete” market information, as represented by our market prices matrix. The resulting optimal solution is shown in Table 3 below. Table 3. The best overall fit to the matrix of market call prices in Table 1 Expiry

FXopt 2

FXopt 3

FXopt 4

FXopt 5

73.44406

47.58264

26.25862

11.46718

3.54044

89.40880

53.47140

24.14783

6.74194

0.80012

98.25958

56.37568

21.92059

3.97862

0.14489

4y

103.30069

58.59928

21.35842

3.08668

0.05267

5y

105.40284

59.02163

19.80163

2.06586

0.00988

1y 2y 3y

FXopt 1

The stochastic volatility model brings about a significant improvement over the deterministic volatility one in any case. Prices of mid-maturity FX calls (i.e., 3y and 4y) are reproduced with a very good precision. For very short- and long-maturity options (i.e., 5y), however, the goodness of fitting worsens suggesting that the assumption of a unique volatility process V (t ) common to all forward forex rates might be too restrictive, especially when calibrating to a very wide range of maturities. 80

In order to capture certain features of a given exotic instrument, one could alternatively try to achieve the best fit to only a specific portion of the matrix of market prices sacrificing the remaining part of it. The quality of the partial calibration will be governed by the, in this case, non-constant weights ω assigned to the elements of the ∆ 2 distance function. The choice of the weights will mostly depend on the particular pricing problem. Conclusions

We proposed an integrated Cross-Currency LIBOR Market Model under a uniform probability measure in a multifactoral environment. The chief purpose of our paper has been the development of a viable pricing framework for exotic cross-currency interest rate instruments that is at the same time flexible enough to allow for the incorporation of as much available market information as possible. In terms of FX markets, on which the main focus of this work has been placed, fulfilling this purpose in a satisfying manner required the calibration to the whole range of FX options prices across all available maturities and strikes. This line of modelling has been reinforced by the significant volatility skew typically observed with FX options and the fact that it seems impossible for the most cross-currency derivatives to choose a particular strike, or a specific expiry of an FX option since they usually depend on a variety of strikes and maturities. The procedure eventually culminated in an extension of the lognormal-type dynamics of the forward forex rate beyond the geometric BROWNian motion and the postulation of a stochastic volatility evolution in conformity with Heston (1993) based on a unique volatility, common to all forward FX rates, that follows a meanreverting square-root process. After determining an analytical expression for the conditional characteristic function in a mildly approximate fashion for the sake of mathematical tractability, closed-form formulas for FX option prices have been obtained and numerically estimated with the aid of the GaussLaguerre Quadrature. The model prices computed in this manner served as the backbone of the ensuing calibration routine, by means of which we matched the model parameters so that the distance between the model and market prices matrices has been minimized. We have seen that introducing stochastic volatility led to a substantial improvement over the deterministic model in any case, although the fitting quality slightly worsened for very short- and especially long-dated options as compared to midmaturity ones. This feature has been ascribed to the possibility of our unique volatility process common to all forward FX rates being too restrictive, especially when faced with a wide maturity spectrum to

Banks and Bank Systems, Volume 3, Issue 4, 2008

calibrate to. More flexibility could be introduced by considering a different stochastic volatility process for the dynamics of each forward FX rate, however, inevitably making the calibration more cumbersome because of the additional volatility and correlation parameters and raising the potential problem of overfitting the model due to the increased number of

parameters. Above all, the pricing of cross-currency exotic interest rate products would become a very difficult task since the drift functions within the dynamics of both the foreign LIBOR and the forward forex rate would, aside from the unpleasant stochastic dependence on LIBOR rates, involve extra intra- and intercorrelated volatility processes.

References 1. Abramowitz, M. and I. A. Stegun (1972). Handbook of Mathematical Functions with Formulas, Graphs, and Mathematical Tables, Dover, New York, pp. 890 and 923. 2. Andersen, L., and J. Andreasen (2000). “Volatility Skews and Extensions of the LIBOR Market Model”, Applied Mathematical Finance, 7, 1-32. 3. Andersen, L., and J. Andreasen (2002). “Volatile Volatilities”, RISK December, 163-168. 4. Andersen, L. and R. Brothertone-Ratcliffe (2005). “Extended LIBOR Market Models with Stochastic Volatility”, The Journal of Computational Finance, 9, 1. 5. Bakshi, G., and D. Madan (2000). “Spanning and Derivative Security Valuation”, Journal of Financial Economics, 55, 205-238. 6. Benner, W., Zyapkov, L., and S. Jortzik (2007). “A Multifactoral Cross-Currency LIBOR Market Model”, SSRN working paper. 7. Brigo, D., and F. Mercurio (2002). “Calibrating Libor,” RISK January, 117-121. 8. Carr, P., and D. Madan (1999). “Option Valuation using the Fast Fourier Transform”, Journal of Computational Finance, 2, 61-73. 9. Daniluk, A., and D. Gatarek (2005). “A fully lognormal LIBOR Market Model”, RISK September, 115-119. 10. Heston, S. (1993). “A Closed-Form Solution for Options with Stochastic Volatility with Applications to Bond and Currency Options”, The Review of Financial Studies, 6, 2, 327-343. 11. Gil-Pelaez, J. (1951). “Note on the Inversion Theorem”, Biometrika, 38, 3/4, 481-482. 12. Glasserman, P., and N. Merener, 2003a, “Numerical Solution of Jump-Diffusion LIBOR Market Models”, Finance and Stochastics, 7, 1, 1-27. 13. Glasserman, P. and S.G. Kou (2003b). “The Term Structure of Simple Forward Rates with Jump Risk”, Mathematical Finance, 13, 3, 383-410. 14. Joshi, M., and R. Rebonato (2003). “A Displaced-Diffusion Stochastic Volatility LIBOR Market Model: Motivation, Definition and Implementation”, Quantitative Finance, 3, 6, 458-469. 15. Oksendal, B. (2000). Stochastic Differential Equations. An Introduction with Applications, Springer Verlag, Berlin Heidelberg New York. 16. Piterbarg, V. (2003). “A Stochastic Volatility Forward LIBOR Model with a Term Structure of Volatility Smiles”, SSRN Working Paper. 17. Piterbarg, V. (2005a). “Stochastic Volatility Model with Time-Dependent Skew”, Applied Mathematical Finance, 12, 2, 147-185. 18. Piterbarg, V. (2005b). “Time to Smile”, RISK May, 71-75. 19. Piterbarg, V. (2006). “Smiling Hybrids”, RISK May, 66-71. 20. Press, W.H., Teukolsky, S.A., Vetterling, W.T., and B.P. Flannery (1996). Numerical Recipes in C. The Art of Scientific Computing, Cambridge University Press, Cambridge, pp. 425-430. 21. Rebonato, R. (2002). Modern Pricing of Interest Rate Derivatives. The LIBOR Market Model and Beyond, Princeton University Press, New Jersey. 22. Rubinstein, M. (1983). “Displaced Diffusion Option Pricing”, Journal of Finance, 38, 213-217. 23. Schloegl, E. (2002). “A Multicurrency Extension of the Lognormal Interest Rate Market Models”, Finance and Stochastics, 6, 173-196. 24. Schoenmakers, J., and B. Coffey, 2003, “Systematic Generation of Correlation Structures for the LIBOR Market Model”, International Journal of Theoretical and Applied Finance, 6, 4, 1-13. 25. Scott, L. (1997). “Pricing Stock Options in a Jump-Diffusion Model with Stochastic Volatility and Interest Rates: Application of Fourier Inversion Methods”, Mathematical Finance, 7, 413-426. Appendix A. FX option pricing formula

Based on a derivation of the inversion theorem by Gil-Pelaez (1951), we determine the probability of finishing in-themoney, where ln ( FX i (t ) ) = Yi (t ) and ln( K ) = k , as follows:

With φ (u ) = i t0



∫e

−∞

∞ iuY

dP(Y ) =

∫ ( cos(uY ) + i sin(uY ) )dP(Y ),

−∞

81

Banks and Bank Systems, Volume 3, Issue 4, 2008

eiukφti0 (−u ) − e − iukφti0 (u ) iu

=

⎡ e −iukφti0 (u ) ⎤ and Re ⎢ ⎥= iu ⎢⎣ ⎥⎦





−∞

−∞

2i sin(uk ) ∫ cos(uY )dP(Y ) − 2 cos(uk ) ∫ i sin(uY )dP(Y ) iu ∞



−∞

−∞

cos(uk ) ∫ i sin(uY )dP (Y ) − i sin(uk ) ∫ cos(uY )dP(Y ) iu



∞ ⎡ e − iukφti0 (u ) ⎤ 1 1 du = − ∫ Re ⎢ ⎥ du ⇒ ∫0 iu iu 2 π 0 ⎢⎣ ⎦⎥ ∞ ⎡ e− iukφti0 (u ) ⎤ 1 1 P {Yi (t ) > k} = 1 − P {Yi (t ) < k} = + ∫ Re ⎢ ⎥ du 2 π 0 ⎢⎣ iu ⎥⎦

1 1 P {Yi (t ) < k} = + 2 2π



eiukφti0 (−u ) − e − iukφti0 (u )

(A12)

The delta of the option is somewhat convoluted, though it can be determined by similar techniques. For any positive numbers λ and ε , we have:

⎡ e − iukφti0 (u − i ) ⎤ 1 Re ⎢ ⎥du = ∫ π ε ⎣⎢ iu π ⎦⎥ 1

λ

λ ∞

λ ∞ ⎡ e − iuk ei (u −i )Y dP(Y ) ⎤ ⎡ eY eiu (Y − k ) dP(Y ) ⎤ 1 ⎥du = ∫ ∫ Re ⎢ ⎥du = ∫ε −∞∫ Re ⎢⎣ iu π ε −∞ ⎣ iu ⎦ ⎦

λ



λ



1 sin u (Y − k ) ⎡ cos u (Y − k ) + i sin u (Y − k ) ⎤ e dP(Y ) ∫ Re ⎢ du = ∫ eY dP(Y ) ∫ du ∫ ⎥ π −∞ iu π −∞ u ⎣ ⎦ ε ε 1

Y

Proceeding from this expression now, we obtain the step function by simply letting ε tend to zero and λ tend to infinity as shown in Gil-Pelaez (1951). Thereafter, we conveniently arrive at the wanted conditional expectation in the following manner: λ





sin u (Y − k ) 1 1 1 e dP(Y ) lim ∫ du = ∫ sign(Y − k )eY dP(Y ) = ∫ eY dP(Y ) − ∫ eY dP(Y ) ∫ λ →∞ π −∞ u 2 −∞ 2 Y >k 2 Y k

φti (−i ) 0

2

+

⎡ e − iukφti0 (u − i ) ⎤ Re ⎢ ⎥du, π ∫0 ⎢⎣ iu ⎥⎦ 1



essentially meaning that the delta of the option is defined by: ∞ ⎡ e −iukφti0 (u − i ) ⎤ 1 1 + ∫ Re ⎢ ⎥du 2 π 0 ⎢⎣ iuφti0 (−i ) ⎥⎦

Thus, additionally observing that

(A13)

φti (−i ) = E 0

Pti

(e

i ( − i )Yi ( t )

)

| F t0 = FX i (t0 ) , the option price is readily computed in

terms of the conditional characteristic function like:

FXopti (t0 ) + P = E ti ⎡( FX i (t ) − K ) | F t0 ⎤ ⎣ ⎦ B(t0 , ti ) i ∞ ⎛ 1 1 ∞ ⎡ e− iukφti (u ) ⎤ ⎞ ⎡ e − iukφti0 (u − i ) ⎤ ⎞ FXopti (t0 ) ⎛ φt0 (−i ) 1 0 (A14) =⎜ + ∫ Re ⎢ ⎥du ⎟ − K ⎜ + ∫ Re ⎢ ⎥ du ⎟ ⎜ 2 ⎟ ⎜ ⎟ π π B(t0 , ti ) iu iu 2 ⎢ ⎥ ⎢ ⎥ 0 0 ⎣ ⎦ ⎠ ⎣ ⎦ ⎠ ⎝ ⎝ − − iuk i iuk i ⎛ 1 1 ∞ ⎡ e φt (u − i ) ⎤ ⎞ ⎛ 1 1 ∞ ⎡ e φt (u ) ⎤ ⎞ FXopti (t0 ) 0 0 ⎟ du K − + ∫ Re ⎢ = FX i (t0 ) ⎜ + ∫ Re ⎢ ⎜ ⎥ ⎥ du ⎟ . i ⎜ ⎜ ⎟ π iu φ i π iu ( ) 2 B(t0 , ti ) − 2 ⎢ ⎥⎦ ⎟⎠ ⎢ ⎥ t 0 0 ⎣ 0 ⎣ ⎦ ⎠ ⎝ ⎝

82

Banks and Bank Systems, Volume 3, Issue 4, 2008

Appendix B. Solution of the ordinary differential equations

Starting with the Riccati equation in (10):

1 1 σ (t0 , ti ) ∂D ρ14 + iuσ i fx (t0 )ξρ34 − α , = aD 2 + bD + c, with a = ξ 2 , b = ξ 2 2 ∂t0 V (t0 )

(B15)

1 1 c = − u 2σ i fx (t0 ) 2 − iuσ i fx (t0 ) 2 , 2 2

it has been led back to a second-order linear differential equation by substitution:

Put aD = v ⇒ v ' = aD ' ⇒ v ' = v 2 + bv + ca Substitute v =

u' u '' and sin ce v ' = − + v 2 ⇒ u '' = −bu ' − acu. u u

The solution ansatz is of an exponential type and is plugged into the ODE to be solved:

e z ( t −t0 ) ⇒ z 2 − bz + ac = 0 ⇒ z1,2 =

b ± b 2 − 4ac , 2

ultimately arriving at the following general solution along with the respective boundary condition:

Ae

b + b 2 − 4 ac ( t − t0 ) 2

+ Be

b − b 2 − 4 ac ( t − t0 ) 2

From D(0) = 0 and D = A

u' ⇒ u ' (t − t0 ) = 0 ⇒ t = t0 au

b + b 2 − 4ac b − b 2 − 4ac +B = 0. 2 2

However, we are not aimed at finding an explicit solution of the second-order linear differential equation. We do not need to compute A and B separately, which is by the way based on a single boundary condition impossible, the ratio A / B would suffice since we actually seek to determine D(t − t0 ) . Having made this crucial observation, we obtain:

With

A b − b 2 − 4ac =− ⇒ B b + b 2 − 4ac

b + b 2 − 4ac b + A e ' u 2 D(t − t0 ) = =− b+ au aAe

b 2 − 4 ac ( t −t0 ) 2 b 2 − 4 ac ( t −t0 ) 2

b − b 2 − 4ac b − e +B 2 + aBe

b − b 2 − 4 ac ( t − t0 ) 2

2

b + b − 4 ac

2

b 2 − 4 ac ( t −t0 ) 2

:B

:B

b − b − 4 ac

b − b 2 − 4 ac

( t −t0 ) b − b 2 − 4ac ( t − t0 ) −b + b 2 − 4ac e 2 + e 2 :e 2 2 2 D(t − t0 ) = − b + b 2 − 4 ac b − b 2 − 4 ac b − b 2 − 4 ac 2 ( t −t0 ) ( t −t0 ) ( t −t0 ) −b + b − 4ac 2 2 + ae :e 2 a e 2 b + b − 4ac

−b + b 2 − 4ac b − b 2 − 4ac b2 − 4 ac ( t −t0 ) + e 2 2 D(t − t0 ) = ⇒ −b + b 2 − 4ac b2 − 4 ac ( t −t0 ) a+a e b + b 2 − 4ac ⎛ ⎞ 2 ⎜ ⎟ b − 4 ac ( t − t0 ) b 2 − 4ac − b ⎜ 1− e ⎟. D (t − t0 ) = ⎜ b − b 2 − 4ac 2 ⎟ 2a b − 4 ac ( t − t0 ) ⎜ 1− ⎟ e ⎜ ⎟ 2 ⎝ b + b − 4ac ⎠

( t −t0 )

(B16)

83

Banks and Bank Systems, Volume 3, Issue 4, 2008

The second ODE:

∂C 1 = α _ θ .D, with α _ θ = αθ + ξ V (t0 )σ (t0 , ti ) ρ14 ∂t0 2

(B17)

is solved by direct integration: t

With C (0) = 0 and j = b 2 − 4ac ⇒ C (0) − C (t − t0 ) = α _ θ ∫ D(t − u )du ⇒ t0

t − t0

D(u )du = α _ θ

1−

C (t − t0 ) = α _ θ

j −b 2aj

b+ j

b − j j ( t −t 0 ) 1− e b+ j



1−

j −b 2a

t − t0

1 − e ju b − j ju ∫0 ∫0 b − j ju du. Put 1 − b + j e = y ⇒ 1− e b+ j b − j j ( t −t 0 ) (1 − y )(b + j ) e 1− 1− b+ j j −b b− j C (t − t0 ) = −α _ θ dy ∫ 2aj y (1 − y ) b− j C (t − t0 ) = α _ θ

b− j b+ j

b+ j j −b dy − α _ θ y (b − j ) 2aj

1−

b − j j ( t −t 0 ) e b+ j



1−

b− j b+ j

1 dy y (1 − y )

b − j j ( t −t 0 ) ⎛ b − j j ( t −t0 ) ⎞ 1− e 1− e b+ j ⎜ ⎟ ⎛1 1 ⎞ j −b b+ j j −b b+ j ⎟ −α _θ ln ⎜ C (t − t0 ) = α _ θ + ⎜ ⎟dy ∫ 2aj b − j ⎜ 1 − b − j 2aj y 1− y ⎠ ⎟ ⎝ b− j 1− ⎜ ⎟ b+ j b+ j ⎝ ⎠ ⎛ b − j j (t −t0 ) ⎞ ⎛ b − j j (t −t0 ) ⎞ 1− 1− e e ⎟ ⎜ ⎟ α _ θ (b + j ) α _ θ ( j − b) ⎜ b + j b+ j ⎟ ⎟− ln ⎜ ln ⎜ C (t − t0 ) = − 2aj 2aj ⎜ 1− b − j ⎟ ⎜ 1− b − j ⎟ ⎜ ⎟ ⎜ ⎟ b+ j b+ j ⎝ ⎠ ⎝ ⎠

C (t − t0 ) + α _ θ

C (t − t0 ) = −

C (t − t0 ) = −

j −b 2aj

α _θ a

α _θ a

1−

b − j j ( t −t 0 ) e b+ j



1−

b− j b+ j

1 d (1 − y ) 1− y

⎛ b − j j ( t −t0 ) ⎞ ⎜ 1− b + j e ⎟ ⎟ +α _θ ln ⎜ ⎜ 1− b − j ⎟ ⎜ ⎟ b j + ⎝ ⎠ ⎛ b − j j ( t −t0 ) ⎞ ⎜ 1− b + j e ⎟ ⎟ +α _θ ln ⎜ ⎜ 1− b − j ⎟ ⎜ ⎟ b+ j ⎝ ⎠

j −b 2aj

1−

b − j j ( t −t 0 ) e b+ j



1−

b− j b+ j

j −b (t − t0 ). 2aj

1 d (1 − y ) ⇒ 1− y

(B18)

In conclusion, we hold (B16) and (B18) to be the solutions of the ordinary differential equations in (10) being integral part of the characteristic function (9), whereby a , b , c and α _ θ are the substitutes defined previously by (B15) and (B17) respectively.

84

Banks and Bank Systems, Volume 3, Issue 4, 2008

Andreas Burger (Germany), Juergen Moormann (Germany)

Productivity in banks: myths & truths of the Cost Income Ratio Abstract

Understanding productivity is essential for banks when considering the fierce international competition. Yet, how do banks perform in terms of their productivity? And how can productivity be measured? A popular measure for productivity and efficiency in banking is the Cost Income Ratio (CIR). But this measure is misleading in both terms. This article discusses the difficulties in measuring productivity in banks and criticizes the inadequate usage of the CIR. In order to derive an approximation of a bank’s productivity an adjusted CIR measure is proposed. The elimination of unwanted effects is conducted in a pragmatic way and is based on publicly available data. This approach is illustrated using large European stock exchange-listed banks as an example. Furthermore, new opportunities for measuring the banks’ productivity are outlined on the basis of introducing efficiency measurements on a process level. Keywords: Cost Income Ratio (CIR), efficiency measurement, productivity in banks, productivity measurement, Data Envelopment Analysis (DEA). JEL Classification: G21, L25, M16.

Introduction∗

Competition in the banking industry has intensified enormously in recent years, a trend that can be observed particularly in the fragmented European banking market. Accordingly, the consolidation of market participants has proceeded at a steady rate and has crossed national borders and dimensions. In fact, the pace has even intensified as a result of the current financial market crises. After establishing large enterprises in several countries, the industry is witnessing the emergence of banks with a value and profitability exceeding any size known thus far. For example, the market capitalization of the five biggest European banks represented $ 193 Bn. for HSBC, $ 92 Bn. for Banco Santander, $ 84 Bn. for BNP Paribas, $ 64 Bn. for Intesa Sanpaolo and $ 48 Bn. for Unicredit (30.09.2008). In many countries the prices for banking products – in terms of interest rates, commissions and fees – are under pressure. A general decline of margins as well as a far reaching assimilation within Europe is expected for the future. This process is accelerated by the harmonization efforts of the European Commission for the financial services market. The profitability of banks is particularly influenced by two factors: The respective market conditions regarding competition and price levels as well as service production capability (Varmaz, 2006). The main indicators for evaluating service capability are productivity and efficiency. Studies of efficiency show that there are large differences between different banks’ service capabilities. The room for improvement in comparison to best-practice banks is usually estimated at 15% to 25% (Berger and Humphrey, 1997; Beccalli, Casu, and Giradone, 2006). These inefficiencies offer opportunities for increasing productivity and, consequently, for improving the banks´ profitability. © Andreas Burger, Juergen Moormann, 2008

Thus, the question arises how banks perform in terms of their productivity and who will belong to the successful banks in the next years. The following sections deal with the current development in a European context and the measurement of productivity and efficiency. The analysis starts with the traditional Cost Income Ratio (CIR) – a popular and critical measure for a bank’s productivity. In the course of this paper the adjustment of the CIR is explained and new approaches for measuring efficiency in banks on a process level are introduced. 1. Harmonization of the European banking market

The European Commission pursues a consequent policy to reduce inefficiencies in national markets and oligopolistic structures with the goal of establishing a harmonized market for financial services (European Commission, 2005). According to the European Commission, a “Level Playing Field” does not exist at this point in time. In its analysis of the retail banking sector, the European Commission ascertained a latent inelasticity of prices in local markets, which is caused by a lacking demand power. The reason is restricted competition resulting from market participants’ efforts to close off their markets and to create market entry barriers. This in turn results in significant differences in prices for deposits and loans as well as prices for additional banking services (EU Commission, 2006). Consequently, significant differences in profitability among European banks can be observed. The creation of an integrated, open and efficient European market for financial services is the central mission of the European Commission. An efficient banking system generates stability and is beneficial for the consumer. From an economic point of view, only highly productive, i.e. efficient suppliers will survive in conditions of fair and transparent competition, as the margins will continuously decrease and inefficient market participants will thus vanish from 85

Banks and Bank Systems, Volume 3, Issue 4, 2008

the market. In such an environment it is almost impossible to set up oligopolistic structures to achieve excessive yields at the expense of the customers. Considering the aforementioned harmonization efforts, a successive assimilation of margins and at the same time a decrease of margins are to be expected in the European financial services sector. This development also affects increasingly the East- and Southeast-European markets. The opportunity to achieve a high profitability by realizing high margins in local markets with low levels of competition will decrease step by step. This awareness plays a decisive role in the strategies of banks. Clearly, established banks which have benefited from the conditions until now are hesitant when it comes to abandoning their convenient source of income. Yet, it seems unavoidable that in the future the productivity of banks will gain increased importance for generating profitability (Vennet, 2002; Rose and Hudgins, 2004; Poddig and Varmaz, 2005). 2. Productivity and efficiency in banks

The aspects of measuring, analyzing and optimizing operational performance play a vital role when the decrease of margins is considered. Especially, the evaluation of productivity and efficiency of banks is critically important (Burger, 2008).

Productivity expresses the relation of output and input. The measurement is directly based on quantities. Productivity is an operational ratio which can be easily calculated and compared. Its strong relatedness to the production process and the consideration of specific input and output qualities allows for a measurement of the “success” of transforming input into output. Additionally, productivity can also be measured under consideration of price components. Thus, several factors with different dimensions can be aggregated. But monetary assessment of the factors represents only a “support calculation”. Experience in banks shows that it is extremely difficult to compare productivity of different banks as distinct and accepted definitions for the main terms (e.g., order volume, card transaction) do not even exist. Measurement of productivity is particularly crucial in process management in order to determine service capability and to identify improvement opportunities. A bank is more productive than its competitors if, for instance, a security transaction is settled and cleared with fewer resources, i.e. either fewer working hours or lower costs. The term efficiency is often used as a synonym for productivity, but according to Cooper, Seiford and Zhu (2004), Coelli et al. (2005), and Sherman and Zhu (2006) this is not accurate. There are many 86

discussions and public announcements of improvement programs in business journals and in the banking community. Yet, “efficiency” is neither precisely defined nor measured. Efficiency can be understood as a comparative concept. The result of transforming input into output is compared to a benchmark which is basically represented by the best-practice case. The precise definition of the underlying elements, however, depends on the particular case at hand (Forsund and Hjalmarsson, 1974). An evaluation of efficiency is impossible if only a single measurement point or several measurement points without an according benchmark exist. A scientific definition of efficiency usually follows the Pareto-Koopmans concept. “Full (100%) efficiency is attained for an object […] if and only if none of its inputs or outputs can be improved without worsening some of its other inputs or outputs” (Cooper, Seiford and Zhu, 2004). A bank, a branch or a business process is efficient if and only if it utilizes – in comparison to other similar objects – the technical facilities and input factors in the optimal way (technical efficiency), uses the resources in the best possible way (allocative efficiency) and produces at an optimal scale (scale efficiency) (Coelli et al., 2005). The measurement of efficiency represents an advancement of productivity analysis. The concept of efficiency is based – in simple terms – on the calculation of total productivity under consideration of different input and output factors. The position and functional form of the efficiency line, which is represented by the sum of all best-practice cases, are usually not known so that estimation is necessary. 3. CIR – the productivity ratio for banks?

In scholarly journals and business practice, including evaluations of rating companies, the discussion about productivity and efficiency in banks is mostly based on the Cost Income Ratio (CIR), which is also known as efficiency ratio. Even though the predication power of the CIR is not clear at all, this ratio is widely regarded as a yardstick when comparing productivity and efficiency of banks. The commonly held notion claims that a high CIR is equivalent to low productivity and low efficiency and vice versa. However, the limited predication power of the CIR becomes apparent in the next two subsections. Consequently, an adjusted CIR is suggested afterwards. The procedure allows for an indicative and pragmatic measurement of productivity in banks. 3.1. Structure of the cost income ratio. The cost income ratio puts expenses (administrative costs) and earnings (operating income) of a bank in relation to each other. The CIR shows how many Euros

Banks and Bank Systems, Volume 3, Issue 4, 2008

(or dollars etc.) were needed in a given period of time to generate one Euro (or dollar etc.) in revenue. Consequently, the CIR measures the output of a

bank in relation to its utilized input. Figure 1 shows the components needed to determine the CIR. Work force Personal Labor costs aufwand € 2 Bn.

x

Sachaufwand Material costs

Administrative Verwaltungscosts aufwand

+

€ 5 Bn

CIR

=

/

25.000 25,000

Labor costs per head € 80,000 80.000

€ 2,7 Bn Abschreibung Depreciation (AfA) € 0,3 Bn.

Interest generating assets

62,5% Zinsüberschuß Interest surplus

Operating Operative income Erträge

+

x

€ 4 Bn.

Interest margin

Commission Provisions-surplus überschuß

€ 8 Bio.

€ 2,5 Bn.

Trading Handelsresult ergebnis € 1 Bn. Sonstige Other income Erträge € 0,5 Bn.

€ 200 Bn.

2%

x

Number Anzahlof accounts Konten 2,5 Bn. Mill.

Price per Preis pro account Konto € 1,000 1.000 Exemplarily Exemplarisch

Source: ProcessLab Fig. 1. Scheme of CIR calculation and quantitative example

The provision for risks, which decreases the earnings in the profit and loss calculation, is usually not included. Non-recurring earnings or expenses are handled differently: (1) Gross earnings (interest and commission surplus) are put in relation to administrative expenses. (2) All operative income components (gross earnings plus trading result and other income) are placed in relation to administration expenses. Some banks also disclose an adjusted CIR which eliminates non-recurring effects. The comparison of banks based on the CIR is fast and easily feasible, and the result appears to be intuitive. The simplicity is certainly an advantage of the CIR, and this might be a reason for its popularity. The ratio is considered to be meaningful for investors. Practically every bank discloses the ratio in its company reports. 3.2. Factors influencing the CIR. A closer look at the CIR calculation reveals that price components (interest rates, commission fees and factor costs) influence the determination of earnings and expenses and consequently distort the predication power of the CIR. While the determination of earnings is based on sales quantities, which are assessed on the basis of prices, the determination of administrative costs requires costs of production factors (in particular, labor costs per head). Particularly the

consideration of prices on the earnings side seems to be problematic for the measurement of productivity. The purpose of measuring productivity is to detect the level of a bank’s production and settlement capability. Therefore, market conditions reflected in prices as well as sales revenues of a bank should not be included in the measurement of productivity. The ability to achieve higher prices by no means improves the productivity of a bank. Comparisons of banks in different countries reveal significant differences in interest rates, commission fees and factor costs. As these elements are incorporated in the CIR calculation, banks situated in a country with comparatively high interest margins ceteris paribus appear to be more productive than others. In order to perform a more detailed analysis, the CIR and interest margins of 62 stock exchangelisted European banks were compared. The data were extracted from a periodically published report (Deutsche Bank AG, 2008). Figure 2 shows the correlation between the interest margin and the CIR for banks in selected European countries. Because of the vital meaning of net interest income of European banks – which accounts for almost 50% of the earnings in 2007 – the interest margins have a significant impact on the CIR. Net interest margins 87

Banks and Bank Systems, Volume 3, Issue 4, 2008

within the respective countries – with the exception of Switzerland (decline of the interest margin > 0.5%) – were relatively robust in the period of 20042007. When analyzing the impact of the interest margin on the CIR in 2006, there is a significant

correlation (R² = 56.0%). The higher the interest margin in the local markets, the lower is the CIR. This correlation highlights the strong influence of this price component on the CIR.

CIR v. Net Interest Margin for European Banks 85,0 80,0 Switzerland

75,0

CIR (in %)

70,0

Germany

Austria

60,0

2004

Italy

65,0

2005 2006

France

2007

Greece

55,0 Scandinavia

Ireland

50,0

40,0 0,00

Great Britain

Spain / Portugal

45,0

0,50

1,00

1,50

2,00

2,50

3,00

Net Interest Margin (in %)

3,50

4,00

4,50

R-Sq = 27.3%, for 2004 until 2007 R-Sq = 40.1%, for 2007

Source: ProcessLab, Data: Deutsche Bank AG (2008). Fig. 2. Correlation of national net interest margins and national cost income ratios in selected European countries in the period 2004 until 2007

The CIR is affected by additional factors which further decrease the predication power concerning productivity. These factors are not related to and therefore independent of the level of service production. However, they have a direct influence on the earnings and expenses of a bank and consequently influence the CIR: ♦ Business model: The specific business model of a bank has a direct effect on the CIR. Therefore, significant differences in the average CIR are the result (e.g., in 2007: private banks: 50.6%, multiregion-banks: 49.8%, universal banks: 60.1%, and corporate banks with focus on capital markets: 79.0% (Deutsche Bank, 2008)). Welch (2006) identifies drivers depending on business models to explain differences in CIR in British banks. ♦ Regional focus: As shown above via the example of interest margins, commission fees and factor costs differ markedly in individual countries as well. 88

♦ Cyclic improvements of income: The CIR seems to be more favorable in boom times because of over-proportionately high earnings. Times characterized by an economic downturn usually generate a decrease in earnings resulting in a less favorable CIR. ♦ Non-recurring effects: Non-recurring income, such as selling holdings or non-recurring costs caused by restructuring programs, is included in the CIR calculation. Banks rarely disclose adjusted CIR values. ♦ Risk affinity: The risk affinity of a bank concerning granting loans has an important impact on the CIR. A higher risk affinity leads to higher interest margins because of higher risk premiums. Thus, interest earnings increase and the CIR decreases. Deferred risk adjustments are not considered in the CIR calculation. Yet, they are reflected in the profitability of a bank. This means that a bank can disclose a favorable CIR even though it needs to write off billions of assets – for instance because of the subprime cri-

Banks and Bank Systems, Volume 3, Issue 4, 2008

sis. A good example is UBS, which deteriorated its CIR of 69.7% (31.12.2006) to 110.3% (31.12.2007) within one year (Fourth Quarter 2007 Report of the UBS Group). The jump was not caused by a sudden decrease in productivity but it was due to the impairments on US mortgage loans with low or no collateral (subprime). ♦ Balance sheet management: The balance sheet policy of a bank affects the refinancing costs along the yield curve. These costs are considered in the interest earnings and thus have an impact on the CIR as well. 3.3. Adjustment of the CIR. To determine the productivity of a bank there is a demand for a ratio which represents the actual performance – namely a ratio that considers the production and settlement of bank services, i.e. transforming resources (inputs) such as human resources, IT systems etc. into products and services (outputs). If productivity is understood and viewed in this way, then the CIR is an unsuitable measure for productivity.

A direct calculation of quantity-based productivity is hardly possible due to the lack of publicly available information in the banking business. But if the price components on the earnings side and the expenses side can be eliminated, the focus can be placed on the quantity components of performance. This approach is also applied in the adjustment of profitability indices which are based on profitability ratios, e.g. the Total Factor Productivity (TFP)-index (Coelli et al., 2005). However, such adjustments are difficult as a result of lacking price information. An extensive comparison of price structures in different countries has not been successfully performed yet, since transparency is low and product supply and consumer behavior are very different. Comparative studies are only available for selected business fields, e.g. performance of account-based services in worldwide retail banking (Capgemini, EFMA and ING, 2007). The following procedure of adjusting price components within the CIR is based on available data and includes most of the banks´ earnings and expenses. A complete adjustment is not possible because of incomplete or missing data related to income from commission business and due to lacking data to adjust for differences in the area of material costs. The adjustment can be carried out as follows: ♦ On the earnings side, the differences in interest margins of the observed banks have to be excluded. As the interest income of European banks represents almost 50% of the total earnings, about half of the earnings can be adjusted. The adjustment of interest margins is easy since the needed data can be calculated based on publicly available information.

♦ On the expenses side, the differences in national labor costs have to be eliminated. This entails adjusting a large expense position as labor costs account for more than 60% of the expenses of European banks.

The effects of adjusting the CIR are shown in the following by means of a comparison of European stock exchange-listed banks for the year 2007. Differences in business models etc. are neglected. The data of German banks are taken as reference data for the adjustment. Thus, the underlying mathematical assumption is that all European banks have the same interest margins and labor costs as German banks. The adjustment leads to interesting results (Figure 3). The starting point for the analysis is the unadjusted CIR. Here, Iberian banks hold the leading position (unadjusted CIR of 45.3%). Greek banks with a CIR of 48.5% are ahead of the UK (50.6%) and of the European average (55.9%). Swiss banks trail with a CIR of 81.8%. If the CIR is considered as a productivity ratio, Swiss banks are the least productive in Europe. French and German banks show unfavorable values as well. The adjustment of price effects involves two steps. Firstly, the discrepancies in market prices regarding interest are eliminated and adapted to the German level. For this purpose the average interest margin of German banks (0.73%) is used. Through this adjustment the CIR of each country changes noticeably. Now French banks are at the top with a CIR of 62.2%, followed by German banks (63.5%) and Nordic banks (64.0%), while the situation of Greek banks (105.5%) has worsened seriously. Austria holds the last position in the ranking of European banks with an adjusted CIR of 122.0%. Secondly, the CIR is adjusted for national differences in labor costs. The ranking changes again. The labor cost level in Germany is slightly above the European average whereas Greek labor costs lie 40% behind the average (Eurostat, 2006).1 As a result of the adjustment of labor costs to the German level, the CIR of Greek banks changes even more. The supposedly high productivity of Greek banks worsens to a CIR of 150.5%. In other words, it would cost 3 Euros to earn 2 Euros. An adjustment of CIR for price components on the earnings and expenses side leads to a different perception of productivity in European banks. Based on the classic CIR, Iberian and Greek banks appeared to be the most productive. After the adjustment, 1 The basis of the calculation are the medians of gross year income of banks and insurers in European regions, e.g. for the Iberian region the data of Spanish and Portuguese banks have been weighted according to the percentage of employees in both countries.

89

Banks and Bank Systems, Volume 3, Issue 4, 2008

however, German and Swiss banks assume the top positions in the ranking. The high degree of automation in these countries may help explain this change

in the ranking, in spite of the very low interest margin and the high labor costs.

CIR 2007

81,8% 63,5%

59,3% 45,3%

Iberian

50,6%

48,5%

UK

Greek

Austrian

52,5%

51,1%

Nordic

Irish

55,9%

55,3%

European

Italian

German

68,8%

French

Swiss

… adjusted about differences in interest margin Net interest margin , Deutsche Bank AG (2008) 122,0% 105,5% 93,0%

91,1%

81,0% 76,2% 72,3%

71,9% 64,0%

Iberian

UK

Greek

Austrian

Nordic

Irish

European

Italian

63,5%

62,2%

German

French

Iberian UK Greek Austrian Nordic Irish European Italian German French Swiss

2.01% 3.13% 4.00% 3.58% 1.12% 1.93% 1.60% 2.18% 0.73% 0.52% 0.43%

Swiss

… plus adjustments about differences in labor costs 150,5%

Labor costs index, Eurostat (2006)

130,2%

102,4%

100,6%

91,2% 83,2% 77,5% 73,6%

71,3% 67,9% 63,5%

Iberian

UK

Greek

Austrian

Nordic

Irish

European Italian

German

French

Iberian UK Greek Austrian Nordic Irish European Italian German French Swiss

0.71% 1.20% 0.53% 0.89% 0.83% 0.83% 0.87% 0.84% 1.00% 0.77% 1.09%

Swiss

Source: ProcessLab

Fig. 3. CIR for European banks before and after the adjustment of differences in interest margins and labor costs

90

Banks and Bank Systems, Volume 3, Issue 4, 2008

3.4. Summary concerning the CIR. This section has shown that the CIR is not adequate for measuring productivity and efficiency in banks. The CIR embodies the character of a profit ratio. Price components on the earnings side have an essential impact on the CIR. They distort the predication power of performance concerning the actual production and settlement of bank products and services (Fiorentino, Karmann and Koetter, 2006). Thus, banks operating in countries with high interest margins seem to be highly productive. Yet, a bank’s ability to achieve high prices for its products does not increase its productivity. In order to arrive at an approximate determination of productivity, the CIR needs to be adjusted. Price components have to be eliminated and adjustments have to be made for the respective interest margin and costs of labor in different countries. The results always lead to the same information and exhibit robustness. The procedure shown here is pragmatic and based on publicly available data. However, the adjusted CIR is only suitable for a direct comparison of several banks, i.e. to determine a ranking in terms of better than or worse than another bank. However, an evaluation with respect to measuring and comparing efficiency to the best-practice case cannot be performed by simply comparing CIRs. The missing indicator function of the CIR for the efficiency of banks has already been shown in empirical studies (e.g., Bikker, 1999). 4. Process-based analysis of efficiency

An adjusted CIR provides an indication about valuebased productivity, i.e. how much input was needed to achieve an adjusted income. Yet, even an adjusted CIR cannot replace a well-grounded analysis of efficiency in banks. Real progress can only be made if the efficiency of business processes represents the focus point of the analysis instead of the productivity of a bank as a whole. 4.1. Necessity for a process approach. Business processes are the basis of enterprises’ productivity and efficiency (Hammer and Champy, 1993). They are relatively solid over time; furthermore, they can be compared between different enterprises. Analysis of processes is a key element for evaluating service capability or performance (Kueng, Meier, and Wettstein, 2001).

Although process orientation is widespread in banks, such a mindset is neither fully understood nor applied continuously. So far, only selected processes are measured and controlled. Yet, to assess productivity of banks on the level of processes, a thorough understanding of the banks’ processes, e.g. in the form of a process architecture (Österle, 1995), is essential. Furthermore, a clear understand-

ing of the relevant input and output factors to generate bank products and services is required. Moreover, standards for the definition and accounting of the measurement of quantities are needed. Obviously, banks face challenges and significant requirements for performance measurement on a process level, due to the characteristics of bank products and services and the complex IT architectures that contain a multitude of applications along the process chains. Besides performing productivity analysis, banks should strive for measuring the efficiency on the level of business processes. Simple analyses of productivity are descriptive and should only serve as a starting point. However, empirical studies show that ratios based on a process level are still rarely conducted and only used in the field of business process management (Kueng, Meier, and Wettstein, 2001; Heckl, 2007). In contrast to productivity analysis, the measurement of efficiency features a normative character (Ray, 2004). The goal is to determine a potential increase in performance compared to the best possible case. This comparison to other processes or to the same process of another bank – in terms of benchmarking – enables an assessment of the bank’s own performance. In efficiency measurement, it is essential to consider several factors simultaneously in the analysis. Only this approach ensures the accurate determination of the multi-dimensional character of performance within a business process. In the context of process management, the factors costs, quality, time, and operational risks have to be balanced. These factors must not be analyzed separately from one another, but need to be considered collectively. 4.2. Approaches for a process-based efficiency analysis. Three benchmarking approaches are available to analyze business process efficiency (Figure 4):

(1) Comparison of a bank’s business process with similar processes within the same organization or with other banks. (2) Comparison of bank processes with processes in another industry as far as they possess a comparable structure or comparable activities. The objective of these two approaches is to analyze the one’s own process performance with respect to inefficiency. The analysis reveals opportunities for improvement compared to the best possible execution of a comparable process. The approaches are related to the strategic level of business process management and allow for a comparison of performance in consideration of a different process design. 91

Banks and Bank Systems, Volume 3, Issue 4, 2008

Starting points for the analysis are the input and output factors of a business process. This entails, on the one hand, considering the resources for process execution in terms of working hours or needed IT systems (inputs). On the other hand, the quality of the product or service as well as the adherence to delivery dates and the operational risks, which are involved in the process execution, need to be included (outputs). (3) Comparison of single transactions within a particular business process. This approach aims to identify opportunities for improvement while executing a certain process Approach 1

Approach 2

Bank process A

Bank process A

Costs?

Quality? Time? Operational risks?

Bank process B

(e.g., securities transaction process). The “intrinsic” inefficiency of a process is caused by differences in the execution of single activities within a process chain (Burger and Moormann, 2008). To identify this type of inefficiency, transactions which are cleared and settled within a business process need to be compared to their output. This approach is related to the operational level of business process management. As is the case in the other approaches, this approach of measuring performance requires consideration of the factors of costs, time, quality, and operational risks. Approach 3

Intrinsic inefficiency

Quality? Costs? Time? Operational risks?

TX

TX Time?

Process B* of a different industry

Quality?

Costs? Operational risks?

* with comparable structure

Benchmarking of bank processes

Bank process A

Benchmarking between bank and other industries processes

Benchmarking of transactions within a process

Objective: Identification of optimization Objective: Identification of optimization Objective: Identification of optimization potential in comparison with a best-practice potential in comparison with a best-practice potential in the process execution through a process within the banking industry. comparison of each single transaction with process of a different industry. the best-practice transaction within the same process.

Fig. 4. Approaches for efficiency analysis of banks’ business processes

4.3. Techniques for measuring efficiency. Methods and tools for the above described analyses of efficiency exist. Academic literature offers several measurement techniques which enable benchmarking with simultaneous consideration of different factors.

business process runs completely efficiently if it utilizes the technical possibilities and input factors optimally (technical efficiency), allocates the resources in a best possible manner (allocative efficiency) and produces at an optimal scale (scale efficiency).

The key to efficiency analysis is the identification of a particular production function of the observed process. According to production theory, the production function represents all best possible inputoutput relations and therefore represents the benchmark for a process comparison. The divergence from the production function can be interpreted as inefficiency. The detected inefficiency thus illustrates the opportunity for improvement in comparison to the best possible case (Farrell, 1957).

The techniques to measure efficiency can be separated into two groups – parametric and nonparametric methods (Lovell, 1993). Yet, there is no best possible method to measure efficiency. The choice of the method has to correspond to the problem and the given conditions (Bauer et al., 1998). In order to perform a measurement, parametric methods require a-priori assumptions concerning the development of the production function. Here, the Stochastic Frontier Analysis (SFA) is the most widely applied method. For non-parametric methods, the development of the efficiency line is determined by empirical data. Here, Data Envelopment

Subject to the available data, the existing inefficiency can be separated into different components. A 92

Banks and Bank Systems, Volume 3, Issue 4, 2008

Analysis (DEA) is commonly used (Cooper, Seiford, and Zhu, 2004). Particularly the efficiency analysis of business processes through DEA seems promising, because the measurement can be conducted on the basis of only a few assumptions. The production function is determined by empirically measured data. The method enables “fair” benchmarking, as each observed object (e.g., a transaction) can present itself in the best possible way. DEA helps to identify realistic opportunities for improvement as this method compares each observed object to a similar peer-object or even to a combination of several peer-objects. Furthermore, DEA is very flexible in its application. Comprehensive business processes as well as single transactions can be examined regarding (in-) efficiency. Several discussions on the strengths and weaknesses of this method have been published (Coelli et al., 2005). The approaches to efficiency analysis deliver important information which contributes to a better understanding of performance in banks. As a result of the on-going search for process improvements, methods such as DEA will gain further relevance for the analysis of process efficiency. Conclusion

Measurement of productivity and efficiency in banks is still in its infancy. As illustrated in this article, the traditional cost income ratio is not a suitable ratio to determine productivity. Interest margins as well as labor costs of a country significantly influence the CIR, and therefore this measurement does not appear to be appropriate when analyzing performance in terms of service production and

settlement. This article suggests a procedure based on publicly available data and which enables an approximate evaluation of productivity in banks. The procedure eliminates price components to focus the analysis on the performance. The adjustment of the CIR leads to remarkable changes in the assessment of the productivity in European banks. Supposed productivity advances in various countries disappear. In particular, those banks which currently operate in markets with high interest margins lose top positions in contrast to banks operating in highly competitive markets with low interest margins. High “real” productivity rates of banks serve as an essential starting point for the expected consolidation in the European financial market and the harmonization of margins. Banks that are currently benefiting from high interest margins have to direct their attention to making the necessary improvements in their own service capabilities so that they are able to compensate for the decreasing income. Measuring productivity of a bank at the meta-level, i.e. for the bank as a whole, is not precise enough to develop specific recommendations for process improvement. Instead, a modern analysis of productivity should rather focus on banking processes. For this purpose several requirements have to be met. Methods such as Data Envelopment Analysis deliver a good insight into banking productivity and efficiency. Modern IT tools like Workflow Management Systems generate detailed data and enable the application of new analysis methods. Hence, exciting opportunities emerge for a future-oriented and effective process management in banks.

References

1.

Bauer P.W., A.N. Berger, G.D. Ferrier, D.B. Humphrey. Consistency Conditions for Regulatory Analysis of Financial Institutions: A Comparison of Frontier Efficiency Methods // Journal of Economics and Business, 1998. – №2. – pp. 85-114. 2. Beccalli E., B. Casu, C. Girardone. Efficiency and Stock Performance in European Banking // Journal of Business Finance & Accounting, 2006. – №1/2. – pp. 245-262. 3. Berger A.N., D.B. Humphrey. Efficiency of Financial Institutions: International Survey and Directions for Future Research // European Journal of Operational Research, 1997. – №2. – pp. 175-212. 4. Bikker J.A. Efficiency in the European Banking Industry: An exploratory Analysis to rank countries // De Nederlandsche Bank Working Paper, Amsterdam 1999. – №18. – 24 pp. 5. Burger A. Produktivität und Effizienz in Banken – Terminologie, Methoden und Status quo // Frankfurt School of Finance & Management Working Paper, 2008. – №92. – 96 pp. 6. Burger A., J. Moormann. Detecting Intrinsic Inefficiency on Process Level – Benchmarking of Transactions in Banking // 4th Workshop on Business Process Intelligence (BPI 08), Milano, 01.09.2008. 7. Capgemini, EFMA, ING. World Retail Banking Report 2007 // Capgemini, EFMA, ING Group, Paris 2007. – 64 pp. 8. Charnes A., W.W. Cooper, E. Rhodes. Measuring the efficiency of decision making units // European Journal of Operational Research, 1978. – №4. – pp. 429-444. 9. Coelli T.J., D.S.P. Rao, C.J. O'Donnell, G.E. Battese. An Introduction to Efficiency and Productivity Analysis // 2nd Edition, New York: Springer, 2005. – 366 pp. 10. Cooper, W.W., L.M. Seiford, J. Zhu. Data Envelopment Analysis: History, Models and Interpretations. In: W.W. Cooper, L.M. Seiford, J. Zhu (Eds.): Handbook on Data Envelopment Analysis // Boston: Kluwer Academic, 2004. – pp. 1-39.

93

Banks and Bank Systems, Volume 3, Issue 4, 2008

11. Deutsche Bank AG. European Banks – Running the Numbers: Spring edition. // DB Global Markets Research, London 2008. – 180 pp. 12. European Commission. Financial Services Policy 2005-2010. // European Commission White Paper, Brussels 2005. – 15 pp. 13. European Commission. Interim Report II – Current Accounts and Related Services. // European Commission Competition DG, Brussels 2006. – 193 pp. 14. Eurostat. Verdienstunterschiede zwischen europäischen Ländern und Regionen. Statistik kurz gefasst // Eurostat, Luxembourg 2006. – 12 pp. 15. Farrell M.J. The Measurement of Productive Efficiency // Journal of the Royal Statistical Society, 1957. – №III. – pp. 253-281. 16. Fiorentino E., A. Karmann, M. Koetter. The cost efficiency of German banks: a comparison of SFA and DEA // Deutsche Bundesbank Discussion Paper, Series 2: Banking and Financial Studies 2006/10, Frankfurt 2006. – 40 pp. 17. Hammer M., J. Champy. Re-engineering the Corporation: A Manifesto for Business Revolution // London: Nicholas Brealey, 1993. – 257 pp. 18. Heckl D. Steuerung von Kreditprozessen. Studie // Frankfurt am Main: Bankakademie-Verlag, 2007. – 102 pp. 19. Kueng P., A. Meier, T. Wettstein. Performance Measurement Systems must be engineered // Communications of AIS, 2001. – №1. – pp. 1-27. 20. Lovell C.A.K. Production Frontiers and Productive Efficiency. In: H.O. Fried, C.A.K. Lovell, S.S. Schmidt (Eds.): The Measurement of Productive Efficiency: Techniques and Applications // Oxford: Oxford University Press, 1993. – pp. 3-67. 21. Österle H. Business Engineering. Prozeß- und Systementwicklung I. Entwurfstechniken // 2nd Edition, Berlin et al.: Springer, 1995. – 375 pp. 22. Ray S. Data Envelopment Analysis: Theory and Techniques for Economics and Operational Research // Cambridge: Cambridge University Press, 2004. – 376 pp. 23. Rose P.S., S.C. Hudgins. Bank Management & Financial Services // 6th Edition, Irwin: McGraw-Hill, 2004. – 782 pp. 24. Varmaz A. Rentabilität im Bankensektor // Wiesbaden: Deutscher Universitäts-Verlag, 2006. – 314 pp. 25. Welch P. Measuring efficiency in the finance factory: time for a rethink // Journal of Financial Transformation, 2006. – №2. – pp. 51-60.

94

Banks and Bank Systems, Volume 3, Issue 4, 2008

Jean Perrien (Canada), Raoul Graf (Canada), Fabien Durif (Canada), Lionel Colombel (France)

The role of norms in the evolution of a relationship: the case of an asymmetrical process in the banking industry Abstract

Few attempts have been made to empirically get an understanding of “what should be going on” at each stage of a relationship between two partners of any dyad. This research investigates the evolution of a relationship in a B2B context using the Nominal Group Technique approach to collect the data and puts the emphasis on the actions to be taken. The data coming from account managers in commercial banking are applied to Macneil’s contractual norms to portray this evolution. Results emphasize the importance of role integrity through most of the stages in this evolution. Furthermore, a relationship is seen as a combination of transactional, relational as well as business related factors, which is rarely taken into account in the literature. Thus, relying only on the relational dimension to understand a marketing relationship has to be challenged. Keywords: relationship marketing, norms, integrity. JEL Classification: M31.

1. Modelling the evolution of a relationship♦

The relationship between a buyer and a seller is a potential source of competitive advantage, when: a) the potential for differentiation on the core of the offering is limited, and b) when both partners gain some tangible advantages from a long-lasting relationship (e.g., Day, 2000; Garbarino and Johnson, 1999). But a relationship between the two partners of any dyad is also a dynamic process. As Dwyer, Schurr and Oh (1987) clearly demonstrate, the exchange process underlying a relationship evolves through various stages, from the selection process to an eventual dissolution. Few attempts have been made to empirically understand this evolution and to get an understanding of “what should be going on” at each of these stages. The objective of this paper is to provide a better understanding of this evolution of a relationship between two organizations, by emphasizing actions to be taken. More specifically we focus on the commercial banking arena. Previous empirical researches have justified the selection of this field of investigation in order to explore the relationships between two organizations (e.g., Perrien, Filiatrault and Ricard, 1993). Within this research, and following conceptual stages in the evolution of a relationship as depicted by Dwyer, Schurr and Oh (1987), we identify four main stages in the evolution of a commercial relationship (see Figure 1 in the Appendix). At stage 1 the two potential partners evaluate the feasibility of a banking relationship. We name that phase the selection process (which gathers together phase 1, awareness, and phase 2, exploration, in the Dwyer Schurr and Oh model); phase 2 is identified as the ♦

© Jean Perrien, Raoul Graf, Fabien Durif, Lionel Colombel, 2008.

beginning of a relationship, phase 3 refers to the consolidation of a relationship (what Dwyer Schurr and Oh (1987) mention as phase 4, commitment); and finally, phase 4 refers to the dissolution phase….unfortunately, as many marriages… a strong relationship may end up in a divorce. Our customization of the Dwyer, Schurr and Oh model is reflecting the empirical situation in the banking industry: exploration = awareness + exploration, namely there must be a preliminary formal analysis of the fit between a bank and a commercial customer. Furthermore, our model of the evolution of a relationship is simple and tends to be as exhaustive as the Dwyer, Schurr and Oh model. 2. Macneil’s norms as a frame of reference

Having introduced the stages involved in the development of a relationship and bearing in mind our normative intent, we have to select a framework of analysis. Its contribution is to provide us with a well-grounded and exhaustive conceptualization of the investigated evolution of a relationship to clearly understand “what should be going on” at each stage, as stated earlier. Therefore, we decided to select Macneil’s analysis of contractual norms as a frame of reference (1974, 1978, 1980, 1981, 1983, and 1985). Among the identified published empirical papers (coming from JMR, JM, JBR, IJSM, and IJBM) on IBI and Emerald databases from 1980 to 2007, which made explicit references to Macneil’s works, few of them took into account the full set of Macneil’s norms: marketers tend to follow a piecemeal approach which significantly departs from Macneil’s contribution. As an indication, the average number of contractual norms which have been included in published empirical researches (explicitly referring to Macneil) is 2.72 when, as we will see, the author identifies many more contractual norms. 95

Banks and Bank Systems, Volume 3, Issue 4, 2008

3. Macneil’s contractual norms

First, let us stress that Macneil is a jurist and that his main objective was to improve the paradigm of contract law. In his straightforward definition of modern contract law Macneil emphasizes the notions of exchange and relationship between parties, a decade ahead of marketers (e.g., Berry and Thompson, 1982). But being ahead, from a conceptual standpoint, is not enough. One has to acknowledge that the very notion of a contractual exchange gains additional value when Macneil defines the critical features of what contractual norms are, and moreover, breakdowns these norms according to their contributions to the contract (i.e. the exchange). Macneil is the very first to introduce transactional and relational norms. Up to that moment, a contract as well as an exchange had been defined on purely transactional norms (as in transaction cost analysis and classical contract law). By widening the scope of the commitment between two partners, Macneil significantly improves our knowledge of a relationship. Hence, his work became the foundations of relationship marketing (Dwyer, Schurr and Oh, 1987), although in subsequent work marketing researchers often refer to Macneil without always depicting a clear understanding of his contribution. 4. Contractual norms: a typology

When referring to norms, Macneil gives us the basic parameters on the various forms contracts, and therefore exchanges between partners may take place. A norm is a component of an exchange and consequently contributes to the implementation of an effective relationship. These norms are therefore basic social and organizational ways on which to implement an effective relationship. Hence, they give us indications of what should be taken into account in the evolution of a relationship. According to Macneil, there are 10 common contract norms. Five norms are transactional: reciprocity, implementation of planning, effectuation of consent, linking norms, creation and restraint of power. Five norms are relational: role integrity, preservation of the relation (including contractual solidarity), harmonization of relational conflict (including flexibility), supracontractual norms and propriety of means. In the marketing literature several researchers include another relational norm, namely communication that Macneil perceives as a component of role integrity (e.g., Morgan and Hunt, 1994; Paulin, Perrien and Fergusson, 1997). The underlying hypothesis to include an additional norm such as communication is straightforward: communication between partners, from a marketing standpoint, is a crucial issue in the development of a relationship to such an extent it must be viewed as a specific rela96

tional norm that can be leveraged by marketing actions to strengthen a relationship. Table 1 (see the Appendix) gives a definition of the 11 norms we analyzed as well as examples of items related to each norm. Measurement issues are not a concern for Macneil and some norms are fairly broad in scope. Nevertheless, they provide us with a unique and fairly exhaustive grid to understand a relationship (Graf and Perrien, 2005; Arnold and Joshi, 1997; Dwyer, Schurr and Oh, 1987). As we think, it is the most advanced framework to have been built up to portray a relationship, we intend to apply this grid to understand the evolution of a relationship in the commercial banking arena. 5. Methodology

To understand what actions have to be conducted at each step of the evolution of a relationship (see Figure 1 in the Appendix), we rely on the expertise of bank account managers, which means information is coming from the supply side of the relationship process. Such an investigation is backed up by research in the commercial banking arena depicting a commercial banking relationship as an asymmetrical process: the responsibilities involved in managing relationships rest in the hands of the banker (Ricard and Perrien, 1993). Furthermore, such an approach fits into the stream of managerial research such as in the development process of innovation, following classical measures of “decision calculus”. We also have to acknowledge that sellers, through the expertise they gain by managing diversified relationships, have a more accurate perspective of what should be done simply because they gather together customers’ expectations and managerial conditions of effective performance. Bank account managers are in charge of a portfolio of commercial accounts (i.e. businesses) and their key responsibility is to manage the relationships with commercial customers to such an extent that several Canadian institutions no longer speak of account managers but of relationship managers. Nevertheless, the common practice is still to refer to account managers. For each of the four stages in the development of a relationship we used the Nominal Group Technique approach (Claxton, Ritchie and Zaichkowsky, 1980) the objective of which is to generate the maximum number of ideas. This data collection technique has never been used before as far as the analysis of Macneil’s norms is concerned (Durif and Perrien, 2006). In a first step, participants were presented with a simple question customized to each of the four stages that were introduced earlier. The format of ques-

Banks and Bank Systems, Volume 3, Issue 4, 2008

tions was standardized: “In your opinion, how should bank identify a good customer to build up a long-term relation (relational approach)?” (Stage 1: the selection process); “In your opinion, how can banks build up a long-term relation with a good new customer (relational approach)?” (Stage 2: the beginning); “In your opinion, how can banks improve long-term relations with good customers (relational approach)?” (Stage 3: the consolidation), “In your opinion, how do you explain banks are loosing what were good long-term customers?” (Stage 4: the dissolution). Obviously, only the structure of the last question (regarding the dissolution process) was slightly different because of the very nature of the stage. Each respondent was only involved in one stage of the development of a relationship.

conducted by two independent researchers having a good understanding of both commercial banking and Macneil’s contractual norms. A preliminary analysis resulted in the addition of three dimensions at the first stage (selection): bank’s characteristics (e.g., Credit policies), customer’s profile (e.g., Management style) and business sector (e.g., Potential for growth). This inductive process resulted in a grid of 13 dimensions through which items extracted from the NGT process were analyzed. The reliability of the content analysis as measured by the Perreault and Leigh coefficient was 97% (an excellent score), only 15 items were misclassified by the two judges. The final agreement came from a consensus between the two judges.

5.1. The NGT technique. Once they had been introduced with the question, the NGT approach participants went through six steps (see Table 2 in the Appendix). The results from this final vote were computed as input data for our analysis.

The following paragraphs present an overall evaluation of the evolution of a relationship in commercial banking as extracted from both the deductive (Macneil’s norms) and inductive (business characteristics) grid we introduced earlier. The results are depicted in Table 3 of the Appendix.

5.2. Participants. 239 account managers coming from the six major Canadian banks as well as the leading foreign bank (HSBC) were involved in the NGT sessions. All sessions were managed by the same moderator. Participants took part in meetings held by the Institute of Canadian Bankers in Western Canada, Ontario and Montreal. Sessions were conducted in English, as French speaking account managers had a good monitoring of English (the moderator was bilingual in order to solve any potential linguistic problem). All in all, it meant that we had an accumulated experience of slightly less than 20000 relationships (at the time of the investigation, the average size of the portfolio was 80 accounts per manager). We did not collect information on profiles of account managers (i.e. education, number of years of experience, employer) as this information was part of the confidential agreement we had with the ICB.

The breakdown of participants with regards to the four stages of the development of a relationship was as follows: 58 for Stage one (“the selection”), 70 for Stage 2 (“the beginning”), 51 for Stage 3 (“the consolidation”) and 50 for Stage 4 (“the dissolution”). Hence the overall sample consisted of 229 account managers from the top seven Canadian banks. Discrepancies between stages were due to the level of account managers’ participation in ICB meetings. Groups of respondents were randomly assigned to each stage. A group consisted of 12 to 15 participants. 6. Content analysis and reliability of results

For each stage of the evolution process, a content analysis of items resulting from NGT sessions was

7. Results

The selection process is mainly the outcome of a fit between the two members of the dyad with a strong emphasis on the customer’s economic and managerial characteristics such as the financial position of the firm and its management. Relational factors rank second with the ability to build a long-term relationship (solidarity) as the leading norm, closely followed by role integrity which mainly concerns the understanding of the potential customer and the professionalism with which to handle his needs. At this stage of the process transactional norms are paramount; although it must be emphasized they are less important than business and relational norms. At the beginning of the relationship, relational factors lead the process, especially role integrity (professionalism, knowledge). Business issues are no longer playing a role, which may mean that before a relationship is set up, the business analysis is “final”. Let us also stress the key role of communication between members of the dyad at this early stage (communication is a sub-dimension of role integrity). Transactional issues decrease through the evolution of the relationship. It means that building up a relationship (stage 2) mainly entails role integrity and communication between partners. The consolidation of a relationship is, by far, a question of role integrity which explains over 50% of this consolidation process. At the same time, effectuation of consent (delivering promises), a transactional norm, impacts the consolidation stage. Hence consolidation is not only a matter of relational skills, it becomes also a matter of transaction 97

Banks and Bank Systems, Volume 3, Issue 4, 2008

handling. It is at this stage that communication as well as solidarity are the most effective through the various stages in the evolution of a relationship although well beyond role integrity and communication. The dissolution of a relationship is the stage which is the least explained by our framework (see the scores) although once more, failures in role integrity play a critical issue. Bearing in mind the fact that the question on the dissolution process explicitly referred to a “good” customer one could easily suspect that external factors to the relationship may explain the dissolution (i.e. competitive pressures). The only available investigation on dissolution is clear proof standing of the above stated argument: competitive pressures are the leading defection purpose (Perrien, Banting and Paradis, 1985). Role integrity is, by far, the leading relational norm which means that suppliers need to act with professionalism and expertise to build an effective relationship. Role integrity is the leading explaining factor at the development stage of a relationship. And is also an explaining factor of a divorce between the two partners. In Table 1 we define role integrity as “underlying actions and attitudes that favor more intimate relationships”, from a more pragmatically standpoint it means role integrity encompasses: 1 – knowledge of customers, their organizations and business environments, 2 – the supplier (the organization as well as front line people) skills, 3 – the supplier ability to meet its responsibilities, 4 – the supplier coordination capability. These four dimensions of role integrity are extracted from items generated through our NGT sessions. Table 4 (see the Appendix) presents some of the major items we identify as dimensions of role integrity through the various stages of the evolution of a relationship. Because the total number of items illustrating this norm is large, we only keep the most significant ones (based on final of scores.) Interestingly enough, in the review of published marketing researches, only 8 of them measured role integrity, whereas solidarity is, by far, the leading norm, especially if we include the body of researches conducted on trust as a sub-component of solidarity. Conclusion and recommendations

A relationship is not – and will never be – the sole outcome of relational norms. Although marketers tend to emphasize, if not over-emphasize, relational dimensions in the effectiveness of a dyad, they have to acknowledge that, several stages in the development of an effective relationship cannot be restricted to relational norms. Actually early stages of a rela98

tionship are significantly influenced by nonrelational norms. There is always, through the evolution of a relation, a mix of relational-transactional and business related factors, although relational norms could be prevalent. As far as the banking industry is concerned, a relation can only start if there is some kind of economic fit between partners…which is well beyond the scope of a relational exchange: the question is not “Do I fit with you priorities?” but “Do you meet my expectations?”… This simply means that a relational strategy at its early stages is mainly supplier driven. From our investigation, role integrity is the leading relational norm. As mentioned by Durif and Perrien (2006), recent studies published regarding Macneil’s Norms clearly highlight that this norm is typical of the phenomenon of interaction in business relationships (Kaufmann and Stern, 1988). It also has considerable impacts on: trust, satisfaction and quality in these relationships and could even represent a competing advantage for organizations (Dant and Schul, 1992). Furthermore, it also influences the perception of ethical and unethical behaviors of individuals in exchange relationships (Pelton, Chowdhury and Vitell, 1999). In this study, we propose a four dimensional definition of role integrity as extracted from items generated by managers. No doubt that to improve our understanding of relationships between buyers and suppliers more attention should be devoted to this norm. Whether in terms of management or measurement, role integrity deserves more attention. Our data are coming from account managers in the banking industry. Although we think that through their experiences they gain a good knowledge of how to develop an effective relationship, their “supplier” vision remains a limit. Indeed, even if a marketing relationship is an asymmetrical process where the main responsibility depends on suppliers, it has to be pointed out that it does not reflect a perfect vision of what should be done to implement and develop effective relationship with customers. Longitudinal investigations involving both members of the dyad should provide a more accurate picture. Unfortunately, it also means, in the realm of relationship marketing, an investigation which should start from the selection process to a possible dissolution…a story which may spread over many years if not many decades, especially in B2B. Meanwhile, contributions based on knowledge of experts may still provide some valuable information. It is the objective this paper aims to achieve.

Banks and Bank Systems, Volume 3, Issue 4, 2008

References

1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21.

22.

Berry, L.L, Thompson, T.W. (1982). “Relationship Banking: the Art of Turning Customers into Clients”. Journal of Retail Banking, Vol. 4, n. 2, pp. 64-74. Claxton, J.D., Ritchie, J.R.B., Zaichkowsky, J. (1980). “The Nominal Group Technique: its Potential for Consumer Research”. Journal of Consumer Research, Vol. 7, n. 3, pp. 308. Day, G.S. (2000). “Managing Market Relationships”. Journal of the Academy of Marketing Science, Vol. 28, n. 1, pp. 24-30. Dant, R.P., Schul, P.L. (1992). “Conflict Resolution Processes in Contractual Channels of Distribution”. Journal of Marketing, 56, 38-54. Durif, F., Perrien, J. (2006). “Can Cognitive Mapping Enrich our Understanding of Macneil’s Contractual Norms? The Specific Case of Role integrity”, Acte de colloque, 14ème colloque de l’ICRM (International Colloquium in Relationship Marketing), 2006 (Leipzig). Durif, F., Ricard, L., Perrien, J. (2006). “The Underestimated Importance of Macneil’s Social Contract Theory in the Field of Relationship Marketing”, 14ème colloque de l’ICRM (International Colloquium in Relationship Marketing), 2006 (Leipzig). Dwyer, F.R., Schurr, P.H., Oh, S. (1987). “Developing Buyer-Seller Relationships”. Journal of Marketing, Vol. 51, April, pp. 11-27. Garbarino, E., Johnson, M.S. (1999). “The Different Roles of Satisfaction, Trust, and Commitment in Customer Relationships”. Journal of Marketing, Vol. 63, April, pp. 70-87. Graf, R., Perrien , J. (2005). “The Role of Trust and Satisfaction in a Relationship: the Case of High Tech Firms and Banks”, 34th EMAC Conference, Milan, 2004. Joshi, A.W., Arnold, S. (1997). “The Impact of Buyer Dependence on Buyer Opportunism in Buyer-Supplier Relationships: the Moderating Role of Relational Norms”. Psychology & Marketing, Vol. 14, n. 8, pp. 823-845. Macneil, I.R (1974). “The Many Futures of Contracts”. Southern California Law Review, Vol. 47, pp. 691-816. Macneil, I.R. (1978). “Contracts Adjustments of Long Term Economic relations under Classical, Neo-Classical and Relational Contract Law”. Northwestern University Law Review, Vol. 72, n. 6. pp. 854-905. Macneil, I.R. (1980). “The New Social Contract: An Inquiry into Modern Contractual Relations”. New Haven, CT: Yale University Press. Macneil, I.R. (1983). “Values in Contract: Internal and External”. Nothwestern University Law Review, Vol. 78, pp. 340-418. Macneil, I.R. (1985). “Relational Contract: What Can We Do and Do Not Know”. Wiisconsin Law Review, pp. 483-525. Morgan, R.M., Hunt, S.D. (1994). “The Commitment-Trust Theory of Relationship Marketing”. Journal of Marketing, Vol. 58, July, pp. 20-38. Paulin, M., Perrien, J., Ferguson, R. (1997). “Relational Contract Norms and the Effectiveness of Commercial Banking Relationships”. International Journal of Service Industry Management, Vol. 8, n. 5, pp. 435-452. Pelton, L.E., Chowdhury, J., Vitell, S.J.Jr. (1999). “A Framework for the Examination of Relational Ethics: An Interactionnist Perspective”. Journal of Business Ethics, 19 (3), pp. 241-253. Perrien, J., Filiatrault, P., Ricard, L. (1993). “The Implementation of Relationship Marketing in Commercial Banking”. Industrial Marketing Management, Vol. 22 pp. 141-148. Perrien, J., Graf, R., Colombel, L. (2005). “Le rôle des normes dans l’évolution d’une relation”. Cahier de recherches, ESG-UQÀM, 18 pages. Perrien, J., Ricard, L. (1995). “The Meaning of a Marketing Relationship: a Pilot Study”. Industrial Marketing Management, Vol. 24, n. 1, pp. 37-43. Perrien, J., Paradis, S., Banting, P.M. (1995). “Dissolution of a Relationship: the Salesforce Perception”. Industrial Marketing Management, Vol. 24, n. 4, pp. 317-318.

Appendix Stage theselection selection Stage 1:1:the (Can we in ain a (Can wecommit commitourselves ourselves relationship?) relationship?)

Stage thebeginning beginning Stage 2:2:the forfor anan enduring (Building the thefoundations foundations enduring relationship) relationship)

Stage theconsolidation consolidation Stage 3:3:the (The institutionalizationofofa arelationship) (The institutionalization relationship

Stage thedissolution dissolution Stage 4:4:the (Unfortunately, nolonger longerworks...) works...) (Unfortunately, ititno

Fig. 1. The evolution of a relationship: the stages

99

Banks and Bank Systems, Volume 3, Issue 4, 2008

Table 1. Macneil’s contractual norms RECIPROCITY

NORM

CHARACTERISTICS Bilateral improvement of the situation of both partners

IMPLEMENTATION OF PLANNING

Specifications of conditions Actions take place on due time

EFFECTUATION OF CONSENT

Meeting contractual objectives

RESTITUTION, RELIANCE AND EXPECTATIONS INTERESTS CREATION AND RESTRAINT OF POWER ROLE INTEGRITY

Clearly stated obligations and conditions

ITEMS Desire for reciprocal relation Total relationship profits Schedule visits Timely delivery Established deadlines Achieving objectives Do it right Avoidance of errors Guaranties

A fair balance of power between buyer and seller

Customer’s position as a centre of influence Client’s competitive position on the market Good knowledge of industry Clear understanding of customers’ needs Develop long-term strategies Stability of account manager

Underlying actions and attitudes that favor more intimate relationships Finding a common ground of accord Continuity of the relationship

PRESERVATION OF THE RELATION (CONTRACTUAL SOLIDARITY) HARMONIZATION OF RELATIONAL CONFLICT (FLEXIBILITY) SUPRA CONTRACTUAL NORMS

Meeting social and political rules (justice…)

COMMUNICATION

Effectiveness and timeliness of exchange of information

Capacity to adapt the contracts depending on the clients Offer a wide variety of product needs Have a flexible offer Ethical selling Involvement in the protection of the environment Rapid response Regular customers contact Make information available

Table 2. Steps used in the NGT 1.

Generation of ideas: after a clarification of the question presented to the group, participants were asked to write their individual responses (ideas) to the question.

2.

Round robin recording of ideas: On an individual basis, each manager was requested to verbally present his or her response to the group. Once the answer had been given, we proceeded to interrogate the next manager of the group in sequence. The “round robin” procedure continued until all the individual ideas had been exhausted.

3.

Discussion for clarification purposes: The objective was to ensure that all participants interpreted the responses in the same manner and to remove any possible duplication of responses.

4.

Preliminary vote: On an individual basis, managers were asked to select eight responses they judged to be the most meaningful with regards to the question (a screening process).

5.

Discussion on the preliminary vote: The purpose of the discussion was to have a general understanding of the vote in the preceding step.

6.

Final vote: Individually, each manager was asked to retain only five statements from the preliminary vote and to assign a mark to each, based on a ten point scale.

Table 3. Results STAGE 1 NORM

Role integrity Preservation of the relation (Solidarity) Harmonization of relational conflict (Flexibility) Supra contractual norms Communication Reciprocity Implementation of planning Effectuation of consent Reliance Power Bank characteristics Customer’s characteristics Business sector Total Relational (1) Total Transactional (1) Total Business (1)

100

Relational, Transactional Business R R R R R T T T T T B B B

STAGE 2

STAGE 3

STAGE 4

Frequency

Score

Frequency

Score

Frequency

Score

Frequency

Score

7 10 3

209 293 194

15 7 2

851 224 54

41 11 6

1630 401 114

23 10 0

496 272 0

5 3 2 0 4 1 2 2 24 5

44 50 48 0 179 51 17 63 696 123 790 358 882

1 16 0 5 4 0 0 0 0 0

39 567 0 134 176 0 0 0 0 0 1735 310

0 21 0 1 5 0 0 0 0 0

0 723 0 38 245 0 0 0 0 0 2868 283 0

0 6 0 0 4 0 0 0 3 0

140 0 0 56 0 0 0 50 0 908 106 50

Banks and Bank Systems, Volume 3, Issue 4, 2008

Table 4. Role integrity depending on the stage of the relationship 1.

The selection ¾ ¾ ¾ ¾ ¾

2.

The beginning ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾

3.

Good knowledge of the industry Knowledge of the competition Account manager motivation Knowledge of the business industry Clear understanding of relationship needs

Well trained staff Provide efficient consistent service Have knowledgeable staff so clients feel comfortable when you talk to them Treat the client with respect/appreciate their business Efficient and accurate service Act professionally on first meeting Knowledge of industry Account manager experience Learn about customer business Understand customer expectation Identify customers needs and goals Product/service knowledge

The consolidation ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾ ¾

Adequate training of personnel Define the type of relationship wished for Employees should know the clients Friendly personnel Meet the clients behind closed doors Better product knowledge for employees Provide high touch personal service Get it right the first time Know top customers well Upgrade account manager skills Become more familiar with the industry Know the company better Know the customers expectations Empathetic and responsive to the customers needs Anticipate customers needs Get to know the customers market Be professional Be proactive instead of reactive

4. The dissolution ¾ ¾ ¾ ¾ ¾ ¾

Inadequate knowledge of needs for the client industry Poor/lack of negotiating skills Lack of attention to customers Inadequate training of bank staff Lack of banker understanding of customers needs Account manager doesn’t understand clients business

101

Banks and Bank Systems, Volume 3, Issue 4, 2008

Authors of the issue Hai-Chin Yu

− Professor of Finance, Department of Intl Business, Chung Yuan University, Taiwan, Department of Finance, Rutgers University (USA)

Ken H. Johnson

− Assist. Prof., Department of Finance and Real Estate, Florida International University (USA)

Der-Tzon Hsieh

− Prof., Department of Economics, National Taiwan University (Taiwan)

Kuan-Min Wang

− Ph.D., Associate Professor of Finance, Overseas Chinese Institute of Technology (Taiwan)

Thanh-Binh Nguyen Thi

− Ph.D., Assistant Professor of Accounting, Chaoyang University of Technology (Taiwan)

Shu-Hui Wu

− Ph.D. program in Graduate Institute of International Economics, National Chung Cheng University (Taiwan)

Alexandra Lai

− , Bank of Canada (Canada)

Raphael Solomon

− Bank (Canada)

Chung-Hua Shen

− Department of Finance, National Taiwan University (Taiwan)

Meng-Fen Hsieh

− Ph.D., Department of Finance, National Taichung Institute of Technology (Taiwan)

Chien-Chiang Lee

− Department of Applied Economics, National Chung Hsing University (Taiwan)

John Mylonakis

− Tutor, Hellenic Open University (Greece)

Seok Weon Lee

− Assoc. Prof., Division of International Studies, Ewha Womans University (Korea)

Christos Floros

− Dr., Senior Lecturer in Banking and Finance, Department of Economics, University of Portsmouth (UK)

Georgia Giordani

− Ph.D. candidate in Banking, Department of Economics, University of Portsmouth (UK)

Izah Mohd Tahir

− Universiti Darul Iman Malaysia (Malaysia)

Nor Mazlina Abu Bakar

− Universiti Darul Iman Malaysia (Malaysia)

Sudin Haron

− Kuala Lumpur Business School (Malaysia)

Wolfgang Benner

− Dr., University of Goettingen (Germany)

Lyudmil Zyapkov

− Dr., University of Goettingen (Germany)

Andreas Burger

− MBE, Senior Advisor Business Development, Central Service Depart-

102

of

Canada

Banks and Bank Systems, Volume 3, Issue 4, 2008

ment Transaction Banking Markets, Commerzbank AG (Germany) Juergen Moormann

− Ph.D., Professor of Banking and Director of ProcessLab, Frankfurt School of Finance & Management (Germany)

Jean Perrien

− Assoc. Prof., Marketing Department, ESG-UQAM (Canada)

Raoul Graf

− Assist. Prof., Marketing Department, Faculty of Business Administration, University Laval (Canada)

Fabien Durif

− Assist. Prof., Faculty of Business Administration, University of Sherbrooke (Canada)

Lionel Colombel

− Assist. Prof., Faculté des Sciences Economiques et de Gestion, Université d’Auvergne (France)

103

Banks and Bank Systems, Volume 3, Issue 4, 2008

Submission guidelines for authors The cover page of a manuscript should contain the title and name(s) of the author(s). The author’s name, degree, position and the place of work as well as contact details (phone number, job or/and personal e-mail) should be provided at the bottom of this page. 1. Abstract preparation guidelines

1.1. The abstract (150-200 words) should reflect the conceptual content of the article. 1.2. Journal of Economic Literature (JEL) classifications are necessary. 2. The paper main body preparation guidelines

2.1. The paper should present the result of independent original research, undertaken by the author; it also should contain the data never published before. 2.2. The paper should contain a clear description of research objective and its subject. 2.3. The methodology of research should be described in detail. 2.4. The author’s personal scientific contribution must be grounded in the paper. 2.5. The paper should contain basic suggestions on how to solve the problem under study. 3. References in the text

3.1. References in the text are made as follows: (Myers, 2000), the former being name of the author, the latter edition year. 3.2. Examples of references: Alchian A., S. Woodward. Reflections on the Theory of the Firm // Journal of Institutional and Theoretical Economics, 1987. – №143. – pp. 110-136. Berle A.A., G.C. Means. The Modern Corporation and Private Property. – New York: Macmillan, 1932. – 418 рр. 4. Manuscript length

4.1. The paper should not be less than 2000 words and should not exceed 6000 words. 5. Submission guidelines

Please send your paper using one of the formats listed below: 5.1. A soft copy as an MS Word .doc file (all versions accepted) or as .rtf file to: Journal “Banks and Bank Systems”: Executive Editor (Lyudmyla Kozmenko): [email protected] 5.2. A hard copy supplied with a disc containing a soft copy of the paper to: Publishing company “Business Perspectives” Dzerzhynsky lane, 10 Sumy 40022 Ukraine 6. Reviewing process

All papers are refereed by the international competent researchers using a “double-blind” review which is the best practice in papers reviewing. 7. Acceptance fee

We offer a very democratic fee policy to our contributors. We only ask for payment from those authors whose papers have already been reviewed and accepted for publication in the journal.

104

Banks and Bank Systems, Volume 3, Issue 4, 2008

A joint subscription form 2009 Follow guidelines to complete it. Journal “PROBLEMS AND PERSPECTIVES IN MANAGEMENT”: For institutional subscribers

USD

Print Online Print version + online Full back issues online

545 350 650 560

For individual subscribers

Print version Online Print version + online Full back issues online

USD 230 200 350 250

Journal “INVESTMENT MANAGEMENT AND FINANCIAL INNOVATIONS”: For institutional subscribers

USD

Print version Online Print version + online Full back issues online

525 340 630 480

For individual subscribers

USD

Print version Online Print version + online Full back issues online

225 190 330 260

For individual subscribers

USD

Journal “INNOVATIVE MARKETING”: For institutional subscribers

USD

Print version Online Print version + online Full back issues online

485 315 580 420

Print version Online Print version + online Full back issues online

215 180 310 220

Journal “BANKS AND BANK SYSTEMS”: For institutional subscribers

USD

Print version Online Print version + online Full back issues online

700 455 840 390

For individual subscribers

Print version Online Print version + online Full back issues online

USD 240 190 330 210

Underline one of the payment methods you prefer, and write the amount to pay (if you prefer, you can pay by one check/bank transfer to subscribe to all journals): 1. I enclose a check for USD / EURO _____________; 2. Send me an invoice for USD / EURO ____________. Write your contact details here: Name __________________________________________________________________________ Institution ______________________________________________________________________ Address ________________________________________________________________________ E-mail __________________________________ Tel ___________________________________ ____________________________________________________________ Signature Please, send this form (with the check if you prefer to pay by check) at: Ms. Lyudmyla Kozmenko Publishing company “Business Perspectives” Dzerzhynsky lane, 10 Sumy, 40022 Ukraine E-mail: [email protected] 105