DEREGULATION, CONSOLIDATION AND BANKS EFFICIENCY IN SINGAPORE: EVIDENCE FROM EVENT STUDY WINDOW APPROACH AND TOBIT ANALYSIS

Int. Rev. Econ. (2007) 54:261 283 DOI 10.1007/s12232-007-0017-2 DEREGULATION, CONSOLIDATION AND BANKS EFFICIENCY IN SINGAPORE: EVIDENCE FROM EVENT STUDY WINDOW APPROACH AND TOBIT ANALYSIS FADZLAN SUFIAN and MUHAMED-ZULKHIBRI ABDUL MAJID Springer-Verlag 2007 Abstract: A three-year window analysis together with the Data Envelopment Analysis (DEA) approach is employed to investigate the effects of mergers and acquisitions on the Singapore banking groups efficiency. The results suggest that the merger has resulted in a higher Singapore banking groups mean overall efficiency. We do not find evidence of more efficient acquirers compared to the targets and that the acquiring banks mean overall efficiency tends to improve from the merger with a more efficient bank. The Tobit regression results suggest that bank profitability has positive impact on bank efficiency, whereas poor loan quality has negative influence on bank performance. (JEL: G21, D24) Keywords: bank mergers, Data Envelopment Analysis, Singapore, Tobit model 1. Introduction Examining banking performance has been a common practice among banking and finance researchers for a number of years. The main reason for Planning and Research Department (BCB), Bumiputra-Commerce Bank Berhad and Department of Banking and Finance, University of Malaya, Kuala Lumpur (Malaysia). E-mail: fadzlan.sufian@cimb.com Central Bank of Malaysia and Monetary and Financial Policy Department, Bank Negara Malaysia, Kuala Lumpur (Malaysia). E-mail: muhamedz@bnm.gov.my All findings, interpretations, and conclusions are solely those of the authors and do not necessarily represent the views of the institutions to which they belong. We would like to thank the anonymous referees for their comments and suggestions. The remaining errors are of our own.

262 F. Sufian, M.-Z. Abdul Majid continued interest in this area of research is the ever-changing banking business environment throughout the world. Many countries that adopted financial deregulation policies are now experiencing aggressive banking practices. Singapore is no exception and is becoming a competitive and important market for financial as well as for other products. Singapore banking is a considerable component in Asian financial activities, and has not been subjected to substantial research as compared to other countries in the developed world. As efficient banking systems contribute in an extensive way to higher economic growth in any country, studies of this nature are very important for policy makers, industry leaders and other agents reliant on the banking sector. The analysis of bank efficiency continues to be important from both the microeconomic and macroeconomic points of view as is documented by its long tradition in the literature 1. From the microeconomic perspective, the issue of bank efficiency is crucial, given increasing competition and measures to further liberalize the banking system. This renders the issue of increasing the efficiency as one of the main priorities of the regulators towards the sector. From the macroeconomic perspective, the efficiency of the banking sector influences the costs of financial intermediation and the overall stability of the financial markets. The motivation of this study comes firstly from the fact that despite the importance of the Singapore banking sector to the domestic, regional, and international economy, there are only a few microeconomic studies performed in this area of research. The present study thus addresses an important gap in the literature. Secondly, in order to appraise the effectiveness and success of mergers and acquisitions activities among the domestic incorporated Singapore commercial banks, it is therefore essential to conduct a formal analysis. This study thus attempts to provide empirical evidence on the efficiency changes of Singapore commercial banks arising from mergers and acquisitions over the past decade. Utilizing the nonparametric Data Envelopment Analysis (DEA) methodology, the overall pure technical and scale efficiency of all domestic incorporated Singapore commercial banks that were involved in mergers and acquisitions will be investigated. The role of mergers in efficiency changes will be examined by comparing relative efficiency scores before and after the merger program. To the best of our knowledge, this will be the first study in the literature to examine this important issue within the context of the Singapore banking sector. 1 For an excellent overview, see Berger et al. (1993), Berger and Humphrey (1997).

Banks efficiency in Singapore 263 The paper raises four important fundamental questions: 1) Did mergers and acquisitions result in the improvement of the mean overall efficiency levels of the Singapore banking system post-merger? 2) Did a less efficient bank become the target for acquisition? 3) Did a less (more) efficient target result in the deterioration (acceleration) in the acquirer s mean overall efficiency level post-merger? 4) What determines the relative performance of banks in Singapore? The paper is structured as follows: the next section gives a brief overview of the Singapore banking system. Section 3 reviews the main literature regarding bank efficiency. Section 4 outlines the approaches to the measurement and estimation of efficiency change and the Tobit regression model. Section 5 discusses the results, and finally, Section 6 provides some concluding remarks. 2. Brief Overview of the Singapore Banking System The development of Singapore as a financial centre was the result of deliberate government policy to broaden the country s economic base in the 1970s. With the introduction of the Monetary Authority of Singapore (MAS) in 1970, the government introduced fiscal incentives, removed exchange controls, and encouraged competition to spur the financial sector development. Supported by its sound macroeconomic fundamentals and prudent policies, today Singapore ranks among the leading international financial centres, and is one of the key centres in Asia. Singapore lags only behind London, New York and Tokyo in foreign exchange trading. Growth in the financial services sector has contributed significantly to its economic development, which today accounts for approximately 13% to 15% of its GDP. This is evidenced by the presence of a wide network of financial institutions providing a range of services that facilitate domestic, regional, and international flow of funds for trade and investments. The Singapore domestic banking sector was closely regulated and largely protected until the later half of the 1990s. The entry of foreign banks had been restricted to the wholesale banking markets since 1971. While the locally incorporated banks are given permission to expand their branch networks, foreign incorporated full licensed banks admitted prior to 1971 are subjected to restrictions in terms of opening up new branches and re-locating existing branches. As such, locally incorporated banks are relatively sheltered from foreign competition. The result is a banking industry with many international players and domestically incorporated commercial banks

264 F. Sufian, M.-Z. Abdul Majid dominating the local banking market. During the Asian Financial Crisis of 1997-1998, Singapore sound economic and financial fundamentals have enabled the sector to weather the crisis relatively well. Despite the losses from defaulted loans, which occurred during the crisis, Singapore commercial banks were adequately capitalized and insolvency was not an issue. Nonetheless, the financial turmoil in local financial institutions has called for strong incentives for banks to merge, which would create large institutions able to cope with international competition. 2.1. Mergers and acquisitions in the Singapore banking sector. - A regional financial centre can be defined as a central location, where there is a high concentration of financial institutions and capital markets that allow financial transactions in the region to take place efficiently. The Singapore government has been actively undertaking financial liberalization and reforms since the 1960s. As a result of its endeavours, Singapore has facilitated greater financial intermediation in the South East Asian region, contributing to the development of capital markets, cross border trade and business investment. Rather than becoming more inward looking, as did some countries affected by the crisis, Singapore hastened financial liberalization in order to create a more resilient financial sector, which could compete in an increasingly globalized environment. The liberalization has involved strengthening domestic banks through consolidation and increasing foreign participation in the financial sector. Since 1998, when Development Bank of Singapore (DBS) acquired the Post Office Savings Bank (POSB) and Keppel Bank merged with Tat Lee Bank, the Singapore government has been encouraging domestic banks to consolidate to prepare them for stiffer competition from foreign banks. In fact, for Singapore banks to compete successfully in the new era of globalization, the government intended to merge the domestic financial institutions into two super banks.

Banks efficiency in Singapore 265 The recent mergers and acquisitions (Ms & As) activities among domestic incorporated Singapore banks were 2 : On June 12, 2001, Singapore s third largest bank, Overseas-Chinese Banking Corporation (OCBC) announced a S$4.8 billion bid (voluntary general offer) for Keppel Capital Holdings (KCH), which owns Singapore s smallest bank, Keppel Tat Lee Bank. On June 29, 2001 Singapore s second largest lender, United Overseas Bank (UOB) made a competing bid for Overseas Union Bank (OUB), Singapore s fourth largest bank, after DBS Holdings Group s unsolicited bid of S$9.4 billion for OUB. UOB s bid succeeded in August 2001 forming Singapore s largest bank in terms of assets. 3. Related Studies Bank mergers and acquisitions may enable banking firms to benefit from new business opportunities created by changes in the regulatory and technological environment. Berger et al. (1999, p. 136) pointed to the consequences of mergers and acquisitions, which may lead to changes in efficiency, market power, economies of scale and scope, availability of services to small customers and payments systems efficiency. Besides improvement in cost and profit efficiency, mergers and acquisitions could also lead banks to earn higher profits through leveraging loans and deposit interest rates in the banking market. Prager and Hannan (1998) found that bank mergers and acquisitions have resulted in higher banks concentration, which in turn leads to significantly lower rates on deposits. Some evidence also suggested that U.S. banks that are involved in mergers and acquisitions improved the quality of their outputs in the 1990s in ways that increased costs but still improved profit productivity by increasing revenues more than costs (Berger and Mester, 2003, p. 88). A note of caution however: encouraging or forcing banks to merge in 2 Characteristics of Singapore s Commercial Banks after the Ms & As in 2001 DBS UOB + OUB OCBC + KEP Total Assets (S$ billion) 111.0 113.7 83.0 Total Loans (S$ billion) 54.2 61.5 50.4 Total Deposits (S$ billion) 92.8 96.6 71.1 Total Shareholders Fund (S$ billion) 8.4 13.1 8.3 Number of Branches 107 93 74 Number of ATMs 900 426 381 Source: Banks Annual Reports

266 F. Sufian, M.-Z. Abdul Majid times of severe banking crisis as a measure to reduce bank failure risk, would not only possibly create a weaker bank, but could also worsen the banking sector crisis. As shown by Shih (2003), merging a weaker bank into a healthier bank in many cases would result in a bank even more likely to fail than both the predecessors bank. On the other hand, he found that mergers between relatively healthy banks would create banks that are less likely to fail. 3.1. Studies on Singapore banks efficiency. - Using DEA with three inputs and two outputs, Chu and Lim (1998) evaluated the relative cost and profit efficiency of a panel of six Singapore listed banks during the period 1992-1996. They found that during the period the banks have exhibited higher overall efficiency of 95.3% compared to profit efficiency of 82.6%. They found that large Singapore banks have reported higher efficiency of 99.0% compared to 92.0% for small banks. They also suggested that scale inefficiency outweighs pure technical inefficiency during the period of study. Rezvanian and Mehdian (2002) used parametric and non-parametric approaches to examine the production performance and cost structure of a sample of Singaporean commercial banks. The results of the parametric methodology suggest that the average cost curve of these banks is U-shaped and there are economies of scale for small and medium-size banks. Further analysis provides evidence of economies of scope for all banks regardless of their size. The non-parametric results indicate that Singapore banks could have reduced cost by 43% had they all been overall efficient. This cost inefficiency seems to be caused equally by allocative and technical inefficiencies. More recently, Randhawa and Lim (2005) utilized DEA to investigate the locally incorporated banks in Hong Kong and Singapore X-efficiency during the period 1995 to 1999. They found that during the period the seven domestic incorporated Singapore banks have exhibited an average overall efficiency score of 80.4% under the intermediation approach and 97.2% under the production approach. They suggest that large Singapore banks have reported higher overall efficiency compared to small banks under the production approach, while on the other hand, small banks have exhibited higher overall efficiency under the intermediation approach. They also suggest that pure technical inefficiency outweighs scale inefficiency under both approaches during the period of study.

Banks efficiency in Singapore 267 4. Estimation Methodology 4.1. Data envelopment analysis. - The present study employs the nonparametric frontier approach DEA to estimate the input-oriented technical efficiency of Singapore banks. This approach measures the efficiency of a decision-making unit (DMU) relative to other similar DMUs with the simple restriction that all DMUs lay on or below the efficiency frontier. The purpose of DEA is to empirically characterize the so-called efficient frontier (surface) based on the available set of DMUs and project all DMUs onto this frontier. If a DMU lies on the frontier, it is referred to as an efficient unit; otherwise, it is labelled as inefficient. The data are enveloped in such a way that radial distances to the frontier are minimized. In practice, efficiency scores are calculated by solving a linear programming problem (see Appendix A and B). The analysis under DEA is concerned with understanding how each DMU is performing relative to others, the causes of inefficiency, and how a DMU can improve its performance to become efficient. In that sense, DEA calculates the relative efficiency of each unit in relation to all other units by using the actual observed values for the inputs and outputs of each DMU. It also identifies, for inefficient DMUs, the sources and level of inefficiency for each of the inputs and outputs. The DEA is carried out by assuming either constant returns to scale (CRS), or variable returns to scale (VRS). The estimation with these two assumptions allows the overall technical efficiency (TE) to be broken down into two collectively exhaustive components: pure technical (PTE) and scale efficiency (SE) i.e. TE = PTE x SE. The former relates to the capability of managers to utilize the firms given resources, whereas the latter refers to exploiting scale economies by operating at a point where the production frontier exhibits constant returns to scale. A useful feature of VRS models as compared to CRS models is that they report whether a DMU is operating at increasing, constant, or decreasing returns to scale. Constant returns to scale will apply when CRS and VRS efficiency frontiers are tangential with each other; in other words, when the slope of the efficiency frontier is equal to the ratio of inputs to outputs (Cooper et al., 2000). Increasing returns to scale must apply below that level, as the slope of the efficient frontier which reflects the marginal rate of transformation of inputs to outputs will be greater than the average rate of conversion. Likewise, decreasing returns to scale must apply above the zone in which constant returns to scale apply. DMUs that are not on the efficient frontier must first be projected onto the efficient frontier before their returns to scale status can be assessed.

268 F. Sufian, M.-Z. Abdul Majid Five useful features of DEA are: first, each DMU is assigned a single efficiency score, hence allowing ranking amongst the DMUs in the sample. Second, it highlights the areas of improvement for each single DMU. For example, since a DMU is compared to a set of efficient DMUs with similar input-output configurations, the DMU in question is able to identify whether it has used input excessively or its output has been under-produced. Third, there is possibility of making inferences on the DMU s general profile. We should be aware that the technique used here is a comparison between the production performances of each DMU to a set of efficient DMUs. The set of efficient DMUs is called the reference set. The owners of the DMUs may be interested to know which DMU frequently appears in this set. A DMU that appears more than others in this set is called the global leader. Clearly, this piece of information gives huge benefits to the DMU owner, especially in positioning its entity in the market. Fourth, DEA does not require a preconceived structure or specific functional form to be imposed on the data in identifying and determining the efficient frontier, error, and inefficiency structures of the DMUs 3 (Bauer et al., 1998; Evanoff and Israelvich, 1991; Grifell-Tatje and Lovell, 1997). Finally, Avkiran (1999) acknowledges the advantage of DEA by stating that this technique allows the researchers to choose any kind of input and output of managerial interest, regardless of different measurement units. There is no need for standardization 4. The main weakness of DEA is that it assumes that data are free from measurement errors. Furthermore, since efficiency is measured in a relative way, its analysis is confined to the sample set used. This means that an efficient DMU found in the analysis cannot be compared with other DMUs outside the sample. 4.2. Multivariate Tobit regression analysis. - It is also of considerable interest to explain the determinants of technical efficiency scores derived from DEA models. As defined in equations (A1) and (A2), the DEA score falls between the interval 0 and 1 (0 < h * 1) making the dependent variable a limited dependent variable. A commonly held view in previous studies is that the use of the Tobit model can handle the characteristics of the distribution of (in)efficiency measures and therefore provide results that can guide policies to improve performance. As the dependent variable inefficiency score is bounded between 0 and 1, an appropriate theoretical specification is a Tobit model with two side censoring. However, firms with an inefficiency score of 3 Avkiran (1999) provides a relatively thorough discussion of the merits and limits of DEA. 4 An additional advantage according to Canhoto and Dermine (2003) is that the DEA technique is preferred to parametric methods when the sample size is small.

Banks efficiency in Singapore 269 1 will never be observed in practice. Therefore, the results of the empirical analysis will not be different if one specifies a one or two side Tobit model. Accordingly, DEA efficiency scores obtained in the first stage are used as dependent variables in the second stage of one side censored Tobit model in order to allow for the restricted [0, 1] range of inefficiency values. Coelli et al. (1998) suggested several ways in which environmental variables can be accommodated in a DEA analysis. The term environmental variables is usually used to describe factors, which could influence the efficiency of a firm. In this case, such factors are not traditional inputs and are assumed to be outside the control of the manager. Hence, the two-stage method used in this paper involves the solution of the DEA problem in the first stage analysis, which comprises mainly the traditional outputs and inputs. In the second stage, the efficiency scores obtained from the first stage analysis are regressed on the environmental variables. The standard Tobit model can be defined as follows for observation (bank) i : (1) * i y ' xi i * * i i if i y y y 0 and yi 0 otherwise 2 where i ~ N(0, ), xi and are vectors of explanatory variables and unknown parameters, respectively. The y * i is a latent variable and y i is the DEA score 5. 4.3. Inputs, outputs definition, and the choice of variables. - The definition and measurement of inputs and outputs in the banking function remains a contentious issue among researchers. To determine what constitutes inputs and outputs of banks, one should first decide on the nature of banking technology. In banking theory literature, there are two main approaches 5 The likelihood function (L) is maximized to solve and based on 20 observations (banks) of y i and x i is 1 2 i [1/( 2 )]( y x ) i L (1 F) e 2 1/2 (2 ) where, 2 y 0 y 0 i i x / 1 2 i t /2 F e dt i. The first product is over the observations for which the banks are 1/2 (2 ) 100% efficient (y = 0) and the second product is over the observations for which banks are inefficient (y >0). F i is the distribution function of the standard normal evaluated at x i /.

270 F. Sufian, M.-Z. Abdul Majid competing with each other in this regard: the production and intermediation approaches (Sealey and Lindley, 1977). Under the production approach, a financial institution is defined as a producer of services for account holders, that is, they perform transactions on deposit accounts and process documents such as loans. Hence, according to this approach, the number of accounts or related transactions is the best measure output, while the number of employees and physical capital are considered as inputs. Previous studies that adopted this approach are, among others, Sherman and Gold (1985), Ferrier and Lovell (1990), and Fried et al. (1993). The intermediation approach on the other hand assumes that financial firms act as an intermediary between savers and borrowers and posits total loans and securities as outputs, whereas deposits along with labour and physical capital are defined as inputs. Previous banking efficiency studies that adopted this approach are, among others, Charnes et al. (1990), Bhattacharyya et al. (1997), and Sathye (2001). For the purpose of this study, a variation of the intermediation approach or asset approach originally developed by Sealey and Lindley (1977) will be adopted in the definition of input and output 6. According to Berger and Humphrey (1997), the production approach might be more suitable for branch efficiency studies, as at most times bank branches basically process customer documents and bank funding, while investment decisions are mostly not under the control of branches. Furthermore, Sathye (2001) also noted that this approach is more relevant to financial institutions, as it is inclusive of interest expenses, which often account for one-half to two-thirds of total costs depending on the phase of the interest rate cycles. The aim in the choice of variables for this study is to provide a parsimonious model and to avoid the use of unnecessary variables that may reduce the degree of freedom 7. All variables are measured in millions of Singapore Dollars (SG$). Given the sensitivity of efficiency estimates to the specification of outputs and inputs, we estimate two alternative models. In Model 1, we model Singapore commercial banks as multi-product firms, producing an output by employing two inputs. Accordingly, Total Deposits (x 1 ) and Interest Expense (x 2 ) will be used as input variables. Total Loans (y 1 ) will be included as an output in Model 1. In Model 2, we follow the approach by Avkiran (1999), to include Total Deposits (x 1 ) as an input vector to 6 Humphrey (1985) presents an extended discussion of the alternative approaches of what a bank produces. 7 For a detailed discussion on the optimal number of inputs and outputs in DEA, see Avkiran (2002).

Banks efficiency in Singapore 271 produce Total Loans (y 1 ) and Interest Income (y 2 ). Table 1. Descriptive statistics Variable Mean Std. Dev. Minimum Maximum Total Loans (y 1 ) 45,348.21 18,845.16 12,713.56 71,021.00 Interest Income (y 2 ) 3,201.95 1,153.90 944.39 5,298.00 Total Deposits (x 1 ) 56,598.01 30,090.08 12,089.23 113,206.00 Interest Expense (x 2 ) 1,674.51 736.21 568.64 3,501.26 Note: Model 1 Outputs = (y 1 ), Inputs (x 1, x 2 ) Model 2 Outputs = (y 1, y 2 ), Inputs (x 1 ) 4.4. Data. - For the empirical analysis, all domestically incorporated Singapore commercial banks will be incorporated in the study. In the spirit of maintaining homogeneity, only commercial banks that make commercial loans and accept deposits from the public are included in the analysis. Therefore, Investment Banks are excluded from the sample. The annual balance sheet and income statement used to construct the variables for the empirical analysis were taken from published balance sheet information in annual reports of each individual bank. Three banks were omitted from the study, namely, Bank of Singapore, Far Eastern Bank and Industrial and Commercial Bank, which are all wholly owned subsidiaries of the OCBC and UOB groups. As for the potential determinants in the Tobit regression, the following variables from the published annual report of individual banks from 1998 to 2004 are used: first, we determine the impact of bank size on Singapore banking groups efficiency, and the impact of efficiency on the Singapore banking groups profitability. Bank size is measured by the amount of total assets, and bank profitability is measured by net operating income to total assets. Second, there are various bank specific characteristics, which may have an impact on efficiency. Three variables are utilized to explain the Singapore banking groups efficiency: 1) capitalization is measured by the amount of share and supplementary capital divided by total assets; 2) asset quality is measured by provision over loans, and 3) overhead costs are measured by personnel expense over the number of employees. 5. Empirical Results In the spirit of Rhoades (1998), we develop a [ 3, 3] event window, to investigate the effect of mergers and acquisitions on the Singapore banking

272 F. Sufian, M.-Z. Abdul Majid groups efficiency. The choice of the event window is motivated by Rhoades (1998, p. 278), who pointed out that there has been unanimous agreement among the experts that about half of any efficiency gains should be apparent after one year and all gains should be realized within three years after the merger. The whole period (i.e. 1998-2004) is divided into three sub-periods: 1998-2000 refers to the pre-merger period, 2001 is considered as the merger year, and 2002-2004 represents the post-merger period, when mergers and acquisitions are expected to have some impact on Singapore banking groups efficiency. We expect to be able to capture the effects of mergers and acquisitions on the efficiency of Singapore banks during this period. During all periods the targets and acquirers mean overall efficiency along with its decomposition of pure technical and scale efficiency scores are compared. This could help to shed some light on the sources of inefficiency of the Singapore banking system in general, as well as to differentiate between the target and acquirers efficiency scores. 5.1. Pre-merger. - Panels A and B of Table 2 illustrate the overall efficiency estimates for both DEA Model 1 and DEA Model 2, along with its decomposition into pure technical and scale efficiency. It is apparent that during the pre-merger period Singapore banks have exhibited a mean overall efficiency of 91.68% for DEA Model 1, while the results for DEA Model 2 suggest a lower mean overall efficiency of 88.59%. Overall, the results suggest that the Singapore banking system has performed relatively well in its basic function transforming deposits to loans, with relatively minimal input waste (i.e. 8.32% for DEA Model 1, and 11.41% for DEA Model 2). Thus, the results imply that during the pre-merger period, in the case of DEA Model 1 for instance, the Singapore banking groups could have produced the same amount of outputs with only 91.68% of the amount of inputs used. In other words, Singapore banking groups could have reduced their inputs by 8.32% without affecting the amount of outputs produced. The findings are in line with Chu and Lim (1998) who found that Singapore banks have exhibited average overall efficiency of 95.30% during the period 1992-1996, while Randhawa and Lim (2005) found 19.60% input waste among seven Singapore domestic banks during the period 1995-1999. The results also compare favourably with Fukuyama s (1993) study on Japanese banks (14%), and the 14% to 25% averages of Indian commercial banks (Bhattacharyya et al., 1997). The decomposition of overall efficiency into its pure technical and scale efficiency estimates for DEA Model 1 suggests that during the pre-merger period, Singapore banks inefficiency was largely due to scale (5.44%) rather than to pure technical efficiency

Banks efficiency in Singapore 273 scores (3.08%). The empirical findings imply that although the Singapore banking groups were managerially efficient in controlling their operating costs, they have been operating at a non-optimal scale of operations prior to the merger. The results from DEA Model 1 are further confirmed by the results from DEA Model 2, which suggest that during the pre-merger period, the Singapore banking groups inefficiency was largely due to scale rather than to pure technical efficiency scores. Table 2. Summary of mean efficiency levels of Singapore banks Panel A Model 1 Pre-Merger* During Merger** Post-Merger*** Bank OE PTE SE OE PTE SE OE PTE SE KEP 99.30 100.00 99.30 - - - - - - OCBC 86.50 96.30 89.60 100.00 100.00 100.00 100.00 100.00 100.00 OUB 97.20 100.00 97.20 - - - - - - UOB 84.70 88.30 96.00 100.00 100.00 100.00 93.20 93.20 100.0 DBS 90.70 100.00 90.70 72.40 100.00 72.40 91.60 100.0 91.60 Mean 91.68 96.92 94.56 90.80 100.00 90.80 94.93 97.73 97.20 Panel B Model 2 Pre-Merger* During Merger** Post-Merger*** Bank OE PTE SE OE PTE SE OE PTE SE KEP 99.23 100.0 99.23 OCBC 96.23 100.0 96.23 100.0 100.0 100.0 100.0 100.0 100.0 OUB 100.0 100.0 100.0 UOB 76.40 78.43 97.73 88.80 100.0 88.80 100.0 100.0 100.0 DBS 71.10 100.0 71.10 88.20 100.0 88.20 75.53 100.0 75.53 Mean 88.59 93.69 92.86 91.82 100.0 92.33 91.84 100.0 91.84 Note: * 1998-2000; ** 2001; *** 2002-2004 OE = Overall Efficiency; PTE = Pure Technical Efficiency; SE = Scale Efficiency 5.2. Post-merger. - The findings from DEA Model 1 (Panel A of Table 2) clearly suggest that the merger has resulted in the improvement of Singapore banking groups overall efficiency during the post-merger period. It is apparent from Panel A of Table 2 that the Singapore banking groups have exhibited mean overall efficiency of 94.93% during the post-merger period, higher than the 91.68% recorded during the pre-merger period. It is also interesting to note that all Singapore banking groups have exhibited a higher mean overall efficiency post-merger. Similar to DEA Model 1, the results from DEA Model 2 (Panel B of Table 2) also suggest that the merger has resulted in the improvement of Singapore banking groups mean overall efficiency. It is clear from Panel B of Table 2 that the Singapore banking groups mean overall efficiency improved to 91.84%, post-merger from

274 F. Sufian, M.-Z. Abdul Majid 88.59% recorded during the pre-merger period. A closer look at the decomposition of efficiency into its pure technical and scale efficiency components reveals that while for DEA Model 1 the overall efficiency improvement during the post-merger period was mainly attributed to scale efficiency, the opposite seems true for DEA Model 2, which suggests that the improvement in overall efficiency was solely attributed to the improvement in pure technical efficiency. It is also interesting to note that despite earlier evidence implying that the lack of competition may result in lower technical efficiency (see Sathye, 2001, and Walker, 1998), it is apparent from Panel B of Table 2 that all Singapore banking groups have reported a 100% mean pure technical efficiency score post-merger. Walker (1998) states that the high degree of concentration in Australian banking, which was dominated by four major banks, may result in the quiet life hypothesis to come into play. The quiet life hypothesis predicts a reverse causation, i.e. as firms enjoy greater market power and concentration, inefficiency follows due to the relaxed environment with less incentive to minimize costs. Hence, the findings suggest that during the period of 1998-2004 the source of inefficiency among Singapore domestic incorporated banks was solely due to scale inefficiency. 5.3. Is the acquirer a more efficient bank? - We now turn to the assessment of how the mergers and consolidation process affects the mean overall efficiency of the involved banks. First, we analyze the pre-merger performance of the banks concerned. Theoretically, the more efficient banks should acquire the less efficient ones. A more efficient bank is assumed to be well organized, and have a more capable management. The idea is that since there is room for improvement concerning the performance of the less efficient bank, a takeover by a more efficient bank will lead to a transfer of the better management quality to the inefficient bank. This will in turn lead to a more efficient and better performing merged unit. In order to see whether indeed it is the case that more efficient banks acquire inefficient ones, we calculate the difference in overall efficiency between an acquiring and an acquired bank. This efficiency difference is measured as the overall efficiency of the acquiring bank, minus the mean overall efficiency of the acquired banks for the last observation period before consolidation. For DEA Model 1, it is clear from Table 3 that during the pre-merger period KEP s (the target) overall efficiency level of 99.30% is relatively higher compared to OCBC s (the acquirer) overall efficiency of 86.50%. Similarly, from Table 3 it is clear that during the pre-merger period, for DEA Model 1, UOB s (the acquirer) overall efficiency level of 84.70% is lower

Banks efficiency in Singapore 275 compared to OUB s (the target) overall efficiency of 97.20%. Thus, the results from DEA Model 1 reject the hypothesis that the targets were less efficient relative to the acquirers. Similarly, the results for DEA Model 2 suggest that KEP s mean overall efficiency of 99.23% is higher compared to its acquirer s, OCBC s, mean overall efficiency level of 96.23% during the pre-merger period. Likewise, it is clear from Table 3 that during the premerger period, UOB s mean overall efficiency of 76.40% is lower compared to its target, OUB s, mean overall efficiency of 100.0%. Again, the results from DEA Model 2 reject the hypothesis that the acquirers were more efficient than the targets during the pre-merger period. Table 3. Summary of mean efficiency levels of targets and acquiring banks Model 1 Model 2 Target/ Pre-Merger* Pre-Merger* Bank Acquirer OE PTE SE OE PTE SE OCBC+KEP KEP Target 99.30 100.0 99.30 99.23 100.0 99.23 OCBC Acquirer 86.50 96.30 89.60 96.23 100.0 96.23 UOB+OUB OUB Target 97.20 100.0 97.20 100.0 100.0 100.0 UOB Acquirer 84.70 88.30 96.0 76.40 78.43 97.73 Note: 1998-2000 OE = Overall Efficiency; PTE = Pure Technical Efficiency; SE = Scale Efficiency The font in bold indicates banking group that is relatively more efficient 5.4. Implications of mergers on acquiring banks efficiency. - We now turn our discussions on the impact of the mergers and acquisitions on the Singapore banking groups efficiency. The issue at hand is whether a more (less) efficient target resulted in the improvement (deterioration) in the acquirers efficiency levels post-merger. For DEA Model 1, KEP s (the target) mean overall efficiency level of 99.30% is higher compared to OCBC s (the acquirer) mean overall efficiency of 86.50% during the premerger period. It is apparent from Panel A of Table 4 that the merger between OCBC and KEP has resulted in the improvement of OCBC s mean overall efficiency during the merger and subsequently post-merger, when OCBC has been operating as a fully efficient bank. Similarly, from Panel A of Table 4 it is clear that during the pre-merger period, UOB has exhibited a lower overall efficiency level of 84.70% for Model 1 compared to its target, OUB s overall efficiency of 97.20%. Again, the results suggest that UOB s overall efficiency improved to 93.20% post-merger. Based on the results for

276 F. Sufian, M.-Z. Abdul Majid DEA Model 1 we can conclude that a more efficient target resulted in the improvement of the acquirers mean overall efficiency post-merger. Table 4. Summary of mean efficiency levels of targets and acquirers banks Panel A Model 1 Pre-Merger* During Merger** Post-Merger*** Bank OE PTE SE OE PTE SE OE PTE SE OCBC+KEP KEP 99.30 100.0 99.30 - - - - - - OCBC 86.50 96.3 89.60 100.0 100.0 100.0 100.0 100.0 100.0 UOB+OUB OUB 97.20 100.0 97.20 - - - - - - UOB 84.70 88.3 96.0 100.0 100.0 100.0 93.20 93.20 100.0 Panel B Model 2 Pre-Merger* During Merger** Post-Merger*** Bank OE PTE SE OE PTE SE OE PTE SE OCBC+KEP KEP 99.23 100.0 99.23 - - - - - - OCBC 96.23 100.0 96.23 100.0 100.0 100.0 100.0 100.0 100.0 UOB+OUB OUB 100.0 100.0 100.0 - - - - - - UOB 76.40 78.43 97.73 88.80 100.0 88.80 100.0 100.0 100.0 Note: * 1998-2000; ** 2001; *** 2002-2004 OE = Overall Efficiency; PTE = Pure Technical Efficiency; SE = Scale Efficiency The font in bold indicates changes that are beneficial for the banking group Similarly to DEA Model 1, it is apparent from Panel B of Table 4 that the results for DEA Model 2 suggest that KEP s (the target) mean overall efficiency of 99.23% is higher compared to OCBC s (the acquirer) mean overall efficiency of 96.23%. The empirical findings again suggest that OCBC s mean overall efficiency improved and the bank has been operating as a fully efficient bank since the merger year, and subsequently during the post-merger period. Similarly to the merger between KEP and OCBC, for DEA Model 2, it is clear from Panel B of Table 4 that during the pre-merger period UOB s (the acquirer) overall efficiency of 76.40% is lower compared to OUB s (the target) overall efficiency of 100.0%. The results suggest that UOB s mean overall efficiency level improved to 88.80% during the merger year, and it has been operating at 100.0% efficiency level post-merger. Similarly to the findings for DEA Model 1, the results from DEA Model 2 again support the hypothesis that the acquirers efficiency improves from the

Banks efficiency in Singapore 277 acquisition of a more efficient target. 5.5. Results of second stage Tobit regression. - To further investigate the determinants of the Singapore banking groups efficiency over time, the efficiency scores for DEA Model 1 and DEA Model 2 are estimated by using the censored Tobit model. Unlike a conventional Ordinary Least Square (OLS) estimation, in cases with limited dependent variables, the Tobit model is known to generate consistent estimates of regression coefficients. The results of the estimation are presented in Table 5. A positive coefficient implies an efficiency increase whereas a negative coefficient reflects deterioration in efficiency. Table 5. Second stage Tobit regression of the efficiency measures and bank characteristics it = + 1 SIZE it + 2 PROFITABILITY it + 3 CAPITALISATION it + 4 PROVISION/LOANS it + 5 OVERHEADS it + it Explanatory Variables DEA 1 DEA 2 CONSTANT 0.477 ( 0.694) 0.612 (0.791) Bank Characteristics SIZE 0.022 ( 0.949) 0.045 ( 1.358) PROFITABILITY 0.003 (0.295) 0.037*** (3.551) CAPITALIZATION 0.375 (1.586) 0.265 (1.254) PROVISIONS/LOANS 1.478*** 0.101 ( 3.710) OVERHEADS 2.478*** (2.828) ( 0.182) 0.662 (0.663) Log likelihood 23.39 23.84 R2 0.39 0.70 Note: The dependent variable is bank s efficiency scores derived from DEA Model 1 and DEA Model 2; SIZE is a measure of bank s market share calculated as a natural logarithm of total bank assets; PROFITABILITY is a measure of bank s profit calculated as the ratio of net operating income to bank s total assets; CAPITALIZATION is the bank s specific characteristics measured as the ratio of the amount of share and supplementary capital divided by total assets; PROVISIONS/LOANS is a measure of bank s assets quality calculated as the ratio of total loan loss provisions divided by total loans; OVERHEADS is a measure of overhead costs calculated as personnel expense over numbers of employees. DEA 1 refers to DEA scores generated from Model 1 and DEA 2 refers to DEA scores generated from Model 2. ***, **, and * indicate significance at 1, 5 and 10% levels; z-statistics are in parenthesis. It is apparent from Table 5 that bank size has a negative effect on efficiency for both models but it is insignificant at any conventional level,

278 F. Sufian, M.-Z. Abdul Majid indicating that the larger banks have lower efficiency, which could be due to a complex organizational structure and moral hazard behaviour. However, the insignificant coefficient indicates that efficiency is independent of the size of the bank. On the other hand, the results suggest that profitability has a significant positive relationship with bank efficiency for DEA Model 2, indicating that the more profitable banks tend to exhibit higher efficiency scores. Banks reporting higher profitability are preferred by clients and attract the biggest share of deposits as well as the best potential borrowers particularly in the Singapore banking sector, which corresponds with the study by Jackson and Fethi (2000) on the Turkish banking sector. Now we turn to the analysis of bank characteristics and their influence on efficiency. As can be seen from Table 5, the capitalization variable yields positive impact but is insignificant at any conventional level in explaining bank performance for both models. Theoretically, better capitalized banks should enjoy a higher level of efficiency. In performing further investigation, we have treated loans as homogenous with respect to risk. We were forced to make such an assumption because we could not correct the model for risk without a thorough investigation of the causes of bad loans (Berger and DeYoung, 1997). If a bank has a poor quality loan portfolio, it should entail additional costs associated with monitoring and enforcement of loan repayment. The significant negative coefficient of the provisions over loans variable in DEA Model 1 supports the above prediction. The findings of the effect of overhead costs on bank efficiency seem counterintuitive at a first glance, whereby higher overhead costs seem to pay off and result in higher bank performance as indicated in DEA Model 1. Although theoretically consolidation should reduce the amount of back office personnel, the reductions could be offset by increases in the front office personnel, implying a better customer service. Furthermore, as suggested by Sathye (2001), management that is more professional might require higher remuneration and thus a highly significant positive relationship with efficiency measure is natural. The result is also consistent with Claessens et al. (2001) who show that overstaffing of domestic banks in middle-income countries has always led to deterioration in bank efficiency, but not in highincome countries. 6. Conclusions Applying a non-parametric frontier approach Data Envelopment Analysis (DEA), the paper attempts to investigate the effects of mergers and

Banks efficiency in Singapore 279 acquisitions on the efficiency of domestic incorporated Singapore banking groups. The sample period is divided into three sub-periods, i.e. pre-merger, during merger and post-merger periods, to compare the differences in Singapore banking groups mean efficiency levels during all periods. Given the sensitivity of efficiency estimates to the specification of inputs and outputs used, we adopted a variant of the intermediation approach to two models. For DEA Model 1, the results suggest that Singapore banking groups have exhibited a commendable overall efficiency level of 91.68%, suggesting a minimal input waste of 8.32%. We found that during the merger year, Singapore banking groups overall efficiency level deteriorated slightly to 90.80%, which was solely due to scale inefficiency. Despite that, during the post merger period Singapore banking groups have exhibited higher mean overall efficiency levels compared to the pre-merger period. Similar to the pre-merger period, the findings suggest that scale inefficiency outweighs pure technical inefficiency in the Singapore banking sector post-merger. Similar to the results from DEA Model 1, the results from DEA Model 2 suggest that Singapore banking groups were relatively efficient in their intermediation role, exhibiting a relatively minimal input waste of 11.41% during the premerger period. In contrast to the results from DEA Model 1, the results from DEA Model 2 suggest that Singapore banking groups mean overall efficiency levels were higher during the merger year and improved further during the post-merger period. Although mergers have resulted in a more efficient banking system, as it may appear from the results for DEA Model 1 and DEA Model 2, ironically the size increase has become the biggest factor resulting in the inefficiency of the Singapore banking system. Hence, from the scale efficiency perspective, both results do not support further consolidation in the Singapore banking sector to create two super banks. The findings from DEA Model 1 and DEA Model 2 suggest that further increase in size would only result in a smaller increase of outputs for every proportionate increase in inputs. This results from the fact that Singapore banking groups have been operating at declining returns to scale (DRS) during the post-merger period. The empirical results from both models do not support the hypothesis of a less efficient bank becoming a merger target, as both the targets were found to be more efficient compared to the acquirers during the pre-merger period. The findings further support the hypothesis that the acquiring banks mean overall efficiency improved from the merger with a more efficient target bank. The explanation of the efficiency scores using Tobit regressions offers

280 F. Sufian, M.-Z. Abdul Majid useful economic insights. We interpret the significance of profitability as an indication of the ability to attract the biggest share of deposit as well as the best potential borrowers. The significance of the level of loan quality portfolio, proxy by provision of bad loans, should entail additional costs associated with monitoring and enforcement of loan repayment, hence negatively related to efficiency. Not surprisingly due to the high complexity of the banking environment in Singapore, overhead costs tend to contribute positively to bank performance, which might be attributed to highly skilled personnel with high remuneration packages. APPENDIX A DEA CCR Model The term Data Envelopment Analysis (DEA) was first introduced by Charnes, Cooper and Rhodes (1978), (hereafter CCR), to measure the efficiency of each Decision Making Unit (DMU) that is obtained as a maximum of a ratio of weighted outputs to weighted inputs. This denotes that the more the output produced from given inputs, the more efficient is the production. The weights for the ratio are determined by a restriction that the similar ratios for every DMU have to be less than or equal to unity. This definition of the efficiency measure allows multiple outputs and inputs without requiring pre-assigned weights. Multiple inputs and outputs are reduced to single virtual input and single virtual output by optimal weights. The efficiency measure is then a function of multipliers of the virtual input-output combination. Formally, the efficiency measure for DMU j can be calculated by solving the following mathematical programming problem: (A1) max 0 0 subject to: 0 n jyrj y r 0 j 1 n 0x i 0 0 jxij j 1 n 1 0 j j 1 0 j 0 (r =1,..,n) (i =1,..,n) (j =1,..,n) where x ij is the observed amount of input of the ith type of the jth DMU (x ij > 0, i = 1, 2, m, j = 1, 2, n) and y rj is the observed amount of output of the rth type for the jth DMU (y rj > 0, r = 1, 2, s, j = 1, 2,, n).

Banks efficiency in Singapore 281 APPENDIX B DEA BCC Model Banker, Charnes and Cooper (1984) extended the CCR model by relaxing the CRS assumption. The resulting BCC model was used to assess the efficiency of DMUs characterized by variable returns to scale (VRS). The VRS assumption provides the measurement of pure technical efficiency (PTE), which is the measurement of technical efficiency devoid of scale efficiency effects. If there appears to be a difference between the TE and PTE scores of a particular DMU, then it indicates the existence of scale inefficiency. The input oriented BCC model for the DMU j can be written as: (A2) max 0 0 subject to: 0 n jyrj y r 0 j 1 n 0x i 0 0 jxij j 1 n 1 0 j j 1 0 j 0 (r =1,..,n) (i =1,..,n) (j =1,..,n) The BCC efficiency scores are obtained by running the above model for each DMU. These scores are called pure technical efficiency scores, since they are obtained from the model that allows variable returns to scale and hence eliminates the scale part. Generally, the CCR efficiency score for each DMU will not exceed the BCC efficiency score, which is intuitively clear since the BCC model analyses each DMU locally rather than globally. Once pure technical efficiency (PTE) estimates are available, scale efficiency (SE) is computed from the following formula: SE = Technical Efficiency (CRS)/ Pure Technical Efficiency (VRS). REFERENCES AVKIRAN N.K., The Evidence on Efficiency Gains: The Role of Mergers and the Benefits to the Public, Journal of Banking and Finance, 1999, 23, pp. 991-1013., Productivity Analysis in the Service Sector with Data Envelopment Analysis, 2 nd edn., Camira, Qld.: N.K. Avkiran, 2002, 242 pp.