Impact of policy changes on the efficiency and returns-to-scale of Japanese financial institutions: An evaluation

Accepted Manuscript Title: Impact of Policy Changes on the Efficiency and Returns-to-Scale of Japanese Financial Institutions: An Evaluation Author: A.S.M. Sohel Azad Suzuki Yasushi Victor Fang Amirul Ahsan PII: DOI: Reference:

S0275-5319(14)00026-9 http://dx.doi.org/doi:10.1016/j.ribaf.2014.05.001 RIBAF 310

To appear in:

Research in International Business and Finance

Received date: Revised date: Accepted date:

29-11-2013 1-4-2014 13-5-2014

Please cite this article as: Azad, A.S.M.S., Yasushi, S., Fang, V., Ahsan, A.,Impact of Policy Changes on the Efficiency and Returns-to-Scale of Japanese Financial Institutions: An Evaluation, Research in International Business and Finance (2014), http://dx.doi.org/10.1016/j.ribaf.2014.05.001 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Impact of Policy Changes on the Efficiency and Returns-to-Scale of Japanese Financial Institutions: An Evaluation

ip t

A.S.M. Sohel Azad*, Suzuki Yasushi, Victor Fang and Amirul Ahsan

Suzuki Yasushi, PhD, LLM, MSc, BA

Lecturer in Finance

Professor of Finance/Economics

School of Accounting, Economics and Finance

College of International Management

Faculty of Business and Law, Deakin University

Graduate School of Management

us

cr

A.S.M. Sohel Azad, PhD, MSc, MBA, BBA (Hons)

221 Burwood Highway, Burwood, Vic-3125, Ritsumeikan Asia Pacific University Australia;

Phone:

+61392446873;

Fax: 1-1 Jumonjibaru, Beppu, Oita 874-8577, Japan E-mail: [email protected]

Victor Fang, PhD, MSc, BSc

Amirul Ahsan, PhD, MBA, MCom, BCom (Hons)

Associate Professor of Finance

Lecturer in Finance

School of Accounting, Economics and Finance

School of Accounting, Economics and Finance

Faculty of Business and Law, Deakin University

Faculty of Business and Law, Deakin University

M

an

+61392446283 ; E-mail: [email protected]

Australia

d

221 Burwood Highway, Burwood, Vic-3125, 221 Burwood Highway, Burwood, Vic-3125, Australia Phone: +61392446571; Fax: +61392446283;

E-mail: [email protected]

E-mail: [email protected]

Ac ce p

te

Phone: +61392446919; Fax: +61392446283;

Abstract:

This paper examines the impact of FSA’s (Financial Services Agency) recent policy changes on the efficiency and returns-to-scale (RTS) of Japanese financial institutions including banks, securities companies and bank holding companies. Three kinds of efficiency are investigated namely, technical efficiency (TE), pure technical efficiency (PTE) and scale efficiency (SE) using the non-parametric methodology named data envelopment analysis (DEA). The DEA analysis shows a substantial improvement in the overall efficiency of Japanese banks, albeit a significant difference of efficiency scores between the major/city banks and the regional banks. Results are robust to alternative specifications of efficiency and scale changes. Key words: Policy changes; DEA; Efficiency; Returns-to-scale; Malmquist Index; JEL Classification: C 14; G 21;

1

Page 1 of 27

Impact of Policy Changes on the Efficiency and Returns-to-Scale of Japanese Financial Institutions: An Evaluation 1. Introduction

ip t

The corporate Japan is historically dominated by the bank-based financial system. However, the prolonged trouble of the financial sector followed by burst of bubble

cr

challenged the bank-based financial system and its whole financial industry until recently.

us

Because of the predominance of the financial institutions, the financial sector’s revival is considered crucial for its economic growth. The Financial Services Agency (FSA) of Japan

an

recommended numerous policy measures towards the revival of this sector. Between 2001 and 2002, several policies were suggested by the FSA. The period has also undergone a

M

strong wave of the mergers and acquisitions. However, the existing literature does not shed light on the effectiveness of those measures on the efficiency of Japanese financial

d

institutions. This paper fills that gap by looking into the impact of those measures and policy

te

changes on different proxies of efficiency and returns-to-scale. Of the several policies adopted, the noteworthy are as follows. In June 2001, the

Ac ce p

FSA suggested two major policies: (i) drastic resolution of the non-performing loans (NPLs) and (ii) structural reforms in the securities markets. The government also passed legislation in 1998 that set April 2002 as the beginning of a changed relationship between Postal Saving System (PSS) and the Fiscal Investment and Loan Program (FILP) with a view to enhance the financial intermediation and strengthen the financial stability. In October 2002, the FSA announced more concrete policies aiming to revive the financial industry as well as the corporate sector because of their historical interdependences. Those policies focused on a wide range of areas including the tightening of loan/asset assessment, enhancing capital adequacy and strengthening governance. Policies were also suggested for strengthening small and medium size enterprises (SMEs) so that they are able to repay back their loans to banks

2

Page 2 of 27

that suffered from valuation losses. The policies mainly emphasized the monitoring, assessment, resolution and disclosure of the real picture of the NPLs. Since 2002, the FSA has begun to issue the “Business Improvement Administrative Order” to those financial

ip t

institutions which have failed to support the revival of SMEs. In December 2002, the Deposit Insurance Law and other related laws were also amended. In June 2003, the government

cr

opted to provide financial support to solve the capital adequacy problem of the ailing Resona

us

Bank, which was temporarily nationalized in November 2003. In March 2004, the UFJ, one of the four mega banks, recorded a huge loss. As a consequence, UFJ was merged with Tokyo

an

Mitsubishi Bank in October 2005. Since December 2004, Japan’s financial system has entered into a new forward-looking phase aiming at establishing a more stable financial

M

system for the future. In addition, the FSA has implemented the “Program for Financial Revival” and other measures to tackle the non-performing loans problem. This new phase

d

changed the attitude of Japan’s financial administration by putting emphasis on “financial

te

system vitality” from a previous emphasis on “financial system stability”. Additionally, following the increased usage of IT services in the financial industry, the Regional Banks

Ac ce p

Association of Japan implemented a credit risk assessment tool namely, credit risk information total system (or CRITS). The CRITS, which was initially implemented in December 2004, has now been updated to the latest version in April 2010. As of January 1, 2012, the tool is found to be used by all of its 64 member banks as well as other regional banks, financial institutions, and nonfinancial companies. The purpose of this large-scale IT based credit risk management model is to improve the capacity to manage the credit risks associated with the loans to SMEs (small and medium enterprises). The recent period has also undergone a strong wave of the mergers and acquisitions. The trend of consolidation started in the late of the last century and continued until recently. However, in most cases, the mergers were not aimed at increasing the competitiveness but were the products of adversity including take-over battles, adverse mergers and acquisitions. 3

Page 3 of 27

For instance, the consolidation of UFJ and MTFG was hindered by UFJ’s prior contract to sell its trust bank to Sumitomo but finally UFJ and MTFG formed the largest bank both in Japan and in the world in 2005. Table 1 summarises the merger wave and the resulting new

ip t

banking group. The financial consolidation in Japan has been due to several reasons. These include:

cr

(i) the pressure from the regulatory authorities (e.g., FSA) to acquire the failing banks by

us

major banks after the bubble burst (ii) the value maximizing motives to increase the market share – both retail and corporate banking (Berger, Demsetz and Strahan (1999))

(iii)

an

non-value maximizing motives e.g. enhancing managerial status and executive compensation (Berger et al. (1999)) (iv) the impact of external competition and deregulation (Drake, Hall

M

and Simper (2006)) and (v) poor profitability. In Japan, merger and acquisition is becoming popular beyond the banking tie-up. A full-fledged conglomerate in the early 2006 between the

d

SMFG (third largest banking group) and Daiwa Securities (second largest brokerage firm in

te

Japan) has altered the landscape of Japan’s financial industry. This is considered to be the first ever large-scale cross tie-up beyond the banking mega merger in Japan. The

Ac ce p

consolidation has enabled them to take mutual advantages of their customer bases and diversified services including bond and share issuances, and mergers and acquisitions that the brokerage specializes in.

Table 1: Trend of recent mergers in the Japanese financial world

The data in this table are compiled from Financial Times (May 19, 2003) and The Nikkei Weekly (July 19, 2004), The Asahi Shimbun/International Herald Tribune (August 13, 2004) and Major Banks’ Annual Reports. Chronology of events

Banks and FIs merged

New Banking Group

2001 (April)

Bank of Tokyo Mitsubishi, Nippon Trust and Mitsubishi Trust

MTFG

2001 (April)

Sakura Bank and Sumitomo Bank

SMFG

2002 (Jan)

Sanwa bank, Tokai Bank and Toyo Trust

UFJ

2002 (April)

Dai-Ichi Kangyo Bank, Fuji Bank, Industrial Bank of Japan and

Mizuho

Yasuda Trust 2003 (March)

Asahi Bank and Daiwa Bank

Resona

2005 (Oct)

UFJ and MTFG

MUFJ

4

Page 4 of 27

It remains to be seen how effective such a cross tie-up would be in improving the efficiency, competitiveness and profitability of Japan’s financial industry. In particular, based

ip t

on the changing outlook of the financial environment described above, it is necessary to further analyze whether those reforms have enhanced the Japanese financial institution’s

cr

technical and scale efficiencies as well as any change in returns-to-scale following those

us

tie-ups and mergers. For investigating efficiency, we use traditional but intensively used non-parametric approach named data envelopment analysis (DEA), which was originally

an

developed by Charnes, Cooper and Rhodes (1978) to evaluate the efficiency of public sector non-profit organizations (e.g., education sector). The DEA efficiency is a measure of

M

performance of an organization. There are voluminous studies that applied DEA in banking industry. Existing studies including, Fukuyama (1993), Altunbas et al. (2000), Fukuyama

d

(2000), Sueyoshi (2001), Fukuyama and Weber (2002), Drake and Hall (2003), Drake, Hall

te

and Simper (2009) and Avkiran and Morita (2010) use DEA to measure the efficiency and returns-to-scale of Japanese financial institutions. Other recent studies including Barros,

Ac ce p

Managi and Matousek (2012) and Yang and Morita (2013) also investigate the efficiency of Japanese banks using DEA and other methodologies. In this paper, we mainly focus on DEA efficiency measures followed by a robustness

check with Malmquist indices. Using DEA, we measure three kinds of efficiency. These are technical efficiency (TE), pure technical efficiency (PTE) and scale efficiency (SE). In addition to these efficiency measures, we also focus on whether a particular financial institution exhibits variable returns-to-scale or constant returns-to-scale. The technical efficiency examines the operational efficiency (OE) of a bank or financial institution. Pure technical efficiency (PTE) also evaluates the operational efficiency. But unlike the TE, the PTE takes into scale effect and variable returns-to-scale. Scale efficiency depends on the size of a bank. A bank can achieve higher scale efficiency with mergers, acquisition and 5

Page 5 of 27

expansions. The analysis includes city banks, regional banks and trust banks. In total, we focus on 75 banks and financial institutions to analyze the efficiency changes in the financial

ip t

industry. Further, since the central issue of this paper is to analyze the impact of the policy changes on the efficiency, we calculate the efficiencies in two financial years: 2002 (or 2003)

cr

and 2005 (or 2006), where the former is used as the benchmark. Measuring efficiency of two

us

years yields a total of 150 DMUs.

The DEA analysis shows a substantial improvement in the overall efficiency of

an

Japanese banks albeit a significant difference of efficiency scores between the major/city banks and the regional banks. To corroborate DEA findings, we calculate Malmquist indices

M

corresponding to the efficiency scores of DEA and, total factor productivity (TFP) change. These indices also show an improvement of the efficiency and productivity change thus

d

supporting the views that the FSA’s policy changes have brought about the expected outcome

te

on the performance of the Japanese banks.

The remainder of the paper is organized as follows. Section 2 discusses the factors

Ac ce p

that affect the efficiency of financial institutions. Section 3 discusses the data and methodology. Section 4 reports the empirical findings and analysis thereof. Section 5 concludes this paper.

2. Factors affecting performance and efficiency of financial institutions: A review of literature

There are several factors that affect or enhance the overall performance of the banks

and financial institutions. Berger and Mester (1997), Unite and Sullivan (2003), Patti and Iimi (2004), Hardy (2005), Williams and Nguyen (2005) and Drake, Hall and Simper (2006) find that the reforms in the banking sector enhance the performance of the banks. Further, the improvement of overall performance enhances the competition and efficiency in the banking

6

Page 6 of 27

industry as a whole. Prior studies indicate that following factors can affect the efficiency of financial institutions of a country: service quality [Roth and Jackson III (1995)], effective institutional framework [Eichengreen and Iversen (1999)], choice of human resource,

ip t

technology and process management techniques, locational advantage and the product complementarities [Clark (2002)], moral hazard and even macroeconomic factors [Park and

cr

Weber (2006)]. The recent Japanese experience provides the evidence that the institutional

us

framework accompanied with bankruptcy law and the related ordinance to deal the problems of non-performing loans and their effective implementation can help the financial supervisory

an

agency of a country to improve banks’ efficiency. Moreover, an efficient (financial) institution reduces the transaction cost related to the non-performing/problem loans.

M

While moral hazard is not a serious concern in Japan, transparency is sometimes a problem for Japanese financial institutions. The regional banks are found to not clearly

d

disclose their financial information, particularly the status of bad loan (non-performing loan).

te

Even the major banks are sometimes alleged to have concealed the extent of bad loans. Consequently, in 2005 FSA (Financial Supervisory Agency) issued business improvement

Ac ce p

order to one of the mega banks for not disclosing the real figure of the non-performing loans. Financial reform is an important factor leading to cost and profit and overall

efficiency. Patti and Hardy (2005), for instance, find that financial consolidation, financial deregulation or liberalization, privatization and structural changes improve the productivity and efficiency of financial sectors. Kumbhakar and Lozano-Vivas (2005) find that deregulation results in higher productivity. Casu, Girardone and Molyneux (2004) and Williams and Nguyen (2005) show that productivity growth is determined by technological progress. The liberalization could also bring negative result on banks’ efficiency. For example, Bauer, Berger and Humphrey (1993) and Berger and Humphrey (1997) provide the evidence that financial deregulation raises bank costs which offset potential productivity gains. Iimi 7

Page 7 of 27

(2004) argues that very rapid industrialization and uniform liberalization of the banking industry may not bring the optimal results as expected but lead to a financial catastrophe. Although in many instances profit (and revenue) performance improves more than cost

ip t

efficiency, liberalization does not necessarily lead to a reduction in the dispersion of relative efficiency (Patti and Hardy, 2005).

cr

A large number of industrialized, developing and transition economies have

us

liberalized their banking and financial systems over the past few years (Fanelli and Medhora (1998)). In the early 1990s, financial liberalization programs were popular in some Asian

an

countries like Indonesia, Korea, Malaysia, the Philippines and Thailand. Unfortunately, the period experienced one of the worst financial crises (i.e., Asian Financial Crisis) in the region.

M

As a result, there are mixed views from among the academics, finance practitioners and policy makers that the deregulation can result in efficiency changes in banks. Moreover,

d

banking sector conditions prior to deregulation is also important in determining future bank

te

performance (Berger and Humphrey (1997)). Drake et al. (2006) show that any unfavorable macroeconomic policy changes pertinent to the financial industry adversely affects the

Ac ce p

efficiency of the banks.

Up until 1970s, the postwar Japanese financial system was heavily regulated. But

due to the external economies, Japan needed to adopt deregulation measures since the mid-1970s. The deregulation was widespread particularly during the period 1980-96 (Shimizu (2000)). This widespread deregulation later caused significant banking problems to the major economies including Japan (Posen, 2000). The liberalization in the 1980s was associated with the removal of binding portfolio constraints. This allowed banks and other depository institutions to take riskier investment and loan portfolios and hold high loan to value ratios. This resulted in huge non-performing assets (NPA). Nonetheless, financial deregulation was important for the Japanese economy but the deregulation measures lacked an effective implementation. In most cases, the deregulation in 8

Page 8 of 27

Japan was lopsided and very slow and, mostly associated with the conservative approach remaining unchanged (Hoshi and Patrick (2000)). For example, the deregulation approach did not bring much free entry, new institutions and vigorous innovative activity. However, Hoshi

ip t

and Patrick (2000) contend that deregulation measures strengthened the process of transformation from the bank-centered financial system to the market-oriented one. Together

cr

with other changes, following measures were implemented: relaxation of interest rates, easy

us

flow of international capital, expansion of the activities of financial institutions, introduction of new/alternative financial instruments. These measures enabled large corporate industrial

an

borrowers to raise funds by domestic and foreign bond markets and other capital issues instead of entirely depending on bank based financing. In this course, large banks, losing

M

their huge clients, began lending to small and medium size enterprises and to real estate on a speculative basis that resulted in stock and land price bubbles in the late 1980s. Due to

d

simultaneous bursts of the bubbles in stock and land price, the small and medium sized firms

te

had difficulty paying back their loans and thus the non-performing assets (NPA) of the banking system rose rapidly. Following this, the FSA has taken several strict measures to

Ac ce p

reduce NPA/NPL.

The structural reforms, for example, governance issues (management and ownership

structure, like cross-holding in the case of Japan) also accompany some impacts on the performance of the financial institution. Altunbas et al. (2000) argue that reforms like encouraging intense competition in the sector or tightening of prudential regulations (on recognition of bad loans, loan loss provision and disposal of bad loans/assets) are likely to change the productivity of banks. To summarize, there are many factors that determine the level of efficiency of a financial institution. While the purpose of this study is not to examine the role of each factor on the efficiency gain, this paper examines whether the efficiency of Japanese financial institutions have altered after the implementation of the reforms as discussed in section 1. 9

Page 9 of 27

Hence, our hypothesis is simple. We expect that there is a positive impact of policy changes on the efficiency of Japanese financial institutions. This is the major benefit of the DEA that allows one to employ a model free from assumptions in investigating the efficiency changes.

ip t

3. Data and methodology

cr

3.1 Data

To facilitate the empirical analysis, we use the following input and output variables.

us

Our input variables are: (1) non-performing loan (NPL) (2) deferred tax assets (DTA) and (3) general and administrative expenses and fixed assets, while our output variables are: (1)

an

liquid assets and securities (2) loans (net of loan loss provision) (3) fees and other income. The selection of these inputs and outputs are guided by the hypothesis related to policy

M

changes as well as by the prior literature on Japanese banks efficiency including that of

Drake and Hall (2003).

d

Fukuyama (1993), Altunbas et al. (2000), Sueyoshi (2001), Fukuyama and Weber (2002) and

te

It is to be noted that most prior studies take deposit as an input. We argue that given

Ac ce p

the low interest rate environment, deposit is not a problem for Japanese banks. Banks do not need to offer any incentives to encourage the people to make deposits and hence, deposits do not attract any direct or indirect costs. Instead, we take NPL and DTA. The reasons to include these variables are obvious and adhere to the objectives of this paper. Inclusion of NPL is motivated by FSA’s pressure on the resolution of the non-performing loans (NPLs), while the inclusion of DTA is followed by an allegation that the regional banks show excessive DTA (Deferred Tax Assets) to inflate their capital. Thus, instead of taking deposit as input, taking NPL and DTA would better reflect our objectives to investigate the changes in efficiency and returns-to-scale that are attributed to measures and policies related to NPL and DTA. The selection of outputs, however, is consistent with the prior literature and related to the objectives of this study. To explain it further, we take “Loan” as one of the outputs to

10

Page 10 of 27

examine the lending performance of the regional banks. Some studies argue that inefficient banks need to emphasize on increasing the fees and other business income. Accordingly, we take “Fees and other income” as an output to examine whether the regional banks are

ip t

increasingly focusing on the transaction banking while keeping their traditional relationship banking unaffected.

cr

We focus on 75 banks and financial institutions meaning that almost all financial

us

institutions are included to analyze the efficiency and returns-to-scale changes in the financial industry. Since the central issue of this paper is to analyze the impact of the changes in

an

policies and market conditions on the efficiency of the Japanese financial institutions, we calculate the efficiencies in two financial years: either 2002 (or 2003) and 2005 (or 2006),

M

where the former is used as the benchmark and the latter is used to examine the changes in efficiency and returns to scale. Including UFJ Holdings as a separate entity, we have 150 [=

d

(75 x 2)] decision making units (DMUs). The data are collected from the Nikkei NEEDS

te

database and Japanese Bankers Association.

While more emphasis is given on the regional banks, the study also covers the city

Ac ce p

banks, trust banks and non-bank financial institutions. The major reason to put emphasis on the regional banks is that these banks play a great role in the development of capital market for small and medium size enterprises. The lack of profitability and financial soundness of regional banks adversely affects these enterprises. The regional banks operate in principal cities of the prefectures and provide bulk of their loans to small and medium size companies. These banks also invest in the local stock market and lend to the local money market. Such lending and investment peaked in the late 1980s and contributed to the stock price bubbles and burst. It is worth noting that both the city and regional banks were burdened with huge non-performing assets/loans following the burst of the bubble. This is why it is important to measure the efficiency and returns to scale of these banks with special focus on the impact of policy changes relating to dealing with non-performing loans that accumulated after the burst 11

Page 11 of 27

of the bubble.

3.2 Methodology

ip t

3.2.1 Selection of input-output matrix for DEA analysis The selection of input-output matrix is motivated by the degree of freedom

cr

requirement for the DEA estimation such that the number of observations or , where i = number of inputs and o = number of outputs.

DMUs

us

To estimate the efficiency, of the several DEA methods, we use CCR and BCC, named after Charnes et al. (1978) and Banker, Charnes and Cooper (1984), respectively.

an

While the CCR model is used to examine the technical efficiency (TE), the BCC model is used to examine the pure technical efficiency (PTE). The scale efficiency (SE) is measured as

d

discussed in the next sub-sections.

M

the ratio of CCR to BCC efficiency scores, respectively. The CCR and BCC models are

te

3.2.2 Measuring TE with CCR

Let us define the efficiency of a financial institution in a particular year (every year

Ac ce p

as separate DMU) as follows:

(1)

Efficiency =

According to the CCR model, the virtual inputs and outputs can be calculated such and the virtual output

that the virtual input

determine the respective

weights to maximize the

(2)

Efficiency score, where,

and

are the weights of inputs and outputs, respectively;

quantities of input and output variables, respectively. A DMU is efficient if exists at least one optimal (

,

) with

and

12

and

are the and there

, otherwise the DMU is said to

Page 12 of 27

be inefficient (

). Based on the input-output matrix (X, Y), the linear programming for input multiplier (weights) and row vector

(LP) problem, with row vector

as output

multiplier (weights) can be formulated as:

ip t

(Primal LP) max

cr

subject to

us

(3)

an

The above primal LP problem can be transformed into the dual form with a real variable

of variables as follows

M

and a non-negative vector

CCR, min

d

(Dual LP)

te

subject to

Ac ce p

(4)

The primal-dual correspondence is shown in Table 2.

13

Page 13 of 27

Table 2: The Primal-Dual Transformation This table shows the primal-dual correspondence of the linear programming problems for measuring TE under CCR. PLP stands for primal linear programming problem and DLP stands for dual linear programming problem. Dual variable (DLP)

Constraint (DLP)

Primal variable (PLP)

ip t

Constraint (PLP)

cr

The DLP gives a feasible solution  = 1 (cannot be greater than 1 and must be

and output shortfalls

terms of input excesses

,

for any feasible solution (

LP solution:

te

ii) Solving for max

of (DLP). This requires a two-phase

d

i) Solving for DLP and

(5)

M

with

and identify them as ‘slack’

an

vector by

us

greater than *), o = 1,  = 0 (j  0), i.e. 0 < *  1. The inefficiency can be expressed in

Ac ce p

subject to

and

where, e = (1,…..,1) (a vector of 1) so that

(6) and

. The

objective term can be replaced with any weighted sum of input excesses and output shortfalls so that

. An optimal value (*, S*, S*) is called max-slack solution,

which is referred to as zero slack. A DMU having zero slack is called CCR efficient. 3.2.3 Measuring PTE

14

Page 14 of 27

To measure pure technical efficiency (PTE) of each financial institution, we follow Banker et al. (1984) or BCC model. We formulate an input oriented BCC model for every year of a financial institution by solving the following (envelopment form) linear

ip t

programming problem:

cr

BCC, min

(7)

, a scalar, indicates the BCC or PTE efficiency score. The dual multiplier form of

M

where,

an

us

subject to

the above linear programming problem can be expressed as:

d

max

Ac ce p

te

subject to

,

free in sign

indicates the returns-to-scale. Assuming that

(8) is on the efficient

frontier the following conditions show the returns-to-scale at this point, i) Increasing returns-to-scale prevails at

if and only if

for all optimal

if and only if

for all optimal

solutions.

ii) Decreasing returns-to-scale prevails at solutions. iii) Constant returns-to-scale prevails at

if and only if

for some optimal

solutions. 3.2.4 Measuring SE

15

Page 15 of 27

Once obtaining TE and PTE scores, we compute the SE by a ratio of technical efficiency and the pure technical efficiency such that

; where,

is CCR

is BCC efficiency score. Similar to other types of efficiency, the

SE also cannot be greater than 1. Using these concepts, relationship or,

. This allows us to decompose the

cr

explained as

can be

ip t

efficiency score and

us

overall/global technical efficiency into pure technical efficiency and scale efficiency. The decomposition of TE illustrates the sources of inefficiency, i.e., whether it is due to inefficient

scale efficiency (SE) or by both.

M

3.2.5 Most productive scale size (MPSS)

an

operation locally (PTE, i.e. BCC efficiency) or disadvantageous conditions displayed by the

The decomposition of TE into PTE and SE not only discovers the sources of

d

inefficiency but also tells whether a DMU is operating at the most productive scale size

te

(MPSS). A bank/DMU is treated to be operating at the most productive scale size (MPSS) if it is fully efficient (100%, or a score of 1) in both the CCR and BCC scores. If a bank/DMU

Ac ce p

has the full BCC efficiency but a low CCR score, then it is operating locally efficiently but not globally efficiently due to the scale size of the DMU (see also, Cooper, Seiford and Tone (2000), p. 136)).

4 Results and analysis

4.1 Impact of policy changes on efficiency and returns to scale Table 3 shows the descriptive statistics of the efficiency scores estimated through DEA. The table suggests that there is a clear evidence of improvement in efficiency in 2006. The average TE score in 2002 is found to be 0.82, while in 2006 this score is 0.85, clearly indicating an improvement in the technical efficiency of the banks and financial institutions.

16

Page 16 of 27

Likewise, the pure technical efficiency has also improved in 2006 (0.94) compared to 2002 (0.91), but interestingly, the scale efficiency has remained unchanged (i.e., 0.90) throughout the study period. This indicates that the M&A may not be a solution to resolve the banking

ip t

sector problem or not a way to improve the efficiency of Japanese banks and financial institutions. Or, it may be too early to evaluate the impact of M&A on the efficiency changes.

cr

Cooper et al. (2000) use the DEA-based scale elasticity measurement in estimating the impact

us

of merger on Japanese banks taking 11 regional banks and 9 city banks using both CCR and BCC models to compare the results of scale efficiencies. CCR results indicate that the

an

regional banks performed worse than the city banks when evaluated on the constant returns-to-scale (CRTS). They conclude that by merging two local (BCC) and efficient banks

M

together, it would increase the returns-to-scale. However, the increase in the returns-to-scale characteristics would result the merged bank becomes an inefficient local bank. Tone (1996)

d

also argues that “when two locally (BCC) efficient DMUs merge to form a new DMU, the

te

new DMU is neither locally (BCC) nor globally (CCR) efficient, if increasing returns-to-scale prevails at all three DMUs.”

Ac ce p

Table 3 shows that , of the other descriptive statistics, the median scores for TE, PTE

and SE are found to be 0.86 (0.87), 0.94 (0.98) and 0.95 (0.96) respectively in 2002 (2006), while the minimum scores of the TE, PTE and SE are 0.44 (0.48), 0.54 (0.67) and 0.44 (0.48) respectively in 2002 (2006). These findings clearly reveal the improvement of various measures of efficiency in Japanese financial institutions following the reforms in this sector.

17

Page 17 of 27

Table 3: Descriptive statistics of efficiency scores This table shows the efficiency scores obtained from DEA as discussed in equations 1-8. TE_02 (TE_06) indicates TE for 2002 (2006), PTE_02 (PTE_06) indicates PTE for 2002 (2006) and SE_02 (SE_06) indicates

ip t

SE for 2002 (2006). TE_06

PTE_02

PTE_06

SE_02

SE_06

Mean

0.818133

0.852667

0.905600

0.943467

0.902533

0.904667

Median

0.860000

0.870000

0.940000

0.980000

Maximum

1.000000

1.000000

1.000000

1.000000

Minimum

0.440000

0.480000

0.540000

0.670000

Std. Dev.

0.158248

0.137402

0.104238

Skewness

-0.48

-0.63

-1.17

Kurtosis

2.010344

2.381351

3.924035

Jarque-Bera

5.894685

6.214789

Probability

0.052479

Sum

0.960000

1.000000

1.000000

0.440000

0.480000

0.077206

0.126492

0.126463

-1.44

-1.54

-1.43

4.363494

4.908210

4.188912

19.64375

31.59010

40.90174

30.12220

0.044717

0.000054

0.000000

0.000000

0.000000

61.36000

63.95000

67.92000

70.76000

67.69000

67.85000

Sum Sq. Dev.

1.853139

1.397067

0.804048

0.441099

1.184019

1.183467

Observations

75

75

75

75

75

75

d

an

us

0.950000

M

cr

TE_02

te

Table 4 shows the correlation matrix of the efficiency scores. It shows that there is a

Ac ce p

significant positive correlation between the TE and PTE and, the TE and SE. However, there is no significant correlation between PTE and SE. Instead PTE (2006) is negatively correlated with the SE of 2002 and 2006. It implies that global efficiency is more important than the local efficiency and hence, targeting local efficiency may result in reduced scale efficiency or vice versa. So, obtaining global efficiency is more important than the local efficiency to improve the scale efficiency of Japanese financial institutions.

Table 4: Correlations between efficiency scores This table shows the correlation among various DEA efficiency scores. TE_02

TE_06

PTE_02

PTE_06

SE_02

SE_06

TE_02

1.00

0.89

0.67

0.39

0.80

0.76

TE_06

0.89

1.00

0.49

0.44

0.80

0.85

18

Page 18 of 27

0.67

0.49

1.00

0.73

0.09

0.13

PTE_06

0.39

0.44

0.73

1.00

-0.07

-0.08

SE_02

0.80

0.80

0.09

-0.07

1.00

0.93

SE_06

0.76

0.85

0.13

-0.08

0.93

1.00

ip t

PTE_02

Table 5 shows the efficiency scores, return-to-scale and the impact of changes in

cr

regulation and market conditions. The score card indicates that the number of technically efficient (TE) banks have increased from 16 (in 2002) to 18 (in 2006) and the number of pure

us

technically efficient (PTE) banks have increased from 23 (in 2002) to 36 (in 2006), indicating a substantial progress in both measures of efficiency. Another important observation is that

an

the measure “most productive scale size (MPSS)” also has increased from 16 to 18. Table 5: Impact of policy changes on efficiency measures for 2002 and 2006 PTE efficient

SE efficient

MPSS

M

TE efficient

16

23

24

16

2006

18

36

24

18

Statust quo

Negative

Positive

12

16

47

te

d

2002

Impact of policy changes

Ac ce p

Table 5 also shows that the policy changes have improved the efficiency of 47 banks and reduced the efficiency of 16 financial institutions. The policy changes have not had any adverse impact on 11 banks and financial institutions, which have maintained their performance throughout the study period. Thus, we argue that the policy changes have improved the efficiency of Japanese financial institutions. Interestingly, however, returns-to-scale have not been changed during the study period. As noted earlier, prior literature finds that mergers could have little influence on the scale efficiency in Japanese financial industry. Moreover, Berger and Humphrey (1997) argue that TE and PTE are more important than the SE. Berger and Humphrey (1997) also argue that the bulk of inefficiency is attributable to PTE rather than SE. Consistent with these arguments, our finding on improvement of PTE efficiency implies that policy measures have brought about positive

19

Page 19 of 27

results to the efficiency of Japanese financial industry. Related to SE, Drake and Hall (2003) find no (or insignificant) improvements in the efficiency from mergers especially of large banks. They find that as compared to regional

ip t

banks, major banks and special types of banks (e.g. LTCB) operate above the minimum efficient scale and can get few benefits from eliminating technical inefficiencies. Because of

cr

such features, they criticize the merger wave in the Japanese financial world in the recent past.

us

They also study size-efficiency relationship, which says that majority of banks can make significant reductions in input usage.

an

In regards to returns-to-scale, we find that city banks exhibit decreasing returns-to-scale (DRTS), while most regional banks show increasing returns-to-scale (IRTS)

M

and few of them show constant returns-to-scale (CRTS). These results are consistent with the prior findings of Fukuyama (1993), Altunbas et al. (2000), Drake and Hall (2003) and Amel

d

et al. (2004). Altunbas et al. (2000) report that regional banks have much scope for increasing

te

efficiency thereby demonstrating IRTS, while city banks show significant diseconomies of scale in particular after merging with the failed or small banks thereby indicating DRTS. The

Ac ce p

returns-to-scale behavior was, however, different in the 1980s and 1990s. For instance, Amel et al. (2004) provide that Japanese banks (of all size) exhibited an increasing returns-to-scale during 1980s through 1990s. In an earlier study, Fukuyama (1993) finds that roughly 93% of Japanese banks exhibit non-constant returns-to-scale (7% shows CRTS), 81% of them operate with increasing returns-to-scale and the rest (12%) shows decreasing returns-to-scale. McKillop, Glass and Morikawa (1996) detect “appreciable scale of economies” in the city banks due to factors like cross-shareholdings between banks and other financial institutions in Japan. They argue that cross-shareholdings provide banks with lower management and monitoring costs for their lending portfolio.

20

Page 20 of 27

4.2 Slacks (input excesses and output shortfalls), weights, referent and projection s

and s  indicate input slacks (excesses) and output slacks (shortfalls)

DMUs do not have slacks, while the inefficient ones have.1

ip t

respectively that a DMU needs to emphasize on gaining full efficiency (θ* = 1). The efficient

Our DEA analysis suggests that of the input slacks it is largely the NPLs and DTAs

cr

that contribute to the inefficiency of the banks. The contribution of general and administrative

us

expenses (our third input variable) to the inefficiency of those inefficient banks is very minimum or negligible. In regards to output augmentation, our DEA analysis reveals that the

an

shortage of all three output variables contribute almost equally to the inefficiency of the less-efficient banks. The inefficient banks have enough room to reduce the NPLs and DTA

M

and augment outputs like liquid assets and securities, loans/investments (net of loan loss provision) and to a lesser degree of fees and other income. This finding is consistent with the

d

relevant measures of FSA related to managing NPL and DTA.

te

The DEA provides other estimates of the efficiency analysis. These are the weights, the referents and the projection for the potential improvement of the inefficient banks and

Ac ce p

financial institutions. The weighted value is obtained by multiplying the weights2 of inputs (and of outputs) by corresponding input values (and output values). The weighted values of input and output can be defined as follows: Input weight, I = xij v(i ) , and Output weight, O = yrju(r)

Reference set (or referent, denoted by lambda, see equations 3-7) is taken only for

the inefficient DMUs, because the lambda value for the efficient DMU works as reference set

1

The slacks values are not provided to conserve the space.

2

Optimal weights ν(i) and u(i) for inputs and outputs are calculated first. ν(o) in our findings corresponds to the

constraints

 n j

j

l

and µ(o) to

 n j

j

 u . In the BCC model where l  u  1

holds, u(o) stands

for the value of the dual variable for this constraint.

21

Page 21 of 27

for itself. It does not require to be evaluated by other DMUs (efficient or inefficient). The larger the value of lambda (see, Eq. 3-7), the more the influential a referent is. In most cases, CCR based efficiency does not discriminate between the reference sets for inefficient DMUs.

ip t

But BCC based efficiency provides slightly different reference sets for the inefficient DMUs. Since with the few exceptions, almost all the major banks are efficient, these banks are found

cr

to be referent for the inefficient banks (DMUs). Some of the regional efficient banks are also

us

found to be referents for the inefficient regional banks. The table for reference set is quite large and hence not provided here.

an

Another important aspect of the DEA methods is that these methods provide the extent of required improvement of the inefficient DMUs (the table of projection is very large

M

and hence not supplied here). The score projection and the percentage (negative in case of input and positive in case of output) explain the level of required improvement of the DMUs.

d

The improvement can be achieved by reducing the input and increasing the projected output.

te

This improvement takes the inefficient DMUs into the efficient frontier. Since the projections of efficient banks remain the same, the DEA analysis does not give the improvement targets

Ac ce p

for these banks.

4.3 Robustness check using Malmquist index (MI) While our main objective of this study is to investigate the efficiency changes, we

also look at the productivity changes. It is argued that a DMU can be productive or performing well, but still it may lack the efficiency (Shu and Lee (2003)). However, for a robustness check and to corroborate our DEA findings, we measure total factor productivity (TFP) changes, technological changes, technical efficiency changes and scale efficiency changes obtained through Malmquist index (MI).3 Similar to DEA, Malmquist indices can be estimated in both input- and output- orientation (Grosskopf (1993)). It is worth noting that the 3

We thank an anonymous referee for suggesting us to include the Malmquist indices results.

22

Page 22 of 27

choice of VRS/CRS has no influence on the calculation of Malmquist indices. Moreover, DEA efficiency score should be equal to 1 for a DMU to be efficient, while any score above 1 indicates an improvement of the efficiency according to Malmquist index. This index shows

ip t

TFP growth from period t to t+1. We use the methodology proposed by Färe et al. (1994) and hence, the methodology is not detailed here to conserve the space.

cr

Table 6: Malmquist index (MI)

This table shows the Malmquist index obtained using Färe et al. (1994). TE stands for technical efficiency, Tech

us

stands for the Technological, PTE stands for the pure technical efficiency, SE stands for the scale efficiency and TFP stands for the total factor productivity. The analysis covers 75 banks (both city and regional banks). Tech change

PTE change

SE change

TFP change

1.106

1.522

1.069

1.035

1.684

an

Mean MI summary

TE change

M

Table 6, which summarises the Malmquist indices for 75 banks (both city and regional) using a balanced panel data for 2002 and 2006, reveals that the all the Malmquist

d

indices are over 1 indicating that the results obtained through DEA are robust. Thus, we

te

conclude that the FSA’s much needed policy changes had had positive consequences on the

Ac ce p

performance of the Japanese banks and financial institutions.

5. Conclusions and policy implications This paper re-examines the technical, pure technical and scale efficiencies and

return-to-scale in details covering almost entire sample of Japanese banks and financial institutions. The objective of this investigation was to measure improvements in efficiency, if any, due to recent policy changes adopted by Japan’s FSA. Inefficiency is a universal phenomenon and it exists because it is not possible for a DMU to operate at the optimal level or on the production frontier perpetually. The same DMU can be fully efficient in a period followed by efficiency or inefficiency in the subsequent period. Our analysis provides this evidence for the Japanese financial institutions. The results suggest that policy changes

23

Page 23 of 27

pertinent to non-performing loans, deferred tax assets and other relevant measures brought about the substantial improvement in the efficiency and returns-to-scale of Japanese banks and financial institutions. For robustness check, the study calculates Malmquist indices for

ip t

the corresponding efficiency scores of DEA. Before an empirical analysis is undertaken, the paper also summarises the key changes in Japan’s financial industry, particularly the changes

cr

that occurred during the 2002 -2004 periods. The efficiency results indicate an overall

us

improvement of the efficiencies of the Japanese financial institutions, while some of the city and regional banks still suffer from inefficiency.

an

The banks, which are not operating at the efficient frontier, are found to have more NPLs and DTAs with lesser extent of general and administrative expenses and fixed assets

M

and require them to augment fees and other income (including gains on sales of stocks and other securities, gains on money held in trust and others). Thus, the inefficient banks have

d

enough room to reduce the NPLs and DTA and augment outputs like liquid assets and

te

securities, loans/investments (net of loan loss provision) and fees and other income. This finding is consistent with the relevant measures of FSA related to managing NPL and DTA.

Ac ce p

Further, we find that banks are using more than optimal resources to produce their desired outputs. Inefficient banks have to reduce the expenses on general and administrative purposes and on the premises and equipment. But reducing the input use or augmenting the output or both is rather a difficult job on the part of policy makers/management of the banks. An input reduction means cutting job to reduce overhead expenses or closing the branches to reduce fixed costs. An output augmentation requires the introduction of a new or alternative financial tool. Given the excessive fixed costs, this is rather challenging for a bank in an efficient and competitive market.

References Altunbas, Y., Liu, M., Molyneux, P. & Seth, R. (2000). Efficiency and risk in japanese banking. Journal of Banking and Finance, 24(10), 1605-1628. 24

Page 24 of 27

Ac ce p

te

d

M

an

us

cr

ip t

Amel, D., Barnes, C., Panetta, F. & Salleo, C. (2004). Consolidation and efficiency in the financial sector: A review of the international evidence. Journal of Banking and Finance, 28(10), 2493-2519. Avkiran, N. K. & Morita, H. (2010). Predicting japanese bank stock performance with a composite relative efficiency metric: A new investment tool. Pacific-Basin Finance Journal, 18(3), 254-271. Banker, R., Charnes, A. & Cooper, W. (1984). Some models for estimating technical and scale inefficiencies in data envelopment analysis. Management science, 30(9), 1078-1092. Barros, C. P., Managi, S. & Matousek, R. (2012). The technical efficiency of the japanese banks: Non-radial directional performance measurement with undesirable output. Omega, 40(1), 1-8. Bauer, P. W., Berger, A. N. & Humphrey, D. B. (1993). Efficiency and productivity growth in us banking. The measurement of productive efficiency: Techniques and applications, 386-413. Berger, A. N., Demsetz, R. S. & Strahan, P. E. (1999). The consolidation of the financial services industry: Causes, consequences, and implications for the future. Journal of Banking & Finance, 23(2-4), 135-194. Berger, A. N. & Humphrey, D. B. (1997). Efficiency of financial institutions: International survey and directions for future research. European Journal of Operational Research, 98(2), 175-212. Berger, A. N. & Mester, L. J. (1997). Inside the black box: What explains differences in the efficiencies of financial institutions? Journal of Banking & Finance, 21(7), 895-947. Casu, B., Girardone, C. & Molyneux, P. (2004). Productivity change in european banking: A comparison of parametric and non-parametric approaches. Journal of Banking & Finance, 28(10), 2521-2540. Charnes, A., Cooper, W. & Rhodes, E. (1978). Measuring the efficiency of decision making units. European Journal of Operational Research, 2(6), 429-444. Clark, G. (2002). London in the european financial services industry: Locational advantage and product complementarities. Journal of Economic Geography, 2(4), 433. Cooper, S., Seiford, L. & Tone, K. (2000). “Data envelopment analysis: A comprehensive text with models, applications, references and dea-solver software”: Kluwer Academic Publishers. Drake, L. & Hall, M. (2003). Efficiency in japanese banking: An empirical analysis. Journal of Banking and Finance, 27(5), 891-917. Drake, L., Hall, M. & Simper, R. (2006). The impact of macroeconomic and regulatory factors on bank efficiency: A non-parametric analysis of hong kong’s banking system. Journal of Banking and Finance, 30(5), 1443-1466.

25

Page 25 of 27

Ac ce p

te

d

M

an

us

cr

ip t

Drake, L., Hall, M. J. B. & Simper, R. (2009). Bank modelling methodologies: A comparative non-parametric analysis of efficiency in the japanese banking sector. Journal of International Financial Markets, Institutions and Money, 19(1), 1-15. Fanelli, J. M. & Medhora, R. (1998). Financial reform in developing countries: McMillan Press, London. Färe, R., Grosskopf, S., Norris, M. & Zhang, Z. (1994). Productivity growth, technical progress, and efficiency change in industrialized countries. The American Economic Review, 66-83. Fukuyama, H. (1993). Technical and scale efficiency of japanese commercial banks: A non-parametric approach. Applied Economics, 25(8), 1101. Fukuyama, H. (2000). Returns to scale and scale elasticity in data envelopment analysis. European Journal of Operational Research, 125(1), 93-112. Fukuyama, H. & Weber, W. (2002). Estimating output allocative efficiency and productivity change: Application to japanese banks. European Journal of Operational Research, 137(1), 177-190. Grosskopf, S. (1993). Efficiency and productivity. The measurement of productive efficiency: techniques and applications, 160-194. Hoshi, T. & Patrick, H. (2000). Crisis and change in the japanese financial system: Kluwer Academic Publishers, p-1. Iimi, A. (2004). Banking sector reforms in pakistan: Economies of scale and scope, and cost complementarities. Journal of Asian Economics, 15(3), 507-528. Kumbhakar, S. C. & Lozano-Vivas, A. (2005). Deregulation and productivity: The case of spanish banks. Journal of Regulatory Economics, 27(3), 331-351. McKillop, D., Glass, J. & Morikawa, Y. (1996). The composite cost function and efficiency in giant japanese banks. Journal of Banking and Finance, 20(10), 1651-1671. Park, K. & Weber, W. (2006). A note on efficiency and productivity growth in the korean banking industry, 1992–2002. Journal of Banking and Finance, 30(8), 2371-2386. Patti, B. d. E. & Hardy, D. (2005). Financial sector liberalization, bank privatization, and efficiency: Evidence from pakistan. Journal of Banking and Finance, 29(8-9), 2381-2406. Roth, A. & Jackson III, W. (1995). Strategic determinants of service quality and performance: Evidence from the banking industry. Management Science, 1720-1733. Shimizu, Y. (2000). Convoy regulation, bank management, and the financial crisis in japan. In R. Mikitani & A. S. Posen (Eds.), Japan’s financial crisis and its parallels to us experience: Institute for International Economics, Washington D.C., September, p. 57. Shu, W. S. & Lee, S. (2003). Beyond productivity—productivity and the three types of efficiencies of information technology industries. Information and software technology, 45(8), 513-524.

26

Page 26 of 27

Ac ce p

te

d

M

an

us

cr

ip t

Sueyoshi, T. (2001). Extended dea-discriminant analysis. European Journal of Operational Research, 131(2), 324-351. Tone, K. (1996). A simple characterization of reterns to scale in dea. Journal of Operations Research Society of Japan, 39, 604-613. Williams, J. & Nguyen, N. (2005). Financial liberalisation, crisis, and restructuring: A comparative study of bank performance and bank governance in south east asia. Journal of Banking and Finance, 29(8-9), 2119-2154. Yang, X. & Morita, H. (2013). Efficiency improvement from multiple perspectives: An application to japanese banking industry. Omega, 41(3), 501-509.

27

Page 27 of 27