Does Financial Deregulation Increase the Banking Efficiency in Bangladesh?
2014 Cambridge Conference Business & Economics ISBN : 9780974211428
Does financial deregulation increase the banking efficiency in Bangladesh?
Iftekhar Robin PhD Research Fellow, Curtin Business School (CBS), School of Economics & Finance, Curtin University, Australia Phone: +61 469 773 842 Email: Iftekhar.Robin@postgrad .curtin.edu.au Ruhul Salim Associate Professor, Curtin Business School (CBS), School of Economics & Finance, Curtin University, Australia Phone: +61 8 9266 4577 Email: Ruhul.Salim@cbs .curtin.edu.au Harry Bloch John Curtin Distinguished Emeritus Professor, Curtin Business School (CBS), School of Economics & Finance, Curtin University, Australia Phone: +61 8 9266 2035 Email: Harry.Bloch@cbs .curtin.edu.au July 1-2, 2014 Cambridge, UK 1
2014 Cambridge Conference Business & Economics ISBN : 9780974211428 Abstract The paper examines the impact of financial reform policy on the cost efficiency of the commercial banks in Bangladesh. Following a translog cost function model, it employs the
single-stage stochastic frontier analysis (SFA) model of Battese and Coelli (1995) to estimate
the cost efficiency with a particular focus on investigating the impact of the financial reform programme. The study uses a unique balanced data set comprising annual banking data from the 12 largest commercial banks in Bangladesh for the period 1983-2012. Findings from this study show that financial deregulation contributes in reducing bank cost although a slightly increasing trend in cost inefficiency has been observed in the post-reform era. The estimated average cost efficiency scores reflect that both public and private sector banks have gained efficiency as a consequence of financial deregulation. Another interesting observation is that political influence on the bank board has negative effect on efficiency. On the other hand, the presence of independent director in the bank board helps reducing cost inefficiency. The findings of this study may contribute in policy decision in the banking sector in Bangladesh since this is the first of its kind estimating the banking efficiency after two decades of reform initiatives. Keywords: Commercial banks, Financial deregulation, Efficiency, Stochastic frontier analysis July 1-2, 2014 Cambridge, UK 2
2014 Cambridge Conference Business & Economics ISBN : 9780974211428
Does financial deregulation increase the banking efficiency in Bangladesh?
ABSTRACT The paper examines the impact of financial reform policy on the cost efficiency of the commercial banks in Bangladesh. Following a translog cost function model, it employs the
single-stage stochastic frontier analysis (SFA) model of Battese and Coelli (1995) to estimate
the cost efficiency with a particular focus on investigating the impact of the financial reform programme. The study uses a unique balanced data set comprising annual banking data from the 12 largest commercial banks in Bangladesh for the period 1983-2012. Findings from this study show that financial deregulation contributes in reducing bank cost although a slightly increasing trend in cost inefficiency has been observed in the post-reform era. The estimated average cost efficiency scores reflect that both public and private sector banks have gained efficiency as a consequence of financial deregulation. Another interesting observation is that political influence on the bank board has negative effect on efficiency. On the other hand, the presence of independent director in the bank board helps reducing cost inefficiency. The findings of this study may contribute in policy decision in the banking sector in Bangladesh since this is the first of its kind estimating the banking efficiency after two decades of reform initiatives.
1. INTRODUCTION
Banking efficiency and productivity growth continue to be important issues in the economics and finance literature, especially at the onset of financial liberalization and globalization of financial markets. From the mid-1970s onwards many developing countries, most notably in Latin America (e.g., Argentina, Brazil, Columbia, Mexico, Uruguay, and Chile) and Asia (e.g., Malaysia, Indonesia, South Korea, Thailand, India, Sri Lanka, Philippines and Pakistan) July 1-2, 2014 Cambridge, UK 3
2014 Cambridge Conference Business & Economics ISBN : 9780974211428 have implemented various Financial Sector Reform Programmes (FSRPs). Bangladesh also initiated financial reform programme in the late 1980s. This process has had a number of phases. The measures that have already been taken under the reform process include the introduction of a market determined interest rate, privatization of state-owned commercial banks and greater freedom for the operation of private sector commercial banks and other financial institutions. In this backdrop, this study is going to examine the impact of financial deregulation on the banking efficiency of the commercial banks in Bangladesh. Cost efficiency has been considered as a performance measure of banks in most of the
banking efficiency literature over the past three decades (e.g.,Boucinha, Ribeiro, & Weyman-
Jones, 2013; Du & Girma, 2011; Ferrier & Lovell, 1990; Fries & Taci, 2005; Humphrey,
different countries, there is no comprehensive study on the banking sector in Bangladesh investigating the impact of financial deregulation in terms of cost efficiency until today. There are few studies on efficiency for individual banks or problem banks in Bangladesh, but these
have been lacking appropriate data and technique (Akther, Fukuyama, & Weber, 2012;
Hassan, 1999; Khanam & Khandoker, 2005; Perera, Skully, & Wickramanayake, 2007).
Therefore, this is the first of its kind in investigating the impact of financial deregulation on the cost efficiency of banks in Bangladesh. Applying the parametric technique, stochastic frontier analysis (SFA), developed by
Aigner et al.(1977) and Meeusen and Van den Broeck (1977), the estimation uses a unique
balanced panel dataset for the period 1983-2012 for 12 major commercial banks in Bangladesh. To estimate the stochastic cost frontier we follow the maximum likelihood (ML)
procedure of Battese and Coelli (1995) model that permits single-stage estimation of the
July 1-2, 2014 Cambridge, UK 4
2014 Cambridge Conference Business & Economics ISBN : 9780974211428 parameters of the cost function and correlates of bank inefficiencies. We do not follow the two-stage procedure due to its limitations. The main caveat of the two-stage analysis is the violation of the assumption made in the first stage that the inefficiency component of the composite error term of the cost frontier is independently and identically distributed. The second-stage involves the specification of a regression model for the predicted technical inefficiency effects, which contradicts the assumption of identically distributed inefficiency
effects in the stochastic frontier (Fries & Taci, 2005).
The remainder of the paper is organized as follows: Section 2 discusses the theoretical and empirical literature on financial deregulation and banking efficiency. Section 3 describes the structure of the financial system in Bangladesh. Section 4 explains research design, empirical model, data sources and variable construction. Section 5 provides the estimation results and interpretation and finally, Section 6 concludes.
2. LITERATURE REVIEW
Research on banking efficiency is date back to the 1960s. Farrell (1957) is the pioneer in
constructing the measurement of technical efficiency in terms of realized deviations from an idealized, frontier isoquant. In an increasingly competitive environment, efficiency and productivity of financial institutions has become critically important. As such, there is growing literature on the efficiency of financial institutions. Efficiency can be measured by comparing observed output to maximum potential output obtainable from given inputs, or comparing observed inputs to minimum potential
(1957), following the ideas of Debreu (1951) and Koopmans (1951). This work provides a
measure of total economic efficiency containing two components: technical efficiency and July 1-2, 2014 Cambridge, UK 5
2014 Cambridge Conference Business & Economics ISBN : 9780974211428 allocative efficiency. Technical efficiency reflects the ability of a firm to produce possible output from a given set of inputs and technology, and allocative efficiency reflects the capacity of a firm to use inputs in optimal proportions, given their respective prices and
production technology, i.e., equate marginal value products with marginal input cost (Coelli,
Rao, O' Donnell, & Battese, 2005; Heshmati, 2003).
There are several difficulties with Farrell’s measure. It measures technical efficiency relative to an isoquant rather than to an efficient subset, which may identify a decision making unit (DMU) as technically efficient when it is not. Moreover, it is a radial measure as it assumes a given input mix although there is no reason to measure technical efficiency radially, even for homothetic technologies. Furthermore, Farrell’s restrictive assumptions on
the departure of the actual cost shares from the optimum shares. The definition does not explain the relationship between allocative inefficiency and increase in cost from such
inefficiency. Bauer (1990) termed this problem as the ‘Greene problem’. However,
Kumbhakar (1997) identifies both allocative and technical inefficiency in a cost-minimizing
framework, and establishes an exact relationship in cost share equations as well as in the translog cost function. The empirical studies on deregulation and banking efficiency provide mixed results. For instance, banking efficiency in the U.S. has remained relatively unchanged after the
Cambridge, UK 6
2014 Cambridge Conference Business & Economics ISBN : 9780974211428
initial stage of deregulation; achieved an improvement at the later stage (Grifell-Tatj`e &
Lovell, 1999; Kumbhakar, Lozano-Vivas, C.A.Knox Lovell, & Hasan, 2001). However,
industry conditions prior to deregulation may explain these unexpected consequences (Berger
& Humphrey, 1997). In contrast, Isik and Hassan (2003) finds substantial improvement in
productivity in Turkish commercial banking after deregulation. Norwegian banks’
The impact of deregulation on banking efficiency varies with the size and ownership structure. Thai small banks prospered less compared to other Thai banks after financial
liberalization (Leightner & Lovell, 1998). Drake et al. (2003) find that large banks in Japan
are generally operating above the minimum efficient scale, and opposite results are found for the smaller banks. Indian medium-sized public sector banks performed reasonably well, and are more likely to operate at a higher levels of technical efficiency during post-reform period
efficient employing two competing approaches DEA and SFA for 70 Indian commercial banks during the period 1986-1991. In a deregulated Australian financial environment, establishment of new banks (both
domestic and foreign) provides an important contribution towards efficiency gains (Sturm &
Williams, 2004). Sathye (2001) observed low levels of overall efficiency in commercial banks
in Australia compared to the banks in Europe, and in the USA. The relative technical efficiency or variations in efficiency levels have been observed
due to the influence of environmental factors. Ariff and Can (2008) investigates the sources of
July 1-2, 2014 Cambridge, UK 7
2014 Cambridge Conference Business & Economics ISBN : 9780974211428 bank inefficiency of 28 Chinese commercial banks considering the influence of ownership type, risk profile, bank size, profitability and key environmental changes on the banking efficiency employing Tobit regression. The authors find that joint-stock banks (national and city-based) on average appear to be more cost and profit efficient than state-owned banks; while medium-sized banks are significantly more efficient than small and large banks. Environmental conditions contribute significantly to the difference in efficiency scores between countries. The cost-efficiency scores of Spanish banks are quite low compared to those of the French banks due to the exclusion of environmental variables from the
specification of the common frontier (Dietsch & Lozano-Vivas, 2000).
Until today, there is hardly any comprehensive study examining the impact of financial deregulation on the efficiency and productivity of the banking sector in Bangladesh since the financial reform programme initiated in the late 1980s. However, there are few studies on efficiency for individual banks or problem banks in Bangladesh, but these have
been lacking appropriate data and technique (Akther et al., 2012; Hassan, 1999; Khanam &
Khandoker, 2005; Perera et al., 2007).The paucity of empirical studies on banking efficiency
in Bangladesh justifies the scope and originality of this research. There is ample opportunity to explore all possible investigations in such an area which remains unexplored for long time.
3. FINANCIAL SECTOR IN BANGLADESH
Since the independence in 1971, Bangladesh has experienced a variety of development approaches in different economic and political regimes. The command economy structure prevailing in the 1970s, the administrative price setting practices lacked flexibility and responsiveness to relative scarcities with attendant inefficiency in resource allocation. Low administered interest rates on savings in the inflationary environment discouraged financial savings and retarded financial intermediation. However, the negative effect of command July 1-2, 2014 Cambridge, UK 8
2014 Cambridge Conference Business & Economics ISBN : 9780974211428 economy regime had on the financial sector been recognized by the late1970s. Therefore, the interest rates rationalization of 1980 with general upward revision, licensing new private banks and privatization of two state-owned commercial banks in the early 1980s were significant but piecemeal and ad hoc reform steps taken by the Central Bank and the Government. After several reviews on the financial sector of Bangladesh, ‘Financial Sector Reform Program (FSRP)’ was implemented during 1989-95, supported by technical assistance from the USAID and the IMF and a balance of payments assistance loan from the IDA. The programme addressed issues on broad fronts, including transition from directed sectoral lending at directed interest rates to unified credit markets with market-based interest rates, transition to indirect tools for monetary management, revision of loan classification and provisioning criteria, revision of legal provisions and procedures for enforcing loan recovery, availability of credit information for loan risk assessment, transition from segmented exchange markets with multiple exchange rates to unified foreign exchange market with a single market-clearing exchange rate and up-gradation of technology and human resources skills in banks. The financial system of Bangladesh constitutes commercial banks, development banks and financial institutions (FIs), co-operative banks, microfinance institutions (MFIs), insurance companies, credit rating agencies and two stock exchanges. While Bangladesh Bank has regulatory and supervisory jurisdiction over the entire banking sector, the Bangladesh Securities and Exchange Commission (BSEC) exercises similar functions for the stock exchanges and the merchant banks 1 . July 1-2, 2014 Cambridge, UK 9
2014 Cambridge Conference Business & Economics ISBN : 9780974211428 There are four categories of banks operating in Bangladesh. These are: four state owned commercial banks, five government-owned development banks dedicated to agriculture, small and medium enterprises (SMEs), housing and industrial lending, 39 private sector commercial banks, of which nine foreign commercial banks and eight Islamic banks based on Islamic Shariah. Out of the 31 non-bank financial institutions, only two have significant government ownership, and the rest are in the private domain. In April 2012, the government approved nine new private sector commercial banks and they came into operation in 2013. Apart from this, 580 microfinance institutions including Grameen Bank (the largest microfinance institution) have been providing microcredit to the hardcore poor especially
.
4. RESEARCH DESIGN
Following the bank efficiency literature, we analyse the cost efficiency estimates for three time periods, 1983-1990 as the pre-reform period, 1991-1995 as the transition period and 1996-2012 as the post-reform period in order to investigate whether banking reform policies
2012 because it is assumed that efficiencies do not fluctuate markedly over short periods of time. Several other studies also suggest that efficiency is reasonably persistent over time
(Berger & Humphrey, 1991; Eisenbeis, Ferrier, & H, 1996). Furthermore, a relatively long
period is required for reform initiatives (e.g., regulatory changes) and other macro-financial
developments to exert their influence upon the banking technology (Isik & Hassan, 2002).
Nevertheless, we have tried to estimate separate cost frontiers for each of the three periods, however, separate frontier estimates do not provide us with any meaningful results due to less July 1-2, 2014 Cambridge, UK 10
2014 Cambridge Conference Business & Economics ISBN : 9780974211428
number of observations. Fries and Taci (2005) suggest that there is minimum requirement of
data period to distinguish reliably between random noise and bank inefficiency.
4.1 Estimation method
A variety of functional forms (of SFA) such as Cobb-Douglas, transcendental logarithmic (translog), generalized Leontief, constant elasticity of substitution (CES) have been followed in the banking literature to estimate cost efficiency. Each functional form has merits and limitations. For example, Cobb-Douglas forms are first-order flexible, while the other functional forms are second-order flexible. A preferred specification for estimating efficiency can be determined by conducting a residual analysis and comparing different efficiency models with respect to statistical significance. All other things equal, we usually prefer functional forms that are second-order flexible, although increased flexibility comes at the cost of more parameters to estimate, and this may lead to econometric difficulties,
multicollinearity for example (Coelli et al., 2005).
The SFA specifies a translog form of cost function for estimation. It assumes a composite error term that contains inefficiencies. The inefficiency component of the error term follows an asymmetric distribution (usually a truncated or half normal distribution) and the random component of the error term follows a symmetric distribution (usually standard normal distribution). The key reason for such structure of the composite error term is that, by definition, inefficiencies cannot be negative. Further, both inefficiencies and random error are assumed to be orthogonal to input prices, outputs and bank-specific variables as specified in
the cost function (Fries & Taci, 2005).
Following Coelli et al.(2005) and Wang and Kumbhakar (2009) the translog cost
frontier model can be written as: July 1-2, 2014 Cambridge, UK 11
2014 Cambridge Conference Business & Economics ln
C it
0 1 / 2
n j
M
m
1
nj
ln
m
ln
w nit
ln
y mit w jit
n N
n
ln 1
m n w nit
mn
ln 1 /
y mit
2
m k
ln
w nit mk
ln 0
t y mit
ln
y kit
1 / 2 00
t
2
m
mt
ln
y mit t
n
nt
ln
w nit t
Q
q
q z qit
it
(
v it
u it
) ISBN : 9780974211428 m ,k=1,………,M; n,j=1,………,N; i=1,……….,I; t=1,……..,T (1) where
y mit
is the m-th output, price,
z qit
represents the q-th explanatory variables that affect the total cost, t is time trend accounts for technological change, and , , , , , and are a vectors of unknown parameters, The components of the composite error term
it
(
u it
v it
) , inefficiency and higher bank inefficiency is associated with higher cost, and is assumed to be normally distributed with truncation below zero. The random error
v
is assumed to be distributed independently and identically according to standard normal distribution,
N
( 0 , 2 ) . In estimating the Equation (1), we assume that the cost function is non-decreasing, linearly homogenous and concave in input prices, if
n
are non-negative and satisfy the homogeneity constraint,
n N
1
n
1 and also impose constraint on symmetry,
mk
km
and
nj
jn
.
4.2 Data
This study uses a unique balanced panel data set constructed from the balance sheets, income statements and other financial statements of the sample banks. The sample contains data of 12 largest commercial banks in Bangladesh for 30 years, 1983-2012 consisting of 360 July 1-2, 2014 Cambridge, UK 12
2014 Cambridge Conference Business & Economics ISBN : 9780974211428 observations. We collect the hard copies of annual reports and other financial statements through contacting each individual bank. The aggregate data on the banking sector have been collected from the Central Bank of Bangladesh (Bangladesh Bank). The macro-financial data have been collected from the national statistical department, Bangladesh Bureau of Statistics (BBS), Ministry of Finance, the Government of Bangladesh, Bangladesh Security and Exchange Commission (BSEC), International Financial Statistics (IFS) of the IMF and World Development Indicator (WDI) of the World Bank.
4.3 Definition of inputs and outputs: different approaches
Efficiency predictions require appropriate definition of input, output and input price variables. Specification of inputs and outputs for banks is not straightforward. There is long-standing debate in the banking literature about the specification of inputs and outputs of banking firms.
Favero and Papi (1995) propose five approaches for input-output specification: production,
intermediation, asset, user cost and value-added approach. Sealy and Lindley (1977) first
develop a positive theory for the behaviour of financial institutions. They propose two different views, the technical view and the economic view of financial institutions. They argue that financial firms are involved in financial intermediation through transferring funds from surplus unit to deficit unit. In the banking literature, mainly two competing approaches dominate: the production approach and the intermediation approach. Both approaches follow the application of traditional microeconomic theory of firms to banks and differ only in the
specification of banking activities (Das & Kumbhakar, 2012).
Both production and intermediation approaches have merits and limitations.
According to Berger and Humphrey (1997), the production approach may be more
appropriate for evaluating branch-level efficiency (since this approach deals with customers’ July 1-2, 2014 Cambridge, UK 13
2014 Cambridge Conference Business & Economics ISBN : 9780974211428 transaction accounts, and thus has little influence on overall bank policy), while the intermediation approach is better for measuring the efficiency of banks as a whole. Like many
other studies on banking efficiency (e.g.,Berger & Mester, 1997; Boucinha et al., 2013; Casu,
Girardone, & Molyneux, 2004; Du & Girma, 2011; Isik & Hassan, 2002; Kumbhakar &
Lozano-Vivas, 2005; Resti, 1997; Wang & Kumbhakar, 2007, 2009), we adopts the
intermediation approach to define inputs and outputs to examine the bank-level cost efficiency. The intermediation approach views banks as the mediator of funds between savers and investors. According to this approach, banks transform various financial and physical resources, such as deposits and other liabilities, into interest-earning assets such as loans,
securities and other investments (Sealey & Lindley, 1977). Thus, the intermediation approach
treats both operating and interest expenses as inputs and loans and other assets as outputs.
4.4 Construction and rationale of the variables
Measurement of cost efficiency requires data on total costs, outputs and input prices. The dependent variable is total cost (TC), which includes both interest expenses and operating costs. Following intermediation approach we model commercial banks as multi-product firms
producing two outputs employing three inputs (e.g.,Boucinha et al., 2013). Apart from output
and input prices, we consider some bank-specific control variables and environmental variables both in the cost function and inefficiency function as independent variables. The details about the definition of the variables are reported in Appendix I.
4.4.1 Input and output variables
The input vectors are (1) labour, the number of full-time employees; (2) physical capital, the book value of premises and fixed assets; and (3) loanable funds, the sum of deposit (demand July 1-2, 2014 Cambridge, UK 14
2014 Cambridge Conference Business & Economics by total loanable funds. ISBN : 9780974211428 and time) and non-deposit funds (borrowed funds). All input prices are calculated as flows over the year divided by the corresponding stock: (1) price of labour (
w
1 ) equals total expenditure on employees, such as salaries and allowances, divided by the total number of employees; (2) price of capital (
w
2 ) equals total expenditure on premises and fixed assets, i.e., total operating expenses (except salary and allowances and charges on loan/investment losses) divided by the book value of physical capital and other fixed assets; and (3) price of loanable funds ( The output vectors are: (1) total loans and advances (
y
1 ), which include loans, cash credits and overdrafts and bills discounted and purchased; (2) other earning assets (
y
2 ) that comprise government securities, treasury bills, shares (fully paid), debentures, bonds and other investments (gross total assets less loans and physical capital/fixed assets).
4.4.2 Control variables and correlates of inefficiencies
Apart from inputs and outputs, there is another set of variables that characterizes the operations of individual banks. These control variables may affect banking technology and service quality, which shifts the cost frontier. Several bank characteristics may be the
determinants of efficiencies as well. Kumbhakar and Lovell (2003) suggest that bank-specific
exogenous variables may belong in the frontier along with inputs/input prices and outputs or they may belong in the one-sided error component (inefficiency function), i.e., as the determinants of efficiency or in both locations. In our empirical model, we include several bank-specific variables in the cost function as control variables which may affect the bank cost directly through shifting the cost frontier. These are the level of equity, financial intermediation (the ratio of total loans to total July 1-2, 2014 Cambridge, UK 15
2014 Cambridge Conference Business & Economics ISBN : 9780974211428 deposits), time trend, and deregulation period dummy variable. We also include some of these control variables in the inefficiency function, such as time trend and deregulation period dummy variables. Some other additional variables are also included in the inefficiency function as the determinants of efficiency. The variation in efficiency level may be associated with these factors that affect incentives and/or managerial selection at the bank level. These environmental factors/correlates are bank size, ownership structure, degree of market concentration and board composition: independent and political directors in the bank board.
4.5 Model specification and estimation procedure
Following the methodology described in Sub-section 4.1, we employ a translog form of cost function for the estimation. Appendix II describes the model specification tests for the selection. We consider total cost (TC) as the dependent variable and two outputs,
y
1 and
y
2 , and three input prices
w
1 , w 2 and The theoretical restrictions stemming from duality theory, i.e., symmetry and linear homogeneity in prices are implicitly imposed in the specification of the estimated equation. Since the cost function is homogenous of degree one in input prices, we need to impose parametric restrictions to ensure that the estimated cost function satisfies this property. In practice, linear homogeneity restrictions are automatically satisfied if the cost and input prices are expressed as a ratio of one input price. We use physical capital price,
w
2 , as numeraire. The data are expressed as deviations from the sample mean, dividing each variable by its geometric mean. The translog form represents a second-order Taylor approximation,
around the geometric mean to any generic cost frontier (Orea & Kumbhakar, 2004). The
normalization of variables allows the first-order coefficient of the translog cost function correspond to the elasticity evaluated at the sample mean (the point of approximation). The July 1-2, 2014 Cambridge, UK 16
2014 Cambridge Conference Business & Economics ISBN : 9780974211428 normalization reduces heteroscedasticity and allows banks of any size to have comparable
residual terms (Du & Girma, 2011).In calibrating an appropriate translog cost frontier model,
we estimate four different translog cost frontier models as defined in Equation (1). The detail of the models and estimated results are presented in Appendix III. The translog cost frontier (Eq.1), as constrained to ensure linear homogeneity, can be expressed as: ln(
C it
/
w
2 ) 0 1 ln
y
1 2 ln
y
2 1 ln(
w
1
w
2 ) 2 ln(
w
3
w
2 ) 1 2 11 (ln
y
1 ) 2 12 ln
y
1 ln
y
2 1 2 22 (ln
y
2 ) 2 1 2 11 ln(
w
1
w
2 ) 2 12 ln(
w
1
w
2 ) ln(
w
3
w
2 ) 1 2 22 (
w
3
w
2 ) 2 11 ln
y
1 ln(
w
1
w
2 ) 12 ln
y
1 ln(
w
3
w
2 ) 21 ln
y
2 ln(
w
1
w
2 ) 22 ln
y
2 ln(
w
3
w
2 ) 1
t
1 2 2
t
2 1
t
ln
y
1
t
2
t
ln
y
2
t
1
t
ln
w
1
w
2
t
2
t
ln
w
3
w
2
t
1
EQ
2
FI
3
DTr
4
DPs
it
(
u it
v it
) (2) where, the random error term
v
is assumed to be independently and identically distributed,
N
( 0 , 2 ) and the symmetry conditions,
mk
km
and
nj
jn
are imposed. The theoretical requirements, the monotonicity properties, i.e., non-decreasing in output
y
[ ln
C it
/ ln
y mit
0 ] and non-decreasing in input prices
w
[ ln
C it
/ ln
w nit
0 ] can be examined by the elasticity of outputs and input prices expressed in terms of coefficients of the fitted cost function. The SFA assumes that the inefficiency component
inefficiency is associated with higher cost. The Battese and Coelli (1992,1995)
July 1-2, 2014 Cambridge, UK 17
2014 Cambridge Conference Business & Economics ISBN : 9780974211428 parameterisation of the inefficiency term
u it
is used to allow for time-varying inefficiency where the technical inefficiency effects are assumed to be an exponential function of time:
u it
{exp[ (
t
T
)]}
u i
where, u is assumed to have truncated normal distribution and only one parameter is to be estimated. The inclusion of a time trend into the cost function as well as the inefficiency function permits the estimation of both trend technical change and trend changes in the
technical inefficiencies over time (Cornwell, Schmidt, & Sickles, 1990).
Allowing for influences beyond time, the inefficiency (
u it
) function can be specified as follows:
u it
0 8
CR
3 1
OWN e it it
2
ID it
3
PD it
4
SIZE it
5
DTr it
6
DPs it
7
t
(3) where with zero mean and variance 2 . The point of truncation is,
z it
, i.e.,
e it
z it
. These assumptions are consistent with
N
(
z it
, 2 ) distribution. Therefore, the inefficiency component of the composite error term of the translog cost function (5.2) has a truncated normal distribution, whose point of truncation depends on the bank-specific characteristics so that the inefficiency terms are non-negative. The parameters of both the Equations (2) and (3) can be estimated simultaneously following the maximum likelihood method using the frontier econometric programme
FRONTIER 4.1 developed by Coelli (1996). The likelihood function can be expressed in
terms of variance parameters
s
2
v
2
u
2 and
u
2
s
2 . If equals zero, then the model July 1-2, 2014 Cambridge, UK 18
2014 Cambridge Conference Business & Economics ISBN : 9780974211428 reduces to a traditional mean response function in which
z it
can be directly included into the cost function. A measure of cost efficiency is the ratio of minimum cost to observed cost. The cost efficiency ( CE ) of the translog cost function (5.2) for i-th bank at t-th observation can be
it
defined as follows:
CE it
f
(
y it
,
w it
,
c it
) exp(
v it
)
f
(
y it
,
w it
.
c it
) exp(
v it
) exp(
u it
) exp(
u it
) (4)
5. EMPIRICAL RESULTS AND ANALYSIS
The translog cost function specified in Eq.(2) and the inefficiency function specified in Eq.(3) have been simultaneously estimated following the maximum likelihood (ML) procedure of
Battese and Coelli (1995) that permits single-stage estimation of the parameters of both the
functions. As discussed in Sub-section 4.5, four different SFA models are estimated together with specification tests for choosing the final model appropriate for the sample. ML estimators are popular in empirical work irrespective of the type of model being estimated because of several desirable large sample (i.e., asymptotic) properties if the assumptions
underlying the model are valid (Coelli et al., 2005). For example, if heteroskedasticity appears
in the symmetric noise error component, we still obtain unbiased estimates of all parameters
describing the structure of the production frontier (Kumbhakar & Lovell, 2003). Maximum
likelihood and least squares estimators are still unbiased and consistent in the presence of heteroskedasticity and autocorrelation, provided all the other assumptions of the model are true. Although translog cost function is more flexible than other functional forms, July 1-2, 2014 Cambridge, UK 19
2014 Cambridge Conference Business & Economics ISBN : 9780974211428 multicollinearity may exist among the variables due to second order terms. However, multicollinearity may not be a severe problem as efficiency scores are used for predictions.
2
5.1 Parameter estimates: cost function
The maximum likelihood parameter estimates for both the translog cost function (Eq.2) and inefficiency function (Eq.3) are reported in Table 1. The detailed estimates for all the four models are described in Appendix IV. Since total cost (dependent variable) and input prices and output variables in the cost function are normalized, dividing by their respective sample mean, estimated parameters can be interpreted as cost elasticity evaluated at the sample mean. Table 1 presents the estimated cost elasticity with respect to each output and input price term evaluated at the sample mean. The non-negative coefficients for outputs (y 1 and y 2 ) and input price terms (w 1 /w 2 and w 3 /w 2 ) satisfy the theoretical requirements for a valid cost function. The cost function must satisfy the required theoretical properties: cost is non-
decreasing in outputs and input (factor) prices and concave in factor prices (Coelli et al.,
2005; Rezvanian & Mehdian, 2002).
The estimated positive and statistically significant coefficient for total loans and advances (output y 1 ) suggests that, on average, a one percent increase in the amount of loans and advances increases total costs by about 0.82 percent. Similarly, the estimated cost elasticity of other earning assets (output y 2 ) is 0.34 suggesting an increase in total cost by about 0.34% due to a one percent increase in other earning assets. However, the sum of the estimated cost elasticities for the two outputs is more than one (
e c
1 ), which indicates the
presence of diseconomies of scale (Jha, Murty, & Paul, 1991; Mahesh & Bhide, 2008).
July 1-2, 2014 Cambridge, UK 20
2014 Cambridge Conference Business & Economics ISBN : 9780974211428 The statistically significant and positive magnitude for the estimated cost elasticity of the input price terms are 0.51 and 0.60 respectively for w 1 /w 2 and w 3 /w 2, which indicates that
Table 1: Maximum likelihood estimates of the translog cost frontier Variables Cost function
Intercept Lny 1 Lny 2 Ln(w 1 /w 2 ) Ln(w 3 /w 2 ) Lny 1 Lny 2 Control variables t (time trend)
t
2 z 1 (equity) z 2 (Financial intermediation) DTr (1= transition period, 0=otherwise) DPs (1= post-reform period, 0= otherwise) Inefficiency function Intercept OWN (1=Public, 0=otherwise) ID (1=Independent director in the bank board, 0=otherwise) 0 1 2 0 1 2 1 2 12 1 11 1 2 3 4
Parameters Estimated value
0.1291*** 0.8284*** 0.3432*** 0.5133*** 0.6047*** -0.0076 -0.0011 0.0004 0.0029** -0.1798*** -0.0247 -0.1301*** 1.1008*** -0.0337** -0.0151
Standard error
0.0415 0.0662 0.0749 0.0723 0.0399 0.1081 0.0054 0.0003 0.0014 0.0139 0.0204 0.0298 0.2493 0.0159 0.0715 1.4563 1.9892 -12.8812 -1.2085 -4.3658 4.0422 -2.1214 -0.2115
t-statistics
3.1068 12.5053 4.5777 7.1047 15.1653 -0.0709 -0.1970 July 1-2, 2014 Cambridge, UK 21
2014 Cambridge Conference Business & Economics ISBN : 9780974211428 PD(1=Political director in the bank board,0=otherwise) SIZE
Variables
DTr (1= transition period, 0=otherwise) DPs (1= post-reform period, 0= otherwise) t (time trend) CR3: 3-bank deposit concentration ratio Sigma-squared Gamma 3 0.0933*** 4 -0.0592**
Parameters
5
Estimated value
-0.0794*** 6 7 0.2889*** -0.0277*** 8 2
s
2
v
u
s
2 2 -0.5270***
u
2 0.0017*** 0.5492*** 0.0135 0.0236
Standard error
0.0301 0.0527 0.0058 0.1874 0.0003 0.1124 6.9066 -2.5096
t-statistics
2.6346 5.4767 -4.7896 -2.8127 6.7497 4.8809 Note: Log-likelihood function=701.2986; LR test of the one-sided error= 950.5364; total number of observations
*** denotes statistical significance level at 1% ** denotes the level of statistical significance at 5% * denotes statistical significance level at 10% Source: Author’s calculation July 1-2, 2014 Cambridge, UK 22
2014 Cambridge Conference Business & Economics ISBN : 9780974211428 total cost is very sensitive to both the labour price (w 1 )and total loanable funds or borrowed funds (w 3
).This is consistent with the study on Portuguese banking (Boucinha et al., 2013).
The estimated cost elasticity with respect to total borrowed funds (w 3 ) is relatively higher than the other two input prices, labour price (w 1 ) and physical capital price (w 2 ).
3 As a consequence, banks are increasing their non-funded credit exposures, for example issuing guarantees, letter of credit and so on with a view to increasing non-interest income from non traditional banking activities (off-balance sheet items) and thus reduce costs. This trend has been observed in international banking as well. In Turkish banking, “off-balance sheet items
earned recently from off-balance sheet activities (Wang & Kumbhakar, 2009).
The estimated coefficient on the interaction term between two outputs (Lny 1 Lny 2 ) is negative indicating that there are scope economies in the joint production of loans and advances (y 1 ) and other earning assets (y 2 ). This is also consistent with the empirical study on
Portuguese banking (Boucinha et al., 2013).However, the estimate is not found statistically
significant. The interaction terms which do not have direct interpretations are not shown in the table1. They are shown in Appendix IV .
The negative magnitude of the estimated time trend (t) coefficient suggests for cost reducing technological progress. The banks in Bangladesh have been adopting cost-effective advanced technologies through offering new products and services, for example, on-line banking, mobile-banking, credit card, debit card, ATM services and so on. However, a very July 1-2, 2014 Cambridge, UK 23
2014 Cambridge Conference Business & Economics ISBN : 9780974211428 small magnitude (0.0011) and also lack of statistical significance of the estimated coefficient indicates that the rate of technological change may not directly affect banking costs. The variation in technical change may have come through diffusion of existing technology from the most efficient banks to other banks rather than reflecting new technological innovations or
change, as suggested by Humphrey (1993). The estimated positive coefficient of the square of
the time trend variable (
t
2 ) indicates an increasing trend in cost in adopting new technologies over time. However, the estimate is not statistically significant and also the magnitude is very small (0.0004). The estimated positive and statistically significant coefficient for equity capital (z 1 ) suggests that more capitalized bank is less cost efficient. Raising equity is associated with
higher costs (Lozano-Vivas & Pasiouras, 2010). However, the magnitude of the cost elasticity
in respect to equity capital at the sample mean is very small (0.0029). This is consistent with
several empirical studies (e.g.,Boucinha et al., 2013; Das & Ghosh, 2006). In compliance to
the Basel Accord II and III (Basel recommendations), banks are required to increase equity capital to increase capital adequacy (equity to total risk-weighted assets) resulting higher borrowings at the cost of interest expenses.
4 Furthermore, higher equity capital requirement is emanated from lower asset quality, non-performing loans (NPLs). However, Fries and Taci
(2005) find opposite result. A higher bank capitalization may lower the bank cost through
greater incentive for sound banking and efficiency. The proposed Basel III accord suggests for strengthening bank capital framework further. Therefore, banks will be required to hold more equity capital against their risk-weighted assets. July 1-2, 2014 Cambridge, UK 24
2014 Cambridge Conference Business & Economics ISBN : 9780974211428 The estimated negative and statistically significant coefficient for another control variable financial intermediation ratio (z 2 ) indicates that higher intermediation ratio lowers
bank costs. Fries and Taci (2005) find similar results in a cross-country study on cost
efficiency. The ratio of total loans to total deposits (intermediation ratio) is a regulatory provision for each bank in order to ensure efficient financial intermediation. Since banks collect deposits at the cost of interest expenses, the higher the capacity to convert the deposited money to loans and advances (interest earning output), the lower the cost of the bank. The estimated negative coefficient for the transition period dummy variable (DTr) indicates that bank cost reduces in the transition period (while implementing the financial sector reform programme) due to financial deregulation compared to the pre-reform period. For the post-reform period dummy variable (DPs), the estimated coefficient is negative and statistically significant suggesting reduced cost in the post-reform period compared to the pre reform period. This may occur due to the adoption of more cost reducing banking regulations and policy initiatives in the banking sector in Bangladesh during the post-reform period.
5.2 Parameter estimates: inefficiency function
The parameter estimates of the inefficiency function (Eq.3) estimated simultaneously with the cost function are reported in Table 1 as well.The estimated coefficient of the inefficiency function reflects the extent of cost efficiency differences across the sample banks over time. The estimated negative and statistically significant coefficient for the ownership dummy variable (OWN) indicates that state-owned banks may have more scope to control cost and thereby reduce cost inefficiency compared to its private sector counterparts. The reasons for private banks being less cost efficient are initial capital expenditure, borrowings at July 1-2, 2014 Cambridge, UK 25
2014 Cambridge Conference Business & Economics ISBN : 9780974211428 high interest rates and other costs such as high salaries paid to their employees and
decorations (Mahesh & Bhide, 2008).
The estimated negative coefficient for the dummy variable for independent director in the bank board (ID) suggests that banks with independent director in the board are more cost efficient than the banks with no independent voice in the board. Since independent directors have reputations and also have no other involvement in bank’s internal affairs, they can better monitor the banking activities and performance without any bias. The is consistent with other
empirical studies (e.g.,Isik & Hassan, 2002; Pathan et al., 2007). The regulatory provision of
independent director in bank board in Bangladesh was introduced in 2006. The main objective of this provision is to ensure good governance in the banking sector. However, only few banks have appointed independent directors in the bank board so far. Conversely, the estimated positive and statistically significant coefficient for the dummy variable for political director (PD) indicates that banks with more political directors in the board are less cost efficient than the banks with political connection. Due to political influence bank finances are directed to investment projects which are not viable, and thus, non-performing loan (NPL) is a common phenomenon in the banking sector in Bangladesh. A significant amount of cost is involved in improving the quality of loan portfolio with such NPLs, for example, default loan monitoring expenses, provisioning against loan losses and so on. Many other empirical studies find similar evidence that political connections or political
members in the bank board affect negatively in banking performance (Carretta et al., 2012;
Imai, 2009; Khwaja & Mian, 2005; Onder & Ozyildirim, 2011).
The estimated negative and statistically significant coefficient for bank size (SIZE) indicates that large size banks can control cost more efficiently. This is consistent with Berger
and Humphrey (1992).The authors suggest that there are cost advantages in expanding the
July 1-2, 2014 Cambridge, UK 26
2014 Cambridge Conference Business & Economics ISBN : 9780974211428 bank size in terms of asset portfolio in the presence of economies of scale. Moreover, large size banks have credibility and thus attract more competent managers who are efficient in delivering cost-effective banking services. Large banks also have the capacity to invest in adopting new technological innovations and thus provide technology driven cost-effective
and Kumbhakar (2007) find contrary results arguing that bureaucratic problem associated
with large size bank does harm to the performance and thus it may be detrimental to the overall technical efficiency. In fact, the literature demonstrates no agreement on the
relationship between size and efficiency (Chen et al., 2005).
The estimated negative and statistically significant coefficients for the transition period dummy (DTr) indicates that banks in Bangladesh have become more cost efficient while adopting reform policies and regulations compared to pre-reform period. This is consistent with other empirical studies examining the impact of financial reform programme
on bank efficiency (Das & Kumbhakar, 2012; Hao, Hunter, & Yang, 2001).However, the
estimated positive and significant coefficient for post-reform period dummy variable (DPs) suggests that cost efficiency may decrease at a later stage of deregulation process. The environmental condition prior to deregulation and other incentives may be liable for such
efficiency differences (Berger & Humphrey, 1997). Perhaps early stages of banking reform
are associated with greater cost efficiency while the level of efficiency deteriorates in
advanced stages. Similar result has been observed by Fries and Taci (2005) while examining
the impact of banking reform on cost efficiency of 289 banks in 15 post-communist countries. July 1-2, 2014 Cambridge, UK 27
2014 Cambridge Conference Business & Economics ISBN : 9780974211428 The estimated negative and statistically significant coefficient of time trend (t) indicates that sample banks reduce cost and thus become more cost efficient over time. Das
and Kumbhakar (2012) find similar declining inefficiency over time.
The estimated negative and statistically significant coefficient of 3-bank deposit concentration ratio (CR3) suggests that banks with larger share of the deposit market are more cost efficient than other banks.The significant positive association between concentrated banking markets and greater cost efficiency validates the fact that concentration is the result
of either superior management or greater efficiency of the production process (Demsetz,
1973; Dietsch & Lozano-Vivas, 2000; Grigorian & Manole, 2002).
The estimated parameter (gamma) corresponds to the proportion of bank inefficiency in the composite error term
it
(
v it
u it
) . The estimated coefficient for gamma suggests that approximately 55 percent of the composite error term is attributed to inefficiency. The estimated coefficient of is statistically significant at 1% level of significance. 5.3 Cost efficiency estimates Table 2 presents the summary statistics of the estimated cost efficiency scores obtained from estimating the cost frontier model (Eq.2). The detail bank-wise cost efficiency scores are presented in Appendix V. Individual bank efficiency indices are measured by computing the deviations of costs from the cost frontier (constructed based on ‘best practice’ banks). The cost efficiency score shows the rate of saving cost without reducing outputs, measured against the best practice frontier. July 1-2, 2014 Cambridge, UK 28
2014 Cambridge Conference Business & Economics ISBN : 9780974211428 Table 2: Summary statistics of the cost efficiency scores estimated, 1983-2012
Pre-reform
(1983-1990) Public Private Mean 0.9410 0.9378 All Public
Transition
(1991-1995) Private All Public
Post-reform
(1996-2012) Private All 0.9389 0.9689 0.9739 0.9722 0.9685 0.9663 0.9671 S.E Max 0.0304 0.0373 0.9874 0.9858 0.0350 0.0202 0.0225 0.0217 0.0322 0.0359 0.0347 0.9874 0.9915 0.9959 0.9959 0.9954 0.9967 0.9967 Min 0.8733 0.8311 0.8311 0.9222 0.8976 0.8976 0.8746 0.8402 0.8402 Note: Efficiency scores estimated by using a panel data set of major commercial banks in Bangladesh over the
period, 1983-2012 (360 observations) using computer programme FRONTIER 4.1(Coelli, 1996)
Since the main objective of this study is to investigate the consequences of the financial reform programme, we analyse the estimated average cost efficiency scores in July 1-2, 2014 Cambridge, UK 29
2014 Cambridge Conference Business & Economics ISBN : 9780974211428 respect of three periods, the pre-reform period 1983-1990, the transition period 1991-1995 and the post-reform period 1996-2012 as mentioned in the empirical design (Section 4). 5.3.1 Pre-reform and post-reform period Table 2 shows that the average cost efficiency for the sample banks increases about 3%, from 93% in the pre-reform period to 96% in the post-reform period. This suggests that although efficiency improves, still banks can reduce costs by 4% in the post-reform period in order to become fully efficient. The efficiency gain of private banks is higher than public banks across the periods. The increase in average cost efficiency of private banks is about 3% while the rate of increase is about 2% for public banks. Although the average cost efficiency of public banks was higher during the pre-reform period, the efficiency difference converges in the post-reform period. Perhaps private sector banks adopted reform measures rapidly and offered technology driven cost-effective banking products and services to compete with other competitors in the industry. There may be several reasons for the slower efficiency gains of public banks compared to private banks. First, public banks fall behind in offering technology-driven and cost-effective products and services, for example, debit card, credit card, on-line banking, mobile-banking and so on. Second, differences in behaviour and objectives which can be explained by related theories such as property rights theory, agency cost theory and
transaction costs theory (Wei, Varela, & Hassan, 2002). These theories suggest that publicly
owned firms should perform less efficiently and less profitably than private firms. The objective of private banks is profit maximization while public banks do not necessarily pursue profit rather they pursue whatever the government demands. The government uses the public banks to support its political objectives to build their own fortunes at the expense of funds that banks owe to depositors. To control the banks, government retains a large number of July 1-2, 2014 Cambridge, UK 30
2014 Cambridge Conference Business & Economics ISBN : 9780974211428 unskilled or redundant employees. In fact, labour intensive policies in public banks are cost inefficient. The government even appoints the chief executive officer (CEO) according to
their choice (not competitive basis) in order to promote politicians’ welfare (Boycko, Shleifer,
& Vishny, 1996; Reaz & Arun, 2006).
The third reason is the difference in budget constraints. The private banks are subject to a relatively ‘hard’ budget constraint and under the monitoring of private owners and board
they are virtually not accountable to anyone. Therefore, public sector bank would direct investment funds to the politically desirable projects. In fact, most of the lending in public banks is directed by the politicians who are directly involved or have link with ruling political
The efficiency estimates also show that private banks were less efficient during the pre-reform period. This may happen since most of the capital expenditures were incurred
rate at the initial stage which incurs significant costs. 5.3.2 Pre-reform- transition period and transition- post-reform period The average cost efficiency of the sample banks during the transition period increases about 3.33% compared to the pre-reform period. During this period, private banks’ efficiency increases about 3.61%, while public banks’ efficiency increases about 2.79%. However, the level of average cost efficiency remains more or less same for both private and public banks between transition period and post-reform period. A marginal decline in average cost July 1-2, 2014 Cambridge, UK 31
2014 Cambridge Conference Business & Economics ISBN : 9780974211428 efficiency for both private banks (about 0.7%) and public banks (0.04%) has been observed during the post-reform period compared to the transition period. Perhaps the early stages of reforms are associated with greater cost efficiency while costs tend to increase at the later stages. This may happen due to the compliance of regulatory requirements (e.g., Basel recommendations). Figure 1 depicts the trend in average cost efficiency of the sample banks over the periods, the pre-reform the transition, and the post-reform. The average cost efficiency of public banks was higher than its private sector counterpart during the pre-form period (1983 1990).After the implementation of the financial reform programme in 1991, both private and public sector banks gained efficiency and eventually the average cost efficiency of the sample banks increased by 3.33% during the transition period (1991-1996) compared to the pre reform period. However, the average cost efficiency level marginally declines (0.5%) during the post-reform period compared to the transition period. July 1-2, 2014 Cambridge, UK 32
2014 Cambridge Conference Business & Economics ISBN : 9780974211428
Figure 1: Average cost efficiency estimates over time
It is evident from the figure 1 that the cost efficiency of the sample banks increases significantly while the financial reform programme implemented (transition period). However, the average cost efficiency tends to decline slightly (0.5%) in the post-reform period, as already discussed, perhaps to maintain the asset quality and the required capital July 1-2, 2014 Cambridge, UK 33
2014 Cambridge Conference Business & Economics ISBN : 9780974211428 against the risk-weighted assets. This is really a serious concern for the banking sector in Bangladesh where a large volume of the risk-weighted assets is non-performing loans (NPL). Due to political influence in the bank governance both in public and private sector banks in Bangladesh (e.g., political directors in the bank board) the volume of NPLs increases over time. As a regulatory requirement (Basel recommendation), to ensure asset quality, banks are required to monitor the NPL and maintain the required provisioning against the potential loan losses which involves substantial costs. Against this backdrop, it would be a challenge for the banks in Bangladesh to be cost efficient, particularly after the full implementation of the new Basel III regulations. The implementation of this accord has already been started from
expensive to enforce. The most obvious costs of Basel III is compliance costs (Kupiec, 2013).
Before the recent global financial crisis, securitization and off-balance sheet exposures were the favoured way to reduce bank capital requirements. However, currently many large banks have reacted to tighter capital regulation by “optimizing” their internal models to produce lower risk weights and lower minimum capital requirements.
6. CONCLUSION
This paper analyses the cost efficiency estimates of 12 major commercial banks in Bangladesh evaluating the impact of financial reform programme. Following the single-stage
SFA model of Battese and Coelli (1995), we estimate several translog cost frontier models
using a unique balanced panel dataset for the period 1983-2012.We find that the sample heterogeneity significantly influences the SFA estimates. After conducting the required specification tests, we obtain the appropriate cost frontier model which describes the sample data including outputs, input prices and explanatory variables in the function. July 1-2, 2014 Cambridge, UK 34
2014 Cambridge Conference Business & Economics ISBN : 9780974211428 The statistically significant parameter estimates of the preferred SFA model provides us the evidence that financial deregulation contributes in reducing bank cost although a slightly increasing trend in cost inefficiency has been observed in the post-reform era. Besides, new innovations in banking products and services applying advanced technology contribute in reducing bank cost. However, political influence on both public and private banks is prevalent. A political director in the bank board is inconsistent with good corporate governance since the estimated coefficient for political director dummy variable reflects increasing cost inefficiency across the sample banks. Alternatively, we find that independent directors in the bank board may be conducive to good corporate governance. The estimated coefficient for independent director dummy variable reflects reducing cost inefficiency, perhaps through monitoring banking activities independently without involving with the internal banking affairs. It is evident from the analysis of the estimated average cost efficiency scores that both public and private sector banks have gained efficiency as a consequence of financial deregulation. The implementation of reform policies such as introduction of market determined interest rates, freedom in investment decisions instead of directed credit, adoption of regulatory changes in compliance with the international banking regulations, e.g., Basel recommendations (Basel I-III) for strengthening capital base of the banks, risk management etc., privatization and new entry of private banks - all these initiatives have ensured a competitive banking environment in Bangladesh. Although the cost efficiency gap between public and private banks converges in the post-reform period, the gap was wider in the pre-reform period. There are several reasons for such wider gap, as mentioned in Section 5.3.1. The sample private banks commenced their operation in the pre-reform period. Since they incurred most of the capital expenditures July 1-2, 2014 Cambridge, UK 35
2014 Cambridge Conference Business & Economics ISBN : 9780974211428 during the initial years of their operation, the fixed cost of the banks increased substantially
(Isik & Hassan, 2003). Apart from this, new private banks borrow most of their investment
funds at the initial stage of their operation at a higher interest rate which incurs a significant cost. However, the gap reduces during the transition period while the reform policy was implemented because private banks adopt reform initiatives relatively quicker and also introduce technology driven cost-effective banking products and services to be competitive compared to the public sector banks, and eventually efficiency differences converge. The cost efficiency scores show a slightly declining trend in the post-reform period for both groups of banks perhaps due to the cost involved for complying the international banking regulations (e.g., maintaining asset quality, capital adequacy and required provisions against loan losses) as per the Basel recommendations. Although the banking in Bangladesh has become relatively cost efficient after the implementation of the financial reform programme, however, there are still room for improvement. In the backdrop of recent global financial crisis, the main challenge of the banks in Bangladesh is the compliance of the new Basel III accord, the recommendations of Basel committee on banking supervision on fundamental banking issues: liquidity coverage, capital adequacy, corporate governance etc. to ensure an efficient and stable financial sector.
References
Aigner, D., Lovell, C. A. K., & Schmidt, P. (1977). Formulation and estimation of stochastic frontier production function models. Journal of Econometrics, 6(1), 21-37. doi: 10.1016/0304-4076(77)90052-5 Akther, S., Fukuyama, H., & Weber, W. L. (2012). Estimating two-stage network slacks based inefficiency: an application to Bangladesh banking. Omega, 41(1), 88-96. doi: 10.1016/j.omega.2011.02.009 Al-Muharrami, S., Matthews, K., & Khabari, Y. (2006). Market structure and competitive conditions in the Arab GCC banking system. Journal of Banking & Finance, 30, 3487-3501. Andres, P. d., & Vallelado, E. (2008). Corporate governance in banking: the role of the board of directors. Journal of Banking and Finance, 32, 2570-2580. July 1-2, 2014 Cambridge, UK 36
2014 Cambridge Conference Business & Economics ISBN : 9780974211428 Ariff, M., & Can, L. (2008). Cost and profit efficiency of Chinese banks: a non-parametric analysis. China Economic Review, 19, 260-273. Avkiran, N. (2000). Rising productivity of Australian trading banks under deregulation 1986 1995. Journal of Economics and Finance, 24, 122-140. Avkiran, N. K. (1999). The evidence on efficiency gains: the role of mergers and the benefits to the public. Journal of Banking & Finance, 23(7), 991-1013. Banker, R. D., Chang, H., & Lee, S.-Y. (2010). Differential impact of Korean banking system reforms on bank productivity. Journal of Banking & Finance, 34(7), 1450-1460. doi: 10.1016/j.jbankfin.2010.02.023 Basel committee on banking supervision. (2011). Basel III : a global regulatory framework for more resilient banks and banking systems. Basel: Bank for International Settlements. Bathala, C. T., & Rao, R. P. (1995). The determinants of board composition:an agnency theory perspective. Managerial and decision economics, 16, 59-69. Battese, G. E., & Coelli, T. J. (1992). Frontier production functions, technical efficiency and panel data: with application to paddy farmers in India. The Journal of Productivity Analysis, 3, 153-169. Battese, G. E., & Coelli, T. J. (1995). A model for technical inefficiency effects in a stochastic frontier production function for panel data. Empirical Economics, 20(2), 325-332. Bauer, P. W. (1990). Recent developments in the econometric estimation of frontiers. Journal of Econometrics, 46, 39-56. Bauer, P. W., Berger, A. N., & Humphrey, D. B. (1993). Efficiency and productivity growth in U.S. banking. In H. O. Fried, C. A. K. Lovell & S. S. Schmidt (Eds.), The Measurement of Productive Efficiency: Techniques and Applications (pp. 386-413). New York: Oxford University Press. Benston, G. J. (1965). Branch banking and economies of scale. Journal of Finance, 20, 312 331. Berg, S. A., Forsund, F. R., & Jansen, E. S. (1992). Malmquist indices of productivity growth during the deregulation of Norwegian banking. Scandinavian Journal of Economics, 94, 211-228. Berger, A. N., DeYoung, R., Genay, H., & Udell, G. F. (2000). Globalization of financial institutions: evidence from cross-border banking performance. Brookings-Wharton Papers on Financial Service, 3, 23-120. Berger, A. N., Hasan, I., & Zhou, M. (2009). Bank ownership and efficiency in China: What will happen in the world’s largest nation? Journal of Banking & Finance, 33(1), 113 130. doi: 10.1016/j.jbankfin.2007.05.016 Berger, A. N., & Humphrey, D. B. (1991). The dominance of inefficiencies over scale, product mix economies in banking. Journal of Monetary Economics, 28, 117-148. Berger, A. N., & Humphrey, D. B. (1992). Measurement and efficiency issues in commercial banking. In Z. Grilliches (Ed.), Output measurement in the service sector (Vol. 56, pp. 256-279). Chicago: University of Chicago. 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. doi: 10.1016/s0377-2217(96)00342-6 Berger, A. N., Kick, T., & Schaeck, K. (2012). Executive board composition and bank risk taking Working Paper (Vol. 12). Bangor: Bangor Business School. July 1-2, 2014 Cambridge, UK 37
2014 Cambridge Conference Business & Economics ISBN : 9780974211428 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. doi: 10.1016/s0378-4266(97)00010-1 Bhattacharyya, A., Lovell, C. A. K., & Sahay, P. (1997). The impact of liberalization on the productive efficiency of Indian commercial banks. European Journal of Operational Research, 98, 332-345. Boucinha, M., Ribeiro, N., & Weyman-Jones, T. (2013). An assessment of Portuguese banks' efficiency and productivity towards euro area participation. Journal of Productivity Analysis, 39(2), 177-190. Boycko, M., Shleifer, A., & Vishny, R. W. (1996). A theory of privatization. The Economic Journal, 106, 309-319. Burki, A. A., & Niazi, G. S. K. (2010). Impact of financial reforms on efficiency of state owned, private and foregin banks in Pakistan. Applied Economics, 42, 3147-3160. Carretta, A., Farina, V., Gon, A., & Parisi, A. (2012). Politicians 'on board': do political connections affect banking activities in Italy? European Management Review, 9, 75 83. 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, 2521-2540. Chen, X., Skully, M., & Brown, K. (2005). Banking efficiency in China: application of DEA to pre- and post-deregulation eras: 1993-2000. China Economic Review, 16, 229-245. Choi, S., & Hasan, I. (2005). Ownership, Governance, and Bank Performance: Korean Experience. Financial Markets, Institutions & Instruments, 14(4), 215-242. doi: 10.1111/j.0963-8008.2005.00104.x Coelli, T. J. (1996). A guide to FRONTIER version 4.1: a computer program for stochastic frontier production and cost function estimation. CEPA Working Paper University of New England. Armidale. Coelli, T. J., Rao, D. S. P., O' Donnell, C. J., & Battese, G. E. (2005). An Introduction to Efficiency and Productivity Analysis (second ed.). New York: Springer. Cornwell, C., Schmidt, P., & Sickles, R. C. (1990). Production frontiers with cross-sectional and time-series variation in efficiency levels. Journal of Econometrics, 46(1–2), 185 200. doi: 10.1016/0304-4076(90)90054-w Cowing, T. G., & Stevenson, R. E. (1981). Introduction: productivity measurement and regulated industries. In T. G. Cowing & R. E. Stevenson (Eds.), Productivity Measurement in Regulated Industries (pp. 3-14). New York: Academic Press. Das, A., & Ghosh, S. (2006). Financial deregulation and efficiency: an empirical analayis of Indian banks during the post reform period. Review of Financial Economics, 15, 193 221. Das, A., & Kumbhakar, S. C. (2012). Productivity and efficiency dynamics in Indian banking: an input distance function approach incorporating quality of inputs and outputs. Journal of Applied Econometrics, 27, 205-234. Debreu, G. (1951). The coefficient of resource utilization. Econometrica, 19(3), 273-292. Demsetz, H. (1973). Industry structure, market rivalry, and public policy. Journal of Law and Economics, 16, 1-9. DeYoung, R., & Nolle, D. E. (1996). Foreign-owned banks in the United States: earning market share or buying it? Journal of Money, Credit and Banking, 28(4), 622-636. July 1-2, 2014 Cambridge, UK 38
2014 Cambridge Conference Business & Economics ISBN : 9780974211428 Dietsch, M., & Lozano-Vivas, A. (2000). How the environment determines banking efficiency: a comparison between French and Spanish industries. Journal of Banking &Finance, 24, 985-1004. Dong, Y. (2009). Cost efficiency in the Chinese banking sector:a comparison of parametric and non-parametric methodologies. (Ph.D thesis), Loughborough University. Drake, L., & Hall, M. J. B. (2003). Efficiency in Japanese banking: an empirical analysis. Journal of Banking & Finance, 27(5), 891-917. doi: 10.1016/s0378-4266(02)00240-6 Du, J., & Girma, S. (2011). Cost economies, efficiency and productivity growth in the Chinese banking industry: evidence from a quarterly panel dataset. Empirical Economics, 41, 199-226. Eisenbeis, R. A., Ferrier, G. D., & H, K. S. (1996). An empirical analysis of the
informativeness of programming and SFA efficiency scores: efficiency and bank
performance. Working paper. University of North Carolina. Chapel Hill NC. Elyasiani, E., & Mehdian, S. (1995). The comparative efficiency performance of small and large U.S. commercial banks in the pre and post deregulation eras. Applied Economics, 27, 1069-1079. Färe, R., & Lovell, C. A. K. (1978). Measuring the technical efficiency of production. Journal of Economic Theory, 19(1), 150-162. doi: 10.1016/0022-0531(78)90060-1 Farrell, M. (1957). The measurement of productivity efficiency. Journal of the Royal Statistical Society 120, 253-281. Favero, C. A., & Papi, L. (1995). Technical efficiency and scale efficiency in the Italian banking sector: a non-parametric approach. Applied Economics, 27(4), 385-396. Ferrier, G. D., & Lovell, C. A. K. (1990). Measuring cost efficiency in banking: econometric and linear prgramming evidence. Journal of Econometrics, 46, 229-245. Fried, H. O., Lovell, C. A. K., & Schmidt, S. S. (2008). Efficiency and productivity. In H. O.Fried, C. A. K. Lovell & S. S.Schmidt (Eds.), The measurement of productive efficiency and productivity growth (pp. 3-91). Oxford: Oxford University Press. Fries, S., & Taci, A. (2005). Cost efficiency of banks in transition: evidence from 289 banks in 15 post-communist countries. Journal of Banking & Finance, 29, 55-81. Greene, W. H. (1980). On the estimation of a flexible frontier production model. Journal of Econometrics, 13, 101-115. Grifell-Tatj`e, E., & Lovell, C. A. K. (1999). Profit and productivity. Management Science, 45(9), 1177-1193. Grigorian, D., & Manole, V. (2002) Determinants of commerical bank performance in transition : an application of data envelopment analysis. Vol. 2850. World Bank Policy Research Working Paper. Washington: World Bank. Gujarati, D. N. (2003). Basic Econometrics. New York: McGraw-Hill. Hao, J., Hunter, W. C., & Yang, W. K. (2001). Deregulation and efficiency: the case of private Korean banks. Journal of Economics and Business, 53, 237-254. Hasan, I., & Marton, K. (2003). Development and efficiency of the banking sector in a transitional economy: Hungarian experience. Journal of Banking & Finance, 27, 2249-2271. Hassan, M. K. (1999). Islamic banking in theory and practice: the experience of Bangladesh. Managerial Finance, 25(5), 60-113. Hermalin, B. E., & Weisbach, M. S. (1991). The effects of board composition and direct incentives on firm performance. Financial Management, 20(4), 101-112. Heshmati, A. (2003). Productivity growth, efficiency and outsourcing in manufacturing and service industries. Journal of Economic Surveys, 17(1), 79-112. July 1-2, 2014 Cambridge, UK 39
2014 Cambridge Conference Business & Economics ISBN : 9780974211428 Humphrey, D. B. (1993). Cost and technical change: effects from bank deregulation. The Journal of Productivity Analysis, 4, 9-34. Hunter, W. C., Timme, S. G., & Yang, W. K. (1990). An examination of cost subadditivity and multiproduct production in large U.S. banks. Journal of Money, Credit and Banking, 22(4), 504-525. Iannnotta, G., Nocera, G., & Sironi, A. (2007). Ownership structure, risk and performance in the European banking industry. Journal of Banking & Finance, 31, 2127-2149. Imai, M. (2009). Political determinants of government loans in Japan. Journal of Law and Economics, 52, 41-70. Isik, I., & Hassan, M. K. (2002). Technical, scale and allocative efficiencies of Turkish banking industry. Journal of Banking & Finance, 26, 719-766. Isik, I., & Hassan, M. K. (2003). Financial deregulation and total factor productivity change: an empirical study of Turkish commercial banks. Journal of Banking & Finance, 27, 1455-1485. Jha, R., Murty, M. N., & Paul, S. (1991). Technological change, factor substitution and economies of scale in selected manufacturing industries in India Journal of Quantitative Economics, 7(1), 165-178. Khanam, M. D., & Khandoker, M. S. H. (2005). X-efficiency analysis of problem banks of Bangladesh. Pakistan Journal of Social Sciences, 3(3), 521-525. Khwaja, A. I., & Mian, A. (2005). Do lenders favor politically connected firms? rent provision in an emerging financial market. The Quarterly Journal of Economics, 120(4), 1371-1411. Kodde, D. A., & Palm, F. C. (1986). Wald criteria for jointly testing equality and inequality restrictions. Econometrica, 54(5), 1243-1248. Koopmans, T. C. (1951). An analysis of production as an efficient combination of activities. In T.C.Koopmans (Ed.), Activity Analysis of Production and Allocation (Vol. 13). New York: Wiley. Kumbhakar, S. C. (1997). Modeling allocative inefficiency in a translog cost function and cost share equations: an exact relationship. Journal of Econometrics, 76, 351-356. Kumbhakar, S. C., & Lovell, C. A. K. (2000). Stochastic Frontier Analysis. Cambridge: Cambridge University Press. Kumbhakar, S. C., & Lovell, C. A. K. (2003). Stochastic Frontier Analysis. Cambridge: Cambridge University Press. Kumbhakar, S. C., & Lozano-Vivas, A. (2005). Deregulation and productivity: the case of Spanish banks. Journal of Regulatory Economics, 27(3), 331-351. Kumbhakar, S. C., Lozano-Vivas, A., C.A.Knox Lovell, C. A. K., & Hasan, I. (2001). The effects of deregulation on the performance of financial institutions: the case of Spanish savings banks. Journal of Money, Credit, and Banking, 33(1), 101-120. Kumbhakar, S. C., & Wang, D. (2007). Economic reforms, efficiency and productivity in Chinese banking. Journal of Regulatory Economics, 32, 105-129. Kupiec, P. H. (2013). Basel III : some costs will outweigh the benefits: American Enterprise Institute. Lefort, F., & Urzua. (2008). Board independence, firm performance and ownership concentration:evidence from Chile. Journal of Business Research, 61, 615-622. Leightner, J. E., & Lovell, C. A. K. (1998). The impact of financial liberalization on the performance of Thai banks. Journal of Economics and Business, 50, 115-131. Lin, X., & Zhang, Y. (2009). Bank ownership reform and bank performance in China. Journal of Banking & Finance, 33(1), 20-29. doi: 10.1016/j.jbankfin.2006.11.022 July 1-2, 2014 Cambridge, UK 40
2014 Cambridge Conference Business & Economics ISBN : 9780974211428 Lozano-Vivas, A., & Pasiouras, F. (2010). The impact of non-traditional activities on the estimation of bank efficiency: international evidence. Journal of Banking & Finance, 34, 1436-1449. Mahesh, H. P., & Bhide, S. (2008). Do financial sector reforms make commercial banks more efficient? a parametric exploration of the Indian case. The Journal of Applied Economic Research, 2(4), 415-441. Meeusen, W., & Broeck, v. d. (1977). Efficiency estimation from Cobb-Douglas production function with composed error. International Economic Review, 18, 435-444. MRA. (2010). Annual Report. Dhaka: Government of Bangladesh. Onder, Z., & Ozyildirim, S. (2011). Political connection, bank credits and growth: evidence from Turkey. The World Economy, 34(6), 1042-1065. Orea, L., & Kumbhakar, S. C. (2004). Efficiency measurement using a latent class stochastic frontier model. Empirical Economics, 29, 169-183. Pathan, S., Skully, M., & Wickramanayake, J. (2007). Board size, independence and performance: an analysis of Thai banks. Asia-Pacific Financial Markets 14, 211-227. Perera, S., Skully, M., & Wickramanayake, J. (2007). Cost efficiency in South Asian Banking: The impact of bank size, state ownership and stock exchange Listings. International Review of Finance, 7(1-2), 35-60. Pi, L., & Timme, S. G. (1993). Corporate control and bank efficiency. Journal of Banking & Finance, 17, 515-530. Reaz, M., & Arun, T. (2006). Corporate governance in developing economies: perspective from the banking sector in Bangladesh. Journal of Banking Regulation, 7(1&2), 94 105. Resti, A. (1997). Evaluating the cost-efficiency of the Italian banking system: what can be learned from the joint application of parametric and non-parametric techniques. Journal of Banking & Finance, 21, 221-250. Rezvanian, R., Ariss, R. T., & Mehdian, S. M. (2011). Cost efficiency, technological progress and productivity growth of Chinese banking pre- and post- WTO accession. Applied Financial Economics, 21(7), 437-454. Rezvanian, R., & Mehdian, S. (2002). An examination of cost structure and production performance of commerical banks in Singapore. Journal of Banking &Finance, 26, 79-98. Salim, R. A., Hoque, M. Z., & Suyanto. (2010). The role of governance, ICT and badloans in Australian bank efficiency: an empirical study. The Asia Pacific Journal of Economics and Business, 14(1), 18-36. Sathye, M. (2001). X-efficiency in Australian banking: an empirical investigation. Journal of Banking & Finance, 25, 613-630. Sealey, C., & Lindley, J. T. (1977). Inputs, outputs and a theory of production and cost at depository financial institution. Journal of Finance, 32, 1251-1266. Sensarma, R. (2005). Cost and profit efficiency of Indian banks during 1986-2003: a stochastic frontier analysis. Economic and Political Weekly, 40, 1198-1209. Shen, C.-H., & Lin, C.-Y. (2012). Why government banks underperform: A political interference view. Journal of Financial Intermediation, 21, 181-202. Shephard, R. W. (1970). Theory of Cost and Production functions. Princeton: Princeton University Press. Stevenson, R. E. (1980). Likelihood functions for generalized stochastic frontier estimation. Journal of Econometrics, 13, 57-66. July 1-2, 2014 Cambridge, UK 41
2014 Cambridge Conference Business & Economics ISBN : 9780974211428 Sturm, J., & Williams, B. (2004). Foreign bank entry, deregulation and bank efficiency: lessons from the Australian experience. Journal of Banking & Finance, 28, 1775 1799. Sun, L., & Chang, T.-P. (2011). A comprehensive analysis of the effects of risk measures on bank efficiency: evidence from emerging Asian countries. Journal of Banking & Finance, 35, 1727-1735. Suyanto, Salim, R. A., & Bloch, H. (2009). Does foreign direct investment lead to productivity spillovers? firm level evidence from Indoneshia. World Development, 37(12), 1861-1876. Wang, D., & Kumbhakar, S. C. (2007). Economic reforms, efficiency and productivity in Chinese banking. Journal of Regulatory Economics, 32, 105-129. Wang, D., & Kumbhakar, S. C. (2009). Strategic groups and heterogeneous technologies: an application to the US banking industry. Macroeconomics and Finance in Emerging Market Economics, 2(1), 31-57. Wei, Z., Varela, O., & Hassan, M. K. (2002). Ownership and performance in Chinese manufacturing industry. Journal of Multinational Financial Management, 12, 61-78. Weisbach, M. S. (1988). Outside directors and CEO turnover. Journal of Financial Economics, 20, 431-460. Wheelock, D. C., & Wilson, P. W. (1999). Technical progress, inefficiency, and productivity change in U.S. banking, 1984-1993. Journal of Money, Credit and Banking, 31(2), 212-234.
Appendix I: Definition of the variables Variables Dependent variable
TC
Definition
Total cost: the sum of interest expenses and operating expenses. The variable is measured in million Taka (Bangladesh currency), deflated using GDP deflator, base: 1996=100 (WDI, 2013). Independent variables Outputs y 1 y 2 Loans and advances: the sum of total loans and bills discounted. The variable is measured in million Taka (Bangladesh currency), deflated using GDP deflator, base: 1996=100 (WDI, 2013). Other earning assets: total assets less total loans and advances and fixed assets. The variable is measured in million Taka (Bangladesh July 1-2, 2014 Cambridge, UK 42
2014 Cambridge Conference Business & Economics ISBN : 9780974211428 Input prices w w w 1 2 3 Price of labour : total expenditure on employees (salaries and allowances) divided by the total number of employees Price of physical capital: total expenditure on premises and fixed assets, i.e, other operating expenses (except salary and allowances and charges on loans/investment losses) divided by the book value of physical capital and other fixed assets Price of loanable funds: total interest expenses divided by total loanable funds (total deposits plus borrowed funds) Control variables and correlates of inefficiencies EQ FI Equity: the sum of core capital and supplementary capital: the sum of paid up capital, statutory reserve, general reserves, other reserves and general provisions. The value of the variable is measured in million Taka (Bangladesh currency), deflated using GDP deflator, base: 1996=100 (WDI, 2013). Financial intermediation: the ratio of total loans to total deposit t DPr DTr DPs OWN SIZE CR3 currency), deflated using GDP deflator, base: 1996=100 (WDI, 2013). Time trend: t=1 for 1983, t=2 for 1984………T=30 for 2012 Pre-reform dummy variable for the period, 1983-1990. However, pre-reform period is considered as base period Transition dummy variable for the period, 1991-1995. DTr=1 if transition period and zero, if otherwise Post-reform dummy variable for the period, 1996-2012. DPs=1 if post-reform period and zero, if otherwise Bank ownership: dummy variable for bank ownership. OWN=1 if public bank and zero, if otherwise Bank size: natural logarithm of the total assets, as deflated using GDP deflator, base: 1996=100 (WDI, 2013). 3-bank concentration ratio: an annual index measures the deposit share of three major state-owned banks (Sonali, Janata and Agrani) reflecting the market power or competition in the banking July 1-2, 2014 Cambridge, UK 43
2014 Cambridge Conference Business & Economics ISBN : 9780974211428 ID PD industry. Independent director: dummy variable; ID=1 if independent directors are in the bank board and zero, if otherwise Political director: dummy variable; PD=1 if political directors are in the bank board and zero, if otherwise
Appendix II: Model specification test
Given the general translog cost frontier as specified in Equation (2), we test a number of null hypotheses for finding the appropriate model for the sample dataset. The likelihood ratio is defined as L /
R L U
, where
L R
is the maximum of the likelihood function when the restrictions are imposed and restrictions are not imposed. This ratio statistic can be expressed as follows: LR= - 2ln = - 2(ln
L R
ln
L U
) (Coelli et al., 2005). The LR ratio can also be expressed as LR= -2[l(
H O
) - l(
H
1 )], where l(
H O
) denotes the value of likelihood function based on the null hypothesis or the restricted frontier model and l(
H
1 ) is the value of likelihood function in the alternative
hypothesis (Suyanto, Salim, & Bloch, 2009). If the sample size is sufficiently large, the log-
likelihood test statistic is asymptotically distributed as chi-square ( 2 ) distribution with the degrees of freedom equal to the number of restrictions imposed by the null hypothesis. July 1-2, 2014 Cambridge, UK 44
2014 Cambridge Conference Business & Economics ISBN : 9780974211428 Table I presents the relevant null hypothesis tests. The first null hypothesis is to confirm whether the Cobb-Douglas cost frontier is an appropriate specification for the dataset by imposing the following restrictions, the second order parameters each equal zero, i.e.,
mk
=
nj
=
mn
=0 in the Equation (2). The Cobb-Douglas cost frontier model is not an appropriate specification given the translog cost frontier model, as the log-likelihood ratio test indicates a strong rejection of the null hypothesis at 1% level of significance (the test statistics exceeds the critical value as shown in the Table I. Therefore, the translog functional form of the cost frontier is the appropriate model that describes the sample data. The second null hypothesis: the half-normal Model (M1) is an appropriate model compared to the truncated normal Model (M2) imposing the restriction, 0 . The log likelihood ratio test indicates that models are similar, as the test statistics do not exceed the critical values as shown in the Table I. However, we use Model 2 (M2) for conducting further tests with other model specifications in deciding the final model to estimate. This is because Model 2 assumes the inefficiency term distributed truncated normal. Therefore, the third null hypothesis is whether Model 2 (without control variables) is an appropriate model compared to the Model 3 (with control variables) imposing restrictions, 1 2 1
t
2
t
1
t
2
t
1 2 3 4 0 . The log likelihood ratio test indicates that the model without control variables (M2) is not an appropriate specification given the translog cost frontier model including control variables in the cost function (M3). The test statistic exceeds the chi-square critical values at 1% level of significance, and thus rejects the null hypothesis (Table I). The forth null hypothesis is whether the model not including environmental variables in the inefficiency function (M3) is an appropriate model compared to the model including July 1-2, 2014 Cambridge, UK 45
2014 Cambridge Conference Business & Economics ISBN : 9780974211428 environmental variables in the inefficiency function (M4) imposing the restrictions, 1 2 3 4 5 6 7 8 0 . The log-likelihood ratio test indicates a strong rejection of the null hypothesis at 1% level of significance since the test statistics exceeds the critical values as shown in the Table I. Therefore, we find that bank characteristics and time dependent variables significantly influence the cost frontier estimation. Thus, an appropriate efficiency model for the sample banks in Bangladesh should include bank-specific and time-specific control variables in the cost function and also environmental variables in the inefficiency function. July 1-2, 2014 Cambridge, UK 46
2014 Cambridge Conference Business & Economics
Model Description
ISBN : 9780974211428 Table I : Log-likelihood tests for model specification
Restrictions (degrees of freedom) Log-likelihood test statistic(
) Chi-square (
of significance
2
) ;level Decision
Cobb-Douglas vs Translog Translog :Half-normal (M1) vs Truncated(M2)
mk
=
nj
=
mn
=0 (10) 0 (1) 489.8952 0.0546
1% 5% 10%
22.53 17.67 15.38 Cobb-Douglas model rejected 5.412 2.706 1.642 Half-normal model not rejected Translog: Truncated model without control variables (M2) vs with control variable(M3) 1 1 2 2 1
t
3 2
t
4 1
t
0 2
t
(10) 215.2318 22.53 17.67 15.38 Truncated model without control variables (M2) rejected Translog: Model (controlled) not including environmental variables in the inefficiency function (M3) vs including environmental variables in the inefficiency function (M4) 1 5 2 6 3 7 4 8 0 (8) 33.3574 19.38 14.86 12.74 M3 not including environmental variable in the inefficiency function rejected Source: Author’s calculation from the log-likelihood functions based on the restricted (the null hypothesis) and unrestricted models as alternative hypothesis. The critical
values are based on the Chi-squared distribution (Kodde & Palm, 1986)
. July 1-2, 2014 Cambridge, UK 47
2014 Cambridge Conference Business & Economics ISBN : 9780974211428
Appendix III: Different specifications of SFA models
The Appendix III shows the alternative specifications of SFA models discussed in this section in order to derive an appropriate model that describes the sample data.
Models Specifications Inefficiency
u
Half-normal (M1)
f
(
y it
,
w it
)
u it
~
N
( 0 ,
u
2 ) Truncated (M2) Controlled (M3) TE Effects (M4)
Heterogeneity
None
f
(
y it
,
w it
)
u it
~
N
( ,
u
2 ) None
f
(
y it
,
w it
,
c it
)
u it
~
N
( ,
u
2 ) Bank-specific and time specific variables included in the cost function
f
(
y it
,
w it
,
c it
)
u it
~
N
(
z it
,
u
2 ) Bank-specific and time specific variables included in the cost function Heterogeneity in the mean of inefficiency distribution
Note: Models are presented in line with Coelli (1996) and Dong (2009)
Appendix IV: Maximum likelihood parameter estimates of the stochastic cost frontier
July 1-2, 2014 Cambridge, UK 48
2014 Cambridge Conference Business & Economics
Variables
Intercept Lny 1 Lny 2 Ln(w 1 /w 2 ) Ln(w 3 /w 2 ) 0.5Lny1Lny1 Lny1Lny2 0.5LnyLny2 0.5Ln(w 1 /w 2 )Ln(w 1 /w 2 ) Ln(w 1 /w 2 )Ln(w 3 /w 2 ) 0.5Ln(w 3 /w 2 )Ln(w 3 /w 2 ) Lny1Ln(w 1 /w 2 ) Lny1Ln(w 3 /w 2 ) Lny2Ln(w 1 /w 2 ) Lny2Ln(w 3 /w 2 )
Control variables
t (time trend) 0.5t2 Lny1t Lny2t Ln(w 1 /w 2 )t Ln(w 3 /w 2 )t z 1 (equity) 1 11 1
t
2
t
1
t
2
t
0 1 2 1 2 11 12 22 11 12 22 11 12 21 22
Parameters Model 1 (Half- normal)
-0.0525*** (0.0107) 0.5411*** (0.0205) 0.3921*** (0.0257) 0.1425*** (0.0159) 0.8191*** (0.0157) -0.6288*** (0.0835) 0.7658*** (0.0976) -0.9185*** (0.1098) 0.2910*** (0.0570) -0.3151*** (0.4222) 0.3104*** (0.0208) -0.1904*** (0.0625) 0.5573*** (0.0593) 0.1059* (0.0643) -0.4406*** (0.0743) - - - - - - 1 -
Variables
z 2 (Financial intermediation) July 1-2, 2014 Cambridge, UK
Parameters
2
Model 1 (Half- normal)
- ISBN : 9780974211428 -0.0536*** (0.0114) 0.5413*** (0.0205) 0.3925*** (0.0257) 0.1421*** (0.0156) 0.8193*** (0.0153) -0.6301*** (0.0834) 0.7675*** (0.0975) -0.9195*** (0.1103) 0.2917*** (0.0570) -0.3159*** (0.0423) 0.3105*** (0.0205) -0.1919*** (0.0619) 0.5575*** (0.0596) 0.1067* (0.0643) -0.4405*** (0.0749)
Model 2 (Truncated)
- - - - - - -
Model 2 (Truncated)
- 0.2012 (1.1000) 0.7347 (1.1000) 0.3954 (1.1000) 0.3915 (1.1000) 0.6891 (1.1000) 0.2948 (1.1000) 0.0195 (1.1000) -0.2160 (1.1000) 0.5574 (1.1000) -0.4626 (1.1000) 0.3979 (1.1000) -0.2557 (1.1000) 0.1086 (1.1000) 0.3574 (1.1000) -0.1793 (1.1000)
Model 3 (Controlled)
-0.0188 (1.1000) 0.0005 (1.1000) -0.0021 (1.1000) -0.0077 (1.1000) -0.0128 (1.1000) 0.0067 (1.1000) 0.0041 (1.1000)
Model 3 (Controlled)
-0.1821 (1.1000)
Model 4 (Tech. Eff. effects)
0.1291*** (0.0415) 0.8284*** (0.0662) 0.3432*** (0.0749) 0.5133*** (0.0723) 0.6047*** (0.0399) 0.3663*** (0.1176) -0.0076 (0.1081) -0.1615 (0.1035) 0.7493*** (0.1105) -0.5605*** (0.0537) 0.4079*** (0.0203) -0.1652** (0.0808) 0.0624 (0.0770) 0.3104*** (0.0919) -0.1961** (0.0863) -0.0011 (0.0054) 0.0004 (0.0003) -0.0077* (0.0042) -0.0051 (0.0046) -0.0207*** (0.0048) 0.0125*** (0.0029) 0.0029** (0.0014)
Model 4 (Tech. Eff. effects)
-0.1798*** (0.0139) 49
2014 Cambridge Conference Business & Economics ISBN : 9780974211428 DTr (1= transition period, 0=otherwise) DPs (1= post-reform period, 0= otherwise) Correlates of bank inefficiencies/environmental variables Intercept OWN (1=Public, 0=otherwise) ID (1=Independent director in the bank board, 0=otherwise) PD(1=Political director in the bank board,0=otherwise) SIZE DTr (1= transition period, 0=otherwise) DPs (1= post-reform period, 0= otherwise) t (time trend) CR3: 3-bank deposit concentration ratio Sigma-squared 3 4 1 0 3 5 7 8 2
s
2 4 6 2
v
2
u
- - - - - - - - - - - 0.0053*** (0.0019) - - - - - - - - - - - 0.0045 (0.0029) 0.0129 (1.1000) 0.0018 (1.1000) - - - - - - - - - 0.0021 (1.1000) -0.0247 (0.0204) -0.1301*** (0.0298) 1.1008*** (0.2493) -0.0337** (0.0159) -0.0151 (0.0715) 0.0933*** (0.0135) -0.0592** (0.0236) -0.0794*** (0.0301) 0.2889*** (0.0527) -0.0277*** (0.0058) -0.5270*** (0.1874) 0.0017*** (0.0003) Gamma
s
2
u
2 0.5887*** (0.1533) 0.5174* (0.3088) 0.4200 (1.1000) 0.5492*** (0.1124) Log likelihood Number of observations 576.9767 360 577.0039 360 687.8661 360 701.2986 360 Note: Asymptotic standard errors are in parentheses. The pre-reform period is considered as the base period for other two period dummy variables, transition period (DTr) and post-reform period (DPs). The computer
programme FRONTIER 4.1, developed by Tim Coelli (1996) has been used for estimation.
*** denotes statistical significance level at 1% ** denotes the level of statistical significance at 5% * denotes statistical significance level at 10% Source: Author’s calculation July 1-2, 2014 Cambridge, UK 50
2014 Cambridge Conference Business & Economics ISBN : 9780974211428
Year
1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995
Year Agrani
0.9621 0.9032 0.9846 0.9874 0.9365 0.9511 0.9634 0.9696 0.9474 0.9673 0.9785 0.9881 0.9915
Agrani Appendix V: SFA cost efficiency scores for 12 largest banks in Bangladesh, 1983-2012 Janata Rupali
0.9056 0.9146 0.8733 0.8919
Sonali AB National City
0.9239 0.9370 0.9117 0.9630 0.9167 0.9492 0.9619 0.8919 0.9257 0.9289 0.9284 0.9506 0.9680 0.9715 0.9432 0.9591 0.8977 0.9063 0.9202 0.9398 0.9482 0.9847 0.9277 0.9222 0.9399 0.9360 0.9069 0.9453 0.9607 0.9453 0.9500 0.9254 0.9632 0.9656 0.9441 0.9598 0.9755 0.9687 0.9764 0.9833 0.9809 0.9858 0.9652 0.9683 0.9764 0.9758 0.9794 0.9822 0.9746 0.9809 0.9839 0.9583 0.9749 0.9804
Janata Rupali
0.9833 0.9845 0.9868 0.9845 0.9875 0.9882 0.9778 0.9822 0.9903 0.9879 0.9902
Sonali AB
0.9817
National City
0.9044 0.9204 0.9578 0.9701 0.9808 0.9739 0.9526 0.9696
IFIC
0.8741 0.9673 0.9503 0.9531 0.9434 0.9748 0.9800 0.9852 0.9745 0.9817 0.9879 0.992 0.9934
IFIC UCB
0.9183 0.8952 0.8978 0.9227 0.9484 0.9590 0.9779 0.9803 0.9687 0.8976 0.9813 0.9849 0.9864
UCB Pubali
0.8524 0.8669 0.8587 0.8697 0.8798 0.9036 0.9094 0.9270 0.9107 0.9223 0.9346 0.9521 0.9849
Pubali Uttara Islami
0.9611 0.8311 0.9434 0.8971 0.9549 0.9338 0.9087 0.9327 0.9363 0.9452 0.9498 0.9646 0.9643 0.9661 0.9750 0.9827 0.9554 0.9769 0.9666 0.9846 0.9719 0.9884 0.9877 0.9959 0.9874 0.9934
Uttara Islami
July 1-2, 2014 Cambridge, UK 51
2014 Cambridge Conference Business & Economics 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
Year
2010 0.9018 0.9120 0.9269 0.9675 0.9756 0.9795 0.9847 0.9813 0.9822 0.9822 0.9811 0.9879 0.9885 0.9904
Agrani
0.9903 ISBN : 9780974211428 0.8867 0.9034 0.9154 0.9527 0.9639 0.9864 0.9889 0.9717 0.9745 0.8746 0.8931 0.9119 0.9169 0.9299 0.9645 0.9735 0.9664 0.9757 0.9813 0.9854 0.9869 0.9782 0.9826 0.9853 0.9905 0.9919 0.9871 0.9885
Janata Rupali
0.9949 0.9923 0.9041 0.8897 0.9174 0.9227 0.9515 0.9252 0.9358 0.9677 0.9402 0.9726 0.9731 0.9588 0.9735 0.9739 0.9866 0.9790 0.9826 0.9847 0.9839 0.9699 0.9859 0.9826 0.9757 0.9875 0.9865 0.9874 0.9849 0.8783 0.8987 0.9246 0.9481 0.9756 0.9801 0.9865 0.9819 0.9852 0.9921 0.9892 0.9799 0.9908 0.9804 0.9845 0.9928 0.9929 0.9920 0.9887 0.9924 0.9936 0.9872 0.9946 0.9924 0.9890 0.9933 0.9939 0.9925 0.9941
Sonali AB National City
0.9935 0.9923 0.9957 0.9919 0.9219 0.8989 0.9117 0.9808 0.9859 0.9876 0.9864 0.9808 0.9876 0.9561 0.9750 0.9756 0.9848 0.9854
IFIC
0.9833 0.8402 0.9053 0.9199 0.9346 0.9541 0.9731 0.9799 0.9814 0.9854 0.9841 0.9907 0.9919 0.9939 0.9948
UCB
0.9918 0.8827 0.8519 0.8594 0.8923 0.9067 0.9184 0.9272 0.9173 0.9196 0.9230 0.9188 0.9305 0.9629 0.9783
Pubali
0.9732 0.8887 0.9052 0.9019 0.9456 0.9161 0.9554 0.9451 0.9730 0.9624 0.9671 0.9722 0.9731 0.9798 0.9811 0.9756 0.9843 0.9795 0.9887 0.9835 0.9885 0.9866 0.9902 0.9879 0.9923 0.9899 0.9932 0.9933 0.9946
Uttara Islami
0.9923 0.9950 July 1-2, 2014 Cambridge, UK 52
2014 Cambridge Conference Business & Economics ISBN : 9780974211428 2011 2012 0.9928 0.9924 0.9944 0.9939 0.9920 0.9929 0.9954 0.9954 0.9965 0.9933 0.9961 0.9967 0.9947 0.9956 0.9921 0.9944 0.9945 0.9945 0.9904 0.9932 0.9952 0.9960 0.9957 0.9962 Note: Efficiency scores estimated by using a panel data of 12 commercial banks in Bangladesh over the period, 1983-2012 (360 observations) using
computer programme FRONTIER 4.1(Coelli, 1996)
July 1-2, 2014 Cambridge, UK 53
2014 Cambridge Conference Business & Economics
End Notes
ISBN : 9780974211428 Merchant banks operate as issue manager or underwriter for initial public offering (IPO) after BSEC approval. 2 If the multicollinearity problem is created by a strong positive correlation between second order terms in the translog cost function, maximum likelihood estimates are still unbiased and
3 The cost elasticity with respect to physical capital price can be recovered by using the linear homogeneity restriction calculated by (1 1 2 ). 4 BASEL accord is the recommendations made by the committee on banking supervision, Bank of International Settlement (BIS). July 1-2, 2014 Cambridge, UK 54
