Analysis of Banking Industry…… (Banon Amelda; Erna Bernadetta S) 53

ANALYSIS OF BANKING INDUSTRY PERFORMANCE EFFICIENCY

IN INDONESIA USING PARAMETRIC AND NONPARAMETRIC

METHODS

Banon Amelda1; Erna Bernadetta S2

1,2Accounting and Finance Department, Faculty of Economic and Communication, Bina Nusantara Unversity

Jl. K. H. Syahdan No.9, Palmerah, Jakarta 11480, Indonesia
1Banonamelda76@gmail.com; 2ernabs@yahoo.com

ABSTRACT

This research aimed to measure the efficiency performance of the banking industry in Indonesia by

using parametric and nonparametric methods, as measured by the stabilization of bank performance efficiency

based on the time series from year to year and to identify which variables to the value of efficiency. The

analytical method applied the parametric method with cross section approach of Stochastic Frontier Analysis

(SFA) while for nonparametric method used intermediation approach from Data Development Analysis (DEA)

CRS and VRS model. The data of this research was the financial statements of banks listed on the stock exchange

for the period 2012-2016 with 29 databanks processed with the help of Stata 12. From the results of the analysis

using the three measures of efficiency, it is known that the efficiency value with Cross Section Stochastic

Frontier Analysis shows a stable and high efficient conditions for all banks. While nonparametric methods show

different efficiency levels for each bank, which with DEA CRS model not all banks show an efficient

performance, only 26,90% on average each year banks have efficient performance, and 99,31% of banks

perform efficiently according to VRS model.

Keywords: performance, efficiency, Stochastic Frontier Analysis (SFA), Data Development Analysis (DEA)

INTRODUCTION

Performance appraisal is an important thing that must be conducted either by management,

shareholder or employees that achieve company goals. Banking as an intermediary institution or as a

supporter in the financing system (Republik Indonesia, 1998) needs to maintain prudence in managing

and maintaining risks so that business processes can be run, on a sustainable basis and the national

economy can be maintained properly. Therefore, performance assessment of financial condition

becomes the most crucial and must be conducted so that unhealthy financial condition can be

identified and detected earlier.

Efficiency is one of the performance parameters that underlie all the organizational

performance, including banking. Muazaroh et al. (2012) define the efficiency as an organizational

ability to maximize output by using certain inputs or using minimal inputs to produce output. This

agrees with research described by Gordo in Muljawan et al. (2014) that efficiency is the ratio between

input and output. This measure refers to technical or operational efficiency that reflects a company's

ability to obtain optimal output from an input used, or vice versa, a company's ability to utilize at least

an input to produce a certain amount of output. The more efficient the banking operations, the higher

the income or profit of the banking and the more competitive.

In the assessment of the efficiency of banking performance, it can be done through the

traditional approach and frontier approach. A traditional approach is an approach that uses Index

mailto:Banonamelda76@gmail.com
mailto:ernabs@yahoo.com

54 Journal The WINNERS, Vol. 19 No. 1, March 2018: 53-67

Number or Ratio, such as Return on Asset (ROA), Capital Adequacy Ratio (CAR), and Profitability

Ratio. While the frontier approach is an approach based on the optimal company behavior that

maximizes output or minimizes costs. The frontier approach itself is divided into two methods, namely

parametric method and nonparametric method. The parametric method is a method the reckon in the

random error and produces statistical inference. For this type of parametric approach, it consists of

Stochastic Frontier Approach (SFA) and Distribution Free Approach (DFA). The difference between

the two approaches is how to separate the inefficient size of each bank and the random error (Fries &

Taci, 2004).

In this research, the parametric method used is the cross-section approach of SFA single

equation model for data panel with pooled effect assumption which can be considered as cross-section

data (no time). This formula follows the SFA form of equation (2) - (3) of Yekti et al. (2015) as

follows where index i = 1, 2, … with n is the number of observation data.

lnTCi = a0 + a1 lnP1i + a2 lnP2i + a3 lnQ1i + a4 lnQ2i + vi – ui…………………………….. (1)

lnTCi = a0 + a1 lnP1i + a2 lnP2i + a3 lnQ1i + a4 lnQ2i with ei = vi – ui ………….. (2)

While to determine the cost efficiency ratio of a bank using cost frontier follow the form of

CEFF model equation (2,3) from Rahmawati (2011) as follows:

lnTCi = a0 + a1 lnP1i + a2 lnP2i + a3 lnQ1,i + a4 lnQ2,i + ei

TCi = exp [a0 + a1 lnP1i + a2 lnP2i + a3 lnQ1,i + a4 lnQ2,i + ei]

TCi = exp [a0 + a1 lnP1i + a2 lnP2,i + a3 lnQ1,i + a4 lnQ2,i + ei]

Where is

TĈI = exp [a0 + a1 lnP1,i + a2 lnP2,i + a3 lnQ1,i + a4 lnQ2,i + ei]

TĈI = the alleged value of the SFA model to approximate the TCi value.

The formula of the i-frontier cost for i = 1, 2, ..., n with n is the number of observational data.

𝐂𝐄𝐅𝐅𝒊 =
𝐓�̂�𝒎𝒊𝒏

𝐓�̂�𝒊
=

𝐞𝐱𝐩[𝒂𝟎 + 𝒂𝟏 𝐥𝐧𝐏𝟏,𝒊 + 𝒂𝟐𝐥𝐧𝐏𝟐,𝒊 + 𝒂𝟑 𝐥𝐧𝐐𝟏,𝒊 + 𝒂𝟒 𝐥𝐧𝐐𝟐,𝒊 + 𝒆𝒎𝒊𝒏]

𝐞𝐱𝐩[𝒂𝟎 + 𝒂𝟏 𝐥𝐧𝐏𝟏,𝒊 + 𝒂𝟐𝐥𝐧𝐏𝟐,𝒊 + 𝒂𝟑 𝐥𝐧𝐐𝟏,𝒊 + 𝒂𝟒 𝐥𝐧𝐐𝟐,𝒊 + 𝒆𝒊]

CEFF𝑖 =
TĈ𝑚𝑖𝑛

TĈ𝑖
=

exp[𝑎0 + 𝑎1 lnP1,𝑖 + 𝑎2lnP2,𝑖 + 𝑎3 lnQ1,𝑖 + 𝑎4 lnQ2,𝑖 ]. exp[𝑒𝑚𝑖𝑛 ]

exp[𝑎0 + 𝑎1 lnP1,𝑖 + 𝑎2lnP2,𝑖 + 𝑎3 lnQ1,𝑖 + 𝑎4 lnQ2,𝑖 ]. exp[𝑒𝑖 ]

𝐂𝐄𝐅𝐅𝒊 =
𝐓�̂�𝒎𝒊𝒏

𝐓�̂�𝒊
=

𝐞𝐱𝐩[𝒆𝒎𝒊𝒏]

𝐞𝐱𝐩[𝒆𝒊]
=

𝒄𝒎𝒊𝒏
𝒄𝒊

The nonparametric method can be divided into two approaches, namely Data Envelopment

Analysis (DEA) using linear programming and assume there is no random error so that DEA approach

produces more production frontier and Free Disposal Hull (FDH). In this research, the nonparametric

method used is DEA approach. DEA is a method that measures the efficiency of DMUs by employing

linear programming techniques to tightly envelop the input-output vector envelope. DEA allows

multiple input-outputs to be considered at the same time without any assumption in the data

distribution (Ji and Lee, 2010). In each case, efficiency is measured in a pattern of proportional change

in input or output. A DEA model can be divided into two orientations, namely an input-oriented model

Analysis of Banking Industry…… (Banon Amelda; Erna Bernadetta S) 55

that minimizes input when given at least the given output level, and an output-oriented model that

maximizes output without requiring more observed input values.

It is described in Ji and Lee (2010) that the DEA model is also divided into two approaches

from returns to scale by adding weight constraints. Where in the literature mentioned that DEA was

originally proposed by Charnes, Cooper, and Rhodes (1978) to measure the efficiency of DMUs with

constant returns to scale (CRS) showing that all DMUs operate on their optimal scale. After that

Banker, Charnes, and Cooper (1984) introduced a measure of efficiency model with variable returns to

scale (VRS) allowing breakdown efficiency in technical efficiency and scale efficiency within DEA.

The efficiency of observation B is defined as for BBBBCRSinputB 010,, / the input-oriented

DEA CRS model and represents the other one that can obtain the same output by reducing the input by

the ratio CRSinputB ,,1  . The efficiency for the output-oriented DEA CRS model is defined as

CBBBCRSoutputB 33,, / and represents the other one that can obtain the same input by increasing the

output by the ratio CRSoutputB ,,1  .

Based on the relative efficiency of the input-oriented VRS frontier is defined as

BBBBVRSinputB 020,, / . All efficiency measures of DMU C are the same regardless of orientation

due to frontiers.

Figure 1 Concepts of Efficiency and Returns to Scale

(Source: Ji and Lee, 2010)

This may lead to inefficient CRS technical efficiency in scale efficiency and "pure" technical

efficiency. In Figure 1, BB2 contribute to technical efficiency from point B by looking at the VRS

model, and BB1 contribute to technical efficiency from point B by looking at the CRS model; then

BB1 contribute to measuring scale efficiency. Application of DEA was initially introduced through

the ratio form. For each firm, a ratio of all outputs and all inputs is obtained. The optimal weight is

obtained by solving the following mathematical programming problems (Coelli et al., 2005):

jj
vu

xvyuz  /max
,

1/:





xvyutosubject jj

56 Journal The WINNERS, Vol. 19 No. 1, March 2018: 53-67

In practice, there is a problem in the formula part of the formula that has an infinite solution

value. To avoid this, given constraint v'xj = 1. Thus, the formed input-oriented CRS efficiency is

defined as (1).

j
vu

yuz 
,

max
(1)







YuXv

xvtosubject j

A group of observed DMUs is described by DMUj with j = 1, 2, ..., n. The variables xj and yj

are the input vectors and the output vectors respectively. The row vectors u and v are the output

multiplier and the input multiplier respectively. Form (1) is known as a form multiplier. The matrices

X and Y are the input and output matrices, respectively. The purpose of input-oriented DEA is to

minimize the actual input, relative to a given actual output, to limit (subject to) constraints that non-

DMUs can operate across the set of possible production and constraints associated with non-negative

weights. In practiced the most DEA programs are available using a dual form as depicted in (2) in

which this model lowers the load calculation and is amount (1) (Ji and Lee, 2010).


 ,

min
(2)











Xxtosubject

where λ is a positive facet vector in Rk and θ is a real variable. The calculation procedure for (2) can

be expressed as:




min
(3)





 ss

ss ,,
min


(4)














ysY

sXxtosubject

where s +, s-, and λ are semipositive vectors in Rk and θ is a real variable. The Model DEA single-

stage completes (3), while the two-stage DEA model completes (3) followed by (4) as aconsequences

(Ji and Lee, 2010).

The DEA CRS model assumption is appropriate when all firms operate at optimal scales, but

imperfect competition (government regulation or financial constraints) may cause a company not to

run at an optimum scale. Thus, the DEA CRS model needs to consider also the condition of variable

returns to scale (VRS). By using the CRS provisions when all firms are not running at optimal scales,

the results in technical efficiency measures (TE) disrupt scale efficiencies (SE). The use of the VRS

provision allows the calculation of TE to overcome the SE effect. The linear programming problem of

CRS can be easily modified in calculating the VRS form by adding convexity constraint to (2), as to

obtain the following linear programming VRS problems (Coelli et al., 2005).


 ,

min
(5)













Xxtosubject

Analysis of Banking Industry…… (Banon Amelda; Erna Bernadetta S) 57

To calculate scale efficiencies (SE), the ratios of technical efficiencies (TE) CRS and technical

efficiencies (TE) VRS are used; SE = TECRS / TEVRS. The DEA model is output-oriented in contrast

to the DEA input orientation. The following are given the following linear programming problems

(Coelli et al., 2005).


DEA CRS is output-oriented



,
min

(6)













Xxtosubject


DEA VRS output-oriented



,
min

(7)















Xxtosubject

Research to measure the efficiency of bank performance has been done by many researchers.

Ferrier and Lovell (1990) analyzed the efficiency rates of 575 banks in the US using the SFA and

DEA methods, which found that the level of bank efficiency in the US with the DEA method was

higher than that of SFA. Hasan and Hunter (1996) have examined the efficiency issue of Japanese

multinational banks in the United States, which the results found that the mean SFA efficiency was

higher than that of the average TFA efficiency. Fiorentino, Karmann, and Koetter (2006) have

conducted cost-efficiency studies on 34.192 universal banks in Germany using the Stochastic Frontier

Approach and Data Envelopment Analysis methods, which from the results of his research indicate

that the average cost efficiency level substantially higher by SFA compared to DEA. Yassine and

Soumia (2016) have examined the effect of certain bank-specific factors; profitability, bank size, and

ownership status on differences in efficiency by using a parametric and nonparametric approach to

banks in Algeria, whose research results indicate a relative consistency between two approaches and

more frontier methods.

In Indonesia, Rahmi (2008) in her research on Sharia Business Unit in Indonesia in 2005-2007

period using SFA and DEA technique concluded that by using DEA which have high-efficiency level

is SBU BTN and BPD DKI, while by using SFA the most efficient SBUs are BPD JABAR and DKI.

Hartono (2009) has used the Stochastic Frontier Approach (SFA) technique in his study concluded that

there is a high level of efficiency for banks listed on the stock exchange during the period 2004-2009,

especially groups of Non-Foreign Exchange Banks and banks with small capitalization. According to

Siregar, Mariana, and Umanto (2015) researched commercial banks in Indonesia for the period 2009-

2013 with the technique of Data Development Analysis (DEA) concluded that the average of

commercial banks listed on the Stock Exchange during the period 2009-2013 efficient. However, for

that period the Non-Foreign Exchange BUSN group has a higher efficiency level than the other

groups.

The difference in efficiency level results is shown by using two different approaches that have

been applied to US banks and banks in Germany using criteria from Bauer et al. (1998). It includes

efficiency level, efficiency rating, identification of extreme performers, time consistency, and the

correlations consistent with the accounting indicators used make the basis for the authors to conduct

the same research on the overall bank in Indonesia whether efficiency measures produce consistent

58 Journal The WINNERS, Vol. 19 No. 1, March 2018: 53-67

results on efficiency levels, efficiency ratings, extreme performers and time consistency. Because

measuring the consistency of efficiency over time and does not vary from year to year, knowing the

extreme bank performers, as well as the efficiency rating, is very important, it is used as a baseline to

determine a policy for policy developers in this case Bank Indonesia and the Financial Services

Authority.

Based on the results of previous research studies, it is known that there are many studies on

the efficiency of both methods. However, to analyze the stabilization of bank's performance efficiency

by using parametric SFA and nonparametric DEA method still does not exist. Differences in the use of

bank samples, time periods, input and output specifications and different methods as the second reason

for the authors to analyze three measures of efficiency using the same bank sample, same period, the

same input and output specifications that applied to three different efficiency measures. Concerning

these issues, the purposes of this research are to determine the level of bank efficiency measured using

three measures of efficiency, to know the factors that influence the efficiency of bank performance

during the observation period, and to measure the consistency of bank efficiency in Indonesia by using

two comparison methods SFA and DEA. Where in this study, it test the efficiency results developed

from four Bauer et al. (1998) criteria to be able to determine the performance of the bank in terms of

efficiency, efficiency rating, identification of extreme performers, and time consistency.

METHODS

This research is a quantitative research using parametric research method with cross section

approach of Stochastic Frontier Approach (SFA) and non-parametric Data Envelopment Analysis

(DEA) model of constant returns to scale (CRS). It is with assumption when all companies operate at

optimum scale and model variable returns to scale (VRS) when the company operates on an abnormal

scale due to imperfect competition (government regulation or financial constraints). The data used in

this study is secondary data in the form of bank financial statements for the period of the observation

year 2012-2016 with the following criteria: (1) Samples used are banks that have to go public in

Indonesia; (2) Banks used are banks that have been listed on the Jakarta Stock Exchange (BEJ) within

the period of the observation year 2012-2016 and not delisting during the observation period to avoid

bias in the results of research; (3) The sample banks have no loss and have complete financial report

during the observation period and has been audited by audit firm; (4) The sample banks have

published their financial statements. Based on the criteria for determining the sample above, it selected

to be observed were 29 banks, consisting of 14 BUSN foreign exchange, two non-foreign exchange

banks, four state-owned banks, two joint venture banks and seven regional development banks. Where

to take the sample is done by purposive sampling.

According to Hadad et al. (2003) in Purwanto (2011), there are three commonly used

approaches in both parametric Stochastic Frontier Analysis (SFA) and non-parametric Data

Envelopment Analysis (DEA) methods. It defines the relationship between input and output in a

financial activity financial institutions (banks), namely; (1) Asset approach describes the main

function of the financial institution as the lender of the loan, in this approach the output is actually

defined into the asset form; (2) Production Approach, in this approach, consider financial institutions

as deposit accounts and lenders so in this approach the output is defined as the amount of labor, capital

expenditure on fixed assets, and other materials; (3) Intermediation Approach, in this approach

consider the financial institution as an intermediary, change and transfer the financial assets of the

surplus unit to the deficit unit. In this case, the institutional inputs such as labor costs, capital, and

interest financing on the deposit. While the output is measured in the form of credit loans and financial

investment.

Analysis of Banking Industry…… (Banon Amelda; Erna Bernadetta S) 59

This research uses intermediation approach, where banks have a function to collect and

channel funds from people who surplus funds to the people who need funds (deficit). According to

Berger and Humphrey (1997) in Purwanto (2011) stated that the intermediation approach is the most

appropriate approach to evaluate the financial performance of financial institutions in general due to

the characteristics of financial institutions as financial intermediation. The independent variable in this

research is the total operational cost (TC), where TC = (Total interest expense + Expense of estimated

loss of commitment and contingency + Total other operating expenses + Provision allowance expense

+ Amortization expense + Non-operational expense). To be able to generate some output (Oi) the bank

asks for input (ii) by minimizing operational costs (TC).

The dependent variable consists of three input variables and three output variables. The first

input variable is the total savings (I1). Deposits are public deposit funds to banks used by banks for

economic activities with the assurance that banks will return the funds intact to the public. The forms

of deposits include demand deposits, deposits, certificates of deposit, savings, and or other similar

forms. The second input variable is the total fixed asset (I2). It is as a proxy of the size of the bank

because the asset has economic value in the future. The higher the value of bank assets, the bank's

ability to guarantee risks for productive assets and financing or credit becomes higher. The third input

variable is the labor load (I3). The burden of manpower absorbs the greatest burden in the bank's

operational expenses because in the workforce it includes salaries, benefits, employee development.

The output variables in this research include total credit (O1) that is total loan given either to party

related to the bank or not related to the bank. Because of the total credit is the main product of the

bank as an intermediary institution. The second output variable is securities (O2), and the third output

variable is non-credit operating income (O3).

The analytical model used to determine the efficiency scores with the Stochastic Frontier

Approach (SFA) approach is to use the cross-section of the Stochastic Frontier Approach (SFA) single

equation model for the data panel with pooled effect assumptions which can be considered as cross-

(no time), with the following formula:

1nTCit = βOt + β11nO1it + β21nO2it + β31nO3it β41nI1it + β51nI2it + β61nI3it + εit………........................ (1)

where:

TC = Total cost incurred by the bank.

O1 = Total Loans disbursed by the bank, either to parties related to the bank or not.

O2 = Securities owned by the bank.

O3 = Non-Credit Operating Income

I1 = Savings Fund Interest Expense

I2 = Depreciation Expense on Fixed Assets

I3 = Burden of Labor

According to Coelli (1996) in Hartono (2009), the value of inefficiency in the cost function

ranged from 1 to up. While the analytical model used to determine efficiency scores with DEA uses an

output-oriented DEA model with Constant Return to Scale (CRS). It is introduced first by Charnes,

Cooper, and Rhodes (1978) and variable returns to scale (VRS) assuming when all firms do not

running-on an optimum scale, results in measures of technical efficiencies (TE) disrupt scale

efficiencies (SE). The use of the VRS provision allows the calculation of TE to overcome the SE

effect.

The researchers have tested the stabilization of bank performance efficiency in Indonesia by

observing the stability of bank performance over time as a recommendation for policymakers to

organize and set appropriate rules or strategies that face global competition. To that end, the

researchers classify stabilization efficiency into bank performance in Table 1.

60 Journal The WINNERS, Vol. 19 No. 1, March 2018: 53-67

Table 1 Stabilization of Bank Performance Efficiency

Efficiency Condition Performance Bank

Stable Inefficiency (SI) Worst

Stable Low Efficiency (SLE) Bad

Stable Intermediate Efficiency (SIE) Fair

Stable High Efficiency (SHE) Good

Unstable Efficiency Increase (UEI) Good

Unstable Efficiency Decrease (UED) Bad

RESULTS AND DISCUSSIONS

From the number of observation data to 29 banks listed on the Jakarta Stock Exchange in

2012-2016 for 145 observations generated descriptions of each research variables in Table 2.

Table 2 Descriptive Statistics of the Aggregate Data

(29 banks 5 years; 2012-2016)

Variable Mean Std.Dev Min Max

Dependent

Total Operating Cost (TC) 0,08334 0,02988 0,01466 0,26579

Independent

Input

Deposit (I1) 0,03560 0,01613 0,00149 0,07464

Fixed Assets (I2) 0,00966 0,00396 0,00061 0,02039

Labor (I3) 0,01975 0,00890 0,00192 0,02039

Output

Loan (O1) 0,65372 0,07991 0,39383 0,79834

Securities (O2) 0,09097 0,06865 0,00041 0,33631

Non-Credit Operating Income (O3) 0,01923 0,01811 0,00258 0,14249

Total operational cost (TC) in 2012 - 2016 from 29 bank companies shows an average value

of 0,08334. It means that the total operational cost incurred by the bank is 8,33% of total assets with a

minimum value of 1,47 % and maximum 26,58%. Inputs that are obtained by banks in the period 2012

- 2016 in the form of deposits cost of funds (I1) of 3,56%, (I2) of 0,97%, depreciation expenses of fixed

assets and labor expenses (I3) of 1,98% (O1) of 65,37%, securities (O2) of 9,10%, and non-credit

operating income (O3) of 1,92% of the total assets of the bank.

By using SFA method, the efficiency level of each bank in Indonesia can be measured. The

efficiency level is analyzed from the cost function model between total cost (TC) as dependent

variable with independent variable; interest expense of deposit fund (I1), depreciation expense of fixed

asset (I2), labor burden (I3), total loan disbursed by bank (O1), bank-owned securities (O2), and non-

credit operating income (O3) that is transformed into natural logarithms (ln) into translog models

rather than linear models. The SFA for the cross-section model (frontier) is estimated based on a

pattern of production function approaching a balanced efficiency of 1 from below (less than or equal

to 1). In addition, a likelihood-ratio test of sigma_u = 0 is used to test whether the sample data can be

worked with SFA for the cross-section (frontier) or SFA model for the panel model (xtfrontier). If the

result of P-value (Prob> = chibar2) > α with α = 0,05, then the data sample is suitable to be done with

SFA for the cross-section model (frontier). From the result of STATA output, obtained P-value equal

Analysis of Banking Industry…… (Banon Amelda; Erna Bernadetta S) 61

to 1,000 (P-value > 0,05), so the test is concluded that the data sample that is suitable to be done with

SFA for cross-section model (frontier).

Furthermore, multicollinearity test to determine whether in a regression model there is

intercorrelation between independent variables or not. Multicollinearity test is done by deflecting the

value of correlation coefficient between free variable. If the value of the correlation coefficient shows

the number of 0,75 both negative and positive, it can be concluded the existence of multicollinear here

is the result of the correlation between free variable. This result indicates that the correlation between

variables is relatively low. This indicates the absence of multicollinear problems in the model. Table 3

shows the correlation between independent variable, while Figure 2 shows the cost efficiency level for

each bank by using the SFA method.

Table 3 Correlation between Independent Variable

Var lnI1 lnI2 lnI3 lnO1 lnO2 lnO3

lnI1 1,000

lnI2 -0,213 1,000

lnI3 0,199 0,190 1,000

lnO1 0,255 -0,079 0,286 1,000

lnO2 -0,039 -0,101 -0,265 -0,634 1,000

lnO3 0,283 0,044 0,294 0,214 -0,041 1,000

Figure 2 The Cost Efficiency Level for Each Bank by the SFA Method

After examining the problem of multicollinearity and cost efficiency prediction, the next step

is to answer the hypothesis test of the problem. The hypothesis for the simultaneous test of regression

coefficient with Wald test (Chi-Square test).

H0: There is no concurrent or joint effect of lnI1, lnI2, lnI3, lnO1, lnO2, and lnO3 to lnTC.

H1: There are concurrent or joint effects of lnI1, lnI2, lnI3, lnO1, lnO2, and lnO3 to lnTC.

This test is expected to have a simultaneous or joint effect of predictor variables on response

variables based on concurrent tests of regression coefficients. This test is satisfied when the value of

P-value (Sig.) ≤ α with α is set at 5%. From the result of output, it can get the result of statistical test of

Chi-Square test (Wald chi2 (6)) 141,90 and P-value (Prob> chi2) 0,0000 so that the test can be

concluded that there are simultaneous or joint effect of lnI1, lnI2, lnI3, lnO1, lnO2, and lnO3 to lnTC

(due to P-value <0,05). There is the hypothesis to test individual regression coefficients with z test.

62 Journal The WINNERS, Vol. 19 No. 1, March 2018: 53-67

H0: There is no partial or individual influence lnI1, lnI2, lnI3, lnO1, lnO2, and lnO3 to lnTC.

H1: There are partial or individual influences lnI1, lnI2, lnI3, lnO1, lnO2, and lnO3 to lnTC.

This test is conducted to determine whether there is the significant influence of each predictor

variable to the response variable. This test shows a significant influence when the value of P-value

(Sig.) ≤ α where α equal to 5%.

Table 4 The Result of the z Test for the Significance of Stochastic Frontier Model Coefficient

Independent Variable Coeff. Std. Error Statistic z P-value

lnI1 0,08429 0,02845 2,96 0,003*

lnI2 -0,06637 0,04022 -1,65 0,099

lnI3 0,47448 0,04860 9,76 0,000*

lnO1 -0,14358 0,17404 -0,82 0,409

lnO2 0,07435 0,02471 3,01 0,003*

lnO3 -0,05361 0,03644 -1,47 0,141

constanta -0,72907 0,36833 -1,98 0,048

*Significant to a significant level ( ) of 5%

From the above equation and the value of P-value (Sig.) for each variable in Table 4, it is

known that the input variable of deposit interest expense (I1), labor (I3), and securities output variable

(O2) have a significant influence on total cost, ie smaller than the 0,05 trust level. This means that any

increase or decrease in the value of variables I1, I3 and O2 of 1 unit will affect the increase or decrease

in total cost of the coefficient of each variable. The Stochastic Frontier Model by involving all

variables for all banks formulates below.

lnTC = -0,729067 + 0,08429 lnI1 – 0,06637 lnI2 + 0,47448 lnI3 – 0,14358 lnO1 + 0,07435 lnO2 –

0,05361 lnO3+ e

The coefficient of determination (R2) is obtained at 0,49459593 which means that the lnTC

diversity capable of being described in lnI1, lnI2, lnI3, lnO1, lnO2, and lnO3 is 49.46% together with

the remaining 50,54% explained by error (e) or other variables not included in the Stochastic Frontier

Model. Based on the results of cost efficiency output with the SFA cross-section model, Cost

Efficiency in 2012 - 2016 from bank companies shows the average value of all banks. This shows that

the cost efficiency is close to the value 1 so it is concluded that the cost efficiency level for all banks

has a very good efficiency level and is expressed as high efficiency as shown in Table 5.

Table 5 Category Efficiency per Grouping

Score Efficiency (%) Level Efficiency

96-100 High Efficiency

86-95 Intermediate Efficiency

66-85 Low Efficiency

<65 Inefficiency

Source: secondary data is processed

According to the analysis with DEA CRS model, it is known that on average every year

73,10% banks in Indonesia have inefficient performance. Whereas according to the DEA VRS model,

it is only 3,45% only Banks in Indonesia that inefficient performance of the Regional Development

Bank Lampung (BLAM) and it only happened in 2012, the rest is efficient. It is shown in Figure 3.

Analysis of Banking Industry…… (Banon Amelda; Erna Bernadetta S) 63

Figure 3 The Average Score Efficiency per Group of Banks in Indonesia Period 2012-2016

The three measures of efficiency, Mixed Banks and Non-Foreign Exchange National Private

Banks show better levels of operational efficiency compared to the Group of State Owned Enterprises

(BUMN), BPD, and Private Foreign Exchange Banks (BUSN Foreign Exchange).

From the average of each year 73,10% of banks in Indonesia are inefficient 40,69% are banks

of Foreign Exchange BUSN group, 13,79% BPD, 16,34% BUMN, and the remaining 3,45% Bank

Mixed and Non -Foreign Exchange BUSN. BUSN Foreign exchange has the worst performance in

2013 where all BUSN banks have inefficient performance, while state-owned banks have a

performance where all state-owned banks are inefficient in 2012 and 2015. Table 6 shows a list of

inefficient banks during the 2012-2016 observation period according to DEA CRS Model.

Table 6 List of Inefficiency Bank Period 2012-2016 by DEA CRS Model

Group Bank 2012 2013 2014 2015 2016

BUSN Devisa BDMN BBNP BBCA BBCA

BNBA BDMN BBMI BBKP

BNII BNBA BBNP BBMI

BNGA BNGA BDMN BBNP

BPKP BNII BNBA BDMN

BSIM BPKP BNGA BNBA

MEGA BSIM BNII BNGA

SDRA MEGA BPKP BNII

BSIM BSIM

MEGA MEGA

NISP NISP

PNBN PNBN

SDRA

BUSN Non Devisa BTPN BTPN BTPN BTPN BTPN

BUMN BBNI BBNI BBNI BBNI BBRI

BBRI BBTN BBTN BBRI BBTN

BBTN BMRI BMRI BBTN BMRI

BMRI BMRI

BPD BDKI BLAM BLAM BDKI BDKI

BLAM BSLT BSLT BLAM BLAM

BNTT BSMT BSMT BNTT BSMT

BSLT BSMT

BSMT BSSB

0,00%

10,00%

20,00%

30,00%

40,00%

50,00%

60,00%

70,00%

80,00%

90,00%

100,00%

Bank
Campuran

BPD BUMN BUSN Devisa BUSN Non
Devisa

Sc
or

e
Ef

fic
ie

nc
y

SFA CRS DEA VRS DEA Limited Efficiency

64 Journal The WINNERS, Vol. 19 No. 1, March 2018: 53-67

Table 6 List of Inefficiency Bank Period 2012-2016 by DEA CRS Model (Continued)

Group Bank 2012 2013 2014 2015 2016

BPD BSSM

Mixed Bank MCOR MCOR MCOR MCOR

Based on the results of the analysis process with the three measures of efficiency can also be

known bank ratings with the lowest and highest efficiency each year as shown in Table 7.

Table 7 The Highest and Lowest Bank Ranked According to

Value the Efficiency of The Period 2012-2016

Model-rank 2012 2013 2014 2015 2016

SFA

Highest BACA AGRO BNII BSLT BACA

BACA

Lowest BBCA BNII BDKI BBRI BBCA

BBRI BDKI BBRI

SDRA BDKI

DEA CRS

Highest AGRO AGROS SDRA AGRO BACA

BACA SDRA BACA BVIC

BBNP BSLT PNBN

BSBR BVIC

BVIC

Lowest BNBA BDMN BDMN BDMN

DEA VRS

Highest AGRO AGRO AGRO AGRO BACA

BACA BDKI MCOR BACA BSLT

BBNP BVIC SDRA BBTN BVIC

BSBR PNBN BSLT MEGA

BVIC SDRA BVIC PNBN

Lowest BLAM BMRI BDMN BDMN BBNI

The rating of the bank with the highest efficiency value is dominated by banks from the

Foreign Exchange BUSN group while the lowest rank is dominated by the group of foreign exchange

BUSN banks and BUMN. The result of DEA VRS analysis process, it is known that AGRO bank for

the fourth consecutive year ranked highest for efficiency value that is from the 2012-2014 period,

while based on the result of analysis process with DEA CRS only 2012, 2013, and 2015 ARGO bank

reach the highest rank. In contrast to SFA, the ARGO bank is the highest in 2013 alone. Likewise,

with BACA banks with three efficiency measures in 2012 and 2016 is the highest for efficiency.

However, in 2013 only the SFA method alone, BACA is ranked highest while with DEA and DEA

VRS DEA method just occurred in 2015. With DEA CRS and DEA VRS BVIC banks are highest in

2012, 2013, 2015 and 2016.

The lowest ranked bank is achieved by BDMN in 2013-2016 with DEA CRS and DEA VRS

methods. State-owned banks are among the lowest in the period 2012-2016, namely BBNI, BBRI, and

BMRI. Bank BMRI achieves the lowest rank by DEA VRS method in 2013, while BBNI in 2016.

While Bank BBRI reaches the lowest rank three years in a row according to SFA method, that is from

2013-2016.

Analysis of Banking Industry…… (Banon Amelda; Erna Bernadetta S) 65

Table 8 Condition of Bank Performance Period 2012-2016

No Bank
Condition Performance Bank

SFA DEA CRS DEA VRS SFA DEA CRS DEA VRS

1 AGRO SH UDE UDE Good Bad Bad

2 BACA SHE UIE UIE Good Good Good

3 BBCA SHE UDE UDE Good Bad Bad

4 BBKP SHE SI SIE Good Worst Fair

5 BBNI SHE SI SLE Good Worst Bad

6 BBNP SHE UDE UDE Good Bad bad

7 BBMI SHE UDE SIE Good Bad Fair

8 BBRI SHE UDE UEI Good Bad Good

9 BBTN SHE SI UDE Good Worst Bad

10 BDKI SHE UDE UDE Good Bad Bad

11 BDMN SHE SI SLE Good Worst Bad

12 BLAM SHE SI UEI Good Worst Good

13 BMRI SHE SI SLE Good Worst Bad

14 BNBA SHE SI UDE Good Worst Bad

15 BNGA SHE SI UEI Good Worst Good

16 BNII SHE SI UDE Good Worst Bad

17 BNTT SHE SI UEI Good Worst Good

18 BSBR SHE SLE UDE Good Bad Bad

19 BSIM SHE SI UDE Good Worst Bad

20 BSLT SHE UEI UEI Good Good Good

21 BSMT SHE SI UDE Good Worst Bad

22 BSSB SHE SI SIE Good Worst Fair

23 BTPN SHE SI UEI Good Worst Good

24 BVIC SHE UEI UEI Good Good Good

25 MCOR SH UED UED Good Bad Bad

26 MEGA SHE SI UEI Good Worst Good

27 NISP SHE UEI SIE Good Good Fair

28 PNBN SHE UEI UEI Good Good Good

29 SDRA SHE UED UEI Good Bad Good

Based on time series sampling as shown in Table 8, the stabilization efficiency with SFA

cross-section indicates that all banks listed on the Jakarta stock exchanges perform at stable high-

efficiency conditions. This means that all bank companies in Indonesia have a high efficient and good

performance. Meanwhile, according to the non-parametric method of DEA both CRS and VRS, some

banks show unstable performance. DEA CRS notes that 13 banks in Indonesia listed on the Jakarta

stock exchange have an unstable performance. Of the 13 unstable banks, eight banks show a decrease

in efficiency or poor performance, such as AGRO, BBCA, BBNP, BBMI, BBRI, BDKI, MCOR, and

SDRA bank. While five banks show an increase in efficiency level or have good performance, such as

BACA, BSLT, BVIC, NISP, and PNBN. Among the banks with stable performance, one bank is in

stable condition of low efficiency or has poor performance that is BSBR bank, while 14 other banks

are in stable condition inefficient or have very bad performance like BBKP, BBNI, BDMN, BLAM,

BMRI, BNBA, BNGA, BNII, BNTT, BSMT, BSLT, BSSB, BTPN, and MEGA.

DEA VRS identifies seven banks from 29 banks listed on the Jakarta stock exchange with

stable performance with efficiency and performance appraisal that varied, while the remaining 22

banks are in unstable condition. Of the seven banks that have stable performance, four of them are in

the stable condition of middle efficiency or good enough performance, such as BBKP, BBMI, BSSB,

and NISP. While the three banks stable low efficiency or have poor performance, such as BBNI,

BDMN, and BMRI. The 22 unstable banks consisted of 11 banks in unstable conditions, where the

bank's efficiency level decreases or shows poor performance, such as AGRO, BBCA, BBNP, BBTN,

BDKI, BNBA, BNII, BSBR, BSIM, BSMT and MCOR. The rest of the 11 banks are in an unstable

66 Journal The WINNERS, Vol. 19 No. 1, March 2018: 53-67

condition, where the efficiency of the bank has improved or shows good performance, such as READ,

BBRI, BLAM, BNGA, BNTT, BSLT, BTPN, BVIC, MEGA, PNBN and SDRA.

From the comparison of the two DEA CRS and DEA VRS models, it can be concluded that

there are different efficiency conditions in some banks, including BBKP, BBNI, BDMN, BMRI, BNII,

BSSB, AGRO, BBCA, BBMI, BBNP, BBRI, BDKI, MCOR, and SDRA. According to the DEA CRS

model, the condition of these banks is stable and declining to inefficient. Whereas according to the

DEA VRS model some of the banks above are stable and declining in performance, yet, still in the

efficient condition only change level only. This reinforces the statement from Ji and Lee (2010) that

CRS technical efficiency is not efficient in scale efficiency and "pure" technical efficiency. So, the

efficiency conditions of some banks with DEA VRS conditions are efficient, but with DEA CRS can

be inefficient. While the difference in efficiency values of SFA and DEA is due to differences in

determining the efficiency level which for the SFA parametric method of determining the efficiency

level reckon the approach of random error while DEA does not reckon the existence of random error,

but rather the output and input approaches determine DMU.

CONCLUSIONS

Based on the results of the analysis, it can be drawn some conclusions as (1) Total operating

expenses are incurred by banks during the period 2012-2016 average 8,33% of total assets with a

minimum value of 1,47% and a maximum of 26,58%. (2) The amount of bank operational expenses

during the period 2012-2016 is strongly influenced by the decrease or increase of interest expense of

deposit fund equal to 8,5%, labor cost equal to 47,45%, and securities equal to 7,43%. (3) There is a

difference in the level of bank efficiency with the three measures of efficiency, where all banks with

parametric approach cross section SFA show a very efficient performance with high-efficiency level

approaching 100%, while with nonparametric DEA approach, not all banks show efficient

performance, only 26. An average 90% on average per year is a bank that has efficient performance

according to CRS model and 99,31% according to VRS model. (4) Based on the efficiency rating, a

fairly a stable Bank is ranked the highest efficiency during the period 2012-2016 are AGRO, BACA

and BVIC banks, while BDMN and BBRI banks are the lowest for efficiency during 2012-2016. (5)

DEA method CRS identifies banks that perform poorly and very poorly during the period 2012-2016,

Bank AGRO, BBCA, BBNI, BBMI, BBRI, BDKI, MCOR, SDRA BBKP, BBNI, BDMN, BLAM,

BMRI, BNBA, BNG, BNII, BNTT, BSMT, BSLT, BSSB, BTPN, and MEGA. While banks that have

good performance during the period 2012-2016 are banks BACA, BSLT, BVIC, NISP, and PNBN. (6)

The DEA VRS method of familiarizing banks that perform poorly during the period 2012-2016 is

bank AGRO, BBCA, BBNP, BBTN, BDKI, BNBA, BNII, BSBR, BSIM, BSMT, BBNI, BMRI,

BDMN, and MCOR. While banks that perform well during the period 2012-2016 are banks BACA,

BSLT, BVIC, BBKP, BBMI, BSSB, NISP, and PNBN.

REFERENCES

Banker, R. D., Charnes, A., & Cooper, W. W. (1984). Some models for estimating technical and scale

ineffeciencies in data envelopment analysis. Management Science, 30(9), 1078–1092.

Bauer, W. P., Berger, N. A., Ferrier, D. G, & Humphrey, B. D. (1998). Consistency conditions for

regulatory analysis of financial institutions: A comparison of Frontier Efficiency Methods.

Journal of Economics and Busines, 50, 85-144.

Charnes, A., Cooper,W. W., & Rhodes, E. (1978). Measuring the efficiency of decision making units.

European Journal of Operational Research, 2, 429-444.

Analysis of Banking Industry…… (Banon Amelda; Erna Bernadetta S) 67

Coelli, T. J., Rao, D. S. P., O’Donnel, C. J., Battese, G. E. (2005). An introduction to efficiency and

productivity analysis. USA: Springer.

Ferrier, D. G., & Lovell, C. A. K. (1990). Measuring cost efficiency in banking; Econometric and

linear programming evidence. Journal of Econometrics, 46(2), 229–245.

Fiorentino, E., Karmann, A., & Koetter, M. (2006). The cost efficiency of German banks: A

comparison of SFA and DEA. Retrieved on May 15th, 2017 from https://papers.ssrn.com.

Fries, S., & Taci, A. (2004). Cost efficiency of banks in transition: Evidence from 289 banks in 15

post-communist countries. European Bank for Reconstruction and Development Working

Paper, 86, 1-29.

Hartono, E. (2009). Cost efficiency analysis of Indonesian banking industry using parametric method

Stochastic Frontier Analysis. Master Thesis. Semarang: University of Diponegoro.

Hasan, I., & Hunter, W. C. (1996). Efficiency of Japanese multinational banks in the United States.

Research in Finance, 14, 157-173.

Ji, Y. B., & Lee, C. (2010). Data envelopment analysis. The Stata Journal, 10(2), 267-280.

Muazaroh., Eduardus, T., Husnan, S., & Hanafi, M. M. (2012). Determinants of bank profit efficiency:

Evidence from Indonesia. International Journal of Economics and Finance Studies, 4(2), 163-

173.

Muljawan, D., Hafidz, J., Astuti, R. I., & Oktapiani, R. (2014). Determinants of Indonesia banking

efficiency and its impact on credit interest calculation. Working Paper of Bank Indonesia.

Purwanto, R. (2011). Comparative analysis of efficiency of conventional commercial bank (Buk) and

sharia (Bus) commercial bank in Indonesia with Data Envelopment Analysis (DEA) method

(period 2006-2010). Master Thesis. Semarang: University Diponegoro.

Rahmawati, R. (2011). Cost efficiency enhancement strategy in sharia commercial banks based on

Stochastic Frontier Approach And Data Envelopment Analysis. Buletin Monetary and

Banking Economy, 17(4), 457-480.

Rahmi, M. S. (2008). Analysis of efficiency of sharia business unit in Indonesia Data Envelopment

Analysis Method (DEA) and Sthocastic Frontier Approach (SFA). Master Thesis. Bogor:

Tazkia University College of Islamic Economic.

Republik Indonesia. (1998). Undang-undang no.10 tahun 1998 tentang perubahan atas unudang-

undang no. 7 thaun 1992 tentang perbankan. Tambahan Lembaran Negara Republik

Indonesia, No. 3790. Sekretariat Negara. Jakarta.

Siregar, L. M., Mariana., and Umanto. (2015). Analysis of the efficiency of performance of commercial

banks with data development methods analysis. Retrieved on April 18th, 2017.

Yassine, B., & Soumia, A. H. (2016). Assessing cost and profit efficiency by a joint application of

parametric and non-parametric approaches: Evidence from the Algerian banking system.

EconWorld 2016 @ImperialCollage Proceedings. London, UK. Pp. 1-29.

Yekti, A., Darwanto, H. D., Jamhari., & Hartono, S. (2015). Technical efficiency of melon farming in

Kulon Progo: A Stochastic Frontier Approach (SFA). International Journal of Computer

Applications, 132(6), 975 - 8887.

https://papers.ssrn.com/