Siddhant

Volume 15, Issue 2, April-June, 2015, pp- 160-171

IndianJournals.com
A product of Diva Enterp rises Pvt. Ltd.

DOI: 10.5958/2231-0657.2015.00017.8

Determinants of Public Sector Banks Profitability in India: An

Empirical Study
Gurmeet Singh

Downloaded From IP - 103.37.201.68 on dated 12-Oct-2015

www.IndianJournals.com

Members Copy, Not for Commercial Sale

ABSTRACT
The economic development of country depends on the effective banking system. In recent years, there have been
considerable pressures on the profitability of banks. The study investigates the long run equilibrium relationship
between the return on assets (ROA) and nine variables over the period 2005 to 2014. Johansens co-integration test
and vector error correction model have been applied to explore the long-run equilibrium relationship. The analysis
reveals that the returns on assets of the Indian public sector banks are positively co-integrated with all the nine
variables under the study and hence, a long-run equilibrium relationship exists between them. The results of
VECM reveal that secured advances to total advances, burden to total assets, burden to interest income, operating
profits to total assets, non-interest income to total assets and profit per employee causes return on asset in long
run and short run. While, non performing assets, deposits to total liabilities and net interest margins causes return
on assets in short run only. The findings from Granger causality based on the VECM indicate bidirectional
causality exists between all the variables tested and return on assets except non-performing assets. Non-performing
assets granger causes return on assets but not the other way round.
KEYWORDS: Indian commercial banks, Co-integration test, VECM, Causality test, Bank profitability, Nonperforming assets, Augmented Dickey Fuller test

INTRODUCTION
In the economic development of country, banks play
an important role. Their ability to make a positive
contribution in igniting the process of growth depends
on the effective banking system. In recent years, there
have been considerable pressures on the profitability
of banks. Profitability is considered to be an index of
financial health. The term profitability refers to an
indication of the efficiency with which the operation of
the business is carried on. Poor operational performance
may indicate poor selling of bank products and hence
poor profits. A lower profitability may rise due to lack
of control over the expenses.
The importance of bank profitability can be appraised
at the micro and macro levels of the economy. At the

micro level, profit is the essential prerequisite of a

competitive banking institution and the cheapest source
of funds. Without profits, any firm cannot attract outside
capital (Gitman, 2007). Thus, profits play a key role in
persuading depositors to supply their funds on
advantageous terms. By reducing the probability of
financial trouble, impressive profits figures also help
reassure a banks other stakeholders, namely, investors,
borrowers, managers, employees, external product and
service suppliers, and regulators (Anyanwaokoro, 1996).
It is not merely a result, but also a necessity for
successful banking in a period of growing competition
on financial markets. Hence, the basic aim of a banks
management is to achieve a profit, as the essential
requirement for conducting any business (Bobkov,
2003). At the macro level, a sound and profitable banking

Assistant Professor, Unitedworld School of Business, Uvarsad, Gandhinagar, Gujarat, India

Email id: prof.gurmeetsingh@live.com
Siddhant

160

Determinants of Public Sector Banks Profitability in India: An Empirical Study

Downloaded From IP - 103.37.201.68 on dated 12-Oct-2015

www.IndianJournals.com

Members Copy, Not for Commercial Sale

sector is better able to withstand negative shocks and

contribute to the stability of the financial system. The
importance of bank profitability at both the micro and
macro levels has made researchers, academics, bank
managements and bank regulatory authorities to
develop considerable interest on the factors that
determine bank profitability (Athanasoglou et al.,
2005).
Banks are urged to generate sufficient revenue to meet
the rising cost of funds. Profitability is a key result
area where performance and results directly and
virtually affect the survival. Therefore, this study
analyses the profitability performance of public sector
commercial banks in India. The rest of this paper is,
organised in following the order: Section 2 presents
review of literature, Section 3, presents data,
methodology and results, whereas Section 4 concludes
the study.

LITERATURE REVIEW
In the literature, bank profitability is usually expressed
as a function of internal and external determinants.
The internal determinants originate from bank accounts
(balance sheets and/or profit and loss accounts) and
therefore could be termed micro or bank-specific
determinants of profitability. The external determinants
are variables that are not related to bank management
but reflect the economic and legal environment that
affects the operation and performance of financial
institutions.
Bourke (1989) found a significant positive relationship
between capital adequacy and profitability indicating
that banks with higher capital ratio are more profitable
than banks with less capital ratio. Similarly, Berger
(1995) and Angbazo (1997) found that US banks with
relatively high capital adequacy were more profitable
than other banks with lower capital ratio. Also,
Molyneux and Thornton (1992), in a study of 18
European countries for the period 1986-1989, report a
positive relationship for state-owned banks. In
developing countries, Vong and Chan (2006)
investigated the determinants of bank performance of
Macao Banking industry for a 15-year period using
Siddhant

small sample of banks and found a positive relationship.

Also, bank size, as a determinant, reports mixed
conclusions. For example, Sinkey (1992) and Boyd et
al. (1998) both reports an inverse relationship between
large banks and profitability but, interestingly, Sinkey
(1992) found a positive relationship for smaller banks.
For developed countries, Naceur (2003) reports a
negative relationship between bank size and profitability
in Tunisia. Cost to income ratio measures banks
expense management. Revell (1997) introduces the
issue of the relationship between bank profitability and
inflation. He notes that the effect of inflation on bank
profitability depends on whether banks wages and
other operating expenses increase at a faster rate than
inflation. The question is how mature an economy is
so that future inflation can be accurately forecasted
and thus banks can accordingly manage their operating
costs. In this vein, Perry (1992) states that the extent
to which inflation affects bank profitability depends on
whether inflation expectations are fully anticipated. An
inflation rate fully anticipated by the banks
management implies that banks can appropriately adjust
interest rates in order to increase their revenues faster
than their costs and thus acquire higher economic
profits. Buch (1997) asserts that foreign-owned banks
use modern technology from and rely on the human
capital of their parent banks, so that they would be
expected to perform better than government-owned
or domestic private banks in transitional economies.
On similar lines, private banks would be expected to
perform better than government-owned banks. Kraft
and Tirtiroglu (1998) used stochastic frontier analysis
(SFA) to examine the bank efficiency in Croatia in the
mid-1990s and found that the newly organised private
banks were more efficient relative to older state
institutions. Jemric and Vujcic (2002) used data
envelopment analysis (DEA) to analyze bank
efficiency in Croatia in the late 1990s and found that
foreign banks and new banks are more efficient. Nikiel
and Opiela (2002) used distribution-free efficiency
estimation for Polish banks in the late 1990s and found
that foreign banks servicing foreigners and business
customers are more cost-efficient but less profit
efficient than other banks in Poland. Isik and Hassan
161

Downloaded From IP - 103.37.201.68 on dated 12-Oct-2015

www.IndianJournals.com

Members Copy, Not for Commercial Sale

Gurmeet Singh

(2003) examined the Turkish commercial banks during

the deregulation period and found that the Turkish
private banks began to close their gap with those public
banks in the new environment. Claessens et al. (2001)
investigated performance differences between
domestic and foreign banks in eighty countries, both
developed and developing, from the late 1990s to mid
2000s and found that foreign bank entry was generally
followed by a reduction in both profitability and the
overhead expenses of domestic banks, suggesting that
foreign participation improves the efficiency of
domestic banking. Haslem et al. (1999) used data
envelopment analysis to analyze the efficiency of U.S.
banks operating internationally. It was found that
management should focus on overall efficiency, but
with particular attention to inputs, especially cash and
real capital, and to foreign loans among the outputs.
There have been several studies analysing bank
profitability and efficiency in India. Sarkar et al. (1998)
compared public, private and foreign banks in India to
find the effect of ownership type on different efficiency
measures. Ram Mohan (2002; 2003) also used
financial measures for comparing operational
performance of different categories of banks over a
period of time. Bhattacharyya et al. (1997) used DEA
to measure the productive efficiency of Indian
commercial banks in the late 1980s to early 1990s and
to study the impact of policy of liberalising measures
taken in the 1980s on the performance of various
categories of banks. They found that the Indian public
sector banks were the best performing banks, as the
banking sector was overwhelmingly dominated by the
Indian public sector banks, while the new private sector
banks were yet to emerge fully in the Indian banking
scenario. Kumbhakar and Sarkar (2003) found
evidence on Indian banks that while private sector banks
have improved their performance mainly due to the
freedom to expand output, public sector banks have
not responded well to the deregulation measures. Ram
Mohan and Ray (2004) compared the revenue
maximising efficiency of public, private and foreign
banks in India, using physical quantities of inputs and
outputs in the 1990s, using deposits and operating costs
162

as inputs, and loans, investments and other income as

outputs. They found that public sector banks were
significantly better than private sector banks on revenue
maximisation efficiency, but between public sector
banks and foreign banks the difference in efficiency
was not significant. Shanmugam and Das (2004) studied
banking efficiency using stochastic frontier production
function model during the reform period, 1992-1999.
They found that deposits are dominant in producing all
outputs and the technical efficiency of raising interest
margin is varied across the banks. In particular, they
found that the reform measures that had been
introduced since 1992 have not helped the banks in
raising their interest margin. Also, in general, they found
that private/foreign banks performed better than public
banks. Das et al. (2004) analysed the efficiency of
Indian banks using data envelopment analysis using
four input measures and found that, despite liberalisation
measures aimed at strengthening and improving the
operational efficiency of the financial system, Indian
banks were still not much differentiated in terms of
input- or output-oriented technical efficiency and cost
efficiency.
Sanjeev (2006) studied efficiency of private, public,
and foreign banks operating in India during the period
1997-2001 using data envelopment analysis. He also
studied if any relationship can be established between
the efficiency and non-performing assets in the banks.
He found that the there is an increase in the efficiency
in the post-reform period, and that non-performing
assets and efficiency are negatively related. Kumar
and Gulati (2007) studied the technical efficiency of
public sector banks in India using two data envelopment
analysis models. They found that foreign banks are
found to be more cost-efficient but less profit-efficient
relative to domestically owned private banks and stateowned banks. The banks affiliated with SBI group were
found to outperform the nationalised banks in terms of
operating efficiency.
Based on the above discussion, the present study tries
to investigate the long run equilibrium relationship
between the return on assets (ROA) and nine variables,
by considering the following model:
Volume 15, Issue 2, April-June, 2015

Determinants of Public Sector Banks Profitability in India: An Empirical Study

Xt = (ROAt, NPAt, DLRt, SATAt, BTAt, BIIt, PTAt,

NIMt, NTRt, PERt )
where, ROA is the return on assets ratio, NPA is net
NPA to net advances ratio, DLR is deposits to total
liabilities ratio, SATA is secured advances to total
advances ratio, BTA is burden to total assets ratio, BII
is burden to interest income ratio, PTA is operating
profits to total assets ratio, NIM is net interest income
to total assets ratio, NTR is non-interest income to total
assets ratio, PER is profit per employee ratio and X is
a 101 vector of variables.

Downloaded From IP - 103.37.201.68 on dated 12-Oct-2015

www.IndianJournals.com

Members Copy, Not for Commercial Sale

DATA AND METHODOLOGY

The aim of this paper is to investigate the determinants
of public sector banks profitability in India. To
accomplish the research objective yearly data ranging
from 2005 to 2014 are obtained which comprises 260
data points for the analysis. The choice of study period
is based on the availability of data series. Descriptions
of variables and data sources are presented in
Table 1.
Table 1: Description of variables
Acronyms Construction of variable
ROA
Return on assets
NPA
Ratio of net NPA to
net advances
DLR
Ratio of deposits to
total liabilities
SATA
Ratio of secured advances
to total advances
BTA
Ratio of burden to total assets
BII
Ratio of burden to interest
income
PTA
Ratio of operating profits
to total assets
NIM
Ratio of net interest income
to total assets (Net Interest
Margin)
NTR
Ratio of non-interest income
to total assets
PER
Profit per employee
(in Rupees Million)
Siddhant

Data source
RBI Website
RBI Website
RBI Website
RBI Website
RBI Website
RBI Website
RBI Website
RBI Website

RBI Website
RBI Website

The present study employs the time series data analysis

technique to study the relationship between the ROA
and NPA, DLR, SATA, BTA, BII, PTA, NIM, NTR
and PER. In a time series analysis, the results might
provide a spurious if the data series are non-stationary.
Thus, the data series must obey the time series
properties, that is, the time series data should be
stationary, meaning that, the mean and variance should
be constant over time and the value of covariance
between two time periods depends only on the distance
between the two time period and not the actual time at
which the covariance is computed. The most popular
and widely used test for stationary is the unit root test.
The presence of unit root indicates that the data series
is non-stationary. The standard procedures of unit root
test namely the Augmented Dickey Fuller (ADF)
(1979; 1981) is performed to check the stationary nature
of the series. Assuming that the series follows an AR
(p) process the ADF test makes a par ametric
correction and controls for the higher order correlation
by adding the lagged difference terms of the dependent
variable to the right hand side of the regression equation.
In the ADF test null hypothesis is that data set being
tested has unit root. This provides a robustness check
for stationary. The unit root tests also provide the order
of integration of the time series variables. In a
multivariate context if the variable under consideration
are found to be I (1) (i.e. they are non-stationary at
level but stationary at first difference), but the linear
combination of the integrated variables is I (0), then
the variables are said to be co-integrated (Enders,
2004). The ADF (1979; 1981) is performed to check
the stationary nature of the series. The complete model
with deterministic terms such as intercepts and trends
is shown in equation (1).
(1)
Where is a constant, thecoefficientonatimetrend
and pthelagorderoftheautoregressiveprocess.Lag
length for VAR system is, selected based on minimum
sequential modified LR test statistic. The vector
autoregression (VAR) is commonly used for
forecasting systems of interrelated time series and for
analysing the dynamic impact of random disturbances
163

Gurmeet Singh

on the system of variables. The VAR approach

sidesteps the need for structural modeling by treating
every endogenous variable in the system as a function
of the lagged values of all of the endogenous variables
in the system. The mathematical representation of a
VAR is as follows:
(2)

Downloaded From IP - 103.37.201.68 on dated 12-Oct-2015

www.IndianJournals.com

Members Copy, Not for Commercial Sale

Where yt is a k vector of endogenous variables, xt is a

d vector of exogenous variables, A1, ...., Ap and B are
matrices of coefficients to be estimated, and t is a
vector of innovations that may be contemporaneously
correlated but are uncorrelated with their own lagged
values and uncorrelated with all of the right-hand side
variables.
Leg Length Criteria computes various criteria to select
the lag order of an unrestricted VAR (Ltkepohl, 1991).
The sequential modified likelihood ratio (LR) test is
carried out as follows. Starting from the maximum lag,
test the hypothesis that the coefficients on lag are
jointly zero using the 2 statistics:
(3)
Where, m is the number of parameters per equation
under the alternative, note that we employ Sims (1980)
small sample modification which uses (T m) rather
than T. We compare the modified LR statistics to the
5% critical values starting from the maximum lag, and
decreasing the lag one at a time until we first get a
rejection.
With the non-stationary series, co-integration analysis
has been used to examine whether there is any long
run relationship exists. However, a necessary condition
for the use of co-integration technique is that the
variable under consideration must be integrated in the
same order and the linear combinations of the integrated
variables are free from unit root. According to Engel
and Granger (1987), if the variables are found to be
co-integrated, they would not drift apart over time and
the long run combination amongst the non-stationary
variables can be established. To conduct the co164

integration test, the Engel and Granger (1987) or the

Johansen and Juselius (1990) or the Johansen (1991)
approach can be used. The Engel-Granger two step
approaches can only deal with one linear combination
of variables that is stationary. In a multivariate practice,
however, more than one stable linear combination may
exist. The Johansens co-integration method is regarded
as full information maximum likelihood method that
allows for testing co-integration in a whole system of
equations.
The Johansen methods of co-integration can be written
as the following vector autoregressive framework of
order p.
(4)
where, Xt is an n 1 vector of non stationary I(1)
variables, A0 is an n 1 vector of constants, p is the
maximum lag length, Bj is an n n matrix of coefficient
and e is a n 1 vector of white noise terms. The number
of characteristic roots can be tested by considering
the following trace statistic and the maximum
eigenvalue test.
(5)

(6)
where, r is the number of co-integrating vectors under
the null hypothesis, T is the number of usable
observations and

is the estimated value for the jth

ordered characteristic roots or the eigenvalue from the

matrix.
A significantly non-zero eigenvalue indicates a
significant co-integrating vector. The trace statistics is
a joint test where the null hypothesis is that the number
of co-integration vectors is less than or equal to r
against an unspecified general alternative that there
are more than r. Whereas, the maximum eigenvalue
statistics test the null hypothesis that the number of
Volume 15, Issue 2, April-June, 2015

Determinants of Public Sector Banks Profitability in India: An Empirical Study

co-integrating vectors is less than or equal to r against

the alternative of r + 1 (Enders, 2004; Brooks, 2002)
A vector error correction (VEC) model is a restricted
VAR designed for use with non-stationary series that
are known to be cointegrated. The VEC Model has
cointegration relations built into the specification so that
it restricts the behaviour of the endogenous variables
to converge to their cointegrating relationships while
allowing for short-run adjustment dynamics. The
cointegration term is known as the error correction
term since the deviation from long-run equilibrium is
corrected gradually through a series of partial shortrun adjustments.

Downloaded From IP - 103.37.201.68 on dated 12-Oct-2015

www.IndianJournals.com

Members Copy, Not for Commercial Sale

The corresponding VEC model is:

Siddhant

(9)
(10)
For all possible pairs of (x, y) series in the group, the
reported F-statistics are the Wald statistics for the joint
hypothesis:

(7)

For each equation, the null hypothesis is that x does

not Granger-cause y in the first regression and y does
not Granger-cause x in the second equation.

(8)

EMPIRICAL ANALYSIS

In this model, the only right-hand side variable is the

error correction term. In long run equilibrium, this term
is zero. However, if y1 and y2 deviate from the long run
equilibrium, the error correction term will be nonzero
and each variable adjusts to partially restore the
equilibrium relation. The coefficient i measures the
speed of adjustment of the ith endogenous variable
towards the equilibrium.
Table 2: Descriptive statistics of variables

ROA
NPA
DLR
Mean
0.869
1.393
84.004
Median
0.880
1.185
85.205
Maximum
2.010
8.110
90.640
Minimum
-0.990
0.150
18.560
Std. dev.
0.360
1.014
7.226
Skewness
-0.669
2.351
-5.692
Kurtosis
6.027
13.594
44.630
Jarque-Bera 118.614 1455.375 20178.220
Probability
0.000
0.000
0.000
Observations 260
260
260
Source: Authors estimation

Further to examine dynamic relationship between

variables, bi-variate Granger Causality test (Engel and
Granger, 1987) is applied. The bi-variate regressions
of Granger Causality Test are as follows:

SATA
83.481
83.695
97.960
64.940
6.292
-0.224
2.708
3.103
0.212
260

The descriptive statistics for all the variables under

study, namely, ROA, NPA, DLR, SATA, BTR, BII,
PTA, NIM, NTR and PER are presented in Table 2.
The value of skewness and kurtosis indicate the lack
of symmetric in the distribution. Generally, if the value
of skewness and kurtosis are 0 and 3 respectively, the
observed distribution is said to be normally distributed.
Furthermore, if the skewness coefficient is in excess
of unity it is considered fairly extreme and the low
BTA
0.720
0.730
2.090
-0.530
0.391
0.206
4.521
26.890
0.000
260

BII
8.994
8.705
27.930
-7.820
5.161
0.265
4.871
40.959
0.000
260

PTA
1.911
1.900
3.340
0.440
0.464
0.197
3.600
5.578
0.061
260

NIM
2.631
2.640
3.980
0.230
0.598
-0.668
4.537
44.937
0.000
260

NTR
1.030
0.980
2.520
0.160
0.321
1.153
6.043
157.978
0.000
260

PER
0.462
0.420
1.320
-0.740
0.273
0.301
4.404
25.295
0.000
260

165

Downloaded From IP - 103.37.201.68 on dated 12-Oct-2015

www.IndianJournals.com

Members Copy, Not for Commercial Sale

Gurmeet Singh

Table 3: Result of augmented Dickey- Fuller unit root test

Variable
Trend
Trend & Intercept
None
t-Statistic Prob.* t-Statistic Prob.* t-Statistic Prob.*
D(ROA) Augmented Dickey- Fuller test statistic
-10.1778
0.0000
-10.3471
0.0000
-10.0967
0.0000
Test critical values:
1% level
-3.4568

-3.9958

-2.5744

5% level
-2.8731

-3.4282

-1.9421

10% level
-2.5730

-3.1375

-1.6158

D(NPA) Augmented Dickey- Fuller test statistic

-17.0159
0.0000
-17.0581
0.0000
-17.0454
0.0000
Test critical values:
1% level
-3.4559

-3.9945

-2.5741

5% level
-2.8727

-3.4275

-1.9421

10% level
-2.5728

-3.1371

-1.6159

D(DLR) Augmented Dickey- Fuller test statistic

-10.5781
0.0000
-10.5805
0.0000
-10.5863
0.0000
Test critical values:
1% level
-3.4570

-3.9960

-2.5744

5% level
-2.8731

-3.4283

-1.9421

10% level
-2.5730

-3.1375

-1.6158

D(SATA) Augmented Dickey- Fuller test statistic

-12.7097
0.0000
-12.7995
0.0000
-12.7343
0.0000
Test critical values:
1% level
-3.4566

-3.9955

-2.5743

5% level
-2.8730

-3.4280

-1.9421

10% level
-2.5730

-3.1374

-1.6158

D(BTA) Augmented Dickey- Fuller test statistic

-11.8085
0.0000
-11.8014
0.0000
-11.8322
0.0000
Test critical values:
1% level
-3.4561

-3.9947

-2.5741

5% level
-2.8728

-3.4277

-1.9421

10% level
-2.5728

-3.1372

-1.6159

D(BII)
Augmented Dickey- Fuller test statistic
-10.7684
0.0000
-10.7534
0.0000
-10.7865
0.0000
Test critical values:
1% level
-3.4562

-3.9949

-2.5742

5% level
-2.8728

-3.4278

-1.9421

10% level
-2.5729

-3.1372

-1.6159

D(PTA) Augmented Dickey- Fuller test statistic

-11.9629
0.0000
-11.9382
0.0000
-11.9770
0.0000
Test critical values:
1% level
-3.4561

-3.9947

-2.5741

5% level
-2.8728

-3.4277

-1.9421

10% level
-2.5728

-3.1372

-1.6159

D(NIM) Augmented Dickey- Fuller test statistic

-12.3499
0.0000
-12.3485
0.0000
-12.3654
0.0000
Test critical values:
1% level
-3.4561

-3.9947

-2.5741

5% level
-2.8728

-3.4277

-1.9421

10% level
-2.5728

-3.1372

-1.6159

D(NTR) Augmented Dickey- Fuller test statistic

-9.7223
0.0000
-9.7078
0.0000
-9.6939
0.0000
Test critical values:
1% level
-3.4567

-3.9956

-2.5744

5% level
-2.8730

-3.4281

-1.9421

10% level
-2.5730

-3.1374

-1.6158

D(PER)
Augmented Dickey- Fuller test statistic
-8.5416
0.0000
-8.5573
0.0000
-8.5492
0.0000
Test critical values:
1% level
-3.4566

-3.9955

-2.5743

5% level
-2.8730

-3.4280

-1.9421

10% level
-2.5730

-3.1374

-1.6158

*MacKinnon (1996) one-sided p-values.

Source: Authors estimation
166

Volume 15, Issue 2, April-June, 2015

Determinants of Public Sector Banks Profitability in India: An Empirical Study

(high) kurtosis value indicates extreme platykurtic

(extreme leptokurtic). From the table it is observed
that the frequency distributions of underlying variables
are not normal. The significant coefficient of JarqueBera statistics also indicates that the frequency
distributions of considered series are not normal.

Downloaded From IP - 103.37.201.68 on dated 12-Oct-2015

www.IndianJournals.com

Members Copy, Not for Commercial Sale

To check the stationarity of the underlying data series,

we follow the standard procedure of unit root testing
by employing the ADF test. The results are presented
in Table 3. On the basis of the ADF test, all the series
are found to be non-stationary at level with intercept.
However, after taking the first difference these series
are found to be stationary at 1, 5 and 10% significance
level. Thus the stationary test indicates that all series
are individually integrated of the order I (1).
Table 4 shows Karl Pearsons Correlation Co-efficient
Matrix. The correlation matrix highlighted that there is
negative correlation between ROA and NPA, SATA,
BTA and BII. While, there is a positive correlation
between ROA and DLR, PTA, NIM, NTR and PER.
The presence and the number of co-integrating
relationships among the underlying variables are tested
through the Johansen procedure, that is, Johansen and
Juselius (1990) and Johansen (1991). Specifically, trace
statistic and the maximum eigenvalue are used to test
for the number of co-integrating vectors. The result of
VAR leg order selection criteria are presented the in

the Table 5. Leg order selected for the study is based

on LR, FPE and AIC criterion. The results of both
trace statics and the maximum eigenvalue test statistics
are presented in Table 6. The trace statistic indicates
four co-integrating equations and the maximum
eigenvalue statistics identify three co-integrating
equations. The results show that long-run equilibrium
relationship exists between the ROA and NPA, DLR,
SATA, BTA, BII, PTA, NIM, NTR and PER.
Assuming one co-integrating vector, the short run and
long run interaction of the underlying variables the
VECM has been estimated based on the Johansen cointegration methodology. The results are presented in
Table 7. The results show that a long-run equilibrium
relationship exists between the return on equity (ROE)
and NPA, DLR, SATA, BTA, BII, PTA, NIM, NTR
and PER. The estimated co-integrating coefficients for
the ROE are based on the first normalised eigenvector
are as follows. These values represent long term
elasticity measures. Thus the co-integration relationship
can be re-expressed as:
ROA = -4.6937 + 0.2022*NPA + 0.0582*DLR + (0.0083)*SATA + (-0.064)*BII + (-12.6943)*PTA +
12.4769*NIM + (-0.1482)*NTR + (-0.4823)*PER
The t-statistics are given in [] brackets while the error
term are given in () brackets. The coefficients of
secured advances to total advances ratio, burden to

Table 4: Result of Karl Pearsons Correlation Co-efficient Matrix

ROA
NPA
DLR
SATA
BTA
BII
ROA
1
-0.7304
0.0193
-0.1527
-0.2006
-0.1706
NPA
-0.7304
1
0.0475
0.3349
0.2589
0.2074
DLR
0.0193
0.0475
1
-0.1720
0.4068
0.4305
SATA -0.1527
0.3349
-0.1720
1
-0.0007
-0.0617
BTA
-0.2006
0.2589
0.4068
-0.0007
1
0.9843
BII
-0.1706
0.2074
0.4305
-0.0617
0.9843
1
PTA
0.6239
-0.2418
0.1891
0.0846
-0.0307
-0.0060
NIM
0.3530
-0.0181
0.4129
0.0659
0.6301
0.6391
NTR
0.3687
-0.1593
-0.1297
-0.0022
-0.2641
-0.2058
PER
0.4730
-0.3562
-0.2962
0.0151
-0.5114
-0.5400
Source: Authors estimation
Siddhant

PTA
0.6239
-0.2418
0.1891
0.0846
-0.0307
-0.0060
1
0.7568
0.5986
0.0985

NIM
0.3530
-0.0181
0.4129
0.0659
0.6301
0.6391
0.7568
1
0.2921
-0.2581

NTR
0.3687
-0.1593
-0.1297
-0.0022
-0.2641
-0.2058
0.5986
0.2921
1
-0.1385

PER
0.4730
-0.3562
-0.2962
0.0151
-0.5114
-0.5400
0.0985
-0.2581
-0.1385
1

167

Gurmeet Singh

Table 5: VAR Lag Order Selection Criteria

Lag
LogL
LR
0
-1351.755
NA
1
-1090.708
499.3036
2
-909.9569
331.3777
3
-783.7753
221.3185

FPE
2.34E-08
6.52E-09
3.44E-09
2.82E-09

AIC
10.80758
9.529431
8.888546
8.680756

SC
10.94764*
11.07005
11.82974
13.02251

HQ
10.86394
10.14935
10.07202*
10.42779

Downloaded From IP - 103.37.201.68 on dated 12-Oct-2015

www.IndianJournals.com

Members Copy, Not for Commercial Sale

4
-648.4575
226.6036
2.17E-09
8.400456
14.14278
10.71105
5
-474.6469
277.2693
1.24E-09
7.814658
14.95755
10.68881
6
-352.5011
185.1575
1.08E-09
7.638898
16.18236
11.07661
7
-199.1735
220.2564
7.53E-10
7.215662
17.15969
11.21693
8
-70.06764
175.2150*
6.48e-10*
6.984664*
18.32926
11.54949
*Indicates lag order selected by the criterion
LR: sequential modified LR test statistic (each test at 5% level); FPE: final prediction error; AIC: Akaike information criterion;
SC: Schwarz information criterion; HQ: Hannan- Quinn information criterion
Source: Authors estimation
Table 6: Result of Johansens Co-integration Test
Hypothesised Eigenvalue
Trace
0.05 Critical
Prob.**
Max-Eigen 0.05 Critical
No. of CE(s)
statistic
value
statistic
value
None*
0.253372
329.5611
239.2354
0.0000
73.33913
64.50472
At most 1*
0.214587
256.222
197.3709
0.0000
60.62785
58.43354
At most 2*
0.201441
195.5942
159.5297
0.0001
56.46147
52.36261
At most 3*
0.166521
139.1327
125.6154
0.0058
45.71875
46.23142
At most 4
0.134658
93.41394
95.75366
0.0717
36.30238
40.07757
At most 5
0.086801
57.11155
69.81889
0.3346
22.79125
33.87687
At most 6
0.070863
34.32031
47.85613
0.4844
18.44817
27.58434
At most 7
0.036222
15.87214
29.79707
0.7212
9.26034
21.13162
At most 8
0.018712
6.611796
15.49471
0.6232
4.741291
14.26460
At most 9
0.007425
1.870505
3.841466
0.1714
1.870505
3.841466
Trace test indicates 4 cointegrating eqn(s) at the 0.05 level
Max-eigenvalue test indicates 3 cointegrating eqn(s) at the 0.05 level
*denotes rejection of the hypothesis at the 0.05 level; **MacKinnon- Haug- Michelis (1999) p-values
Source: Authors estimation

total assets ratio, burden to interest income ratio,

operating profits to total assets ratio, non-interest
income to total assets ratio and profit per employee
ratio are negative and statistically significant, while the
coefficient of NPA to net advances ratio, deposits to
total liabilities ratio and net interest margin are positive
and statistically insignificant. The intercept term is
negative. The results revels that the relationship
168

Prob.**
0.0058
0.0299
0.0180
0.0567
0.1253
0.5464
0.4583
0.8109
0.7741
0.1714

between ROA and SATA, BTA, BII, PTA, NTR and

PER is positive while the relationship between the ROA
and NPA, DLR and NIM is negative. The sign of the
error correction coefficient in determination of ROA
is negative (-0.0218) and the t-value is (-0.0881). This
indicates that return on assets do respond significantly
to re-establish the equilibrium relationship once
deviation occurs.
Volume 15, Issue 2, April-June, 2015

Determinants of Public Sector Banks Profitability in India: An Empirical Study

Table 7: Results of Vector Error Correction Model
Panel A: Normalized Co-integration Coefficients
ROA(-1)
1.0000

D(NPA) D(DLR) D(SATA) D(BTA)

0.2022

0.0582

-0.0083

-11.4864

D(BII)

D(PTA) D(NIM) D(NTR) D(PER) Constant

-0.0641

-12.6943 12.4769

-0.1482

-0.4823

-4.6937

(-0.0326) (-0.0092) (-0.0077) (-6.9743) (-0.0359) (-6.9941) (-6.9776) (-0.2061) (-0.1660)

[ 6.1987] [ 6.3380] [-1.0801] [-1.6470] [-1.7838] [-1.8150] [ 1.7881] [-0.7189] [-2.9065]

Panel B: coefficient of Error Correction term

Error Correction:

D(ROA) D(NPA) D(DLR) D(SATA) D(BTA)

CointEq1

-0.0218

-1.0425

-21.7032

-4.5852

-0.8585

D(BII)
-11.1149

D(PTA) D(NIM) D(NTR)

D(PER)

-0.0265

0.3353

-0.8945

0.5373

(-0.2480) (-0.6097) (-4.5043) (-3.9420) (-0.2369) (-3.0882) (-0.3572) (-0.3939) (-0.2206) (-0.1656)

[-0.0881] [-1.7099] [-4.8183] [-1.1632] [-3.6243] [-3.5991] [-0.0741] [-2.2707] [ 2.4358] [ 2.0252]
F-statistic

7.6434

8.3259

12.2082

7.8717

9.9221

9.4322

5.9375

7.3896

4.3344

8.2361

Downloaded From IP - 103.37.201.68 on dated 12-Oct-2015

www.IndianJournals.com

Members Copy, Not for Commercial Sale

Standard errors in ( ) & t-statistics in [ ]

Source: Authors estimation
Table 8: Result of Granger causality Test
Null hypothesis:

Obs

F-statistic

Prob.

Decision

NPA does not Granger Cause ROA

252

3.70726

0.0004

Reject

ROA does not Granger Cause NPA

1.69619

0.1001

Accept

DLR does not Granger Cause ROA

252

7.23047

1.00E-08

Reject

ROA does not Granger Cause DLR

7.00344

3.00E-08

Reject

SATA does not Granger Cause ROA

252

4.64233

3.00E-05

Reject

ROA does not Granger Cause SATA

14.2196

7.00E-17

Reject

BTA does not Granger Cause ROA

252

4.09936

0.0001

Reject

ROA does not Granger Cause BTA

5.35315

3.00E-06

Reject

BII does not Granger Cause ROA

252

4.41863

5.00E-05

Reject

ROA does not Granger Cause BII

5.58611

2.00E-06

Reject

PTA does not Granger Cause ROA

252

3.44355

0.0009

Reject

ROA does not Granger Cause PTA

3.72164

0.0004

Reject

NIM does not Granger Cause ROA

252

3.62133

0.0005

Reject

ROA does not Granger Cause NIM

5.65998

0.0000

Reject

NTR does not Granger Cause ROA

252

5.41871

0.0000

Reject

ROA does not Granger Cause NTR

4.27098

0.0001

Reject

PER does not Granger Cause ROA

252

2.84633

0.0049

Reject

ROA does not Granger Cause PER

2.21243

0.0273

Reject

Source: Authors estimation

Siddhant

169

Gurmeet Singh

The co-integration results indicate that causality exists

between the co-integrated variables but it fails to show
us the direction of the causal relationship. The pairwise Granger Causality test (1987) is performed
between all possible pairs of variables to determine
the direction of causality. The rejected hypotheses are
reported in Table 8. The results shows that there is
bidirectional causality exists between all the variables
tested (i.e. DLR, SATA, BTA, BII, PTA, NIM, NTR
and PER) and ROA except NPA. NPA granger causes
ROA but not the other way round.

Downloaded From IP - 103.37.201.68 on dated 12-Oct-2015

www.IndianJournals.com

Members Copy, Not for Commercial Sale

CONCLUSION
This study examined the determinants of public sector
banks profitability in India by finding inter-linkage
between the return on assets and net NPA to net
advances, deposits to total liabilities, secured advances
to total advances, burden to total assets, burden to
interest income, operating profits to total assets, net
interest margin, non-interest income to total assets and
profit per employee by using Johansens co-integration
test. The analysis used yearly data over the period 2005
to 2014 which is obtained from RBI website.
To conclude, the Augmented Dickey Fuller test
suggests that all the series are found to be nonstationary at level with intercept. However, after taking
the first difference these series are found to be
stationary at 1, 5 and 10% level of significance. The

result of Karl Pearsons correlation matrix suggests

that there is negative correlation between ROA and
NPA, SATA, BTA and BII. While, there is a positive
correlation between ROA and DLR, PTA, NIM, NTR
and PER. The analysis revealed that the return on
assets in study formed significant long-run equilibrium
relationship with all the nine variables tested. The
Johansens co-integration test suggests that the return
on assets of public sector banks of India have cointegrated with the nine variables under the study in
the long-run. The findings from Granger causality based
on the VECM indicate there is bidirectional causality
exists between all the variables tested and return on
assets except non performing assets. Non performing
assets granger causes return on assets but not the other
way round. It is observed from the findings that secured
advances to total advances ratio, burden to total assets
ratio, burden to interest income ratio, operating profits
to total assets ratio, non-interest income to total assets
ratio and profit per employee ratio causes ratio of return
on asset in long run and short run. While, NPA to net
advances ratio, deposits to total liabilities ratio and net
interest margin causes return on assets in short run
only. However, the limitations of the study should not
be over looked. The present study is limited to only
nine selected variables. Inclusion of more variables
with a longer time period may improve the results. A
logical extension of the study can be done by including
more variables and analyze the determinants of
profitability.

REFERENCES
Nikiel E and Opiela T, 2002. Customer type and bank efficiency in Poland: implications for emerging banking market.
Contemporary Economic Policy, Vol. 20.
Perry P, 1992. Do banks gain or lose from inflation. Journal of Retail Banking, Vol. 14, No.2, pp. 2530.
Ram Mohan TT, 2002. Deregulation and performance of public sector banks. Economic and Political Weekly, Vol. 37, No. 5.
Ram Mohan TT, 2003. Long-run performance of public and private sector bank stocks. Economic and Political Weekly, Vol.
38, No. 8.
Ram Mohan TT and Ray SC, 2004. Comparing performance of public and private sector banks: a revenue maximisation
efficiency approach. Economic and Political Weekly, Vol. 39, No. 12.
Revell J, 1979. Inflation and financial institutions. Financial Times.
Sanjeev GM, 2006. Data envelopment analysis (DEA) for measuring technical efficiency of banks. VISION-The Journal of
Business Perspective, Vol. 10, No. 1.
170

Volume 15, Issue 2, April-June, 2015

Determinants of Public Sector Banks Profitability in India: An Empirical Study

Sarkar J, Sarkar S and Bhaumik SK, 1998. Does ownership always matter? Evidence from the Indian banking industry.
Journal of Comparative Economics, Vol. 26.
Shanmugam KR and Das A, 2004. Efficiency of Indian commercial banks during the reform period. Applied Financial
Economics, Vol. 14.
Sims C, 1980. Macroeconomics and reality. Econometrica, Vol. 48, pp. 148.
Sinkey JJ, 1992. Commercial Bank Financial Management in the Financial Services Industry. New York: Macmillan
Publishing Company.

Downloaded From IP - 103.37.201.68 on dated 12-Oct-2015

www.IndianJournals.com

Members Copy, Not for Commercial Sale

Vong IP and Chan HS, 2006. Determinants of bank profitability in Macao. The 30th Anniversary of Journal of Banking and
Finance Conference. Beijing.

Siddhant

171