CHAPTER - IV

RESEARCH METHODOLOGY

77
 Chapter- IV

Research Methodology

4.1 Introduction

This chapter depicts the way of conducting the research for answering the
question of whether CSR expenditure affects the financial performance of the
Central Public Sector Enterprises in India or not. This chapter includes research
questions, a detailed sampling design, source of data, the period covered under
the study, and research methods of the study.

4.2 Research Approach

There are two types of research approaches- one is the deductive research
approach and another is the inductive research approach. The deductive
approach can be done through quantitative analysis and the inductive approach
can be done through a qualitative approach (Soiferman, 2010)1. In a deductive
approach, a researcher needs to start from a generalised or established theory,
rules, or laws, and then the researcher formulates a hypothesis and starts
collecting data. Thereafter, the researcher needs to test the hypothesis by
applying various statistical tools and techniques and lastly, the researcher comes
to a conclusion regarding accepting or rejecting the hypothesis that he/she has
established. The deductive approach is also known as the top to bottom
approach. On the other hand, in the inductive approach, a researcher collects
various facts, figures, and data and then formulates a tentative hypothesis and
after analysing the qualitative data, he/she develops a theory (Thomas, 2006)2.
Here we have applied the deductive approach of research as we have collected
data from CPSEs in India and then have set a hypothesis, i.e., does CSR affect
the financial performance of CPSEs in India? Then we have tried to come to a
definite conclusion after analysing the quantitative data with various statistical
tools and techniques.

78
4.3 Research Strategy

The research strategy is a systematic step-by-step procedure as to how research

is being conducted by a researcher to get the answer to the research question.
There are mainly two types of researches- qualitative research and quantitative
research. Qualitative research helps to understand people‟s beliefs, experiences,
attitudes, behaviours, opinions, and various social issues resembling social
norms, socioeconomic status, gender roles, ethnicity, religion, etc (Pathak et al.,
2013)3. The data collected under qualitative research is the non-numeric type and
the researcher collects data through observation, interviews, questionnaires,
focus groups, various documents, etc. On the other hand, in quantitative research
strategy, the data are numeric type and the data are collected from various data
sources that are available on the internet, various published journals, books, etc.
Quantitative research strategy focuses on collecting data and analysing data to
understand the relationship among various variables empirically with the help of
various tools and techniques. Here, in this research, we have collected
quantitative data and employed a quantitative research strategy to conduct the
research empirically.

4.4 Research Design

Research design is a conceptual structure of a researcher within which research

would be conducted (Kothari, 2004, p.14)4. It is an overall systematic plan to have
the answer to the research problem (Boru, 2018)5. It is basically the decisions of a
researcher as to what, where, when, how, how much, and by which means to
collect and analyse the data which is being studied in research (Kothari, 2004,
p.31)6. There are mainly two types of research designs. These are (i)
experimental research and (ii) non-experimental design.

4.4.1 Experimental Research

Experimental research is a study that is based on scientific research design. An

experimental research study tries to find out an association between two or more
variables where one variable is called the dependent variable and another is
called the independent variable. In this research, one needs to collect quantitative
data, and then the researcher needs to develop a hypothesis. After collecting data
and developing the research hypothesis, the researcher needs to employ various

79
tools and techniques to establish the relationship between the variables by
accepting or rejecting the null hypothesis. In this type of research, the researcher
can calculate, measure, and compare the data as per the requirement of the
researcher because this research can be done in a controlled environment or in a
controlled manner. Applying experimental research is appropriate when the
researcher has a definite research question and a hypothesis to establish the
causal relationship between two variables. The experimental research design is
also called the deductive research method. In this study, we have done
experimental research.

4.4.2 Non-experimental Research

Unlike experimental research, non-experimental research cannot manipulate the

independent variable and it cannot randomly assign conditions or order of
conditions to participants. Non-experimental research is based on observations to
correlate the variables and describe the findings (Bonds-Raacke and Raacke,
2014)7. In a non-experimental research study, one researcher collects data that
naturally occur. In this research, one should not have a definite research question
and he or she cannot develop a hypothesis to establish the causal relationship
between two variables. Unlike experimental research, where the statistical
relationship can be established with two variables, there is one single variable
involved in the research question and also in the hypothesis in the case of non-
experimental research. But there may be two variables to establish a non-causal
statistical relationship. Normally, the research question under this research design
is large and exploratory type. The non-experimental research can be done in two
forms; one is correlational research and another is quasi-experimental research.
In correlational research, one researcher tries to measure two variables, and
understands and establishes the statistical relationship between these two
variables without any influence of extraneous variable. On the other hand, in
quasi-experimental research, one can manipulate the independent variable but
cannot assign conditions or order of conditions randomly to participants.

4.5 Sample Design

The total number of central public sector enterprises was 255 in 2017-18. Out of
these, we have selected 83 CPSEs with the help of simple random sampling

80
without replacement (SRSWOR) method. We have employed the SRSWOR
method because this method gives an equal opportunity for all companies to be
selected in the sample of the research. We have applied Yamane‟s (1967) basic
formula for the determination of sample sizes because the population of the study
is finite. The purpose of using the formula is to have an optimum number of
samples so that we can access the results of the research for the whole
population with good accuracy.

The optimum sample size is determined by using Yamane‟s formula-

𝑁
𝑛=
1+𝑁 𝑒 2

Where, n= optimum sample size, N = population size, and e = margin of error or

sampling error.

Here the adequate sample size is 255 for our study. Since the sample size is a
three-digit one, we have considered three-digit random numbers for the SRSWOR
procedure.

We have started with 83 CPSEs as our sample. But due to unavailability of data,
we considered 73 CPSEs companies as sample size of our research. The lists of
73 CPSEs are given below:

81
 Table 3.1: List of Selected Companies
S. N. Name of the Companies
1 BEML LTD.
2 BHARAT HEAVY ELECTRICALS LTD.
3 BHARAT PETROLEUM CORP. LTD.
4 COAL INDIA LTD.
5 CONTAINER CORP. OF INDIA LTD.
6 ELECTRONICS CORP. OF INDIA LTD
7 GAIL (INDIA) LTD.
8 HINDUSTAN PETROLEUM CORP. LTD.
9 INDIAN OIL [Link].
10 INDIAN RAILWAY FINANCE CORP. LTD.
11 IRCON INTERNATIONAL LTD.
12 KIOCL LTD.
13 MANGALORE REFINERY & PETROCHEMICALS
14 MAZAGON DOCK SHIPBUILDERS LTD.
15 MECON LTD.
16 MMTC LTD.
17 MOIL LTD.
18 MUMBAI RAILWAY VIKAS CORP. LTD.
19 NATIONAL ALUMINIUM CO. LTD.
20 NATIONAL FERTILIZERS LTD.
21 NBCC (INDIA) LTD.
22 NHPC LTD.
23 NLC INDIA LTD.
24 NMDC LTD.
25 NORTH EASTERN ELECTRIC POWER CORP. LTD.
26 NTPC LTD.
27 OIL & NATURAL GAS CORP. LTD.
28 OIL INDIA LTD.
29 ONGC VIDESH LTD.
30 POWER FINANCE CORP. LTD.
31 POWER GRID [Link] INDIA LTD.
32 POWER SYSTEM OPERATION CORP. LTD.
33 RAIL VIKAS NIGAM LTD.
34 RAILTEL CORP. OF INDIA LTD
35 RASHTRIYA CHEMICALS & FERTILIZERS LTD.
36 RASHTRIYA ISPAT NIGAM LTD.

82
37 RURAL ELECTRIFICATION CORP. LTD.
38 SHIPPING CORP. OF INDIA LTD.
39 SJVN LTD.
40 STATE TRADING CORP. OF INDIA LTD.
41 STEEL AUTHORITY OF INDIA LTD.
42 THDC INDIA LTD.
43 ANDREW YULE & CO. LTD.
44 BALMER LAWRIE & CO. LTD.
45 BHARAT COKING COAL LTD.
46 BHARAT DYNAMICS LTD.
47 CENTRAL COALFIELDS LTD.
48 CENTRAL MINE PLANNING & DESIGN INSTITUTE LTD.
49 COCHIN SHIPYARD LTD.
50 COTTON CORP. OF INDIA LTD.
51 DREDGING CORP. OF INDIA LTD.
52 EASTERN COALFIELDS LTD.
53 ENGINEERING PROJECTS (INDIA) LTD.
54 GARDEN REACH SHIPBUILDERS & ENGINEERS LTD.
55 HINDUSTAN SHIPYARD LTD.
56 HINDUSTAN STEELWORKS CONSTRUCTION COMPANY LTD.
57 HLL LIFECARE LTD.
58 INDIA TRADE PROMOTION ORGANISATION
59 INDIAN RAILWAY CATERING & TOURISM CORP. LTD.
60 INDIAN RENEWABLE ENERGY DEVELOPMENT AGENCY LTD.
61 KAMARAJAR PORT LTD.
62 MAHANADI COALFIELDS LTD.
63 MINERAL EXPLORATION CORP. LTD.
64 MISHRA DHATU NIGAM LTD.
65 MSTC LTD.
66 NATIONAL SMALL INDUSTRIES CORP. LTD.
67 NORTHERN COALFIELDS LTD.
68 NUMALIGARH REFINERY LTD.
69 PROJECTS & DEVELOPMENT INDIA LTD.
70 SOUTH EASTERN COALFIELDS LTD.
71 URANIUM CORP. OF INDIA LTD.
72 CENTRAL RAILSIDE WAREHOUSING CO. LTD.
73 EDCIL (INDIA) LTD.

83
4.6 Period of the Study

The study covers a period of eight years from the year 2010-11 to 2017-18. We
started to collect data for 10 years from 2008-09 to 2017-18. But for the years
2008-09 and 2009-10, we could not get data from many companies. So, we have
been bound to choose data period from 2010-11 to 2017-18 financial year. In the
new Companies Act that came into force on 12 September 2013, it has been
mentioned under Sec. 135 that every company, whose net worth is Rs. 500 crore
or more; or has a turnover of Rs. 1,000 core or more; or has a net income of Rs. 5
crore or more, has to spend at least 2% of its average net profit of last three
preceding years on various CSR activities. As every CPSE needs to spend money
on CSR activities after enforcement of the Companies Act, 2013, we could choose
a time period after the year 2013 but even before coming into force of the
Companies Act, 2013, there were some big companies that also had been doing
CSR activity voluntarily. That is why this period (2010-11 to 2017-18) has been
selected for the study.

4.7 Variables used and their meanings

In this research, a total of 14 variables have been used. These variables are CSR,
CEE, CE, CSRTC, DPS, EPS, ICR, NP, ROA, ROE, SAL, SALES, TAX, and VA.
The meanings of these variables are discussed below:

i. CSR: CSR denotes Corporate Social Responsibility. Corporate Social

Responsibility means undertaking social and environmental responsibilities
by companies to save our environment and to provide a better living
standard to the people. How much a company spends on various social
activities, is explained with the CSR variable.

ii. CEE: CEE denotes Capita Employed Efficiency. Capital Employed Efficiency
ratio means how much a company is able to create value after utilising its
capital. The CEE can be calculated as value added divided by capital
employed. If a firm can generate more value by utilising its capital employed
efficiently, then obviously this firm would have money to spend on various
CSR activities (Razafindrambinina and Kariodimejo, 2011)8.

84
iii. CR: CR denotes Cost-Ratio. It is calculated as operating costs divided by
operating profits. This ratio is used to see how much a company is taking
care of its customers.

iv. CSRTC: CSRTC denotes CSR to creditors. It shows the responsibility of a

company towards its creditors. This ratio is calculated as total assets divided
by total liabilities.

v. DPS: DPS denotes Dividend per Share. Here, this variable shows how much
a company takes care of its equity shareholders. Giving more dividends to
equity shares shows that the company is taking care of its equity
shareholders and CSR positively affects DPS (Mittal and Sadhu, 2018)9.

vi. EPS: EPS denotes Earnings per Share. It tells how the company can
generate profit for its equity shareholders. Sometimes CSR has a positive
relationship with EPS (Khan et al., 2016)10. Sometimes CSR has also a
negative relationship with EPS (Rehan et al., 2018)11.

vii. ICR: ICR denotes Interest Coverage Ratio. It is calculated as finance costs
divided by profit before tax. This ratio shows the ability of a company to take
care of its investors i.e., lenders of money supply means that whether a
company can fulfil its responsibility towards its investors by providing due
financial obligations to them (investors) on time. CSR helps to reduce debt-
financing costs (Kordsachia, 2020)12. Doing CSR activities is more relevant
when a company has a low interest coverage ratio (Kordsachia, 2020)13.

viii. NP: NP denotes net profit. It shows a firm‟s ability to earn a profit. Here this
variable is very important because if a firm can earn more profit, then
obviously it can spend more money on various CSR activities. Again, by
doing so, a firm‟s net profit would also be increased [Preston and Bannon
(1997)14, Lu et al. (2009)15, Choi et al. (2010)16, Foote et al. (2010)17] and
vice-versa. This variable is considered as financial performance indicators of
CPSEs in India.

ix. ROA: Return on Assets is denoted by ROA. This variable indicates how
efficiently a company can use its assets to make a profit. A higher value of

85
 ROA indicates that the assets of the company are being utilised efficiently.
This variable is being used as a financial performance indicator of CPSEs in
India. This is an important variable because if the value of ROA of a firm is
high, that indicates the company is spending more money on various CSR
activities, i.e., doing more CSR activities will give rise to higher ROA
[Asatryan and Brezinova (2014)18, Jamali et al. (2015)19].

x. ROE: ROE denotes Return on Equity. This variable shows how a company
can earn profit by utilising its equity. The higher value of ROE means that the
company‟s financial position is good. This is an important variable because it
can be affected by CSR expenditure. It means that if a company is spending
more money on various CSR activities, then that company‟s ROE would be
improved [Jackson and Hua (2009)20, Asatryan and Brezinova (2014)21].

xi. SAL: The salary of employees is denoted by the SAL variable. This variable
shows the responsibility of a firm towards its employees. If a company pays
more salary to its employees, then it would have less money to do CSR
activities (Burbano, 2014)22.

xii. SALES: Sales is an important and significant source of profit earned

through the operational activities of a firm. Sales of a firm can be increased
by doing CSR activities (Nyame-Asiamah and Ghulam, 2019)23.

xiii. TAX: Companies need to pay tax. Paying tax to Government means fulfilling
responsibility towards Government. If a company pays a significant amount
of tax, then that company‟s CSR expenditure will be high (Khan et al,
2014)24. Sometimes CSR is negatively related to corporate tax avoidance
(Gulzar et al., 2018)25.

xiv. VA: VA denotes Value Added. CSR and Value Added are positively related
to each other (Guadano and Pedroza, 2018)26.

4.8 Sources of the data

In this study, secondary data have been used. All the data have been collected
from the annual reports of individual companies, [Link], and CMIE
Prowess Database.

86
4.9 Tools used

To measure the relationship between CSR and the financial performance of

CPSEs in India, we have applied the correlation and panel regression test. We
have also applied the unit root test, panel cointegration test, and VECM Granger
Causality.

4.9.1 Correlation Analysis

Correlation helps to measure the strength and direction of the relationship

between two variables. The two variables are said to be correlated if one variable
gets changed by the change in the other variable (Schober et al., 2018)27. The
value of the degree of correlation ranges from -1 to +1. If the value of the
correlation coefficient is „+1‟, it means that the two variables have a perfectly
positive relationship, whereas, if the value of the correlation coefficient is „-1‟, it
means that the two variables have a perfectly negative relationship between them
and the „0‟ (zero) value of the coefficient of correlation indicates that the variables
are not related to each other (Schober et al., 2018)28.

4.9.2 Panel Regression analysis

Here our data is panel data where we have collected data of 73 companies for 8
years which means we have cross-sectional as well as time-series data. So, the
regression technique that we need to use is panel data regression analysis. In
regression analysis, we can see how the dependent variable gets affected by the
independent variable. In the present study, two-panel regression models, viz.
fixed- effects model and random-effects model have been employed. Further, the
Hausman specification test has been used in order to find out which panel model
(fixed-effects model or random-effects model) should be used in this study. The
fixed effects-model, random-effects model and Hausman specification test are
discussed below:

[Link] Fixed Effects Model

The fixed-effects model is one kind of panel regression model where all
parameters are non-random or fixed types i.e., the values of independent
variables are constant over time. The main goal of the fixed effects model is to
estimate one true effect. The fixed-effects model investigates the relationship

87
between the dependent and independent variables. In the case of using the fixed-
effects model, we assume that there is something within a variable that may affect
or bias the dependent and independent variables and should control this
unobserved variable (Torres-Reyna, 2007)29. The key assumptions of this model
are that the unobserved variables are invariant type over a time period and any
change in dependent variable will be caused by the independent variable (Stock
and Watson, 2006, p. 356-358)30. We need to remove the effect of these time-
invariant characteristics with the help of the fixed-effects model to have the net
impacts of the predictors on the outcome variable (Torres-Reyna, 2007)31. In the
fixed-effects model, the researcher can control omitted variable which varies
across the entities but remains the same over the time period (Stock and Watson,
2006, p. 356-358)32. Baltagi (2005)33 identified that the intercept of fixed-effects
models (FEM) can differ from case to case but it cannot be changed over time. It
means that as per FEM, the slopes of coefficients are constant but the intercept
varies across cross-sectional units. This kind of approach takes into consideration
individuality by allowing the intercept to vary with cases while slope coefficients
are considered as constant across firms.

Yit= α + α1D1i +…+ αnDni+ β1X1it … βmXmit+ γ1 (D1iX1it) +…+ γw(DniXmit) + μit

t = 1…t n = 1…number of dummy variables

i= 1…i m = 1…number of independent variables

w = 1…number of interactions

If we see that any of these interaction results are statistically significant, then we
can say that there is a statistically significant difference between groups. If the
relationship between D1i andX1it is statistically significant, then the net impacts of
a particular case, say i over time, is (β1 + γ1), which suggests that X1 is different
from the comparison case. If we see that all the differential slope coefficients and
the differential intercepts are statistically significant, then we can conclude that all
cases are not similar and they are different from the comparison case.

88
[Link] Random Effects Model

The random-effects model (REM) is one kind of panel regression model where all
the parameters are random, i.e., not-fixed type, and the data are collected from
different populations. The assumption of the random model is that there is no
omitted variable or the omitted variable is not related to the independent or
explanatory variable. The main objective of REM is to estimate the mean of a
distribution of effects of panel dataset. The application of REM will be appropriate
when we can draw N number of individuals from a large population randomly. If
we use the fixed-effects model when the N is large, then there would be a
massive loss of a degree of freedom (Baltagi, 2005)34The REM assumes that the
intercept (α1i) is a random variable with a mean value of α1 (without subscript i)
and the error term (εi) with the mean value of zero and variance of ζε2. This is
shown below:

(1)

The assumption behind the framework is that all cases have been taken from a
large population and the selected variables have the same grand mean value for
the intercept (α1) and the individual differences are shown in the error term (εi).

(2)

Where εi refers to the individual –specific error (cross-section), µit refers to the
combined error, i.e., both time series as well as cross-section error. The
assumptions underlying the composite error term are as follows:

89
 (3)

In general, the above assumptions suggest that there is no correlation between

the error terms and also there is no autocorrelation among both cross-section and
time- series units. The composite error, εi, is hidden, as well as unobservable.

[Link] Hausman Specification Test

Hausman specification test was proposed by Hausman in the year 1978 that
identifies the independent variable in a regression model. Before selecting the
best regression method, one should figure out whether the predictor variables
affect the endogenous variable or not. This test helps us to identify the best model
between the fixed-effects model and the random-effects model. In this test, the
null hypothesis is constructed as “the random-effects model is suitable” and the
alternative hypothesis is “the random-effects model is not suitable”, i.e., in the
case of the alternative hypothesis, the fixed effects model is preferred. In the case
of the Hausman specification test, if we see that the probability value is less than
0.05, then we should reject the null hypothesis which tells us that the fixed-effects
model is appropriate, and if the probability value is more than 0.05, then we
should accept the null hypothesis. It indicates that the random-effects model is
more appropriate.

[Link] Unit Root Tests

Panel unit root is used to check the stationarity of variables (Mahembe &
Odhiambo, 2019)35. We need to do the unit root tests to see whether our data are
stationary or non-stationary, as our data is panel data where the time series
elements are also included. To check the stationarity of the panel data series,
ADF-Fisher Chi-Square test (ADF-1979), Philips-Perron (PP-1998), Levin, Lin and

90
Chu (2002), and Im, Pesaran, and Shin (2003) panel unit root tests have been
used. Now we can start understanding the unit root (stochastic) process from
below:

Random Walk Model-

Yt = ρYt−1 + µt ............................................................................................. Eq. (1)

Where,

Yt= Value of a discrete variable at t time period

Y( t-1)= The lagged value of Yt for one period.

ρ= Rho, the value of ρ belongs to -1 to +1

µt= White-noise error term

Now if we consider the value of ρ as 1, then the above random walk model will be
called as random walk model without drift, which is called a non-stationary
stochastic process. Therefore, we need to regress Yt on its (one-period) lagged
value Y( t-1.) . Now the equation 1 will be as –

Yt- Y( t-1) = ρ Y( t-1) - Y( t-1) + µt

Yt- Y( t-1) =(ρ-1) Y( t-1) + µt.............................................................................. Eq. (2)

We can alternatively write the above equation the value of equation (2) as-

ΔYt = δY(t−1 )+ µt ......................................................................................... Eq. (3)

Where, (ρ-1) = δ and Δ = First difference operator.

Now, if we consider δ= 0, then, ρ will be 1, which means there is a unit root that
tells us that the time series is non-stationary.

Now, putting the value of δ= 0, we get-

ΔYt= 0 x Y(t−1 )+µt [If we multiply the Y(t−1 ) with zero, it will be zero]

= 0+ µt

= µt

91
Therefore,

ΔYt= (Yt- Y( t-1))= µt

As we know that µt is a white error term, the time series now becomes stationary
which indicates that at the first difference, the random walk time series will be
stationary (Gujarati and Porter, 2020)36.

Levin, Lin and Chu Test

Levin, Lin, and Chu (LLC) simplify Quah‟s procedure to allow for heterogeneity of
individual deterministic trends; heterogeneous auto corrected errors, and fixed
effects. They assumed that - both N, i.e., number of cross-section units, and T,
i.e., the length of the time-series - tend to infinity, though, T increases at a faster
rate than cross-section unit, N. LLC reveals that unit root test based on panel
data gives more powerful result than doing unit root test for individual cross-
section data. The null hypothesis of the LLC mechanism of unit root test is that
each individual time series has a unit root or non-stationary and the alternative
hypothesis is that each time series has no unit root or each time series is
stationary (Balatgi, 2005, pp 240-241)37.

pi

Δ Yit= ρ Yi,( t-1)+ ∑ θiLΔYit-L + αmidmt+ ɛit

L=1

Where,

Δ Yit= Value of Y at first difference

ρ= Rho

Yi,( t-1)=The lagged value of Yt for one period.

αmi=The corresponding vector of coefficients for model m=1,2,3.

dmt = The vector of deterministic variables

ɛit= Error term

Pi= The lagged order

92
Im, Pesaran and Shin Unit Root Test

Im, Pesaran, and Shin (2003) test permits heterogeneous deterministic intercept
terms and develops another testing procedure which is based on averaging
individual unit root test statistics. If we see that the error term µit is serially
correlated with serial correlation characteristics of all cross-sectional units, then
we can use an average of the Augmented Dickey-Fuller test as suggested by IPS.
This test can also be used for unbalanced panel data. The null hypothesis of the
IPS test is that each series is non-stationary, i.e., each series has the unit root and
the alternative hypothesis is that some series, but not all series, are non-
stationary, i.e., some series have unit roots. We can have null hypothesis- H0: 1
= 2 = … = N = 0 and the alternative hypothesis H1: Some but not necessarily all
i< 0. We need to find out t-ratios for the αi, t , after considering the dtr and the
i
pi, then we also need to calculate their arithmetic average i.e., arithmetic mean-

t NT  iN1t i / N .IPS illustrates that t NT can be adjusted to yield an

asymptotic statistic [N (0, 1)] under the null hypothesis:

 N 
N 1 / 2  t NT  N 1  E (t i ) 
*
t NT   i 1 
1/ 2
 1 N 
 N Var(t i )
 i 1 

The E (t ) and Var(t ) have been found out by simulation of IPS.
i i

4.9.3 VECM Granger Causality

The vector error correction model (VECM) to test the Granger causality was
proposed by Granger in 1988 (Mahembe and Odhiambo, 2019)38. The Granger
causality test has been developed in the year 1969. Granger causality test is an
important tool to analyse the causal relationship between two time series where
the test helps us to forecast a one time series with the help of another time series
(Wei, 2016)40. As per the concept of Granger causality test, „ X Granger causes
Y ‟ if the variable X has past records to predict the change of Y variable and „ Y

93
Granger causes X‟ if and only if Y variable has past records to predict the change
of X variable ([Link]). We can do this process with the help of the
Vector Autoregression (VAR) model. But, if the variables are cointegrated into
each other, then these variables must have a short-run and long-run causal
relationship, and this relationship is not established by the standard first difference
VAR model. In this case, the Vector Error Correction Model (VECM) can be used
which was proposed by Granger in 1988. The VECM framework is as follows:

𝑝 𝑝 𝑝

1𝑛𝐶𝐸𝐸 = ∝1 + 1 1𝑛𝐶𝐸E𝑡−1 + 2 1nCR 𝑡−1 + 3 1nCSR 𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝 𝑝

+ 4 1nCSRTC𝑡−1 + 5 1nDPS𝑡−1 + 6 1nEPS𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝 𝑝

+ 7 1nICR 𝑡−1 + 8  1nNP𝑡−1 + 9  1nROA𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝 𝑝

+ 10  1nROE𝑡−1 + 11  1nSAL𝑡−1 + 12  1nSALES𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝

+ 13  1nTAX𝑡−1 + 14  1nVA𝑡−1 + Ѵ1t

𝑖=1 𝑖=1

𝑝 𝑝 𝑝

1𝑛𝐶𝑅 = ∝2 + 1 1nCR 𝑡−1 + 2 1𝑛𝐶𝐸E𝑡−1 + 3 1nCSR 𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝 𝑝

+ 4 1nCSRTC𝑡−1 + 5 1nDPS𝑡−1 + 6 1nEPS𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝 𝑝

+ 7 1nICR 𝑡−1 + 8  1nNP𝑡−1 + 9  1nROA𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝 𝑝

+ 10  1nROE𝑡−1 + 11  1nSAL𝑡−1 + 12  1nSALES𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝

+ 13  1nTAX𝑡−1 + 14  1nVA𝑡−1 + Ѵ2t

𝑖=1 𝑖=1

94
 𝑝 𝑝 𝑝

1𝑛𝐶𝑆𝑅 = ∝3 + 1 1nCSR 𝑡−1 + 2 1𝑛𝐶𝐸E𝑡−1 + 3 1nCR 𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝 𝑝

+ 4 1nCSRTC𝑡−1 + 5 1nDPS𝑡−1 + 6 1nEPS𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝 𝑝

+ 7 1nICR 𝑡−1 + 8  1nNP𝑡−1 + 9  1nROA𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝 𝑝

+ 10  1nROE𝑡−1 + 11  1nSAL𝑡−1 + 12  1nSALES𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝

+ 13  1nTAX𝑡−1 + 14  1nVA𝑡−1 + Ѵ3t

𝑖=1 𝑖=1

𝑝 𝑝 𝑝

1𝑛𝐶𝑆𝑅𝑇𝐶 = ∝4 + 1 1nCSRTC𝑡−1 + 2 1𝑛𝐶𝐸E𝑡−1 + 3 1nCR 𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝 𝑝

+ 4 1nCSR 𝑡−1 + 5 1nDPS𝑡−1 + 6 1nEPS𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝 𝑝

+ 7 1nICR 𝑡−1 + 8  1nNP𝑡−1 + 9  1nROA𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝 𝑝

+ 10  1nROE𝑡−1 + 11  1nSAL𝑡−1 + 12  1nSALES𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝

+ 13  1nTAX𝑡−1 + 14  1nVA𝑡−1 + Ѵ4t

𝑖=1 𝑖=1

𝑝 𝑝 𝑝

1𝑛𝐷𝑃𝑆 = ∝5 + 1 1nDPS𝑡−1 + 2 1𝑛𝐶𝐸E𝑡−1 + 3 1nCR 𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝 𝑝

+ 4 1nCSR 𝑡−1 + 5 1nCSRTC𝑡−1 + 6 1nEPS𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝 𝑝

+ 7 1nICR 𝑡−1 + 8  1nNP𝑡−1 + 9  1nROA𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝 𝑝

+ 10  1nROE𝑡−1 + 11  1nSAL𝑡−1 + 12  1nSALES𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝

+ 13  1nTAX𝑡−1 + 14  1nVA𝑡−1 + Ѵ5t

𝑖=1 𝑖=1

95
 𝑝 𝑝 𝑝 𝑝

1𝑛𝐸𝑃𝑆 = ∝6 + 1 1nEPS𝑡−1 + 2 1𝑛𝐶𝐸E𝑡−1 + 3 1nCR 𝑡−1 + 4 1nCSR 𝑡−1

𝑖=1 𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝 𝑝

+ 5 1nCSRTC𝑡−1 + 6 1nDPS𝑡−1 + 7 1nICR 𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝 𝑝

+ 8  1nNP𝑡−1 + 9  1nROA𝑡−1 + 10  1nROE𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝 𝑝

+ 11  1nSAL𝑡−1 + 12  1nSALES𝑡−1 + 13  1nTAX𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝

+ 14  1nVA𝑡−1 + Ѵ6t

𝑖=1

𝑝 𝑝 𝑝 𝑝

1𝑛𝐼𝐶𝑅 = ∝7 + 1 1nICR 𝑡−1 + 2 1𝑛𝐶𝐸E𝑡−1 + 3 1nCR 𝑡−1 + 4 1nCSR 𝑡−1

𝑖=1 𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝 𝑝

+ 5 1nCSRTC𝑡−1 + 6 1nDPS𝑡−1 + 7 1nEPS𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝 𝑝

+ 8  1nNP𝑡−1 + 9  1nROA𝑡−1 + 10  1nROE𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝 𝑝

+ 11  1nSAL𝑡−1 + 12  1nSALES𝑡−1 + 13  1nTAX𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝

+ 14  1nVA𝑡−1 + Ѵ7t

𝑖=1

𝑝 𝑝 𝑝 𝑝

1𝑛𝑁𝑃 = ∝8 + 1  1nNP𝑡−1 + 2 1𝑛𝐶𝐸E𝑡−1 + 3 1nCR 𝑡−1 + 4 1nCSR 𝑡−1

𝑖=1 𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝 𝑝

+ 5 1nCSRTC𝑡−1 + 6 1nDPS𝑡−1 + 7 1nEPS𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝 𝑝

+ 8 1nICR 𝑡−1 + 9  1nROA𝑡−1 + 10  1nROE𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝 𝑝

+ 11  1nSAL𝑡−1 + 12  1nSALES𝑡−1 + 13  1nTAX𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝

+ 14  1nVA𝑡−1 + Ѵ8t

𝑖=1

96
 𝑝 𝑝 𝑝

1𝑛𝑅𝑂𝐴 = ∝9 + 1  1nROA𝑡−1 + 2 1𝑛𝐶𝐸E𝑡−1 + 3 1nCR 𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝 𝑝

+ 4 1nCSR 𝑡−1 + 5 1nCSRTC𝑡−1 + 6 1nDPS𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝 𝑝

+ 7 1nEPS𝑡−1 + 8 1nICR 𝑡−1 + 9  1nNP𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝 𝑝

+ 10  1nROE𝑡−1 + 11  1nSAL𝑡−1 + 12  1nSALES𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝

+ 13  1nTAX𝑡−1 + 14  1nVA𝑡−1 + Ѵ9t

𝑖=1 𝑖=1

𝑝 𝑝 𝑝

1𝑛𝑅𝑂𝐸 = ∝10 + 1 1𝑛𝐶𝐸E𝑡−1 + 2 1nCR 𝑡−1 + 3 1nCSR 𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝 𝑝

+ 4 1nCSRTC𝑡−1 + 5 1nDPS𝑡−1 + 6 1nEPS𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝 𝑝

+ 7 1nICR 𝑡−1 + 8  1nNP𝑡−1 + 9  1nROA𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝 𝑝

+ 10  1nROE𝑡−1 + 11  1nSAL𝑡−1 + 12  1nSALES𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝

+ 13  1nTAX𝑡−1 + 14  1nVA𝑡−1 + Ѵ10t

𝑖=1 𝑖=1

𝑝 𝑝 𝑝

1𝑛𝑆𝐴𝐿 = ∝11 + 1  1nSAL𝑡−1 + 2 1𝑛𝐶𝐸E𝑡−1 + 3 1nCR 𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝 𝑝

+ 4 1nCSR 𝑡−1 + 5 1nCSRTC𝑡−1 + 6 1nDPS𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝 𝑝

+ 7 1nEPS𝑡−1 + 8 1nICR 𝑡−1 + 9  1nNP𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝 𝑝

+ 10  1nROA𝑡−1 + 11  1nROE𝑡−1 + 12  1nSALES𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝

+ 13  1nTAX𝑡−1 + 14  1nVA𝑡−1 + Ѵ11t

𝑖=1 𝑖=1

97
 𝑝 𝑝 𝑝

1𝑛𝑆𝐴𝐿𝐸𝑆 = ∝12 + 1  1nSALES𝑡−1 + 2 1𝑛𝐶𝐸E𝑡−1 + 3 1nCR 𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝 𝑝

+ 4 1nCSR 𝑡−1 + 5 1nCSRTC𝑡−1 + 6 1nDPS𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝 𝑝

+ 7 1nEPS𝑡−1 + 8 1nICR 𝑡−1 + 9  1nNP𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝 𝑝

+ 10  1nROA𝑡−1 + 11  1nROE𝑡−1 + 12  1nSAL𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝

+ 13  1nTAX𝑡−1 + 14  1nVA𝑡−1 + Ѵ12t

𝑖=1 𝑖=1

𝑝 𝑝 𝑝

1𝑛𝑇𝐴𝑋 = ∝13 + 1  1nTAX𝑡−1 + 2 1𝑛𝐶𝐸E𝑡−1 + 3 1nCR 𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝 𝑝

+ 4 1nCSR 𝑡−1 + 5 1nCSRTC𝑡−1 + 6 1nDPS𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝 𝑝

+ 7 1nEPS𝑡−1 + 8 1nICR 𝑡−1 + 9  1nNP𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝 𝑝

+ 10  1nROA𝑡−1 + 11  1nROE𝑡−1 + 12  1nSAL𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝

+ 13  1nSALES𝑡−1 + 14  1nVA𝑡−1 + Ѵ13t

𝑖=1 𝑖=1

𝑝 𝑝 𝑝 𝑝

1𝑛𝑉𝐴 = ∝14 + 1  1nVA𝑡−1 + 2 1𝑛𝐶𝐸E𝑡−1 + 3 1nCR 𝑡−1 + 4 1nCSR 𝑡−1

𝑖=1 𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝 𝑝

+ 5 1nCSRTC𝑡−1 + 6 1nDPS𝑡−1 + 7 1nEPS𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝 𝑝

+ 8 1nICR 𝑡−1 + 9  1nNP𝑡−1 + 10  1nROA𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝 𝑝 𝑝

+ 11  1nROE𝑡−1 + 12  1nSAL𝑡−1 + 13  1nSALES𝑡−1

𝑖=1 𝑖=1 𝑖=1
𝑝

+ 14  1nTAX𝑡−1 + Ѵ14t

𝑖=1

98
Selected References

1. Soiferman, L. K. (2010). Compare and Contrast Inductive and Deductive

Research Approaches. Inductive and Deductive Research Approaches, 1-
23. [Link] Retrieved on 08.11.2020.

2. Thomas, D. R. (2006). A General Inductive Approach for Analysing

Qualitative Evaluation Data. [Link]
Retrieved on 08.112020.

3. Pathak, V., Jena, B. & Kalra, S. (2013). Qualitative Research. Perspective

in Clinical Research, 4(3). Doi: 10.4103/2229-3485.115389. Retrieved on
08.11.2020.

4. Kothari, C. R. (2004). Research Methodology Methods and Techniques

(Second Revised Edition). New Age International (p) Limited. Available at
[Link] Retrieved on
10.12.2020.

5. Boru, T. (2018). Chapter Five Research Design and Methodology.

Available on [Link] Retrieved on
08.11.2020.

6. Kothari, C. R. (2004). Research Methodology Methods and Techniques

(Second Revised Edition). New Age International (p) Limited. Available at
[Link] Retrieved on
10.12.2020.

7. Bonds-Raacke, J. M. & Raacke, J. D. (2014). Non-experimental Research

Methods from Research Methods: Are you Equipped? Available at
[Link]
ontent/Higher_Education/Uploads/Bonds_Raacke_Research_2e_Ch4.pdf.
Retrieved on 09.12.2020.

8. Razafindrambinina, D. & Kariodimedjo, D. (2011). Is Company Intellectual

Capital Linked to Corporate Social Responsibility Disclosure? Findings
from Indonesia. IBIMA Publishing. DOI: 10.5171/2011. 511442. Retrieved
on 12.10.2020.

99
9. Mittal, A. & Sandhu, N. (2018). Relationship between Corporate Social
Responsibility, Dividend Payout, EBITDA, and Earnings per Share: An
Empirical Analysis of 40 Listed Companies in NSE. Journal of Advances
and Scholarly Researches in Allied Education, 15(7), 83-91.
DOI: 10.29070/15/57797. Retrieved on 11.10.2020.

10. Khan, F., Rahman, M. M., Ullah, M. W., & Tanu, M. T. (2016). Impact of
Corporate social responsibility expenditures on bank‟s financial
performance: A study on the Southeast Bank Ltd. International Journal of
Multidisciplinary Research and Development, 3(2), 271-276.

11. Rehan. M., Khan, M. I. & Khan, M. K. (2018). Effect of Corporate Social
Responsibility on Profitability of Banks. European Academic Research, VI
(7), 3763-3782.

12. Kordsachia, O. (2020). A risk management perspective on CSR

and the marginal cost of debt: empirical evidence from Europe.
[Link] Retrieved on 11.11.2020.

13. Kordsachia, O. (2020). A risk management perspective on CSR

and the marginal cost of debt: empirical evidence from Europe.
[Link] Retrieved on 11.11.2020.

14. Preston, L.E. & O‟Bannon, D.P. (1997). The corporate Social –Financial
Performance Relationship A Typology and Analysis. Business
Society,36(4), 419-429.

15. Lu, C., Change, C. & Tu, C. (2009). Corporate Social Responsibility and
Organisational Performance in Container Shipping. International Journal of
Logistics Research and Applications, 12(1), 119-132.

16. Choi,J-S; Kwak,Y. M., & Choe,C. (2010). Corporate Social Responsibility
and Corporate Financial Performance: Evidence from Korea. Australian
Journal of Management. DOI 10.1177/0312896210384681.

17. Foote, N., Gaffney, N., & Evans, J. R. (2010). Corporate Social
Responsibility: Implications for performance excellence. The Quality
Management & Business Excellence, 21(8), 799-812.

100
18. Asatryan, R. & Brezinova, O. (2014). Corporate Social Responsibility and
Financial Performance in the Airline Industry in Central and Eastern
Europe.
ActaUniversitatiesAgriculturaeetSilviculturaeMendelianaeBrunensis, 62(4),
633-639.

19. Jamali, H., T, S. Subekti, I. & Assih, P. (2015). The Influence of Corporate
Governance and Corporate Social Responsibility on Financial Performance
with Efficiency as Mediating Variable. International Journal of Business and
Management Invention, 4(5), 01-10.

20. Jackson, L. & Hua, N. (2009). Corporate Social Responsibility and financial
performance: A Snapshot from Lodging and Gaming Industries. Journal of
Hospitality Management, 17(1), 63-78.

21. Asatryan, R. & Brezinova, O. (2014). Corporate Social Responsibility and

Financial Performance in the Airline Industry in Central and Eastern
Europe.
ActaUniversitatiesAgriculturaeetSilviculturaeMendelianaeBrunensis, 62(4),
633-639.

22. Burbano, V. C. (2014). Can Firms pay less and Get more...By Doing Good?
The Effect of Corporate Social Responsibility on Employee Salary
Requirements and Performance. [Link]
content/uploads/[Link]. Retrieved on 12.10.2020.

23. Nyame-Asiamah, F. & Ghulam, S. (2019). The relationship between CSR

activity and sales growth in the UK retailing sector.
[Link] Retrieved on 11.11.2020.

24. Khan, M., Yousaf, Z., Khan, Z. A., & Yasir, M. (2014). Analysis of the
Relationship between CSR and Tax Avoidance: Evidence from Pakistan.
The International Journal of Business & Management, 2(7), 53-57.

25. Gulzar, M. A., Cherian, J., Sial, M. S., Badulescu, A., Thu, P. A.,
Badulescu, D. & Khuong, N. V. (2018). Does Corporate Social
Responsibility Influence Corporate Tax Avoidance of Chinese Listed

101
 Companies? Sustainability.[Link]
[Link]. Retrieved on 11.11.2020.

26. Guadano, J. F. & Pedroza, J, H. (2018). Impact of Corporate Social

Responsibility on Value Creation from a Stakeholder Perspective.
Sustainability. [Link]
02062%20(1).pdf. Retrieved on 12.10.2020.

27. Schober, P., Boer, C., & Schwarte, L. A. (2018). Correlation Coefficients:
Appropriate Use and Interpretation. Anesthesia& Analgesia. 126(5). DOI:
10.1213/ANE.0000000000002864. Retrieved on 12.11.2020.

28. Schober, P., Boer, C., & Schwarte, L. A. (2018). Correlation Coefficients:
Appropriate Use and Interpretation. Anesthesia & Analgesia. 126(5). DOI:
10.1213/ANE.0000000000002864. Retrieved on 12.11.2020.

29. Torres-Reyna, O. (2007). Panel Data Analysis Fixed and

Random Effects using Stata (v. 4.2). Available at
[Link] Retrieved on
08.12.2020.

30. Stock, J. H. & Waston, M. W. (2006). Introduction to Econometrics.

Available at [Link]
[Link]. Retrieved on
13.11.2020.

31. Torres-Reyna, O. (2007). Panel Data Analysis Fixed and

Random Effects using Stata (v. 4.2). Available at
[Link] Retrieved on
08.12.2020.

32. Stock, J. H. & Waston, M. W. (2006). Introduction to Econometrics.

Available at [Link]
[Link]. Retrieved on
13.11.2020.

33. Baltegi, B. H. (2005). Econometrics Analysis of Panel Data. John wiley&

Sons, Ltd. Available at

102
 [Link]
econometric-analysis-of-panel-data_himmy.pdf. Retrieved on 10.12.2020.

34. Baltegi, B. H. (2005). Econometrics Analysis of Panel Data.

John wiley & Sons, Ltd. Available at
[Link]
econometric-analysis-of-panel-data_himmy.pdf. Retrieved on 10.12.2020.

35. Mahembe, E. & Odhiambo, N. M. (2019). Foreign aid,

poverty and economic growth in developing countries: A dynamic panel
data causality. Cogent Economics & Finance, 7, 1-17.
[Link] Retrieved on 01.12.2020.

36. Gujarati, D. N. & Porter, D. c. (2009). Basic Econometrics. McGraw-

Hill/Irwin. Available at
[Link] Retrieved
on 13.12.2020.

37. Baltegi, B. H. (2005). Econometrics Analysis of Panel Data. John wiley &
Sons, Ltd. Available at
[Link]
econometric-analysis-of-panel-data_himmy.pdf. Retrieved on 10.12.2020.

38. Mahembe, E. & Odhiambo, N. M. (2019). Foreign aid, poverty and

economic growth in developing countries: A dynamic panel data
causality. Cogent Economics & Finance, 7(1), 1-17.
[Link] Retrieved on 01.12.2020.

103