Tax Policy and Economic Growth in the Developing and Developed Nations

Full text

International Journal of Finance and Managerial Accounting, Vol.4, No.14, Summer 2019
15
With Cooperation of Islamic Azad University – UAE Branch
Tax Policy and Economic Growth in the Developing
and Developed Nations
Masud Shahmoradi
Ph.D.Student in Department of Accounting, Sanandaj Branch, Islamic Azad University, Sanandaj, Iran.
shahmoradi299@yahoo.com
Ata Mohamadi Molqarani
Assistant Professor in Department of Accounting, Sanandaj Branch, Islamic Azad University, Sanandaj, Iran.
(Corresponding author)
Ataata.mm68@yahoo.com
Farzad Moayri
Assistant Professor in Department of Economic, Sanandaj Branch, Islamic Azad University, Sanandaj, Iran.
f.Moayeri46@gmail.com
ABSTRACT
Fiscal policy is a policy that tries to achieve certain economic goals through instruments such as changes in
government expenditure and taxation. The financial policy uses two instruments of government revenue (tax) and
government spending (spending) to influence the economy. And in the economic literature, they consider
economic growth to be equal to GDP. The impact of financial policies on economic growth is a matter of many
economists. The source of finance used by the government of majority of the world countries is mainly tax
revenue. Tax paid by the public is an effective instrument for developing financial policy by the nations. This
study aims to investigate the impact of financial policy (tax) on GDP of the developed and developing countries
using the panel data technique. The data is collected from World Bank database for the period 2008-2016. The
results of analysis revealed that there is a negative and significant relationship between the logarithm of the ratio
of tax revenues and GDP in the developed countries; however, there is no significant relationship between tax
revenues and economic growth in developing countries.
Keywords:
Tax Policy, Tax Revenue, Economic Growth, GDP, Panel Data
1. Introduction

16 / Tax Policy and Economic Growth in the Developing and Developed Nations
Vol.4 / No.14 / Summer 2019
In the economic literature, gross domestic product
(GDP) has been considered the measure of economic
growth. Almost all countries aim to achieve higher
economic growth and suitable conditions are also
required to achieve this goal. One of the most
important determinants of economic growth of a
country is governments’ tax policy.Setting levels of
taxation is one of the significant factors to drive the
economic policy of nations.
The expansion and diversification of economic
activities and the growing role of governments to
create and expand public services, social security, the
expansion of government commitments in the
economic and social spheres, and efforts to achieve
economic growth and fair distribution of income
depend significantly on indirect taxes and receive tax
payments. Accordingly, tax policies can influence
economic development signfifcantly.
The government gets most of its expenditure
through tax revenue and resource allocation decisions
according to the national priorties higly depend on the
the amount of tax collected. Also, economic planning
is conducted through evaluating the level of
imaginable resources required for future investment
based on revenues earned through tax. Tax may affect
economic choices and ultimately economic growth via
its impact on return of physical and human capital.
In most developed countries, the level of taxes has
risen significantly over the course of the current
century, (level of
taxes have been increased from about 5–10 per
cent of GDP at the turn of the century to 20–30 per
cent at present). Such a significant increase in taxes
has raised questions concerning the impact of tax
structure on economic growth (Lee and Gordon,
2005).
In this paper we shall try to describe how taxes
revenue impact on GDP of the developed and
developing countries separately?
2. Literature Review
In their paper entitled “The Impact of Taxes and
Government Consumption Expenditure on Economic
Growth of Selected MENA Islamic Countries,”
Asghari and Zonnouzi (2013) investigated the effect of
taxes and Government Consumption Expenditure on
economic growth of selected Islamic countries from
MENA region during 1995-2011, using Panel Smooth
Transition Regression (PSTR) model. The results
indicate that taxes and government consumption
expenditure have negative impact on economic
growth. As threshold of GDP for government
consumption expenditure and taxes increases, the
positive effects of investment and export revenues on
economic growth decrease.
In their paper, “An Investigation of the
Relationship between Taxation and Economic Growth
(The Case Study: Iran, OPEC and OECD Countries),”
Faramarzi and et al (2015) investigated the
relationship between the taxes and economic growth in
Iran, OPEC and OECD. The results of this research for
Iran during 1963-2011 indicated that there is no
significant causal relationship between taxes and
economic growth. The results also in 26 OECD
member countries during 1998 to 2011 indicate a long-
term causal relationship between taxes and economic
growth and taxes has negatively affected the economic
growth. However, these results from Pedroni and Kao
tests also indicate negative causalrelationship between
taxes and economic growth in OPEC selected
countries during 1994-2011 which is the result of the
oil-dependent economy.
In their paper, Poulson & Kaplan explored the
impact of tax policy on economic growth in the states
within the framework of an endogenous growth model.
Regression analysis was used to estimate the impact of
taxes on economic growth in the states from 1964 to
2004. The analysis revealed a significant negative
impact of higher marginal tax rates on economic
growth.
In his paper, Matthew Ocran (2009) examined the
effect of fiscal policy variables including, government
gross fixed capital formation, tax revenue and
government consumption expenditure as well as
budget deficit, on economic growth in South Africa
during 1990 to 2004. Quarterly data was used in the
estimation with the aid of vector regressive modeling
technique. The outcome indicated that there is a
positive relationship between government
consumption expenditure and establishment of gross
capital investment entities and economic growth;
however, there is a negatve relathionship between
government consumption expenditure and economic
growth.
In their paper, Canicio & Zachary (2014), the
effects of government tax revenue growth on
economic growth were investigated for Zimbabwe

International Journal of Finance and Managerial Accounting / 17
Vol.4 / No.14 / Summer 2019
during the period of 1980-2012. The study applied the
Granger Causality test, Johansen’s cointegration test
and vector autoregressive model to serve the purpose.
Findings of this study clearly revealed that there is an
independence relationship between economic growth
and total government tax revenue with 30% speed of
adjustment in the short run towards equilibrium level
in the long run. This implies that there is fiscal
independence between tax revenue and growth.
Investigating the short-run and long-run relationship
between the tax revenue and economic growth in
Zimbabwe reveled that taxes affect the allocation of
resources and often distort the economic growth.
Gondor & Ozpence (2014) conducted an empirical
study on fiscal policy in crises time in Romania and
Turkey. Providing some empirical basis for the
argument, they reveal that pro-cyclical fiscal policy
does not assist in dampening the GDP shocks. Being
focused on empirical, contextualized analysis, their
study concentrates on the cyclical dynamics of
macroeconomic aggregates and only offers conjectures
as to the reasons behind the behavior of fiscal policy
and its influence on the macroeconomic output.
Binti Saidin et al., (2016) studied the role and
impact of tax in economic growth in 27 selected Asian
countries for 5 year time period (2011-2015) using
panel data. The relationship between the dependent
variables (GDP per capita and FDI rate) and
independent variables (individual income tax,
corporate tax, and consumption tax) was investigated
in order to identify the role of tax in economic growth.
Descriptive analysis and regression analysis will be
adopted to analyze the data. Their research verified the
negative relationship between personal income taxes
and economic growth while there is a significant
relationship between personal income taxes and
foreign direct investment rate. They also found out that
there is a negative relationship between corporate
taxation and the dependent variables of the
research.Ojong et al, (2016) examined the impact of
tax revenue on the Nigerian economy during the
period 1986–2010.In their study, they examined the
impact of company income tax and the effectiveness of
non oil revenue on the Nigerian economy. Ordinary
least square (OLS) of multiple regression models was
used to establish the relationship between dependent
and independent variables. The finding revealed that
there is a significant relationship between petroleum
profit tax and the growth of the Nigeria economy;
however, there is no significant relationship between
company income tax and the growth of the Nigeria
economy .
In their paper, Ahmad et al., (2016) examined the
relationship between total tax revenues and economic
growth in Pakistan using annual time series data from
1974 to 2010. Auto Regressive Distributed Lag
(ARDL) bounds testing approach for co-integration,
was applied to estimate, the long run and short run
relationship, among the variables. The results showed
that total tax revenues have negative and significant
effect, on economic growth, in long run. Due to one
percent upsurge in total taxes, economic growth would
be reduced by -1.25 percent.
Eugene and Chineze (2016) studied the impact of
taxation policies on the overall economic growth of
Nigeria during the period 1994-2013 using OLS
method. The results of the study confirmed the
positive impact of a tax on Nigeiran economic growth.
Babatunde et al., (2017) investigated the impact of
taxation on the economic growth during the period
2004-2013 in 16 African states using Panel Data. The
results revealed a significant and positive relationship
between tax revenues and GDP and tax revenues
accelerates the economic growth of African states. The
results also cinfirmed a positive and significant
relationship between foreign direct investment and
economic growth while there is a negative and
significant relationship between inflation and
economic growth.
In their paper, Egbunike et al., (2018) examined
the effect of tax revenue on economic growth of
Nigeria and Ghana during the period 2000-2016 (for
17 years). Researchers used multiple regressions as
tools of analysis. The results confirmed a positive
impact of tax revenue on the gross domestic product
(GDP) of Nigeria and Ghana.
3. Methodology
3.1. Research Hypotheses
The research hypotheses would be as follows:
1) Tax policies impact on economic growth of the
developed countries efficiently.
2) Tax policies impact on economic growth of the
developing countries efficiently.
3.2. Research Population
Research Population includes almost all Member
States of the United Nations. Published on November

18 / Tax Policy and Economic Growth in the Developing and Developed Nations
Vol.4 / No.14 / Summer 2019
4, 2010 by the Human Development Report Office of
the United Nations Development Programme (UNDP),
The Human Development Report is is an annual
milestone that displays the Human Development Index
in different countries. The states of the list include 167
(out of 192) UN members together with Hong Kong
and China. 24 Member States of the UN have been
excluded in this research due to lack of sufficient
information on the human development index.
Human Development Index is a relative estimation
by which the health dimensions, life expectancy, the
education dimension and, in general, the standard of
living dimensions are assessed. The assessment is
conducted based on the level of walfare among
children and minors and the results can be used to
evaluate the imoact of economic policy on standards of
living, as well as the degree of development. The HDI
was developed by Pakistani economist Mahbub ul
Haq, and an Indian economist Amartya Kumar Sen,
and was further used to measure the country’s
development by the United Nations Development
Program (UNDP).
All Member States (169) being studied in this
investigation are ranked into two groups including the
developing and developed countries, and random
selection method is applied to choose sample states
from each group.
1) 42 states are chosen from the developed
countries
2) 127 states are chosen from developing
countries
Sample population include 29 states (15 developed
countries and 14 developing countries) selected based
on the access to their data on the World Bank data
base. The current study covers the period from 2008 to
2016.
3.3. Model and Research Variables
The research model is designed as follows:
Y=AEα1 Dα2 Gα3 Cα4 Xα5 eut (1)
Where Y is GDP, A is coefficient, E is the ratio of
tax revenues to GDP, D is the ratio domestic credits
granted to the private sector, G is the ratio of
government expenditure to GDP, C is the capital stock,
X is the import and exports ratio to GDP, ut is a error
term. Expressed in logarithm form, the specification
can be rewritten as:
(2)
LYit=LA+α1LEit+ α2LDit+ α3LGit+ α4LCit+ α5LXit+ut
Where t is index of time-year and i stands as cross-
country.
4. Results
4.1. Descriptive Statistics
Descriptive statistics are brief descriptive
coefficients that summarize, classify and describe a
given data set. In fact, this type of analysis describes
the data and research information, and provides a
general plan or pattern of data for faster and better use.
Descriptive statistics provides information about
central parameters and the distribution of research
data. In general, descriptive statistics can be used to
describe the characteristics of a category of
information which in turn contributes to better
understanding the results of a test. Furthermore, it can
facilitate the comparison of the results of the test with
other tests and observations. The descriptive statistics
of the main variables of the model using Eviews
software are as follows.
Table 1 shows descriptive statistics of study
variables. Mean is the most commonly used measure
of central tendency representing the balancing point of
a data distribution and is a good indicator of the
centrality of the data. Another descriptive parameter is
the standard deviation, which indicates the dispersion
of the data. Also, the minimum and maximum
parameters in the table above show the range of data
variations. The median is the middle point of data, of
which half of the data is smaller and half larger than
the data.
Standard deviation (SD) is the most commonly
used measure of dispersion. It is the average squared
distance to the mean measuring the spread of data
about the mean. Skewness is a measure of the
asymmetry or symmetry of the probability distribution
of a real-valued random variable about its mean. If the
distribution is symmetric, then the mean is equal to the
median, and the distribution has zero skewness. For a
unimodal distribution, negative skew commonly
indicates that the tail is on the left side of the
distribution, and positive skew indicates that the tail is
on the right.
The kurtosis of any univariate normal distribution is 3.

International Journal of Finance and Managerial Accounting / 19
Vol.4 / No.14 / Summer 2019
Distributions with large kurtosis represent tail data
exceeding the tails of the normal distribution while
distributions with low kurtosis represent tail data that
is generally less extreme than the tails of the normal
distribution. Kurtosis represents the height of a
distribution and it measures maximum values in either
tail. For example, in the distribution t, where the
dispersion of the data is greater than the normal
distribution, the height of the curve is shorter than the
normal curve.
The Jarque–Bera test is applied to testing the
normality. The large value of Jarque–Bera indicates
that the distribution is more distanced from normal
distribution. A value of 0 Jarque–Bera indicates the
data is normally distributed. The data does not come
from a normal distribution, if Jarque–Bera value is
lower than the 5% significance level. If the value is
greater than 5%, the data are normally distributed.
Table 1. Descriptive statistics of study variables
Statistics
Country
LY
LE
LD
LG
LC
LX
Mean
Developed
27.17
2.79
0.09
-1.72
25.63
-0.16
Developing
26.53
2.84
-0.35
-1.89
25.40
-0.27
Median
Developed
26.84
2.68
0.18
-1.68
25.49
-0.17
Developing
26.50
3.18
-0.19
-1.47
25.58
0.18
Maximum
Developed
30.45
3.32
0.92
-1.32
28.85
1.48
Developing
29.88
7.55
3.89
2.88
29.06
4.70
Minimum
Developed
23.86
2.07
-2.09
-2.41
21.76
-1.47
Developing
22.76
-1.41
-4.63
-6.22
21.45
-4.31
Std.Dev.
Developed
1.62
0.33
0.63
0.22
1.68
0.79
Developing
1.77
2.24
2.30
2.32
1.76
2.20
Skewness
Developed
0.00
-0.12
-2.16
-1.09
-0.19
0.37
Developing
-0.35
-0.50
-0.40
-0.49
-0.27
-0.27
Kurtosis
Developed
2.82
1.71
7.97
4.64
2.74
2.27
Developing
2.62
2.38
2.18
2.28
3.01
2.39
Jarque-Bera
Developed
0.16
9.6
244
42.19
1.27
6.10
Developing
3.42
7.29
6.92
7.90
1.61
3.53
Probability
Developed
0.91
0.00
0.00
0.00
0.52
0.04
Developing
0.18
0.02
0.03
0.01
0.44
0.17
Sum
Developed
366
377
12.35
-233
3460
-21
Developing
3343
358
-44
-238
3201
-35
Observations
Developed
135
135
135
135
135
135
Developing
126
126
126
126
126
126
Source: Author‟s computation using E-views 8.0 (2016).
4.2. Unit Root Test (tets of stationarity)
Prior to the analysis of a time series, a unit-test
must be performed to find out whether the series is
stationary. Therefore, tets of stationarity or unit root
test is performed for model variables. The results are
analyzed using the Eviews software. Levin-Lin-Chu
unit-root test is performed for the developing and the
developed countries. The results for the panel unit root
test are presented below.
4.3. Unit root test analysis
Since the probability value (p value) for the unit
root tests for all countries is less than 0.05, it can be
concluded that series is stationary the unit root
hypothesis is rejected. So with a very low probability
of spurious regression, that provides misleading
statistical evidence of a linear relationship between
independent non-stationary variables, it can be
concluded that variables has no unit root, and the
regression model for countries can be made.
Table 2. The result of panel unit root test

20 / Tax Policy and Economic Growth in the Developing and Developed Nations
Vol.4 / No.14 / Summer 2019
Variable
Country
Statistics
Prob
Result
LY
Developed
-6.76
0.00
stationary
Developing
-8.76
0.00
stationary
LE
Developed
-3.10
0.00
stationary
Developing
-6.84
0.00
stationary
LD
Developed
-33.52
0.00
stationary
Developing
-6.48
0.00
stationary
LG
Developed
-4.57
0.00
stationary
Developing
-4.42
0.00
stationary
LC
Developed
-6.12
0.00
stationary
Developing
-4.42
0.00
stationary
LX
Developed
-15.75
0.00
stationary
Developing
-4.69
0.00
stationary
Source: Author‟s computation using E-views 8.0 (2016).
4.4. Estimation of Regression Model and
Hypothesis Testing
Using Eviews software, regression analysis is used
to test the hypothesis. To do so, first, regression
coefficients in a regression model were estimated and
the research model was tested based on interpretion of
the estimations. The model is defined as follows:
4.5. Model
(3)
LYit=LA+α1LEit+ α2LDit+ α3LGit+ α4LCit+ α5LXit+ut
Chow test or Limer's F-statistics is used to
determine which one of the pooled or panel models are
appropriate for estimating the regression models of
research. In studies of time-series data, before
estimating a model, test should be performed to select
fixed effects pattern against random effects pattern.
4.6. Testing the Pooled or Panel Model
Various tests are used in order to determine the
type of panel data model. The most general test is the
F-Limer test for using the fixed effects model against
the estimated model of pooled data.
Pooled data analysis is conducted when we have
time series of cross sections with the same y-intercept,
while Panel data refers to samples of the same cross-
sectional units observed at multiple points in time with
different y-intercepts. In the following, the F-limer test
will be checked to choose between pooling and panel
data.
Panel date models are divided into two sets of
assumptions: the random effects model and the fixed
effects model. In the fixed effects model, the
individual-specific effect is a random variable that is
allowed to be correlated with the explanatory
variables. In the random effects model, the individual-
specific effect is a random variable that is uncorrelated
with the explanatory variables. Therfeore, test should
be performed to select fixed effects pattern against
random effects pattern before estimating a model
.Chow test or Limer's F-statistics is performed to
determine which one of the pooled or panel models are
appropriate for estimating the regression models of
research.
4.7. F Limer Test
F Limer Test is used to choose between methods
of paneling and pooled data. In F Limer Test, H0 is
sameness of cross from origins (pooled data method)
that is situated in front of non-sameness of cross from
origins (paneling data method). If according to the
obtained result H0 is accepted, pooled data model is
the preferred method and combined regression model
(pooled) is statistically verified. Thus, research
hypothese are tested using pooled method. However, if
the H0 hypothesis is rejected, the panel data method is
accepted and the research hypotheses are tested using
the panel data method. The results of F Limer test
analyzed by EViews software are as follows:
Table 3. The result of F Limer test
Country
T-
statistic
Prob
Result
Developed
161.95
0.00
Panel model (with fixed or
random effects).
Developing
237/82
0.00
Panel model (with fixed or
random effects).
Source: Author‟s computation using E-views 8.0 (2016).
If possibility of F-test is less than significance
level of ≤0.05, null hypothesis for pooled data model
(regression without fixed or random effects) is
rejected. Therefor, the appropriate pattern to estimate
the models is associated with either fixed or random
effects and using pooled model can be rejected. As the
results indicate, the panel model data is used for all the
models.
4.8. Hausman Test

International Journal of Finance and Managerial Accounting / 21
Vol.4 / No.14 / Summer 2019
Now we turn to conducting the Hausman test to
see whether a fixed-effects or random effects model is
more appropriate for the data that we consider. When
the individual-specific effect is a random variable that
is uncorrelated with the explanatory variables, H0 is
confirmed. H1 is confirmed when model fixed-effects
model is appropriate and individual-specific effect is a
fixed variable that is correlated with the explanatory
variables. Using Hausman test, the panel data model
should be testet to reject or confirm fixed-effects or
random effects model. The Hausman test results are as
follows:
Table 4. Hausman Test Results
Country
T-
statistic
Prob
Result
Developed
161.95
0.00
The model does not have random
effects (it has fixed effects).
Developing
237.82
0.00
The model does not have random
effects (it has fixed effects).
Source: Author‟s computation using E-views 8.0 (2016).
As Table 4 shows, based on the calculated
probability value for the Hausman test, the p-value is
less than 0.05 and we find out that the random effects
model must be rejected to estimate the model.
Accordingly, the test results indicate that all models
have cross-sectional fixed effects for all the countries
included in the panel data.
4.9. Model Estimation
Estemiated model and coefficents for the developed
countries are as follows:
Table 5. Regression model for the developed countries
Variable
Coeff
Std.Err
t-stat
Prob
LE
-0.062195
0.071721
-3.509566
0.0007
LD
-0.049772
0.027485
-1.810892
0.0732
LG
-0.425644
0.058488
-7.277509
0.0000
LC
-0.132006
0.018854
7.001587
0.0000
LX
0.012356
0.015789
-0.782580
0.4357
C
24.16489
2.072057
11.66227
0.0000
AR(1)
0.986085
0.028928
34.08743
0.0000
R-squared
0.999964
Adjusted R-squared
0.999956
Durbin-Watson stat
1.940467
F-statistic
136136
Prob(F-statistic)
0.0000
Source: Author‟s computation using E-views 8.0 (2016).
4.10. Analysis of regression model for the
developed countries
The F-statistic and p value for general model are
136136 and 0.0000, respectively, indicating the
significant model (as p value of the statestics is less
than 0.05). The commonly used goodness of fit is the
coefficient of determination which is the square of the
correlation (r) between predicted y scores and actual y
scores; thus, it ranges from 0 to 1. If the the coefficient
of determination is close to 1, model fits the data well,
while the negative values will likely happen if R2 is
close to zero indicating that model does not fit the data
well. In the above table, the coefficient of
determination is 0.99, which indicates that the model
fits the data well. Testing the residuals from least
squares regressions, Durbin & Watson statistic values
in the range of 1.5 to 2.5.
4.11. Hypothesis testing for the developed
countries model
The first hypothesis of this study is as follows:
Tax policies impact on economic growth of the
developed countries efficiently.
Given that the coefficient of LE variable in the
model is significant (since p value is less than 5%),
this hypothesis is confirmed. In other words, if the tax
revenues increase by one percent, economic growth in
the developed countries is reduced by as much as 6
percent.
However there is no significant relationship
between LD variable and economic growth, there is a
negative and significant relationship between
government spending and economic growth. More
precisely, an increase in government expenditure by
1% would decreas economic growth by 42%. The LC
coefficient is also negative. This indicates that one
percent increase in the capital stock results in decrease
of economic growth by thirteen percent. Indeed, there
is no significant relationship between LX (import and
export) and economic growth. Hence, the estimated
regression model for the developed countries is as
follows:
LY= LA+ 24.16489 -0.062195 LE- 0.425644 LG –
0.132006 LC
Accordingly, estimating the alpha coefficients
(LA) for the developed countries, a regression model

22 / Tax Policy and Economic Growth in the Developing and Developed Nations
Vol.4 / No.14 / Summer 2019
for predicting GDP in sample countries would be
possible. The alpha coefficients for the sample
countries are presented in the table below:
Table 6. Alpha coefficient for the developed countries
Effect
COUNTRY
Row
Effect
COUNTRY
Row
1.48
Germany
8
1.37
Australia
1
-1.38
Ireland
9
2.33
United States
2
0.96
Singapore
10
1.83
Japan
3
1.57
Korea, Rep.
11
0.36
Argentina
4
-1.40
Luxembourg
12
-0.37
Norway
5
0.06
Spain
13
-0.88
Austria
6
-2.38
Greece
14
0.00
Netherlands
7
-3.56
Cyprus
15
Source: Author‟s computation using E-views 8.0 (2016).
4.12. Estimation of Model for Developing
Countries
Based on the estemiated model and coefficents for
developing countries are as follows:
Tabel 7: Regression model for developing countries
Variable
Coeff
Std.Err
t-stat
Prob
LE
-0.064069
0.037354
-1.715160
0.0897
LD
-0.118027
0.039297
-3.003460
0.0034
LG
-0.657112
0.039639
-16.57720
0.0000
LC
0.119555
0.023298
5.131659
0.0000
LX
-0.158067
0.028824
-5.483845
0.0000
C
22.99882
0.613588
37.48249
0.0000
AR(1)
0.942886
0.019815
47.58562
0.0000
R-squared
0.999890
Adjusted R-squared
0.999868
Durbin-Watson stat
2.019479
F-statistic
44110
Prob(F-statistic)
0.0000
Source: Author‟s computation using E-views 8.0 (2016).
4.13. Analysis of regression model for
developing countries
The F-statistic and p value for general model are
44110 and 0.0000, respectively, indicating the
significant model (as p value of the statestics is less
than 0.05). The commonly used goodness of fit is the
coefficient of determination which is the square of the
correlation (r) between predicted y scores and actual y
scores; thus, it ranges from 0 to 1. If the the coefficient
of determination is close to 1, model fits the data well,
while the negative values will likely happen if R2 is
close to zero indicating that model does not fit the data
well. In the above table, the coefficient of
determination is 0.99, which indicates that the model
fits the data well. Testing the residuals from least
squares regressions, Durbin & Watson statistic values
in the range of 1.5 to 2.5.
4.14. Hypothesis testing for developing
countries model
The second hypothesis of this study is as follows:
Tax policies impact on economic growth of
developing countries efficiently.
Given that the coefficient of LE variable in the
model is not significant (since p value is more than
5%), this hypothesis is rejected. In other words, there
is a not a significant relathionship between the tax
revenues and economic growth of developing
countries. There is a significant relationship between
the LD variable (the credits granted to the private
sectors) and economic growth. That is, if LD increases
by 1%, the economic growth in developing countries
decreases by 11%.
There is a negative and significant relationship
between government spending and economic growth.
More precisely, an increase in government expenditure
by 1% would decreas economic growth by 65%. The
LC coefficient is also signifiant. This indicates that 1%
increase in the capital stock results in decrease of
economic growth by 11%. Actually, there is negative
and significant relationship between LX (import and
export) and economic growth. An increase in import
and export of the developing countries gevornment by
1%, the ecponomic growth decreases by 15%
Hence, the estimated regression model for the
developed countries is as follows:
LY=LA+22.99882- 0.1180 LD – 0.6571LG + 0.1195
LC – 0.158 LX
Consequently, estimating the alpha coefficients
(LA) for developing countries, a regression model for
predicting GDP in sample countries would be possible.
The alpha coefficients for the sample countries are
offered in the table below:
Table 8. Alpha coefficient for the developed countries
Effect
COUNTRY
Row
Effect
COUNTRY
Row

International Journal of Finance and Managerial Accounting / 23
Vol.4 / No.14 / Summer 2019
-0.61
Peru
8
-1.03
Turkey
1
2.12
China
9
-0.05
Malaysia
2
-0.90
Egypt, Arab Rep.
10
1.35
Brazil
3
0.23
Indonesia
11
-1.86
Lebanon
4
0.37
Philippines
12
0.38
Thailand
5
1.43
India
13
-3.01
Mauritius
6
-0.43
Iran
14
0.72
Mexico
7
Source: Author‟s computation using E-views 8.0 (2016).
Model Fit Analysis
4.15. Linear Independence of Variables
(Variance inflation factor-VIF)
The correlation coefficient table was used to study
the existence or non-existence of lineraity between
independent variables. Since correlation coefficient
can be used when two predictor variables in a multiple
regression have a non-zero correlation, which is called
collinearity, variance inflation factor (VIF) test was
conducted to exam multicollinearity. The test results
analyzed by Eviews software are as follows:
Table 9. Variance inflation factor (VIF) test
Variable
VIF- Developed
Countries
VIF- Developing
Countries
LE
1.29
94.95
LD
1.16
1.78
LG
3.8
83.58
LC
3.17
2.31
LX
1.94
68.32
Source: Author‟s computation using E-views 8.0 (2016).
If the VIF exceeding 10, you can assume that the
regression coefficients are poorly estimated due to
multicollinearity. However, the VIF value less than 10
indicates low correlation among variables and it is
considered acceptable. Variance inflation factor (VIF)
test results revealed that nonlinear correlations among
variables for the developed countries but
multicollinearity exist among variables for the
developing countries which don’t fit well.
4.16. White’s test for Heteroskedasticity
The following are results of White’s test for
Heteroskedasticity:
Table 10. Heteroskedasticity test
Country
Chi-Sq.
statistic
Prob
Result
Developed
116.36
0.0000
There is no uniformity of
variance.
Developing
46.3
0.0000
There is no uniformity of
variance.
If the test statistics are greater than the critical
value then we reject the null hypothesis of constant
variance in favor of heteroscedasticity. The test
showed that the value of the F statistic is significant
(the p-value is smaller than 0.05). This means that
there is evidence of heteroscedasticity in the model, so
the null hypothesis is rejected. The following change
has been administrated to correct the model with
evidence of heteroscedasticity (this also has been done
for the previously estimated model): Specifying a
method for computing coefficient covariances, Cross-
section weights (PCSE) option has been selected form
the Eviews menue when the panel model is running.
This will change the method of calculating the
coefficient standard errors and thus the t-statistics and
p-values are corected for Heteroskedasticity.
4.17. Normality of the Model Residuals
The histogram plot of the model residual for the
developed countries is as follows:
0
2
4
6
8
10
12
-0.025 0.000 0.025 0.050
Series : Standardized Resi duals
Sam ple 2009 2016
Obs ervatio ns 120
Mea n -8.89e -11
Medi an -0.001302
Maxi mum 0.052186
Mini mum -0.039335
Std. De v. 0.016156
Ske wnes s 0.212552
Kurtos is 2.981501
Jarqu e-Be ra 0.905275
Proba bil ity 0.635949
Figure 1. Histogram plot of the model residual for the developed countries

24 / Tax Policy and Economic Growth in the Developing and Developed Nations
Vol.4 / No.14 / Summer 2019
The almost bell-shaped plot of the histogram in the
figure above and the Jarque–Bera statestics indicate
that the model gives an adequate fit to the data and the
typical assumption of normally distributed residuals is
satisfied (as the Jarque–Bera statestics is more than
0.05). Accordingly, the model prediction for the
developed countries is fitted to the data effectively and
the results are reliable.
The histogram plot of the model residual for the
developing countries is as follows:
Histogram plot of the model residual for
developing countries is almost bell-shaped, as it is
obvious in the figure above, and the Jarque–Bera
statestics indicate that the model gives an satisfactory
fit to the data and the typical assumption of normally
distributed residuals is contented (as the Jarque–Bera
statestics is not less than 0.05). Therefore, the model
prediction for the developing countries is adequetly
fitted to the data and the results are reliable.
0
2
4
6
8
10
12
-0.04 -0.02 0. 00 0.02 0.04
Series : Standardized Resi duals
Sa mple 2009 2016
Obs ervatio ns 112
Mea n -4.77e- 09
Medi an -1.35e-05
Maxi mum 0.054686
Mini mum -0.046075
Std. D ev. 0.023325
Ske wnes s 0.175074
Kurtos is 2.445457
Jarq ue-Be ra 2.007236
Proba bil ity 0.366551
Figure 2. Histogram plot of the model residual for the developing countries
5. Discussion and Conclusions
This study examined the impact of tax revenues on
GDP (economic growth) in the developed and
developing countries. The dependent variable is GDP
while tax revenue is independent variable for this
study and all these variables are expressed in logarithm
form. The control variables considered are: domestic
credits granted to the private sectors, on going
government expenditures, capital stock, and imports
and exports.The results of analysis showed that there is
a negative and significant relationship between the
logarithm of the ratio of tax revenues and GDP in the
developed countries; however, there is no significant
relationship between tax revenues and economic
growth in developing countries.The results also
showed no significant relationship existed between the
domestic credits granted to the private sectors and
economic growth in the developed countries, while
there is a significant and negative correlation between
the domestic credits granted to the private sectors and
economic growth in developing countries. The results
indicate that there is a negative ans significant
correlation between government expenditure and
economic growth in both developing and the
developed countries, although much stronger
correlation exists in developing countries.The
relationship between capital stock and economic
growth in the developed countries is a significant and
negative, while there is a positive and significant
relationship in developing countries. Finally, the
results revealed no significant relationship between
total sum of import and export in the developed
countris with economic growth, whereas this
relationship in the developing countries is negative and
significant.
References
1) Ahmad, S., Sial, M. (2016). Taxes and economic
growth: An empirical analysis of Pakistan.
European Journal of Business and Social Sciences,
5(2), pp. 16-29.
2) Asghari, R., Mohseni Zenouzi, S.J. (2013). The
Effect of Taxes and Government Consumption

International Journal of Finance and Managerial Accounting / 25
Vol.4 / No.14 / Summer 2019
Expenditures on Economic Growth in Islamic
Countries of the MENA Region. Journal of
Economic Development Research, 3(11), pp.1-22
3) Babatunde A., Ibukun, O., Oyeyemi, G. (2017).
Taxation revenue and economic growth in Africa,
journal of accounting and taxation, 9(2), 11-22.
4) Binti Saidin, N., Basit, A., Hamza, S. (2016). The
Role of Tax on Economic Growth. Internation
Journal of Accounting & Business Management,
4(2), pp.232-250.
5) Canicio, D., Zachary, T. (2014). Causal
relationship between government tax revenue
growth and economic growth: A case of
Zimbabwe. Journal of economics and sustainable
development, 5(17), pp 9-22.
6) Eugene, N., Chineze Abigail, E., (2016). Effect of
Tax Policy on Economic Growth in Nigria (1994-
2013). International Journal of Business
Administration, 7(1), pp. 50-58.
7) Faramarzi, A., Dastgib Farooja, M., Hakimipour,
N., Alipour, S., Jabbari, A. (2015). Investigating
the Relationship between Taxation and Economic
Growth, Case Study of Iran and OPEC and
Organization of Economic Cooperation (OPEC)
and (OECD). Journal of Economic Sciences,
9(32).pp. 103-122.
8) Göndör, M., Ö. Özpençe. (2014). An Empirical
Study on Fiscal Policy in Crises Time: Evidence
from Romania and Turkey. Procedia Economics
and Finance, 15.pp. 975-984.
9) Lee, Y., Gordon, R. H. (2005). Tax Structure and
Economic Growth. Journal of Public Economics,
89(5-6).pp. 1027-1043
10) Ocran, K. (2009). Fiscal policy and economic
growth in South Africa. Emerald Group
Publishing Limited, 38(5),pp. 604-618.
11) Ojong, M., Anthony O., Arikpo F. (2016). The
impact of tax revenue on economic growth:
Evidence from Nigeria. IOSR Journal of
economics and finance, 7(1).pp. 32-38.
12) Poulson, W., Kaplan, G., (2008). State income
taxes and economic growth. Cato journal,
28(1).pp. 53-71