International Journal of Scientific Research and Management (IJSRM)

||Volume||11||Issue||05||Pages||2783-2795||2023||
Website: www.ijsrm.in ISSN (e): 2321-3418
DOI: 10.18535/ijsrm/v11i05.el03

The Effect of Government Expenditure on Education on the

Enrollment Rate of Different Educational Levels in Selected OECD
Countries
Elnaz Hajebi1, Carol Billing2, Mahtab Hajebi3
1
Ferdowsi University of Mashhad
2
University of Idaho, College of Education, Health and Human Science
3
University of Central Florida

Abstract
One of the most important factors in economic, social, and cultural development is to create a
development plan to acquire knowledge, and govern the spirit of learning in the society, leading to the
strengthening of human capital, and increasing the quality of the educational system in accordance with
the needs of economic, and social development.
Regarding this, and also based on the recognition of the importance of education in economic growth,
economists, and economic development planners have paid special attention to the basics of resource
allocation in the education sector.
Therefore, according to the fact that government expenditure on education can be an effective factor to
encourage people to acquire knowledge, and improve their level of education, and on the other hand,
based on the fact that the improvement of people's level of education in the society shows the increase of
human capital in the country, which itself has a positive effect on the economic growth of the country, and
the inequality reduction, this paper aims to investigate the effect of government expenditure on education
on the enrollment rate in different educational levels in selected OECD countries from 2010 to 2019.
The results show that the effect of all coefficients of the variables is expected based on the theoretical
foundations, and the government expenditure on education has a significant positive effect on the
enrollment rate in the primary, secondary, and tertiary levels.

Keywords: Government Expenditure on Education, Enrollment rate, OECD, GDP

1. Introduction
Fair distribution of resources is one of the most important goals in modern economy. These distribution
policies are not only related to the distribution of the results, and the ultimate benefits of the economy (such
as income), but also include items, such as social security, and the development of equal opportunities for
progress. Therefore, programs, and policies related to education, health, social services, unemployment
insurance, retirement rights, and children's rights can also be considered as another part of policies that can
have a significant distributional role on the economic distribution in the society.

Many empirical evidence indicate the importance of education in economy. Education not only affects
human development, health improvement, and labor market conditions, but also affects the economic
performance of countries, so that when economy enters a knowledge-oriented stage, education, and human
capital play a critical role in the economic growth of the countries, because government's high expenditure
on education often has a significant impact on people's income, and economic growth, as well as poverty
reduction. (Coulomb, 2004)

Most of the empirical studies have investigated the effect of two important parts of the government
expenditure (education, and health expenditures) in the form of causal relationships, or their effect on other
indicators, but due to the use of statistical methods, and data from different countries they have had different

Elnaz Hajebi, IJSRM Volume 11 Issue 05 May 2023 [www.ijsrm.in] EL-2023-2783

Electronic copy available at: https://ssrn.com/abstract=4497453

results. On the other hand, the concerns related to human development indicators, whose manifestation can
be seen in the Millennium Development Goals of the United Nations, have prompted governments to
provide a major part of the expenditures related to these indicators.

Therefore, regarding the importance of education in economy, this research sought to investigate the effect
of government expenditure on education on the enrollment rate in different educational levels in selected
OECD1 countries using the panel data approach from 2010 to 2019, and the required statistics, and data were
extracted from the OECD Statistics website.

In the following, the theoretical foundations, and research background are presented in the second part. In
the third part, the research method is given. The fourth part deals with the estimation of the model, and in
the fifth part, conclusions, and suggestions are presented.

2. Theoretical foundations and literature review

2.1 Theoretical framework
The increase in the role of education in socio-economic dynamics has increased the need to pay attention to
the economic effects of education, and its cost-benefit analysis for the government, and people. The constant
changes in the technology, and the competitiveness of countries to achieve economic development, and
social welfare as quickly as possible, and the impact of these changes attracted more attention to the role,
and functions of public, and tertiary education. In this regard, it is necessary for the governments to pay
more attention to their education system, and allocate a significant budget to it to achieve the desired level of
development at the right time. Many experts in the field of human development believe that education is the
first, and most important step in planning human resources development, and the foundation of the
economic, and social development of any society is formed by skilled, and trained human resources. In other
words, training people by using their talent provide the basis for the growth, and development of the
country.

Education is a process during which knowledge, information, and cultural heritage, and traditions are
transferred to a person. In this process, individuals generally prepare themselves mentally for social life, and
in other words, they become sociable. In other words, the main axis of education is the expansion, and
evolution of human wisdom, science, information, and knowledge.

Education empowers the human through knowledge, training, skills, and smooth’s her/his personality,
attitudes, talents, and behaviors. It creates awareness, patience, self-esteem, and confidence in people
because of their rights. Education is one of the valuable factors to evaluate the level of a society (UNDP,
1990). There are several advantages of education, and thereby has positive effect on overall quality of our
life

Regardless of which aspect of development is considered, education is the root of all of the developments,
and it improves the quality of human skills, and talents, which in turn causes the formation of human capital,
and accelerates the process of economic, and social progress, and development of the society.

Achieving the millennium development goals will require serious attention to the efficiency, and
effectiveness of government expenditure on education, and health sectors. Examining the state of education,
and health from a macroeconomic point of view shows that the growth in human capital will increase the
return period of investment in human resources, and ultimately, the investment of the entire economy; as it
increases the years of exploitation of this resource, and the number of years that the investment has a
positive return. With the increase in life expectancy, private sector savings, and investment, as well as
foreign direct investment are encouraged, and a great step will be made toward the improvement of the
country's economic conditions. This improvement in the next stage will have a positive effect on the
improvement of health, and educational indicators. On the other hand, low health, and education costs will
reduce the rate of return on investment in human resources. This negative effect will be revealed first on the
profitability of investment in health, and education, and in the next stage on the whole economy, and a

1
Organization of Economic Cooperation and Development
Elnaz Hajebi, IJSRM Volume 11 Issue 05 May 2023 [www.ijsrm.in] EL-2023-2784

Electronic copy available at: https://ssrn.com/abstract=4497453

closed circuit will be formed related to the lack of investment growth in human capital. One of the most
important solutions for this problem is the government's involvement in providing the necessary costs to
improve the health, and education status in the society.

The organization of education systems varies between countries, as does the length of time for a student to
complete an educational level. This makes primary, and secondary programs more costly in some countries
than others. The length of the program thus affects the amount of educational investment per education
level, and does not necessarily reflect a country’s policy to place more importance on one part of the
education system than another.

On the same note, although participation in primary, and secondary education is very high in most OECD
countries, the proportion of students enrolled in tertiary programs varies considerably, which obviously
affects spending differentials across countries (see OECD Education at a Glance 2021 for more detailed
information).

All OECD countries invest a substantial proportion of national resources in education. In 2018, OECD
countries spent, on average, 3.4% of GDP on primary, secondary, and post-secondary non-tertiary
education, with 3.1% coming from public sources, and further 0.3% from private sources

In the same year, OECD average spending on tertiary education came to 1.4% of GDP, of which 1% came
from public sources, and 0.4% from private sources. On average, OECD countries spend, each year, USD
11 700 per student enrolled in education. However, the actual amount spent can differ substantially across
levels of education, with per-student spending often far higher for students in tertiary education than for
students in primary, and secondary education. In fact, in several OECD countries (Canada, Estonia,
Luxembourg, Mexico, Sweden, Turkey, the United Kingdom, and the United States), the average per-
student spend in tertiary education is at least double the average per-student spend in primary, secondary,
and post-secondary non-tertiary education.

However, benefits of education are more likely to have a long-term effect, not only within a country’s
growth, and development, but also with regard to social sector, and personal development. If investments on
education generate educated workers, they will be able to make other workers also more productive. Thus,
the overall level of education can benefit society because human capital spillovers are likely to increase
productivity above the direct effect on individual productivity (NBER, 2005). Therefore, the magnitude of
social returns to education is a crucial tool to assess the efficiency of public investment in education (NBER,
2005). Education rewards, not only the individual, but it also creates benefits that are shared by society at
large (NBER, 2005). Since there are evidences in regard to human capital increasing productivity, we can
say that education is productivity-enhancing.

Education affects both short run, and long run growth, although with a different degree of effectiveness. It is
estimated that one year more of education would increase the per capita output by between three, and six
percent (augmented neo-classical growth theory), and by an over one percentage point (new growth
theories) (Sianesi & Van, 2003).

2.2 Literature review

Gupta et al. (1997) in a research titled "The Efficiency of Government Expenditure: Experiences from
Africa" investigated the effect of government expenditure on education, and health in 38 African countries
from 1984 to 1995, and their results show that the government expenditure of African countries are less
efficient than Asian countries, and the improvement of access to education, and health needs to allocate
more funds to them.

McMahon (1999) finds a negative, and significant relationship between per pupil expenditures, and the
primary gross enrolment rate, and a positive, and significant impact of total education expenditure as a
proportion of GNP. The results of the McMahon study suggest that increasing primary education
expenditure while holding per pupil expenditures constant, has a positive, and significant impact on the

Elnaz Hajebi, IJSRM Volume 11 Issue 05 May 2023 [www.ijsrm.in] EL-2023-2785

Electronic copy available at: https://ssrn.com/abstract=4497453

primary gross enrolment rate. However, this study does not include income per capita as a separate
explanatory variable, and it may be the case that these resource variables are proxy for income per capita.

Bergh and Frink (2004) in a study titled "Tertiary education: Does Public Expenditure Increase Enrollment
rate?" investigated the effect of educational expenditure on the enrollment rate of students in tertiary
education. They conducted their study in 132 countries, and concluded that public spending on primary, and
secondary education has a positive impact on the enrollment rate in tertiary education; while the subsidies
paid to the tertiary education do not have a significant impact on the enrollment rate in tertiary education.
Therefore, it is recommended that more attention be paid to the allocation of public resources to education,
especially in less developed countries.

Anyanwu and Erhijakpor (2007) in a research titled "Education expenditures, and school enrolment in
Africa: Illustrations from Nigeria, and other SANE2 countries" investigated the effect of government
expenditure on education on school enrollment rate from 1990 to 2002, and their results show that
government expenditure on education has a significant positive effect on the enrollment rate in primary, and
secondary education, and among the SANE countries, the government expenditure on education in Nigeria
had the biggest positive, and increasing effect on the enrollment rate in primary, and secondary education. In
addition, in their study, they concluded that other variables such as political interventions such as
unification, and supported freedom, national income, and international charity associations have positive
impact on achieving the MDG3, and education expenditure alone cannot provide high-quality human capital
for this purpose.

Using panel data from 118 developing countries in 1971–2000, Baldacci et al. (2008) estimate a non-linear
model to capture the spending-outcome relationship. They account for the interaction between education,
and health, and control for governance, and the higher growth attributable to better human capital, and
country income levels. The fixed-effects model is utilized to make the most out of limited cross-country
time series data, and minimize distortions from heterogeneity. Baldacci et al. find strong evidence that
public expenditure on education directly results in increased better educational outcomes.7 However, the
positive effects of education spending are reduced in countries suffering from poor governance. The authors
further find that higher spending alone is insufficient; other policy interventions, such as improving
governance, and taming inflation, must be incorporated to achieve the MDGs.

Okezie A. et al. (2019) investigated Government education expenditure, and primary school enrolment in
Nigeria for the period of 1970-2017 using ARDL. It was observed that an insignificant relationship exists
between government education expenditure on primary school enrolment while a positive relationship exists
between remittances, and primary school enrolment. Population growth has positive relationship in the short
run, but a negative relationship in the long run. In this country, Primary schooling is managed, and received
funds mainly from the local governments (indirectly through deducting teacher salaries from their
entitlement from the Federation Account), and from the state governments. The result also shows that
parents/guardians play more role than the government in providing primary education in the country.

Idrees et al. (2021) conducted a study on analysis of the effect of government expenditure on school
enrollment rate in Pakistan. In this study data was used over the period of 2000 to 2017, and the method of
Least Square is applied to obtain empirical results. The result shows that national income, and government
expenditure have positive effects on school enrollment rate, indicating when national, and government
expenditure increase, ultimately school enrolment also increases in Pakistan.

Shafuda, and De (2020) examined the impact of public spending on human capital on human development
indicators such as health care outcomes, educational attainment, and national income growth in Namibia
using time series data from 1980 to 2015. The results were mixed, revealing a significant positive long-run
relationship of public spending on education with primary net enrollment rate, and tertiary gross enrollment

2
South Africa, Algeria, Nigeria, Egypt
3
Millennium Development Goals
Elnaz Hajebi, IJSRM Volume 11 Issue 05 May 2023 [www.ijsrm.in] EL-2023-2786

Electronic copy available at: https://ssrn.com/abstract=4497453

rate. In contrast, no cointegration was observed between public education spending, and primary, and
secondary gross enrollment rate.

Nenbee and Danielle (2021) carried out a research with the subject of the Primary School Enrolment, Public
Spending on Education, and Economic Growth in Nigeria. Using ARDL over the period of 1987-2017 they
recommend that there should be re-design of educational strategies by the government to include enrolment
campaigns, and alternative learning programs especially at the basic education level, and also increased her
budgetary allocation to education sector in line with UNESCO recommendation of about 26%.

2.3- Research hypotheses

- Government expenditure on education has a significant positive effect on the enrollment rate in primary
education.
- Government expenditure on education has a significant positive effect on the enrollment rate in secondary
education.
- Government expenditure on education has a significant positive effect on the enrollment rate in tertiary
education.

3. Research methodology
Given that the purpose of this research was to investigate the effect of government expenditure on education
on the enrollment rate in different educational levels in selected OECD countries4; the effect of government
expenditure on education on the enrollment rate was investigated by panel data, and Anyanwu and
Erhijakpor (2007) model in primary, secondary, and tertiary education in the form of three econometric
models according to equations 1, 2, and 3, respectively. The data used is for the time period ......., and on an
annual basis.

(1)

(2)

(3)
Lenrp: Logarithm of education (primary) enrolment;
Lenrs: Logarithm of education (primary) enrolment;
Lenrt: Logarithm of education (primary) enrolment;
β1 : Regional/Country-specific effect;
Ledexpp: Logarithm of government expenditure on primary education as percent of GDP;
Ledexps: Logarithm of government expenditure on secondary education as percent of GDP;
Ledexpt: Logarithm of government expenditure on tertiary education as percent of GDP;
Ldemoc: Lograrithm of democracy index;
Lpopr: Logarithm of population rate;
Lgdpper: GDP per capita in international dollars;
Uit: Error term

4
List of Countries Included in the sample used in the estimations: Austria, Australia, Canada, Chile, Denmark, Finland, France,
Germany, Iceland, Korea, Mexico, Norway, Sweden, Switzerland, Turkey, USA, UK.
Elnaz Hajebi, IJSRM Volume 11 Issue 05 May 2023 [www.ijsrm.in] EL-2023-2787

Electronic copy available at: https://ssrn.com/abstract=4497453

Cross-sectional data, time series, and panel data consisting of time series or the combination of cross-
sectional data, and time series are used for analyzing the relationship between economic variables
statistically, and econometrically. The functional form of panel data econometrics is as follows;

t= 1,2, … N
t= 1,2, … N

As it can be seen in the functional form of panel data econometrics, t shows the time, and i shows the
sections. In this equation, an individual effect exists. This effect cannot be observed by independent
variables, does not change depending on time, but includes characteristics peculiar to sections (Baltagi,
2005). In panel data econometrics, the next step after converting cross-sectional data, and time series data to
the panel system is to determine if the cross-section, and period effects can be explained by the fixed effects
model or the random-effects model. The fixed effects model creates a different fixed value for each cross-
sectional unit. In the fixed effects model, it is assumed that the slope coefficients do not change, but fixed
coefficients show differences among only cross-sectional data or time data or among both types of data. If
the differentiation occurs only depending on time, these types of models are named as one-way time
dependent fixed effect models. If a differentiation occurs in panel data depending on both time, and section,
these models are named as two-way fixed effects model. However, because the cross-sectional effect is
generally investigated more in panel data studies, panel data models are generally considered as one-way
models (Hsiao, 2002). One-way, and two-way fixed effects models can be seen in the Eq. (2), and (3) given
below:

In this equations, it is considered that the error terms are distributed independently, and identically in such
that their variances equal to zero. In the fixed effects model, the fixed effects estimator allows the fixed
constant to differ across cross-section units by estimating different constants for each cross-section (Baltagi,
2005). The changes that occur depending on cross-sections or both cross-sections, and time are observed
when they are integrated into the model as a component of the error terms. The advantage of random effects
model over the fixed effects model is that, without loss of degree of freedom, the random effects model
allows the inclusion of the effects that are out of the sample to the model. The functional relation for the
mentioned models can be demonstrated as follows:

4. Experimental findings
According to Granger, and Newbold (1974), a regression analysis between the variables does not provide
reliable results in case non-stationary data is used. For this reason, stationary should be checked before
performing the regression analysis. The studies conducted by Levin, and Lin (1992, 1993), Breitung, and
Meyer (1994), Quah (1994), Maddala, and Wu (1999), Hadri (2000), and Im et al. (2003) suggest the use of
unit root tests in panel data models.

Contrary to what is customary in the case of time-series data, in the case of panel data, Dickey-Fuller, and
augmented Dickey-Fuller tests cannot be used to test the reliability, but it is necessary to test the collective
reliability of the variables. Although the time course of this research is not very long, the stationary test of
the variables was performed first to prevent the development of spurious regression for the estimation of the
models. The most common tests used for this purpose were Levin, Lin, and Chu, Im, Pesaran, and Shin,
Breitung, and Hadri. In the first three mentioned tests, the null hypothesis is based on non-stationarity, and
Elnaz Hajebi, IJSRM Volume 11 Issue 05 May 2023 [www.ijsrm.in] EL-2023-2788

Electronic copy available at: https://ssrn.com/abstract=4497453

in Hadari's test statistic, the null hypothesis is based on stationarity. In this research, the test of Levin et al.
(2002) was used, which is more applicable to examine the reliability of variables in panel data, and the tests
were examined using EViews 9. If the calculated P-value is less than five percent, the hypothesis of a single
root for that variable is rejected. The results of the stationarity test of variables according to Levin et al.
(2002) method is given in Table (1). Given that the probability values are less than 0.05, H0 related to the
existence of a single root is rejected at 95% confidence level, and except for the Lpopr (population rate)
variable, all the variables of the model are stationary at the I(0) level5.
Table1- Unit Root Test
Stationary test Differentiation Significance Test
Variable
Levin, Lin and Chu order I(d) level statistic

Lenrp Individual intercept, and trend I(0) 0.000 -6.7873

Lenrs Individual intercept, and trend I(0) 0.000 -10.417

Lenrt Individual intercept, and trend I(0) 0.0004 -3.386
Ledexpp Individual intercept, and trend I(0) 0.000 -73.704
Ledexps Individual intercept, and trend I(0) 0.001 -3.0838
Ledexpt Individual intercept, and trend I(0) 0.000 -26.625
Lgdpper Individual intercept, and trend I(0) 0.0412 -1.7365
Lpopr Individual intercept, and trend I(1) 0.000 -7.8941
Ldemoc Individual intercept, and trend I(0) 0.0013 -3.0132

4.1. Cointegration test

Cointegration is the study of the stability of long-term correlations between non-stationary time series
variables. Cointegration approach is a solution to spurious regression problem in time series. Cointegration
is when two, or more time series variables are related based on a theoretical foundations to form a long-term
balancing relationship, although these time series may have a random trend (would be non-stationary), they
follow each other well over time, so that the difference between them is stationary. Therefore, the concept of
cointegration implies the existence of a long-term balancing relationship towards which the economic
system moves over time. (Noferesti, 1999, p. 76). Since panel data may be non-stationary, cointegration,
checking its presence, and testing it in this type of data is also very important. Like stationarity test,
cointegration test in panel data is stronger than cointegration tests for individual cross-sectional units,
because these tests can be used even in situations where the time period is short, and the sample size is
small. (Baltagi, 2005, p. 252) In this regard, in this part of the research, cointegration test in panel data is
discussed.

According to the results of the stationarity test in Table (1), the Lpopr variable is not at a significant level,
and with one-time differentiation it becomes stationary. To investigate the presence, or absence of spurious
regression, cointegration test is performed before model estimation. In this study, Pedroni's test was used to
examine the cointegration between research variables. The results of this test are presented in Table (2).

Table (2) shows the results of Pedroni's cointegration test for panel data in the three models of this study.
The significance of most of the statistics indicates that the null hypothesis of this test, that there is no
cointegration relationship, is rejected, and the existence of cointegration relationship between the variables
is confirmed, and it can be said that there is a long-term balancing relationship between the enrollment rate,
and the independent variables.

Table 2- results of Pedroni's cointegration test for panel data in the three models

5
It was investigated in all three states of Individual Intercept, Individual Intercept and Trend, and None and became stationary
at Individual Intercept and Trend.
Elnaz Hajebi, IJSRM Volume 11 Issue 05 May 2023 [www.ijsrm.in] EL-2023-2789

Electronic copy available at: https://ssrn.com/abstract=4497453

 Intraclass Interclass
Model 1 Weighted
Statistic Prob. Prob. Statistic Prob.
Statistic
Group rho-
Panel v-Statistic -1.059849 0.8554 -1.638188 0.9493 4.190175 1.0000
Statistic
Group PP-
Panel rho-Statistic 2.704958 0.9966 3.140876 0.9992 -26.79305 0.0000*
Statistic
Group ADF-
Panel PP-Statistic -4.008542 0.0000* -7.960458 0.0000* -9.046550 0.0000*
Statistic
Panel ADF-Statistic -4.006446 0.0000* -4.903564 0.0000*
Intraclass Interclass
Model 3 Weighted
Statistic Prob. Prob. Statistic Prob.
Statistic
Group rho-
Panel v-Statistic 3.847851 0.0001* -0.649301 0.7419 4.245036 1.0000
Statistic
Group PP-
Panel rho-Statistic 3.384512 0.9996 3.499733 0.9998 -10.14775 0.0000*
Statistic
- Group ADF-
Panel PP-Statistic 0.0000* -6.336743 0.0000* -3.976031 0.0000*
8.880119 Statistic
-
Panel ADF-Statistic 0.0009* -2.263396 0.0118
3.107731
Intraclass Interclass
Model 2 Weighted
Statistic Prob. Prob. Statistic Prob.
Statistic
Group rho-
Panel v-Statistic 5.369321 0.0000* 4.377392 0.0000* 3.489290 0.9998
Statistic
Group PP-
Panel rho-Statistic 2.214401 0.9866 2.794959 0.9974 -4.832892 0.0000*
Statistic
Group ADF-
Panel PP-Statistic -3.299969 0.0005* -2.398313 0.0082* -1.465612 0.0714
Statistic
PanelADF-Statistic -2.691923 0.0036* -0.970711 0.1658

* They are significant at 95% confidence level.

4.2. Selection of the model type

Before fitting the models, three tests were used to determine the appropriate model:

• In the first stage, F-test was used to distinguish between the integrated model (data integration), and
panel data;

• Then, if there was panel data, in the second step, the type of fixed, or random effects was
determined by the Hausman test;

• If the effect of the model was random in the second stage, the Breusch-Pagan test was used to
choose between the integrated model, or the random effect.

4.2.1. Chow test or Limer's F-statistics

Limer's F-statistics was used to choose between panel data, and integrated data methods. In this test, H0
indicates the equality of the intercepts (integrated data), and the opposite H1 indicates the heterogeneity of
the intercepts (panel data).

Elnaz Hajebi, IJSRM Volume 11 Issue 05 May 2023 [www.ijsrm.in] EL-2023-2790

Electronic copy available at: https://ssrn.com/abstract=4497453

 Integrated data methods
Panel data methods (fixed effects)
If the calculated P-value is greater than 0.05% error level, the null hypothesis is not rejected, and the
estimation is performed by integrated methods model; otherwise, panel data with fixed effects are used.

Table 3- confirmation of the panel data model with fixed effects

value p-value Test result
Model 1 11.38 0.000 Rejection of the integrated model, and confirmation
of the panel data model with fixed effects
Model 2 27.46 0.000 Rejection of the integrated model, and confirmation
of the panel data model with fixed effects
Model 3 112.69 0.000 Rejection of the integrated model, and confirmation
of the panel data model with fixed effects

Since the F-statistics is significant in the three studied models, with a probability level of more than
99%, H0 is rejected, and panel data with fixed effects can be used to estimate the model.

4.2.2. Hausman Test

In the second step, the Hausman test was used to determine which of the fixed, and random effects models
could be expressed, and investigated. The null hypothesis of the Hausman test indicates the use of the
random effects method, and the opposite hypothesis indicates the use of the fixed effects method.
If the P-value is greater than 5% error level, the H0 is not rejected, and the random method will be used;
otherwise, H1 is confirmed, and fixed effects method is used.

Table 4- confirmation of the panel data model with random effect

x2 -value p-value Test result
Model 1 1.59 0.8099 Rejection of the panel data model with fixed
effect, and confirmation of the panel data model
with random effect
Model 2 23.91 0.0001 Confirmation of the panel data model with fixed
effect
Model 3 2.35 0.6718 Rejection of the panel data model with fixed
effect, and confirmation of the panel data model
with random effect

4.2.3. Breusch-Pagan test

Since in the previous stage random effect model was determined for the three models according to the
Hausman test, Brioche-Pagan test was used for the final model selection between integrated data, and panel
data with random effects, the result of which is summarized in Table (5) for the three studied models. The
null hypothesis of this test indicates the use of integrated data method, and the opposite hypothesis indicates
the use of panel data with random effect.

Table 5- final model selection between integrated data, and panel data with random effects
x-2 value p-value Test result
Model 1 172.08 0.0000 Rejection of the integrated model, and
confirmation of the panel data model with
random effects
Model 2 In model 2, fixed effects were selected
Model 3 209 0.000 Rejection of the integrated model, and
confirmation of the panel data model with
random effects

Elnaz Hajebi, IJSRM Volume 11 Issue 05 May 2023 [www.ijsrm.in] EL-2023-2791

Electronic copy available at: https://ssrn.com/abstract=4497453

 If P-value is greater than 5% error level, the null hypothesis is not rejected, and the estimation is
performed by the integrated methods model; otherwise, panel data with random effects are used.

4.3. Modified Wald test

In the second model, where the fixed effects were determined, to investigate the heterogeneity of variance,
the modified Wald test is used, and the null hypothesis implies the homogeneity of variance. It can be seen
in Table (6) that H0 is rejected with a probability level of 99%, and Model 2 does not have homogeneity of
variance.

Table 6- heterogeneity of variance

LR X2 -value p-value Test result

Model 2 6094.6 0.0000 Variance heterogeneity

4.4. Likelihood-ratio (LR) test

Generally, when the number of periods studied in panel data is greater than the number of time periods, the
probability of variance heterogeneity is increased. Therefore, the heterogeneity of variance in the models of
this study was analyzed by LR test.

Table 7- Likelihood-ratio (LR) test

Model LR X2- value p-value Test result

Model 1 27.46 0.0000 Variance heterogeneity

Model 2 136 0.0000 Variance heterogeneity
Model 3 139.08 0.0000 Variance heterogeneity

4.5. Model Estimation

The required tests were investigated, and the estimation results of models 1, and 3 using the random effect
panel data method, and model 2 using the fixed effect panel data method are summarized in the following
tables:

Table 8- Models Estimation

Variable/Model 1 Coefficient z statistic P-value
The logarithm of government expenditure on
0.0150* -4.81 0.0000
education at primary level, Ledexpp
The logarithm of population growth rate, Lpopr 0.0187* 6.83 0.0000
The logarithm of GDP per capita, constant 2015,
0.0288* -3.29 0.001
lgdpper
Ldemoc, democracy index 0.0536* 3.39 0.001
Intercept 2.1527* 50.98 0.0000
Wald X2 77.85 0.0000
Variable/Model 2 Coefficient z statistic P-value
The logarithm of government expenditure on
0.098* 2.86 0.004
education at secondary level, Ledexps
The logarithm of population growth rate, Lpopr 0.053* 5.12 0
Logarithm of GDP per capita, constant 2015,
0.121* 3.61 0
lgdpper
Ldemoc, democracy index 0.138* 2.9 0.004
Intercept 1.463* 9.34 0
Wald X2 123.62 0.000
Elnaz Hajebi, IJSRM Volume 11 Issue 05 May 2023 [www.ijsrm.in] EL-2023-2792

Electronic copy available at: https://ssrn.com/abstract=4497453

 Variable/Model 3 Coefficient z statistic P-Value
The logarithm of government expenditure on
0.399* 8.27 0.0000
education at tertiary level, Ledexpt
The logarithm of population growth rate, Lpopr -0.039*** -1.59 0.1130
The logarithm of GDP per capita, constant 2015,
0.325* 4.81 0.0000
lgdpper
Ldemoc, democracy index 0.212** 1.8 0.0720
Intercept 0.265 0.84 0.3980
Wald X2 160.13 0.000
* They are significant at 95% confidence level.
** They are significant at 90% confidence level.
* They are significant at 85% confidence level.

In the following, the sign of the coefficients, and its compatibility with the expected theoretical foundations
is interpreted. As seen in Tables 1, 2, and 3, 1% increase in government expenditure leads to an increase of
0.01% in the Enrollment rate in primary level, 0.09% in the Enrollment rate in secondary level, and 0.39%
in the Enrollment rate in tertiary level. On the other hand, one percent increase in the population growth rate
has caused an increase of 0.01%, and 0.05% in the enrollment rate in primary, and secondary levels,
respectively, and a decrease of 0.03% in the enrollment rate in tertiary level. In addition, one percent
increase in GDP per capita has led to an increase of 0.02%, 0.12%, and 0.32% in the enrollment rate in
primary, secondary, and tertiary levels, respectively. One percent increase in the democracy index also leads
to an increase of 0.05%, 0.13%, and 0.21%, respectively, in the enrollment rate in primary, secondary, and
tertiary levels.

Given that logarithm is taken of the model variables, the coefficients also represent elasticity. Accordingly,
the elasticity of the enrollment rate in tertiary level is higher than that of the enrollment rate in other levels
with respect to each of the independent variables (except the population rate). The elasticity of the
enrollment rate in secondary level to government expenditure is higher than that of the enrollment rate in
primary level to the same variable. This issue indicates that to increase the enrollment rate in higher levels,
increase in government expenditure is leveraged.

5. Conclusion and policy suggestions

The purpose of this research was to investigate the effect of government expenditure on education on the
Enrollment rate in different educational levels in selected developed OECD countries. According to the
findings, the research hypotheses were confirmed.

According to the findings, in the three estimated models, the sign of the coefficients is in accordance with
the expected theoretical foundations, and the variable of government expenditure on education has a direct
relationship with the enrollment rate in each of the primary, secondary, and tertiary levels.

Regarded to the positive effect of GDP per capita in the three educational levels, it can be said that as GDP
per capita is a standard to show the economic power of countries, and their people, with the increase of GDP
per capita, these countries will better meet the educational needs, and since the purchasing power of people
increases with the increase of this variable, the enrollment rate increases in the three levels, and given that
the coefficient of this variable is higher in the tertiary level than the secondary, and primary levels; it can be
said that people's interest in enrolling in tertiary level is much higher than the changes in GDP per capita,
because the higher the income, and purchasing power of people, the more capable, and interested they will
be in tertiary level. In relation to the positive effect of urban population on the Enrollment rate in both
primary, and secondary education, it can be said that as with the increase in population, the possibility of
gathering people increases in larger groups, and in certain places, and the possibility of providing services
for people becomes easier, and the number of people who enjoy these services increases, the positive effect
of this variable on the enrollment rate is not far from expected at different education levels, and this is also
confirmed by other researchers.

Elnaz Hajebi, IJSRM Volume 11 Issue 05 May 2023 [www.ijsrm.in] EL-2023-2793

Electronic copy available at: https://ssrn.com/abstract=4497453

In relation to the positive effect of government expenditure on education on the Enrollment rate in the three
educational levels, since people and the economy have realized that the government expenditure in the
education sector is a capital expenditure that will have long-term benefits for the economy, and individuals,
with the increase of the government expenditure in the education sector, people are very motivated, and
interested to enroll in different educational levels, especially in tertiary education. So, in general, it can be
concluded that government expenditure on education, and GDP per capita, and population have played an
important role in improving the status of the investigated educational indicators (enrollment rate in different
education levels). Therefore, according to the results of the economic policy-makers to achieve the goals of
the third millennium, and better development indicators, it is recommended that appropriate policies be
developed to increase the GDP per capita, and the purchasing power, and the ability of people to provide the
necessary expenses for education, especially in tertiary education. In addition, since the population growth is
inevitable over time, and due to the positive effect of this variable on the enrollment rate, it is suggested that
the policy-makers help the country to achieve better development indicators using appropriate plans in the
field of controlling, and guiding the immigrant population to the cities by instilling an appropriate
educational culture; in addition to reducing the negative aspects of the issue. In addition, it is recommended
that government expenditure be increased in the education sector, as according to the findings, and the
basics of macroeconomics, with the increase in government expenditure on education, and the subsequent
increase in the enrollment rate in different educational levels; the growth of human capital increases, and the
period of investment return in human power, and finally the total investment in the economy will increase,
and this increase in investment will improve the economic conditions of the country, and this chain of
improvement will have a positive effect on the improvement of all educational indicators, as with the
reduction of government expenditure on education, which is considered as a long-term investment, the
enrollment rate in different levels, and then the efficiency rate of human resources will decrease, and this
negative effect will put the economy in a closed circle of lack of investment growth in human capital that
one of the most important solutions to get out of this problem is the government intervention in providing
the necessary expenses to improve the educational situation.

This paper also found that democracy has a significant positive effect on the enrolment rate in primary,
secondary, and tertiary levels. Needless to say that the level of democracy, and government expenditure on
education in OECD countries is high enough to be translated into human capital accumulation, and
economic growth.

References
1. Anyanwu, J.C., and Erhijakpor, A.E.O. Education expenditures, and school enrolment in Africa:
Illustrations from Nigeria, and other SANE countries. Economic Research Working Paper. 2007. 92.
2. Appiah, Elizabeth N. The Effect of Education Expenditure on Per Capita GDP in Developing
Countries. International Journal of Economics, and Finance; 2017. 9(10); 136-139.
3. Breitung, J., Meyer W. Testing for Unit Roots in Panel Data: Are Wages on Different Bargaining
Levels Cointegrated?. Applied Economics.1994. 26. 353-361.
4. Baldacci, E., Clements, B., Gupta, S., and Cui, Q. Social spending, human capital, and growth in
developing countries. World Development. 2008. 36(8);1317-1341.
5. Baltagi, B. H. Econometric Analysis of Panel Data. New York: John Wiley, and Sons. 3th edition.
2005.
6. Bergh, A, and Frink, G. Tertiary education: Does Public Expenditure Increase Enrollment rate.
Working Paper. 2004. 84.
7. Gupta, S,. and Honjo, K,. and Verhoeven, M. The Efficiency of Government Expenditure:
Experiences from Africa” IMF Working Paper.1997. 97(153).
8. Di Lupidio, V. How does government expenditure on education, and specific reforms in education
affect economic growth? Evidence from OECD countries, and policy implications. Faculté des
sciences économiques, sociales, politiques et de communication, Université catholique de
Louvain.2019.
9. Education at a Glance, OECD Indicators, OECD Publishing, Paris, (2021).
10. Granger, C.W.J., Newbold, P. Spurious Regressions in Econometrics. Journal of Econometrics.
1974. 2; 111-120.

Elnaz Hajebi, IJSRM Volume 11 Issue 05 May 2023 [www.ijsrm.in] EL-2023-2794

Electronic copy available at: https://ssrn.com/abstract=4497453

 11. Hadri, K. Testing For Stationarity in Heterogeneous Panel Data. The Econometrics Journal. 2000.
3;148-161
12. Hajebi, E., & Razmi, M.J. Effect of Women Higher Education on Economic Development in Some
OPEC Countries and North Africa, Journal of Economic Modeling Research. 2016. 6 (24).175-200.
13. Idrees M., Khan, F. and Omar Fauzee, M.S.B. Analysis of the Effect of Government expenditure on
School Enrollment rate in Pakistan, Responsible Education, learning, and Teaching in Emerging
Economies. 2021. 3(1); 27-35.
14. Im, K.S., Paseran M.H., Shin Y. Testing For Unit Roots in Heterogeneous Panels, Journal of
Econometrics. 2003. 115. 53- 74.
15. Levin, A., Lin C. Unit Root Tests in Panel Data: New Results. University of California, San Diego
Working Paper. 1993. 56- 93.
16. Maddala, G. S. ,Wu, S. A Comparative Study of Unit Root Tests With Panel Data and A New
Simple Test. Oxford Bulletin of Economics and Statistics, Special Issue. 1999. 631-652.
17. McMahon W. Education, and Development: measuring the social benefits, Oxford: Oxford
University Press.1999.
18. Noferesti, M. Unit Root and Co-integration on Econometrics, Rasa Publication. 1999.
19. Naiz, M. M. Human capital development, and economic growth in Bangladesh. Journal of world
economic research. 2018. 52-63.
20. Nenbee, S.G. and Danielle, I. E. Primary School Enrolment, Public Spending on Education, and
Economic Growth in Nigeria, Mediterranean Journal of Social Sciences. 2021. 12(5);103-113.
21. Okezie A., I., Joseph C. U., and Sandralyn, O. Government education expenditure, and primary
school enrolment in Nigeria: An impact analysis, Journal of Economics, and International Finance.
2019. 11(3); 24-37.
22. Quah, D. Exploting Cross-Section Variations For Unit Root Inference in Dynamic Data. Economic
Letters, Volume: 44, Issue. 1994.1(2); 9- 19.
23. Quiggin, J. Human Capital Theory And Education Policy In Australia. Australian Economic
Review.1999. 32(2); 130-144.
24. Shafuda, Christopher P. P., and Utpal Kumar De. Government expenditure on human capital, and
growth in Namibia: A time series analysis. Journal of Economic Structures.2020. 9(21).
25. Sianesi, B,. & Van, R. The Returns to Education: Macroeconomics, Journal of Economic Surveys.
2003.17. 157-200.
26. Villela, R, and Paredes, J.J. Empirical Analysis on Public Expenditure for Education, Human
Capital, and Economic Growth: Evidence from Honduras. Economies. 2022. 10(241).
27. OECD Education Database, https://www.oecd.org/education/database.htm
28. OECD Social Expenditure database, www.oecd.org/els/social/expenditure OECD Education at
a Glance 2021: OECD Indicators, OECD Publishing, Paris
29. http://epp.eurostat.ec.europa.eu/portal/page/portal/education/data/
30. UNESCO International Standard Classification of Education, www.uis.unesco.org/isced

Elnaz Hajebi, IJSRM Volume 11 Issue 05 May 2023 [www.ijsrm.in] EL-2023-2795

Electronic copy available at: https://ssrn.com/abstract=4497453