Content uploaded by Ebaidalla Mahjoub Ebaidalla
Author content
All content in this area was uploaded by Ebaidalla Mahjoub Ebaidalla on Feb 09, 2018
Content may be subject to copyright.
African Journal of Economic Review, Volume VI, Issue I, January 2018
160
Understanding Household Education Expenditure in Sudan: Do Poor and Rural
Households Spend Less on Education?
Ebaidalla M. Ebaidalla
27
Abstract
This paper examines the factors that influence households’ expenditure on education in Sudan, using
the National Baseline Household Survey (NBHS) data (2009) for national, urban and rural levels. The
results of Tobit model indicate that household income, head education, head age, household size,
number of school-age children and residing in urban areas are the most significant factors affecting
education expenditure. Interestingly, the results show that the income elasticity of education in urban
sample model is greater than that of rural model, implying that households residing in urban areas are
likely to spend more on education. In addition, the effect of household income is found to be positive
and significant in the highest income quintile. Overall, the results revealed that households with
higher income, whose heads are educated and reside in urban areas tends to spend more on education
compared to poor and rural households. These results signify the lack of inter-generational
educational and income mobility in Sudan, implying that children from poor households are caught
permanently in low income and educational levels, and are not able to “catch up” their peers in high
income families.
Keywords: Education Expenditure, Tobit models, Sudan
JEL Classification: I21, I22, I24, C24
27
Assistant Professor of Economics, University of Khartoum,Email: ebaidallamahjoub@yahoo.com
African Journal of Economic Review, Volume VI, Issue I, January 2018
161
1. Introduction
Education has been considered as a key factor for supporting economic growth and
development and alleviating poverty in developing countries. According to human capital
theory, education allows individuals to gain better skills and knowledge needed to access
jobs, hence enhances productivity and economic growth; which in turn help in eradicating
extreme poverty and hunger (Bryant, 1990; Becker, 2009 Mincer, 1970 and Schultz, 1961).
Therefore, the issue of education expenditure by both governments and households has
gained a sizable attention from researchers and international development organizations.
In Sudan, the education system has been affected by many economic transformations that the
country undergone in the last three decades. Specifically, the adoption of liberalization and
free market policies in early 1990s have resulted in reducing public spending on education.
Since then, the size of private investment in education has expanded remarkably.
Accordingly, households’ expenditure on education has gone up although basic education
such as, primary and secondary education, is still delivered through public sector. Moreover,
the reduction of government expenditure on education has contributed greatly in lessening the
quality of public education; hence a large segment of population is pushed into private
education. This leads to a significant increase in household education expenditure,
particularly in urban areas and among high income households.
Against this backdrop, many questions can be raised in accordance with the aims of this
study, including: What are the key determinants of the households’ education expenditure in
Sudan? Does the poor and rural household spend less than urban and rich household? To
what extent could the factors that affecting education expenditure vary across rural and urban
areas as well as among different categories of income groups?
Regarding the importance and policy relevance, the empirical investigation to be undertaken
by this study is useful for several reasons. First, investigating household education
expenditure is crucial to provide evidence, which can be used to formulate relevant policies
targeting planning and reforming education system in Sudan. Second, understanding the
factors that affecting educational spending in Sudan may help policymakers and key
stakeholders (i.e. national and international NGOs) to design effective strategies that ensure
better access to education so as to create more jobs and reduce poverty. Finally, by
identifying the factors affecting education expenditure among different areas (i.e. urban and
rural) and income quintiles, the study would place strong foundation in designing effective
education programs for disadvantaged groups of population.
The remainder of this paper is organized as follows. The next section outlines some stylized
facts about education system and its finance in Sudan. Section three discusses the theoretical
and empirical literature on the determinants of household educational spending. While
section four outlines data and research methodology, section five presents the empirical
results and discussions. Section six ends with a conclusion and possible ways forward.
2. Education in Sudan: An Overview
Gaining its independence in January 1956 from British colony, Sudan inherited an education
system designed to provide civil servants and professionals to serve the colonial
administration. The distribution of education facilities such as, teachers and enrollment was
biased in favor of the needs of the British administration and Western curriculum. Thus, the
African Journal of Economic Review, Volume VI, Issue I, January 2018
162
education services were clustered in urban cities, although about 70 percent of population
resides in rural areas. However, at that time the education was fully sponsored by government
and the public expenditure on education was about 20 percent (Nour, 2012). Most of
education during the colonial era was focused on the basic education (i.e. primary,
intermediate and secondary), while tertiary education was limited to the University of
Khartoum. In addition, a few number of students of wealthier parents received secondary and
university education abroad.
After the independence, the education system in Sudan has received considerable attention
from national governments. The national education policies concentrated on the target of
achieving universal and compulsory education with aim of equitable distribution of facilities
among urban and rural areas. Therefore, the education system has experienced a significant
change in terms of years of schooling and distribution of schools. For instance, the Nimeiri
regime (1969) considered the education system as inadequate for the needs of social and
economic development, hence reorganized the education system in the 1970s as a result
(Elmagboul, 2014). The basic education system was changed from 4-4-4 to 6-3-3 (6 primary
years compulsory, 3 year for intermediate and 3 for secondary). The technical and vocational
education also has gained more attention during Nimeiri government. Moreover, during the
era of 1970s the tertiary education has expanded by establishment two new universities (i.e.
University of Juba and University of Gaziera) in addition to oldest one: the University of
Khartoum.
During the 1980s, Sudan underwent a remarkable expansion in basic education with the
opening of hundreds of primary and secondary schools, despite economic and political
instability. The technical and vocational education also increased remarkably. All these
efforts led to a significant increase in the rate of enrollment from 1980.
In early the 1990s, the education system in Sudan witnessed a great transformation. First, it
was further reorganized into eight years of primary education followed by three years
secondary schooling. In addition, Arabic language was adopted as instruction language in all
public universities. Moreover, the tertiary education expanded and more than thirty
universities were established. The number of private schools grew rapidly following
economic policies lifting government subsidies to service sectors, including education.
Regarding financing education in Sudan, the country inherited a tax-based education system
from the British colony, in which the state provides free educational services for the entire
population. Thus, successive national governments adopted free education and this continued
until the adoption of free market policies in the decade of the early 1990s. However, after the
implementation of the Structural Adjustment Program (SAP), the government began its
sudden withdrawal from the provision of educational services. The austerity measures
adopted in 1992 resulted in a great reduction in public spending on education. To fill the gap
in financing education resulting from these policies, the government provided licenses to
private schools. In line with this system, parents were requested to pay some fees for public
schools in order to utilize education.
To understand the contribution of government in education, Table 1 below presents the public
spending on education in Sudan and a sample of Sub-Saharan African countries. The table
shows that public spending on education in Sudan is accounted for a small proportion from
the country’s GDP compared to other countries in the sample.
African Journal of Economic Review, Volume VI, Issue I, January 2018
163
Table 1: Public Education Expenditure (% of GDP) in Sudan and a sample of SSA countries
Country
1990-1999
2000-2009
2010-2014
Angola
2.6
2.7
3.5
Botswana
6.3
9.7
9.6
Cameroon
3.1
3.0
3.1
Cote d'Ivoire
4.8
4.1
4.7
Ethiopia
2.6
4.6
4.5
Ghana
4.1
6.0
6.9
Kenya
6.0
6.3
5.5
South Africa
5.8
5.0
6.0
Sudan
1.0
1.8
2.1
Uganda
2.5
3.6
2.5
Source: World Bank, World Bank Indicator (2016)
Table shows that Sudan has the smallest public education spending ratio to its GDP compared
to other SSA countries in our sample. Specifically, the government expenditure on education
(% of GDP) remained rotating around 1 percent during 1990-1999. During 2000-2009, it
increased to the rate of 1.8 percent, indicating the expansion in education expenditure, which
may be due to oil revenue at such a period. Moreover, during the last period (2010-2014) the
spending on average progressed to 2.1 percent. However, in all periods, the public spending
on education in Sudan lags far behind the levels of public expenditure in SSA countries.
Regarding the contribution of government education spending to the total public spending,
Table 2 below presents data on public spending on education as a percentage of total
government expenditures for Sudan and a sample of SSA.
Table 2: Public Education Expenditure (% of Total Government Expenditure) in Sudan and a
sample of SSA countries
Public Education expenditure (% of Total Government Spending)
Country
1990-1999
2000-2009
2010-2014
Angola
6.1
6.9
8.7
Botswana
20.0
24.3
21.0
Cameroon
11.6
18.7
15.7
Cote d'Ivoire
19.0
21.9
20.7
Ethiopia
14.0
20.6
26.7
Ghana
15.0
22.3
27.9
Kenya
24.0
25.0
20.6
South Africa
20.0
19.4
19.2
Sudan
9.1
8.9
11.0
Uganda
10.0
14.8
11.5
Source: World Bank, World Bank Indicator (2016)
Table 2 indicates that Sudan has the second lowest percentage of public education spending
(percentage of total government spending) after Angola. For instance, during the period
(1990- 1999), Kenya holds the highest rate of public spending on education, which is about a
threefold of that of Sudan. The low rate of public educational spending as a percentage of
GDP and total government expenditure implies low public investment in education in Sudan.
This also indicates that public education spending falls below the standardized international
African Journal of Economic Review, Volume VI, Issue I, January 2018
164
adequacy criterion, which was earlier adopted in the 1960s and related to the supply side and
implies the allocation of either 8 percent of GDP on education or 20 percent of total
government or public spending on education (Nour, 2013). The reduction in government
spending on education resulted in a significant deterioration in efficiency indicators like
education attainment and enrollment.
Regarding the demand for education, Table 3 presents the gross enrolment ratio for the three
educational levels, primary, secondary and tertiary, respectively
28
.
Table 3: Gross Enrolment Ratio by Educational Level in Sudan and a Sample of SSA Countries
(%)
Primary level
Secondary Level
Tertiary Level
1990-
1999
2000-
2009
2010-
2014
1990-
1999
2000-
2009
2010-
2014
1990-
1999
2000-
2009
2010-
2014
Angola
18.4
115.4
85.7
11.6
18.9
28.8
0.6
2.3
8.4
Botswana
15.2
17.0
16.9
55.0
77.5
83.3
5.3
10.6
21.4
Cameroon
10.5
16.6
29.1
25.9
30.0
51.6
3.6
6.1
11.5
Cote d'Ivoire
1.7
2.9
5.3
24.3
25.6
40.1
4.6
9.1
8.5
Ethiopia
1.4
2.3
14.2
11.5
24.2
35.7
0.8
2.6
7.4
Ghana
83.7
75.0
113.7
37.5
47.3
60.9
1.2
6.9
13.5
Kenya
36.9
48.1
67.4
38.5
48.1
67.6
3.1
South Africa
26.1
42.1
75.8
79.6
88.9
92.9
13.1
16.3
19.3
Sudan
16.2
23.6
34.3
33.2
36.2
39.3
6.3
11.0
15.4
Uganda
9.1
11.4
12.4
10.4
21.0
80.8
1.5
3.5
4.2
Source: World Bank, World Bank Indicator (2016)
The table shows that the enrolment ratio for primary education in Sudan was close to some
African countries that belonging to poor income group like Angola. However, the primary
enrolment ratio falls below some of SSA counties like Kenya and Ghana. Regarding the
secondary enrolment ratio, Sudan also has a lower rate compared to some SSA countries like
Botswana, Ghana and Kenya. The low enrolment ratio in primary and secondary education in
Sudan may be attributed to poverty and economic instability. During the period under
consideration, the tertiary enrolment ratio in Sudan has the second highest ratio during all
periods after South Africa. This high tertiary enrolment ratio may be due to expansion in
tertiary education over the last three decades.
Regarding the educational attainment, Figure 1 below shows the average years of total
schooling in Sudan and a sample of SSA countries
29
. As indicted from the Figure, Sudan has
the lowest rate of educational attainment among other SSA countries in comparison. The low
level of educational attainment confirms the relatively low level of school enrollment. This
also supports the high rate of illiteracy in Sudan, which is about 26 percents in 2013 (World
Bank, 2013). However, there are many factors that may be held responsible for low
educational attainment, including the high cost of education, poverty and unemployment. In
general, the low rate of educational attainment and enrollment indicates low commitment to
28
Gross primary or secondary school enrolment ratio - The number of children enrolled in a level (primary or secondary),
regardless of age, divided by the population of the age group that officially corresponds to the same level (World Bank,
2016).
29
Educational attainment refers to the highest level of schooling that a person has reached. Here we use average years of
total schooling as calculated by Baroo and Lee (2010).
African Journal of Economic Review, Volume VI, Issue I, January 2018
165
the standardized international adequacy and equity criterions in the demand side as measured
by the lack of adequacy in enrollment rate in primary, secondary and tertiary education and
literacy rate of population (Nour, 2013).
Figure 1: Education Attainment of population aged 15 and older in Sudan and a sample of SSA
countries (average of years of total schooling)
Source: World Bank, World Bank Indicator (2016)
3. Literature Review
Given the importance of education in economic growth and development, the determinants of
household educational expenditure have gained a considerable attention from both
researchers and policy makers in last decades. However, most of the exiting literature has
focused on the macroeconomic perspective and government expenditure on education. On the
other hand, the issue of household' expenditure on education has gained a few attention
particularly in developing countries. In this section we briefly review some empirical studies
on this issue.
The empirical literature indicates that household education expenditure is influenced by many
variables, including household characteristics, parents’ education level and household
income, among them. However, the main consensus among most of empirical studies is that
household income is the most significant factor affecting education expenditures (e.g.
Hashimoto and Heath, 1995; Panchamukhi, 1965); and Kothari, 1966 and Tilak, 2002).
Huston (1995) analyzed the impact of income and household characteristics on education
expenditure in US. Using a sample from the 1990-1991’ Consumer Expenditure Survey, he
found that head age, education level, income, region, race, and family size are the most
significant factors affecting household education expenditure.
Kanellopoulos and Psacharopoulos (1997) investigated the factors that affect private
expenditure on education in Greece, using household Expenditure Survey of 1988. They
found that household size and number of children under six years of age have negative effect
on private spending on education, while the head’s years of education and income have a
positive impact on education expenditure. In the same vein, Psacharopoulos and
Papakonstantinou (2005) examined the household expenditure on university education in
Greece, using a sample of 3000 university freshmen. They argued that private education is
African Journal of Economic Review, Volume VI, Issue I, January 2018
166
highly inelastic, indicating its importance in Grecian household budget. They also found that
private out of pocket spending to prepare for the entrance exams and study at college exceeds
that of public spending. In addition, they found that poorer families spend a higher share of
their income on education of their children. Moreover, using data from household surveys for
1990 and 1992, Psacharopoulos et al. (1997) examine the extent of private expenditure on
education in Bolivia and calculate an income elasticity of 0.23. They conclude that education
expenditure is not a luxury good for Bolivian families.
Tilak (2002) studied the household education expenditure in rural India using the national
survey on Human Development in rural India (HDI) (1994). The paper also examines the
household expenditure on education by different groups of population. He found that there is
nothing like free education in India and household expenditure on education represents a
considerable portion of household budget. In addition, households from lower socio-
economic background and low income groups spend considerable amounts on acquiring
education, including specifically elementary education, which is expected to be provided free
to all by the State. His results also indicate that household income, educational level of the
head of household and the household size are among the most significant factors affecting
education expenditure. Interestingly, he found that education is income inelastic in India by
compiling time series of household expenditure estimates over the period 1960-61 and 1984–
85.
Glewwe and Jacoby (2004) examined the relationship between household resources and
demand for education in Vietnam using household panel survey data covering the period
1993-1998. They found a positive relationship between household income and demand for
education, even after controlling for locality-specific factors such as change in education
returns, supply and quality of schools, and opportunity costs of schooling.
Tansel and Bircan (2006) studied the demand for private tutoring in Turkey, using household
expenditure survey (1994). Adopting tobit model, the authors showed that private tutoring is
neither a luxury nor a necessity item in a household’s budget. They also found that parents’
educational level, especially of mothers have positive and significant affect on private
tutoring expenditures, which means inequity in the intergenerational distribution of
education. Moreover, the results indicate that private tutoring expenditures increase at a
decreasing rate with the age of the household head, hence implying lifecycle considerations.
Their results also indicate that urban families spend more than rural household residents.
Finally, household private tutoring expenditures are found to be declined with the number of
children in the household.
Qian et al. (2011) examined parents’ expenditure on their children’s education, using
household survey data from 32 selected cities across China in 2003. Their results show that
household income has significant effects on both domestic and overseas educational
expenditures. The results also indicate that households whose mothers have secondary school
or college education and fathers who are working in professional occupations are likely to
spend more on education. Moreover, their study found that households belonging to the
highest income group, with a college-educated father, a mother who is a cadre or middle
professional and living in coastal areas, are most probable to spend on children’s education
abroad.
African Journal of Economic Review, Volume VI, Issue I, January 2018
167
Sulaiman et al (2012) examined the determinants of household expenditure on education in
Malaysia. Using household survey data, they found that household characteristics such as
parents’ income and educational level, mother’s work status, job category of head of
household and parents’ awareness of globalization in respect of their children’s education are
the most significant factors affecting education expenditure. Specifically, their results show
that the elasticity of income is very high (approximately 1 percent) indicating the importance
of household income in education expenditure.
Vu Quang (2012) investigated the factors that affecting household expenditure on children’s
education in Vietnam. Using the Vietnamese Household Living Standards Survey (VHLSS
2006) and adopting tobit model, he found that household income has a positive and
significant effect on household education expenditure. Meaning, increase in the income of the
household is always associated with an increase in educational expenditure. His result also
revealed that households whose heads have a higher level of education or with professional
jobs are more likely to spend more on education. Moreover, households with more primary-
school-age or secondary school-age children are likely to spend more on education compared
to households with pre-school-age or college-age children. Vu Quang shows that families
with more resources and better human capital are those who are able to spend more income
on their children’s education.
Andreou (2012) investigated the determinants of household education expenditure in Cyprus,
using expenditure surveys of 1996-1997, 2002-2003 and 2008-2009. He found that the level
of education expenditure increases with income across years. In addition, his results pointed
out that household income, number of children in household, region of residence and heads’
age and education level are the most important factors affecting the level of household
expenditure on education.
Recently, Acar et al (2016) using Turkish household budget surveys from 2003, 2007 and
2012, investigated the determinants of household education expenditures, adopting an Engel
curve framework. In particular, they estimate Tobit regressions of real educational
expenditures by income groups to examine if and to what extent the determinants of
educational expenditures differ by income groups. Their results indicate that the estimated
expenditure elasticity is low for the top- and the bottom-income quartiles while it is high for
the middle-income quartiles. The results also show that for all income groups the expenditure
elasticity of education increases over time, indicating that Turkish household allocates greater
share of their budgets to education expenditures.
The above discussion has made clear that there is a dearth of empirical studies on household
educational expenditure in Africa in general and Sudan in particular. Therefore, this study
would contribute to the existing literature by examining the factors that affecting household
health expenditure across national, urban and rural areas. Moreover, unlike the previous
studies, this paper emphasizes the role of income and regional disparities in household
educational expenditure.
4. Data and Methodology
4.1 Data and Variables
The data used in this study is sourced from the national baseline households survey (NBHS)
conducted by the Central Bureau of Statistics in 2009. The survey contains data on all
household' expenditures (e.g. food, education, health, utility, etc..) as well as demographic
African Journal of Economic Review, Volume VI, Issue I, January 2018
168
and socio-economic characteristics of households and individuals. The survey comprises
48825 individuals of 7913 households and covering 15 states. However, information on
education expenditure for each individual in household is not exist, thus, we use household as
a unit of the analysis. The data include expenditure of the household in past 12 months (year).
Following previous studies (e.g. Qian and Smyth, 2010 and Quang, 2012) we focus on
households with dependent children of age is not older than 22, as most of households'
members are graduated from university by that age. Accordingly, there was 7,257 valid
households who hold such criteria
30
. Therefore, we ensure that there is no sample selection
problem, because most of the households with children have positive education expenditure.
Based on the literature review discussed in the previous section, the dependent variable in our
analysis is household education expenditure on education. The dependent variable is
explained by a vector of explanatory variables, which include household income and socio-
economic characteristics. The socio-economic characteristics include a set of variables that
are hypothesized to influence household education expenditure such as, household size,
education level of head of household, gender, age of the head of household, marital status and
dummy variables indicating region of residence, and occupation. Regional and seasonal
factors are also considered. The definition and descriptive statistics of the variables used in
the analysis is presented in table 4.
4.2 Estimation Technique
To analyze the factors affecting household education expenditure, this paper uses tobit model,
which is appropriate technique to estimate household expenditure with zero observations
(Tobin, 1958). That is, because not all the households spend on education services, numerous
zero observations will exist in the data and we are facing with the so-called censored sample
problem (Barslund, 2007; Czarnitzki and Stadtmann, 2002; Dardis et al., 1994). The tobit
model was originally developed by Tobin in 1958 to accommodate censoring in the
dependent variable. This model also overcomes the bias associated with assuming a linear
functional form in the presence of such censoring. The tobit model considers that all zeros are
attributable to standard corner solutions. Negative values of the dependent variable are
assumed to be exist but are considered to be unobservable and bunched at zero. Based on the
Tobin’s model, it is assumed that a latent variable that measures the consumer’s propensity to
spend money on education (yh) is related to the vector of explanatory variables (Xh) and
undetectable influences, as specified in the following:
It is assumed that a household h spends ( ) on education if the latent variable ( ) is
positive. In contrast to the observed expenditure of household h ( ), the value of the
unobservable value ( ) can be negative. Negative values of the latent variable imply that
household will not spend any money on education:
30
The study does not discriminate between private and public education expenditure because there is no
information on the type of schooling and/or education expenditure in the NBHS' data.
African Journal of Economic Review, Volume VI, Issue I, January 2018
169
The conventional estimators for these types of models are based on maximum likelihood
estimation (MLE). The MLE produces consistent estimates of the parameters of the tobit
model, under appropriate assumptions such as, homoscedasticity and normality of the error
terms. The likelihood function consists of two parts: the product of the probabilities that
households do not spend any money on education [Pr ( )] and the product of the
probabilities that households spend on education [Pr ( )]:
Assuming standard normal distributed errors ( , the likelihood function of censored model
can be rewritten using a probability density function ( ) and cumulative distribution function
( ) of the standard normal distribution as (Tobin, 1958):
Equation (3) will be estimated via the maximum likelihood (ML). The estimation is run for
different samples, namely full, urban and rural household samples, as well as for different
household income groups.
5. Empirical Results and Discussion
This section presents the empirical results and discussions. First, we present some descriptive
statistics about the variables that used in the analysis and then report the econometric results.
5.1 Descriptive Statistics
Before analyzing the factors influencing household educational expenditure in Sudan, it is
useful to present some descriptive statistics. Thus, table 4 below describes the definition and
mean as well as the standard deviation of variables employed in the analysis.
As can be read from the table, the reported statistics indicate that the mean of total household
income is SDG 6,846 per annum. This is somewhat consistent with the national statistics as
reported by NBHS (2009). However, the higher standard deviation of the total income point
to the prevalence of income inequality in Sudan. The mean of health expenditure is about
SDG 472 per month, representing about 17 percent out of non-food expenditure. This
suggests that a considerable portion of Sudanese households’ income is spent on education.
The standard deviation of household education expenditure is also high, indicating a great
disparity among households in terms of educational expenditure.
The table also indicates that the average of gender variable is very high (about 90 percent),
indicating the dominance of males in heading households. Regarding the mean and standard
deviation of education variables, the table shows that most heads of households and spouses
have low levels of educational attainment, confirming the widespread illiteracy in Sudan.
African Journal of Economic Review, Volume VI, Issue I, January 2018
170
Table 4: Summary Statistics of Variables used in the Analysis
Variable
Definition
Mean
Std. Dev.
Education Expenditure
Household expenditure on education
472.501
4644.570
Income
Household total disposable income in SDG
6846.134
24416.660
Household's Head Characteristics
Age
Age of head of household in years
45.811
14.810
Gender of head
Gender of the head of household (1 = male; 0 =
female)
0.896
0.305
Education level of Household head
Primary
Primary school, dummy
0.192
0.394
Secondary
Secondary school, dummy
0.078
0.268
University
University, dummy
0.042
0.201
Education level of
Spouse
Spouse Primary
Primary school, dummy
0.191
0.393
Spouse Secondary
Secondary school, dummy
0.070
0.255
Spouse University
University, dummy
0.032
0.176
Number of children in household
Pre-school
The number of children aged 1 to 6 living in the
household.
0.967
1.046
Primary school
The number of children aged 6 to 14 living in the
household
1.500
1.512
Secondary school
The number of children aged 15 to 17 living in the
household
0.404
0.628
University level
The number of children aged 18 to 22 living in the
household
0.971
1.099
Profession of Household's Head
Agriculture
A dummy variable where 1 =household’s head
being engage in agricultural activities, 0 otherwise.
0.072
0.258
Industry
A dummy variable where 1 =household’s head
being engage in industrial activities, 0 otherwise.
0.003
0.053
Service
A dummy variable where 1 =household’s head
being engage in industrial activities, 0 otherwise.
0.925
0.262
Household Type of Dwelling
House
A dummy variable where 1 = being a resident in
house, 0 otherwise.
0.995
0 .068
Apartment
A dummy variable where 1 = being a resident in
apartment, 0 otherwise.
0.006
0.108
villa
A dummy variable where 1 = being a resident in
villa, 0 otherwise.
0.005
0.126
Household Characteristics
Household size
Number of household' members
6.173
2.806
Room
Number of rooms
3.265
1.869
Married head
A Dummy variable, (1= married; 0= unmarried)
0.895
0.306
Electricity
A Dummy variable, (1= electrified; 0= un-
electrified)
0.391
0.488
African Journal of Economic Review, Volume VI, Issue I, January 2018
171
Moreover, as can be read from the table, the mean of number of heads engaging in agriculture
and industry is very small, while the mean for service activity is very high. This implies that a
considerable portion of household income is generated from service activities, confirming the
dominance of the service sector in Sudan’s economy. Moreover, as can be fairly read from
the table, the average household size is about six persons, which is consistent with the 2009
NBHS. Interestingly, the mean of dummy variable (married) is high, implying that most
household heads are married. Finally, the mean of electricity is found to be relatively small,
demonstrating the weakness of infrastructure in Sudan, particularly in rural areas.
5.2 Econometrics Results
A. Determinants of Household' Health Expenditure
First, the results of tobit estimation of equation (3) for the full, urban and rural sample are
presented in Table 1 in Appendices. As can be observed from the table, most of the variables
carry their expected signs and in line with the theory. The result reveals that the coefficient of
household income is positive and significant in all estimated models. However, the results
show some differences in income coefficients across models, indicating variations in terms of
income impact on education between regions. For instance, the elasticity of income is higher
in urban sample compared to rural sample. This result indicates that households residing in
urban areas spend about 6 percent more on children’s education than those living in rural
areas. This result suggests that urban households devote a considerable portion of their
budget to children’s education. This can be explained by the fact that the extremely poor
quality of education in the country, led most of urban households to switch their children to
private institutions, which supply better educational services than their public counterparts.
On a national level, an increase in household income by a 1 percent elevates its education
spending by 8.4 percent. This strong association between household income and education
expenditure indicates the absence of free provision of education in Sudan. Alternatively
stated, due to the withdrawal of government from financing education, households are
pressed to cover education spending relying on their own resources. Furthermore, quality
deterioration of public schools pushes a considerable part of the population to private
institutions.
Regarding the household head characteristics, the results show that age of head has positive
and significant impact of education expenditure. This result confirms many previous
empirical studies (e.g. Suliaman, 2012 and Andreos, 2012). Also, the coefficients of
education level of head and spouse are found to be positive and significant in full, urban and
rural sample models. This means that a household whose head received university degree or
diploma is likely to spend the more on their children’s education. This result indicates that
educated heads and mothers are likely to spend more in education. This finding is in line with
the previous studies of Acar (2016) and Vu Quang (2012).
The number of secondary school and university age children has positive and significant
impact on education expenditure. This implies that household with children in high education
institutions tend to spend more on education compared to those with more children in low
education levels. In addition, household head who engage in the service activities tends to
spend more in education compared to those participating in agricultural activities. This is
because most of service activities are located in urban areas, where households have higher
opportunity to spend more on education compared to rural households who engage in
agricultural sector.
African Journal of Economic Review, Volume VI, Issue I, January 2018
172
Moreover, the results show that the coefficients of household size, number of room and
access to electricity have positive and significant impact on household education expenditure
in Sudan. This can be justified by that fact that larger household with urbanized facilities
tends to spend more in education. This finding is also confirms the positive and significant
coefficient of urban dummy variable, which indicates that households residing in urban areas
tend to spend more in education than those living in rural area.
In terms of geography, households residing in the Northern, eastern, central and Kurdofan
regions are likely to spend less on their children’s education than households residing in the
capital city (Khartoum). This confirms the fact that households in Khartoum devote a large
investment for their children’s education. Expectedly, the coefficient of Darfur region is
found to be negative but not significant. This finding can be justified by the fact that people
of Darfur suffer from civil war and a large portion of Darfur population live in IDP camps
and spend nothing in education, as most of education services provided by government and
non-governmental organizations.
Overall, households with higher income and residing in urban areas tend to spend more on
education of their children. This finding confirms our hypothesis that rural and poor
household spend less in education in Sudan. In addition, households whose head and mother
have higher education level are likely to invest more on education.
Regarding the factors affecting education expenditure by income quintile, Table 2 in
Appendices reports the marginal effects for the tobit estimates. As can be read from the table,
the coefficient of household income in the bottom four income quintiles are insignificant. On
the other hand, the effect of household income in the highest (fifth) income quintile is found
to be positive and statistically significant. This indicates that households belonging to high
income quintile are likely to spend more on children education. This result confirms the
previous results of full, urban and rural models. This also implies that children’s education is
an important investment for rich population. However, the result suggests that an increase in
income of household that belonging to low income quintiles does not raise the education
expenditure, as poor households devote a greater part of their budget to food and health
expenses.
Similar to the results obtained from the full, urban and rural samples, the education level of
household head is found to be very significant in influencing household expenditure,
particularly for the highest income group. This finding supports the previous analysis that
household with higher income and educated head tends to spend more on education than poor
and less educated heads. In addition, the number of secondary and university-age children
increases household education expenditure in both fourth and fifth quintile. In addition,
households whose head is working in service sector and belonging to third and fourth income
quintile spend more on education compared to other income quintiles. Moreover, the results
shows that households reside in other regions than Khartoum spend less. Finally, the
coefficient of Darfur is not significant, confirming the pervious analysis.
6. Conclusion and Policy Implications
This paper examines the factors influencing household educational expenditure, with
emphasis on the role of household income. The study used the NBHS data (2009) for
national, urban and rural levels and employed a tobit model. For further understanding of the
African Journal of Economic Review, Volume VI, Issue I, January 2018
173
impact of income on children’s education, the analysis is executed for different income
groups.
The results of the tobit estimation reveal that household income, heads’ educational level,
heads’ age, household size, number of school-age children and residing in urban areas are the
most significant factors affecting educational expenditure in full, urban and rural samples of
the surveyed households. Interestingly, the empirical results show some variations between
the effects of household income on educational expenditure across urban and rural areas.
Specifically, the income elasticity of education in the urban sample model is greater than that
of the rural model, implying that household resides in urban areas tends to spend more on
education than rural households. In addition, the effect of household income is found to be
positive and significant in the highest income quintile, implying that rich households tend to
spend more than poor households.
Overall, our results indicate that households with higher income, residing in urban areas tend
to spend more on education in Sudan. In addition, household whose head and mother have
higher education level are likely to spend more on education than the others. These results
signify the weakness of inter-generational educational and income mobility in Sudan. This
also suggests that children from poor household are caught permanently in low income and
low education levels and are not able to “catch up” their peers of high income families.
Accordingly, education policies in Sudan need to take into account the equality of
opportunity in education to ensure that children from low education families have as much
access to education as their richer counterparts; thus leading to higher intergenerational
mobility in Sudan. Accordingly, liberalization of education that adopted in 1992 should be
revised with cautions so as to achieve income and educational equality.
References
Acar, E.Ö., Cilasun, S.M. and Günalp, B. (2016), 'An Analysis of Education Expenditures in
Turkey by Income Groups', Economic Research Forum, Working Paper Series, No.
991, Cairo, Egypt.
Andreou, S.N. (2012), 'Analysis of Household Expenditure on Education in Cyprus', Cyprus
Economic Policy Review, Vol 6, No. 2, 17-38.
Becker, G.S. (2009), Human capital: A theoretical and empirical analysis, with special
reference to education, University of Chicago press, Cicago, USA.
Bryant, W. (1990), The Economic Organization of the Household, Cambridge University
Press, Cambridge, UK.
Czarnitzki, D. and Stadtmann, G. (2002), 'Uncertainty of outcome versus reputation:
Empirical evidence for the First German Football Division', Empirical
Economics, Vol. 27, No. 1,101-112.
Dardis, R., Soberon-Ferrer, H. and Patro, D. (1994), 'Analysis of leisure expenditures in the
United States', Journal of Leisure Research, Vol. 26, No. 4, 309.
African Journal of Economic Review, Volume VI, Issue I, January 2018
174
Donkoh, S.A. and Amikuzuno, J.A. (2011), 'The determinants of household education
expenditure in Ghana', Educational Research and Reviews, Vol. 6, No. 8, 570- 579.
Glewwe, P. and Jacoby, H.G. (2004), 'Economic growth and the demand for education: is
there a wealth effect?' Journal of Development Economics, Vol. 74, No. 2, 33-51.
Hashimoto, K. and Heath, J.A. (1995), 'Income elasticities of educational expenditure by
income class: The case of Japanese households', Economics of Education
Review, Vol. 14, No. 1, 63-71.
Huston, S.J. (1995), 'The household education expenditure ratio: Exploring the importance of
education', The Journal of Family Economics and Resource Management Biennial
division of the AAFCS, Vol. 1, 51-56.
Kanellopoulos, C. and Psacharopoulos, G. (1997), 'Private education expenditure in a ‘free
education country: the case of Greece', International Journal of Educational
Development, Vol. 17, No. 1, 73-81.
Mincer, J. (1970), 'The distribution of labor incomes: a survey with special reference to the
human capital approach', Journal of Economic Literature, Vol. 8, No. 1, 1-26.
Nour, S. M. (2013), Education, Training and Skill Development Policies in Sudan: Macro–
Micro Overview. In Technological Change and Skill Development in Sudan (pp. 313-
407). Springer, Berlin Heidelberg. Germany.
Psacharopoulos, G. and Papakonstantinou, G. (2005), 'The real university cost in a “free”
higher education country', Economics of Education Review, Vol. 24, No. 1, 103-108.
Qian, J.X. and Smyth, R. (2011) 'Educational expenditure in urban China: income effects,
family characteristics and the demand for domestic and overseas education', Applied
Economics, Vol. 43, No. 24, 3379-3394.
Schultz, T.W. (1961), 'Investment in human capital', The American economic review, Vol. 51,
No. 1, 1-17.
Tansel, A. and Bircan, F. (2006), 'Demand for education in Turkey: A tobit analysis of
private tutoring expenditures', Economics of Education Review, Vol. 25, No. 3, 303-
313.
Tilak, J.B. (2002) 'Determinants of household expenditure on education in rural India',
National Council of Applied Economic Research, Working Paper, No. 88, New Delhi,
India.
Tobin, J. (1958), 'Estimation of relationships for limited dependent variables', Econometrica,
Vol. 26, No. 1, 24-36.
World Bank (2016), World Bank’s Development Indicators. Accessed 10, July, 2016.
African Journal of Economic Review, Volume VI, Issue I, January 2018
175
Appendices
Table 1: Tobit Estimation Results for Household Education Expenditure in Sudan (Full, Urban
and Rural Sample)
Variable
Full Sample
Urban
Rural
Income
0.084***
(0.000)
0.130***
(0.001)
0.067***
(0.003)
Household's Head
Characteristics
Age
0.007***
(0.000)
0.006**
(0.043)
0.008***
(0.000)
Gender of head
-0.106
(0.177)
-0.138
(0.332)
-0.071
(0.445)
Married
-0.056
(0.471)
-0.090
(0.516)
-0.046
(0.625)
Education level of
Household head
Primary
0.153***
(0.002)
0.156*
(0.062)
0.151**
(0.011)
Secondary
0.337***
(0.000)
0.390***
(0.000)
0.238**
(0.016)
University
0.654***
(0.000)
0.526***
(0.000)
0.864***
(0.000)
Education level of the
Spouse
Spouse Primary
0.136***
(0.006)
0.045
(0.578)
0.202***
(0.001)
Spouse Secondary
0.371***
(0.000)
0.311***
(0.004)
0.360***
(0.002)
Spouse University
0.439***
(0.000)
0.374***
(0.009)
0.408**
(0.038)
Number of children in
household
Pre-school
-0.155***
(0.000)
-0.153***
(0.001)
-0.149***
(0.000)
Primary school
-0.006
(0.790)
-0.063*
(0.075)
0.034
(0.190)
Secondary school
0.151***
(0.000)
0.118**
(0.027)
0.177***
(0.000)
University level
0.181***
(0.000)
0.159***
(0.000)
0.196***
(0.000)
Profession of household's
head (agriculture as
reference)
Service
0.338***
(0.001)
0.375
(0.498)
0.350***
(0.000)
Industry
-0.186
(0.745)
-0.412
(0.636)
0.107
(0.889)
Household type of
dwelling (house as
reference)
Apartment
0.375**
(0.024)
0.479**
(0.015)
-0.157
(0.682)
African Journal of Economic Review, Volume VI, Issue I, January 2018
176
villa
-0.152
(0.233)
-0.335
(0.249)
-0.078
(567)
Other Household
characteristics
Household size
0.079***
(0.000)
0.093***
(0.001)
0.065***
(0.002)
Room
0.037***
(0.001)
0.062***
(0.002)
0.021
(0.128)
Electricity
0.353***
(0.000)
0.425***
(0.000)
0.306***
(0.000)
Urban
0.273***
(0.000)
Region (Khartoum as
reference)
Northern
-0.555***
(0.000)
-0.720***
(0.000)
-0.260
(0.109)
Eastern
-0.444***
(0.000)
-0.524***
(0.000)
-0.183
(0.274)
Central
-0.660***
(0.000)
-0.481***
(0.000)
-0.538***
(0.001)
Kordufan
-0.534***
(0.000)
-0.625***
(0.000)
-0.319*
(0.053)
Darfur
-0.081
(308)
-0.103
(0.347)
0.110
(0.497)
Constant
0.962***
(0.000)
0.997***
(0.000)
0.827***
(0.000)
Observations
7257
2230
5027
Pseudo R2
0.113
0.093
0.092
LR chi2
1589.20 (0.000)
491.25 (0.000)
767.91 (0.000)
Log likelihood
-6198.213
-2378.344
-3784.186
Note: p-values in parentheses
***p<0.001,**p<0.01,*p<0.05
African Journal of Economic Review, Volume VI, Issue I, January 2018
177
Table 2: Tobit Estimation Results for Household Education Expenditure by Income quintile
Variable
1st quintile
2nd quintile
3rd quintile
4th quintile
5th quintile
Income
0.030
(0.431)
-0.050
(0.224)
-0.026
(0.551)
0.007
(0.875)
0.152***
(0.001)
Household's head
characteristics
Age
0.004
(0.185)
0.005
(0.130)
0.007*
(0.052)
0.004
(0.193)
0.008**
(0.021)
Gender of head
0.209
(0.161)
-0.088
(0.569)
-0.230
(0.187)
-0.174
(0.322)
-0.064
(0.721)
Married
-0.140
(0.352)
-0.105
(0.502)
-0.382**
(0.027)
0.070
(0.681)
0.006
(0.970)
Education level of
household head
(illiterate as reference)
Primary
0.025
(0.846)
0.084
(0.420)
0.241**
(0.015)
-0.046
(0.633)
0.343***
(0.001)
Secondary
-0.181
(0.476)
0.070
(0.705)
0.370**
(0.013)
0.059
(0.679)
0.545***
(0.000)
University
1.107
(0.207)
1.381***
(0.005)
0.313
(0.277)
0.094
(0.623)
0.696***
(0.000)
Education level of
the Spouse (illiterate
as reference)
Spouse Primary
0.244*
(0.060)
0.026
(0.815)
-0.055
(0.579)
0.115
(0.235)
0.206**
(0.043)
Spouse Secondary
0.904***
(0.009)
-0.101
(0.631)
0.202
(0.245)
0.330**
(0.041)
0.378***
(0.003)
Spouse University
1.519***
(0.004)
0.821**
(0.015)
-0.089
(0.812)
0.509**
(0.017)
0.285*
(0.094)
Number of children
in household
Pre-school
-0.157**
(0.015)
-0.211***
(0.001)
-0.181***
(0.002)
-0.126**
(0.015)
-0.153***
(0.003)
Primary school
0.109**
(0.043)
-0.164***
(0.003)
0.002
(0.966)
0.007
(0.874)
-0.010
(0.799)
Secondary school
0.224***
(0.005)
0.037
(0.619)
0.158**
(0.018)
0.228***
(0.000)
0.098
(0.115)
University level
0.136**
(0.026)
0.006
(0.917)
0.144***
(0.008)
0.209***
(0.000)
0.209***
(0.000)
Profession of
household's head
(agriculture as reference)
Service
0.201
(0.123)
0.250
(0.149)
0.455**
(0.041)
0.808***
(0.006)
0.241
(0.485)
Industry
-0.991
(0.255)
-0.241
(0.801)
0.116
(0.915)
1.137
(0.320)
Household Type of
Dwelling (house as
reference)
Apartment
-0.416
(0.179)
-0.228
(0.503)
0.624***
(0.006)
Villa
-0.348
-0.149
-0.027
African Journal of Economic Review, Volume VI, Issue I, January 2018
178
(0.283)
(0.316)
(0.921)
Other Household
Characteristics
Household size
0.004
(0.925)
0.212***
(0.000)
0.107***
(0.009)
0.037
(0.279)
0.033
(0.266)
Room
0.062*
(0.061)
0.005
(0.871)
-0.003
(0.916)
0.015
(0.531)
0.006
(0.774)
Electricity
0.092
(0.585)
0.228***
(0.002)
0.212**
(0.018)
0.258***
(0.005)
0.542***
(0.000)
Region (Khartoum
and reference)
Northern
-1.096***
(0.005)
-0.809***
(0.000)
-0.459**
(0.010)
-0.564***
(0.000)
-0.768***
(0.000)
Eastern
-0.971***
(0.009)
-0.645***
(0.001)
-0.349*
(0.050)
-0.387**
(0.021)
-0.614***
(0.000)
Central
-1.246***
(0.001)
-0.947***
(0.000)
-0.689***
(0.000)
-0.692***
(0.000)
-0.660***
(0.000)
Kordufan
-0.818**
(0.028)
-0.625***
(0.001)
-0.521***
(0.005)
-0.599***
(0.001)
-0.714***
(0.000)
Darfur
-0.573
(0.113)
-0.299
(0.113)
0.074
(0.681)
-0.095
(0.562)
-0.090
(0.540)
Constant
1.577***
(0.001)
1.844***
(0.000)
2.080***
(0.000)
1.565***
(0.000)
1.173***
(0.001)
Observations
1419
1507
1671
1211
1440
Pseudo R2
0.110
0.087
0.079
0.066
0.103
LR chi2
155.61
(0.000)
182.7
(0.000)
214.17
(0.000)
205.94
(0.000)
402.61
(0.000)
Log likelihood
-625.912
-952.165
-1244.937
-1442.129
-1740.427
Note: p-values in parentheses
***p<0.001,**p<0.01,*p<0.05