Determinants of Child Schooling Progress in
Rural Ethiopia
Degnet Abebaw1,*,
Andinet Delelegn2 and Assefa Admassie3
1 Senior
Researcher, Poverty and Human Resources Division, EEA/EEPRI
2 Assistant Researcher, Poverty
and Human Resources Division, EEA/EEPRI.
3 Director, EEA/EEPRI.
Abstract
Despite rapid
expansion of primary enrollment in Ethiopia,
many children continue to remain behind normal schooling progress. Current
national figures indicate that nearly 45% of the children who enroll in grade
one drop out school before completing grade five. The purpose of this paper is,
therefore, to examine the main micro-level determinants of schooling progress
of primary school-age (7-14 years old) children using data from rural Ethiopia.
Using demand side and supply side factors, estimation results of the Poisson
regression provide various explanations for this delay in schooling. The major
demand side factors determining schooling progress include poverty, parental
education, land and non-land asset ownership, village fixed effects and a
child’s demographic characteristics. On the supply side, differences in availability
of primary and junior schools in the village significantly explain variation in
children’s primary education achievement. The analysis also shows that the
importance of these factors vary between boys and girls. The paper concludes by
highlighting some policy interventions required to raise child schooling progress
as well as close the gender gap in primary education in rural Ethiopia.
JEL Classification: D13; I21; R21.
Keywords: child schooling progress; Poisson
regression; Ethiopia.
1. Introduction
Education is a
process through which mankind transmit experience, new findings, and values
accumulated over time with the aim of enabling individuals and societies to
make all rounded participation in the development process. In this regard, education
plays a key role in enhancing economic progress, improving individual welfare
and social development (Hannum and Buchmann, 2005). Available evidence shows
that there are several channels through which such effects may arise. For
instance education raises labor productivity (Welch, 1970), increases
technological innovation and adaptation (Rodriguez and Wilson, 2000),
contributes to better health (World Bank, 1993) and gives greater ability to
deal with shocks (World Bank, 2001). As a result, education is to date a basic
ingredient for creating a competitive and knowledge based economy (Hanushek and
Kimko, 2000).
In recognition
of these multidimensional benefits, governments, international community and
development agencies have placed increasingly high attention on education
(Pritchett and Filmer, 1997). In fact, the United Nations World Conference of
Education in Jomtien, Thailand
in 1990, issued a declaration of “education for all” and recalled that access
to education is a fundamental human right. In 2000, in Now York, the UN set
down Millennium Development Goals (MDGs) which not only reiterated Universal
Primary Education (UPE) as one primary goal but also re-enforced that there
should be gender equality in access to primary education in all countries. In
this respect, the submit calls for countries to plan and implement appropriate
policies and strategies to ensure UPE for all children by 2015. Achieving the
goal of UPE by 2015 requires not only children are in school but also complete
primary education. Furthermore, the MDGs require development agencies and
developed countries to re-affirm their commitment to provide financial and
technical assistance to poor countries to implement these plans.
Notwithstanding
the global agreement that ensuring UPE is desirable, there is wide disagreement
on how to achieve it. As a recent World Development Report (World Bank, 2004)
points out, even though expanding access is important it is not enough to
ensure that all children from different backgrounds are enrolled and progress is
made in the education system. In a recent study on possibilities of UPE for
Sub-Saharan Africa, Bennell (2002) has noted that in countries where poverty is
deep-rooted and education is de facto
non-compulsory, the supply of education may not create its own demand. This
means that apart from physical expansion of school infrastructure, context
specific policy measures are required to create effective demand for education
among poor households and individuals (Ray, 2003; Bedi and Marshal, 2002;
Al-Samarrai and Reilly, 2000).
In the
literature, child schooling has been extensively studied by several researchers
using different indicators. Many researchers use either current status of
enrollment (e.g. Cockburn and Dostie, 2007; Lire, 2005; Assefa, 2002), number
of school days attended in a reference period (e.g. DeGraff and Bilsborrow,
2003; Bedi and Marshal, 2002), or the last completed grade of schooling (e.g.
Ray, 2003; Tansel, 2002). However, as pointed out by Behrman and Knowles (1999,
p. 2175), relying on these indicators fails to account for “…the age of
starting school, grades passed per year in school, and performance on
examinations in the last completed grade”. In other words, these and similar
other studies do not explain cumulative education performance of children.
Some recent
studies, which have tried to fill this information gap, are found for Honduras
(Gitter and Barham, 2007), Peru
(Pal, 2004; Patrinos and Psacharapoulos, 1997) and Mozambique
(Handa and Simler, 2005). In analyzing cumulative education performance, these
studies apply a composite performance index by combining information on
official school entry age and a child’s current age and his/her actual grade
completed.
In this paper,
we build on the latter approach and try to examine whether and to what extent
rural Ethiopian children are able to keep up normal schooling progress in the
primary education system. More specifically, the study estimates the impact of
individual, parental, household and community factors on child educational
progress. The data used for the analysis were extracted from the two rounds of
the Ethiopian Rural Household Survey, conducted in 1997 and 2004 by the
Economics Department of Addis Ababa University in collaboration with the Oxford
Center for African Economies.
The paper is
organized as follows. In Section 2, the paper presents background information
and problem statement. Review of the relevant literature is found in Section 3.
In Section 4, the paper outlines the theoretical framework. Data sources and
research methodology are discussed in Section 5. In Section 6, the paper
presents the descriptive and the analytical results and the discussions.
Finally, in Section 7, the paper provides conclusions and draws some policy
implications.
2. Background to the Ethiopian Education
System and the Problem Statement
In Ethiopia,
as in many other developing countries, education has received an increased
government attention. Adopted in 1994, the current education and training
policy of the Ethiopian government stresses boosting coverage and ensuring
equitable access to education for all. In transferring these policy objectives
into action, the government has formulated and adopted a twenty year education
sector development program in 1997, with a rolling period of five years. It is
also interesting to note that the education sector currently receives the
largest share of public spending on pro-poor sectors in
the country. For instance, between
2001/02 and 2004/05, the sector received about 18% of the government’ total
expenditure on pro-poor sectors. As a result of the new education policy
environment and increased public budget allocation, school enrollments have
improved significantly over the last decade. For instance, as shown in Figure
1, between 1998/99 and 2004/05, primary school gross
enrollment rate has increased from 56% to 88% for boys and from 35% to 72% for
girls (MOE, 2005). These figures indicate that slowly, but incrementally, the
gender gap in children’s primary school participation in Ethiopia
has narrowed down.
Figure 1. Trends
in Primary School Enrollment, 1998/99-2004/05
Despite various efforts
taken by the Ethiopian government to increase coverage and equitable access to
all children, the education system still faces several problems. According to
UNESCO (2006), about 59% of the country’s adult population (15 years and over) is
illiterate indicating insufficient access to education in the past. Available
evidences also suggest that many children do not get the chance to be in school
at an appropriate age at present. For instance, according to the MOE (2005), of
the total children who enrolled in grade one in 2004/05 academic year, only
about 60% are seven years of age. Given the low enrollment records and lack of
access to schools in the past, the remaining (40%) of the current enrollees in
grade one are older than seven years of age.
Apart from this,
children’s primary schooling in Ethiopia
encounters high rates of dropout and of grade failures. For instance, according
to the official data sources (MOE, 2005), nearly one among four grade-one
children drops out of school and about 3% of them repeat grade-one. Moreover,
primary education in Ethiopia has low rates of completion with around 43% and
66% of children quitting school before completing, respectively, the lower
primary education (grade 5) and upper primary education (grade 8), in 2004/05. The
MOE figures further reveal that whereas dropout rates are higher for boys,
girls repeat grade more often than boys.
As noted by
Bennell (2002) for many other Sub-Saharan countries, the high gross enrollment
figures in Ethiopia
mask the fact that nearly half of children do not complete the full primary
education. Moreover, as noted by Bedi et
al. (2004), an increase in gross enrollment rate could be due to a rise in
the number of children repeating grades. In fact, the critical educational
problem for many countries around the world is not the availability of school facilities
but children dropping out or not attending available schools (World Bank, 2004).
The economic
literature on costs of late entry to school, grade repetition and delays in
grade completion are discussed by several authors (see for e.g. Behrman and
Knowles, 1999). By starting school late, a child encounters delays in
post-school returns. The cost of delayed schooling then equals the difference
in the present discounted value of future income with the delay and without the
delay. Late school entry is also a deterrent to full school participation
(Behrman and Knowles, 1999). A similar result has been reported by recent
empirical studies for Rural Tanzania (Bommier and Lambert, 2000; Burke and
Beegle, 2004), Ghana
(Ray, 2003), and Honduras
(Bedi and Marshall, 2002), and the Philippines
(Tan et al., 1997), to cite but a few
examples. The main factor for this effect is that as children grow older they
are more likely to participate in the labor market or on farm activities or in
the domestic work. On the other hand, failure to complete grades at a proper
pace results in two types of costs. In the first place, like delays in
schooling, it causes delays in post-school returns. Second, it increases both
private costs for parents and internal inefficiency of primary education for
the government (Colclough and Al-Samarrai, 2000). Others (e.g. Bedi and
Marshall, 2002) also note that low school progression has a negative influence
on final educational achievement of the child. As noted by Binder and Scrogin
(1999), poor education performance such as grade failures would induce children
to reduce their academic effort and increase work effort.
Thus, there is
ample concern for seeking desired school progress corresponding to a child’s
age. Given the paucity of empirical studies on this subject in Ethiopia,
the main objective of this paper is to estimate the principal determinants of
desired
schooling progress of children in rural Ethiopia.
Unlike enrollment rate, which is a flow measure, grade-for-age score measures a
child’s stock of human capital accumulation at a given age. As discussed later,
apart from enrollment this score shows if a child has encountered problems such
as late entry, failed grades or dropped out of school in the past. Stated in
other words, the grade-to-age score shows whether a child has kept pace with
grades expected to his or her age. In view of these features, an econometric
analysis of its main determinants can generate more information for policy
makers and school administrators. As mentioned previously, this is what this
paper intends to address by using a rich dataset from Ethiopian rural children.
3. Literature Review
Research on
child schooling is vast and this paper reviews only a limited part of it, which
is relevant for developing countries. Previous studies (e.g. Patrinos and
Psacharopoulos, 1997; Gitter and Barham, 2007; Jacoby and Skoufias, 1997; Bedi
and Marshall, 2002, Gitter and Barham, 2007; Lire, 2005; Assefa, 2002) find
several explanations for the inadequate schooling and educational attainments
of children particularly in developing countries of Africa,
Asia, and Latin America. A common
thread running through these studies is that child schooling experience in
rural areas is related negatively with household poverty, and child age.
Regarding gender effect, most studies just mentioned above find that girls are
more likely to get less schooling than boys and that parental education has a
positive and significant influence on enrollment and level of educational
attainment.
Using a panel
data from Tanzania,
Burke and Beegle (2004), show that child school attendance is determined by a
host of factors including household, child, and community characteristics. But
they also noted that there are important gender differences in the factors
influencing child schooling attendance and participation. In their analysis of
the determinants of primary enrollment in Kenya,
Bedi et al. (2004) have shown the key
role played by child age, parental education, household wealth and school
inputs on parents’ decision to enroll
their children.
In a recent
study in Ethiopia,
Assefa (2002) provides evidence that while child labor is a common phenomenon
in rural Ethiopia,
many children are able to combine schooling with work, particularly in the
agricultural sector. Similar evidence is obtained by Cockburn and Dostie (2007)
who finds that child time allocation decision among schooling, and work in
rural Ethiopia
is strongly determined by a combination of household income and wealth, family
composition and asset ownership. In both studies, extent of child time
allocation for these activities has not been explored since they have adopted a
categorical dependent variable. Furthermore, these authors focused their
attention on current status of child schooling and as such did not analyze
schooling progression.
4. Theoretical Framework
The economic
literature on schooling attainment has its root on human capital theory
pioneered by Schulz (1960), Becker (1964), and Mencer (1974). According to this
theory, investment in child schooling is justified by taking into account costs
and returns of schooling. To put it another way, the level of schooling that
parents choose for their children are determined by private marginal costs and benefits.
In Ethiopia,
primary education is free and parents are not obliged to pay fees. However,
they face various other costs of sending their children to school. These costs include
direct outlays for school uniforms, transport, books and school supplies (pen,
pencil, and notebooks). Again, in Ethiopia,
children participate in different activities on-or off-farm and domestic
chores. This implies that children’s time spent in schooling has an opportunity
cost since it involves forgoing children’s contributions to household
production and to the labor market. Opportunity costs of
schooling may vary from one child to the next depending on the type of activity
performed and labor productivity in alternative activities (Song et al., 2006). To capture this effect it
is essential to use child labor wage in the estimation. However, data on this
variable is not available. One means of addressing this limitation is to
incorporate variables such as a child’s
own demographic characteristics (i.e. age, sex) and location and community
characteristics into the empirical model as these variables could determine the
child’s opportunity cost of schooling (Song et
al., 2006; Bedi and Marshal, 2002).
As previously
mentioned, schooling benefits include increased future earnings. Since costs of
and returns to current schooling decisions occur at different time periods, an
individual household invests on education of his/her family member as far as
expected returns are greater than costs. Following this notion, many authors
have researched schooling investments and its determinants both in developed
and developing countries.
From the
viewpoint of the Beckerian proposition, well-functioning markets (e.g. credit,
labor, etc) are required for individuals and families to acquire optimal level
of schooling. When these markets are absent or malfunctioning, individuals and
families may under-invest in education. In fact, as pointed out by several
authors markets are imperfect (de Janvry et
al., 1991) and as reported by several authors it is often difficult or is
impossible to borrow for human capital investments (Bardhan and Udry, 1999).
Recent empirical studies have also shown that investments on education are
hampered by credit market imperfections (see for e.g. Gitter and Barham, 2007;
Jacoby and Skoufias, 1997). Stated in other words, beside market factors, one’s
schooling decision is closely associated with household resource endowments
such as wealth and income (Berham and Knowles, 1999).
By embedding
human capital within Becker’s (1981) household production model, one obtains a
theoretical basis to evaluate determinants of investment on schooling. Assuming
that a parent’s main objective is to maximize expected utility function in
consumption and education over two-periods, subject to a full-income
constraints a reduced form schooling demand function can be specified as:
,
where, 
is number of completed
grades by a school-age child in the family, 
is a vector of market
wages for adults and children, 
is a vector of expected
future earnings conditional on current enrollment, 
is unearned income, 
is a vector of child, parental and household characteristics,
and Z is a vector of community and regional factors.
5. Data and Empirical Model
5.1 Data Sources
The data used in
this study came from the Ethiopian Rural Household Survey (ERHS) collected by
the department of Economics, Addis Ababa
University in collaboration with
the Center for the Study of African Economics of Oxford University. The survey
has been conducted on rural households for six rounds between 1994 and 2004,
i.e., 1994a, 1994b, 1995, 1997, 1999/2000, and 2004, in fifteen villages in
Amhara, Oromia, Tigray and Southern Nations Nationalities and Peoples Regions.
The four regions
included in the sample represent the main sedentary faming system in the
country. The main farming systems found in the study areas include the
plough-based cereals faming system of the Northern and Central Highlands, the Mixed
Plough/Hoe Cereals Farming System and enset farming system in the southern parts of the country, and cash
crop farming such as Chat and coffee. Furthermore, sample size in each village
was chosen so as to approximate a self-weighting sample, when considered in
terms of the farming system.
The survey is
aiming at generating a multi-purpose dataset comprising a range of household
level demographic, consumption, education, health, income, and asset variables
on one hand and, community, market and other infrastructural variables on the
other hand during each survey period. The questionnaire is comprehensive in a
way that there are a number of modules included in the questionnaire.
More
specifically, in this paper we use data from 2004 and 1997 ERHS. The 1997
survey data were specifically to control for initial socioeconomic differences
affecting child education. In fact, according to Ethiopia’s
policy, a primary school-age child in grade 8 in 2004 is expected to have
started his/her primary schooling in grade 1 in 1997. Our empirical analysis
for this paper was based on sample of 1290 primary school-age rural children
living in 622 households.
5.2. Empirical Model and Hypotheses
In this section,
we specify the empirical model and identify important explanatory variables for
the empirical analysis. As mentioned previously, the main aim of the paper is
to identify the key determinants of child schooling progression in rural Ethiopia.
Hence, our unit of analysis is a primary school-age child, even though the data
used in the analysis were as reported by the child’s household head.
The dependent
variable, which we want to explain in this paper is education gap, 
, experienced by primary school-age rural children in Ethiopia.
An increase in value of indicates a decline in
a child’s schooling progress. As specified below, is measured by
matching grades completed by a child to the child’s current age (Patrinos and
Psacharopoulos, 1997). In notation, let 
and 
are, respectively, the
appropriate grade in relation to a child’s age and the actual grade completed
by a child in a household. To put it
another way, is a desired level of schooling progress and can be computed
as difference between a child’s current age and official enrollment age as:
Then, the
dependent variable , is constructed as follows:
.
As shown above,
the dependent variable stands for count of grades delayed and is a non-negative
integer. If a child enrolls at official entry age and keeps up normal progress
in school, then the value of will be zero. However,
due to a variety of problems such as late entry, drop outs and grade failures, will have a positive integer value. In other words, in such
circumstances children remain some grades behind the appropriate school level
for their age. Thus, modeling this kind of a dependent variable requires the use
of certain kind of count data models (see Cameron and Trivedi, 1998). In the
count data analysis, the probability that the 
child is grades behind a desired
grade in schooling can be expressed as:
.
In specifying
such a discrete probability function, a common starting point is to apply a
Poisson regression, which is a count data model. The Poisson model can be
specified as follows:
,
Where,
is the mean of the
distribution, 
is a vector of Poisson
regression coefficients to be estimated and 
. The exponential function has merit in that it restricts the
dependent variable to the non-negative range. In the statistics literature, the
model is also known as log-linear model since the logarithm of the conditional
mean is linear in parameters:
.
In the Poisson
model, it is assumed that the conditional mean, 
, of the dependent variable given the exogenous variables is
equal to its variance, which is represented as follows:
.
To test this
assumption, Cameron and Tivedi’s (1990) regression approach can be used as
follows:
.
Here it is
important to note that a value of 
, which is statistically different from zero is a sign of the
presence of either over-dispersion 
or under-dispersion 
in the data. To relax
this limitation, several approaches are proposed in the literature. Among
these, a Negative Binomial regression, which allows the conditional mean of the
dependent variable to exceed from the conditional variance have been widely
used (Cameron and Trivedi, 1998).
In Table 1, we
identify the main explanatory variables along with their anticipated impact on
the expected number of grades lag among primary school-age children in rural
areas of Ethiopia.
Our hypotheses on the variables have been guided by economic theory, previous
empirical studies and field context. As mentioned previously, the explanatory
variables are composed of child, household and community characteristics.
In agrarian
societies, such as Ethiopia,
children are important contributors to family income especially in rural areas
where agriculture employs most of the child labor force in the country.
Moreover, in the Ethiopian culture, domestic chores are usually performed by
women and children. This means that child schooling entails some positive
opportunity cost for parents and this costs may vary depending on the age and
sex of the child. A study by Patrinos and Psacharopoulos (1997) for Peru
indicates that a child’s age has a positive and significant effect on his/her
schooling progression. A recent case study from Nepal,
Peru and Zimbabwe
(Lire, 2005) also confirms that child schooling is negatively associated with a
child’s age. The main mechanism underlying this effect is that as a child gets
older, it is more likely that he/she is sent to work than school. In a similar
vein, child age is expected to increase child schooling delay in the present
study. Consistent with previous studies (e.g. Patrinos and Psacharopoulos,
1997; Grira, 2004) in this study boys are expected to have more schooling progress
than girls.
Previous
empirical studies on the effect of the number of children on child schooling provide
mixed results. Empirical findings corroborate with two differing arguments. The
first is based on resource dilution hypothesis which says that as the number of
children in the household increases, the per capita schooling resources for
each child decreases. Consistent with this theory, Dayioglu, 2005) reports that
child labor increases with the number of children in the family. In a study of Nepal,
Peru, and Zimbabwe,
Lire (2005) also finds that the probability of child schooling increases with
the number of young children. The second argument on the effect of number of
children on schooling is based on “specialization” in the family. According to
this argument, parents put some children to work, others to education as the
number of children increases in the family. Again, this proposition is
supported by empirical evidence from Botswana
(Chernichovsky, 1985 cited in Patrinos and Psacharopoulos, 1997). In order to
test these competing hypotheses in the present study, two variables, namely,
number of young boys and number of young girls in the household, are
incorporated into the model, but with no a
priori signs.
As noted
earlier, it has been anticipated that household and parental factors play a key
role in determining child schooling progress. Education of parents influences
child education through a variety of pathways. Educated parents may have more
information about advantages of educating their child and thus provide better
schooling environment at home such as helping children with their homework.
Secondly, educated parents may be more capable to deal with disequilibria when
shock exists to the family with out dislocating children out of school.
Thirdly, educated parents may have higher permanent income or wage, better
health and thus better education to their children. Fourthly, and related to
one or more of the above mentioned factors, educated parents are more
productive and hence need minimal or no child labor. In fact, several empirical
studies (e.g. DeGraff and Bilsborrow, 2003; Grira, 2004; Lire, 2005; Gitter and
Barham, 2007) concur with these arguments and find a positive and statistically
significant effect on child schooling of parental education. As noted by
Kurosaki et al. (2006), this effect
is consistent with the wealth effect hypothesis (educated parents are richer
than uneducated parents) and the preference effects hypothesis (educated
parents value education more than uneducated parents). As such, this study
hypothesizes for rural Ethiopia
a similar impact on child schooling of parental education using two proxy
variables, namely education of the household head and his/her spouse.
The literature
on the effect of poverty on child schooling and/or labor provides different
arguments and mixed findings. On one hand, according to the “Luxury Axiom” of
Basu and Van (1998), child labor is expected to decline as parents’ income
raises. Along this notion, Edmonds (2005) for Vietnam
reports household expenditure has negative and significant effect on the
incidence of child labor. On the other hand, according to the “wealth paradox”,
parents with productive assets to work with are more likely to send their
children to work than school. However, this effect again depends on the type
and size of assets owned by parents. Thus, it is important to use different
variables measuring wealth in empirical studies such as ours. For instance,
using survey data from Ghana and Pakistan, Bhalotra and Heady (2003) find the
incidence of child labor is higher among land-rich parents, whereas Cockburn and Dostie (2007) and Assefa (2002),
both using data from rural Ethiopia, report that child schooling enrollment is
influenced negatively by the size of landholding, and positively by other
assets such as farm equipment. In fact, the effect of land holding on schooling
or child labor depends on availability of competitive markets for land and labor.
If labor markets are available and a farmer can use his land as collateral to
borrow money from bank to hire in labor, then large holdings would not exert a
negative effect on child schooling (Bhalotra and Heady, 2000). However, in Ethiopia,
land belongs to the state and farmers cannot use land as collateral. Moreover,
labor markets are imperfect and farmers cannot hire in labor as they wish. One
way to reduce the impact of imperfect labor market is to sell out some of the
land they have. However, this option is unlikely since they do not have the
right to sell land. Therefore, farmers have the incentive to employ their own
family labor including school-age children for farm activities.
Consistent with
these diverse findings, this study incorporates several explanatory variables
related to household poverty and wealth. We use household expenditure, measured
by an adult equivalent unit, as a proxy for household poverty status. Since
this variable may be potentially endogenous, we used its lagged value in 1997.
Our hypothesis is that child schooling progression is positively related with
household expenditure. To capture wealth effect, we included into our model two
variables, namely land and value of all consumer durables and farm equipment.
Specifically we anticipate that child schooling progression will be influenced
negatively by the size landholding and positively by the value of household
durable assets. Because of the lack of government land redistribution during
the last ten years on one hand and the absence of land markets in the country,
in this study we assume that household land ownership is more or less fixed and
it is less likely that land ownership causes problem of endogeneity with recent
child schooling behavior. Likewise, in using the Birr value of consumer durables and farm equipment as
proxy for wealth we assume that child labor is not used as a source of finance
for the accumulation of household durable assets.
In addition, we
anticipate that parents’ access to credit and remittance have a positive effect
on child schooling progress. These variables could help relax a resource poor
farmer’s budget constraint and enable him to finance schooling expenses of his
children. A recent study from Honduras
(Gitter and Barham, 2007) indicates a positive and significant association
between child schooling progress and household access to credit. Likewise,
Beegle et al. (2006) show that access
to credit has a negative and significant effect to reduce child labor in Tanzania. Regarding remittances, Edwards and Ureta
(2003) indicate that children of households who received remittances show
significantly lower hazards of leaving school at all grade level in El
Salvador.
Child
educational achievement is also influenced by school availability and village
characteristics. On this point, Lire (2005) finds evidence that a higher number
of schools in a community tends to improve school attendance and to reduce
child labor in Nepal.
Gitter and Barham (2007) have also reported a negative and significant effect
on child educational attainment of travel time to nearest school in rural Honduras.
In this regard, we incorporate dummy variables for whether primary and junior
schools are available in a child’s village and hypothesize that these variables
will have a positive influence on a child’s educational attainment. A main
econometric concern in using the school supply variables here is that they are
likely to become endogenous with child schooling. In other words, if the
placement of schools in the villages is determined by the presence of excess
delayed enrollment, then school supply becomes endogenous with the dependent
variable. To avoid this problem, we used lagged values of the school supply
variables.
In the
literature, one also finds evidence that lack of insurance markets or its close
substitutes drives children out of school and work in the labor market (de
Janvry et al., 2006; Jacoby and
Skoufias, 1997). It is thus important to expect that children residing in
different villages face different risk of schooling delay, especially if
villages vary from one another in terms of cultivated crops, agro-ecology and farming
systems. Thus, we incorporated village level variables to account for village
level fixed effects on child educational progress in rural areas. Besides,
regional dummy variables are included to capture regional differences in labor
market availability, school supply and other factors.
6. Results and discussion
6.1 Descriptive results
This section
describes important individual, household, and community level characteristics of
children in our sample. The dependent
variable, the educational gap,
is the difference between the ideal number of grades/classes that has to be
completed and the actual number of grades/classes completed by the child at
that school-age. The higher the gap is the poorer the educational performance.
The descriptive
results demonstrate that about 94% of the children experienced certain amount
of grade delay in schooling. In our sample the education gap experienced by a
child seems to vary with the age and sex of the child (see Table 2). In most of
the cases girls are more likely to start schooling later and also be enrolled
in lower grades than boys. The difference in gender gaps in primary schooling
progress becomes statistically significant for 12- to 14-years old children.
Table 2.
Age-Specific Education-Gap by Sex
|
Age
|
Boys (N = 657)
Mean Std. Dev
|
Girls (N = 633)
Mean Std. Dev
|
Mean difference between boys’ and girls’ education gap
|
|
7
|
0.83
|
0.38
|
0.88
|
0.33
|
0.05
|
|
8
|
1.66
|
0.61
|
1.61
|
0.66
|
0.05
|
|
9
|
2.31
|
0.97
|
2.27
|
0.99
|
0.05
|
|
10
|
2.84
|
1.16
|
3.0
|
1.19
|
0.15
|
|
11
|
3.54
|
1.48
|
3.14
|
1.64
|
0.41
|
|
12
|
3.74
|
1.86
|
4.16
|
1.8
|
0.42**
|
|
13
|
4.48
|
1.99
|
4.94
|
2.03
|
0.46*
|
|
14
|
4.61
|
2.27
|
5.39
|
2.36
|
0.78**
|
** and * represent
statistically significant means at 5% and 10% level, respectively.
On average,
children in our sample are 3 grades behind the desired school level appropriate
for their age. It is only 6% of the children who are enrolled at the
appropriate schooling level. For nearly 70% of the children the incidence of
education gap ranges from one to four grades (see Figure 2). Taken together,
the sample children completed only 2.5 grades. One
of the contributing factors to high incidence of child education gap or grade
delays is the low level of enrollment. Table 3 describes current enrollment
rate and grades completed according to age of children in the sample. The
result indicates children’s primary enrollment rate stands at 53%, which is
less than the national net enrollment rate of 57% for the same age group in
2003/04 (MOE, 2005). In fact, the problem is more severe for certain age
cohorts than others. For instance the lowest current enrollment rates are
observed for children aged 7 (26%) and 8 (34%) year olds suggesting that late
school entry is a serious problem in rural Ethiopia.
Figure 2: Kernel Density of Number
of Grades lagged among Primary School-Age Rural Children, 2004
|
Table3:
Current enrollment and grades attained by primary school-age rural children, 2004
|
|
Age
in Years
|
Number
of Counts
|
Count
Currently enrolled
|
Enrollment
Rate by Age
|
Average
grade currently completed
|
|
7
|
177
|
40
|
0.23
|
0.28
|
|
8
|
178
|
57
|
0.32
|
0.63
|
|
9
|
|