183

Document Sample
183 Powered By Docstoc
					        JOURNAL OF AFRICAN ECONOMIES, VOLUME 18, NUMBER 2, PP. 183– 211
              doi:10.1093/jae/ejn015 online date 7 August 2008


Universal Primary Education and School Entry in Uganda
                     Louise Grogan*
  Department of Economics, University of Guelph, Guelph,
                      ON, Canada


This paper examines the initial effects of the introduction of Universal
Primary Education (UPE) in January 1997 on school entry in Uganda.
Given that advanced age at school entry has historically been associated
with primary school dropout, the paper focuses on the the effects of fee
elimination on the age at which a child enters school. Data from the
2000 Uganda Demographic and Health Survey and 2001 Education
Data Survey are employed to examine the effects of UPE on the probability
that a child begins attending school before age nine. School fee elimination
under UPE is found to cause a 3% increase in this probability on average.
Effects are found to be particularly pronounced for girls and children
living in rural areas.

JEL classification: H41, H43, J18, O55




1. Introduction
A major aim of the United Nations’ Millenium Development Goals
(United Nations, 2000) is to reduce the number of uneducated
African youth. This manifesto sets 2015 as the target year for all
children in the world to complete primary school, and for boys
and girls to have equal access to education at all levels. In the
past ten years, several Sub-Saharan African countries have insti-
tuted measures aimed at this goal by eliminating primary school
fees in government-aided (public) schools. Malawi eliminated
* Corresponding author: Louise Grogan, Department of Economics, University of
  Guelph, MacKinnon Building, Room 743, Guelph, ON, Canada. Telephone: þ1
  (519) 824 4120 ext. 53473. Fax: þ1 (519) 763 8497. e-mail: lgrogan@uoguelph.ca

# The author 2008. Published by Oxford University Press on behalf of
the Centre for the Study of African Economies. All rights reserved.
For permissions, please email: journals.permissions@oxfordjournals.org
184     Louise Grogan

these fees in 1994, Uganda in 1997, Tanzania in 2000, and Cameroon,
Burundi, Ghana, Rwanda and Kenya in 2003.
   The elimination of school fees has been undertaken differently in
different African countries. While Universal Primary Educa-
tion (UPE) was introduced for all grades of primary schooling
in Malawi in 1994, prior to this it had been introduced on a
grade-by-grade basis since 1991. In Uganda, UPE was introduced
for all primary grades simultaneously in 1997.1 Lesotho began elimi-
nating school fees gradually in 2000, phasing in UPE by first elimi-
nating fees at the first year of primary school. Kenya, like Uganda,
eliminated school fees at all levels of primary schooling simul-
taneously in 2003.
   In all countries in which UPE was instituted, the elimination of
the direct costs of schooling created an instantaneous large increase
in school enrolment. Enrolment increased by nearly 70% in Malawi
in the first year of implementation, and by 58% in Uganda (Uganda
Ministry of Education and Sports, 1999). When school fees were
eliminated in Lesotho for students in the first year of primary
school, enrolment increased by 75%. In Kenya, enrolment increased
by 22% in the first year of its programme. These aggregate increases
in enrolment after the elimination of fees reflect both increases in
school attendance among the primary school-age population and
increases due to adults and teenagers attending school for the
first time.
   Whereas school fees and subsidies have been much examined in
the US education literature, to date few studies have examined the
impact of the elimination of school fees in countries of Sub-Saharan
Africa on pupil enrolment, learning outcomes or retention in
schooling. Primarily, this is because the elimination of school fees
is such a recent phenomena that data are not yet available.2 The lite-
rature on school vouchers in the US (see, for example, Epple and
Romano, 1998; Ladd, 2002; Neal, 2002) has documented the large
behavioural effects of reducing the cost of schooling. Several
papers have used natural experiment techniques to identify the

1
    In both Uganda and Malawi, the elimination of school fees at all levels of
    primary schooling was prompted by a national-level election campaign.
2
    The most recent Demographic and Health Survey (DHS) and Educational Data
    Survey (EdData) for Malawi, which adopted UPE first in 1994, were carried
    out on a different sample than the main DHS survey. This means that important
    household information cannot be related to pupil information.
                                        UPE and School Entry in Uganda            185

effects of subsidising post-secondary education on enrolment at the
college level (see, for example, Kane, 1995; Heckman et al., 1998;
Ichimura and Taber, 2002). To date, only one published study has
examined school fee elimination in any African country. Deininger
(2003) shows that there were substantial increases in overall enrol-
ment rates in Uganda following school fee elimination, and that
fee elimination reduced socio-economic disparities in access to
primary schooling. However, Deininger finds that there were notice-
able reductions in the quality of education following the influx of
UPE entrants.
   This paper focuses on one margin on which the elimination of
school fees might be expected to have had an impact: the age at
which children enter schooling. This is a very important margin in
the Ugandan case. As will be demonstrated, school entry at ages
above eight is very strongly associated with early school dropout.
The Uganda DHS Survey and EdData Survey, which were under-
taken in late 2000 and in the first half of 2001, together comprise
one of the first sources of data with which an examination of poten-
tial effects of school fee elimination on this margin is feasible.


2. Background
Uganda eliminated school fees amid the background of a spiralling
AIDS crisis and civil war in the north of the country. However,
several previous crises since the country’s first elections as an inde-
pendent nation in 1962 had shattered the economy and education
                                  ´
system. The successful coup d’etat launched by Idi Amin in 1971
stalled the educational progress achieved since independence. By
1985, government expenditures on education amounted to about
27% of the levels of the 1970s (World Bank, 1993).
   Recognising the education crisis, the Government of Uganda in
1987 convened an Education Policy Review Commission (EPRC),
which had the mandate to make policy recommendations for all
levels of education. In their report to the government in 1989, the
Commission recommended the universalisation of primary edu-
cation as soon as feasible, stating:
          ‘Only when every child is enrolled at the right age and does not leave
          school without completing the full cycle of primary education it would be
          possible to ensure that all the citizens have the basic education needed for
186   Louise Grogan

           living a full life. Also, it will help in achieving a transformation of the
           society leading to greater unity among the people, higher moral standards,
           and an accelerated growth of the economy’. (Uganda Ministry of
           Education and Sports, 1999).


In 1993, the Government of Uganda and the United Nations
Children’s Fund (UNICEF) began a series of initiatives to increase
school enrolment in Uganda. The Primary Education and Teacher
Development Project was begun in 1993. It comprised the following
major goals: (i) to reform the education of primary teachers, (ii) to
prepare for reforms in the primary school curriculum, (iii) to
reform the pupil examination system, (iv) to improve the provision
of textbooks and reading materials in classrooms, (v) to introduce a
system of assessing the quality of education provided and (vi) to
introduce a framework for country-wide assessments of the
overall progress in education.
   The rapid elimination of school fees at the primary level was
likely accelerated by the first direct elections for President of
Uganda, which took place in the spring of 1996. The eventual
winner of these elections, the current president Yoweri Museveni,
made a campaign promise to provide free primary schooling to
four children per Ugandan family. In December 1996, after being
elected, President Musveni announced that school fees would be
eliminated in January 1997, coincident with the new school year.
An enumeration and advertising campaign was undertaken, and
the new school entrants began learning within two months of the
presidential announcement. In practice, school fees were waived
for all primary school students, regardless of how many siblings
were also attending school.
   The announcement of UPE in late 1996 committed the govern-
ment to paying tuition fees at the rate of 5000 Ugandan shillings
per pupil per annum in the first three years of schooling, and 8100
Ugandan shillings for the fourth to the seventh classes. Other
costs of schooling, such as transportation and uniforms, remained
the responsibility of families. To put these fees in the context
of local salaries, in 1999 a teacher in a government-aided school
in Uganda earned about 75,000 shillings per month (Uganda
Ministry of Education and Sports, 1999).
   The elimination of school fees removed one of the main sources
of funding for government-aided schools and replaced it with a
                                          UPE and School Entry in Uganda        187

commitment to very substantial increases in funding from the
government. According to the Uganda Ministry of Education and
Sports (1999), parental contributions were providing up to 90% of
recurrent and capital expenditures made by schools just prior to
the elimination of school fees. The introduction of UPE initially
increased the reliance of government-aided schools on the receipt
of money from Kampala.
   Evidence suggests that funding from the Ugandan government
was relatively unlikely to reach schools in the mid-1990s, before
the introduction of UPE. Reinikka and Svensson (2004) examine
the receipt of school grants in Uganda using school-level panel
data. They find that only 13% of capitation grants for non-wage
expenditures actually reached schools in Uganda during 1991 –5.
Most of the grant was absorbed by local politicians and administra-
tors. As well, they find that schools in better-off communities
received larger fractions of the original grant money. These results
suggest that, prior to the introduction of UPE, government-aided
schools in poorer areas of Uganda were more dependent on
revenue from school fee collection than were those in wealthier areas.
   The central government devolved responsibilities for schools to
District Councils during 1997 and began publicising the level of
capitation grants through local newspaper campaigns. This
resulted in an increase from 13 to about 80% of capitation grants
reaching schools by 1999, according to the Public Expenditure
Tracking Survey (Ojoo, 2005; Reinikka and Svensson, 2005).
   Because school fees were eliminated before infrastructural
improvements in the school system had been carried out, the access
shock created by the elimination of fees resulted in a substantial
initial decrease in resources available per pupil and a large increase
in the pupil–teacher ratio. To address this, the Ugandan goverment
approved an Education Sector Investment Plan (ESIP) in December
1998. According to The Uganda Millenium Development Goal
Report (2003), the textbook-to-pupil ratio in Ugandan schools had
risen to 1:4 in 2002.3 However, by 2003, the textbook-to-pupil ratio

3
    The report states, ‘Although the sectoral budget allocation increased from 20.6
    billion Ugandan shillings at the start of UPE to 46.7 in 2003, this increase has
    not resulted in a proportional improvement in the pupil teacher ratio, or the
    quality of education’. However, international evidence does not show conclus-
    ively that pupil –teacher ratios are a key factor in learning outcomes. See, for
    example, Hanushek (1986) and Hanushek (1997) for evidence against strong
188   Louise Grogan

had risen to 1:3, the classroom-to-pupil ratio to 1:55 and the
desk-to-pupil ratio to 1:3 (Ministry of Education and Sports of
Uganda, 2005). Aside from further large-scale infrastructure invest-
ments, the government of Uganda has committed itself to training
large numbers of new teachers under the Primary Education and
Teacher Development Project operating since 1995 and to achieving
a textbook-to-pupil ratio of 1:1.
   Clearly, the elimination of fees in government-aided schools in
Uganda had substantial consequences across the education system
and at several levels of government. Household survey data, such
as the DHS and EdData surveys, can clearly not speak of school
quality or funding issues associated with UPE. However, the over-
arching goal of school fee elimination was to get Ugandan children
into school. The DHS and EdData surveys offer the opportunity to
provide a first assessment of the effectiveness of UPE in furthering
this goal.

3. Data
The 2000 DHS and 2001 DHS EdData surveys were, respectively,
the third and fourth DHS surveys undertaken in Uganda.
Previously, surveys had been undertaken in 1988 and 1995
(Uganda Bureau of Statistics and ORC Macro International, 1988a,
1995b). The 1995 survey provides useful information on the edu-
cational situation in Uganda just before the introduction of UPE.
This survey suggests that, in 1995, 70% of school-age boys attended
primary education, compared with 67% of girls (Uganda Bureau of
Statistics and ORC Macro international, 1995b). Owing to the preva-
lence of grade repetition, common to many countries in the
Sub-Saharan Africa, and to high drop-out rates, only 11% of
males and 9% of females of secondary school age were found to
attend school. About 26% of the population aged 15 or above had
successfully completed primary school. In the 1995 DHS, 32% of
primary school children were in the appropriate grade for their
age, with the vast majority being too old for their grade.
   The main DHS survey is a stratified random sample of Ugandan
households. This survey was conducted in late December 2000, and

 effects of school resources on learning outcomes, and Card and Kreuger (1990)
 and Case (1999) for counter-evidence.
                                         UPE and School Entry in Uganda        189

January and February 2001. After employing appropriate weights,
the survey can be considered representative at the national level.
The primary purpose of this survey, common to most DHS
surveys, is to provide information on education, nutrition, child
and adult mortality, fertility, maternal and child health and know-
ledge of HIV-AIDS. Health questionnaires were administered to
women and men, and detailed information on the living circum-
stances of each household was recorded.
   Within six months of the completion of the main DHS survey, the
specially constructed Ugandan EdData survey was administered.
Households containing individuals aged 5 –18 in the main survey
were targeted for this second survey, with some exceptions.
Households were excluded if an under-19 member had been ident-
ified as the household head in the main DHS survey. Also excluded
were households in which children were not de jure. The EdData
survey collected information on the age of children at the beginning
and end of their schooling, educational attainment and reasons for
non-attendance. From parents and guardians, the information on
their knowledge of UPE was collected. Adults were asked to give
their assessments of the qualities and failings of the schools in the
local area.4 Using the sample weights constructed for the second
survey, the sample is representative of Uganda as a whole. In total,
4,217 households were re-surveyed for the EdData survey, and the
information on the education of children was reported by 4,246
parents or guardians. The sample weights of the EdData survey,
which make the sample nationally representative, are employed
throughout these analyses.
   Given that UPE affected primarily children who were at risk of
attending government-aided schools in Uganda, it is of interest to
understand the basic structure of the education sector in the
country. Table 1 presents summary statistics on the percentage of
children attending different types of primary schools. Column 2
presents results for the full DHS sample. More than 85% of those
who have attended primary school attended government-aided
schools. The second most commonly attended school type is private
non-religious (10.5%), followed by private religious schooling
(3.8%). Columns 3 and 4 of Table 1 reveal, however, that private

4
    For more on the 2000 DHS and 2001 EdData surveys, the reader is referred to the
    website of DHS Macro, www.measuredhs.com
190   Louise Grogan

        Table 1: Types of Schools Attended by Children in 2000 DHS Survey


Financing of primary school      Full sample        Urban             Rural
                                % of students    % of students     % of students


Government aided                      83.87            62.44             91.78
Community                              1.19             0.04              1.62
Private non-religious                 10.50            28.02              4.03
Private religious                      3.82             8.90              1.94
Others                                 0.06             0.14              0.03
Missing                                0.56             0.46              0.60
Number of observations            10,496            2,830             7,666

Source: Author’s calculations using Uganda 2001 Educational Survey and DHS
2000 Main Survey. These data are calculated among the DHS subsample who have
attended school.

non-religious schooling is very concentrated in urban areas, where
28% of students attend such schools. In contrast, only 4% of chil-
dren in rural areas attend private non-religious schools. Similarly,
a majority of private religious education takes place in urban areas.
In rural areas, nearly 92% of children who have attended primary
schools attended the government-aided schools, for which fees
were eliminated in 1997. This suggests that the elimination of
school fees in government-aided schools might be expected to
have a greater impact in rural areas.
   Next we examine summary statistics for the sample of children
aged 5– 18 at the time of the EdData survey. These are presented
in Table 2. Columns 2 and 3 compare the individual and household
characteristics of children who have attended school with those
who have not. Slightly more boys than girls have attended
school. In this sample, children who have attended school are
slightly more likely than those who have not to have deceased
parents. This is likely because the children who have not attended
school are, on average, younger than those who have. In comparing
the presence of key household consumer durables across these two
groups, it is apparent that children who have not attended school
are also more economically disadvantaged. For example, about
8% of school attendees’ households have electricity, versus about
1% of non-attendees. Non-attendees are also far more likely to
                                      UPE and School Entry in Uganda      191

                      Table 2: Summary Statistics (Means)


                           Attended        No           Attended     Attended
                            school      schooling     government-     private
                                                      aided school    school


Female                      0.4872         0.5096        0.4873       0.4870
                            (0.005)       (0.018)       (0.005)      (0.014)
Year of birth               1998.48       1989.77       1988.48      1988.47
                            (0.038)       (0.241)       (0.040)      (0.117)
Mother dead                 0.0808         0.0423        0.0785       0.1036
                            (0.003)       (0.007)       (0.003)      (0.009)
Father dead                 0.1422         0.0766        0.1435       0.1294
                            (0.003)       (0.009)       (0.004)      (0.010)
Child of household head     0.7264         0.7946        0.7279       0.7117
                            (0.004)       (0.014)       (0.005)      (0.013)
Household size              7.7085         7.1870        7.7118       7.6758
                            (0.032)       (0.100)       (0.034)      (0.109)
No toilet                    0.1188        0.2602        0.1237       0.0690
                            (0.003)       (0.015)       (0.003)      (0.007)
Telephone                   0.0298         0.0007        0.0226       0.1020
                            (0.002)       (0.001)       (0.002)      (0.009)
Bicycle                     0.4933         0.4328        0.5105       0.3216
                            (0.005)       (0.017)       (0.005)      (0.013)
Motorcycle                  0.0416         0.0217        0.0396       0.0615
                            (0.002)       (0.005)       (0.002)      (0.007)
Truck                       0.0322         0.0138        0.0287       0.0674
                            (0.002)       (0.004)       (0.002)      (0.007)
Electricity                 0.0789         0.0130        0.0622       0.2454
                            (0.003)       (0.004)       (0.003)      (0.012)
No floor                     0.7862         0.9468        0.8181       0.4675
                            (0.004)       (0.008)       (0.004)      (0.014)
Mud walls                   0.5491         0.6829        0.5707       0.3335
                            (0.005)       (0.016)       (0.005)      (0.013)
Thatched roof               0.3129         0.5353        0.3289       0.1537
                            (0.005)       (0.018)       (0.005)      (0.010)
Urban                       0.1066         0.0428        0.0801       0.3699
                            (0.003)       (0.007)       (0.003)      (0.014)
Number of observations      10,496          807           9,275        1,221

Source: DHS Education Survey 2001.
Note: DHS Sample weights used. Standard errors in parentheses.
192     Louise Grogan

reside in rural areas. Whereas about 11% of attendees live in urban
areas, only about 4% of non-attendees do.
   It is also of interest to contrast the living circumstances of chil-
dren who attend private schools versus government-aided (state)
schools. This is done in columns 4 and 5 of Table 2. Here, the
term ‘private’ refers to all schools which are not government-aided,
including religious schools. To summarise, the greatest differences
among these two groups of children appear to be in housing quality
and consumer durables possessed by the household. In general,
children who attend private school are more likely to live in house-
holds with indoor toilets, telephones, floors and electricity. Private
school attendees are also far more likely to reside in urban areas
than those attending government-aided schools.
   Late school start in Uganda (particularly after age 8) is associated
with greater risks of dropping out before the completion of primary
school.5 In Table 3, the association between late school start and
primary school dropout is demonstrated for the 1982 and 1983
cohorts interviewed in the 2001 Uganda EdData Survey. These
cohorts were aged 10 and 11, respectively, by January 1994. This
implies that those who began school by age 10 should have finished
at least seven years of school at the time of the EdData interviews in
early 2001.
   Table 3 presents the probit marginal effects of regressions exam-
ining the probability that individuals born in 1982 or 1983 have
completed at least seven years of schooling at the time of the
EdData survey, conditional on having started school by age 10.6
About 87% of the sample had completed this amount of schooling.
Panel A presents results in which dummies are used to denote the
age at which the child began school. The reference age at school
start is 6, the normal age of school entry in Uganda. Column 2 pre-
sents results which do not control for individual or household
characteristics other than the child’s age at school start. Column 3
includes a full set of controls for these characteristics. The results
clearly show that school entry between ages 5 and 8 is not associated
5
    Education in Uganda follows the British system. The normal age for school entry
    in Uganda is 6, although many students begin at age 5 or 7. It consists of seven
    years of primary school, four years of secondary school (O-level), two years of
    upper secondary school (A-levels), followed by university. Before an individual
    can enter secondary school, he or she must successfully complete the Primary
    school Leaving Examination, or PLE.
6
    In the 2001 EdData, only 1.3% of school attendees began school after this age.
                                       UPE and School Entry in Uganda         193

Table 3: Probability of Completing At Least Seven Years of Schooling (Probit
                               Marginal Effects)


                                                                 jtj-Test equality
                                                                 coefficient


Panel A
  Start age 5                           20.0392     20.0276      0.15
                                         (0.063)      (0.041)
  Start age 7                           20.0233     20.0125      0.03
                                         (0.036)      (0.024)
  Start age 8                           20.0740     20.0599      0.17
                                         (0.065)      (0.052)
  Start age 9                           20.2480**   20.1721**    0.56
                                         (0.096)      (0.097)
  Start age 10                          20.3858**   20.2240**    1.05
                                         (0.109)      (0.108)
  Public school                                       0.0696**
                                                      (0.042)
  Female                                            20.0186
                                                      (0.023)
  Mother completed primary education                  0.1342**
                                                      (0.029)
  Father completed primary education                  0.1345**
                                                      (0.031)
  Mother dead                                         0.0063
                                                      (0.039)
  Father dead                                         0.0381
                                                      (0.023)
  Child of household head                             0.0301
                                                      (0.026)
  Number of boys in household                         0.0212**
                                                      (0.010)
  Number of girls in household                        0.0316**
                                                      (0.009)
  Household size                                    20.0169**
                                                      (0.005)
  Birth order                                       20.0028
                                                      (0.010)
  Other controls                        No          Yes
    Urban                               No          Yes

                                                           (continued on next page)
194    Louise Grogan

Table 3: Continued

                                                              jtj-Test equality
                                                              coefficient


    Regional dummies                  No           Yes
    Household wealth index            No           Yes
    Number of Observation             712          712
    Pseudo R 2                          0.0655       0.3094
    Obsered P-value                     0.8699       0.8699
    Predicted P-value                   0.8808       0.9415
Panel B
  Started before 9                      0.2628**    0.1551** 1.18
                                        (0.065)     (0.064)
  Other controls as in
      Panel A
    Number of                         712          712
      observations
    Pseudo R 2                           0.0589      0.3046
    Observed P-value                     0.8699      0.8699
    Predicted P-value                    0.8796      0.9397

Source: Author’s calculations using Uganda 2001 Educational Supplement and
DHS 2000 Main Survey.
Notes: Reference age at school start is six in Panel A.
*Significant at 10% level; **Significant at 5% level.


with differential probabilities of completing at least seven years of
schooling. However, beginning school at age 9 or 10 is associated
with strongly reduced probabilities.
   Clearly, there may be unobserved factors contributing both to
later ages at school entry and to higher dropout propensities
among children who enter school later. However, the coefficients
relating to these ages of school start tell a similar story whether
or not controls for other observable characteristics are included
(compare columns 2 and 3, and the t-tests of equality of coefficients
presented in column 4). Coefficients relating to the age at school
entry are statistically the same across specifications in which only
age at school start is controlled for and those in which all observa-
ble features of individuals, their households and fixed effects are
controlled for. Although this by no means proves that later school
entry causes dropout, it does strongly suggest that unobserved
                                          UPE and School Entry in Uganda        195

hetergeneity is not the main factor behind the observed association.
One would expect that, if unobserved heterogeneity were driving
these results, coefficients would have changed significantly with
the inclusion of such an extensive list of controls.7
   Panel B of Table 3 presents estimates in which dummies repre-
senting age at school start are now replaced by a single dummy indi-
cating whether or not the child began school before age 9. Column 2
presents results without controls for other observable characteristics
of the individual, while column 3 presents results which include a
full set of controls. As in Panel A, results are essentially the same
across the two specifications. Starting school before age 9 is associ-
ated with a 16– 26% increase in the probability that an individual
completes at least seven years of schooling.
   The results of Table 3 underline the importance of providing a
disaggregated analysis of the effects of school fee elimination on
school enrolment. If a majority of new school entrants are young,
the eventual effects on educational attainment in the population
will likely be very different from the case in which a majority of
them are older. This makes sense according to standard theories
about incentives to invest in education: as the opportunity cost of
being in school rises in age (labour market opportunities improve
in age), so do the incentives to drop out when families face
income shocks. Because people become stronger as they approach
adulthood, their productivity in physical labour rises. Children’s
ability to make up family income lost due to the sickness of an
adult in the household rises in age. It is possible that this relationship
has been somewhat altered in Uganda by the advent of UPE.
However, it is unlikely that the labour market factors which raise
the opportunity cost of time spent in school with age have changed
much. For this reason, the remainder of this paper focuses on
the effects of school fee elimination under UPE on the probability
of entering school before age 9.

7
    For example, child ability may be an unobservable driving the association
    between late entry and early dropout. It is argued that this cannot explain the
    association for the following reason. Assume that parental education were corre-
    lated with child ability, which seems reasonable. If unobservable differences in
    ability across children were actually driving the observed association between
    age at school entry and school dropout, including dummies for whether the
    mother and father have completed primary education should have significant
    effects on the coefficients of interest. As shown in column 4 of Table 3, this is
    not the case.
196   Louise Grogan

4. Fee Elimination and School Entry Ages
Did the elimination of school fees cause a discontinuity in the prob-
ability of a child entering school before his or her ninth birthday?
As has been demonstrated, an increase in this probability should
have positive implications for primary school completion rates.
Figure 1 provides a graphical presention of the results of a simple
regression of the probability that a child begins school before age
9, by year of birth. Marginal effects of a probit regression including
only year of birth and a gender dummy as regressors are graphed,
as is the 95% confidence interval.
   The results show that, relative to the reference birth year of 1982,
there were no significant differences in the probability that a child
born in 1983 through 1987 entered school before age 9. However,
those born between 1988 and 1992 experienced significantly
greater probabilities of entering school before this age than those
born in the reference year. Individuals born in 1988 would have
been eight years old when school fees were eliminated in January
1997. They were thus the first birth cohort who could have been
affected on this margin. However, to show that this association is




               Figure 1: Probability of School Entry Before Age 9
                                          UPE and School Entry in Uganda           197

a causal effect of school fee elimination under UPE, more controls
are clearly necessary.
   Table 4 presents the probit marginal effects of an examination of
the probability that an individual begins attending school before
age 9, by birth year. The oldest cohort in the sample was born in

Table 4: Probability of Beginning School before Age 9 by Birth Cohort (Probit Marginal
                 Effects: Probability of Beginning School before Age 9)


Born 1983                                         0.0080                  0.0077
                                                  (0.026)                 (0.026)
Born 1984                                         0.0377                  0.0363
                                                  (0.021)                 (0.022)
Born 1985                                         0.0442                  0.0406
                                                  (0.020)                 (0.020)
Born 1986                                         0.0217                  0.0234
                                                  (0.0210)                (0.021)
Born 1987                                         0.0296                  0.0287
                                                  (0.019)                 (0.019)
Born 1988                                         0.0884**                0.0873**
                                                  (0.015)                 (0.016)
Born 1989                                         0.0735**                0.0740**
                                                  (0.017)                 (0.018)
Born 1990                                         0.0961**                0.0980**
                                                  (0.015)                 (0.015)
Born 1991                                         0.1339**                0.1353**
                                                  (0.011)                 (0.011)
Born 1992                                         0.1202**                0.1215**
                                                  (0.013)                 (0.013)
Female                                           20.0081                 20.0047
                                                  (0.010)                 (0.010)
Mother completed primary education                0.0584**                0.0767**
                                                  (0.011)                 (0.010)
Father completed primary education                0.0616**                0.0794**
                                                  (0.010)                 (0.010)
Mother dead                                      20.0102                 20.0054
                                                  (0.017)                 (0.016)
Father dead                                       0.0072                  0.0051
                                                  (0.013)                 (0.013)
Child of household head                           0.0953**                0.0903**
                                                  (0.013)                 (0.012)
Number of boys in household                       0.0018                  0.0054
                                                  (0.005)                 (0.005)

                                                                (continued on next page)
198    Louise Grogan

Table 4: Continued
Number of girls in household                 0.0120**            0.0140**
                                             (0.005)             (0.005)
Household size                              20.0064**           20.0054**
                                             (0.002)             (0.002)
Birth order                                 20.0264*            20.0263**
                                             (0.014)             (0.005)
Other controls
  Urban dummy                                 Yes                 Yes
  Regional dummies                            Yes                 Yes
  Consumer durables dummies                   Yes                 No
  Household wealth index                      No                  Yes
  Observed P-value                            0.8241              0.8241
  Predicted P-value                           0.8484              0.8454
  Number of observations                      8,206               8,206
  Pseudo R 2                                  0.0958              0.0820

Source: Author’s calculations using Uganda 2001 Educational Supplement and
DHS 2000 Main Survey.
**Significant at 5% level; *Significant at 10% level.

1982 (the reference year) and the youngest in 1992. The results show
that individuals born in the years 1988 through 1992 have signifi-
cantly greater probabilities of beginning school before age 9 than
those in the reference group. Those born between 1983 and 1987,
in contrast, do not have different probabilities than the reference
group of beginning school before age 9.
   Like Figure 1, these results show that the elimination of school
fees under UPE caused discontinuity in the probability that an
individual attends school before age 9. An individual who was
born in, say, October 1988 would have been eight years old at
the onset of UPE in January 1997. The fact that those born in
this year and later have higher probabilities of entering school
before age 9 appears attributable to school fee elimination
under UPE. In the language of the programme evaluation litera-
ture, such children would have been eligible for the ‘treatment’,
which is entering school under UPE before his or her ninth birth-
day. In contrast, a child born in October 1987 would not have
been eligible for this treatment due to his or her earlier birth.
Year of birth is clearly exogenous, and while it is possible that
individuals born in different years have different characteristics,
these are now controlled for in the regression. The initial
                                         UPE and School Entry in Uganda       199

treatment effect of UPE can be identified by comparing the prob-
ability of entering school before age 9 among students born close
to the cutoff of January 1988.
   The results in Table 4 are also robust to the type of household
wealth controls used. Column 2 presents results using consumer
durables dummies and column 3 employs a linear term in the
DHS/World Bank-constructed household wealth index. This
household wealth index was created in cooperation with the
World Bank for the purpose of comparing living standards across
households in circumstances in which income measures are either
unavailable or unreliable. The index employs principal components
analysis to divide the population into quintiles of wealth distri-
bution on the basis of assets possessed by the household. Each
household asset, such as a car, radio, television or dwelling charac-
teristic such as type of roof or floor, is assigned a factor score. This
enters into a household’s total asset score. Households with higher
wealth index scores generally have more assets and better living
conditions than do those with lower wealth index scores.8
   As can be seen, results are essentially the same when the DHS/
World Bank wealth index is employed to control for household
wealth instead of the asset dummies.
   These findings suggest that regression discontinuity analysis is an
appropriate technique for identifying the magnitude of this discon-
tinuous effect. Regression discontinuity analysis is a programme
evaluation technique in which sample members are assigned to
treatment or control groups on the basis of a cutoff score, or pre-
programme measure. In these data, there was a discontinuous
increase in the probability of entering school by age 9 coincident
with the introduction of UPE. Although there may be a secular
trend in the probability of entering school by age 9 across people
of different birth years, it is the treatment cutoff in January 1998
which provides identification. Similar types of regression disconti-
nuity analyses have been used widely in the evaluation of edu-
cational programmes and school quality (see, for example,
Thistlethwaite and Campbell, 1960; Angrist and Lavy, 1999; Black,

8
    For more information on the DHS/World Bank household wealth index, the
    reader is refered to Filmer and Pritchett 1998 and Filmer 2001 , who show that
    this asset-based index compares well with expenditure data in measuring house-
    hold living standards.
200   Louise Grogan

1999). Where randomised trials are inappropriate, regression dis-
continuity design can provide robust identification.9
   There are two key variables used for identification in the
regression discontinuity estimates: a continuous, linear variable
for the year of birth (YBIRTH) and a dummy variable indicating
whether or not school fees had been eliminated before a child
attained the age of 9 (UPELEQ8). The first of these variables
(YBIRTH) accounts for any potential secular, linear trend across
cohorts in the probability of entering school before age 9.10 Given
that both the Ugandan government and UNICEF had been
working for several years to increase school enrolment by the
time school fees were eliminated, allowing such a secular trend
seems appropriate. The second variable, UPELEQ8, assigns
sample members to treatment and control groups on the basis of
the year of birth. While the assignment to the treatment or control
group might here be considered exogenous to the individual, con-
trols for observable personal and household characteristics are also
included in the regression.
   Formally, define yi as the age of school entry of a respondent in
the DHS EdData survey. In the probit model, the variable yi is not
observed. An observation rule defines the relationship between
the latent variable yi and the observed variable yi*.
   Given the observation rule, yi* takes the form:
          
            0 iff yà ¼ 0 Has not entered school by age 9;
    yà ¼             i                                           ð1Þ
      i     1 iff yà ¼ 1 Enters school at age 8 or earlier:
                     i


   Using the regression discontinuity estimator, the effects of being
in a UPE-affected cohort on the probability of school entry before
age 9 are identified by the UPELEQ8 term.
              yà ¼ b0 þ b1  ðYBIRTHÞ þ b2  ðUPELEQ8Þ
               i

                    þ l  ðCONTROLSÞ þ ei                                         ð2Þ

  Table 5 presents regression discontinuity estimates of the effect of
school fee elimination in January 1997 on the probability that a
9
   For a detailed analysis of the restrictiveness and robustness of regression discon-
   tinuity design as a programme evaluation mechanism, see, for example, Hahn
   et al. (2001).
10
    Note that this is not the case in the results presented in Table 4 or Figure 1.
      Table 5: Regression Discontinuity Estimates of the Probability of Beginning School Before Age 9 (Probit Marginal Effects)


                                  Full         Full      Females       Males        Urban        Rural        Lowest         Highest
                                sample       sample        only        only                                 50% wealth     50% wealth
                                                                                                               index          index



Year of birth, YBIRTH         0.0133**     0.0136**     0.0110**     0.0169**     0.0101**    0.0140**      0.0122**       0.0142**
                              (0.003)      (0.003)      (0.004)      (0.004)      (0.005)     (0.003)       (0.004)        (0.004)
UPE started   8, UPELEQ8      0.0328*      0.0309*      0.0513**     0.0108       0.0047      0.0342*,      0.0284**       0.0284**
                              (0.017)      (0.018)      (0.025)      (0.025)      (0.031)     0.0386**      (0.017)        (0.017)
                                                                                              (0.019)
Female                       20.0081     20.0049                                20.0039      20.0050     20.0002         20.0157
                              (0.010)     (0.010)                                (0.017)      (0.012)     (0.014)         (0.015)
Mother completed              0.0591**    0.0773**     0.0962**     0.0554**     0.0715**     0.0790**    0.0501**        0.0623**




                                                                                                                                          UPE and School Entry in Uganda
  primary education           (0.011)     (0.010)      (0.014)      (0.015)      (0.019)      (0.011)     (0.015)         (0.015)
Father completed primary      0.0632**    0.0809**     0.0797**     0.0837**    20.0032       0.0902**    0.0607**        0.0688**
  education                   (0.010)     (0.010)      (0.014)      (0.014)      (0.017)      (0.011)     (0.014)         (0.014)
Mother dead                  20.0112     20.0067       0.0172      20.0301      20.0564**     0.0018     20.0465**        0.0159
                              (0.017)     (0.017)      (0.021)      (0.025)      (0.028)      (0.018)     (0.025)         (0.022)
Father dead                   0.0076      0.0052       0.0047       0.0073       0.0354**     0.0002      0.0176         20.0024
                              (0.013)     (0.013)      (0.018)      (0.018)      (0.017)      (0.014)     (0.016)         (0.019)
Child of household head       0.0954**    0.0906**     0.0906**     0.0933**     0.0838**     0.0938**    0.0666**        0.1295**
                              (0.013)     (0.012)      (0.017)      (0.018)      (0.020)      (0.014)     (0.017)         (0.019)
Number of boys in household   0.0015      0.0052       0.0118*      0.0001       0.0064       0.0043      0.0155**       20.0108*
                              (0.005)     (0.005)      (0.006)      (0.007)      (0.007)      (0.005)     (0.007)         (0.006)
Number of girls in household  0.0119**    0.0139**     0.0169**     0.0100       0.0139**     0.0135**    0.0139**        0.0076
                              (0.005)     (0.005)      (0.007)      (0.007)      (0.007)      (0.005)     (0.006)         (0.007)
Household size               20.0064**   20.0054**    20.0104**    20.0005      20.0066**    20.0046*    20.0100**       20.0010
                              (0.002)     (0.002)      (0.003)      (0.003)      (0.003)      (0.003)     (0.003)         (0.003)




                                                                                                                                          201
                                                                                                               (continued on next page)
                                                                                                                                          202
                                                                                                                                          Louise Grogan
Table 5: Continued
                                   Full          Full       Females       Males         Urban       Rural       Lowest         Highest
                                 sample        sample         only        only                                50% wealth     50% wealth
                                                                                                                 index          index



Birth order                  20.0267       20.0267      20.0300**     20.0238**    20.0267      20.0307**   20.0085        20.0200
                              (0.024)       (0.024)      (0.006)       (0.007)      (0.024)      (0.005)     (0.019)        (0.021)
Other controls
  Urban dummy                   Yes          Yes           Yes          Yes           No          No          Yes            Yes
  Regional dummies              Yes          Yes           Yes          Yes           Yes         Yes         Yes            Yes
  Consumer durables             Yes          No            No           No            No          No          Yes            Yes
  dummies
  Household wealth index        No           Yes           Yes          Yes           Yes         Yes         No             No
  Number of observations        8,206        8,206         4,063        4,143         2,185       6,021       3,955          4,251
  Pseudo R 2                    0.0925       0.0788        0.0958       0.0698        0.0723      0.0793      0.0813         0.1182
  Observed P-value              0.8241       0.8241        0.8284       0.8201        0.8795      0.8177      0.8396         0.8115
Predicted P-value               0.8472       0.8442        0.8526       0.8379        0.8959      0.8379      0.8604         0.8400

Source: Author’s calculations using Uganda 2001 Educational Survey and DHS 2000 Main Survey.
**Significant at 5% level, *significant at 10% level.
                                  UPE and School Entry in Uganda   203

child entered school before age 9. The reported coefficients are the
marginal effects of a probit regression in which the dependent vari-
able is a binary variable indicating that a child entered school at age
8 or earlier. Results for the full sample are presented in columns 2
and 3. Column 2 uses dummy variables to control for household
wealth, whereas column 3 includes a linear term in the DHS/
World Bank wealth index. Results are similar across the two speci-
fications. Both indicate that the discontinuous, positive effect of
school fee elimination on the probability of entering school before
age 9 is about 3%. As well, the coefficients on the YBIRTH term
show a statistically significant secular trend in this probability
across the cohorts included in the DHS.
   The effects of school fee elimation likely differed across popu-
lation subgroups. Columns 4 through 7 of Table 5 present results
which disaggregate the sample by gender and area of residence.
In column 4, results are presented for females, and in column 5
for males. Comparing the two estimates, it is apparent that the
effects of school fee elimination on this margin are concentrated
on females. Among girls for whom school fees were eliminated
before the ninth birthday, the probability of entering school
before this age is 5% higher. No such effect of UPE is found
for boys.
   Columns 6 and 7 of Table 5 present results for children in urban
and rural areas, respectively. In both areas, a positive secular trend
is observed across cohorts. The probability of entering school before
age 9 is increasing significantly over time. However, the effect of
school fee elimination appears to be concentrated in rural areas.
In rural areas, which comprise two-thirds of Uganda’s population,
a 3.4% increase in the probability of attending school before age 9 is
associated with the advent of UPE in January 1997. In urban areas,
no significant jump in this probability is associated with the intro-
duction of UPE. There are several possible explanations for the
observed lack of effect of UPE on the probability of a child entering
school before the ninth birthday in urban areas. This may reflect
the fact that school enrolment in urban areas was much higher
than in rural areas before the elimination of school fees. There is
also some anecdotal evidence that urban schools continued to
charge fees after 1997.
   The final two specifications in Table 5 divide the sample at 50th
percentile of the household wealth index. The goal is to ascertain
204     Louise Grogan

whether or not the introduction of UPE had differential effects on
the probability of a child entering school before age 9 across the
wealth distribution. The results provide evidence of a greater
effect of the introduction of UPE on the probability of children
from a poorer household entering school before age 9. The coeffi-
cient on UPELEQ8 is larger for the poorer 50 percentile than for
the richer 50 percentile (0.0386 versus 0.0284), and a t-test of the
equality of coefficients rejects the null hypothesis at the 10% level
(jtj ¼ 4.24). This finding is consistent with the finding of Deininger
(2003), that UPE acted to reduce the gap in school in attendance
across socioeconomic groups.
    The regression discontinuity specification presented in Table 5
assumes that any secular trend across birth cohorts in the prob-
ability of a child entering school before age 9 is linear. In fact, this
assumption is more strict than is necessary for the identification
of effects using regression discontinuity (Hahn et al., 2001).
However, more flexible specifications allowing for a quartic trend
in the date of birth result in statistically unchanged coefficients
of the variable of primary interest, UPELEQ8.11 This is likely
because the data span only 11 different birth years, so trends
across cohorts are adequately captured by the linear YBIRTH term.
    The presence of information on school type in the EdData survey
provides an opportunity to check the identification strategy
employed above. The results of Table 5 suggest that school fee elimi-
nation mainly affected children whose financial backgrounds put
them on the margin for school attendance before January 1997.
The elimination of school fees at the primary level in January 1997
was undertaken only in government-aided schools. In general, chil-
dren who attend private schools come from wealthier backgrounds,
as was seen in Table 2. If the identification strategy used in Table 5 is
indeed valid, fee elimination in government-aided schools should
have had a negligible impact on children whose families could
afford private schooling. That this is indeed the case in the DHS
EdData is shown in Table 6.

11
     These specifications are not presented here but are available on request from the
     author. Given the lack of difference in coefficients on UPELEQ8 across specifica-
     tions employing a linear trend in YBIRTH and a quartic in YBIRTH, it is unlikely
     that allowing for a fully non-parametric specification would yield different
     results. See, for example, Ludwig and Miller (2007) for recent work employing
     non-parametric regression discontinuity estimators.
Table 6: The Effect of School Fee Elimination on Private School Entry Probit Marginal Effects: Probability That School
                                              Attended Is Private—includes unschooled


                                  Full        Females         Males          Urban          Rural          Lowest            Highest
                                sample          only          only                                       50% wealth        50% wealth
                                                                                                            index             index



Year of birth, YBIRTH       20.0014        20.0041*        0.0014         0.0002        20.0016        20.0023           20.0004
                             (0.002)        (0.002)        (0.002)        (0.007)        (0.002)        (0.002)           (0.002)
UPE   8, UPELEQ8             0.0102         0.0230        20.0042        20.0122         0.0118         0.0096            0.0090
                             (0.010)        (0.014)        (0.014)        (0.041)        (0.009)        (0.012)           (0.012)
Sex                          0.0041                                       0.0199         0.0051         0.0008            0.0106
                             (0.006)                                      (0.025)        (0.006)        (0.006)           (0.008)




                                                                                                                                         UPE and School Entry in Uganda
Mother dead                  0.0187**       0.0200         0.0158         0.0033         0.0196**       0.0137            0.0203
                             (0.010)        (0.014)        (0.014)        (0.033)        (0.011)        (0.012)           (0.014)
Father dead                 20.0140**      20.0205**      20.0076        20.0019        20.0148**      20.0139**         20.0121
                             (0.006)        (0.008)        (0.009)        (0.030)        (0.006)        (0.006)           (0.009)
Related to household head    0.0061         0.0040         0.0091        20.0108         0.0068         0.0078            0.0032
                             (0.006)        (0.008)        (0.008)        (0.026)        (0.006)        (0.007)           (0.008)
Number of boys in           20.0044*       20.0085**      20.0004         0.0032        20.0033        20.0084**          0.0017
 household                   (0.003)        (0.004)        (0.003)        (0.011)        (0.003)        (0.003)           (0.004)
Number of girls in          20.0098**      20.0047        20.0155**      20.0102        20.0103**      20.0133**         20.0062
 household                   (0.003)        (0.004)        (0.004)        (0.010)        (0.003)        (0.003)           (0.004)
Household size               0.0035**       0.0042**       0.0027*        0.0102**       0.0020         0.0076**         20.0016
                             (0.001)        (0.002)        (0.002)        (0.005)        (0.001)        (0.002)           (0.002)
Other controls
  Urban dummy                     Yes            Yes           Yes            No             No              Yes              Yes




                                                                                                                                         205
                                                                                                              (continued on next page)
                                                                                                                                         206
                                                                                                                                         Louise Grogan
Table 6: Continued
                                   Full        Females          Males          Urban          Rural          Lowest           Highest
                                 sample          only           only                                       50% wealth       50% wealth
                                                                                                              index            index



  Consumer durables                yes            Yes            Yes            Yes             Yes            Yes               Yes
  dummies
  Pseudo R 2                   0.2085         0.2257          0.2043         0.2166         0.1372         0.3182           0.1747
  Observed P-value             0.0803         0.0826          0.0782         0.2817         0.0568         0.0798           0.0807
  Predicted P-value            0.0467         0.0453          0.0458         0.2204         0.0369         0.0352           0.0472
  Number of observations       8,206          4,063           4,143          2,185          6,021          3,955            4,251

Source: Author’s calculations using Uganda 2001 Educational Survey and DHS 2000 Main Survey.
Note that only students born before 1993 are included in the estimation sample. Birth order and regional dummies are included.
**Significant at 5% level, *significant at 10% level.
                                  UPE and School Entry in Uganda   207

   Table 6 presents marginal effects from probit estimation of the
probability that a child attends a private school. Both school atten-
dees and non-attendees are included in the sample. Estimates for
the full sample and for each of the subsamples of Table 5 are pre-
sented. To summarise, it is found that there is no discontinuity in
the probability of attending private school associated with school
fee elimination. This is true for the groups whose probability
of attending school before age 9 increased under UPE: poorer
students, girls and those in rural areas. This finding suggests that
school fee elimination affected only children who were sufficiently
financially disadvantaged to not have been at risk of attending
private school, either before or after the elimination of fees in
government-aided schools. It also suggests that UPE did not
initially lead to significant numbers of parents switching children
from public to private schooling because of fears about the
quality of public education under UPE.
   The findings in Table 6 may be also related to anecdotal evidence
that fee elimination was not, in fact, undertaken in urban areas. Most
private schools are located in urban areas and the incentives to
attend them would not have changed much if UPE were not actually
being implemented. Resources per student would not have dimin-
ished much in government-aided schools if few new students
attended school and they were still collecting fees. In rural areas,
there are very few private schools, so even parents worried about
quality under UPE would not have had many local alternatives.
One might expect that over time new alternatives to overcrowded
government-aided schools will emerge if some parents are willing
to pay more. However, these results pertain only to the first four
years of UPE, and thus cannot inform us about the longer term
effects of UPE on the private education sector in Uganda.

5. Conclusions
Overall, the results of this paper suggest significant positive effects
of school fee elimination on the timely enrolment of girls and chil-
dren living in rural areas of Uganda. Given the strong historical
association between age at school start and retention in schooling
in Uganda, school fee elimination should promote the completion
of primary education among these two disadvantaged groups.
The results of this analysis show that looking at gross or net
208   Louise Grogan

enrolment in primary education may provide a very limited picture
of both the quantitative effects of the elimination of school fees and
of the effects specific to socioeconomic groups.
   Clearly, however, fee elimination under UPE changed more than
ages at school entry. In all countries which eliminated fees, including
Uganda, the sudden increase in enrolment led to shortages of tea-
chers and textbooks. Wherever UPE was instituted, large fractions
of new students in the first year were adults or far above the
normal age at school entry. Classrooms became overcrowded, some-
times necessitating multiple school ‘shifts’ during the day. These
factors are likely to have had negative effects on retention in school-
ing. Nakibuuka (2004) reports that 2003 registrations for the Primary
School Leaving Examination, which should have been written then
by the first UPE cohort, were far below the levels that UPE enrolment
figures would predict. A Poverty Elimination Action Plan revision
paper, written by Uganda Ministry of Education (2003) found that
only 33% of the 1997 UPE cohort had reached Primary 6 by 2002,
and only 22% had reached Primary 7 by 2003. Massive investments
in teacher education, textbooks and school construction do,
however, appear to have markedly improved the resources available
to students and teachers in more recent UPE cohorts.
   There is a clear need to gather longitudinal data at the individual
level on the educational trajectories of students and on the quality
of learning outcomes under UPE. In order to ascertain the specific
effects of the elimination of school fees on AIDS orphans, a large
and particularly disadvantaged socioeconomic group in Uganda,
data must be collected which include specific questions on the
timing of parental sickness and death. There is also a need for
survey data which permits an examination of the effects of the
elimination of school fees on the resources available at the school
level in Uganda. With data of this nature, it would be possible to
provide a comprehensive picture of the costs and benefits of
school fee elimination in Uganda.



Acknowledgements
The author is grateful for helpful comments of participants at the
2005 Centre for the Study of African Economies (CSAE)
Conference in Oxford, to the editors and referees of this Journal
                                 UPE and School Entry in Uganda   209

for very helpful comments and to ORC Macro for making the
Demographic and Health Survey (DHS) data and Education
Survey (EdData) for Uganda publicly available.


Funding
The author is grateful to the Social Sciences and Humanities
Research Council of Canada (SSHRC) for financial support.


References
Angrist, J. and V. Lavy (1999) ‘Using Maimonides’ Rule To Estimate
   the Effect of Class Size on Scholastic Achievement’, Quarterly
   Journal of Economics, 114 (2): 533– 75.
Black, S. (1999) ‘Do Better Schools Matter? Parental Valuation of
   Elementary Education’, Quarterly Journal of Economics, 114 (2):
   577– 99.
Card, D. and A. Kreuger (1990) ‘Does School Quality Matter?
   Returns to Education and the Characteristics of Public Schools
   in the United States’, NBER Working Paper No. 3358.
Case, A. and A. Deaton (1999) ‘School Inputs and Educational
   Outcomes in South Africa’, Quarterly Journal of Economics, 114
   (3): 1047– 84.
Deininger, K. (2003) ‘Does Cost of Schooling Affect Attendance by
   the Poor?’ Economics of Education Review, 22 (3): 291– 305.
Epple, D. and R. Romano (1998) ‘Competition between Private and
   Public Schools, Vouchers, and Peer Group Effects’, American
   Economic Review, 88 (1): 33 –62.
Filmer, D. and L. Pritchett (1998) ‘Estimating Wealth Effects without
   Expenditure Data or Tears: With an Application to Educational
   Enrollments in States of India’, World Bank Policy Research
   Working Paper No. 1994.
Filmer, D. and L. Pritchett (2001) ‘Estimating Wealth Effects without
   Expenditure Data—or Tears: An Application to Educational
   Enrollments in States of India’, Demography, 38 (1): 115– 32.
Hahn, J., P. Todd and W. van der Klaauw (2001) ‘Identification and
   Estimation of Treatment Effects with a Regression Discontinuity
   Design’, Econometrica, 69 (1): 201 –9.
210   Louise Grogan

Hanushek, E. (1986) ‘The Economics of Schooling: Production and
   Efficiency in Public Schools’, Journal of Economic Literature, 24 (3):
   1141–77.
Hanushek, E. (1997) ‘Assessing the Effects of School Resources on
   Student Performance: An Update’, Educational Evaluation and
   Policy Analysis, 19 (2): 141– 64.
Heckman, J., L. Lochner and C. Taber (1998) ‘General Equilibrium
   Treatment Effects: A Study of Tuition Policy’, American
   Economic Review, 88 (2): 381 –6.
Ichimura, H. and C. Taber (2002) ‘Semiparametric Reduced Form
   Estimation of Tuition Subsidies’, American Economic Review, 92
   (2): 286– 92.
Kane, T. (1995) ‘Rising Public College Tuition and College Entry:
   How Well Do Public Subsidies Promote Access to College?’
   NBER Working Paper No. 5164.
Ladd, H.F. (2002) ‘School Vouchers: A Critical View’, Journal of
   Economic Perspectives, 16 (4): 3–24.
Ludwig, J. and D. Miller (2007) ‘Does Head Start Improve Children’s
   Life Chances? Evidence from a Regression Discontinuity Design’,
   Quarterly Journal of Economics, 122 (1): 159–208.
Ministry of Education and Sports of Uganda (2005) ‘Ministerial
   Seminar on Education for Rural People in Africa: Policy
   Lessons, Options and Priorities’, Ministerial Seminar, Addis
   Ababa, Ethiopia, 7 –9 September 2005.
Nakibuuka, S. (2004) ‘We Are Not Benefiting from UPE’, The
   Monitor (Kampala), 13 October.
Neal, D. (2002) ‘How Vouchers Could Change the Market for
   Education’, Journal of Economic Perspectives, 16 (4): 25– 44.
Ojoo, Z. (2005) Public Expenditure Tracking Survey Uganda,
   http://info.worldbank.org/etools/docs/library/152802/.
Reinikka, R. and J. Svensson (2004) ‘Local Capture: Evidence from a
   Central Government Transfer Program in Uganda’, Quarterly
   Journal of Economics, 119 (2): 679– 705.
Reinikka, R. and J. Svensson (2005) ‘The Power of Information:
   Evidence from a Newspaper Campaign to Reduce the Capture
   of Public Funds’, mimeo, World Bank.
                                UPE and School Entry in Uganda   211

The Uganda Millenium Development Goal Report (2003) MDG
  Reports, United Nations Development Group.
Thistlethwaite, D. and D. Campbell (1960) ‘Regression
  Discontinuity Analysis: An Alternative to the Ex Post Facto
  Experiment’, Journal of Educational Psychology, 51 (6): 309– 17.
Uganda Bureau of Statistics and ORC Macro International (1988a)
  Uganda DHS Educational Supplement Education Profile, www.
  measuredhs.com.
Uganda Bureau of Statistics and ORC Macro International (1995b)
  Uganda DHS Educational Supplement Education Profile. www.
  measuredhs.com.
Uganda Ministry of Education (2003) Poverty Elimination Action
  Plan, Uganda Ministry of Education Sector Revision Paper.
Uganda Ministry of Education and Sports (1999) The Ugandan
  Experience of Universal Primary Education, Government of the
  Republic of Uganda.
United Nations (2000) ‘Millenium Development Goals’, Millenium
  Summit of the United Nations.
World Bank (1993) Uganda Growing Out of Poverty, World Bank
  Country Study, http://www.worldbank.org/data/countrydata/.

				
DOCUMENT INFO