Does full-day schooling reduce educational inequality in Vietnam?
Tran Ngo Minh Tama* and Laure Pasquier-Doumerb
a
Centre for Analysis and Forecasting, Vietnam Academy of Social Sciences, Hanoi,
Vietnam
Postal address: Center for Analysis and Forecasting, 14th Floor, 1 Lieu Giai, Ba Dinh,
Hanoi, Vietnam
Tel: (84) 04 62730475/ Mobile: 84-9 1228 1922
Email: [email protected]
b
IRD, UMR 225 DIAL, Paris, France
Université Paris-Dauphine, LEDa-DIAL, Paris, France
Postal address: DIAL, 4 rue d’Enghien, 75010, Paris
Tel : +33 1 53 24 14 50
Email: [email protected]
*
Corresponding author. Email: [email protected]
Does full-day schooling reduce educational inequality in Vietnam?
Abstract: Full-day schooling has been implemented in Vietnam to deal with
the lack of instructional time in primary education in regard to international
standards. Yet full-day schooling may impact educational inequality by filling the
gap in terms of instructional time between children with different family
background. Indeed, well-off families use to compensate the lack of instructional
time by private tutoring. This paper investigates whether the huge development of
full-day schooling in recent years can improve equality of opportunity in Vietnam
by using the data from the Young Lives School Survey 2011-2012. Firstly, it
examines the variation in instructional time and school resources between pupils
with and without full-day schooling from different social background. It then
investigates the learning progress associated with full-day schooling across social
background. Evidence implies that full-day schooling cannot narrow the inequality
of opportunity in learning achievement at this stage of implementation. On the
contrary, it contributes to rise the gap in learning progress among children of
different social background.
Keywords: full-day schooling; education; inequality of opportunity; learning
progress; social background; Vietnam
1. Introduction
Shortages of financial and human resources have constrained many developing
countries to limit the instructional time to half of the day. With economic development and
the achievement of universal enrolment in primary education, many of these countries have
implemented policies to lengthen the instructional time in order to reach international
standards for education. Vietnam belongs to these countries. Although the full-day
schooling policy has been decided very early in Vietnam (1956), its implementation
remained marginal until the 2000’s and has accelerated since 2009 with the School
Education Quality Assurance Program.
By lengthening the instructional time, full-day schooling is expected to improve the
learning achievement of the children. This is confirmed by most of the studies, although the
observed effect of increasing the instructional time on achievement is often small (see
Holland et al. 2015 or Orkin 2013 for a review). Pieces of evidence show for Vietnam a
strong association between full-day schooling and achievement (Tran, 2014; World Bank,
2011), although no study has addressed the causal inference.
Expanding the instructional time has also been proposed as a policy option to reduce
the inequality of opportunity, especially in developed countries like USA (Biernat, 2011).
Inequality of opportunity is defined here as the association between educational
performance and family background. The argument putting forward is that this policy
allows children with disadvantaged social background to narrow the gap in instructional
time compared to well-off children. Indeed, well-off families are able to compensate the
lack of instructional time by private tutoring and extra-classes. In addition, poorest families
rely mostly on schools to provide academic learning for their children; in contrast to
children from well-off or middle-class families, who rely on school for only a portion of
their learning and are more likely to receive intellectual stimulation during the time they
spend at home. By substituting time spent at home with time at school, children from
2
different social background will benefit from a more similar learning environment.
However, narrowing the gap in instructional time for children with different social
background will reduce inequality of opportunity in learning achievement only if the
quality of instructional time is the same for all groups of children.
Yet, very few studies have analyzed the effect of expanding the instructional time on
inequality of opportunity, particularly in developing countries, and these studies reach
different conclusions. Looking at the association between social background and
performance in mathematics in 54 countries, Schutz et al. (2008) found no significant
difference in the equality of opportunity between half-day and whole-day school systems.
Their study relies on comparable data from the Third International Mathematics and
Science Study (TIMSS). Due to data constraints, their study presents some limitations. In
particular, the measurement of social background that is the number of books at home is
simplistic. In addition, they are able to control for unobserved country heterogeneity but
only when it does not vary across time as they used country fixed effects.
Hindicapie (2016) shows that switching from a half school day to a full school day
improves more the achievement of the lowest-income pupils in Colombia and therefore
reduces the inequality of opportunity in learning achievement. Their demonstration relies
on an identification strategy that exploits within school variation in the length of the school
day which is supposed to be exogenous. Using PISA 2006 data from over 50 countries,
Lavy (2010) found also a greater effect of instructional time on academic achievement for
pupils from low socioeconomic status families. The identification of the effect of
instructional time is based on a comparison of the performance of the same pupils in
different subjects in order to overcome potential selection and endogeneity problems. Lavy
(2012) came to the same conclusion for Israel by using another identification strategy
which relies on natural experiment that provided an opportunity to exploit a sharp
change in the method of funding schools in Israel.
Finally, Orkin (2013) study the effect of a large switch to full-day schooling in a
developing country - Ethiopia. The causal inference is made by using a natural experiment
that is the variation in the time of implementation of the reform by the districts. This
variation is supposed to be exogenous. Her result differs from other study as she concluded
to an increase of inequality of opportunity: effects of full-day schooling are larger among
better-off children defined as children who are not stunted, children from richer households
and children in urban schools.
This paper aims at contributing to this question by tackling the case of Vietnam, a
developing country with particularly interesting context. First, inequality of opportunity in
education is a raising concern for the country. Inequality in education across socioeconomic groups or across ethnic groups has increased during the last decades, especially
in secondary education attainment (World Bank 2011). Second, the implementation of fullday schooling in Vietnam presents some particularity that may change the effect of full-day
schooling on inequality of opportunity. The participation of the families in funding the fullday schooling policy is far from being negligible. This may impede the positive impact of
this policy on equality of opportunity. On the other hand, Vietnam has conducted an
ambitious policy from 2009 to 2015 to support schools in the most disadvantaged provinces
in their switch to full-day schooling.
3
This paper therefore examines whether full-day schooling is associated with higher
equality of opportunity in educational achievement. Full-day schooling reduces the
inequality of opportunity in learning achievement only if the pupils with disadvantaged
social background benefit more from this policy than their well-of counterparts in terms of
learning time and resources at school. To test it, we inspect the variations in instructional
time and school resources between pupils with and without full-day schooling from
different social background. Data from the School Survey 2011 collected under the Young
Lives Project in Vietnam are used. The learning progress associated with full-day schooling
across social background is investigated by estimating value-added models with the
inclusion of interaction terms between social background and full-day schooling. Past
educational inputs and unobserved characteristics of the pupils are partially captured by the
lagged achievement, but unobserved time-varying characteristics are not controlled for.
Thus, this paper provides an array of evidence on the relationship between full-day
schooling and inequality in learning progress across social backgrounds rather than causal
inferences.
The results show that full-day schooling only partially narrows down the gap in
instructional time of students from different level of social background. Pupils from welloff families attend both full-day schooling and extra classes, and poorest children have still
less access to full-day schooling. In addition, while full-day-schooling implementation is
associated with higher level resources at the school and the class levels, resources are
unequally distributed across students benefiting from the full-day-schooling and with
different social background. The higher the level of social background of the students, the
higher the resources they have access to. Finally, full-day-schooling is not associated with
better learning progress, especially for children with disadvantaged social background. This
evidence depicts very minor contribution of FDS in the reduction of inequality of
opportunity in learning achievement. It even seems to contribute to this inequality.
The paper is organized as follows. Section 2 provides an overview of the
implementation of full-day schooling in Vietnam. The data and the definitions used in the
analysis are described in Section 3. Section 4 scrutinizes whether full-day schooling
reduces the instructional time between students with different social background. The
distribution of additional resources devoted to full-day schooling across social background
is analyzed in section 5. Section 6 analyses whether the effect of full-day schooling on
learning achievement varies according to the social background of the children. Section 7
summarizes the results and presents their policy implications.
2. Full-day schooling in the context of Vietnam
Instructional time in Vietnamese schools lags far behind the one of advanced
economies but other Southeast Asian countries as well (Figure B2 and B3 in Appendix).
The time constraint does not allow teachers for ensuring the basic primary curriculum.
Concurrently private tutoring has been widespread, with the share of pupils in primary
school that attended extra classes reaching 35 percent in 20122, increasing the instructional
time of the well-off children. Being aware of these disparities, the Government has been
2
Authors’ estimations from the Vietnam Household Living Standards Survey 2012
4
promoting the transition to full-day schooling with a minimum national standard of 35
periods per week by 2025, each period including from 35 to 40 minutes of instructional
time (Vietnam Education Development Strategic Plan 2008-2020). This means an increase
of at least 40 percent in instructional time to shift from half-day schooling to full-day
schooling at the primary level.
Since the first introduction in 19563, full day schooling in primary education has been
implemented in Vietnam spontaneously and mostly in the urban areas until the 2000’s. The
adoption of full-day schooling is decided at the school level with limited support from the
government. Local authority and parents share the cost of infrastructure and additional
salary of teachers induced by full-day schooling. In addition, full-day schooling is either
applied to the whole school or some classes in the school, and the extension of learning
time could vary to reach from 30 to 40 periods per week, depending on the willing and the
resources of the schools. Consequently, the first schools which lengthened the learning time
were those where parents were willing and able to pay extra costs. The proportion of pupils
attending full-day schooling in poorest districts has remained very low until 2009. It was 30
per cent in 2008 compared to 59 per cent nationwide (World Bank, 2011).
In late 2009, the Government inaugurated 4 the School Education Quality Assurance
Program 5 (SEQAP) to enhance the quality of primary education by supporting the
transition to full-day schooling of primary schools. The program lasted for 6 years (till
2015) and has been implemented in the 36 most disadvantaged provinces of the country out
of 63 provinces. Around 1770 schools have benefited from this program. The total cost of
SEQAP has been 186 million USD. For all schools moving to full-day schooling, SEQAP
subsidized the training and the additional working time of teachers and school managers.
The support was concentrated on the schools that had 23 periods of instructional time per
week only and wished to shift to full-day schooling, especially in the communities of ethnic
minorities. According to the Ministry of Education and Training, the number of schools
switching to full-day schooling has been growing quickly, with 73 per cent of pupils were
attending full-day schooling in primary school in 2012, and 70 per cent of pupils in the
poorest districts (Bodewig et al., 2014). The great extension of full-day schooling in recent
time, especially for most disadvantaged children, calls for the investigation of its
relationship with educational inequality.
3. Data and definitions
The data source used in this paper is the School Survey conducted in Vietnam during
the school year 2011-2012. This is the dedicated survey of the Young Lives Project, a
longitudinal study of childhood poverty in Ethiopia, Peru, India and Vietnam based on
3
Decree No. 596-NĐ dated 30 August 1956 by the Ministry of Education on issuing the Regulation for
the 10-year general education enabled the full-day schooling for students in schools having enough number of
classrooms, with no further guidance for the implementation of full-day schooling.
4
Decision No. 12/QĐ-BGDĐT dated 4 January 2010 by the Ministry of Education and Training on
Approval of the content and decision on investment in the SEQAP; Decision No. 483/QĐ-BGDĐT on
approval of Handbook on guidance of implementation and Handbook on financial administration of SEQAP.
5
A multi-donor grant funded program which is co-financed and co-implement by the Government of
Vietnam and its partners (World Bank, DFID and Belgium).
5
household surveys of 12,000 children born in 1994-1995 (Older cohort) and 2001-2002
(Younger cohort) in 15 years. Young Lives in Vietnam follows 2000 children in Younger
cohort and 1,000 in Older cohort.
Although the sampling of the School Survey is not representative of the population in
Grade 5 in Vietnam and is pro-poor biased, it covers the diversity of children in the country
in a wide variety of attributes and experiences. Indeed, the sampling of the School Survey
relies on the one of the Young Lives longitudinal surveys. In these surveys, a sentinel-site
sampling design is employed, comprising twenty purposely selected sites chosen to
represent diversity, but with a pro-poor bias (Nguyen, 2008). These sites are clustered
within five provinces6, each province contains four sites, and each site is formed of one or
two communes, totaling 36 communes in all. The selection of sentinel sites followed a
complex consultative process with a pro-poor selection rule based on the poverty ranking of
each commune. At the site level, children in the Younger cohort were selected randomly in
2001, so that the data are representative of the birth cohort in each site. On this basis, any
school that was attended by at least one child of the Younger cohort in Grade 5 in 2011 was
included in the school survey sample. The sample of each surveyed child was then
expanded by adding 20 of his/her classmates randomly selected for the purpose of school
and class level analyses. The School Survey sample consists of 3,284 Grade 5 pupils in 176
classes of 52 schools.7
The Young Lives School Survey 2011 is very suitable source of data to investigate the
effect of full-day schooling on inequality of opportunity as it provides rich information on
instructional time and on characteristics of schools, principals, teachers and students.
Indeed, in addition to students, the survey also interviews their school principals and head
teachers. More importantly, students’ academic achievement recorded at two points in time
allows the analysis of student’ academic progress.
The data are used for the analysis based on the core definitions as follows:
Full-day schooling (FDS)
According to the data, the availability of FDS to a student could be determined in four
ways. Indeed, the school principal, the head teacher and the student him/herself have been
directly asked whether FDS is available at the school level, at the class level or for the
student respectively. The fourth way to define FDS is to use the number of periods that
students have in one week according to the head teacher.8 We retain this last variable for
several reasons. First, FDS has been implemented either for a whole school, or some
selected classes, or even for some students in one class. Thus, positive answer of the
6
Ben Tre, Da Nang, Hung Yen, Lao Cai and Phu Yen
The Young Lives Younger Cohort children included in the School Survey (1,138 children) constitute
57 per cent of the Younger Cohort sample as a whole. For an analysis of the selection bias among Young
Lives Younger Cohort due to the fact that those children who were not studying in Grade 5 in the selected
sites in 2011–12 are not included in the School Survey, see Rolleston et al. 2013. Due to higher probability of
late enrollment or repetition of the most disadvantaged children, the School Survey sample slightly overrepresents more advantaged children in terms of wealth of the Young Lives Younger Cohort. However, this
selection bias is compensated by the selection of the classmates.
8
For each day of the week, the head teacher has been asked “How many periods does this grade 5 class
have on each of these days?”
7
6
principal on the availability of FDS in the school may hide various situations across classes
and do not guarantee that the student into consideration benefits from FDS. To a less
extent, the same applies for the teacher’s answer. The student’s answer to the question “Do
you attend full-day schooling (two shifts- morning and afternoon)?” can be biased if the
student confuses FDS with extra-classes organized by the teacher in the classroom. In
addition, these three variables based on direct declaration do not provide a definition which
guarantees that students with FDS have longer instructional time than students without
FDS. Thus, FDS is defined according to the class timetable reported by the head teacher: a
student is considered as attending FDS if the head teacher declares that the student benefit
from at least 30 periods per week. The threshold of 30 periods has been retained firstly
because it is the minimum required by the government to be considered as full-day
schooling. Second, it is the most consistent with the common declaration of the head
teacher and the principal. 9 With this definition, 40.5 percent of students benefit from
FDS.10
Extra classes
Extra classes refer to classes that students take in addition to their formal schooling
day. Responses by principals in the survey indicate there were less than 4 percent of the
schools which arranged non-compulsory additional classes by the same teacher who taught
during the normal schooling day. Head teachers only gave information on their private
tuition in general rather than for the students in the class. Hence this study defines extra
classes according to the pupils’ declaration on whether they attend unofficial/noncompulsorily extra classes. 11 There are 39 percent of the students who attended extra
classes.
Learning achievement
Student academic achievement is measured by test scores in mathematics and
Vietnamese recorded at two points in time, one at the beginning and the other at the end of
9
According to this alternative criterion, FDS is defined when the head teacher’s answer to the question
“Do the pupils in your class attend full day schooling?” is “Yes, all pupils attend full day schooling” and the
principal’s answer to the question “Does your school offer Full-day schooling to any pupils in Grade 5?” is
“yes, all pupils”. When the two criteria are compared, the exclusion errors of the criteria based on 30 periods
includes 14% of the students and the inclusion errors only 5%. When the threshold is increased to 35 periods
(the targeted objective of the government), exclusion errors include 25% of the students.
10
This proportion differs from the 73% of pupils attending full-day schooling in primary schools
according to the Ministry of Education and Training for three main reasons: (i) School Survey sample is propoor biased as mentioned in the description of the data; (ii) only Grade 5 pupils are considered in the School
Survey and not all the pupils of primary school as in MOET figures; (iii) definition of full-day schooling may
be different in the two sources of data because of the wide variation in FDS practice and of the unclear
distinction between formal full-day schooling and informal extra classes.
11
For information on extra classes, the principals are asked “Does your school offer non-compulsory
additional classes available outside of the normal school day, provided at an additional charge to any students
who wish to participate, on the premises?” and “If yes, are these additional classes taught by the same
teachers who teach during the normal school day?”. Teachers are just asked “Do you do private tuition to
supplement your income?”. For students, the question is “Do you attend unofficial/non-compulsorily extra
classes, whether at school or not?” and “How many hours do you attend classes each week in each of these
subjects – Math, Vietnamese and Other subjects”.
7
the school year 2011-2012. Each test is a 45 minute test, which is characterized by 30
multiple-choice items. These tests were designed by the Vietnamese National Institute of
Educational Sciences to cover key subject areas of the Grade 5 curriculum. The first tests
cover the Grade 4 curriculum to measure learning level at the Grade 5 entry. The second
tests reflect G5 curriculum, more advanced G4 questions and common items replicated
from the first tests. There are 15 and 12 such common items in the tests and retests in
mathematics and Vietnamese respectively.12 This allows transforming the test and retest
scores to a common interval scale using IRT.13 The two rounds of tests enable tracking
students’ academic achievements during the year.
Instructional time
Time of instruction that a pupil has received from his/her attendance at formal
schooling and extra classes constitutes his/her instructional time. In this study the
instructional time is measured by the number of periods per week.
Social background
The pupils’ social background 14 is measured by a composite index to approximate
socioeconomic characteristics of the pupil. Social background level is constructed using
information collected in the survey that is related to student’s home and family, including
educational levels of parents, the number of meals per day, asset-based factors and factors
related to student’s learning environment at home (see detailed description in Table A1 in
appendix). Because these factors are categorial variables, the social background index is
computed applying the Multiple Correspondence Analysis (MCA).
MCA shows that 78 percent of the variation in social background factors is explained
by the first dimension (see Table A2 and A3 in Appendix), signaling the justification for
using only the first dimension in ranking the levels of student’s social background. Among
factors with highest contribution, having computer, internet, air conditioner at home, or
having mother or father with higher secondary level of education contribute positively to
the index while having no study desk, no study chair, no study lamp, or no study place
contribute negatively. In other words, the higher the index, the higher the socio-economic
status of the student. The average social background indexes of students having these
factors are highlighted in Table 1. Distribution of social background index is provided in
the Figure B1 in Appendix.
12
See (Rolleston et al., 2013) for detailed description of the tests
Item Response Theory is “the model-based measurement in which trait level estimates depends on
both person’s responses and the properties of the items that were administered” (Embretson & Reise, 2000)
13
14
Also referred to as family background or socio-economic status in relevant empirical literature
8
Table 1: Average social background indexes of students with selected factors
Characteristic
No
Yes
A computer at home
-0.3646
1.1969 ***
Internet at home
-0.3030
1.3198 ***
Air conditioner at home
-0.1887
1.4459 ***
Mother completed higher secondary education
-0.2718
0.8708 ***
Father completed higher secondary education
-0.2937
0.7821 ***
A study desk at home
-1.6068
0.2175 ***
A study chair at home
-1.5065
0.2105 ***
A study lamp at home
-0.9590
0.3594 ***
Own place to study at home
-1.0483
0.2637 ***
Source: Authors' calculations based on the YLs School Survey 2011
Note: difference significant at 1% (***), 5% (**), 10%(*)
4. Does FDS reduce the gap in instructional time between students with different
social background?
Through the lengthened school day, FDS may raise instructional time for children from
low social background while keeping the time constant for high social background children
who should substitute full-day schooling for extra-classes instructional time. Calculations
from the data show that instructional time is still low in Vietnam and that the objective of
the government to reach 35 periods per week is fulfilled only for the most advantaged
quintile of students (Figure 1). In addition, the instructional time increases with the level of
social background and the gap in instructional time across social background is still huge.
On average the students from the first quintile of social background have only 29 periods
per week, while the figure for the highest social background is 40.
9
Figure 1: Instructional time across quintile of social background
40
No of periods per week
38
36
34
32
30
28
26
Q1
Q2
Q3
Q4
Q5
Quintle of Social background
Source: Authors' calcula ons based on the YLs School Survey 2011
This is firstly because students of higher social background have better access to FDS
(Figure 2). Less than one quarter of students with the most disadvantaged social
background (quintile 1) benefit from the FDS. This proportion is twice as large for the 40
percent of students with the highest social background (quintiles 4 and 5). Thus, despite the
implementation of the School Education Quality Assurance Program, students with low
social background are still lagging behind in the access to the FDS. Poor targeting of
SEQAP may partially explain the remaining low access to FDS for disadvantaged children.
Indeed, our data shows that 86 percent of students from quintile 1 have not benefitted from
SEQAP, while more than half (57 percent) of students in SEQAP belongs to the three most
advantaged quintiles of social background (Table 2).15
Table 2: Access to SEQAP schools by social background
Quintile of
social background
Q1
Q2
Q3
Q4
Q5
Total
Obs
Non-SEQAP
school
19.92
19.64
19.22
19.78
21.44
100
2 841
SEQAP school
20.54
22.57
25.51
21.22
10.16
100
443
Source : Authors’ calculations based on the YLs School Survey 2011
15
The principals have been asked in the School Survey whether their school or their commune is, or has
ever been part of SEQAP.
10
Figure 2: Access to FDS across quintile of social background
Non FDS
FDS
80%
Percentage of students
70%
60%
50%
40%
30%
20%
10%
0%
Q1
Q2
Q3
Q4
Q5
Quintile of social background
Source: Authors' calculations based on the YLs School Survey 2011
The second reason explaining the gap in instructional time across social background
relies on the attendance of well-off students in extra classes. Children with positive social
background index, which correspond more or less to the two highest quintiles of social
background, are more likely to attend both extra classes and FDS than attending only FDS
(Figure 3). For these children, there is no substitution between FDS and extra-classes as
expected. At the bottom of the social ladder, the likelihood to attend half-day schooling
only is the highest, while the attendance in only FDS is observed at lower proportion than
the attendance in extra classes. For intermediary levels of social background (the third
quintile), FDS without extra classes is the most common scheme.
Thus, types of schooling differ widely across social background. As each type of
schooling is associated with different instructional time, the gap in instructional time
remains. Indeed, Figure 4 shows that FDS and extra classes together provide the highest
instructional time of nearly 45 periods per week, which is 79 percent higher than only halfday schooling (with neither FDS nor extra classes). It is evident that FDS helps to limit the
students’ lack of instructional time. By attending FDS, students have an increase of 35
percent in instructional periods in comparison with half-day schooling. However, extra
classes go with higher instructional time for students than FDS. Extra classes increase the
instructional time by 48 percent, which is 13 points of percentage stronger than for FDS. It
appears from Figure 4 that FDS only partially narrows down the gap in instructional time of
students from different level of social background.
11
Figure 3: Distribution of student social background by schooling type
Source: Authors’ calculations based on the YLs School Survey 2011
A third reason why instructional time still highly differs across social background could
be due to the large variations of the amount of additional time included in FDS. Figure 4
also examines the number of periods embedded in each type of schooling for each quintile
of social background index in order to assess whether time related to FDS is the same for
children with different social background. It shows that this third explanation does not hold
up under scrutiny. Indeed, the number of periods of those attending only FDS is almost the
same for each quintile of social background index. Only the amount of periods related to
extra-classes increases with the level of social background.
In sum, FDS does narrow down the differences in instructional time from attending
extra classes, but still does not nullify the gap. Since well-off students have higher access to
both FDS and extra classes, students with low social background lag behind in the number
of instructional hours.
12
Figure 4: Mean instructional time across schooling types, whole sample and by
student social background
Without FDS & EC
Without FDS & with EC
With FDS & without EC
With FDS and EC
50
Number of periods per week
45
40
35
30
25
20
15
10
5
0
Whole
Q1
Q2
Q3
Q4
Q5
Quin le of social background
Source: Authors' calcula ons based on the YLs School Survey 2011
5. Do students from different social background have the same access to resources
devoted to full-day schooling?
The implementation of FDS requires the enhancement of school resources to ensure
that FDS is functioning and can be translated into learning achievement. It is thus important
to know whether FDS goes hand in hand with more resources at the school level and
whether these resources are equally distributed according to the socio-economic
background of the children. Additional resources mostly include school infrastructures like
new or rehabilitated classrooms, class facilities and incremental teacher salaries to cover
the additional workload. SEQAP also provides teacher and school principal with trainings
to improve teaching and management practices. Therefore, the definition of school
resources is limited here to school infrastructure, class and school facilities, and teacher and
principal qualification and experience. Additional teacher salaries are already reflected in
the instructional time.
Table 3 shows the higher level in almost all aspects of school resources when FDS is
available. Students with FDS have better access to facilities both at school and class levels
compared to students without FDS. This is especially true for resources related to
infrastructures: the needs for major repairs of the school halve for students with FDS; the
proportion of students with access to clean drinking water is nearly five times higher with
13
FDS but still remain low, showing that it is not a common standard of schools in Vietnam;
most of the FDS students have latrines. School and class facilities are also higher. The
likelihood to have a library at school is 1.4 times higher for students with FDS. The ones to
access to computer or to internet are almost two times higher with FDS. Classes with FDS
are more often furnished with modern facilities like TV or electric fan. However, these
facilities are not encompassed in the specifications of FDS. It thus may reflect the selfselection process in the implementation of FDS rather than an improvement of facilities due
to FDS: schools with higher resources and financial contribution of the families are more
likely to have implemented FDS.
Table 3: Frequency of students with school resources – FDS vs. Non-FDS (%)
School facilities
No
FDS
Class facilities
No FDS
FDS
Major repairs needed
29,9%
12,1% ***
Black board or white board
99,0%
100,0% ***
Separate room for G5
96,8%
97,0%
Wall map
70,7%
67,8% *
Have library
64,4%
89,3% ***
Teacher's cabinet
83,1%
92,4% ***
Computers for students
32,1%
55,9% ***
Teacher’s desks
98,4%
100,0% ***
Internet for students
28,3%
51,4% ***
Sufficient electric lights
96,0%
98,5% ***
Electricity
95,1% 100,0% ***
Electric fan
87,0%
97,0% ***
Working electricity today
91,9%
95,5% ***
TV
4,1%
15,1% ***
Latrines for students
87,1%
98,5% ***
Video player or DVD
1,0%
0,0% ***
Separate latrines for boys/girls
83,6%
87,9% ***
Radio
3,0%
0,0% ***
5,8%
32,0% ***
Overhead projector
7,0%
Computer
6,1%
16,6% ***
69,3%
84,2% ***
Clean drinking water for students
FDS
Books other than text books
Principal
No
FDS
Teacher
FDS
No FDS
6,0%
FDS
Years of education
19,8
20,5 ***
Years of education
18,9
19,1 ***
Years of working in school
25,9
25,5
Years of teaching
18,5
16,1 ***
31,2%
55,9%
In-service training for more than 15 days
***
In-service training for more than 5 days
49,0%
55,1%
Source: Authors' calculations based on the YLs School Survey 2011
Note: difference significant at 1% (***), 5% (**), 10%(*)
The training qualification measured by the number of years of education16 or by the inservice training of both principals and head teachers are higher for students with FDS than
for others, but the difference is very slight. In addition, teachers or principals are not more
experienced when FDS is implemented.
Thus, FDS students have access to better school resources than the non-FDS group. But are
these resources equally distributed across students benefiting with FDS but with different
social background? To answer this question, we created two indexes to synthetize the
16
The number of years of education refers to the sum of the number of years for obtaining the highest
level of general education completed and the number of years for obtaining the highest level teacher training
qualification received.
14
***
information on school facilities and class facilities by using MCA analysis.17 The higher
values of these indexes, the more facilities the student has access to in his/her school and
class.
Figure 5: School resources across social background level for students attending FDS
A. Facilities
B. Principals and Teachers
28.00
0.70
School facilities
0.60
26.00
Class facliities
Principal's years of working
24.00
0.40
Year
Index
0.50
0.30
22.00
Principal's years of education
20.00
Teacher's years of education
0.20
18.00
0.10
Years being teacher
16.00
0.00
Q1
Q2
Q3
Q4
Q5
14.00
-0.10
Q1
Q2
Q3
Q4
Q5
Level of social background
Level of social background
Source: Authors' calculations based on the YLs School Survey 2011
Figure 5 shows the upward trend of school and class facilities across social background
– the higher the level of social background of the students, the higher the level of school
and class facilities they have access to. In addition, more advantaged social background
also allows students to access teachers with higher education level: on average, students in
the lowest quintile of social background are faced with teacher with one year of education
less than students in the highest quintile of social background. On the other hand, no
correlation is found between experience of teacher or principal and social background, nor
between social background and education level of the principal.
These results show that students are sorted into classes or school with different level of
resources according to their social background, and more importantly, that FDS does not
counteract this selection process. Thus, FDS does not overcome inequality of opportunity
in accessing resources that has been already observed by VASS (2012) and Hoang et al.
(2012).
6. Does the effect of full-day schooling on learning achievement vary across social
background?
This section tests whether the students of disadvantaged social background have higher
learning progress from attending FDS than the advantaged group. We estimate a Valueadded (VA) model where the test scores of students measured at the end of the academic
year are regressed by the test scores at the beginning of the academic year and by
interaction terms between FDS dummy and social background index. To reduce inequality
17
See detailed MCA results in Table A4 to Table A9 in Appendix.
15
of opportunity in learning achievement, the positive effect of FDS on learning progress
should be higher for children from disadvantaged social background.
More precisely, we estimate Value-added model using Ordinary Least Square (OLS).
Value-added models were initially developed to identify the contribution of school and
teacher characteristics to student achievement. These models are derived from a structural
“education production function” in which learning achievement is modeled as a function of
child, family and schooling inputs, and where learning is viewed as both dynamic and
cumulative process. Empirical estimation of this function raises difficulties due to the lack
of information regarding past inputs and due to the non-random assignment of students to
schools and teachers. Value-added models circumvent these difficulties by assessing the
difference in the learning achievement of student at the beginning and the end of an
academic year. Given that achievement is produced by a combination of the student’s
innate ability and all inputs accumulated until the time of reference, the model links current
student learning outcome to current educational inputs and lagged outcome, which is
assumed to capture all educational inputs in the past as innate ability of the students. As a
result, VA models have been widely used to deal with bias due to omitted past educational
inputs and time-invariant unobserved characteristics of the children as innate ability
(Mizala & Romaguera, 2002; Todd & Wolpin, 2007; Guarino et al., 2015).
Applying the review of Mizala & Romaguera (2002) and Todd & Wolpin (2007) on
VA model, a retest score in Math or Vietnamese of student i is a function of a vector of
current educational inputs , a dummy on full-day schooling
, the lagged test score
and a residual
that sums up unobservable factors as follows:
=
+
+
+
,
(1)
Cunha and Heckman (2009) additionally provide empirical evidence on crossproductivity effects of skills, in other words different skills substitute for each other. This
feature implies ability in Math might contribute to ability in Vietnamese and vice versa, so
(1) can be specifically expressed as:
=
=
+
+
+
+
+
,
,
+
+
,
,
+
(2)
(3)
in which the subscripts M and V stand for Mathematics and Vietnamese respectively, and
other symbols hold the same meaning as in (1).
Because the implementation of FDS is not random and may depend on unobserved
characteristics of schools and students, the coefficients
and
of FDS might be biased.
The problem of self-selection comes from the unavailability of school or student-related
factors that are correlated with both FDS and learning achievement. For example, students
with high motivation or ability are more likely to choose FDS schools and in the meantime
to have better achievement. The same holds true at the school level if schools which choose
to implement FDS are for example the most dynamic schools or schools where parents are
highly involved. The introduction of the lagged test scores in the right-hand side of the
equation controls for grouping and assignment mechanisms as far as those grouping
mechanisms are correlated with prior achievement (Guarino et al. 2015). In other words,
and coefficients are unbiased if two students with the same test scores will obtain the
same retest scores, regardless of their unobserved characteristics, and school-related
16
unobserved characteristics. Although this is a strong assumption, we expect that potential
bias in case this assumption is not verified will be mitigated by the introduction of variables
on school resources (school facilities, class facilities, qualification of teacher and principal)
and of social background variables. These variables may partially capture the heterogeneity
of schools and students that could impact the dynamic of learning between the test and the
retest. However, self-selection bias due to time-varying unobserved characteristics might
remain in theory as in other studies that do not rely on randomized trial in the
implementation of instructional time lengthening policies. Therefore, the coefficients
and
of FDS should be interpreted as the association between learning progress and FDS
and not as causal inference.
We estimate first equations (2) and (3) by defining the vector of current educational
inputs
as the characteristics of student and his/her family including student’s gender,
ethnicity, level of social background, number of siblings, number of older siblings, location
(urban/rural area), and school and class inputs including student’s attendance in FDS,
school and class facilities, qualification and experience of teacher and principal. The
corresponding models are labeled as Model (1). Description of the variables is provided in
Appendices, Table A10.
The heterogeneity in the effect of FDS on student achievement for students from
different level of social background could be addressed by including interaction terms
between explanatory variables in the achievement production function. Therefore the
interaction term between student social background and FDS will be added as additional
variables of the VA model to enable examining how the effect of FDS on learning progress
differs across levels of social background. The corresponding models are labeled as Model
(2).
The OLS estimation results of VA models (1) and (2) are shown in Table 4. The first
two columns provide the estimation results for Mathematics, and the remainder for
Vietnamese.
The results of the test for multicollinearity reject the null hypotheses that there are such
problems in models (1) (see Appendices, Table A11). Robustness check is also conducted
to test whether the coefficients
and
of FDS are sensitive to the introduction of school
facilities, and class facilities. The model appears to be robust to these changes of
specification (Appendices, Table A12).
17
Table 4: Estimation results of VA model of learning achievement (OLS)
Retest scores in Math
(1)
Attend FDS
Social background
-22.3391***
-23.6983***
Minority
Number of siblings
(1)
(2)
-24.0159***
-25.7633***
(3.0526)
(3.0886)
(3.3296)
(3.4054)
1.9378**
-1.0720
10.3063***
6.4459***
(1.7651)
(2.2009)
(1.9301)
(2.3087)
FDS x Social background
Boy
(2)
Retest scores in Vietnamese
8.2084***
10.4311***
(3.1412)
(3.4457)
0.0605
-0.2938
-16.5946***
-17.0314***
(2.8028)
(2.8088)
(3.0015)
(3.0008)
15.9974***
13.7276***
18.9412***
16.0769***
(4.9736)
(5.0364)
(5.5608)
(5.5893)
1.5408**
1.1337
4.1751*
3.6666*
(1.9191)
(1.9255)
(2.0550)
(2.0643)
-5.8800***
-6.2354***
-2.0808**
-2.4922**
(1.7920)
(1.8039)
(1.9146)
(1.9137)
0.5542***
0.5553***
0.0464***
0.0479***
(0.0181)
(0.0181)
(0.0172)
(0.0171)
0.0481***
0.0458***
0.4100***
0.4071***
(0.0166)
(0.0166)
(0.0186)
(0.0187)
6.5002***
6.6914***
2.2217**
2.4601**
(1.3600)
(1.3619)
(1.5338)
(1.5391)
Class facilities
6.9425***
7.3009***
2.9253**
3.4010*
(1.6745)
(1.6828)
(1.8536)
(1.8554)
Teacher's years of education
0.7100**
0.6276**
2.9808***
2.8771***
(0.6566)
(0.6568)
(0.7803)
(0.7753)
Urban areas
-10.1042***
-10.7194***
-8.0587*
-8.8391*
(3.7740)
(3.7868)
(4.0084)
(4.0204)
Constant
250.2823***
254.0716***
253.2825***
257.9335***
(17.7477)
(17.7534)
(19.3574)
(19.3427)
Number of observations
3 080
3 080
3 081
3 081
Adjusted R2
0.392
0.393
0.277
0.279
Number of older siblings
Math test score
Vietnamese test score
School facilities
Note: standard error in parentheses; *** p<0.01, ** p<0.05, * p<0.1
Source: Authors' calculations based on the YLs School Survey 2011
18
Figure 6: Average marginal effects with 95% Confidence Intervals
For both Math and Vietnamese, attending FDS does not mean higher learning progress
ceteris paribus, and even it is negatively and significantly associated with learning progress.
Compared to students attending FDS, students not in FDS but with same observable
characteristics have 22 and 24 points more in their retest scores in Math and Vietnamese
respectively (Table 4, Models (1)).
This result goes against the evidence found in other contexts (Holland et al. 2015,
Orkin 2013). Several reasons may explain the negative effect of FDS in this paper. First,
learning progress is measured through tests in Mathematics and Vietnamese. Yet, the
increase of instructional time due to FDS is very weak in these subjects: it includes only 1.5
and 1.7 additional periods per week in Vietnamese and Mathematics respectively, and these
two subjects account for around one third of the total increase of instructional time
(Erreur ! Source du renvoi introuvable. in Appendix). Thus, measuring learning progress
in Vietnamese and Mathematics only may largely underestimate the positive effect of FDS
on learning achievement. Second, the implementation of FDS may have side effect on the
learning in these subjects: children may lose concentration by spending all the day at school
if the curricula is not adapted to this change; additional instruction time may crowd out the
time spent by the teacher with/for the students out of the teaching time (interactions with
parents, revision of homework, etc.). While these hypotheses are difficult to test, Erreur !
Source du renvoi introuvable. in Appendix shows some evidence supporting them:
teachers with FDS complain much more about major problem in their class than teachers
without FDS and they interact less with the parents.
In addition, Models (2) show that the coefficients of interaction between FDS and
social background are positive and significant: for each unit increase in the social
background index, FDS students with same observable characteristics have 8 and 10 points
more in their retest scores in Math and Vietnamese respectively, holding other things
constant. As illustrated in Figure 6, FDS is negatively associated with learning progress for
students with low social background only. For student with an index of social background
higher than 1, which more or less corresponds to the fifth quintile of students, there is no
significant difference between FDS and non FDS students in terms of learning progress.
19
This result suggests that FDS contributes to inequality of opportunity in education, because
its conditions of implementation may be more detrimental to students with low social
background than for the well-off students. As an illustration, the lower the social
background is, the lesser the increase of instructional time in Vietnamese and Math
(Erreur ! Source du renvoi introuvable. in Appendix). However, as the selection process
in FDS implementation is not perfectly controlled for in our estimates, it may also be
interpreted as unobserved heterogeneity of students that is not compensated by the increase
of instructional time. It could be so if students with low social background and who benefit
from FDS are more likely to face shocks that affect their learning progress but that are not
captured by their initial score than students without FDS or students with higher social
background. Whatever the interpretations of these results, they show that FDS did not
provide students with low social background with enough additional instructional time or
with enough resources to compensate their disadvantage in learning progress compared to
better-off students.
Estimation results also depict the significantly positive correlation between other
school resources and learning progress. The correlation is probably due to the selection into
schools with better resources of the high social background students. Students with access
to better school and class facilities have higher learning progress. Students taught by higher
qualified teachers often obtain higher learning progress.
With same observable characteristics as the majority students, the students of ethnic
minorities have higher learning progress. The ethnic gap in scores is averaged at 14 to 16
points for the retest scores in Math and 16 to 19 points for Vietnamese. This matches what
was observed in Rolleston et al., (2013) on more learning progress of the ethnic minority
students. The reason might be attributed to the lower starting point of the minorities, which
leaves them more room for making progress than that of the majority.
Holding other things equal, boys seem to have lower progress in Vietnamese than girls.
For each additional older sibling of a student, he/she has six and two points less in their
retest scores in Math and Vietnamese. This might be a consequence of the lower level of
parental investment or care devoted to him/her. Moreover, students living in rural areas
obtain more learning progress than urban group. The situation is probably due to the same
reason for ethnic gap in retest scores: the rural students have more room for progress then
the urban ones.
In addition, both the self-productivity and cross-productivity effects are observed from
the estimation results as the retest score of one subject is positively correlated with the test
score of both the subject and the other subject.
7. Conclusions and discussion
FDS has been applied initially to deal with the current lack of instructional time. By
increasing instructional time, FDS should help to move forwards higher quality of
education. FDS may also help to restrict the negative effects of widespread extra classes in
Vietnam on inequality of opportunity in education. However, these two objectives have not
been reached yet: learning progresses are not better with FDS and the gap between children
with different social background remains high although they benefit from FDS.
20
As the nature of a transition process, the implementation of FDS has been varyingly
conducted to adapt to the physical and human resources of the schools. These resources
mainly rely on the affordability of communities and families. Consequently, the
instructional time still varies a great deal among FDS students. FDS still supplies students
with less instructional time than the extra classes do. In other words, the number of periods
provided for FDS students does reduce, but not enough to cancel out, the gap in
instructional time by attending extra classes.
The inequality of opportunity is apparent among students for the two following results.
Firstly, students from advantaged social background seem to attend both FDS and extra
classes. In addition, the increase of instructional time in Mathematics and Vietnamese due
to FDS is not equally distributed across students, and students with low social background
are disadvantaged in this respect. Both imply disparities in instructional time between
students from different levels of social background. Secondly, the extent to which students
access to more well-equipped school and class facilities, more qualified and more
experienced teachers rises with student social background level. Briefly speaking, FDS
students with low social background have lower access to school resources compared with
their counterparts from more advantaged social background.
Estimation results do not provide evidence on the positive relationship between FDS
and learning progress. At this stage of implementation, it appears that FDS has not reduced
the inequality of opportunity in learning achievement. On the contrary, it seems to magnify
the effect of social background on learning progress, thereby widening the gap in learning
achievement.
This above evidence gives hints of policy for the implementation of FDS towards the
target of educational equality in Vietnam. First of all, it is important to introduce a more
coherent regulation of FDS. The regulation should guarantee the same instructional time to
all the students. But additional school resources, training and adapted curricula should be
provided to limit reverse side effects of increasing the instructional time. Secondly, the
support on FDS should be extended to the students of social background disadvantage to
compensate for their gaps in instructional time from the students with higher social
background. Finally, since current FDS in essence is private tutoring at a lower cost, the
universal coverage of free FDS should be advocated in the long term to deal with the
inequality of opportunity in education.
21
Acknowledgement
This paper has received funding from the Privatisation in Education Research Initiative (PERI,
http://www.periglobal.org/) and from the NOPOOR project (www.nopoor.eu) under the FP7 of the European
Commission. The authors would like to thank anonymous reviewers for their comments on earlier drafts of
the paper.
References
Biernat, L. (2011). Reducing the Achievement Gap: There is Only One Real Solution. SSRN Electronic
Journal. http://doi.org/10.2139/ssrn.1936041
Bodewig, C., Badiani-Magnusson, R., Macdonald, K., Newhouse, D., Rutkowski J. (2014). Skilling up
Vietnam: preparing the workforce for a modern market economy. Washington, DC: World Bank.
Cabrera-Hernandez, F. (2015). Does lengthening the school day increase students’ academic achievement?
Evidence from a natural experiment. Working Paper Series No. 74-2015, Department of Economics,
University of Sussex. Retrieved from http://www.sussex.ac.uk/economics/documents/wps-74-2015.pdf
Cunha, F., & Heckman, J. (2009). The Economics and Psychology of Inequality and Human Development
(No. w14695). Cambridge, MA: National Bureau of Economic Research. Retrieved from
http://www.nber.org/papers/w14695.pdf
Dang, H.-A. (2007). The determinants and impact of private tutoring classes in Vietnam. Economics of
Education Review, 26(6), 683–698. http://doi.org/10.1016/j.econedurev.2007.10.003
Embretson, S. E., & Reise, S. P. (2000). Item response theory for psychologists. Mahwah, N.J: L. Erlbaum
Associates.
Guarino, C. M., Reckase, M. D., Stacy, B., Wooldridge, J. M. (2015). A comparison of Student Growth
Percentile and Value-Added Models of Teacher Performance, Statistics and Public Policy, 2:1, 1-11,
DOI: 10.1080/2330443X.2015.1034820..
Hincapie D. (2016). Do Longer School Days Improve Student Achievement? Evidence from Colombia. InterAmerican Development Bank Working Paper 679.
Hoang, X. T., Nguyen, T. P., Vu, V. N., Do, T. Q., Nguyen, T. H., Dang, T. H., & Nguyen, T. G. (2012).
Inequality Perception Study in Vietnam". Background paper for the 2012 Poverty Assessment.
Holland P., Alfaro P., Evans K. (2015) Extending the School Day in Latin America and the Caribbean. World
Bank Policy Research Working Paper 7309, http://dx.doi.org/10.1596/1813-9450-7309
Lavy, V. (2010). Do differences in school instruction time explain international achievement gaps in math,
science and reading? Evidence from developed and developing countries. National Bureau of Economic
Research Working Paper 16227. Retrieved from http://www.nber.org/papers/w16227.pdf
Lavy, V. (2012). Expanding school resources and increasing time on task: Effects of a policy experiment in
Israel on student academic achievement and behavior. National Bureau of Economic Research Working
Paper 18369. Retrieved from http://www.nber.org/papers/w18369
Mizala, A., & Romaguera, P. (2002). Equity and Educational Performance. Economía, 2(2), 219–273.
http://doi.org/10.1353/eco.2002.0009
Nguyen P. Ngoc (2008) An assessment of the Young Lives Sampling Approach in Vietnam, Young Lives
Technical Note n°4.
Orkin, K. (2013). The Effect of Lengthening the School Day on Children’s Achievement in Ethiopia. Young
Lives Working Paper 119. Retrieved from http://www.younglives-ethiopia.org/files/workingpapers/wp119_orkin_ethiopia
Patall, E. A., Cooper, H., & Allen, A. B. (2010). Extending the School Day or School Year: A Systematic
Review of Research (1985-2009). Review of Educational Research, 80(3), 401–436.
http://doi.org/10.3102/0034654310377086
Schutz, G.,Ursprung H., & Wossman L. (2008). Education Policy and Equality of Opportunity. KYLOS,
61(8): 279-308.
Roemer, J. E. (1998). Equality of opportunity. Cambridge, Mass.: Harvard Univ. Press.
Rolleston, C., James, Z., Tran, N. T. M. T., & Pasquier-Doumer, L. (2013). Making Progress: Report of the
Young Lives School Survey in Vietnam. Young Lives Working Paper 100. Retrieved from
22
http://www.younglives.org.uk/publications/WP/making-progress-report-of-the-young-lives-schoolsurvey-in-vietnam/yl-wp100_rolleston
Todd, P. E., & Wolpin, K. I. (2007). The Production of Cognitive Achievement in Children: Home, School,
and Racial Test Score Gaps. Journal of Human Capital, 1(1), 91–136. http://doi.org/10.1086/526401
Tran, N. T. M. T., & Pasquier-Doumer, L. (2015). Does full-day schooling increase educational inequality in
Vietnam? PERI Young Lives ESP Working Paper Series 2015 No. 72.
Tran, T. B. L. (2014). Full Day Schooling Performance of Primary Schools in Disadvantaged Areas in
Vietnam: A Comparative Case Study. VNU Journal of Sciences: Education Research, 30(No. 4), 17–30.
VASS. (2012). Opportunities for Children in Vietnam. (Vietnam Academy of Social Sciences) Background
paper for the 2012 Poverty Assessment.
World Bank. (2011). Viet Nam: High Quality Education for All by 2020 (Vol. 2): Analytical report (Working
paper No. 68092).
23
Appendices
A. Tables
Table A1: Descriptions of variables used for the social background index
Variable name
Mother's educational level
Father's educational level
Number of meals per day
Home possession
A television at home
A radio at home
An electric fan at home
A mobile phone at home
A bicycle at home
A motorcycle at home
A car at home
Air conditioner at home
A study lamp at home
A study desk at home
A study chair at home
A pocket calculator at home
Own place to study at home
Number of books in the home
A computer at home
Internet at home
Category
Primary or less
Lower secondary
Higher secondary
Primary or less
Lower secondary
Higher secondary
1
2
3 or more
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
None
1 to 5
6 to 10
More than 10
No
Yes
No
Yes
Proportion (%)
53.12
23.15
23.73
52.02
20.73
27.25
0.31
16.57
83.12
5.39
94.61
73.45
26.55
10.93
89.07
8.65
91.35
14.40
85.60
10.99
89.01
91.93
8.07
88.92
11.08
27.83
72.17
12.33
87.67
12.70
87.30
68.33
31.67
20.20
79.80
20.68
20.28
13.57
45.47
77.13
22.87
81.79
18.21
Source: Authors' calculations based on the YLs School Survey 2011
24
Table A2: MCA for Social background index
Dimension
Principal inertia
Cummulation
percent
78.31
78.31
Percent
dim1
0.033287
dim2
0.003861
9.08
87.40
dim3
0.000396
0.93
88.33
dim4
0.000127
0.30
88.62
dim5
0.000058
0.14
88.76
dim6
0.000005
0.01
88.77
dim7
0.000001
0.00
88.78
dim8
0.000000
0.00
88.78
Total
100.00
0.042505
Source: Authors' calculations based on the YLs School Survey 2011
Table A3: MCA Statistics of columns – Social background index
mass
Mother's educational level
Primary or less
0.0279
Lower secondary
0.0122
Higher secondary
0.0125
Father's educational level
Primary or less
0.0274
Lower secondary
0.0109
Higher secondary
0.0144
Number of meals per day
1
0.0002
2
0.0087
3 or more
0.0438
A television at home
No
0.0027
Yes
0.0499
A radio at home
No
0.0387
Yes
0.0140
An electric fan at home
No
0.0057
Yes
0.0469
A telephone at home
No
0.0045
Yes
0.0481
A bicycle at home
No
0.0076
Yes
0.0451
A motorcycle at home
No
0.0057
Yes
0.0469
A car at home
No
0.0484
Yes
0.0042
Air conditioner at home
No
0.0467
Yes
0.0059
A study lamp at home
No
0.0146
Yes
0.0380
A study desk at home
No
0.0064
Yes
0.0462
A study chair at home
No
0.0066
Yes
0.0460
A pocket calculator at home
No
0.0358
Yes
0.0168
Own place to study at home
No
0.0107
Yes
0.0420
Number of books in the home
None
0.0109
1 to 5
0.0106
6 to 10
0.0072
More than 10
0.0240
A computer at home
No
0.0405
Yes
0.0121
Internet at home
No
0.0430
Yes
0.0096
Source: Authors' calculations based on the YLs School Survey 2011
Overall
quality
0.7307
0.4128
0.8274
0.7056
0.3878
0.8182
0.3561
0.9319
0.9314
0.9232
0.9232
0.9904
0.9904
0.9672
0.9672
0.9145
0.9145
0.9547
0.9547
0.9067
0.9067
0.9067
0.9067
0.8952
0.8952
0.9363
0.9363
0.8423
0.8423
0.8621
0.8621
0.9758
0.9758
0.9234
0.9234
0.9778
0.8872
0.6487
0.9779
0.8434
0.8434
0.8339
0.8339
%inert
0.0216
0.0093
0.0441
0.0203
0.0086
0.0404
0.0004
0.0059
0.0012
0.0338
0.0019
0.0061
0.0168
0.0294
0.0036
0.0214
0.0020
0.0195
0.0033
0.0238
0.0029
0.0021
0.0240
0.0073
0.0573
0.0505
0.0194
0.0759
0.0105
0.0670
0.0096
0.0142
0.0302
0.0477
0.0121
0.0305
0.0032
0.0014
0.0173
0.0241
0.0802
0.0182
0.0811
coord
0.8489
-0.0537
-1.8465
0.8179
0.1328
-1.6595
1.1020
0.8832
-0.1792
3.5435
-0.1951
0.4399
-1.2181
2.3915
-0.2899
2.2753
-0.2125
1.5931
-0.2676
2.1655
-0.2642
0.1983
-2.2658
0.3865
-3.0500
1.9943
-0.7668
3.3351
-0.4611
3.1183
-0.4475
0.6911
-1.4730
2.2064
-0.5615
1.8304
0.3556
-0.1655
-0.9381
0.7595
-2.5308
0.6273
-2.7951
Dimension 1
sqcorr
0.7302
0.0030
0.7553
0.7053
0.0176
0.7660
0.3559
0.8948
0.8961
0.7996
0.7996
0.9659
0.9659
0.8672
0.8672
0.8509
0.8509
0.7710
0.7710
0.8823
0.8823
0.7088
0.7088
0.7530
0.7530
0.9008
0.9008
0.7338
0.7338
0.7503
0.7503
0.9462
0.9462
0.8542
0.8542
0.9357
0.3238
0.1099
0.9563
0.7601
0.7601
0.7277
0.7277
contrib
0.0201
0.0000
0.0425
0.0183
0.0002
0.0395
0.0002
0.0068
0.0014
0.0345
0.0019
0.0075
0.0207
0.0325
0.0039
0.0233
0.0022
0.0192
0.0032
0.0268
0.0033
0.0019
0.0217
0.0070
0.0551
0.0581
0.0224
0.0711
0.0098
0.0642
0.0092
0.0171
0.0365
0.0520
0.0132
0.0364
0.0013
0.0002
0.0211
0.0233
0.0778
0.0169
0.0754
coord
0.0645
-1.8530
1.6746
0.0469
-1.7880
1.2709
0.0684
-0.5282
0.1045
4.0906
-0.2252
-0.2059
0.5702
2.3840
-0.2890
1.8269
-0.1707
2.2831
-0.3835
1.0576
-0.1290
-0.3076
3.5153
-0.4932
3.8922
1.1621
-0.4468
3.7660
-0.5207
3.5351
-0.5074
-0.3589
0.7649
1.8440
-0.4693
1.1403
-1.3771
-1.0757
0.4145
-0.7382
2.4600
-0.7034
3.1345
Dimension 2
sqcorr
0.0005
0.4098
0.0721
0.0003
0.3702
0.0521
0.0002
0.0371
0.0353
0.1236
0.1236
0.0246
0.0246
0.1000
0.1000
0.0636
0.0636
0.1837
0.1837
0.0244
0.0244
0.1979
0.1979
0.1422
0.1422
0.0355
0.0355
0.1085
0.1085
0.1118
0.1118
0.0296
0.0296
0.0692
0.0692
0.0421
0.5634
0.5388
0.0217
0.0833
0.0833
0.1062
0.1062
contrib
0.0001
0.0420
0.0350
0.0001
0.0349
0.0232
0.0000
0.0024
0.0005
0.0460
0.0025
0.0016
0.0045
0.0323
0.0039
0.0150
0.0014
0.0395
0.0066
0.0064
0.0008
0.0046
0.0523
0.0114
0.0897
0.0197
0.0076
0.0907
0.0125
0.0826
0.0118
0.0046
0.0098
0.0363
0.0092
0.0141
0.0201
0.0083
0.0041
0.0221
0.0735
0.0213
0.0948
25
Table A4: MCA for School facilities index
Principal
Cummulation
Percent
inertia
percent
dim1
0.043530
62.92
62.92
dim2
0.007214
10.43
73.35
dim3
0.000758
1.10
74.44
dim4
0.000030
0.04
74.49
Total
0.069183
100.00
Source: Authors' calculations based on the YLs School Survey 2011
Dimension
Table A5: MCA Statistics of columns - School facilities index
mass
Major repairs needed
No
0.0814
Yes
0.0186
Separate room for G5
No
0.0022
Yes
0.0978
Have library
No
0.0230
Yes
0.0770
Computers for students
No
0.0580
Yes
0.0420
Internet for students
No
0.0601
Yes
0.0399
Electricity
No
0.0008
Yes
0.0992
Working electricity today
No
0.0042
Yes
0.0958
Latrines for students
No
0.0028
Yes
0.0972
Separate latrines for boys/girls
No
0.0117
Yes
0.0883
Clean drinking water for students
No
0.0817
Yes
0.0183
Source: Authors' calculations based on School Survey 2011
Overall
quality
0.8740
0.8740
0.6273
0.6273
0.8092
0.8092
0.7155
0.7155
0.7197
0.7197
0.5742
0.5742
0.6756
0.6756
0.7740
0.7740
0.7851
0.7851
0.6327
0.6327
%inert
0.0036
0.0156
0.0207
0.0005
0.1206
0.0361
0.0955
0.1320
0.0878
0.1320
0.0434
0.0004
0.0535
0.0024
0.0611
0.0018
0.1270
0.0169
0.0091
0.0403
coord
0.1804
-0.7906
-2.8603
0.0656
-2.5556
0.7646
-1.3108
1.8113
-1.2377
1.8608
-5.6986
0.0468
-3.1637
0.1391
-4.7584
0.1380
-3.5429
0.4703
0.0430
-0.1915
Dimension 1
sqcorr
0.4688
0.4688
0.5569
0.5569
0.7846
0.7846
0.6567
0.6567
0.6592
0.6592
0.3842
0.3842
0.4958
0.4958
0.6570
0.6570
0.7290
0.7290
0.0105
0.0105
contrib
0.0026
0.0116
0.0183
0.0004
0.1504
0.0450
0.0997
0.1377
0.0920
0.1383
0.0265
0.0002
0.0422
0.0019
0.0638
0.0019
0.1471
0.0195
0.0002
0.0007
coord
0.4120
-1.8052
2.4969
-0.0573
1.1102
-0.3322
-0.9637
1.3318
-0.9215
1.3854
9.8453
-0.0809
4.6798
-0.2058
4.9329
-0.1431
2.4155
-0.3206
0.8132
-3.6204
Dimension 2
sqcorr
0.4052
0.4052
0.0703
0.0703
0.0245
0.0245
0.0588
0.0588
0.0606
0.0606
0.1900
0.1900
0.1798
0.1798
0.1170
0.1170
0.0562
0.0562
0.6222
0.6222
contrib
0.0138
0.0605
0.0140
0.0003
0.0284
0.0085
0.0539
0.0745
0.0510
0.0767
0.0790
0.0006
0.0922
0.0041
0.0686
0.0020
0.0684
0.0091
0.0540
0.2404
Table A6: School facilities index by facility
Facility
No
Yes
Major repairs needed
-0.0087
-0.6770
Separate room for G5
-2.3537
-0.0924
Have library
-1.6952
0.3552
Computers for students
-0.9065
0.8829
Internet for students
-0.8651
0.9841
Electricity
-3.8004
-0.0540
Working electricity today
-2.5800
0.0094
Latrines for students
-2.7507
0.0675
Separate latrines for boys/girls
-1.5159
0.2523
Clean drinking water for students
-0.1639
-0.1592
Source: Authors' calculations based on the YLs School Survey 2011
26
Table A7: MCA for Class facilities index
Dimension
Principal inertia
Percent
Cummulation
percent
50.60
62.63
69.01
69.06
69.06
dim1
0.013994
50.60
dim2
0.003327
12.03
dim3
0.001765
6.38
dim4
0.000013
0.05
dim5
0.000000
0.00
Total
0.040056
100.00
Source: Authors' calculations based on the YLs School Survey 2011
Table A8: MCA Statistics of columns - Class facilities index
Overall
mass
quality
Black board or white board
No
0.0005
0.1064
Yes
0.0828
0.1064
Wall map
No
0.0252
0.8983
Yes
0.0581
0.8983
Teacher's cabinet
No
0.0109
0.8406
Yes
0.0725
0.8406
Teacher’s desks
No
0.0008
0.7895
Yes
0.0825
0.7895
Sufficient electric lights
No
0.0025
0.7803
Yes
0.0809
0.7803
Electric fan
No
0.0074
0.8269
Yes
0.0759
0.8269
TV
No
0.0762
0.4830
Yes
0.0071
0.4830
Video player or DVD
No
0.0828
0.1064
Yes
0.0005
0.1064
Radio
No
0.0819
0.5150
Yes
0.0015
0.5150
Overhead projector
No
0.0779
0.3372
Yes
0.0055
0.3372
Computer
No
0.0748
0.5498
Yes
0.0086
0.5498
Books other than text books
No
0.0209
0.7757
Yes
0.0624
0.7757
Source: Authors' calculations based on School Survey 2011
%inert
0.0304
0.0002
0.0345
0.0150
0.0976
0.0146
0.0669
0.0007
0.1174
0.0036
0.0696
0.0068
0.0121
0.1293
0.0002
0.0304
0.0001
0.0051
0.0064
0.0907
0.0191
0.1665
0.0621
0.0208
coord
1.1121
-0.0068
-1.5174
0.6596
-3.8067
0.5705
-10.5465
0.1038
-8.1539
0.2482
-3.8683
0.3789
-0.3096
3.3218
-0.0068
1.1121
-0.0113
0.6281
-0.1575
2.2489
-0.4096
3.5701
-2.0254
0.6795
Dimension 1
sqcorr
0.0105
0.0105
0.8529
0.8529
0.8157
0.8157
0.6831
0.6831
0.7057
0.7057
0.8086
0.8086
0.3067
0.3067
0.0105
0.0105
0.0573
0.0573
0.1540
0.1540
0.3323
0.3323
0.6995
0.6995
contrib
0.0006
0.0000
0.0581
0.0253
0.1574
0.0236
0.0903
0.0009
0.1636
0.0050
0.1113
0.0109
0.0073
0.0784
0.0000
0.0006
0.0000
0.0006
0.0019
0.0276
0.0125
0.1093
0.0859
0.0288
coord
6.9081
-0.0423
0.7181
-0.3121
1.3636
-0.2043
8.5399
-0.0840
5.4397
-0.1656
1.1921
-0.1168
-0.4814
5.1643
-0.0423
6.9081
0.0655
-3.6405
-0.3525
5.0323
-0.6796
5.9237
1.3712
-0.4600
Dimension 2
sqcorr
0.0959
0.0959
0.0454
0.0454
0.0249
0.0249
0.1065
0.1065
0.0747
0.0747
0.0183
0.0183
0.1762
0.1762
0.0959
0.0959
0.4577
0.4577
0.1833
0.1833
0.2175
0.2175
0.0762
0.0762
contrib
0.0242
0.0001
0.0130
0.0057
0.0202
0.0030
0.0592
0.0006
0.0728
0.0022
0.0106
0.0010
0.0177
0.1895
0.0001
0.0242
0.0004
0.0195
0.0097
0.1382
0.0345
0.3010
0.0394
0.0132
27
Table A9: Class facilities index by facility
Class facilities
No
Yes
Black board or white board
0.4870
-0.0030
Wall map
-0.6645
0.2888
Teacher's cabinet
-1.6670
0.2498
Teacher’s desks
-4.6185
0.0454
Sufficient electric lights
-3.5707
0.1087
Electric fan
-1.6940
0.1659
TV
-0.1356
1.4547
Video player or DVD
-0.0030
0.4870
Radio
-0.0049
0.2751
Overhead projector
-0.0690
0.9848
Computer
-0.1794
1.5634
Books other than text books
-0.8869
0.2976
Source: Authors' calculations based on the YLs School Survey 2011
Table A10: Description of variables
Description of variables
Obs.
Mean
SD
Min
Max
Attend FDS
3 264
0.41
0.49
0
1
FDS x Social background
3 264
0.07
0.56
-3.54
2.37
Social background
3 264
-0.01
1.01
-3.54
2.37
Boy
3 264
0.52
0.50
0
1
Minority
3 264
0.13
0.33
0
1
Number of siblings
3 233
0.89
0.74
0
2
Number of older siblings
3 235
0.79
0.79
0
2
Math retest score
3 185
545
99
258
814
Math test score
3 244
500
100
185
750
Vietnamese retest score
3 186
517
96
214
759
Vietnamese test score
3 254
500
100
192
727
School facilities
3 244
-0.16
1.19
-4.63
1.03
Class facilitites
3 264
0.00
1.00
-5.86
1.78
Teacher's years of education
3 244
18.96
1.96
11
21
Urban areas
3 264
0.83
0.38
0
1
28
Table A11: Multicollinearity tests
MATH
VIETNAMESE
No interaction
Variables
Social background
Class facilities
School facilities
Minority
Vietnamese test score
Math test score
FDS
Urban areas
Teacher's years of education
Number of siblings
Number of older siblings
Boy
Mean VIF
VIF
1.60
1.47
1.43
1.42
1.36
1.23
1.19
1.12
1.07
1.04
1.03
1.02
1.25
1/VIF
0.6252
0.6817
0.6983
0.7055
0.7371
0.8106
0.8401
0.8904
0.9384
0.9620
0.9668
0.9758
MATH
No interaction
Variables
Social background
Class facilities
School facilities
Minority
Vietnamese test score
Math test score
FDS
Urban areas
Teacher's years of education
Number of siblings
Number of older siblings
Boy
Mean VIF
VIF
1.59
1.46
1.43
1.41
1.36
1.23
1.19
1.12
1.07
1.04
1.03
1.02
1.25
1/VIF
0.6280
0.6828
0.6983
0.7090
0.7327
0.8133
0.8397
0.8913
0.9383
0.9636
0.9674
0.9767
VIF
2.31
1.61
1.48
1.45
1.44
1.37
1.23
1.23
1.13
1.07
1.04
1.04
1.03
1.34
1/VIF
0.4333
0.6221
0.6777
0.6882
0.6964
0.7306
0.8126
0.8154
0.8880
0.9366
0.9572
0.9627
0.9745
VIETNAMESE
With interaction
Variables
Social background
FDS x Social background
Class facilities
Minority
School facilities
Vietnamese test score
Math test score
FDS
Urban areas
Teacher's years of education
Number of siblings
Number of older siblings
Boy
Mean VIF
VIF
2.31
1.60
1.48
1.46
1.44
1.36
1.23
1.22
1.13
1.07
1.05
1.04
1.03
1.34
1/VIF
0.4333
0.6247
0.6771
0.6843
0.6964
0.7347
0.8101
0.8163
0.8871
0.9366
0.9555
0.9612
0.9735
With interaction
Variables
Social background
FDS x Social background
Class facilities
Minority
School facilities
Vietnamese test score
Math test score
FDS
Urban areas
Teacher's years of education
Number of siblings
Number of older siblings
Boy
Mean VIF
29
Table A12: VA models of learning achievement (robustness check)
Attend FDS
Social background
FDS x Social background
Model 1 (FDS= attending at least 30 periods/week)
Math
Vietnamese
No
With
No
With
interaction
Interaction
interaction
Interaction
-22.3391*** -23.6983*** -24.0159*** -25.7633***
(3.0526)
(3.0886)
(3.3296)
(3.4054)
1.9378**
-1.0720
10.3063***
6.4459***
(1.7651)
(2.2009)
(1.9301)
(2.3087)
8.2084***
10.4311***
(3.1412)
(3.4457)
FDS - attending at least 35 periods/week
Model 2 (FDS= attending at least 35 periods/week)
Math
Vietnamese
No
With
No
With
interaction
Interaction
interaction
Interaction
2.0062**
(1.7704)
1.4111**
(2.0082)
10.4090***
(1.9493)
9.1247***
(2.1972)
-14.2838***
(3.3124)
-14.9302***
(3.4419)
2.5074**
(3.5879)
-15.1998***
(3.5840)
-16.6286***
(3.7562)
5.4316**
(3.8891)
FDS x Social background
FDS - declared by teacher
Model 3 (
Math
No
interaction
In
1.7684**
(1.7743)
1
-8.0333***
(2.8652)
-8
FDS x Social background
Boy
0.0605
-0.2938
(2.8028)
(2.8088)
Minority
15.9974***
13.7276***
(4.9736)
(5.0364)
Number of siblings
1.5408**
1.1337
(1.9191)
(1.9255)
Number of older sisters/brothers
-5.8800***
-6.2354***
(1.7920)
(1.8039)
Math test score
0.5542***
0.5553***
(0.0181)
(0.0181)
Vietnamese test score
0.0481***
0.0458***
(0.0166)
(0.0166)
School facilities
6.5002***
6.6914***
(1.3600)
(1.3619)
Class facilities
6.9425***
7.3009***
(1.6745)
(1.6828)
Teacher's years of education
0.7100**
0.6276**
(0.6566)
(0.6568)
Urban areas
-10.1042*** -10.7194***
(3.7740)
(3.7868)
Constant
250.2823*** 254.0716***
(17.7477)
(17.7534)
Number of observations
3,080
3,080
Adjusted R2
0.392
0.393
Note: standard error in parentheses; *** p<0.01, ** p<0.05, * p<0.1
-16.5946***
(3.0015)
18.9412***
(5.5608)
4.1751*
(2.0550)
-2.0808**
(1.9146)
0.0464***
(0.0172)
0.4100***
(0.0186)
2.2217**
(1.5338)
2.9253**
(1.8536)
2.9808***
(0.7803)
-8.0587*
(4.0084)
253.2825***
(19.3574)
3,081
0.277
-17.0314***
(3.0008)
16.0769***
(5.5893)
3.6666*
(2.0643)
-2.4922**
(1.9137)
0.0479***
(0.0171)
0.4071***
(0.0187)
2.4601**
(1.5391)
3.4010*
(1.8554)
2.8771***
(0.7753)
-8.8391*
(4.0204)
257.9335***
(19.3427)
3,081
0.279
0.4484
(2.8232)
18.7079***
(4.9537)
1.8482**
(1.9303)
-5.8235***
(1.8052)
0.5597***
(0.0181)
0.0600***
(0.0165)
5.3791***
(1.3617)
6.4779***
(1.6709)
0.9577**
(0.6735)
-2.4269
(3.7678)
224.1710***
(17.5410)
3,080
0.385
0.3725
(2.8321)
18.0866***
(5.0212)
1.7698**
(1.9358)
-5.8951***
(1.8119)
0.5596***
(0.0181)
0.0597***
(0.0165)
5.3783***
(1.3625)
6.5081***
(1.6727)
0.9579**
(0.6736)
-2.5038
(3.7707)
224.5809***
(17.5306)
3,080
0.385
-16.1472***
(3.0193)
21.8509***
(5.5969)
4.4648*
(2.0599)
-2.0020**
(1.9288)
0.0524***
(0.0171)
0.4227***
(0.0184)
0.9854
(1.5226)
2.4116**
(1.8674)
3.2381***
(0.8049)
0.1947
(3.9176)
225.3452***
(19.4778)
3,081
0.269
-16.2860***
(3.0210)
20.5205***
(5.6767)
4.3041*
(2.0674)
-2.1374**
(1.9304)
0.0524***
(0.0170)
0.4222***
(0.0184)
0.9824
(1.5236)
2.4747**
(1.8684)
3.2390***
(0.8053)
0.0435
(3.9254)
226.1132***
(19.4918)
3,081
0.269
Source: Authors' calculations based on the YLs School Survey 2011
30
0.2209
(2.8258)
21.6041***
(4.9665)
1.5997**
(1.9315)
-6.1027***
(1.8207)
0.5585***
(0.0182)
0.0582***
(0.0165)
4.2199***
(1.3153)
6.6870***
(1.6673)
0.5360**
(0.6659)
-5.9028**
(3.7787)
236.7292***
(17.9265)
3,080
0.383
21
1
-6
0
0
4
6
0
-
23
(
Table A13: Number of periods by subject and teacher’s perception of class
management by FDS status and quintile of social background
Quintile of social background
Number of periods
in Vietnamese
Number of periods
in Mathematics
Number of periods
in all subjects
Regular visits to
pupil's home (%)
Major problem of
disciplines (%)
1
2
3
4
5
All
No FDS
7.9
7.8
7.7
7.8
7.5
7.8
FDS
8.6
9
9.1
9.4
9.9
9.3
Diff.
0.7
1.2
1.4
1.6
2.4
1.5
No FDS
5
5
5
4.9
4.7
4.9
6.6
FDS
6.3
6.3
6.4
6.6
7.2
Diff.
1.2
1.3
1.4
1.7
2.6
1.7
No FDS
25
25
25.3
25.5
26.2
25.4
FDS
33.8
34.3
34.5
34.5
34.7
34.4
Diff.
8.7
9.3
9.2
8.9
8.5
9
No FDS
54.3
50.6
48.9
41.9
22.6
44.8
FDS
25.9
17.2
21.1
20
10.3
18.1
Diff.
-28.4
-33.4
-27.8
-21.9
-12.3
-12.3
No FDS
5.8
4.5
7.3
11.1
0.9
5.9
FDS
19.6
28
35.6
34.6
28.9
30.5
Diff.
13.8
23.5
28.3
23.5
28.1
24.6
Major problem of poor
attendance among pupils
(%)
No FDS
9.1
23.5
28.3
23.5
28.1
7
FDS
17.1
23.6
27.7
25.7
11.6
21.4
Diff.
8
0.1
-0.6
2.2
-16.5
14.4
Major problem of
interruptions to teaching
(%)
No FDS
3.9
4.2
8.2
10.5
1.8
5.5
FDS
17.7
27.6
32.9
29.5
12.2
24.4
Diff.
13.8
23.4
24.7
19
10.5
18.9
Source : Authors’ calculations based on the YLs School Survey 2011
31
7.1. B. Figures
Figure B1: Distribution of students across social background index
Figure B2
Number of statutory teaching hours per year in public instutions by level
of education (2003)
Primary educa on
Lower secondary educa on
Upper secondary educa on
1,600
1,400
1,200
1,000
800
600
400
200
Vi
et
na
m
Ne
w
Ja
pa
n
a
Ze
al
an
d
Th
ai
la
nd
A
Re
us
tr a
pu
bl
lia
ic
of
Ko
re
a
M
al
ay
sia
La
o
PD
R
Ca
m
bo
di
a
an
c
ia
Sr
iL
In
d
on
es
ia
Ph
ilip
pi
ne
s
In
d
Ba
ng
la
de
sh
0
Source: Adapted from UNESCO 2006
32
Figure B3
Number of statutory teaching hours per year in public institutions, by educational level (2012)
Primary education
Lower secondary education
Upper secondary education, general programmes
Hours per year
1,400
1,300
1,200
1,100
1,000
900
800
700
600
500
400
300
200
Vietnam
Greece
Denmark
Japan
Russian Federation
Iceland
Norway
Finland
Israel
Korea
Turkey
Poland
Estonia
Austria
Slovenia
Hungary
Czech Republic
Slovak Republic
Italy
Belgium
France
Portugal
OECD average
Spain
England
Germany
Ireland
Indonesia
Luxembourg
Canada
Netherlands
New Zealand
Mexico
Australia
Chile
Scotland
0
Argentina
100
Source: Adapted from OECD (2014)
33