학업성취도에 있어서의 기회불평등 - Seoul National...
Transcript of 학업성취도에 있어서의 기회불평등 - Seoul National...
경제학석사학위논문
Inequality of Opportunity
in Educational Achievement:
A Cross-national Analysis using TIMSS data
학업성취도에 있어서의 기회불평등
-TIMSS 자료를 이용한 국제 비교-
2018 년 2월
서울대학교 대학원
경제학부 경제학 전공
옥 승 빈
Abstract
Inequality of Opportunity in Educational
Achievement:
A Cross-national Analysis using TIMSS data
Seungbin Ohk
Department of Economics
The Graduate School
Seoul National University
This study analyzes the trend in the degree of inequality of opportunity in educational
achievement during the past two decades, using Trends in International Mathematics and Science
Study (TIMSS) data set for Grade 8 students from 16 countries. Individual socioeconomic
background is measured using father’s level of education and the home learning environment
index, an aggregate index including factors relevant to parental support and resources. This study
first measures and compares the cumulative distributions, conditional on social background, using
stochastic dominance tools. Then, two inequality opportunity indices, the Gini Opportunity Index
(GOI) and Rags to Riches Index (RRI) are applied to track the extent of the inequality of
opportunity within countries and over time. The cross-national comparison is implemented by
ranking the countries according to the two inequality opportunity indices. The results show that
students from advantaged socioeconomic backgrounds are more likely to achieve academic
success in all countries, suggesting that no country obtains equality in terms of opportunity for
educational achievement. An example of a country with relatively low values of the opportunity
inequality indices is Hong Kong. England and New Zealand appear to be the most unequal
countries during the study period in terms of equality of opportunity. When the complete set of
countries is considered, the global trends in the two opportunity inequality indices show similar
patterns. While inequality in outcomes has increased, the inequality of opportunity has remained
relatively unchanged over the last two decades.
keywords: equality of opportunity, opportunity index, educational achievement, social mobility,
socioeconomic background
Student number: 2016-20156
Contents
1 Introduction ............................................................................................. 3
2 Equal Opportunity: Conception and Measurement ................................ 6
2.1 Definition of Equality of Opportunity ................................................................. 6
2.2 Opportunity Inequality Index ............................................................................... 7
2.2.1 The GOI...................................................................................................... 7
2.2.2 The RRI ...................................................................................................... 8
3 Data ......................................................................................................... 8
3.1 Data Set ................................................................................................................ 8
3.2 Social Background ............................................................................................... 9
3.2.1 Father’s Level of Education ..................................................................... 10
3.2.2 The HLEI.................................................................................................. 10
4 Empirical Results .................................................................................. 11
4.1 Inequality of Opportunity Related to Father’s Education .................................. 11
4.1.1 Conditional Distributions: Mathematics .................................................. 12
4.1.2 Conditional Distributions: Science........................................................... 13
4.1.3 The Stochastic Dominance Test ............................................................... 15
4.2 Inequality of Opportunity Related to the HLEI ................................................. 17
4.2.1 Conditional Distributions: Mathematics .................................................. 17
4.2.2 Conditional Distributions: Science .......................................................... 19
4.2.3 The Stochastic Dominance Test ............................................................... 20
4.3 Opportunity Inequality Index............................................................................. 23
4.3.1 Trend ........................................................................................................ 23
4.3.2 Rankings from 1995 to 2003 .................................................................... 25
4.3.3 Rankings from 2007 to 2015 .................................................................... 28
5 Conclusion ............................................................................................ 30
A Appendices
List of Tables
1 Stochastic Dominance Tests (Father's Education) ....................................................... 15
2 Stochastic Dominance Tests (HLEI) ............................................................................ 20
List of Figures
1. Math score Distributions Conditional on Father's Education ....................................... 13
2. Science score Distributions Conditional on Father's Education ................................... 15
3. Math score Distributions Conditional on HLEI ............................................................ 19
4. Science Score Distributions Conditional on the HLEI ................................................. 20
5. Inequalities in Outcomes and Opportunities conditional on father’s education ........... 24
6. Inequalities in Outcomes and Opportunities conditional on HLEI ............................... 25
7. Differences between Mathematics and Science ............................................................ 26
8. GOI in math 1995-2003 measured by father's education(left) and HLEI(right) ........... 27
9. GOI in Science 1995-2003 Measured by Father's Education (left) and HLEI (right) .. 27
10. RRI in math 1995-2003 measured by father's education(left) and HLEI(right) ........... 28
11. RRI in science 1995-2003 measured by father's education(left) and HLEI(right) ....... 28
12. GOI in math 2007-2015 measured by father's education(left) and HLEI(right) ........... 29
13. GOI in science 2007-2015 measured by father's education(left) and HLEI(right) ....... 29
14. RRI in math 2007-2015 measured by father's education(left) and HLEI(right) ........... 29
15. RRI in science 2007-2015 measured by father's education(left) and HLEI(right) ....... 30
1 Introduction
Education has been considered as a promising way of climbing up the social ladder, functioning
as a powerful weapon for encouraging social mobility. However, recently in Korea for over a
decade higher education has been regarded as an insurance against downward mobility rather than
a promise of upward social mobility. Parents with more economic and social resources support
their children for attaining higher academic achievement and to ensure that their child retain
superior position on the social ladder, leaving students from disadvantaged families with less
chance succeed in academic achievement. When the inequality of opportunity in educational
achievement prevails in a society, that is, the differences in the level of educational achievement
according to social background exist, an individual's socioeconomic background plays an
important role in determining her own social status. Inequality of opportunity results in lower
social mobility. Inequality of educational achievement across more or less advantageous socio-
economic family backgrounds are used to formulate opportunity inequality in educational
achievement. Using international test score data, we access opportunity inequality in Korea and
other countries. We also provide international comparison of opportunity inequalities.
Individual achievements result from circumstances and effort. A distinction can be made
(in terms of the sources of the inequality in achievements) between what individuals can and
cannot be held responsible for. Examples of the latter include inequalities resulting from race,
gender, and family background, which are beyond a person's control. The former are the
remaining factors that are relevant to individual’s choices, such as efforts, time, allocation,
strategies, habit formation, etc. Inequalities due to individual choices can be morally acceptable.
But inequalities due to reasons beyond one's control are not morally acceptable. Therefore,
identifying the sources of inequality is important when designing social policies that lesson such
inequality without reducing incentives for effort.
Roemer (1998) developed a model that determines the social policy that coordinates the
inequality of opportunities among various groups of individuals with different levels of
environmental factors that cannot be attributed to self-responsibility. The conceptual definition of
equality of opportunity in our study follows John Roemer's principles of leveling the playing field
and nondiscrimination principle. His leveling the playing field ensures children from
disadvantaged social backgrounds public supports for compensatory education so that they may
be able to compete for jobs with those from more advantaged backgrounds. Roemer views the
nondiscrimination principle, "as deriving from a particular interpretation of the leveling the
playing field principle". Morally irrelevant personal characteristics such as gender or race should
not affect one's eligibility for a position. According to Roemer, equality of opportunity should
focus on inequalities due to circumstances, compensating for possible disadvantages due to
factors related to social background.
While there is no agreement over how best to measure and compare the extent of equality
of opportunity, several concepts of equal opportunity have been developed. Based on the equal
opportunity policy proposed by Roemer (1998), Lefranc et al. (2008, 2009) use stochastic
dominance techniques to characterize and measure inequality of opportunity. They define equality
of opportunity as a situation in which income distributions that are conditional on social
backgrounds, the opportunities offered by different circumstances, “cannot be ranked according
to stochastic dominance criteria". Lefranc et al. (2008) compared conditional income distributions
across circumstances in OECD countries, defining the circumstances by the parental education
and occupation. They also introduced an index of inequality of opportunity, which is the Gini
index of the levels of well-being for all different social environments.
We investigate inequality of opportunity using data from the Trends in Mathematics and
Science Study (TIMSS), the large international comparative student achievement test. This data
set includes the assessment of student performance and family background questionnaires.
Unfortunately, family income is not included in the TIMSS data set. We indexify family
background by using three aspects of home environment: wealth, educational resources such as
the number of books and parents' education. The number of books in home can be treated as proxy
for family income and for the educational resources supported by parents. Then, students’ social
backgrounds are measured by students' home learning environment index (HLEI), which simply
aggregate the level of parents' education, the number of books, and the existence of home
possessions.
Because parents’ level of education affects the academic performance of their children,
both directly and indirectly, through family income, many previous studies including Lefranc et
al. (2008) also used parental education as a measure of social background. However, the HLEI,
is preferable, for two reasons, when using TIMSS data. First, the HLEI can make the composition
of environments as equal as possible across countries. One shortcoming of using parental
education as a measure of circumstances is that it leads to notable differences in the relative sizes
of groups across countries, which could limit the cross-country comparability of the findings.
When using the HLEI, the relative sizes of the groups become more balanced across countries.
For instance, in Korea, one group with a certain level of father's education represents less than
two percent of the overall population. Because non-parametric stochastic dominance tests require
a sufficiently large number of observations in each group, such measures might provide
unreasonable estimations.
Second, the home learning environment may have a distinct influence on educational
outcomes. For example, parents may have a low level of education, but may nevertheless have
high educational aspirations for their children and, thus, invest heavily in their children's
education. The HLEI ensures that high achievement in one factor compensates for low
achievement in another factor. The HLEI is designed to indicate parental support -the extent of
home resources and parents’ education which may have an important impact on academic
achievement.
Kim and Lee (2008) measured the extent of the equality of opportunities for income in
terms of parents’ level of education and father’s occupation in Korea. They compared the impact
of the fiscal policy in Korea- reducing inequality of opportunity for income - with those in 11
Western countries, which had been analyzed by Roemer et al. (2003). Ko and Lee (2011)
examined the degree of equality of opportunity for academic attainment and income in Korea.
They reported that the father's education accounts for 16 percent to 59 percent of the inequalities
for son's educational attainment and 2 percent to 12 percent of the inequality in the son's income.
Byun and Kim (2010) investigated the trends in educational inequality based on math
performance in South Korea using TIMSS data for three waves (1999, 2003, and 2007). They
compared the changes in the impact of the socioeconomic gap on student achievement in Korea
with those in the United States. Their results suggest that the effects of socioeconomic background
on student achievement have remained unchanged in the United States, but have increased over
time in Korea. Salehi-Isfahani et al. (2014) examined the level of inequality of opportunity in
educational achievement in math and science, focusing on the Middle East and North Africa,
using data from TIMSS (1999, 2003, and 2007). Their results showed that most countries in
Middle East and North Africa are less equal than European countries in terms of opportunity. Kim
et al. (2014) compared the influence of parents' education level on student's achievement in PISA
among OECD countries. They find that parental education is significantly related to students'
academic attainment in every OECD country. In addition, the level of inequality of opportunity
is relatively low in Korea.
Recent findings indicate that educational opportunity is a determinant of inequality and
intergenerational mobility. However, few studies have empirically examined and compared the
trends in the equality of opportunity for academic achievement across countries. In this context,
this study makes two contributions. First, it measures the extent of the inequality of opportunity
in terms of academic performance across time and countries using stochastic dominance tools.
Second, we compare the level of equality of opportunity for academic performances in 16
countries using two opportunity inequality indices: Gini Opportunity Index (GOI), developed by
Lefranc et al. (2008), and the Rags to Riches Index (RRI), introduced by Oh and Ju (2016).
Furthermore, the extent of intergeneration mobility is captured by comparing the values of the
opportunity indices, and particularly the RRI which provides clear information on the chances of
success of students in the lowest socioeconomic classes.
The remainder of this thesis is organized as follows. Section 2 discusses the model and
defines equality of opportunity for academic achievement and opportunity inequality indices.
Section 3 provides information on the datasets and describes the procedures used to estimate the
indices in practice. Section 4 presents the main results, and compares the levels of inequality of
opportunity in 16 countries. Section 5 concludes our investigation.
2 Equal Opportunity: Conception and Measurement
Although there is no single definition of fair equality of opportunity, there is widespread
agreement on its basic principles. While inequalities due to factors for which individuals are
accountable – a person’s level of effort- are considered morally acceptable, inequalities due to
circumstances that are beyond a person's control, but that affect the outcome of interest are unfair
and should be compensated for.
2.1 Definition of Equality of Opportunity
This study follows the same method proposed in Lefranc et al.(2008).1 Consider a society of
individuals, where each individual's academic achievement is determined by circumstances and
effort. Define a finite set T (i.e.,a type) as a set of individuals with similar circumstances.
Individuals within each type can differ only in their level of effort. The distribution of scores
within each type is regarded as the opportunity set open to the individuals of that type. Differences
in opportunity sets indicate the presence of inequalities in opportunities. All individuals with the
same level of effort should have the same chances of achieving the objective, regardless of their
type.
Let c denote a set of circumstances,e be an individual’s level of effort, and F(∙ |c, e)
be the conditional probability distribution of academic achievement, given c and e. Denote the
distribution of the test scores, conditional on circumstances c, by F(y|c). Considering the concepts
of opportunity equality, if the distribution of scores is independent of circumstances, then equality
of opportunities requires that F(∙ |𝑐, 𝑒) = F(∙ |𝑐′, 𝑒). However, this condition seems improbable
in practice. Lefranc et al. (2009) noted this is “a compelling case of equality of opportunity”. They
formalize equality of opportunity using stochastic dominance relations among the conditional
distributions (Lefranc et al.,2009):
Given 𝑐, 𝑐′ and 𝑒 , the first order stochastic dominance(FSD) holds between two
probability distributions F(∙ |𝑐, 𝑒) and F(∙ |𝑐′, 𝑒) if for all 𝑦,
F(𝑦|𝑐, 𝑒) ≥ F(𝑦|𝑐′, 𝑒)
with strict inequality for some y. This coincides with the property that for all 𝑢(𝑦) is
monotonically increasing,
∫ 𝑢(𝑦)𝑑𝐹(𝑦|𝑐, 𝑒) ≥ ∫ 𝑢(𝑦)𝑑𝐹(𝑦|𝑐′, 𝑒)
with strict inequality for at least one monotonically increasing u(y). Then circumstance 𝑐 is
always preferred to circumstance 𝑐′ by any student. When the FSD holds for all 𝑒, we say that
the first order opportunity inequality exists between the two circumstances, 𝑐 and 𝑐′ . The
1 A detailed explanation of the measurement and the GOI are given in Lefranc et al. (2008).
second order stochastic dominance (SSD) holds between the two probability distributions
F(∙ |𝑐, 𝑒) and F(∙ |𝑐′, 𝑒) if for all 𝑥,
∫ 𝐹(𝑦|𝑐, 𝑒)𝑑𝑦 ≥ ∫ 𝐹(𝑦|𝑐′, 𝑒)𝑑𝑦𝑥
0
𝑥
0
with strict inequality for some 𝑥 . This coincides with the property that all students with rick
averse and monotonically increasing utility functions prefer circumstance 𝑐 to circumstance 𝑐′.
When the SSD holds for all 𝑒, we say that the second order opportunity inequality exists between
the two circumstances, 𝑐 and 𝑐′.
The distribution of effort may be affected by circumstances. Under an extremely inferior
circumstance, it may be difficult to make enough effort for achieving high test-scores. Then non-
existence of stochastic dominances defined earlier does not seem to warrant opportunity equality.
To avoid this issue, we follow Roemer (1998) and Lefranc et al. (2008, 2009) and measure effort
in such a way that the distribution of effort does not depend on circumstances. When such a
measure of effort is used, the first and the second order opportunity inequalities have the following
necessary conditions, which are used in our investigation as conditions for opportunity
inequalities.
The First Order Opportunity Inequality Condition: For some circumstances c ,c ' , there exist first
order stochastic dominance between F (× |c ) and F (× |c ') .
The Second Order Opportunity Inequality Condition: For some circumstances c ,c ' , there exist
second order stochastic dominance between F (× |c ) and F (× |c ') .2
2.2 Opportunity Inequality Index
2.2.1 The GOI
Let the population be partitioned into K types by T = {𝑇1, 𝑇2, … , 𝑇𝐾}. Let 𝜇 denote the mean
score of the entire population, the mean of {𝜇1(1 − 𝐺1), 𝜇2(1 − 𝐺2), … , 𝜇𝐾(1 − 𝐺𝐾)},
G indicate the Gini coefficient, P be a population proportion, and the subscript t denote
circumstance t. Then 𝜇𝑡(1 − 𝐺𝑡) represents twice the area under the Generalized Lorentz curve
of circumstance t. Rank all environments in ascending order of 𝜇𝑡(1 − 𝐺𝑡). Then, the GOI is
expressed as follows:
GO =1
𝜇∑ ∑ 𝑃𝑖𝑃𝑗
𝑗>𝑖
𝑘
𝑖=1
(𝜇𝑗(1 − 𝐺𝑗) − 𝜇𝑖(1 − 𝐺𝑖))
The GOI is proportional to the sum of the differences between the areas of the opportunity sets.
2 The proof can be found in Lefranc et al. (2009).
Therefore, the closer the value is to zero, the more equal is the opportunity, and the closer the
value is to one, the more unequal is the opportunity.
2.2.2 The RRI
One of the most important issues related to opportunity equality is the way it increases the
problem of intergenerational mobility. If the probability of high performance is much lower
among students from a disadvantaged background, that is, if a less advantaged environment leads
to a lower possibility of higher educational attainment, then social mobility will decline as a result.
The RRI identifies how many students from the most disadvantaged background can ascend the
social ladder, and is measured as the number of students from the most disadvantaged
environment as a proportion of the highest-performing students.
Denote 𝑐 as the most disadvantaged circumstance, and 𝑞𝑐 as the population
proportion of the most disadvantaged group. Define p, the set of highest-performing students, as
those in top p percentile. Let n𝑝 denote the number of students in p , and n𝑝,𝑐, denote the
number of students from the most disadvantaged group in p.
𝑅𝑅𝑝 = 1 −n𝑝,𝑐, n𝑝⁄
𝑞𝑐
The value of the RRI ranges between zero and one3, with zero representing 0 percent inequality
of opportunity. This indicates that, the share of students from the lowest socioeconomic
background among the top p percent of the overall population by achievement is equal to the
share of students from the lowest socioeconomic background. Then, a value of one indicates the
highest level of opportunity inequality, where students from the lowest socioeconomic
background cannot achieve the top p percent of academic success. In this study, the highest-
performing students are defined as 20 percent of the population. A decrease in the RRI implies an
increase in the number of students from the least advantaged group who have achieved a first
quintile score on the test.
3 Data
3.1 Data Set
The Trends in International Mathematics and Science Study (TIMSS), an international
comparative study of student achievement in mathematics and science, represents the most
comprehensive international comparison. The TIMSS is designed to assess several criteria in
3 If the proportion of high-performing students from lowest social status to total high-performing students exceeds
the proportion of lowest social status population to total population, RRI might have negative value; the negative
values imply that students from lowest social status are more likely to achieve highest performance.
terms of students' characteristics, home environment, learning resources, and science and
mathematics achievements among Grades 4 and 8. The study has been administered every four
years by the International Association for the Evaluation of Educational Achievement (IEA) since
1995.
In this study, data from six waves of TIMSS (1995, 1999, 2003, 2007, 2011, 2015) are
employed to analyze the relationship between students’ academic performance and home
environment, which are drawn mainly from student background questionnaires. This study
focuses on 16 countries: Australia, Canada, Germany, England, Finland, France, Hong Kong, Italy,
Japan, Korea, the Netherlands, New Zealand, Norway, Singapore, Sweden, and the United States.
Consequently, the purpose of this study is to measure and compare the degree of the inequality in
opportunity (in science and mathematics achievement) among Grade 8 students in the 16 countries
who participated in the TIMSS assessment. All of these countries have participated in the TIMSS
assessment at least once, but not all countries participated in every round. For example, Hong
Kong, Korea, Singapore, and the Unites States joined TIMSS in 1995, and France and Germany
participated in the program in 1995 only. The inequality indices are computed using only those
countries that participated in that year. Therefore, when comparing these indices across countries,
this study will separate the data into two parts: from 1995 to 2003, and from 2007 to 2015.
3.2 Social Background
In this study, individual social background is measured by: (1) father’s level of education, as in
Roemer (1998) and Lefranc et al. (2008), and (2) the HLEI, which is derived from selected factors.
Since only restricted circumstances are available in TIMSS, our value of Inequality of opportunity
indices should be considered as lower bounds.
Most studies on the equality of opportunity that use a non-parametric approach, consider
a single variable (e.g., level of education, income, or parents’ occupation) as the social origin of
the inequality. This might be attributed to the requirements of a reasonable partition and large sub-
samples within each type. When partitioning the population into types (subsets of the population
that are homogeneous with respect to circumstances) to test for the existence of opportunity
inequality, the test suffers from a limited number of observations as the number of types increases.
This study develops a circumstance index, the HLEI, taking into account that several factors,
rather than parental income, occupation or education alone, influence the equality of opportunity
in education. The HLEI also solves the problem of a reasonable number of sub samples. The HLEI
is a composite measure that captures a positive home learning environment using three
dimensions: parents’ level of education, home educational materials (resources and facilities), and
the number of books in the home. Assuming that all factors included in HLEI may indicate parents’
aspirations and devotion to their child’s education, partitioning the population according to the
value of HLEI is meaningful, because it reflects the resources and environmental effects provided
by parents.
According to the mean scores conditional on the social background (i.e., father’s
educational attainment and the HLEI) students’ educational performance has been shown to be
related to their social background. In all countries, the mean scores of students from the more
privileged group are always higher than those of students from the less privileged group.
3.2.1 Father’s Level of Education
The data on students’ educational achievements are assigned to one of three categories based on
the first measurement: father’s level of education. The categories Low, Med, and High include
students whose fathers have less than or equal to lower secondary education, short-cycle tertiary
education or below, and a bachelor degree or above, respectively. Table 1 presents the descriptive
statistics (means and standard deviations) conditional on the father’s level of education.4
The proportion of students from disadvantaged socioeconomic backgrounds (Low) has
decreased dramatically in Korea. In 2015, the proportion of Korean students whose fathers have
a low level of education was 1.98 percent, and those whose fathers have a medium level of
education was 46.85 percent. The sum of these two groups is smaller than the percentage of
Korean students whose fathers have high level of education (51.17 percent). In fact, with the
exception of the United States, the proportion of students from low backgrounds has decreased
by more than half in all countries.
3.2.2 The HLEI
The HLEI reflects the level of the learning environment, and is based on three factors: number of
books at home, coded from 1 to 5, where 5 indicates having more books in the home; home
possessions,5 as dichotomous variables (the items asked were not the same in each year); and
parents’ education level. The responses to parents’ highest level of education are re-coded on a
seven-point scale so that high scores indicate higher levels of education. Then, each factor
receives a score from 0 to 1. The overall index value is the sum of the weighted scores for each
factor. In this study, each factor is simply weighted at 1: Factor1’s score × 1 + Factor2’s
score×1+Factor3's score×1. Thus, the HLEI value can range from 0 to 3. A higher score indicates
a more socioeconomically advantaged environment at home. Then, based on the scores derived
from the weighted factors, students are divided into three equally sized groups as much as that is
possible, where category 3(1) denotes the most advantaged (disadvantaged) social background.
When measuring circumstances using the HLEI, each group includes at least 25.92 percent and
4 Descriptive statistics for the remaining countries are presented in Appendix.
5 In 1995, 1997, and 2003: calculator, computer, study desk, and dictionary. In 2007: calculator, computer, study
desk, dictionary, and internet connection. In 2011: computer, study desk, own room, own books and internet
connection. In 2015: digital devices, own computer, computer, study desk, own room, internet connection and own
mobile phone.
at most 41.43percent of the total subpopulation.
4 Empirical Results
The degree of opportunity inequality can be analyzed using the curves of the distributions of the
test scores, conditional on the social backgrounds. When comparing these curves to check for
first-order inequality of opportunity, three situations can occur: the two curves are identical; the
two curves intersect; one curve lies above the other. Equality of opportunity is satisfied in the first
two cases, and is violated in the third case.
According to Lefranc et al.(2008), the distance between the distributions of the groups also
needs to be estimated and compared, because the gap reflects “the magnitude of the advantage
conferred by more privileged backgrounds over less privileged ones.” The gap between the
distributions of the different backgrounds varies between countries, suggesting that the extent of
the advantage given to the privileged group varies as well. In other words, if the three conditional
distributions for each category are close, differences in social background correspond to very
small differences in academic performance.
Differences in partitioning the individual social backgrounds into circumstances correspond
result in different findings. We have to keep in mind that a restrictive approach to partitioning
individual social backgrounds into circumstances is likely to lead to an underestimate of
inequality of opportunity: any inequality associated with unconsidered backgrounds (race, gender
or language) may remain within three types and would be attributed to effort.
Reporting visual inspections in terms of the FSD and second-order stochastic dominance
(SSD) for six rounds of TIMSS and as many as 16 countries for math and science takes a lot of
space. Therefore, this paper concentrates on comparing the dominance results using one or two
countries from each continent: Australia, Korea, Hong Kong, Sweden, and the United States are
presented and compared.6 Whether the gaps between the distributions of the different types have
narrowed can be seen clearly from a comparison of the curves for the first and the last test.
4.1 Inequality of Opportunity Related to Father’s Education
The cumulative distribution functions conditional on father’s education are presented in this
section. According to Figure 1 and 2, the CDF for individuals from the more privileged group
(students whose fathers have higher levels of education) is always below the CDF for individuals
from the less privileged group (students whose fathers have lower level of education). Note that
the distribution of the students in the latter category in 2015 includes samples that are relatively
6 Results for the remaining countries are included in the Appendix.
smaller than they were in 1995 for all countries except the United States. Students from lower
socioeconomic backgrounds are less likely to obtain higher levels of academic performance than
those from higher socioeconomic backgrounds.
4.1.1 Conditional Distributions: Mathematics
Figure 1 presents the conditional distributions for scores in mathematics in 1995 (left) and 2015
(right). The test scores are plotted on the horizontal axis and the percentiles on the vertical axis.
Except for the United States, the gaps between the three CDFs corresponding to the differences
in father’s levels of education appear to be widening further, indicating an increase in opportunity
inequality. In the case of Hong Kong, although it has not escaped the worldwide widening of the
gaps, the three conditional distributions for each category are closer than any of the other countries
in all rounds. Sweden seems to have a more equal distribution than those of other countries in
1995. However, the inequality of opportunity seems to increase to a point where it matches or
exceeds that of the United States. In Korea, the distance between Low and Med is closer than the
distance between Med and High, suggesting that there is more equality of opportunity at the
bottom of the social ladder.
4.1.2 Conditional Distributions: Science
Figure 2 shows the conditional distributions for scores in science in 1995 (left) and 2015 (right).
In nearly all cases, the cumulative distribution functions of science score conditional on father’s
education are similar to those of the mathematics scores. For Korea, the gaps in 1995 are relatively
smaller than those shown in Figure 1, but then become more similar to those in 2015.
Figure 1: Math score Distributions Conditional on Father's Education
Figure 2: Science score Distributions Conditional on Father's Education
4.1.3 The Stochastic Dominance Test
To complement the result offered by the cumulative distributions, the stochastic dominance test
results based on the SSD are presented in Table 3. Here, the top and bottom 2.5 percent of each
group are excluded. Thus, in 2015, the distribution for Korea cannot be ranked using the stochastic
dominance tool, because Korean students from a low social background form only about 1.97
percent of the total population.7 With the exception of this group, the equality of opportunity
measured using the SSD tests appear consistent with the comparisons obtained from the visual
inspection of CDFs. As shown in Table 3, the distributions of the less privileged groups are equal
in Hong Kong in 2003, but the degree of equality of opportunity seems to weaken.
Table 1: Stochastic Dominance Tests (Father's Education)
Mathematics Science
Country Category 1 2 3 1 2 3 Australia
1995
Low ≺1*** ≺1*** ≺1*** ≺1***
Med ≺1*** ≺1***
High
1999
Low ? ≺1***
≺2* ≺1***
Med ≺1*
≺1*
High
2003
Low ≺1* ≺1***
≺2*** ≺1***
Med ≺2***
≺1***
High
2007
Low ? ≺1***
≺2*** ≺1***
Med ≺1***
≺1***
High
7 The result of dominance tests conditional on HLEI - distributions of Korean students from more privileged group
dominates those from less privileged group, indicates that some results from Table3 have problems with small sample
size.
2011
Low ≺2*** ≺1***
≺1*** ≺1***
Med ≺1***
≺1***
High
2015
Low ≺1*** ≺1***
≺1*** ≺1***
Med ≺1***
≺1***
High
Hong
Kong
1995
Low ≺1*** ≺1*** ≺1** ≺2***
Med ≺2*** ?
High
1999
Low ≺1* ≺1**
= ≺1***
Med ?
≺1***
High
2003
Low = ?
= ≺1***
Med ?
≺1***
High
2007
Low ? ≺1**
≺2* ≺1**
Med ≺2**
≺2**
High
2011
Low ≺1** ≺1***
≺1** ≺1***
Med ≺1***
≺1***
High
2015
Low ? ≺1***
≺1** ≺1***
Med ≺1***
≺1***
High
Korea
1995
Low ≺1** ≺1*** ≺2* ≺1***
Med ≺1*** ≺1***
High
1999
Low ≺2*** ≺1***
≺1*** ≺1***
Med ≺1**
≺1***
High
2003
Low ≺1*** ≺1***
≺1** ≺1***
Med ≺1***
≺1***
High
2007
Low ≺2** ≺1***
≺2* ≺1***
Med ≺1***
≺1***
High
2011
Low ≺2** ≺1***
≺2** ≺1***
Med ≺1***
≺1***
High
2015
Low ? ≺1***
? ≺1***
Med ≺1***
≺1***
High
Sweden
1995 Low
≺2** ≺1** ≺2*** ≺1***
Med ≺2* ≺1∗∗
High
2003
Low ≺1* ≺1**
≺2* ≺1**
Med ≺2**
≺1*
High
2007
Low ≺2** ≺1**
≺1*** ≺1***
Med ?
?
High
2011
Low ≺1* ≺1***
≺1* ≺1***
Med ≺1***
≺1***
High
2015
Low ≺1* ≺1***
≺1*** ≺1***
Med ≺1**
≺2**
High
USA
1995
Low ≺1*** ≺1*** ≺1*** ≺1***
Med ≺1*** ≺1***
High
1999
Low ≺1*** ≺1***
≺1*** ≺1***
Med ≺1***
≺1***
High
2003
Low ≺1*** ≺1***
≺1*** ≺1***
Med ≺1***
≺1***
High
2007
Low ≺1*** ≺1***
≺1*** ≺1***
Med ≺1***
≺1***
High
2011
Low ≺1*** ≺1***
≺1*** ≺1***
Med ≺1***
≺1***
High
2015
Low ≺2*** ≺1***
≺1*** ≺1***
Med ≺1***
≺1***
High Notes ≺𝑖 : the column dominates the row for order i stochastic dominance. =: the distributions are equal. ?: the
distributions cannot be ranked using first- and second-order stochastic dominance. Father's education group
corresponds to social background. Low: lower secondary education or below, Med: short-cycle tertiary education or
below, High: bachelor degree or above.
4.2 Inequality of Opportunity Related to the HLEI
4.2.1 Conditional Distributions: Mathematics
Figure 3 shows the conditional distributions for scores in mathematics in 1995(left) and
2015(right). Category 1 is the bottom one-third of the population; Category 2 is the second one-
third of the population; and Category 3 is the upper one-third of the population.
Hong Kong exhibits the lowest degree of opportunity inequality. On the other hand, in
Australia, Korea, and the United States, the gaps between the three distributions are quite large in
both 1995 and 2015, suggesting that more advantages are offered to the privileged group. While
the gap in the United States remains unchanged and narrows somewhat in Korea, the gaps in
Australia, Hong Kong, and Sweden have roughly doubled in size.
Figure 3: Math score Distributions Conditional on HLEI
4.2.2 Conditional Distributions: Science
Figure 4 describes the conditional distributions for scores in science in 1995(left) and 2015(right).
Although the distances between the different types has widened in all countries, Hong Kong and
Korea display relatively small degrees of inequality.
Figure 4: Science Score Distributions Conditional on the HLEI
4.2.3 The SSD Test
The stochastic dominance test results according to the SSD are summarized in Table 4.
For all cases other than Hong Kong in 1995, the more privileged group first-order
stochastically dominates the less privileged group.
Table 2: Stochastic Dominance Tests (HLEI)
Mathematics Science
Country Category 1 2 3 1 2 3 Australia
1995
1 ≺1*** ≺1*** ≺1*** ≺1***
2 ≺1*** ≺1***
3
1999
1 ≺1* ≺1***
≺1*** ≺1***
2 ≺1**
≺1***
3
2003
1 ≺1*** ≺1***
≺1*** ≺1***
2 ≺2***
≺1***
3
2007
1 ≺1** ≺1***
≺1*** ≺1***
2 ≺1***
≺1***
3
2011
1 ≺1*** ≺1***
≺1*** ≺1***
2 ≺1***
≺1***
3
2015
1 ≺2*** ≺1***
≺1*** ≺1***
2 ≺1***
≺1***
3
Hong
Kong
1995
1 ≺2*** ≺1*** = ≺1***
2 ≺2*** ≺2*
3
1999
1 ≺1*** ≺1***
≺1** ≺1***
2 ≺1**
≺2***
3
2003
1 ≺2** ≺1***
≺1* ≺1***
2 ≺1**
≺1*
3
2007
1 ≺1** ≺1***
≺2*** ≺1***
2 ≺1***
≺1*
3
2011
1 ≺1*** ≺1***
≺2*** ≺1***
2 ≺1***
≺1***
3
2015
1 ≺1** ≺1***
≺1* ≺1***
2 ≺1***
≺1***
3
Korea
1995
1 ≺1*** ≺1*** ≺1*** ≺1***
2 ≺1*** ≺1*
3
1999
1 ≺1*** ≺1***
≺1*** ≺1***
2 ≺1**
≺1***
3
2003
1 ≺1*** ≺1***
≺1*** ≺1***
2 ≺1***
≺1***
3
2007 1 ≺1*** ≺1***
≺1*** ≺1***
2 ≺1***
≺1***
3
2011
1 ≺1*** ≺1***
≺1*** ≺1***
2 ≺1***
≺1***
3
2015
1 ≺1*** ≺1***
≺1** ≺1***
2 ≺1***
≺1**
3
Sweden
1995
1 ≺1*** ≺1*** ≺1*** ≺1***
2 ? ?
3
2003
1 ≺1*** ≺1***
≺1*** ≺1***
2 ≺1**
≺1***
3
2007
1 ≺1*** ≺1***
≺1*** ≺1***
2 ≺1**
≺1*
3
2011
1 ≺1** ≺1***
≺1*** ≺1***
2 ≺1***
≺1***
3
2015
1 ≺1*** ≺1***
≺1*** ≺1***
2 ≺1***
≺1***
3
USA
1995
1 ≺1*** ≺1*** ≺1*** ≺1***
2 ≺1*** ≺1***
3
1999
1 ≺1*** ≺1***
≺1*** ≺1***
2 ≺1***
≺1***
3
2003
1 ≺1*** ≺1***
≺1*** ≺1***
2 ≺1***
≺1***
3
2007
1 ≺1*** ≺1***
≺1*** ≺1***
2 ≺1***
≺1***
3
2011
1 ≺1*** ≺1***
≺1*** ≺1***
2 ≺1***
≺1***
3
2015
1 ≺1*** ≺1***
≺1*** ≺1***
2 ≺1***
≺1***
3 Notes ≺𝑖 : the column dominates the row for order i stochastic dominance. =: the distributions are equal.?: the
distributions cannot be ranked using first- and second-order stochastic dominance. Categories refers to social
background. 1: bottom one-third. 2: second one-third. 3: upper one-third.
4.3 Opportunity Inequality Index
In this section, the GOI and the RRI are multiplied by 100 and 1000, respectively. Thus, zero
represents complete inequality and 100(1000) represents complete equalities in terms of the
GOI(RRI). The values of these indices are presented in the Appendix.
4.3.1 Trend
An important point, shown in both Figure 5 and 6, is that countries with a relative low GOI value
do not necessarily have low RRI value, and vice versa. Also, for Korea, the trends in RRI using
the father’s education differ to those measured using the HLEI.
Figure 5 displays inequalities in outcomes and opportunities for Korea, the United States
and 16 countries when the circumstances measure is father’s level of education. In Korea, the
value of the RRI shows an increasing trend, which implies that students whose fathers with low
level of education tend to be less likely to score in the 80th percentile on the test than before. The
average mathematics score for 16 countries in 2015 is higher than that in 2011. However, for
Korea, all 3 groups have a lower score in 2015 than in 2011. The decline has been greatest among
those in the most advantaged group. From the increased RRI, it seems that the slight decrease in
the GOI is due to the narrowing gap between the two more privileged groups. For the United
States, the gap between the students whose father has medium level of education and the students
whose father has low level of education has been narrowed.
Figure 7 shows the differences in the values of inequality indices between mathematics
and science. In most of the countries, there is no appreciable difference in the degrees and patterns
of the GOI regardless of the different subject areas. In terms of GOI, Japan, Korea and Singapore
are the exceptional cases where there are relatively noticeable differences in the level of
inequalities of opportunity according to the subject area. GOI for mathematics are slightly higher
than those for science in Japan and Korea. In Singapore, the case is the exact opposite. In terms
of RRI, the gap regarding to the subject area is somewhat smaller and not constant.
Figure 5: Inequalities in Outcomes and Opportunities conditional on father’s education
Figure 6: Inequalities in Outcomes and Opportunities conditional on HLEI
Figure 7: Differences between Mathematics and Science
4.3.2 Rankings from 1995 to 2003
In this section, the levels of opportunity inequality are compared by subject area, across-time
changes, and within each country. The rest of the paper is divided into two parts, focusing on the
first period (from 1995 to 2003), and on the second period (from 2007 to 2015).
In the following sets of figures, the results on the left represent the average values of the
GOI when circumstances are measured using father’s education level (GOI-edu), and those on
the right are obtained when circumstances are measured using the HLEI (GOI-HLEI). The GOI-
edu rankings are not always the same as the GOI-HLEI rankings, for several reasons. First, when
partitioning students into three categories according to father’s level of education, the differences
in the number of observations in each category underestimate or overestimate the GOI. As
mentioned in section 3.2.1, the relative size of each category based on father’s level of education
varies between countries, and the relative size of each category represents between 25 percent to
42 percent of the students based on the HLEI, which is an aggregate index including parental
education, home possessions, and the number of books in the home. Second, the two factors other
than father’s education influence the academic performance of students indirectly as a proxy for
income, representing the capacity to purchase goods and services for academic support. Third, the
other two factors also influence the academic performance of students directly. The use of devices
such as computers and dictionaries and an abundance of books may foster a higher level of
academic achievement.
Figure 8: GOI in math 1995-2003 measured by father's education(left) and HLEI(right)
Figure 9: GOI in Science 1995-2003 Measured by Father's Education (left) and HLEI (right)
Figure 8 presents the average value of the GOI from 1995 to 2003 for mathematics. As
shown in the graph on the left, Japan, England and the United States are among the most unequal
countries in terms of GOI based on circumstances measured by father’s education while Hong
Kong and Canada rank as the most equal nations. The graph on the right shows that, Hong Kong,
Canada, and France are the most equal countries. A comparison of GOI-edu and GOI-HLEI shows
a aignificant difference in the case of Japan and Korea. Japan ranks first, with a value of 2.3254
(GOI-edu), but ranks ninth out of 16 countries according to GOI-HLEI (2.738). In contrast, Korea
ranks seventh in terms of GOI-edu (2.1359), but ranks second in terms of GOI-HLEI.
Figure 9 shows the average values of the GOI from 1995 to 2003 for science. Similarly,
to the results for mathematics, Hong Kong, Finland, and the Netherlands, show a low degree of
inequality in terms of GOI-edu. The United States ranks first, followed by Singapore, in terms of
both GOI-edu and GOI-HLEI. In Japan and Singapore, unlike in other countries, different features
appear in relation to the subjects. Singapore, which has an intermediate level of inequality of
opportunity in mathematics scores, is one of the most unequal country in terms of science scores.
In Japan, father’s education levels tend to have a significant impact on their children’s
mathematics achievements, but almost no difference in the case of science.
Figure 10: RRI in math 1995-2003 measured by father's education(left) and HLEI(right)
Figure 11: RRI in science 1995-2003 measured by father's education(left) and HLEI(right)
In Figure 10 and 11, the GOI is replaced by the RRI. The findings are similar to those
reported in Figure 8 and 9. With regard to mathematics, the United States and Japan are the least
equal countries according to both GOI-edu and GOI-HLEI, while Hong Kong is the most equal
country. With regard to science, England and the United States have the highest values relative to
the other countries in terms of both RRI-edu and RRI-HLEI. Furthermore, Hong Kong has a
relatively low value of both RRI-edu and RRI-HLEI, and Germany is ranked as the second least
unequal country according to RRI-edu, and fourth most unequal country according to RRI-HLEI.
4.3.3 Rankings from 2007 to 2015
In this section, the figures show the degree of inequality of opportunity measured by the GOI and
RRI in terms of arithmetic average, from 2007 to 2015. France, Germany and the Netherlands did
not participate in the program between 2007 and 2015. Therefore, these three countries are
therefore omitted from the comparison for this period.
Figure 12: GOI in math 2007-2015 measured by father's education(left) and HLEI(right)
Figure 13: GOI in science 2007-2015 measured by father's education(left) and HLEI(right)
Figure 12 and Figure 13 show the average value of the GOI from 2007 to 2015 in
mathematics and science, respectively. England, Australia and New Zealand are the least equal
countries, and Hong Kong, Finland, Sweden and Norway are the most equal countries in both
mathematics and science. Canada, in mathematics, ranks eighth according to GOI-edu, but stand
out as the second most equal country according to GOI-HLEI.
Figure 14: RRI in math 2007-2015 measured by father's education(left) and HLEI(right)
Figure 15: RRI in science 2007-2015 measured by father's education(left) and HLEI(right)
According to Figure 14, showing the average value of the RRI in mathematics, Japan,
England and New Zealand are the most unequal countries, while Italy, Hong Kong, and Canada are
the least unequal countries. The United States ranks second when the social background is partitioned
by HLEI, but is considered to belong in intermediate unequal country when the social background is
interpreted by father’s education. While Japan shows serious inequities of opportunity in achievement
in mathematics, the degree of inequality of opportunity in science is relatively low. For the United
State, the results shown in Figure 14 and 15 are made more interesting by the fact that dividing social
backgrounds by father’s education or HLEI influences the rank of the level of inequality of opportunity
in mathematics achievement. However, such differences do not exist for science.
5 Conclusion
This study examines how the differences in levels of academic achievement are related to social
background, across time and countries. International comparisons of the degree of inequality in
opportunity are based on two opportunity inequality indices.
This study first measures and compares cumulative distributions conditional on social
background using stochastic dominance. Then the comparison is implemented by ranking
countries according to two opportunity inequality indices. The results indicate that not only social
backgrounds attributes to the access of education, but also to the chance to success in education.
The two indices reveal substantial variation in the extent to which different countries achieve
equality of educational opportunity for students from different social backgrounds. In addition, in
some countries, the level of inequality varied by subject areas, suggesting that social background
advantages in education may be related to systemic features of countries’ education systems.
Particularly in Japan and Korea, students from low social backgrounds are more likely to suffer
from lack of opportunities to develop mathematic ability, as both countries have high values of
RRI and GOI in mathematics compared to science. On the other hand, students from low social
backgrounds are more likely to suffer from lack of opportunities learning science than learning
mathematics in Singapore. England and New Zealand continuously display the greatest
opportunity inequality, while the United States ranked as one of the most unequal countries during
the period of 1995-2003, has improved in terms of opportunity inequality during the second phase,
2007-2015.
References
Atkinson, A. B. (1970). On the measurement of inequality, Journal of economic theory, 2(3):244-
263.
Becker, G. (1975). Human Capital: A Theoretical and Empirical Analysis, with Special Reference
to Education (2nd ed.). New York: National Bureau of Economic Research.
Behrman, Jere, Nancy Birdsall, and Miguel Székely. (2000). Intergenerational Mobility in Latin
America: Deeper Markets and Better Schools Make a Difference, in N. Birdsall and C.
Graham (eds), New Markets, New Opportunities? Economic and Social Mobility in a
Changing World, Brookings Institution, Washington DC, 135-167.
Bourguignon, François, Francisco H. G. Ferreira, and Marta Menéndez. (2003). Inequality of
Outcomes and Inequality of Opportunities in Brazil, World Bank Policy Research
Working Paper#3174, Washington DC.
Byun, S. and Kim, K. (2010). Educational inequality in South Korea: The widening socio
economic gap in student achievement. Research in Sociology of Education,17:155-182.
Checchi, D. and V. Peragine. (2005). Regional Disparities and Inequality of
Opportunity: The Case of Italy, IZA Discussion Paper#1874, Bonn, Germany.
Davidson R. and J. Duclos. (2000). Statistical Inference for Stochastic Dominance and for the
Measurement of Poverty and Inequality. Econometrica, 68(6):1435–64.
Djavad Salehi-Isfahani, Nadia Hassine and Ragui Assaad. (2014). Equality of opportunity in
educational achievement in the Middle East and North Africa, The Journal of Economic
Inequality, 12(4):489-515.
Hamada, H. and A, Ishida. (2003). Unequal Society and Equality of Opportunity:A Measurement
Method for the Degree of Inequality in a Society of Equal Opportunity, 社会学評論,
54(3):233-249. (in Japanese)
Ishida, A. (2014). An Index of Relative Deprivation Caused by Inequality of Opportunity.
理論と方法(Sociological Theory and Methods), 29(1):81-97. (in Japanese)
Kim, J., Y. Chun, and B. Lim. (2014). Parents Education and Students' Academic Performance: A
Comparison of OECD Countries,재정학연구, 7(2):27-57. (in Korean)
Kim, W. and W. Lee. (2008). A Study on the Extent to which the Korean Tax Regime Equalizes
Opportunities for Income Acquisition among Citizens. Korea Institute of Public Finance
Research Report#08-03. (in Korean)
Ko, J. and W. Lee.(2011). Father's Education and Son's Achievement,재정학연구,4(2): 47-87.
(in Korean)
Lefranc, A., N.Pistolesi, and A.Trannoy. (2008). Inequality of opportunities vs. inequality of
outcomes:Are Western societies all alike?. Review of Income and Wealth, 54.4:513-546.
_________________________________. (2009). Equality of Opportunity and Luck: Definitions
and Testable Conditions, with an Application to Income in France. Journal of Public
Economics, 93, 2009, 1189-1207.
Oh, S. and B. Ju. (2017). Inequality of Opportunities for Income Acquisition in Korea.재정학
연구, 10(3):1-30. (in Korean)
Oh, S., C. Kang, H. Jeong, and B. Ju. (2016). Equality of Opportunity for Educational
Achievement in Korea.재정학연구, 9(4):1-32. (in Korean)
Rawls, J. (1971). A theory of justice, Cambridge, MA: Harvard University Press
Roemer, J. (1993). A pragmatic theory of responsibility for the egalitarian planner. Philosophy
and Public Affairs, 22(2): 146–166.
________ (1998). Equality of Opportunity, Harvard University Press, Cambridge.
Roemer, J. and A.Trannoy. (2016). Equality of Opportunity: Theory and Measurement. Journal
of Economic Literature, 54(4): 1288-1332.
Roemer, J., R. Aaberge, U. Colombino, et al. (2003). To what extent do fiscal regimes equalize
opportunities for income acquisition among citizens?. Journal of Public Economics, 87:
539-565.
A Appendices
A.1 Descriptive Statistics
A.1.1 Conditional on Father’s Education
Subject Mathematics Science
Country Category N Mean SD Total% N Mean SD Total%
Australi
a 1995
Low 3139 498.7 93.45 34.22 3139 508.3 102.9 34.22
Med. 3626 517.7 93.84 39.53 3626 531.6 102.4 39.53
High 2408 559.5 90.70 26.25 2408 568.6 102.4 26.25
1999
Low 816 516.3 72 34.29 816 534.4 84.09 34.29
Med. 809 535.1 75.72 33.99 809 551 81.30 33.99
High 755 569.5 70.34 31.72 755 579.8 75.25 31.72
2003
Low 639 482.3 78.35 24.29 639 508.9 74.87 24.29
Med. 1330 509.3 76.19 50.55 1330 532.8 67.35 50.55
High 662 548.5 77.40 25.16 662 569.5 67.29 25.16
2007
Low 554 479.5 72.13 27.56 554 501.6 74.67 27.56
Med. 1005 501.0 74.02 50.00 1005 520.6 74.34 50.00
High 451 554.9 74.50 22.44 451 573.2 70.73 22.44
2011
Low 340 462.9 80.89 10.57 340 476.7 78.37 10.57
Med. 1985 496.2 74.40 61.70 1985 516.2 75.99 61.70
High 892 577.3 83.23 27.73 892 590.4 78.98 27.73
2015
Low 482 461.1 84.83 11.77 482 469.8 85.14 11.77
Med. 2239 507.6 75.87 54.68 2239 515.7 75.13 54.68
High 1374 552.3 74.87 33.55 1374 559.9 76.20 33.55
Hong
Kong 1995
Low 3343 577.7 94.93 63.30 3343 510.4 85.27 63.30
Med. 1515 599.1 95.28 28.69 1515 525.6 84.76 28.69
High 423 621.5 90.89 8.010 423 540.5 83.74 8.010
1999
Low 2006 581.3 71.76 54.14 2006 530 69.23 54.14
Med. 1393 592.9 69.23 37.60 1393 534.5 68.70 37.60
High 306 613.2 68.55 8.260 306 565.4 65.33 8.260
2003
Low 1906 584.5 67.56 49.55 1906 554.5 64.21 49.55
Med. 1478 591.8 66.08 38.42 1478 559.8 61.49 38.42
High 463 618 72.09 12.04 463 578 65.15 12.04
2007
Low 1161 565.4 90 46.20 1161 527.9 81.52 46.20
Med. 1000 580.1 87.17 39.79 1000 537.2 77.53 39.79
High 352 613.4 86.79 14.01 352 561.6 73.78 14.01
2011
Low 970 571 80.16 35.69 970 526.2 71.91 35.69
Med. 1237 593.9 76.72 45.51 1237 540 69.69 45.51
High 511 632.5 68.73 18.80 511 574 63.19 18.80
2015
Low 732 580.3 77.45 27.58 732 531.4 72.42 27.58
Med. 1374 596.3 74.63 51.77 1374 546.4 68.26 51.77
High 548 630.5 67.44 20.65 548 581.7 66.90 20.65
Korea
1995
Low 1433 567 103.8 28.02 1433 536.2 91.61 28.02
Med. 2475 594.4 104.0 48.39 2475 552.1 92.48 48.39
High 1207 640.5 99.10 23.60 1207 577.9 90.10 23.60
1999
Low 1098 569.4 77.53 21.45 1098 533.7 81.83 21.45
Med. 2627 588.3 72.74 51.31 2627 552.8 81.45 51.31
High 1395 625.9 72.15 27.25 1395 586.9 83.27 27.25
2003
Low 527 545.4 86.99 11.19 527 525.9 73.54 11.19
Med. 2548 584.7 76.45 54.09 2548 554.6 65.24 54.09
High 1636 618.8 75.16 34.73 1636 583.2 65.80 34.73
2007
Low 189 553.1 96.50 5.500 189 521.6 80.12 5.500
Med. 1574 584.3 83.50 45.82 1574 545.5 72.49 45.82
High 1672 630.2 85.42 48.68 1672 579.8 70.89 48.68
2011 Low 169 573.1 93.18 4.550 169 526.4 78.77 4.550
Med. 1776 599.4 83.25 47.77 1776 553.6 72.52 47.77
High 1773 652.2 78.78 47.69 1773 591 71.58 47.69
2015
Low 66 570.3 83.42 1.980 66 523.3 72.34 1.980
Med. 1560 595.6 79.15 46.85 1560 547.7 70.19 46.85
High 1704 637.3 78.04 51.17 1704 583 73.75 51.17
Sweden
1995
Low 917 514.7 87.46 19.78 917 530.7 95.33 19.78
Med. 2496 527.9 86.09 53.84 2496 542.8 91.05 53.84
High 1223 548.9 89.31 26.38 1223 567.6 97.83 26.38
2003
Low 299 488.8 66.84 17.61 299 518.5 71.58 17.61
Med. 717 505.4 66.42 42.23 717 532.8 70.21 42.23
High 682 529.9 69.96 40.16 682 556.4 72.89 40.16
2007
Low 297 478.3 66.47 16 297 496.3 74.78 16
Med. 976 501.4 63.74 52.59 976 529.8 72.30 52.59
High 583 520.4 66.48 31.41 583 548.5 73.58 31.41
2011
Low 276 474.6 65.96 11.55 276 502.7 81.28 11.55
Med. 1269 495.1 63.12 53.12 1269 527.9 73.56 53.12
High 844 518 63.63 35.33 844 550.7 71.91 35.33
2015
Low 195 472.6 66.55 10.32 195 491.6 88.59 10.32
Med. 916 505.6 64.91 48.47 916 532.3 73.10 48.47
High 779 535.7 65.68 41.22 779 562.9 81.18 41.22
Canada
1995
Low 2737 503.7 81.51 24.02 2737 503.7 91.31 24.02
Med. 4550 516.1 82.75 39.93 4550 518 90.40 39.93
High 4107 532.1 84.02 36.05 4107 540.5 90.48 36.05
1999
Low 942 516.8 72.40 15.15 942 521.9 75.97 15.15
Med. 2518 533.7 70.38 40.49 2518 536.3 74.23 40.49
High 2759 550.2 70.61 44.36 2759 551.3 77.34 44.36
2015
Low 181 504.6 74.27 4.080 181 504.9 75.40 4.080
Med. 2355 523.9 63.88 53.05 2355 526.2 63.24 53.05
High 1903 565.3 62.95 42.87 1903 563.8 66.06 42.87
German
y 1995
Low 1813 498.4 84.19 53.80 1813 518.8 94.92 53.80
Med. 1167 519.6 83.96 34.63 1167 541.3 91.47 34.63
High 390 542.7 87.71 11.57 390 560.3 103.7 11.57
England
2011
Low 59 490.0 79.55 9.770 59 540.5 67.98 9.770
Med. 367 529.0 70.43 60.76 367 576.1 71.93 60.76
High 178 564.3 68.66 29.47 178 613.8 65.70 29.47
2015
Low 150 472.3 85.62 9.750 150 490.8 93.25 9.750
Med. 970 522.3 77.99 63.03 970 554.0 75.92 63.03
High 419 569.1 71.62 27.23 419 596.2 72.59 27.23
2003
Low 209 509.9 78.08 13.05 209 522.7 82.17 13.05
Med. 739 526.5 71.98 46.16 739 550.0 73.59 46.16
High 653 580.0 74.70 40.79 653 600.0 75.76 40.79
Finland
1999
Low 391 511.2 61.86 38.64 391 538.5 74.19 38.64
Med. 497 540.8 62.57 49.11 497 564.0 77.59 49.11
High 124 566.8 61.69 12.25 124 584.3 75.71 12.25
2011
Low 281 491.0 59.74 11.59 281 533.5 64.83 11.59
Med. 1399 517.1 61.48 57.69 1399 558.0 61.51 57.69
High 745 543.3 60.65 30.72 745 583.7 62.44 30.72
France
1995
Low 981 516.9 78.21 39.46 981 474.2 79.05 39.46
Med. 1085 539.5 74.05 43.64 1085 495.3 76.36 43.64
High 420 560 72.82 16.89 420 506.9 74.58 16.89
Italy
1999
Low 1591 465.0 83.45 53.10 1591 476.6 85.90 53.10
Med. 1165 498.8 80.16 38.89 1165 519.3 80.63 38.89
High 240 523.7 73.20 8.010 240 542.9 75.23 8.010
2003
Low 1733 463.1 74.89 45.25 1733 471.1 77.09 45.25
Med. 1486 500.5 71.02 38.80 1486 507.1 73.45 38.80
High 611 506.8 73.29 15.95 611 510.9 76.68 15.95
2007 Low 1484 459.6 75.15 40.85 1484 477.2 79.90 40.85
Med. 1485 500.7 68.74 40.88 1485 514.4 70.01 40.88
High 664 505.5 69.86 18.28 664 524.2 67.13 18.28
2011 Low 1238 483.9 70.74 38.32 1238 480.3 74.45 38.32
Med. 1431 505.8 70.33 44.29 1431 510.2 72.97 44.29
High 562 527.6 65.69 17.39 562 531.6 66.46 17.39
2015
Low 1178 478.8 72.28 35.41 1178 482.5 73.22 35.41
Med. 1600 506.7 68.65 48.09 1600 510.7 71.11 48.09
High 549 528.6 71.14 16.50 549 532.5 71.32 16.50
Netherl
ands 1995
Low 409 510.0 82.75 16.99 409 525.1 82.57 16.99
Med. 1629 539.5 82.54 67.65 1629 550.9 81.21 67.65
High 370 562.7 84.26 15.37 370 566.4 87.35 15.37
1999
Low 230 523.9 70.58 13.31 230 516.4 85.10 13.31
Med. 1227 553.3 64.51 71.01 1227 562.4 66.15 71.01
High 271 569.9 64.77 15.68 271 576.3 74.58 15.68
2003
Low 82 518.6 65.41 4.550 82 509.6 61.66 4.550
Med. 1333 544.1 63.40 73.89 1333 543.4 55.57 73.89
High 389 573.3 63.82 21.56 389 564.4 56.33 21.56
Norway
1995
Low 459 471.2 84.70 14.98 459 493.3 85.19 14.98
Med. 1638 488.9 80.76 53.44 1638 512.8 86.02 53.44
High 968 508.2 84.53 31.58 968 529.5 88.54 31.58
2003
Low 138 425.4 70.42 8.270 138 471.9 73.86 8.270
Med. 519 465.5 64.60 31.12 519 502.2 62.30 31.12
High 1011 489.2 65.60 60.61 1011 519.8 63.87 60.61
2007
Low 119 436.5 72.09 6.470 119 450.9 82.13 6.470
Med. 565 470.6 63.88 30.71 565 493.1 71.41 30.71
High 1156 490.5 60.59 62.83 1156 513.9 69.07 62.83
2011
Low 93 449.8 69.10 5.230 93 448.7 86.03 5.230
Med. 706 476.9 58.77 39.69 706 501.6 67.38 39.69
High 980 503.4 59.23 55.09 980 525.0 66.83 55.09
2015
Low 118 487.1 59.35 5.100 118 483.5 68.72 5.100
Med. 997 506.9 67.65 43.07 997 507.1 75.58 43.07
High 1200 537.8 65.88 51.84 1200 542.9 73.57 51.84
New
Zealand 1995
Low 1151 486.5 81.96 27.67 1151 501.5 92.03 27.67
Med. 1974 493.3 87.69 47.46 1974 512.0 99.41 47.46
High 1034 530.2 91.59 24.86 1034 545.3 100.8 24.86
1999
Low 605 484.7 81.66 28.06 605 501.3 84.95 28.06
Med. 924 489.0 82.99 42.86 924 514.7 82.35 42.86
High 627 536.4 84.71 29.08 627 555.5 88.67 29.08
2003
Low 170 480.6 74.29 11.64 170 500.6 61.07 11.64
Med. 914 500.3 69.66 62.56 914 533.4 66.73 62.56
High 377 540.8 75.45 25.80 377 566.8 66.94 25.80
2011
Low 365 463.7 81.08 20.61 365 487.6 81.86 20.61
Med. 971 497.2 81.54 54.83 971 521.8 81.88 54.83
High 435 543.9 79.77 24.56 435 566.5 80.71 24.56
2015
Low 447 479.1 86.38 18.55 447 499.6 92.05 18.55
Med. 1151 502.1 85.53 47.76 1151 526.7 91.62 47.76
High 812 548.3 88.84 33.69 812 568 91.88 33.69
Singapo
re 1995
Low 2614 601.3 91.47 31.80 2614 550.3 100.5 31.80
Med. 5096 627.9 91.92 62 5096 583.0 100.8 62
High 509 673.6 85.43 6.190 509 640.3 91.73 6.190
1999
Low 1465 592.6 78.42 39.10 1465 551.7 94.84 39.10
Med. 1890 609.3 74.53 50.44 1890 577.8 91.87 50.44
High 392 651.6 73.89 10.46 392 636.0 85.25 10.46
2003
Low 2658 598.0 77.49 63.62 2658 567.5 87.74 63.62
Med. 918 624.9 69.31 21.97 918 608.5 79.16 21.97
High 602 654.5 67.77 14.41 602 642.3 72.07 14.41
2007
Low 734 567.7 89.48 24.75 734 531.0 101.9 24.75
Med. 1587 600.6 85.06 53.51 1587 576.5 93.19 53.51
High 645 648.9 75.32 21.75 645 640 81.29 21.75
2011
Low 841 586.9 84.37 22.85 841 556.0 98.11 22.85
Med. 1837 615.3 74.98 49.92 1837 596.3 86.30 49.92
High 1002 651.2 70.60 27.23 1002 644.8 81.10 27.23
2015
Low 635 590.5 82.83 16.99 635 561.9 87.86 16.99
Med. 1857 617.6 77.19 49.69 1857 594.0 79.09 49.69
High 1245 664.8 63.93 33.32 1245 648.4 68.58 33.32
USA
1995
Low 1362 453 76.34 15.34 1362 483.3 92.97 15.34
Med. 5078 485.3 85.46 57.21 5078 522.3 102.6 57.21
High 2436 526.1 95.37 27.44 2436 557.4 104.2 27.44
1999
Low 739 462 78.11 11.71 739 480.2 89.31 11.71
Med. 3420 505.5 77.11 54.18 3420 518.2 87.65 54.18
High 2153 545.1 83.55 34.11 2153 563.1 90.98 34.11
2003
Low 804 458.6 69.98 13.85 804 482.2 73.19 13.85
Med. 2316 498.2 71.38 39.90 2316 522 72.31 39.90
High 2684 538.6 76.57 46.24 2684 562 76.28 46.24
2007
Low 776 470.8 66.54 16.62 776 472.3 74.41 16.62
Med. 1895 504.3 69.54 40.59 1895 519.8 74.36 40.59
High 1998 540.7 73.14 42.79 1998 553.9 76.86 42.79
2011
Low 968 476.2 70.32 15.47 968 487.5 79.03 15.47
Med. 2593 503.9 67.85 41.45 2593 524.5 73.70 41.45
High 2695 543.9 72.72 43.08 2695 562.7 74.54 43.08
2015
Low 1018 496.1 81.88 15.84 1018 505.1 80 15.84
Med. 2954 512.3 77.04 45.98 2954 525.9 78.01 45.98
High 2453 554.3 78.08 38.18 2453 564.1 76.41 38.18
A.1.2 Descriptive Statistics Conditional on HLEI
Subject Mathematics Science
Country Category N Mean SD Total% N Mean SD Total%
Australi
a 1995
1 3284 488.7 91.64 35.80 3284 497.2 99.28 35.80
2 2960 523.5 91.79 32.27 2960 535.2 101.3 32.27
3 2929 560.4 90.14 31.93 2929 573.9 100.6 31.93
1999
1 794 511.2 73.43 33.36 794 520.9 83.35 33.36
2 814 539.3 71.48 34.20 814 556.3 77.68 34.20
3 772 570.4 71.21 32.44 772 588.8 71.48 32.44
2003
1 901 479.2 78.93 34.25 901 502.4 71.25 34.25
2 881 517.9 77.19 33.49 881 538.2 65.11 33.49
3 849 542.3 72.35 32.27 849 569.8 64.99 32.27
2007
1 744 471.4 74.13 37.01 744 488.2 75.44 37.01
2 611 507.0 67.59 30.40 611 531.6 66.37 30.40
3 655 547.7 72.76 32.59 655 567.1 69.34 32.59
2011
1 1154 465.1 72.47 35.87 1154 478.4 72.00 35.87
2 1008 516.5 70.01 31.33 1008 534.7 67.77 31.33
3 1055 570.4 84.20 32.79 1055 592.2 75.75 32.79
2015
1 1375 476.1 83.09 33.58 1375 478.6 80.43 33.58
2 1400 517.7 70.23 34.19 1400 529.0 69.53 34.19
3 1320 560.4 69.01 32.23 1320 570.9 67.40 32.23
Hong
Kong 1995
1 1916 573.9 96.63 36.28 1916 508.0 84.61 36.28
2 1696 584.6 94.60 32.12 1696 514.6 85.56 32.12
3 1669 605.4 92.86 31.60 1669 530.2 85.01 31.60
1999
1 1287 571.2 71.77 34.74 1287 520.0 70.91 34.74
2 1327 588.3 68.75 35.82 1327 535.2 67.57 35.82
3 1091 609.1 67.66 29.45 1091 551.7 65.47 29.45
2003
1 1321 577.5 68.16 34.34 1321 547.1 65.38 34.34
2 1248 588.4 66.91 32.44 1248 558.8 61.57 32.44
3 1278 608.4 65.99 33.22 1278 572.7 61.15 33.22
2007
1 967 552.2 91.40 38.48 967 518.0 83.48 38.48
2 825 580.6 86.59 32.83 825 537.4 77.88 32.83
3 721 610.5 79.73 28.69 721 560.3 69.08 28.69
2011 1 976 566.1 83.10 35.91 976 521.5 74.66 35.91
2 915 593.1 74.08 33.66 915 539.6 66.48 33.66
3 827 624.6 68.58 30.43 827 567.1 64.15 30.43
2015
1 955 575.7 76.81 35.98 955 528.1 72.69 35.98
2 828 599.1 72.04 31.20 828 548.5 65.36 31.20
3 871 624.3 70.50 32.82 871 574.2 67.59 32.82
Korea
1995
1 1988 558.6 104.2 38.87 1988 527.8 92.11 38.87
2 1517 601.6 99.42 29.66 1517 557.8 90.56 29.66
3 1610 641.6 96.49 31.48 1610 581.8 87.09 31.48
1999
1 1737 561.6 75.02 33.93 1737 528.6 79.37 33.93
2 1700 596.0 69.08 33.20 1700 555.9 80.60 33.20
3 1683 627.1 70.44 32.87 1683 590.8 81.00 32.87
2003
1 1952 560.0 78.71 41.43 1952 535.1 66.57 41.43
2 1221 593.8 72.51 25.92 1221 563.1 62.44 25.92
3 1538 631.4 70.90 32.65 1538 593.1 62.77 32.65
2007
1 1149 560.8 87.62 33.45 1149 526.6 73.38 33.45
2 1203 607.9 78.88 35.02 1203 562.2 67.70 35.02
3 1083 648.6 77.80 31.53 1083 595.9 66.37 31.53
2011
1 1154 465.1 72.47 35.87 1154 478.4 72.00 35.87
2 1008 516.5 70.01 31.33 1008 534.7 67.77 31.33
3 1055 570.4 84.20 32.79 1055 592.2 75.75 32.79
2015
1 1375 476.1 83.09 33.58 1375 478.6 80.43 33.58
2 1400 517.7 70.23 34.19 1400 529.0 69.53 34.19
3 1320 560.4 69.01 32.23 1320 570.9 67.40 32.23
Sweden
1995
1 1607 504.1 83.29 34.66 1607 520.6 91.60 34.66
2 1585 537.7 85.88 34.19 1585 553.3 90.82 34.19
3 1444 553.2 87.93 31.15 1444 569.4 95.21 31.15
2003
1 567 476.8 63.89 33.39 567 503.4 68.02 33.39
2 573 518.0 64.72 33.75 573 543.8 68.40 33.75
3 558 541.5 64.51 32.86 558 571.4 66.02 32.86
2007
1 634 475.8 63.39 34.16 634 495.8 73.41 34.16
2 642 507.4 60.80 34.59 642 535.9 67.73 34.59
3 580 532.2 63.28 31.25 580 564.6 67.25 31.25
2011
1 850 474.0 65.16 35.58 850 499.6 77.00 35.58
2 745 500.3 57.64 31.18 745 532.5 65.89 31.18
3 794 529.0 59.68 33.24 794 568.3 65.41 33.24
2015
1 637 482.5 66.33 33.70 637 501.8 75.71 33.70
2 683 515.5 60.28 36.14 683 539.7 75.64 36.14
3 570 552.9 59.66 30.16 570 590.2 65.51 30.16
Canada 1995
1 4015 504.5 83.10 35.24 4015 495.1 88.21 35.24
2 3619 517.0 81.70 31.76 3619 524.8 91.10 31.76
3 3760 535.7 83.04 33 3760 549.5 87.77 33
1999 1 2187 520.7 71.48 35.17 2187 518.2 74.25 35.17
2 2097 538.1 68.89 33.72 2097 544.5 76.40 33.72
3 1935 557.8 70.16 31.11 1935 561.2 72.84 31.11
2015 1 1506 513.2 67.91 33.93 1506 510.9 66.49 33.93
2 1455 541.9 61.70 32.78 1455 542.9 60.98 32.78
3 1478 568.8 60.04 33.30 1478 572.9 60.75 33.30
German
y 1995 1 1303 480.5 80.30 38.66 1303 499.5 89.88 38.66
2 947 515.1 82.45 28.10 947 537.1 95.21 28.10
3 1120 541.4 83.20 33.23 1120 562.2 92.04 33.23
England 2011
1 203 501.8 72.95 33.61 203 548.4 70.85 33.61
2 202 543.4 66.16 33.44 202 587.7 62.82 33.44
3 199 563.4 69.51 32.95 199 617.2 68.39 32.95
2015 1 603 493.1 82.40 39.18 603 517.8 82.24 39.18
2 447 533.2 71.61 29.04 447 566.3 67.60 29.04
3 489 573.1 67.66 31.77 489 604.4 68.08 31.77
2003 1 547 500.4 72.72 34.17 547 516.2 72.05 34.17
2 529 549.1 68.56 33.04 529 573.4 69.08 33.04
3 525 592.3 67.20 32.79 525 614.5 68.53 32.79
Finland 1999
1 349 506.1 63.89 34.49 349 529.2 77.34 34.49
2 346 534.9 57.56 34.19 346 561.3 70.56 34.19
3 317 558.1 62.57 31.32 317 580.7 76.20 31.32
2011 1 847 498.1 59.97 34.93 847 536.4 61.28 34.93
2 792 524.9 60.18 32.66 792 565.7 59.83 32.66
3 786 544.8 60.30 32.41 786 588.7 59.97 32.41
France 1995
1 844 514.4 78.43 33.95 844 470.8 78.52 33.95
2 843 534.4 74.01 33.91 843 488.4 77.61 33.91
3 799 554.8 73.35 32.14 799 508.8 73.43 32.14
Italy 1999
1 1017 452.3 83.11 33.95 1017 462.7 84.23 33.95
2 1013 484.8 79.74 33.81 1013 497.9 82.21 33.81
3 966 513.2 77.16 32.24 966 537.1 76.24 32.24
2003 1 1415 454.6 73.35 36.95 1415 463.3 75.72 36.95
2 1139 487.8 71.40 29.74 1139 493.6 74.50 29.74
3 1276 514.8 69.33 33.32 1276 520.6 71.84 33.32
2007 1 1300 453.6 74.78 35.78 1300 468.9 78.27 35.78
2 1230 493.3 70.57 33.86 1230 509 71.47 33.86
3 1103 511.8 65.26 30.36 1103 529.5 65.16 30.36
2011 1 1143 472.4 70.30 35.38 1143 470.1 73.97 35.38
2 1073 504.3 67.18 33.21 1073 504.9 69.58 33.21
3 1015 530.8 63.78 31.41 1015 536.8 64.84 31.41
2015 1 1116 467.1 72.84 33.54 1116 468.8 73.88 33.54
2 1109 504.4 63.73 33.33 1109 508.4 63.55 33.33
3 1102 531.6 65.30 33.12 1102 538.3 66.96 33.12
Netherl
ands 1995 1 946 512.1 83.05 39.29 946 528.1 79.62 39.29
2 673 541.2 80.44 27.95 673 550.6 81.37 27.95
3 789 568.1 77.95 32.77 789 573.7 82.37 32.77
1999 1 598 530.5 66.09 34.61 598 532.7 72.28 34.61
2 622 552.8 66.05 36 622 562.3 68.69 36
3 508 575.2 59.75 29.40 508 582.0 68.98 29.40
2003 1 628 520.8 61.84 34.81 628 519.0 55.06 34.81
2 595 551.8 62.13 32.98 595 550.3 52.55 32.98
3 581 578.6 57.18 32.21 581 573.2 49.90 32.21
Norway
1995 1 1073 470.2 79.09 35.01 1073 494.6 84.82 35.01
2 1127 495.4 82.82 36.77 1127 518.5 87.15 36.77
3 865 516.4 82.74 28.22 865 536.8 85.41 28.22
2003 1 561 445.6 67.33 33.63 561 484.5 67.38 33.63
2 606 484.9 62.63 36.33 606 518.7 59.82 36.33
3 501 502.6 62.02 30.04 501 530.6 60.76 30.04
2007 1 618 450.3 65.75 33.59 618 466.6 75.12 33.59
2 668 491.8 55.92 36.30 668 514.6 60.78 36.30
3 554 502.7 58.21 30.11 554 532.3 65.20 30.11
2011 1 643 464.3 60.96 36.14 643 480.8 70.46 36.14
2 643 497.8 56.77 36.14 643 521.3 64.52 36.14
3 493 513.8 56.28 27.71 493 540.1 62.96 27.71
2015 1 820 492.1 66.06 35.42 820 488.9 74.66 35.42
2 751 523.2 65.04 32.44 751 527.1 69 32.44
3 744 553.5 59.41 32.14 744 561.0 68.49 32.14
New
Zealand 1995 1 1402 469.2 82.99 33.71 1402 483.6 93.87 33.71
2 1428 499.4 83.97 34.34 1428 517.4 94.51 34.34
3 1329 535.5 87 31.95 1329 553.5 96.74 31.95
1999 1 738 460.6 79.62 34.23 738 477.1 81.96 34.23
2 705 504.4 79.88 32.70 705 529.1 77.56 32.70
3 713 541.0 79.15 33.07 713 563.6 80.69 33.07
2003 1 491 479.3 68.76 33.61 491 506.1 61.50 33.61
2 503 501.6 69.63 34.43 503 533.9 64.37 34.43
3 467 544.5 69.79 31.96 467 574.7 63.28 31.96
2011 1 607 457.0 85.32 34.27 607 475.7 83.31 34.27
2 606 509.7 74.93 34.22 606 531.4 71.77 34.22
3 558 544.8 71.40 31.51 558 576.4 70.07 31.51
2015 1 822 463.7 87.16 34.11 822 480.6 93.50 34.11
2 800 522.9 80.94 33.20 800 547.9 84.08 33.20
3 788 560.2 73.55 32.70 788 586.0 71.76 32.70
1 2810 590.3 90.99 34.19 2810 536.5 97.08 34.19
Singapo
re 1995
2 2699 624.8 89.96 32.84 2699 575.9 97.54 32.84
3 2710 651.2 88.29 32.97 2710 615.4 97.53 32.97
1999 1 1277 578.6 76.81 34.08 1277 532.5 92.77 34.08
2 1257 606.2 73.96 33.55 1257 571.0 87.75 33.55
3 1213 636.6 71.85 32.37 1213 617.3 86.72 32.37
2003 1 1490 578.2 79.50 35.66 1490 542.2 90.82 35.66
2 1472 616.9 67.85 35.23 1472 593.4 76.03 35.23
3 1216 646.8 67.50 29.10 1216 634.1 71.13 29.10
2007 1 1071 558.3 88.30 36.11 1071 522.8 99.12 36.11
2 939 609.7 75.15 31.66 939 586.0 85.21 31.66
3 956 645.1 77.81 32.23 956 633.7 80.55 32.23
2011 1 1260 583.8 82 34.24 1260 552.3 94.53 34.24
2 1201 620.5 74.11 32.64 1201 602.1 83.61 32.64
3 1219 651.7 66.37 33.13 1219 646.8 75.33 33.13
2015 1 1280 591.5 81.51 34.25 1280 563.5 82.62 34.25
2 1222 629.3 74.09 32.70 1222 606.0 77.30 32.70
3 1235 665.9 61.39 33.05 1235 651.0 65.48 33.05
USA 1995
1 3019 452.6 77.65 34.01 3019 482.3 97.39 34.01
2 2908 487.4 84.96 32.76 2908 526.4 100.6 32.76
3 2949 532.6 89.30 33.22 2949 566.9 97.78 33.22
1999 1 2129 471.0 76.54 33.73 2129 481.1 88.03 33.73
2 2217 516.6 74.52 35.12 2217 533.8 84.33 35.12
3 1966 555.3 78.54 31.15 1966 573.7 83.17 31.15
2003 1 2024 470.3 71.05 34.87 2024 491.6 72.21 34.87
2 1863 513.4 68.86 32.10 1863 538.6 70.37 32.10
3 1917 552.1 73.66 33.03 1917 576.3 71.12 33.03
2007 1 1577 475.2 69.71 33.78 1577 482.0 77.11 33.78
2 1694 516.3 67.68 36.28 1694 530.7 71.28 36.28
3 1398 555.2 66.61 29.94 1398 570.7 68.73 29.94
2011 1 2086 479.9 67.40 33.34 2086 492.0 74.65 33.34
2 2321 516.9 69.31 37.10 2321 537.3 71.45 37.10
3 1849 557.4 67.65 29.56 1849 580.0 67.98 29.56
2015 1 2278 488.9 78.81 35.46 2278 500.0 79.18 35.46
2 2079 524.7 73.12 32.36 2079 537.1 72.91 32.36
3 2068 565.9 73.47 32.19 2068 576.6 71.07 32.19
A.2 Conditional Distributions
A.2.1 Conditional on Father’s Education
A.2.2 Conditional on HLEI
A.3 Opportunity Inequality Index
GOI RRI
Country Year Math Science Math Science
Australia 1995 3.0222 3.0222 65.7218 65.7218
1999 2.6125 2.0723 41.0067 62.3558
2003 2.8628 2.774 44.6446 61.2122
2007 2.3268 2.3268 41.0067 51.6599
2011 3.0222 3.0222 65.7218 63.8491
2015 3.0222 2.8628 63.8491 63.8491
Canada 1995 1.6628 2.6331 44.6446 44.6446
1999 1.7906 1.7015 62.3558 61.2122
2015 2.0723 1.7906 44.6446 47.1885
Germany 1995 2.6125 1.9167 41.0067 51.6599
England 1999 2.6331 2.6125 76.7166 69.5291
2003 2.774 2.6331 80.1462 80.1462
2007 2.774 2.774 61.2122 51.6599
Finland 2011 1.9167 1.7906 51.6599 65.7218
2015 1.7015 1.6628 47.1885 61.2122
France 1995 2.6331 2.3268 51.6599 47.1885
Hong Kong 1995 1.6628 2.6331 44.6446 44.6446
1999 1.7906 1.7015 62.3558 61.2122
2003 2.0723 1.7906 44.6446 47.1885
2007 2.6125 1.9167 41.0067 51.6599
2011
2015
1.7906
1.6628
1.6628
1.6628
37.5703
37.5703
37.5703
37.5703
Italy 1999 2.6331 2.774 44.6446 44.6446
2003 2.6125 2.6331 47.1885 44.6446
2007 1.9167 2.0723 37.5703 41.0067
2011 1.9029 2.3268 41.0067 37.5703
2015 1.9029 1.9167 37.5703 41.0067
Japan 2003 2.774 1.9167 63.8491 76.7166
2007 2.6331 1.9167 63.8491 65.7218
2011 2.3268 1.9167 63.8491 76.7166
2015 2.8628 2.3268 47.1885 80.1462
Korea 1995 2.8628 2.6125 47.1885 69.5291
1999
2003
2.0723
2.3268
1.9167
2.3268
63.8491
62.3558
47.1885
62.3558
2007 1.9029 1.9029 51.6599 62.3558
2011 2.6125 2.0723 61.2122 69.5291
2015 1.7906 1.6628 76.7166 69.5291
Netherlands 1995 1.9029 1.7015 76.7166 76.7166
1999
2003
1.7015
1.7015
1.9029
1.7015
65.7218
65.7218
63.8491
65.7218
Norway 1995 1.9167 1.7906 63.8491 62.3558
2003 2.0723 1.9029 61.2122 63.8491
2007 1.7906 1.7906 47.1885 61.2122
2011 1.7906 1.9029 69.5291 47.1885
2015 1.6628 1.9029 80.1462 76.7166
New Zealand 1995 2.3268 2.0723 62.3558 61.2122
1999 2.3268 2.3268 61.2122 51.6599
2003 1.9029 2.0723 80.1462 47.1885
2011 2.8628 2.774 62.3558 65.7218
2015 2.6125 2.6331 62.3558 62.3558
Singapore 1995
1999 2.0723
1.9029
2.774
2.6125
69.5291
47.1885
63.8491
41.0067
2003 1.9167 3.0222 41.0067 41.0067
2007 2.6125 2.6331 61.2122 47.1885
2011 2.0723 2.8628 51.6599 44.6446
2015 2.6331 3.0222 65.7218 61.2122
Sweden 1995
2003 1.7015
1.7906
1.9029
1.7906
61.2122
51.6599
41.0067
51.6599
2007 1.6628 1.7015 44.6446 44.6446
2011
2015
1.6628
1.9167
1.7015
2.0723
44.6446
69.5291
51.6599
65.7218
USA 1995 2.774 2.8628 80.1462 80.1462
1999
2003
2.774
3.0222
2.6331
2.8628
69.5291
69.5291
69.5291
80.1462
2007 2.0723 2.6125 62.3558 63.8491
2011 2.6331 2.6125 76.7166 62.3558
2015 2.3268 2.6125 51.6599 44.6446