University of Southern California Dissertations and Theses

An empirical analysis of the quality of primary education across countries and over time

(USC Thesis Other)

An empirical analysis of the quality of primary education across countries and over time

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

AN EMPIRICAL ANALYSIS OF THE QUALITY OF PRIMARY EDUCATION

ACROSS COUNTRIES AND OVER TIME

by

Ike I. Song

A Thesis Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree of
MASTER OF ARTS
(ECONOMICS)

May 2010

Copyright 2010 Ike I. Song

ii
Dedication

Dedicated to
my family,
and the millions of
students and teachers.

iii
Acknowledgements

I would like to express my gratitude to the committee members, faculty, graduate
students, and staff of the Department of Economics at USC and scholars outside of USC.
I appreciate the help of Professor Cheng Hsiao for his time and immediate feedback,
Professor Yong J. Kim for his insight and considerate follow-ups, Professor Manochehr
Rashidian and Sripad Motiram for answering many of my questions, and especially,
Professor Jeff Nugent for the countless editing sessions and encouragement. I am still
amazed by Professor Nugent’s knowledge of economics and history of the 123 countries
examined in this research. I will always be grateful for Dr. Nugent’s guidance and
supporting me since day 1 at USC.
Thank you, Heon Jae Song, Bo Kim, and Xiao Huang for your comments and
suggestions. I appreciate the administrative assistance of Young Miller, Morgan Ponder,
and Jennifer Brown. Lastly, I would like to thank Richard Jung and Charles Yoo for their
fantastic input.
Completing this paper would not have been possible without the aid of these
wonderful people and brilliant committee members.

Committee Members:

Jeffrey B. Nugent (Chair)
Professor of Economics
B.A. Amherst College, 1957
M.A. New School for Social Research, 1961
Ph.D. New School for Social Research, 1965

iv

Cheng Hsiao
Professor of Economics
B.A. National Taiwan University, 1965
B.Phil., Oxford University, 1968
M.S. Stanford University, 1970
Ph.D. Stanford University, 1972

Yong Jin Kim
Assistant Professor of Economics
B.A. Trinity College, University of Cambridge, 1996
M.S. London School of Economics, 1997
Ph.D. London School of Economics, 2003

v
Table of Contents

Dedication ii

Acknowledgements iii

Abstract viii

Chapter 1: Introduction 1

Chapter 2: Literature Review and Modeling 5
2.1 Pupil-to-Teacher Ratio (PTR) 6
2.2 GDP Per Capita (GDPpc) 10
2.3 Gini Index 12
Figure 1 – Noteworthy Figures from Appendix D 14
2.4 Inequality and Governance Factors 14
(Democ and Polcon)
2.5 Population Growth (PopGr) 16

Chapter 3: Descriptive Statistics and Preliminary Analysis 18
3.1 Data Information 18
3.2 Regional Statistics by Variables 18
(a) Education Data 20
(b) Living Standard Data 22
(c) Political Equality Data 23
(d) Population Growth Data 24
(e) Primary Export Data 25
(f) Summary 26
3.3 Correlation Matrix 26
(a) Living Standard and Education 26
(b) Political Equality and Education 28
(c) Population Growth and Education 28
(d) Primary Export and Education 29
(e) Summary 29

Chapter 4: Regressions and Empirical Findings 31
4.1 Educational Quality Function 32
4.2 Instrumental Variable 32
(a) Table 3 Results 35
(b) Hausman Test: Fixed-Effects
versus Random-Effects Modeling 37

vi
4.3 Interaction Terms 38
(a) Table 4 Results 40
(b) Table 5 Results 43
(c) Hausman Test: Fixed-Effects
versus Random-Effects Modeling 45
4.4 Summary 46

Chapter 5: Summary and Conclusions 49

References 52

Appendices 55
Appendix A: Correlation Matrix using the same
data-set as in Table 2 55

Appendix B: Testing Residuals of Democ and Polcon 56

Appendix C: Gini*lnGDPpc versus Gini*Governance 57

Appendix D: Country List and Average Figures 59

vii
List of Tables

Table 1 – Averages of Raw Data (top) and Rate of Change (bottom as %) 19
of Primary Educational, Economic, Political, and Demographic
Measures Across the Globe. 1950-2005 at five year intervals

Table 2 – Correlation Matrix of the Five Regions and OECD Countries 27
between 1950 and 2005 in five year intervals. OECD Countries
are in Italics

Table 3 – Regression Outputs of Fixed and Random Effects. PTR as 33
dependent variable

Table 4 – Regression Output of Fixed and Random Effects using Interaction 41
Terms. PTR as dependent variable

Table 5 – Regression Output of Fixed and Random Effects using Interaction 44
Terms and Gini*ln GDP pc. PTR as dependent variable

viii
Abstract

The objective of this study is to empirically examine the effects of economic
conditions, political factors, and population growth on schooling quality at the primary
level across various parts of the world and over time. Quality of education is measured
primarily in terms of pupil-to-teacher ratios. The data employed is a panel data-set of 123
countries representing Asia, Latin America, Middle East and North Africa, Sub-Saharan
Africa, Transitional economies, and developed OECD countries between 1950 and 2005
in five year intervals. The results show that the quality of primary education is positively
affected by increases in GDP per capita and more democratic governance, and negatively
influenced by income inequality and population growth. This study relies on comparisons
of fixed-effects and random-effects regression analyses, involving interaction terms both
between these determinants and region dummy variables, and between income inequality
and other factors in order to measure differences in the effects of these factors across
regions on the quality of primary education. An important finding is that income
inequality tends to be detrimental to primary schooling quality in countries with relatively
high per capita income.

1
Chapter 1: Introduction

Throughout history human capital has been proven to be essential in
promoting long-run economic growth and improving living standards given that a
worker’s earnings are positively correlated with the basic reading, writing and arithmetic
skills that she develops early in life. Hence, we have come to realize the significance of
primary education since it is an important factor in influencing one’s future
achievements. Yet, rather than quantity of education, analysis of quality can be regarded
as a higher priority because at a young age, students generally need sufficient amounts of
guidance and instruction in order to develop their intellectual capacities, cognitive
faculties, and problem solving skills. Without the attention and proper training provided
by experienced instructors, children are not likely to learn well regardless of lengthy class
hours or heavy workloads. What is of more value, therefore, is schooling quality –
measured chiefly by the pupil-to-teacher ratio in this study and instructors’ teaching
abilities or internationally comparable test scores in other research – rather than its
quantity, such as enrollment rates and school hours. With our attention on educational
quality, one may ask: what are the major determinants of primary schooling quality? In
that context, we wish to learn the extent to which these determinants may vary over time
and across several regions since different parts of the world have unique political-
economy and demographic backgrounds that affect education heterogeneously.
Empirical evidence of this paper is drawn from an unbalanced panel dataset of
123 countries from 1950 to 2005 in five year increments and analyzed using a relatively

2
general model of educational quality. Primary educational quality, or the pupil-to-teacher
ratio (PTR), is attributable to a number of relevant and quantifiable determinants, such as
GDP per capita, income inequality as reflected by the Gini coefficient, two governance
indicators (Democracy and Polconiii_2002), and population growth. It is conceivable that
schooling quality can benefit from rises in per capita income and the two political
equality variables, and reductions in income inequality and population growth. To
validate this theory, we assess the level of influence these factors have on the quality of
primary education by regressing PTR on these political-economy variables and others.
Data on PTR, GDP per capita, Gini Index, and population growth rate can be
found from the Statistical Yearbooks of United Nations Educational, Scientific and
Cultural Organization (UNESCO) and its website, but there were missing observations,
in which case they were obtained from the World Development Indicators. Data on the
first political equality variable, democracy, can be found in the works of Robert,
Marshall, and Jaggers (2007), whereas those of Polconiii_2002, the second political
equality variable, is available in the work of Henisz (2002). Unfortunately, not all
observations were available for these governance variables as well. With the data,
regression analyses were conducted involving two sets of specifications: the first one can
be distinguished mainly by the use of an instrumental variable, whereas the second set by
interaction terms. All regression models were estimated using fixed and random-effects
specifications followed by Hausman tests to determine which of the two is the better
estimate. Our choice of using PTR as the regressor is based on several studies that have
demonstrated this variable to be an objective and consistent measure of educational

3
quality and perhaps, the best of those available. Hence, we estimate the influence of
economic-political factors and population growth on PTR. Because governance
characteristics, in principle, can also be determined by quality of education, we estimate
in one model the relation between governance and schooling quality using primary export
as an instrumental variable for the political equality variables.
In another set of specifications, we identify possible differences in the effects
of political-economic and population growth factors across regions by using interaction
terms because the effects of these determinants might vary from one country to another or
from one region to another. Important interaction terms include one that has Gini
interacting with income, income interacting with regions, and each of the two political
variables interacting with individual regions. Regions in this study specifically refer to
developing countries in Asia, Latin America, Middle East and North Africa (MENA),
Sub-Saharan Africa, and Transitional Economies.
1
Interaction terms were limited to these
five locations because countries within these regions were found to be much more
homogenous with respect to political-economic conditions and factor endowments than
across regions.
For example, since just about all oil-producing Middle Eastern countries are
heavily dependent on their primary exports and share a similar political-economic
background, the variation across countries can be analyzed by allowing for the use of
regional dummy variables and their interactions with other explanatory variables.

1
The data-set for the regression models also included three groups – Western Europe, US and
Canada, and Japan, Australia and New Zealand – but interaction terms were assigned only to
developing countries in order to compare the degree to which the determinants of primary
educational quality varies within these countries.

4
Lastly, the paper is organized in the following manner: Chapter 2 covers the
literature review and modeling; Chapter 3 identifies the sources of data and provides a
preliminary analysis of descriptive statistics; Chapter 4 reports the empirical findings of
three regression analyses; and Chapter 5 summarizes the study and conclusions.

5
Chapter 2: Literature Review and Modeling

Education has long been identified as a priority for study because of its role in
economic growth and impact on well-being. As a result, numerous scholars have
identified various factors that they believe have determined both the quantity and quality
of education. Even with respect to quality alone, numerous factors have been examined.
These range from democratic governance (Dreze and Sen, 2002), income inequality
(Gradstein and Justman, 1997), educational priority between vocational skills and
cognitive development (Bertocchi and Spagat, 2004), population growth and fertility
rates (Zsigmond, 1976), obstacles to borrowing on credit (Chen, 2005), children
withdrawing from school to work in crowded, urbanized cities of developing countries
(Gilbert, 1994), and policy implementation struggles in areas of high ethnic diversity
(Easterly and Levine, 1997). Some researchers hold that school-specific factors, such as
teacher salary, teacher training, expenditure per pupil, family background, and even
nutrition are important determinants of educational quality (Barro and Lee, 2000). Others
argue that school inputs are what matters for schooling quality and that they play a more
vital role in less developed countries relative to countries with a mature economy
(Palafox, Prawda, and Velez, 1994). Such studies point to the importance of physical
properties (size of classrooms, availability desks and writing materials), the
organizational structure of schools, student discipline, and motivation to learn (Harbison
and Hanushek, 1992; Rutter, 1983).

6
Although the above arguments have been insightful, there may well be other
factors that are more directly responsible for improvements in the quality of primary
education. Motiram and Nugent (2007) present empirical evidence based on data of Latin
America, suggesting that these determinants include variations in income, political and
economic inequality, and civil war intensities. Based on this assumption, this study
attempts to test the applicability of a political-economy model to the world in general
using GDP per capita, the Gini coefficient, democracy, Polcon indices, and population
growth. In doing so, however, we must keep in mind that these determinants may interact
differently across regions because of its unique histories and factor endowments. In other
words, for instance, one society may be affected by GDP per capita more positively than
another society; this hypothesis is tested by using regional interaction terms in the
regression model of Chapter 4.
The purpose of this study, therefore, is not only intended in identifying the
world-wide determinants of educational quality and assessing the significance of them,
but also to compare the extent to which these effects vary. In essence, we examine three
major categories of independent variables (economic conditions, governance
characteristics, and demographic patterns) that cause changes in the quality of education
as measured by the pupil-to-teacher ratio.

2.1 Pupil-to-Teacher Ratio (PTR)
The PTR, an inverse measure of schooling quality, is chosen as the most
suitable indicator of educational quality for this study for two reasons. First, PTR is more

7
relevant to quality than other commonly available measures of schooling, such as
enrollment rate, which gauges quantity to a greater extent than quality. PTR seems more
indicative of quality since smaller class sizes are associated with signs of improvement in
school achievement, including test scores (Glass and Smith, 1979; Glass, Cahen, Smith,
and Filby, 1982).
2
And as discussed earlier, rather than long school hours and ample
amounts of homework, it is conceivable that a solid educational training depends on other
factors, such as the competence of instructors, school facilities, and most importantly
PTR. Some scholars contend that an appropriate class size is necessary for a decent, if not
superior, education. A general prescription quoted in one study is that if the class size
exceeds 25, the instructor must have an assistant and if there are more than 40 pupils,
there must be two teachers (Angrist and Lavy, 1999). Class size is arguably the most
influential factor because a low PTR allows for a more stimulating, productive learning
environment as well as a pleasant classroom experience for both the instructor and
student (Mueller, Chase, and Walden, 1988).
Second, PTR seemed most suitable for practical reasons since data on it are
widely available both across countries and over time. Indeed, we have found PTR to be
available for as many as 123 countries throughout most of the period between 1950 and
2005. In relevant literature, other proxies for schooling quality include internationally
comparable test scores, grade repetition rates, survival to the fifth grade, teacher salaries,
adequacy of school facilities, availability of textbooks, and existence of computers and

2
Angrist and Lavy, suggest that a positive return on test scores with fourth and fifth graders, but not
third graders, were significant, affirming the positive role of PTR. Other scholars have also noted that
reduced class sizes are beneficial to students (Wright, Shapson, Eason, and Fitzgerald, 1977; Finn and
Achilles, 1990).

8
writing materials. Some of these measures, such as comparable test scores used in the
work of Barro and Lee (2002), may be as valuable in measuring educational quality as
PTR. But data on such variables are not adequately available over time for many
countries, resulting in an unsatisfactory longitudinal analysis. Even when they are
available, data may not be drawn from random samples or may come from only small
samples. Survival to the fifth grade or repetition rates – reported in UNESCO’s
yearbooks – are almost complete and also potentially as useful as PTR, but for the
purposes of measuring quality, they are rendered ambiguous since high repetition rates
and low levels of survival to the fifth grade may be the result of rigorous educational
programs and strict standards for advancing to the next grade level. Hence, high
repetition rates in some cases can imply higher quality in education. Data on repetition
rates is also in rather limited supply, restricting coverage across countries and over time.
Therefore, this study relies chiefly on PTR as a measure of schooling quality at the
primary level.
In addition, the Net Enrollment Rate (NER) and Gross Enrollment Rate
(GER) are frequently used as quality measures and indeed are given priority by UNESCO
as part of the Global Millennium Development goals in evaluating educational systems.
But by themselves, they may be misleading because a country with high enrollment rates
can still have substandard schools.
3
Although data on GER is analyzed in Chapter3 and
given that this variable does quantitatively represent a country’s educational system, GER

3
GER and NER are sophisticated measures of school attendance, but do not necessarily convey
schooling quality since GER is the number of children in school divided by the number of school
age children, whereas NER is the number of school age children who are in school divided by the
number of school age children.

9
is often overstated in developing countries because the repetition rates tend to be high and
the numerator takes into account those children who are repeating a grade. With NER, the
overstated enrollment rate problem is treated, but still NER may not necessarily measure
the quality of education because it reflects more on the prevalence of fortunate
circumstances in which children are allowed to attend school. With this in mind, NER
and GER alone, are not recommended for measuring educational quality, although they
can assist in evaluating school systems when combined with other proxies.
For these reasons, we resort to PTR as the most reliable and available measure
of school quality both longitudinally and across countries. However, PTR does have its
limits and less ideal in some respect. For instance, PTR does not measure a student’s
progress, the caliber of an instructor’s teaching skills, or cultural influences that
emphasize the quality of education families desire. Nevertheless, for want of the most
suitable measure, PTR is used in this study.
Now we turn to the explanatory variables. Based on the literature review
discussed above, we include GDP per capita and Gini index as measures of relevant
economic conditions, democracy or political constraints indices as measures of
governance characteristics, and population growth as a measure of demographic
condition. The importance of these explanatory variables is discussed below in length,
one by one, along with relevant literature.

10
2.2 GDP Per Capita (GDPpc)
GDP per capita, which we treat as the average annual income per capita of the
general population, is made comparable by valuing it in US dollars of purchasing power
parity prices. We use this variable in its natural log form for regression analyses in
Chapter 4 under the assumption that income is the driving influence behind primary
education. However, past studies on human capital have commonly emphasized the
reverse case of per capita GDP influencing education. Lockheed et al. argue that
education leads to higher returns in one’s future earnings since classroom training
enhances one’s cognitive abilities, problem solving skills, and competence in numeracy
and literacy.
4
In another study, it was noted that men who attended schools in the United
States with higher ratings in instructional quality – as judged by PTR, average term
length, and teacher salary – tended to achieve higher earnings in the labor market (Card
and Krueger, 1992).
5

These studies relate current educational quality to future income levels. The
focus in this study, however, is on how immediate income levels affect current primary
schooling since parents’ preferences about their children’s education largely depend on
their income. A rise in quality of schooling can be expected as parents earn higher wages,

4
At the aggregate level, test scores of general intellectual achievement – including ACT, SAT,
Weschler Adult Intelligence Scale Revised (WAIS-R) – have been shown to have some
explanatory power in the decline of the US GNP growth rate as one estimate in a study indicates
that GNP could have been $86 billion higher for a single year in 1967 if test scores and
labor quality did not drop (Bishop, 1989).

5
Paradoxically, students who were enrolled in smaller class sizes did not perform better on
standardized test scores (Card and Krueger, 1992), but in a later study in 1999, Krueger examines
a sample of 11,600 students to find that those in a low PTR class had the effect of improved test
scores, raised by four percentile points during the first year of attendance.

11
but in some cases, the quantity of schooling (measured by enrollment rates) can increase
as well, leading to crowding effects that damage educational quality. However, with
higher incomes, parents would generally demand better quality and send children to
private schools or move into respectable school districts. In such cases, public and private
providers of education will react by expanding schools and appointing more teachers,
thereby lowering the overall PTR.
6

A study examines, after controlling for sensitivity checks of neighborhood
quality, how much more parents in Massachusetts were willing to pay to move into
school districts with superior schooling quality: for a five percent increase in test scores
from the mean, these parents were willing to spend 2.1 percent or $3948 more for a home
in order to send their children to a specific elementary school (Black, 1999). This
illustrates how much people value better quality education, but of course, only if they can
afford it.
However, there are people who have little regard for education because of
their income. Consider the Dobe Ju/’hoansi in the Kalahari Desert of Namibia where
anthropologist Richard Lee conducted field research. Lee reported in the early 1970s, that
of the 60 Ju children who were of school age, not a single child enrolled in school at the
primary level as parents complained that the obligatory school outfit and annual school
fees of $4.50 were too high for them.
7
The Dobe people could not afford schooling and

6
Data covered in a later section indicate that governments respond to the rising enrollment rates
by appointing more teachers, lowering the PTR (see the discussion on Correlation Matrix of
Section 3).

7
Another distinction is that education and associated skills are valued differently depending on
where one is located. In Africa, education fosters vocational oriented needs as the relationship
between schooling and future earnings potential revealed that computational skills as opposed to

12
for this reason, quality of education was of no concern for them. For non-Dobe people in
Namibia, however, average PTR has been 32 with per capita income of $1,841 between
1950 and 2005, suggesting that schooling quality and income levels were quite high
compared to Sub-Saharan Africa where PTR ranges between 28 and 78, and per capita
income ranges between $122 and $3,797. Despite Namibian PTR and income being
strong relative to its neighboring provinces, the Dobe Ju/’hoansi children at one point
were deprived of education because of low income even when schools, higher quality
ones, were available. By contrast, in places such as Massachusetts, rich parents go out of
their ways to send their children to a better elementary school. Therefore, the role of per
capita GDP seems to be significant in influencing how one is exposed to superior
educational quality.

2.3 Gini Index
The Gini coefficient is a commonly used and quite readily available measure
of income inequality. In communities and countries with greater income inequality, for
any given level of per capita income, we can expect a bigger gap in the quality of
education received between the poor and rich because income level can determine which
school a child attends. In such cases, shifts occur between public and private provision of
education depending on the Gini coefficient. Gradstein and Justman’s work reveals that
when income inequality is severe, parents generally favor free provision of education

cognitive abilities were more valuable (Moll, 1996). The Dobe people in Namibia make little
money by selling home-made crafts to tourists and rely on hunting and gathering for food;
therefore, computational skills and writing skills at the expense of obligatory school fees –
however small they may be – are virtually useless.

13
through government intervention and subsidization. In the opposite scenario of little
income inequality in which the median voter is well off, people tend to rely less on the
government and prefer private provisions of education; the median voter will most likely
be able to send their children to private schools if public education lacks greatly in
quality.
Additionally, income inequality creates social stratification, leaving the
community with the arduous task of obtaining agreement on how much to spend on
public education. Since wealthy, powerful elites tend to share different interests with
people in lower classes, not so many rich people will be motivated to fund public
education on behalf of the poor (Bourguignon and Verdier, 2000). Even when proper
funding is available, much of it is allocated to private schools or higher education instead
of its public counterparts (Plank, 1990). As a result, countries with high income
inequality often have weaker educational systems with respect to both quality and
quantity. Robinson (2004) points out that in less developed countries (LDCs) and
Transitional economies where income inequality is high, PTR also tends to be high and
NER low, again confirming the anticipated negative relationship between schooling
quality and the Gini index.
For illustrative purposes, consider two different countries with a substantial
gap in income inequality, such as Austria in relation to Central African Republic or Sub-
Saharan African countries. Measures of education and the Gini index highlight a trend
that PTR and NER are weak when income inequality is high. A few noteworthy figures
from Appendix D (on pages 59-63) denote:

14
Figure 1 – Noteworthy Figures from Appendix D
PTR NER% Gini Index
Austria 18 96 24
Central African
Republic
75 53 58
Chad 70 41 40
Mozambique 78 50 40

The Gini Coefficient of Austria is almost half of that in Central African Republic, Chad,
or Mozambique. As hypothesized, a high PTR and low NER is evident when income
inequality is high. Austria’s PTR (about a fourth) and NER (almost double) reflect a
strong educational system compared to the figures of Sub-Saharan African countries.
This pattern is obvious throughout Appendix D: where income inequality is severe,
children often experience inferior quality of and little access to education.

2.4 Inequality and Governance Factors (Democ and Polcon)
Governance characteristics are measured by two political constraints indices,
Polcon from Polcon III 2002, and democracy index (Democ). The Polcon variable reports
the feasibility of policy changes by taking into account the number of government
branches that can exercise checks and balances, and veto power over each other. Democ
incorporates the characteristics of governing authorities and regimes, ranging from
autocracies to fully developed democratic institutions. Both variables numerically
measure political inequalities, which have pervasive effects on public services, often
sullying the quality of education and consequently, hindering children from receiving

15
decent primary education.
8
Hence, governance characteristics play a significant role in
impacting schooling quality. Even political regime types can produce different outcomes
in the quality of education as democratic nations tend to have a higher demand on
premium education because of increased public goods expenditure that directly shares a
positive correlation with electoral competitiveness (Brown and Hunter, 1999). This is
because in the absence of democracy, a powerful minority can control the provision of
public goods, including primary education. If so, as mentioned earlier through the works
of Bourguignon and Verdier (2000), this small circle of elites may have little incentive to
provide high quality public goods for those in the general public. Hence, they will not
want members of the lower echelon to have a voice in policy making decisions – a
process that more than often yields low school standards.
It is quite obvious as one might assume that deficiencies in governance
characteristics are sufficient causes for producing low quality public goods. Indeed,
studies have shown that political inequalities often result in the reduction of public goods
(Acemoglu and Robinson, 2000). By contrast in a democracy, concerns of the median
voter will be crucial and their needs, such as subsidizing high quality public education at
the primary level, will not be overlooked. Therefore, the level of democracy, political
equality, and even regime type can be major determinants of changing educational
quality.

8
In the case of Latin American countries, excessive accumulation of external debt forced its
governments to reduce domestic spending in order to repay their debt (Haggard and Kaufman,
1995) and one obvious consequence of lowered government expenditures is a decline in public
educational spending.

16
2.5 Population Growth (PopGr)
The balance between student enrollment rates and school quality as reflected
by school inputs – number of instructors and classrooms, and physical properties – can
directly be affected by population growth. That is, the shortcomings of primary
educational quality can be explained in part by PopGr as one study reveals that a rapid
growth in population overwhelmingly adds responsibility to the government, forcing
public services to become substandard (Ladd, 1992). The quality of education is at risk
since, for a given income level and governance characteristics, a sudden growth in
population may swamp the supply of schools and teachers. If the student enrollment rate
exceeds the growth in faculty and facilities, then the quality of education will suffer
because crowding of classrooms reduce the amount of attention that each student would
have otherwise received.
When dealing with issues of the population and educational quality, the
emphasis is on the growth rate of the population rather than the size. Although it was
noted in a study of China that the resources, including education, housing, and
agricultural output, were not enough to support its massive population (Shen, 1998),
China’s population growth rate has been maintained at a steady level. Indeed, primary
schooling quality is one of the best in Asia perhaps because of China’s “one child per
household,” which goes a long way in explaining its relatively low PTR of 29. Asia’s
PTR ranges from 26 to 51. Their low PTR, perhaps, can be attributed to a well
maintained population growth rate of 1.51, which falls in the mid-range of 0.82 to 2.67
for Asian countries. Therefore, the size of population itself is not of direct interest, but

17
what is of more importance is the rate of population growth and how steady it is
maintained.
This hypothesis is revisited and tested at a later point in Chapter 4 where the
impact of population growth can be measured on PTR directly by regressing PTR on
population growth rate along with other explanatory variables. Before diving into
regression models, descriptive statistics are covered in the next section.

18
Chapter 3: Descriptive Statistics and Preliminary Analysis

3.1 Data Information
The data for PTR, GER, and NER can be found from UNESCO Statistical
Yearbooks and UNESCO website, but missing data were obtained from the World
Development Indicators (WDI). Data for GDP per capita, Gini Index, population growth
rate, and primary export are from WDI. The two measures of governance characteristics,
political constraints (Polcon) and democracy indices (Democ), were respectively obtained
from the works of Henisz, The Institutional Environment for Infrastructure Investment
(2002), and the democracy index in the Polity IV Project: Political Regime
Characteristics and Transitions, 1800-2007 by Robert, Marshall, and Jaggers (2007).

3.2 Regional Statistics by Variables
For the purposes of this study, OECD countries include developed countries
(DCs), namely the US, Canada, Japan, Australia, and New Zealand and rich Western
European countries in Appendix D, such as France, Germany, UK, etc.; non-OECD
countries, also referred to as the five regions, are developing countries in Asia, Latin
America, Middle East and North Africa (MENA), Sub-Saharan Africa, and Transitional
Economies. For a listing of the 123 OECD and non-OECD countries and each country’s
statistical figures, please refer to Appendix D. Table 1, on the following page, is a
snapshot of the 123 countries divided into eight groups or regions: consisting of 99

19

Table 1 – Averages of Raw Data (top) and Rate of Change (bottom as %) of
Primary Educational, Economic, Political, and Demographic Measures Across
the Globe. 1950-2005 at five year intervals

Regions &
OECD
PTR GER NER GDPpc Gini Democ Polcon PopGr PrimExport
Asia
34 83% 82% $1,689 0.39 4.17 0.23 2.05% 50%
-2% 13% 5% 17% -0.15% 4% 7% -2% -12%
Latin America
32 91% 85% $2,669 0.5 4.37 0.26 2.17% 74%
-2% 11% 3% 7% 1% 22% 3% -3% -3%
MENA
36 72% 78% $1,380 0.40 1.23 0.11 2.53% 68%
-1% 18% 10% 11% -1% 15% -3% -2% -4%
Sub-Saharan
Africa
45 65% 60% $693 0.48 1.19 0.12 2.48% 73%
2% 28% 12% 5% 0.04% -9% 9% 45% -1%
Transitional
Economies
17 96% 91% $3,108 0.30 4.32 0.25 0.73% 44%
-8% 5% 1% 12% 7% 2% 2% -22% -2%
Japan,
Australia,
& New
Zealand
25 99% 100% $17,131 0.36 10 0.45 1.24% 53%
-4% 4% 0.11% 14% 4% 0% 1% -18% -6%
US & Canada
22 99% 95% $21,256 0.35 10 0.41 1.26% 36%
-3% 2% 0.44% 11% 3% 0% 2% -6% -8%
Western
Europe
21 95% 60% $15,106 0.32 9.04 0.44 0.71% 29%
-6% 6% 1% 15% 1% 0% 4% 30% -5%
MINIMUM
17 65% 26% $693 0.30 1.19 0.11 0.71% 29%
-8% 2% 0.11% 5% -1% -9% -3% -22% -12%
MAXIMUM
45 99% 100% $21,256 0.50 10 0.45 2.53% 74%
2% 28% 12% 17% 7% 22% 9% 45% -1%
MEANS
5 Regions 33 81 79 $1,909 41 3.11 0.19 1.99% 62
OECD 23 98 97 $17,831 34 10 0.43 1.07% 39
Differential 10 17 18 $15,922 7 6.57 0.24 0.92% 22
MEAN GROWTH RATES
5 Regions -2% 15% 6% 10% 1% 7% 4% 3% -4%
OECD -4% 4% 1% 13% 3% 0% 2% 1% -6%
Differential 2% 11% 5% 3% 2% 7% 2% 2% 2%

Data sources: PTR, GER, and NER from UNESCO; Democ by Roberts, Marshall and Jaggers; Polcon by
Henisz; and GDP per capita, Gini, Population Growth Rate, and Primary Export from World Development
Indicators. Note that some observations are missing.

20
countries that represent the five non-OECD regions and 24 countries categorized into
three OECD groups. Towards the bottom of Table 1 are the averages of the five regions
of developing countries taken together and those of OECD countries. Below that are
average growth rates over 55 years in five year increments for the two regional groups
and for each variable in the analysis.
According to Table 1, the measures for primary education quality, economic
conditions, political inequality and demographic patterns are all significantly different
between OECD countries and non-OECD regions – Asia, Latin America, MENA, Sub-
Saharan Africa, and Transitional economies. Not all, but some of the growth rates of
countries from the five regions are unfavorable in comparison to their OECD
counterparts. For instance, the average growth rate of PTR is -2% for the five regions and
-4% for OECD members, implying that class sizes in DCs have been diminishing
approximately twice as quickly as in LDCs between 1950 and 2005. Because reduced
class sizes are suggestive of improved educational quality as teachers can be more
engaging in smaller classes, it’s unfortunate but not surprising that LDCs’ PTR has been
higher during the 55 years and also progressing at a slower rate than OECD countries.

3.2 (a) Education Data
Over the period of 55 years, OECD countries on average have had 10 fewer
pupils for every instructor at the primary level than non-OECD countries, suggesting that
classroom sizes have been almost 50 percent larger in the five regions of developing
countries. This shows that non-OECD regions generally have quality of primary

21
education that is remarkably inferior to that of OECD countries. However, it would be
unfair to assume that schooling quality of all non-OECD countries has always been
poorer. For instance, Transitional economies have had low PTR, ranging from 17 to 25.
The maximum average value of PTR has been 45, which represents Sub-
Saharan Africa’s PTR. Turning to Appendix D, the lowest reported PTR in the Sub-
Saharan African region of 28 belongs to Mauritius and the highest of 78 to Mozambique
between 1950 and 2005. While 28 can be considered a good class size, 78 reflects a
considerable weakness and in fact, is the largest among the other 122 countries in this
study; Central African Republic and Chad also have abnormally high PTRs.
Similarly, the differentials in the Gross Enrollment Rates (GER) have
averaged 17% and Net Enrollment Rates (NER) 18%, illuminating the fact that the five
non-OECD regions have not kept up with OECD countries. MENA and Sub-Saharan
Africa have the lowest GERs of 72% and 65%, and NERs of 78% and 60%, respectively,
and the highest PTR of 36 and 45, respectively. These figures imply a deficiency in
primary education in MENA and especially Sub-Saharan Africa. A noticeable pattern, as
seen in Table 1 and Appendix D, is that countries with a high PTR generally have low
enrollment rates, such as Mozambique, Central African Republic, and Chad, whereas
countries with a low PTR tend to have high enrollment rates, exemplified by Mauritius,
Estonia, New Zealand, and France. Primary schools in non-OECD countries have
suffered low levels of enrollment rates and high levels of PTR, but the growth rates of
GER (15%) and NER (6% ) of the five developing regions have been rising at a faster

22
pace than those of OECD members at 4% and 1%, respectively. Nevertheless, non-
OECD countries are struggling to have primary education readily available to the public.

3.2 (b) Living Standard Data
Economic conditions, measured by GDP per capita and the Gini coefficient,
are also better in OECD countries. For instance, the Gini index of the five regions is
higher by seven points. In particular, income inequality is severe in Latin America (0.50)
and Sub-Saharan Africa (0.48), whereas MENA, Asia and Transitional economies do not
have such high disparities in income inequality during the 1950-2005 period. However,
the average annual per capita GDP was relatively low, with reported figures of $693 in
Sub-Saharan Africa and $1,380 in MENA. Collectively, the average per capita GDP of
the five non-OECD regions was $1,909, whereas OECD countries was $17,831, showing
that the standard of living is dramatically lower in LDCs. Within each of the five regions,
the country with the lowest annual per capita GDP according to Appendix D are Nepal in
Asia ($168), Guyana in Latin America ($786), Yemen Republic in MENA ($511),
Burundi and Ethiopia in Sub-Saharan Africa (both $122), and Tajikistan representing
Transitional economies ($314). With the exception of Tajikistan, the other five countries
have one of the worst education as reflected by an average PTR of 40 and NER of 60%.
Returning to Table 1, the five year growth rates in per capita GDP seem to
have been some 3 % higher in OECD countries, but from a growth convergence
perspective, one may have presumed that LDCs could have had higher rates since they
had more “catching up” to do relative to the mature economies of OECD countries. Many

23
regions seemed to have suffered from low standards of living, although some have
prospered. GDP per capita in reported Asian countries, for instance, has been rising on
average by 17%, which is 2% to 6% higher than that of OECD members. Yet, low
income growth rates of Latin America (7%) and Sub-Saharan Africa (5%) are well below
those of OECD countries as reported in Table 1.

3.2 (c) Political Equality Data
Democ and Polconiii_2002, as discussed earlier, are the two measures of
governance. Both variables are negatively related to PTR, meaning poor governance
impairs the quality of primary education.
Democ has a scoring system based on an ordinal scale of 0 to 10 that ranges
from autocracies to fully developed democratic institutions with zero indicating the
lowest practice of democratic principles and 10 for a perfect attainment of democratic
governance. For example, compare Belgium with Gabon in Appendix D. Belgium has a
perfect Democ index of 10 between 1950 and 2005, whereas Gabon has a reported figure
of 0 during that period. Indeed, the level of correlation between PTR and governance is
strengthened as the average PTR in Belgium is 18, while it is 52 in Gabon. Polcon
iii_2002 scores are based on the probability of policy changes that depend on the number
of government branches entitled to vetoing others and performing checks and balances.
The scale ranges from 0 to 1 with zero being the lowest, indicating a complete lack of
veto power, and checks and balances. While a value of 1 is theoretically attainable, no

24
country has a reported score of 1.
9
Again, comparing Belgium with Gabon in Appendix D
for the period between 1950 and 2005, Polcon values were 0.59 for Belgium and 0 for
Gabon where there is a president with virtually absolute power.
Data on governance characteristics reflect that the five non-OECD regions
have experienced severe political inequality with Democ scores averaging 3.11, whereas
many OECD countries have Democ indices of a perfect 10, indicating an effective,
reliable government. In particular, MENA and Sub-Saharan Africa have the lowest
Democ indices of 1.23 and 1.19 and Polcon values of 0.11 and 0.12, both respectively.
These figures suggest the presence of major political inequality in these regions even
relative to other non-OECD regions. Democ of 7% and Polcon of 4% are the rates of
improvement between 1950 and 2005 in the five regions, showing positive changes, but
scores on governance are still low relative to those of developed countries.

3.2 (d) Population Growth Data
With respect to demography, rich OECD countries have well maintained their
population growth rate at about 1.07% for the past 55 years, whereas the five non-OECD
regions have struggled as shown by an average figure of 1.99%. Excluding Transitional
economies, the population of the remaining four regions has been growing somewhere
between 2.53% and 2.05%, which are almost double those of OECD countries. This

9
Polconiii_2002 observations are between 1800 and 2001. The formula for deriving
Polconiii_2002 is as follows:
]
1 - N
N
n
1) -
n
(
[ - 1
i
i n
=1 i
Σ

where n = the number of parties, n
i
= seats held by nth party and N = total seats. For further details
on the computational methodology, refer to the codebook by Heinz.

25
perhaps explains in part why PTR is so high in the five regions. For instance, Pakistan
found in Appendix D is a classic example of a country with an unusually high PopGr of
2.67% and low GER of 45% and NER of 43% – the worst in Asia.
Returning to Table 1 for the “average growth rate of PopGR,” which measures
the average change in the population growth rate during 1950-2005 in five year
increments, one can infer that non-OECD countries have experienced rapid fluctuations
in the size of population. Crowded classrooms in most areas of the five regions are
common as the population growth rate is three times that of OECD members (3% and
1%, respectively). In such cases, it is evident that population growth can have a
devastating effect on the quality of education.

3.2 (e) Primary Export Data
Primary Export (PrimExport) – exportation of food and other agricultural
goods, unprocessed textiles, oil, logging, and mining as a percent of total exports –
although not of direct interest for measuring educational quality, is examined for the
purposes of the regression model discussed in the next chapter. Regions with the lowest
Democ and Polcon measures generally have the heaviest reliance on exporting their
abundant primary resources. For instance, MENA and Sub-Saharan Africa have low
Democ estimates of 1.23 and 1.19, and high primary export shares of 68% and 73%,
respectively; Western Europe, on the other hand, has Democ of 9.04 and primary export
of 29%, suggesting that poor governance is positively correlated with primary export,
which tends to hinder the quality of primary education.

26
3.2 (f) Summary
The patterns found in the descriptive statistics indicate that primary exports,
poor economic, political, and demographic conditions imply noticeable deficiencies in
the quality and quantity of primary education. The lower the income level or higher the
Gini coefficient, the higher is the PTR and the lower are GER and NER. Weak Democ
and Polcon indices are consistent with higher levels of PTR, and lower values of GER
and NER. In contrast, as the population growth rate inflates, PTR tends to increase, and
GER/NER will likely contract. This correspondence is noticeable in the correlation
matrix.

3.3 Correlation Matrix
Table 2 on the following page reports the correlation matrix of the five
regions and OECD countries (in Italics). This table leaves us with a salient point that as
economic and political factors improve and the population growth rate decreases, school
enrollment rates increase and most important of all, PTR diminishes. Note that Appendix
A, discussed in Chapter 4 with the regression outputs, also displays a correlation matrix,
but reports the combined correlates of all the 123 countries instead of separating OECD
countries and the five regions as in Table 2.

3.3 (a) Living Standard and Education
In Table 2, PTR and the natural log of GDP per capita, which have a
correlation of -0.5336 and -0.3311 for the five regions and OECD countries, respectively,

27
suggest that income has a much stronger association with the quality of education in the
five regions than in OECD countries. Similarly, income inequality has a stronger
correlation with PTR in LDCs. In essence, PTR is higher – indicating that the quality of
education is lower – when GINI is higher or GDPpc is lower.
Table 2 – Correlation Matrix of the Five Regions and OECD Countries between
1950 and 2005 in five year intervals. OECD Countries are in Italics

PTR GER NER
ln
GDPpc
Gini Democ Polcon
PopGr
Lagged
Prim
Export
PTR
1
1
GER
-0.3358 1
-0.4266 1
NER
-0.5316 0.871 1
-0.1957 0.5405 1
lnGDPpc
-0.5336 0.5298 0.6621 1
-0.3311 0.1167 0.2466 1
Gini
0.3002 0.0388 -0.1451 0.0081 1
0.2797 0.1327 0.1623 -0.21 1
Democ
-0.2759 0.3784 0.3805 0.4398 0.0677 1
-0.366 0.04 0.306 0.2723 -0.0819 1
Polcon
-0.2241 0.294 0.2636 0.3398 0.1069 0.7015 1
-0.4263 -0.0276 0.1695 0.2871 -0.1564 0.6544 1
Pop Gr
Lagged
0.1306 -0.0741 -0.162 -0.1445 0.1662 -0.1344 -0.1069 1
0.0249 -0.022 -0.0061 0.0766 -0.0029 0.0041 0.0455 1
PrimExport
0.3104 -0.3505 -0.3587 -0.3261 0.1978 -0.3047 -0.3374 0.2029 1
0.3542 -0.1423 -0.0143 -0.4348 0.214 -0.0526 -0.2352 -0.0402 1

Data sources: PTR, GER, and NER from UNESCO; Democ by Roberts, Marshall and Jaggers; Polcon by
Henisz; and GDP per capita, Gini, Population Growth Rate, and Primary Export from World Development
Indicators. Note that some observations are missing.

Observing the cross tabulation of NER (0.6621) and GER (0.5298) with
natural log of GDPpc for LDCs, each of its strong, positive association demonstrates that
when income is higher, NER and GER are correspondingly larger. This suggests that
parents can better afford to send their children to school when income levels are higher,

28
especially in LDCs where primary education is not always mandatory. As mentioned in
Chapter 2, an underlying assumption is that more teachers are appointed as the
enrollment rates increase, ultimately decreasing PTR. We see that the necessary condition
to support the assumption is met when GER and NER rises with income. Further, NER
and GER both have negative correlations with PTR of -0.5316 and -0.3358, respectively,
indicating that the size of the class reduces as more students enroll for school. This
phenomenon is prevalent in OECD countries and we see very similar outcomes through
other variables, such as Democ, Polcon, and PopGrLagged.

3.3 (b) Political Equality and Education
Polcon and Democ both have negative correlations with PTR and positive
ones with GER and NER as expected for LDCs. In comparison to the -0.4263 correlate of
OECD countries for PTR and Polcon, the five regions have a weaker one of -0.2241. We
can see this weaker association between PTR and Democ for the five regions and a
relatively stronger relationship for OECD countries, leading to a generalization that when
political inequality is alleviated, schooling quality improves more noticeably in OECD
countries.

3.3 (c) Population Growth and Education
The population growth rate is lagged one time period (t-1) of five years
because it generally takes time for the change in population growth to affect schooling
quality. The correlation between PTR and PopGrLagged is 0.1306 for the five regions,

29
which is very weak, and even lower at 0.0249 for OECD countries. Although these
variables are barely correlated, it is still apparent that in parts of the world where the
population growth is high, PTR is lower. In fact, this variable turned out to be significant
in Chapter 4. Even in Table 1, we can see that population growth increases more rapidly
in the five regions at 2.15% as opposed to 0.93% in OECD countries.

3.3 (d) Primary Export and Education
Lastly, the correlation between primary export (PrimExport) and PTR in the
five regions confirms the anticipated relationship that heavy reliance on exportation of
primary goods is positively associated with PTR at 0.3104 for the five regions and 0.3542
for OCED countries. The correlation between GER (or NER) and PrimExport reveals that
when there is heavy reliance on primary exportation, enrollment rates tend to drop. For
instance, NER and PrimExport have correlations of -0.0143 for OECD countries and
-0.3587 for the five regions, highlighting a remarkable disparity between DCs and LDCs.
The gap in degree of association, however moderate the actual correlations are, is
supportive of political-economy arguments and the existence of “banana republics” –
United Emirates, Nicaragua, Honduras, Iran, etc – discussed in the next chapter.

3.3 (e) Summary
Based on the correlation matrix, an obvious pattern emerges: the higher the
Gini coefficient and population growth rate, the higher the PTR; and the higher the
income, Democ, Polcon, GER and NER, the lower the PTR, indicating a smaller class

30
size. We also find that favorable economic conditions, governance characteristics, and
population growth rates have positive correlations with GER and NER, both of which are
negatively associated with PTR.
Therefore, the correlation matrix lends support for the patterns found in Table
1: the higher the Gini coefficient and population growth rate, the higher the PTR, which
is equivalent to the scenario of overcrowded classrooms and poor delivery of high
educational quality. When income levels, Democ, Polcon, GER and NER rise, PTR tends
to drop, allowing teachers to have more interaction with students and thus, resulting in
improved instructional quality. Furthermore, favorable economic conditions and
governance characteristics, and lower population growth rates are positively associated
with GER and NER, which are inversely related to PTR. That is, as positive
circumstances raise school attendance, more instructors are appointed to meet the
demand, ultimately reducing the pupil-to-teacher ratio.

31
Chapter 4: Regressions and Empirical Findings

Schooling quality is estimated using regression models based on an
unbalanced panel data-set that includes all of the five regions and all OECD countries.
The regression models are intended to test the hypothesis that beneficial changes in
economic-political and demographic factors promote primary schooling quality. The first
and foremost assumption is to confirm whether PTR has an inverse relationship with
income, Democ, and Polcon, and whether it shares a positive correlation with the Gini
coefficient and population growth. Once the validity of this theory is ascertained, we
advance to establish that non-OECD countries might be affected by these determinants in
varying degrees since none of the five regions have identical factor endowments, and
political-economic and demographic characteristics.
Both sets of hypotheses are examined with different estimation techniques.
We use fixed-effects (fe) and random-effects (re) specifications and Hausman tests to
identify the appropriate method as these two approaches have clear advantages and
disadvantages. Hence, regression tables 3 through 5, report the results of fe and re
estimates along with Hausman tests of the various parameters of the model. For
specifications using primary exports as an instrumental variable, an assumption made is
that primary export is a valid one. The legitimacy of this assumption is then tested and is
substantiated through F-tests.
Lastly, it is critical to be aware that PTR has its limitations because it does not
measure every dimension of an educational system, including instructor’s teaching skills,

32
cultural differences, and student aptitude; however, PTR had practical advantages of
consistently and objectively measuring schooling quality across a large number of
countries and over time.

4.1 Educational Quality Function
The educational quality function takes the following form:
Qjt = αjt +β1*Economic jt + β2*Political jt + β3*Population j t-1 + εjt + ujt
with Qjt denoting quality of education, or PTR, for country j representing the five regions,
time t between 1950 and 2005 in five year intervals, and the unobserved effects, εjt; the
three determinants are Economic conditions (per capita GDP in natural log form or the
Gini coefficient), Political inequality estimates (either Democ or Polcon used in the
model), and a Demographic characteristic (population growth rate lagged five years).
Variations of the educational function are categorized into two basic specifications that
include an instrumental variable in one model (Table 3) and interaction terms in another
(Tables 4 and 5).

4.2 Instrumental Variable
Table 3, on the following two pages, shows the results of both instrumented
and non- instrumented equations. Primary exports – a percent of total exports – is used as
an instrument to treat the endogeneity problem of two explanatory variables, Democ and
Polcon. Columns B, D, F, and H are the instrumented versions, each of which represents
two-stage least squares estimators, using Primary export as an instrumental variable for a

33

Table 3 – Regression Outputs of Fixed and Random Effects.
PTR as dependent variable

A. Not instrumented
B. Instrument
Democ=PrimExp
C. Not instrumented
D. Instrument
Democ=PrimExp

Fixed Random Fixed Random Fixed Random Fixed Random
effects Effects Effects Effects Effects effects effects effects
ln GDPpc
-7.28*** -6.19*** -7.52*** -4.26*** -7.12*** -5.82*** -9.03*** -5.20***
(0.65) (0.46) (0.89) (0.97) (0.78) (0.49) (1.35) (1.1)
Gini
0.02 0.13** 0.01 0.15**
(0.06) (0.05) (0.08) (0.06)
Democ
-0.32*** -0.30*** -1.23*** -1.83*** -0.18* -0.17* -0.26 -0.8
(0.08) (0.09) (0.46) (0.56) (0.11) (0.10) (0.53) (0.56)
Polcon

PopGrLagged
0.16** 0.19*** 0.21 0.22 0.32*** 0.34*** 0.27** 0.27**
(0.08) (0.07) (0.14) (0.14) (0.11) (0.10) (0.12) (0.11)
Constant
86.11*** 76.62*** 92.42*** 69.49*** 82.99*** 67.68*** 98.61*** 65.09***
(4.75) (3.42) (5.8) (5.24) (6.63) (4.5) (9.91) (7.28)
N 896 896 766 766 616 616 501 501
R
2
Within 0.17 0.17 0.11 0.09 0.17 0.16 0.2 0.1
R
2
between 0.44 0.45 0.42 0.39 0. 4684 0.5 0.44 0.47
R
2
overall 0.43 0.43 0.43 0.39 0.46 0.49 0.5 0.51
Hausman
chi
2
(3) 7.16 chi
2
(3) 6.99 chi
2
(4) 15.21 chi
2
(4) 31.63 Fixed
Random
Test Ho:
Prob>chi
2
0.0668 Prob>chi
2
0.0723 Prob>chi
2
0.0043 Prob>chi
2
0.00
difference in
coefficients
not
systematic

Significance levels of 1%, 5%, and 10% are indicated by ***, ** and * respectively.
Standard Errors are in parentheses.

34

Table 3: continued

E. Not instrumented
F. Instrument
Polcon=PrimExp
G. Not instrumented
H. Instrument
Polcon=PrimExp

Fixed
Random Fixed Random Fixed Random Fixed Random
effects Effects Effects Effects Effects effects effects effects
ln GDPpc
-7.51*** -6.45*** -7.41*** -5.36*** -7.23*** -5.95*** -6.68*** -4.81***
(0.63) (0.45) (0.89) (0.64) (0.78) (0.48) (1.16) (0.72)
Gini
0.01 0.12** 0.11 0.22***
(0.06) (0.05) (0.07) (5.58)
Democ

Polcon
-2.73* -2.85** -13.58*** -18.45*** -1.00 -1.6 -19.30*** -19.96***
(1.4) (1.37) (4.78) (5.03) (1.58) (1.53) (6.18) (5.92)
PopGrLagged
0.17** 0.20*** 0.26* 0.29** 0.36*** 0.35*** 0.24** 0.25**
(0.08) (0.07) (0.13) (0.64) (0.11) (0.10) (0.12) (0.11)
Constant
87.04*** 78.03*** 89.20*** 73.78*** 83.27*** 68.36*** 80.59*** 61.42***
(4.65) (3.35) (5.99) (4.11) (6.62) (4.45) (8.92) (5.58)
N 907 907 775 775 616 616 570 570
R
2
Within 0.17 0.17 0.14 0.12 0.17 0.17 0.05 0.13
R
2
between 0.44 0.45 0.44 0.43 0.47 0.51 0.43 0.45
R
2
overall 0.43 0.43 0.45 0.43 0.46 0.49 0.47 0.47
Hausman
chi
2
(3) 6.63 chi
2
(3) 7.25 chi
2
(4) 15.09 chi
2
(4) 14.76 Fixed
Random
Test Ho:
Prob>chi
2
0.0847 Prob>chi
2
0.0267 Prob>chi
2
0.0045 Prob>chi
2
0.0052
difference in
coefficients
not
systematic

Significance levels of 1%, 5%, and 10% are indicated by ***, ** and * respectively.
Standard Errors are in parentheses.

35
number of reasons. From a statistical standpoint, the F-tests of the residuals for
democracy and polconiii_2002 justify the use of primary export as an instrument and it is
strongly correlated with Democ (-0.4284) and Polcon (-0.4252) as shown in Appendix
A.
10
Secondly, as mentioned in the preceding chapter, banana republics can often be
identified by the level of primary exports rooted in political-economy arguments.
A frequently advanced position is that many LDCs have not made significant
progress in transitioning from an agrarian society to an industrial one because these
countries are ruled by a small circle of land owners and corrupted elites – symptoms of
absence in democracy. Finally, since Democ/Polcon and education reinforce one another
and factor endowments do not have a direct influence over education, utilizing
PrimExport as a proxy seemed to be a suitable choice.

4.2 (a) Table 3 Results
The results as shown in Table 3 above indicate that the instrumental variable
technique has the effect of enlarging the coefficients of the two political variables and
producing a lower p-value for Polcon. In comparison to those of the non-instrumented
model, instrumenting Polcon with PrimExport strengthened the results of Gini and
Polcon for both fe and re as indicated by their significance levels and coefficient size,

10
The p-values of F-test are 0.0764 for democracy and 0.0941 for Polcon, both of which are
significant and validate the instrumentation of PrimExport. The overidentification test involved
two equations under the OLS approach. First, income, population growth and primary export were
regressed on democracy. Then the predicted values of democracy, its residual, income and
population growth were regressed in the education function with PTR as the dependent variable
(outputs are reported in Appendix B). Finally, an F-test of the democracy residual yielded a p-
value of 0.0764 for democracy. The same procedure was used for Polcon with the F-test indicating
a p-value of 0.0941. The residuals in the education function are modest and democracy is captured
by the major impact on PTR.

36
albeit the loss in degrees of freedom. Nevertheless, both estimation methods, whether
using an instrument or not, produced results that supported the initial theory that PTR is
directly affected by political-economic and demographic factors.
The results are robust, confirming the presumed assumption that schooling
quality benefits from higher per capita income, alleviations in income and political
inequality, and reductions in population growth. We review each variable, starting with
per capita income.
The coefficients of log GDP per capita are always highly significant, whereas
the regression results of governance measures, Gini, and population growth are somewhat
sensitive and depend on the estimation method. For instance, we find that PopGrLagged
is generally significant at either the 1% or 5% level with the exception of column B.
Since the population growth rate is heavily affected by fluctuations in fertility, it was
deemed appropriate to lag population growth by one time period of five years in order to
account for the time between birth and primary school attendance.
Democ and Polcon have negative coefficients, showing that governance
characteristics and PTR share an inverse relationship. Thus, the quality of education
improves as political equality is achieved. Democ and Polcon are used as alternate
measures of governance because they capture different aspects of political processes. Yet,
using both variables in one specification was avoided since these political constraints
indices are highly correlated to one another as seen in Appendix A and may lead to
collinearity problems.

37
As for income inequality, the results suggest that class size tends to increase
when inequality rises. Although the results of the regression appear favorable, the
coefficients of Democ and Polcon are larger and statistically more significant when the
Gini index is excluded from the specification. Compare columns D and B, for instance. In
column D, the Democ coefficients of -0.26 for fe and -0.80 for re are not significant,
whereas in column B in which the Gini index is omitted, the Democ coefficients of -1.23
for fe and -1.83 for re are significant at the 1% level. A plausible explanation for this is
that income inequality may be related to political inequality so that after controlling for
political inequality, the remaining effect of income inequality is reduced. Furthermore,
income inequality is difficult to compare across a large number of countries and
including the Gini index in the regression model greatly reduces n. For example, 291
observations are lost switching from column E to G.

4.2 (b) Hausman Test: Fixed-Effects versus Random-Effects Modeling
Closely investigating the fixed and random-effects specifications of Table 3,
the variations in the regressors of PTR appear to be caused by both between-country
effects and within-country effects given that the significance levels of the right-hand side
variables do not greatly vary between fe and re estimates; an exceptional case is column
H, in which the standard errors of Gini are 0.07 for fe and 5.58 for re, indicating some
loss in efficiency. For the most part, however, this tradeoff did not turn out to be an issue
since just about every explanatory variable retained similar significance levels and
standard errors switching from fe to re and vice versa. Most noteworthy would be log of

38
GDP per capita that has a coefficient always larger when using the fe specification and
statistically significant at the 1% level in both fe and re.
Comparing the outputs of Table 3, even though the data is drawn from a large
number of countries and thus, appears legitimate to presume that the regressors and
unmeasured factors, εjt, are uncorrelated, the Hausman test clearly favors the
specifications of fixed-effects throughout all the models in Table 3. This is most likely
the case because an unbalanced panel data-set can be problematic when using the re
method and the individual unobserved country-specific effects are correlated with the
variables of interest. Moreover, the purpose of using a panel data-set in many instances is
to allow the variables to be correlated (Wooldridge 2002). With that in mind, the fixed-
effects specification turned out to be the better choice.

4.3 Interaction Terms
Next, we turn to Table 4, which present another set of specifications that use
interaction terms by determinants and region. These two tables are an extension of Table
3 not only with the intention of further estimating key determinants, but also to compare
the differences in their impacts on primary schooling quality across countries. Dummy
variables were assigned to the five regions and OECD countries, but only the interaction
terms of the five regions appear in the tables since our interest is in developing countries.
Hence, interaction terms of only Asia, Latin America, MENA, Sub-Saharan Africa, and
Transitional economies with key variables are reported in Tables 4 and 5.

39
The objective initially was to estimate the effects of each variable and region
by employing interaction terms (e.g. Asia*lnGDPpc, Asia*Gini, Asia*Democ/Polcon,
Asia*PopGR, and so forth for each region). However, because of collinearity problems,
doing so proved to be impossible. Hence, we proceeded with a few variables at a time,
only saving most of the key ones. In particular, the interactions involving the Gini
coefficient and population growth had to be excluded because the variations within these
variables were from across regions as opposed to over time. Although not reported, the
interactions of Gini with each region and PopGrLagged with each region presented
problems to other interaction terms as results often were not computable; perhaps, this
was due to collinearity issues. When results were attainable by omitting variables one at a
time, interaction terms of these two variables and other ones yielded insignificant results.
Therefore, the interaction terms of Gini with each region and PopGrLagged with each
region had to be removed from the specifications of Tables 4 and 5, although they were
included individually as non-interacting regressors.
Consequently, the equation takes the form:
5
Qjt = αjt + β
1
*Economic jt + β
2
*Political jt + β
3
*PopGr j t-1 + ∑ β
4
region k * Economic

r
=1

5
+ ∑ β
5
region k * Political + εjt + ujt
r
=1

with the notations identical to those in the preceding discussion, except k denotes the
name of the region. Both fe and re estimates are reported with the regressand again being
quality of education, PTR.
11
This equation performs a sensitivity check using a set of

11
Even though OECD countries are part of the panel, interaction terms correspond only to the five
regions since this technique requires a set of dummy variables that are not included as interaction
terms.

40
identical variables to those in Table 3, but also allows the impact of economic factors (ln
GDP pc and Gini) and political variables (Democ and Polcon) to vary across regions as
shown in the results of Table 4 shown on the next page.

4.3 (a) Table 4 Results
Column A of Table 4 reports estimates obtained using Democ instead of
Polcon, whereas column B is reversely arranged using Polcon. In both columns, Sub-
Saharan Africa ln GDP pc was dropped because of collinearity. Nevertheless, the
remaining political-economy and demographic variables of Table 4 have parameter
estimates that are supportive of the theoretical basis for this study: PTR is inversely
related to income and political equality typically at the 1% and 5% significance levels,
and PTR is positively correlated with Gini and population growth in both fe and re
estimates. But since the Hausman tests support the fixed-effects specifications in columns
A and B, results of the fe estimates are examined more closely.
Both fe estimates consistently report regression outputs with similar
significance levels and signs of coefficients. In columns A and B, PTR seems to be
reduced by income in MENA and Latin America to a greater extent than in other regions.
Particularly in column A, the coefficients of MENA lnGDPpc (-21.32) and Latin
America lnGDPpc (-8.86) are both robust and significant at the 1% level. Combining
each of them with the coefficient of ln GDP pc (-3.95), MENA lnGDPpc is -25.27 (-3.95
+ -21.32) and Latin America lnGDPpc becomes -12.81 (-3.95 + -8.86). This means that

41

Table 4 – Regression Output of Fixed and Random Effects using
Interaction Terms. PTR as dependent variable

A B
Fixed-Effects Random-Effects Fixed-Effects Random-Effects
ln GDP pc
-3.95*** -4.86***
ln GDP pc
-4.17*** -5.10***
(1.06) (0.54) (1.07) (0.52)
Gini
0.02 0.05
Gini
0.01 0.05
(0.06) (0.05) (0.06) (0.05)
Democ
-0.65** -0.67***
Polcon
-8.74* -9.80**
(0.28) (0.23) (4.87) (4.23)
PopGrLagged
0.25** 0.25**
PopGrLagged
0.31*** 0.29***
(0.11) (0.10) (0.11) (0.10)
Asia ln GDP pc
-3.32* -1.28***
Asia ln GDP pc
-2.88* -0.89
(1.69) (0.41) (1.77) (0.40)
Asia Democ
0.94** 0.97***
Asia Polcon
8.87 9.47*
(0.39) (0.35) (6.35) (5.61)
Lt America ln
GDP pc
-8.86*** -0.43
Lt America ln
GDP pc
-9.50*** -0.35
(2.93) (0.34) (2.90) (0.33)
Lt America
Democ
0.38 0.31
Lt America
Polcon
5.76 5.24
(0.33) (0.29) (5.63) (5.12)
Mena ln GDP pc
-21.32*** -0.90*
Mena ln GDP pc
-22.45*** -0.88*
(3.85) (0.48) (3.86) (0.47)
Mena Democ
-1.04 -1.08
Mena Polcon
5.15 -0.59
(0.85) (0.74) (7.71) (7.12)
Sub-Africa
Democ
1.15*** 0.97***
Sub-Africa
Polcon
17.19*** 16.57***
(0.36) (0.30) (5.62) (4.85)
Transition ln
GDP pc
-3.91 -2.33***
Transition ln
GDP pc
-4.17 -2.14***
(3.03) (0.38) (2.95) (0.36)
Transition
Democ
0.27 0.49
Transition
Polcon
2.26 6.30
(0.37) (0.33) (6.28) (5.70)
Constant
88.66*** 69.45***
Constant
90.93*** 69.97***
(7.14) (4.44) (7.08) (4.46)
N 616 616 N 616 616
R
2
Within 0.27 0.21 R
2
Within 0.27 0.21
R
2
between 0.06 0.65 R
2
between 0.06 0.65
R
2
overall 0.05 0.58 R
2
overall 0.04 0.58
Hausman fixed
random
chi
2
(13) 67.92
Hausman fixed
random
chi
2
(13) 604.67
Test Ho:
difference in
coefficients not
systematic
Prob>chi
2
0.00
Test Ho:
difference in
coefficients not
systematic
Prob>chi
2
0.00

Significance levels of 1%, 5%, and 10% are indicated by ***, ** and * respectively.
Standard Errors are in parentheses.

42
compared to other regions, each dollar rise in per capita income in MENA and Latin
American countries has a stronger effect on quality of education.
Findings reported in column B using Polcon indicate similar patterns
discussed so far: interaction terms of MENA income and Latin America income are the
most robust and have larger effects than in other regions. Just as in column A, income
seems to play a bigger role than political equality does in column B. Similar to income,
improvements in governance are most beneficial to schooling quality in certain regions as
observed for MENA in column A and Transitional Economies in column B.
Some of the fe results from both columns A and B are surprising. With respect
to governance, both Democ and Polcon have the expected negative and significant effect
on PTR. However, the significant positive interaction terms for Africa and Asia indicate
that for these regions, democratic governance actually raises PTR. Using the same
method as before, the signs of combined coefficient values of Asia Democ/Polcon (0.29
in column A and 0.13 in column B) and Sub-Saharan Africa governance (0.50 in column
A and 8.45 in column B) are unexpectedly positive.
These results hint at the possibility of unobserved factors, such as historical
and cultural influences, that outweigh the positive effects of democratic governance on
educational quality. Sub-Saharan Africa’s backwards response to a positive stimulus, for
example, can be explained by Africa’s fragmentation and ethnic diversity as Easterly and
Levine argue (1997). In other cases, it may be that more progress is needed since
numerous Asian and Sub-Saharan African countries are prone to reductions in public

43
services due to prolonged civil wars. Indeed, these two regions have had the highest civil
war intensity estimates since 1950.
12

4.3 (b) Table 5 Results
On the next page is Table 5, which displays the patterns found in Table 4:
income has a larger impact on PTR in Latin America and MENA relative to other regions
and governance characteristics are most beneficial in MENA and Transitional Economies
as indicated in columns A and B, respectively. More specifically, in column A for the fe
specification, the coefficients of income are the largest and significant at the 1% level for
Latin America (-9.33) and MENA (-21.43); this recurring pattern is found in column B as
well. With respect to governance, the effects of both Democ and Polcon on PTR are
generally negative and significant. But once again, the interaction terms of these variables
with Asia and Sub-Saharan Africa dummies are large, positive, and significant relative to
the other regions.
Although Tables 4 and 5 report similar results, a crucial difference in Table 5 is
that a new interaction term, Gini*lnGDPpc, is introduced for the purposes of determining
whether the effect of inequality in income or governance would vary with GDPpc. It
turned out that at low levels of income, inequality actually helps primary educational

12
The variable, Civil War Intensities, gauges conflicts between countries that have made threats of
using military force. Based on this data, Sub-Saharan Africa and Asia had the highest reported
average Civil War Intensity values of 0.45 and 0.51, respectively, between 1950 and 2005 in five
year intervals among the five regions and OECD countries (see Jones, D.M., S.A. Bremer, and
J.D. Singer, 1996 for more details on data recording methods).

44

Table 5 – Regression Output of Fixed and Random Effects using
Interaction Terms and Gini*ln GDP pc. PTR as dependent variable

A B
Fixed-Effects Random-Effects Fixed-Effects Random-Effects
Gini*ln GDP pc
0.05* -0.003
Gini*ln GDP pc
0.06** 0.004
(0.03) (0.03) (0.03) (0.03)
ln GDP pc
-6.12*** -4.72***
ln GDP pc
-6.73*** -5.20***
(1.70) (1.26) (1.69) (1.24)
Gini
-0.38 0.07
Gini
-0.47** 0.02
(0.25) (0.22) (0.25) (0.22)
Democ
-0.62** -0.68***
Polcon
-8.54* -9.93**
(0.28) (0.23) (4.86) (4.24)
PopGrLagged
0.26*** 0.24***
PopGrLagged
0.32*** 0.29***
(0.11) (0.10) (0.11) (0.10)
Asia ln GDP
pc
-3.19** -1.27***
Asia ln GDP pc
-2.65 -0.88**
(1.69) (0.41) (1.78) (0.40)
Asia Democ
0.94*** 0.98***
Asia Polcon
8.22 9.33*
(0.40) (0.36) (6.45) (5.71)
Lt America ln
GDP pc
-9.33*** -0.43
Lt America ln
GDP pc
-10.05*** -0.35
(2.94) (0.33) (2.90) (0.33)
Lt America
Democ
0.34 0.32
Lt America
Polcon
5.61 5.35
(0.33) (0.29) (5.62) (5.14)
Mena ln GDP
pc
-21.43*** -0.89**
Mena ln GDP
pc
-22.63*** -0.87**
(3.85) (0.48) (3.85) (0.47)
Mena Democ
-1.09 -1.05
Mena Polcon
4.94 -0.42
(0.85) (0.74) (7.69) (7.14)
Sub-Africa
Democ
1.13*** 0.98***
Sub-Africa
Polcon
17.59*** 16.68***
(0.36) (0.30) (5.62) (4.85)
Transition ln
GDP pc
-3.29 -2.34***
Transition ln
GDP pc
-3.45 -2.15***
(3.05) (0.37) (2.96) (0.36)
Transition
Democ
0.24 0.50
Transition
Polcon
2.20 6.57
(0.38) (0.33) (6.27) (5.71)
Constant
105.37*** 68.40***
Constant
110.78*** 70.76***
(12.45) (9.72) (12.37) (9.66)
N 615 615 N 615 615
R
2
Within 0.27 0.21 R
2
Within 0.27 0.21
R
2
between 0.05 0.65 R
2
between 0.05 0.65
R
2
overall 0.04 0.58 R
2
overall 0.04 0.58
Hausman fixed
random
chi
2
(14) 81.71
Hausman fixed
random
chi
2
(14) 34.71
Test Ho:
difference in
coefficients not
systematic
Prob>chi
2
0.00
Test Ho:
difference in
coefficients not
systematic
Prob>chi
2
0.00
Significance levels of 1%, 5%, and 10% are indicated by ***, ** and * respectively.
Standard Errors are in parentheses.

45
quality: as income rises, the impact becomes more negative on PTR.
13
Gini*lnGDPpc is
0.05 and 0.06 with p-values being 0.10 and 0.05 in columns A and B, respectively. These
figures indicate that in countries where GDPpc is high, income inequality tends to
weaken schooling quality, whereas in countries where income is low, income inequality
promotes educational quality. Hence, this distinction must be realized that income
inequality is not always detrimental to schooling quality.

4.3 (c) Hausman Test: Fixed-Effects versus Random-Effects Modeling
Based on the Hausman Test, another point to address is the advantages and
disadvantages of using the random-effects and fixed effects models. Although they
reported similar results, tradeoffs emerged. For instance, the heterogeneity issue
associated with education in developing countries were treated by re, but the assumptions
of the re specification are unwarranted since this estimation requires that the right-hand
side variables be exogenous and uncorrelated with εjt. On the other hand, when using the
fe modeling, the time-invariant effects on PTR were suppressed and thus, problematic.
Still, however, the fixed-effects model was the better choice because it treated potential
endogeneity issues and the influence of unobserved variables. Therefore, its results were
deemed more reliable.
Additionally, some variations in the coefficient size and significance levels
did occur comparing those of re and fe. For example, ln GDP pc for Latin America and
MENA particularly had different results, switching from re to fe with the latter

13
According to the table in Appendix C, the interaction terms, Gini*Democ (p-value 0.21) and
Gini*Polcon (p-value 0.761) proved to be insignificant, whereas Gini*income turned out
significant as shown in Table 5.

46
specification producing larger coefficients and smaller p-values in both tables. Efficiency
was expected to be lost because switching between the two generally creates such losses.
In this case, the standard errors clearly enlarged and the p-values somewhat shrank when
going from re to fe as seen in both columns A and B (eg. see Asia lnGDPpc, Latin
America lnGDPpc, Mena lnGDPpc, and Transition lnGDPpc). These changes in standard
errors and significance levels when switching from re to fe indicate that the variations in
PTR and its regressors arise from differences within-country effects as opposed to
between-country effects. Examining the loss in efficiency, although re and fe appear to
have produced more or less equal results in terms of p-values and coefficient sizes, the
Hausman test favored the fixed-effects specification, invalidating the assumptions
necessitated by the random-effects estimate. One can expect the coefficients of the
regressors to be correlated, which is accounted for by the fe specification. Therefore, the
outputs of fixed-effects estimates are considered to be superior in Tables 4 and 5 as is
evident in Table 3.

4.4 Summary
In review, although efficiency was lost and a tradeoff had to be dealt with
between the two estimation techniques, the fixed-effects estimate proved to be the
reliable specification in Table 3 in which instrumental variables were employed and in
Tables 4 and 5 that made use of interaction terms. The fe estimates of Table 3 provide
strong evidence that poor economic conditions, such as low income levels, and income
inequality – to a lesser extent – impair schooling quality. This can be seen in column H of

47
Table 3, for instance; the coefficient of lnGDPpc is -6.68 with a significance level of 1%
and the coefficient of Gini is 0.11. Likewise, more democratic governance characteristics
and moderate population growth are beneficial to schooling quality. Referring to Table 3,
column H again: Polcon has a large coefficient of -19.30 at the 1% significance level and
population growth of 0.24 at the 5% significance level.
From Table 4, we can conclude that the introduction of regional interaction
terms does not change the interpretations of Table 3. Table 4 reports that the regional
interaction terms improve income and political equality have stronger influences in
certain regions than in others. Looking at the fixed-effects estimates in both columns of
Table 4, the effects of income are much stronger in MENA and Latin America than
elsewhere and weakest in Asia – maybe because people in Asia are already devotes to
education regardless of changes in income levels. Some of the most robust figures are
interactions of lnGDPpc of MENA and Latin America. Among these two, MENA
lnGDPpc is noteworthy, namely its coefficients of -21.32 and -22.45 in columns A and B
of Table 4, respectively (both with 1% significance levels).
We see similar robust outputs again in Table 5 for per capita income in Latin
America and MENA. However, this is not to say that a rise in per capita income is
impervious to PTR in other regions while progress is only realized in MENA and Latin
America. Rather, these findings can be construed as primary schooling quality in these
two regions gaining the most from mitigations in economic-political factors relative to
Asia, Transitional Economies, and Sub-Saharan Africa. Lastly, an important finding in
Table 5 is the distinction that in countries where per capita GDP is high, income

48
inequality is harmful to primary schooling quality and where wages are low, income
inequality actually can be beneficial.

49
Chapter 5: Summary and Conclusions

This paper examines the impact of political-economy and demographic
variables of schooling quality at the primary level for 123 countries between 1950 and
2005 in five year intervals. There is strong evidence from all the countries analyzed that
favorable political-economy and demographic changes are followed by a decline in the
pupil-to-teacher ratios, or improvement in educational quality. That is, PTR shares an
inverse relationship with GDP per capita and governance characteristics (Democ and
Polconiii_2002), whereas it has a positive correlation with the Gini coefficient and
population growth. Empirical findings based on the two educational production functions
proved to support this theory.
The first set of specifications includes instrumental variables to treat potential
reverse causation problems. In particular, we use primary exports as instruments for the
two governance measures. To allow for differences in the effects of some of the
independent variables, the regression model was further developed through the use of
dummy interaction terms to allow for differentiated effects of key determinants of
schooling quality across regions and over time. Interesting findings include the effects of
per capita GDP and income inequality as shown in Table 5, in which we find significant
differences in the effects of inequality across countries of varying income levels.
Under all estimation procedures, the Hausman test favored the fixed-effects
method, and its outputs were convincing of primary educational quality being impacted
mostly by per capita income that always reported the largest and significant parameter

50
estimates. As shown in Appendix C, the interaction between income and Gini proved to
be more significant than the one between political indices and Gini. Perhaps, income has
a larger role in influencing PTR because as the standard of living improves and parents’
earnings rise – unveiling traces of economic growth and a shift away from an agrarian
society – school attendance becomes feasible and the government responds to increasing
enrollment rates by hiring more instructors. Therefore, one can expect positive outcomes
in terms of educational quality by stimulating economic conditions and to a lesser extent,
though not to be overlooked, by restoring political equality and maintaining low levels of
population growth.
This study, however, is not a prescription for improving PTR through
modifications in political-economic factors and population growth since cultural and
historical differences, civil wars, ethnic diversity, and other factors can have a stronger
effect on schooling quality in different parts of the world. In addition, assuming that PTR
alone is a comprehensive indicator of primary educational quality would be unwarranted
because PTR portrays only one dimension of schooling quality. Although PTR has the
advantages of effectively measuring educational quality on a global scale and lending
wide comparability to a vast cross-sectional study, PTR alone does not assess the
effectiveness of schools. There might be unobserved factors – wages of faculty and staff,
teaching skills, the physical characteristics of classrooms, government funding, and so
forth – that the constant term of the fixed-effects estimate measures, but cannot identify.
Therefore, further research can incorporate other proxies or combine them to
estimate other aspects of schooling quality, such as instructors’ training and years of

51
teaching experience, internationally comparable test scores, and WAIS-R that accounts
for country-specific differences. Such measures may better determine the effects of per
capita income, the Gini coefficient, political constraints indices, and population growth.
Lastly, new research to evaluate how schooling quality is affected by civil war intensities,
culture, and income inequality within countries of high and low income, as found in
Table 5, is a recommended topic.

52
References

Acemoglu, D. and Robinson, J., 2000. Political Losers as a Barrier to Economic
Development. The American Economic Review 90, 126-130.

Angrist, J.D. and Lavy, V., 1999. Using Maimonides' Rule to Estimate the Effect of Class
Size on Scholastic Achievement. The Quarterly Journal of Economics 114 (2), 533-575.

Barro, R.J. and Lee, J.W., 2001. Schooling Quality in a Cross-Section of Countries.
Economica 68, 465-488.

Bertocchi, G. and Spagat, M. 2004. The Evolution of Modern Educational Systems.
Technical vs. General Education, Distributional Conflicts, and Growth. Journal of
Development Economics, 73. 559-582.

Bourguignon, F. and Verdier, T., 2000. Is Financial Openness Bad for Education? A
Political Economy Perspective on Development. European Economic Review 44 (4-6),
891-903.

Bishop, J.H., 1989. Is the Test Score Decline Responsible for the Productivity Growth
Decline? The American Economic Review 79 (1), 178-197.

Black, S.E., 1999. Do Better Schools Matter? Parental Valuation of Elementary
Education. The Quarterly Journal of Economics 114 (2), 577-599.

Brown, D.S. and Hunter, W., 1999. Democracy and Social Spending in Latin America.
American Political Science Review 93 (4), 779-790.

Card, D. and Krueger, A.B., 1992. Does School Quality Matter? Returns to Education
and the Characteristics of Public Schools in the United States. The Journal of Political
Economy 100 (1), 1-40.

Chen, H., 2005. Educational Systems, Growth and Income Distribution: A
Quantitative Study. Journal of Development Economics 76, 325-353.

Cummins, J., 1992. Empowering Minority Students. In Policy and Practice in Bilingual
Education: Extending the Foundations, O. Garcia and C. Baker, eds. Bristol, PA:
Multilingual Matters Ltd., 103-117.

Drèze, J., and Sen, A., 2002. Democratic Practice and Social Inequality in India. Journal
of Asian and African Studies 37 (2), 6-37.

Easterly, W. and Levine, R., 1997. Africa's Growth Tragedy: Policies and Ethnic
Divisions. The Quarterly Journal of Economics 112 (4), 1203-1250.

53
Finn, J.D. and Achilles, C.M., 1990. Answers and Questions about Class Size: A
Statewide Experiment. American Educational Research Journal XXVII, 557-577.

Gilbert, A., 1998. The Latin American City. Michigan: Latin America Bureau.

Glass, G.V., L.S. Cahen, M.L. Smith, and N.N. Filby, 1982. School Class Size: Research
and Policy. Beverly Hills, CA: Sage.

Glass, G.V. and Smith, M.L., 1979. Meta-Analysis of Research on Class Size and
Achievement. Educational Evaluation and Policy Analysis I, 2-16.

Gradstein, M. and Justman, M., 1997. Democratic Choice of an Education System:
Implications for Growth and Income Distribution. Journal of Economic Growth 2, 169–
183.

Haggard, Stephan, and Robert R. Kaufman., 1995. The Political Economy of Democratic
Transitions. Princeton, NJ: Princeton University Press.

Lee, R.B., 2003. The Dobe Ju/’hoansi. Toronto, Ontario: Thomson Learning, Inc.

Harbison, R.W. and Hanushek, E.A., 1992. Educational Performance of the Poor:
Lessons from Rural Northeast Brazil by Washington, D.C. Oxford University Press for
the World Bank.

Henisz, W. J., 2002. POLCONIII. The Institutional Environment for Infrastructure
Investment. Industrial and Corporate Change 11 (2): Forthcoming.

Jones, D.M., S.A. Bremer, and J.D. Singer, 1996. Militarized Interstate Disputes, 1816-
1992: Rationale, Coding Rules, and Empirical Patterns. Conflict Management and Peace
Science 15 (2), 163-213.

Krueger, A.B., 1999. Experimental Estimates of Education Production Functions. The
Quarterly Journal of Economics 114 (2), 497-532.
Ladd, H.F., 1992. Population Growth, Density and the Costs of Providing Public
Services. Urban Studies 29 (2), 273-295

Lockheed, M.E., A.M. Verspoor, and associates, 1991. Improving Primary Education in
Developing Countries. Oxford: Oxford University Press for the World Bank.

Moll, P.G., 1998. Primary Schooling, Cognitive Skills and Wages in South Africa.
Economica, New Series 65 (258), 263-284.

54
Motiram, S. and Nugent, J.B., 2007. Inequality, Quality of Primary Education and
Development in Latin America and the Caribbean. Conference on the Quality of
Education in Latin America and the Caribbean. Mexico City, Mexico. February 2007.

Mueller, D., C.I. Chase, and J.D. Walden, 1988. Effects of Reduced Class Size in
Primary Classes. Educational Leadership XLV, 49-50.

Palafox, J.C., J. Prawda, and E. Velez., 1994. Primary School Quality in Mexico.
Comparative Education Review 38 (2), 167-180.

Plank, D.N., 1990. The Politics of Basic Educational Reform in Brazil. Comparative
Education Review 34 (4), 538-559.

Robert, T., M.G. Marshall, and K. Jaggers 2008. Polity IV Project: Political Regime
Characteristics and Transitions, 1800-2007, available at http://www.systemicpeace. org/
polity/polity4.htm

Robinson, J., 2004. When is a State Predatory? mimeo, Harvard University.

Rutter, M., 1983. School Effects on Pupil Progress: Research Findings and Policy
Implications. Child Development 54, 1-29.

Shen, J., 1998. China’s Future Population and Development Challenges. The
Geographical Journal 164 (1), 32-40.

Wooldridge, J.M., 2002, Econometric Analysis of Cross Section and Panel Data - 2
nd
Ed.
Massachusetts: MIT Press.

Wright, E.N., S.M. Shapson, G. Eason, and J. Fitzgerald, 1977. Effects of Class Size in
the Junior Grades. Toronto: Ontario Ministry of Education.

UNESCO, 1999. Institute for Statistics. Statistical Yearbook, Paris: UNESCO.

UNESCO web address: http://stats.uis.unesco.org/ReportForlders/reportfolders.aspx

World Development Indicators, World Bank.

Zsigmond, Z., 1976. Imapct of Projected Population Trends on Post-Secondary Education
1961-2001. The Canadian Journal of Statistics 4 (2), 255-276.

55
Appendices

Appendix A – Correlation Matrix using the same data-set as in Table 2

The following correlation matrix is intended to supplement the regressions of Table 3 in
Chapter 4; we can see that PrimExport is correlated with the endogenous explanatory
variables. Note that the table below is based on data drawn from both OECD countries
and the five regions, just as with the regressions in Table 3. The correlation matrix of
Table 2 is for comparing OECD countries with the five regions and does not combine the
entire 123 countries.

PTR lnGDPpc Gini Democ Polcon
Pop Gr
Lagged
Prim
Export
PTR 1
lnGDPpc -0.6716 1
Gini 0.4247 -0.3257 1
Democ -0.4633 0.6359 -0.2243 1
Polcon -0.369 0.4878 -0.1057 0.7336 1
Pop Gr
Lagged
0.2808 -0.2155 0.219 -0.186 -0.1551 1
PrimExport 0.3937 -0.4945 0.3509 -0.4284 -0.4252 0.2132 1

56
Appendix B – Testing Residuals of Democ and Polcon

Dependent Variable

Dependent Variable

Democ PTR Polcon PTR
ln GDPpc
1.50*** -1.51
ln GDPpc
0.05*** -3.72***
(0.08) (1.21) (0.0048) (0.61)
PopGrLagged
-0.04 0.35***
PopGrLagged
-0.0002 0.45***
(0.03) (0.12) (0.0018) (0.11)
PrimExport
-0.02***
PrimExport
-0.0019***
(0.0042) (0.0002)
Democ
hat
-2.57***
Polcon
hat
-30.50***
(0.68) (7.74)
Democ
residual
-0.22*
Polcon
residual
-3.51*
(0.12) (2.09)
Constant
-4.98*** 53.74***
Constant
-0.0043 65.86***
(0.80) (5.76) (0.05) (2.80)
N 819 766 N 825 775
R
2
0.44 0.46 R
2
0.32 0.46
R
2
Adjusted 0.44 0.45 R
2
Adjusted 0.31 0.46
F-test
Democ Residual
0.0764
F-test
Democ Residual
0.0941
Significance levels of 1%, 5%, and 10% are indicated by ***, ** and * respectively.
Standard Errors are in parentheses.

57
Appendix C – Gini*lnGDPpc versus Gini*Governance

Gini*Democ and Gini*Polcon are insignificant, indicating that income and political
inequality do not have such a big role as Gini*lnGDPpc from Table 5. However, an
interesting pattern is that per capita incomes of MENA and Latin America seem to be the
most robust as seen in Tables 4 and 5. Also, in Appendix C on the next page, column B is
the only place in this study where the random-effects specification is favored by the
Hausman test.

58
Appendix C: continued

A B
Fixed-Effects Random-Effects Fixed-Effects Random-Effects
Gini*Democ
0.01 0.00
Gini*Polcon
0.06 -0.05
(0.01) (0.01) 0.19 (0.18)
ln GDP pc -3.96*** -4.85***
ln GDP pc
-4.16*** -5.09***
(1.06) (0.54) 1.08 (0.52)
Gini
-0.06 0.02
Gini
-0.01 0.06
(0.08) (0.07) 0.08 (0.07)
Democ
-1.15** -0.85*
Polcon
-10.73 -7.85
(0.49) (0.47) 8.16 (8.01)
PopGrLagged
0.26** 0.25***
PopGrLagged
0.31*** 0.29***
(0.11) (0.10) 0.11 (0.10)
Asia ln GDP pc
-3.09* -1.29***
Asia ln GDP pc
-2.85 -0.87**
(1.70) (0.42) 1.78 (0.41)
Asia Democ
0.81** 0.95***
Asia Polcon
8.23 9.36*
(0.41) (0.36) 6.54 (5.71)
Lt America ln
GDP pc
8.82*** -0.42 Lt America ln
GDP pc
-9.48*** -0.36
(2.93) (0.34) 2.90 (0.33)
Lt America Democ
0.16 0.25 Lt America
Polcon
4.86 6.01
(0.37) (0.32) 6.36 (5.65)
Mena ln GDP pc
-21.75*** -0.92*
Mena ln GDP pc
-22.51*** -0.86*
(3.87) (0.48) 3.87 (0.47)
Mena Democ
-1.16 -1.11
Mena Polcon
4.92 -0.41
(0.86) (0.74) 7.75 (7.14)
Sub-Africa Democ
0.94** 0.92*** Sub-Africa
Polcon
16.49*** 17.10***
(0.40) (0.32) 6.09 (5.12)
Transition ln GDP
pc
-3.83 -2.38*** Transition ln
GDP pc
-4.17 -2.11***
(3.03) (0.39) 2.95 (0.37)
Transition Democ
0.37 0.53 Transition
Polcon
2.62 5.96
(0.38) (0.35) 6.41 (5.92)
constant
91.78*** 70.43***
constant
91.53*** 69.25
(7.56) (4.95) 7.36 (4.89)
N 615 615 N 615 615
R2 Within 0.27 0.21 R2 Within 0.27 0.20
R2 between 0.06 0.65 R2 between 0.06 0.65
R2 overall 0.04 0.58 R2 overall 0.04 0.59
Hausman fixed
random
chi
2
(14) 70.47
Hausman fixed
random
chi
2
(14) 7.49
Test Ho:
difference in
coefficients not
systematic
Prob>chi
2
0.00
Test Ho:
difference in
coefficients not
systematic
Prob>chi
2
0.91

Significance levels of 1%, 5%, and 10% are indicated by ***, ** and * respectively.
Standard Errors are in parentheses.

59
Appendix D – Country List and Average Figures

Asia PTR GER% NER% GDPpc Gini Democ Polcon PopGr
Primary
Export
1. Bangladesh 51 76 70 $ 285 39 1.50 0.17 2.15 22
2. China 29 91 99 $ 439 31 0 0 1.51 31
3. Fiji 32 100 96 $ 1,684 46 7.03 0.33 1.94 52
4. India 43 72 79 $ 301 35 8.66 0.42 1.99 39
5. Indonesia 29 101 99 $ 541 31 0 0 1.85 31
6. South Korea 45 91 97 $ 5,428 34 7.28 0.32 0.82 23
7. Laos 28 70 74 $ 288 30 0 0 2.23 94
8. Malaysia 26 82 85 $ 2,255 50 5.73 0.34 2.44 63
9. Mongolia 29 97 87 $ 413 27 2.45 0.05 1.73 64
10. Nepal 33 64 71 $ 168 37 1.65 0.07 2.42 38
11. Pakistan 38 45 43 $ 381 36 3.80 0.48 2.67 39
12.
Papua New
Guinea
31 59 59 $ 564 51 10 0.51 2.20 86
13. Philippines 34 98 96 $ 859 48 4.36 0.30 2.47 53
14. Singapore 29 97 97 $ 11,720 39 2.75 0.06 2.63 40
15. Sri Lanka 41 98 87 $ 536 40 6.62 0.39 1.74 61
16. Thailand 28 89 77 $ 1,158 44 4.86 0.32 2.03 57
AVERAGE 34 83 82 $ 1,689 39 4.17 0.23 2.05 50
MINIMUM 26 45 43 $ 168 27 0 0 0.82 22
MAXIMUM 51 101 99 $ 11,720 51 10 0.51 2.67 94

Latin America PTR GER% NER% GDPpc Gini Democ Polcon PopGr
Primary
Export
17. Argentina 21 100 96 $ 6,710 44 4.95 0.17 1.48 77
18. Bolivia 31 76 88 $ 976 51 3.47 0.21 2.03 91
19. Brazil 25 87 82 $ 2,706 56 4.18 0.39 2.21 64
20. Colombia 32 95 75 $ 1,592 56 6.88 0.33 2.30 79
21. Costa Rica 26 97 89 $ 3,034 50 10 0.35 2.61 67
22.
Dominican
Republic
44 92 79 $ 1,509 47 3.42 0.23 2.51 52
23. Ecuador 33 96 88 $ 1,195 48 5.84 0.15 2.43 96
24. El Salvador 38 71 77 $ 1,880 48 2.57 0.26 2.28 64
25. Guatemala 31 66 66 $ 1,435 58 2.03 0.27 2.74 66
26. Guyana 30 108 88 $ 786 40 2.30 0.24 1.27 38
27. Honduras 34 85 81 $ 850 57 4.18 0.21 3.12 83
28. Jamaica 44 92 94 $ 3,002 43 9.70 0.34 1.18 79
29. Mexico 37 94 91 $ 4,550 52 1.68 0.17 2.38 55
30. Nicaragua 35 77 76 $ 1,032 50 2.08 0.16 3.08 87
31. Panama 27 98 89 $ 3,097 55 3.22 0.26 2.32 43
32. Paraguay 27 100 85 $ 1,234 51 1.78 0.14 2.62 89
33. Peru 34 94 90 $ 2,020 52 3.12 0.27 2.30 83
34.
Trinidad and
Tobago
30 101 90 $ 5,525 42 8.75 0.40 0.81 83

60
Appendix D: continued

35. Uruguay 26 101 89 $ 4,892 43 4.60 0.39 0.78 66
36. Venezuela 31 88 84 $ 5,423 44 7.76 0.32 2.98 98
AVERAGE 32 91 85 $ 2,672 49 4.63 0.26 2.17 73
MINIMUM 21 66 66 $ 786 40 1.68 0.14 0.78 38
MAXIMUM 44 108 96 $ 6,710 58 10 0.40 3.12 98

M.E.N.A. PTR GER% NER% GDPpc Gini Democ Polcon PopGr
Primary
Export
37. Algeria 34 73 66 $ 1,643 37 0.43 0.09 2.23 70
38.
Egypt Arab
Rep
31 84 76 $ 976 36 0.17 0.12 2.22 75
39.
Iran, Islamic
Rep. of
29 72 79 $ 1,563 43 0.60 0.05 2.73 95
40. Jordan 50 78 94 $ 1,759 38 1.15 0.14 4.23 54
41. Morocco 33 59 59 $ 945 39 0.17 0.14 2.09 61
42. Tunisia 37 85 91 $ 1,457 42 0.17 0.01 1.80 50
43. Turkey 37 83 94 $ 2,187 48 6.91 0.31 2.09 59
44.
Yemen
Republic
40 41 63 $ 511 36 0.25 0 2.86 80
AVERAGE 36 72 78 $ 1,380 40 1.23 0.11 2.53 68
MINIMUM 29 41 59 $ 511 36 0.17 0 1.80 50
MAXIMUM 50 85 94 $ 2,187 48 6.91 0.31 4.23 95

Sub-Saharan Africa PTR GER% NER% GDPpc Gini Democ Polcon PopGr
Primary
Export
45. Angola 38 51 49 $ 756 40 0 0.15 1.90 94
46. Benin 44 59 52 $ 292 37 1.84 0.17 2.57 70
47. Botswana 32 86 75 $ 1,554 60 7.77 0.20 2.48 10
48.
Burkina Faso
(Upper
Volta)
52 23 26 $ 197 48 0.20 0.14 2.30 65
49. Burundi 49 48 46 $ 122 35 0.17 0.05 2.22 90
50. Cameroon 50 82 70 $ 666 49 0.50 0.03 2.30 90
51. Cape Verde 44 87 93 $ 1,027 57 0 0.13 2.24 69
52.
Central
African
Republic
75 51 53 $ 297 58 1.50 0.19 2.05 55
53. Chad 70 38 41 $ 202 40 0.28 0 2.26 96
54. Cote d’Ivoire 40 52 52 $ 717 39 0.50 0.01 3.43 79
55. Ethiopia 48 31 33 $ 122 38 0.75 0.06 2.52 96
56. Gabon 52 119 86 $ 3,797 52 0 0 2.41 98
57. Gambia 29 44 43 $ 316 48 2.52 0.15 3.01 84
58. Ghana 30 66 52 $ 247 36 0.92 0.13 2.67 74
59. Guinea 45 33 42 $ 358 44 0.25 0.07 2.78 41
60. Kenya 36 76 68 $ 376 54 0.85 0.14 3.16 84
61. Lesotho 51 104 72 $ 322 62 0 0.05 1.70 8

61
Appendix D: continued

62. Liberia 36 40 54 $ 490 47 0 0 2.77 98
63. Madagascar 58 79 69 $ 312 45 1.62 0.25 2.61 80
64. Malawi 57 71 71 $ 137 58 1.75 0.14 2.85 92
65. Mauritania 36 36 60 $ 357 41 0.07 0.06 2.42 71
66. Mauritius 28 91 89 $ 2,856 40 4.80 0.39 1.67 38
67. Mozambique 78 60 50 $ 189 40 1.50 0.15 2.16 93
68.
Namibia
(South West
Africa)
32 120 74 $ 1,841 72 1.50 0.32 2.52 43
69. Niger 40 19 27 $ 223 43 0.60 0.14 2.81 89
70. Nigeria 33 63 64 $ 361 46 1.53 0.19 2.52 97
71. Rwanda 52 61 66 $ 235 29 0.13 0.04 3.24 100
72. Senegal 48 43 51 $ 435 50 1.07 0.12 2.82 71
73. Sierra Leone 34 44 50 $ 238 62 0.55 0.10 2.15 48
74. South Africa 34 99 91 $ 3,071 61 5.75 0.36 1.95 40
75. Swaziland 35 84 75 $ 1,123 61 0.15 0.00 2.50 44
76. Togo 70 75 71 $ 269 39 0.25 0.00 2.99 70
77. Uganda 35 57 44 $ 211 36 0.92 0.13 2.96 95
78.
Zaire Dem
Rep. (Congo
Kinshasa)
42 76 59 $ 227 54 0.45 0.00 2.91 93
79. Zambia 46 77 68 $ 436 49 0.98 0.12 0.83 91
80.
Zimbabwe
Rhodesia
38 100 83 $ 567 56 1.08 0.19 2.73 64
AVERAGE 45 65 60 $ 693 48 1.19 0.12 2.48 73
MINIMUM 28 19 26 $ 122 29 0 0 0.83 8
MAXIMUM 78 120 93 $ 3,797 72 7.77 0.39 3.43 100

Transitional
Economies
PTR GER% NER% GDPpc Gini Democ Polcon PopGr
Primary
Export
81. Albania 25 99 96 $ 1,125 33 1.80 0.09 1.62 19
82. Azerbaijan 10 93 85 $ 894 31 0.00 0.00 1.82 93
83. Belarus 13 103 91 $ 1,368 25 0.60 0.00 0.35 39
84. Bulgaria 19 92 97 $ 1,635 25 2.00 0.13 0.03 42
85. Croatia 16 93 84 $ 4,295 29 2.50 0.42 0.28 28
86.
Czech
Republic
22 96 88 $ 5,551 24 9.20 0.47 0.21 13
87. Estonia 13 104 96 $ 3,954 36 7.50 0.55 0.33 26
88. Hungary 18 95 94 $ 3,434 24 2.50 0.11 0.24 28
89. Kazakhstan 9 98 90 $ 1,459 33 1.67 0.00 1.49 76
90. Latvia 11 89 88 $ 3,080 31 7.25 0.41 0.19 34
91. Lithuania 11 91 93 $ 3,760 31 9.75 0.43 0.44 42
92. Moldova 17 88 88 $ 558 33 6.75 0.48 -0.03 68
93. Poland 17 94 91 $ 3,178 29 5.27 0.29 0.45 35
94. Romania 21 90 96 $ 1,918 30 1.85 0.16 0.57 57

62
Appendix D: continued

95.
Russian
Federation
19 108 94 $ 12,050 41 6.25 0.09 0.56 68
96. Slovenia 16 106 98 $ 9,236 27 10 0.55 0.42 12
97. Tajikistan 17 96 85 $ 314 33 0.66 0.20 2.58 74
98. Ukraine 15 98 81 $ 913 29 6.60 0.31 0.36 31
AVERAGE 17 96 91 $ 3,108 30 4.32 0.25 0.73 44
MINIMUM 9 84 81 $ 314 24 0 0.00 -0.03 12
MAXIMUM 33 108 98 $ 12,050 41 10 0.55 2.58 93

J.A.N PTR GER% NER% GDPpc Gini Democ Polcon PopGr
Primary
Export
100. Japan 25 98 98 $ 14,847 38 10 0.49 1.69 77
101. Australia 28 95 100 $ 25,241 34 10 0.49 0.73 6
102. New Zealand 21 103 100 $ 11,305 36 10 0.37 1.31 78
AVERAGE 25 99 99 $ 17,131 36 10 0.45 1.24 53
MINIMUM 21 95 98 $ 11,305 34 10 0.37 0.73 6
MAXIMUM 28 103 100 $ 25,241 38 10 0.49 1.69 78

US and Canada PTR GER% NER% GDPpc Gini Democ Polcon PopGr
Primary
Export
103. Canada 21 95 95 $ 17,840 31 10 0.42 1.42 47
104. United States 22 103 95 $ 24,671 38 10 0.40 1.10 25
AVERAGE 22 99 95 $ 21,256 35 10 0.41 1.26 36
MINIMUM 21 95 95 $ 17,840 31 10 0.40 1.10 25
MAXIMUM 22 103 95 $ 24,671 38 10 0.42 1.42 47

Western Europe PTR GER% NER% GDPpc Gini Democ Polcon PopGr
Primary
Export
105. Austria 18 98 96 $ 16,467 24 10 0.44 0.33 19
106. Belgium 18 98 98 $ 15,621 27 10 0.59 0.32 21
107. Cyprus 27 91 88 $ 7,715 30 9.66 0.29 1.00 40
108. Denmark 19 89 98 $ 1,195 35 10 0.53 0.40 52
109. Finland 19 96 99 $ 16,107 30 10 0.54 0.45 27
110. France 21 109 98 $ 16,032 36 8.03 0.36 0.65 23
111. Germany 16 100 100 $ 19,045 28 10 0.47 0.39 11
112. Greece 28 89 97 $ 7,827 36 8.51 0.32 0.69 56
113. Ireland 31 104 100 $ 13,084 36 10 0.43 0.52 38
114. Israel 18 93 89 $ 12,642 34 9.29 0.53 3.02 19
115. Italy 18 85 98 $ 13,182 37 10 0.47 0.38 16
116. Luxembourg 19 87 79 $ 26,008 25 8.00 0.45 0.85 11
117. Netherlands 26 90 95 $ 16,400 32 10.00 0.49 0.87 40
118. Norway 19 94 99 $ 23,871 33 10 0.48 0.80 64
119. Portugal 25 105 93 $ 6,421 36 4.13 0.21 0.73 26
120. Spain 29 99 97 $ 9,645 33 4.13 0.24 0.67 35

63
Appendix D: continued

121. Sweden 16 94 98 $ 20,165 27 10 0.48 0.45 21
122. Switzerland 18 85 98 $ 28,272 34 10 0.62 0.76 8
123.
United
Kingdom
22 98 98 $ 17,319 35 10 0.36 0.28 20
AVERAGE 21 95 96 $ 15,106 32 9.04 0.44 0.71 29
MINIMUM 16 85 79 $ 1,195 24 4.13 0.21 0.28 8
MAXIMUM 31 109 100 $ 28,272 37 10 0.62 3.02 64

GLOBAL Average PTR GER% NER% GDPpc Gini Democ Polcon PopGr
Primary
Export
Five Regions 33 81 79 $ 1,909 41 3.11 0.19 1.99 62

OECD
Countries
23 98 97 $ 17,831 34 10 0.43 1.07 39
Differential 10 17 18 $ 15,922 7 6.57 0.24 0.92 22

Descriptive statistics derived from these soures: PTR, GER, and NER from UNESCO; Democ by Roberts, Marshall and Jaggers;
Polcon by Henisz; and GDP per capita, Gini, Population Growth Rate, and Primary Export from World Development Indicators.
Note that some observations are missing and figures are based on average values taken from 1950-2005 in five year increments.

Abstract (if available)

Abstract The objective of this study is to empirically examine the effects of economic conditions, political factors, and population growth on schooling quality at the primary level across various parts of the world and over time. Quality of education is measured primarily in terms of pupil-to-teacher ratios. The data employed is a panel data-set of 123 countries representing Asia, Latin America, Middle East and North Africa, Sub-Saharan Africa, Transitional economies, and developed OECD countries between 1950 and 2005 in five year intervals. The results show that the quality of primary education is positively affected by increases in GDP per capita and more democratic governance, and negatively influenced by income inequality and population growth. This study relies on comparisons of fixed-effects and random-effects regression analyses, involving interaction terms both between these determinants and region dummy variables, and between income inequality and other factors in order to measure differences in the effects of these factors across regions on the quality of primary education. An important finding is that income inequality tends to be detrimental to primary schooling quality in countries with relatively high per capita income.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Essays in panel data analysis

PDF

The impact of agglomeration policy on CO₂ emissions: an empirical study using China’s manufacturing data

PDF

Political determinants and economic effects of HIV/AIDS: a push for the multisectoral approach

PDF

Essays on health and aging with focus on the spillover of human capital

PDF

Essays on econometrics analysis of panel data models

PDF

Three essays on supply chain networks and R&D investments

PDF

The impact of trade liberlaization on firm performance in developing countries -- new evidence from Pakistani manufacturing sector

PDF

Three essays on linear and non-linear econometric dependencies

PDF

Essays on macroeconomics and income distribution

PDF

Inter-temporal allocation of human capital and economic performance

PDF

Essays on health economics

PDF

Three essays on social policy: institutional development, and subjective well-being as a cause and consequence of labor market outcomes

PDF

An empirical analysis of the relationship between real estate investment and regional economy in China

PDF

Essays on the econometric analysis of cross-sectional dependence

PDF

Intergenerational transfers & human capital investments in children in the era of aging

PDF

Three essays on health & aging

PDF

Essays on education and institutions in developing countries

PDF

Financial crises and trade policy in developing countries

PDF

Three essays on the identification and estimation of structural economic models

PDF

Comparative study on Asian financial capitals' competitiveness: focused on strengths and weaknesss of the city of Seoul

Asset Metadata

Creator Song, Ike I. (author)

Core Title An empirical analysis of the quality of primary education across countries and over time

School College of Letters, Arts and Sciences

Degree Master of Arts

Degree Program Economics

Publication Date 03/25/2010

Defense Date 05/22/2009

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag demographic pattern,governance characteristics,OAI-PMH Harvest,per capita income,political-economic factors,primary education,schooling quality

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Nugent, Jeffrey B. (committee chair), Hsiao, Cheng (committee member), Kim, Yong Jin (committee member)

Creator Email isong@usc.edu,song_ike@hotmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-m2880

Unique identifier UC1461608

Identifier etd-Song-3400 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-291526 (legacy record id),usctheses-m2880 (legacy record id)

Legacy Identifier etd-Song-3400.pdf

Dmrecord 291526

Document Type Thesis

Rights Song, Ike I.

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Repository Name Libraries, University of Southern California

Repository Location Los Angeles, California

Repository Email cisadmin@lib.usc.edu