University of Groningen Faculty of Economics and Business MSc in International Economics and Business
(IE&B) Master’s thesis June, 2017
The Role of Education and Inequality in
Explaining Democracy
Supervisor: Assistant Professor DR. Dirk Akkermans Co-Assessor: Research Associate DR. Tarek Harchaoui Author: Pelin Özgül
Student ID: S2860325
1
ABSTRACT
In this thesis, the impact of education and inequality on democracy is studied. The analysis is based upon the well-known Modernization Theory of Seymour Lipset who argues the need of a continuous and an effective economic development to sustain democratic political institutions. Additionally, following Edward Glaeser’s 2006 work, it is stated that there exists a mediating effect of political participation on the relationship between education and democracy. To investigate the mechanism further, a mediation analysis is conducted where political participation takes the form of the potential mediator. To distinguish this research from other conventional studies, two different democracy indices are used since the literature is not challenged with alternative measurements. The empirical results show that when it comes to investigate such relationship, the choice of democracy measure matters. While Lipset’s modernization hypothesis finds empirical support for one democracy indicator, it is only partly supported for the other. The mediation hypothesis on the other hand, finds no empirical support. Thus, the fact that both education and inequality are not mediated by political participation, in the context of democracy, indicates that education and inequality might be a proxy for other unobserved characteristics.
2
TABLE OF CONTENTS
Page Number
1.Introduction 4
2. Theoretical Framework 8
2.1 As Socio- Economic Drivers: Education and Inequality 8
2.2 The Great Debate: Democracy 9
2.3 Education and Democracy 11
2.4 Inequality and Democracy 14
2.5 Importance of Civic Participation 16
2.6 Education and Civic Participation 16
2.7 Inequality and Civic Participation 18
3. Data and Methodology 19
3.1 Measurement Instruments 19
3.1.1 Dependent Variables 19
3.1.2 Independent Variables 21
3.1.3 Mediating Variable 22
3.1.4 Control Variables 22
3.2 Imputation of missing data 23
3.3 Methodology & Research Model 24
4. Results 26
4.1 Empirical Results 26
3
4.2.1 Descriptive Statistics 27
4.2.2 Correlation Matrix 27
4.2.3 Check for Multicollinearity 27
4.2.4 Check for Normality 27
4.3 Regression Results 29
4.4 Correlation Between Two Democracy Indicators 32
5. Conclusion and Limitations 33
6. Bibliography 35
7. APPENDICES 38
Appendix 1 : List of Countries 39
Appendix 2 : Summary Statistics 39
Table 1 : Summary Statistics of Polity IV 40
Table 2 : Summary Statistics of PEI 40
Appendix 3 : Correlation Matrices 42
Table 3 : Correlation Matrix of Polity IV 42
Table 4 : Correlation Matrix of PEI 43
Appendix 4 : Regression Results for Mediation Analysis 44
Table 8 : Mediation Results 44
Table 9 : Mediation Results 45
Table 10 : Mediation Results 46
4
1- Introduction
The world has experienced multiple transitions over the last decades and moved to a disparate phase. Changes within the socio-economic trends were mainly attributed to the consequences of demographic shifts, industrialization practices and modernization movements as well as democratic transitions. Even though not all countries had experienced the democratization trend in the same way, such as Nigeria, Thailand, Venezuela and especially Iraq and Afghanistan, where Bush administration attempted to establish “democracy”, studies argue that there exist a general upward trend. Many countries moved from autocratic regimes that have both lower participation in political decision making and weak constraints on executive power to more democratic regimes that includes broader political participation and greater limits on the exercise of political power (Murtin & Wacziarg 2014). Because of the differences in population, the plain number of democratic countries might be unable to present how many people in the world actually possess democratic rights. Therefore, looking at the share of people governed by different political regimes reveals a more interesting picture. As shown in Figure 1 below, one can see that throughout 19th century, majority of the world population either lived in autocratic regimes or colonized by imperial powers. Not earlier than the second half of the 20th century, a boost in democratization trend occurred, colonial empires collapsed and eventually, more and more countries moved towards into democratic regimes. Since then, the share of the world population living in democracies is increasing continuously ( Roser, 2016).
Figure 1: Number of world citizens living under different political regimes
5
In his book, “The Third Wave: Democratization in the Late Twentieth Century”, Samuel Huntington discusses this global upward wave of democratization occurring all over Europe, Latin America, Africa and Asia. A wave of democratization was described as group of transitions from non-democratic regimes to democratic systems that take place within a specific time and dominate transitions in the opposite direction, whereas the 3rd wave, refers to the widespread global push towards democracy occurred in 1970s and 1980s (Huntington,1993). While investigating the trend and its underlying causes, Huntington states that democratic transitions, collapses or cases of overthrown all resulted from variety of dynamics. Alongside the changes in political, military factors and religious alterations, the role of socio economic factors seem to have significant impact on this trend. Thus, although analyzing this democratization movement seems interesting all by itself, the underlying socio-economic factors that might have led these transitions would be much more sophisticated for a discussion.
Starting from 18th century, various factors emerged and formed the basis of these structural changes; industrialization, urbanization, health, growth of middle class, change in the occupational structure etc. The impact of all these factors were analyzed and examined in the literature vigorously but the maximum attention is given to the role of socio-economic development in terms of wealth. It seems that modern democracies have come to emerge in more economically developed countries. As also shown in Figure 2 below, where GDP per capita and democracy index follow the same upward trend, studies confirm that democratic transitions partly overlap with the average level of GDP per capita. By looking at this demonstration, it would be unrealistic to accept the view that democratic institutions can be easily established and maintained, from scratch, in anywhere, at any time (Inglehart, & Welzel, 2009).
6
The discussion of these structural and societal conditions rooted in the society, which ensure democracy to emerge, survive and stabilize, are considered as the main factors that are conducive to democracy. Income, education, wealth and other factors were deeply analyzed by Seymour Martin Lipset in his paper “Some Social Requisites of Democracy: Economic Development and Political Legitimacy” where he built up his well-known modernization theory. The core idea of the theory is the necessity of a continuous, effective economic development and a strengthened middle class to form a democratic regime and ensure its survival. As explained later in detail, the theory relies heavily on changing societal conditions to foster democratization. Thereby, once countries become wealthier, more literate, better educated and had developed substantial middle classes, these elements create a self-reinforcing process that transforms both political institutions and social life and eases the democratic political institutions to emerge (Inglehart, & Welzel, 2009).
Moreover, studies that associate modernization to democracy also touch upon the growing importance of their impact on civic participation. For some scholars, it is believed that these “requisites” for democracy does not only shape the institutions, but also civic attitude. In 2006, Glaeser, in his paper “Why does Democracy Needs Education”, takes the modernization argument one step further and links the effectiveness of education in the context of democracy to its positive impact on civic participation. He argues that schooling teaches people to interact with others and raises the benefits of civic participation, voting and organizing political events (Glaeser et al, 2006). Therefore, it’s been argued that by increasing the likelihood of participating in political events for individuals, education increases the likelihood of democratic actions that could possibly be performed within the country.
Yet, there remains one critical issue in this research area. Regardless of the aim of the studies, whether supporting modernization theory or not, the problem lies in the fact that majority of the researches lack from a decent measurement of democracy. How to measure “democracy” remains a contentious issue for quantitative studies. More precisely, the existing literature suffers from narrowness of the measurements. Although studies tend to use different democracy indicators, the core definition seems to remain neither different nor challenged by alternative methods. Thus, when it comes to combine the theory with a quantitative study, the choice of democracy measures seems critical.
7
linkage. For the latter, a mediation analysis is specifically conducted where political participation takes the form of the potential mediator.
8
2- Theoretical Framework
Figure 3: Conceptual Framework
The conceptual framework, visualized in Figure 3 above, is essential and construct the hearth of the paper, therefore, the reader is encouraged to consult in order to follow up the discussions and analyses that had been conducted throughout the paper.
2.1- As Socio- Economic Drivers: Education and Inequality
In order to discuss how education and inequality might affect democracy, the role of education and inequality in the context of socio- economic development is briefly identified below.
9
social outcomes. OECD suggests that education may possibly contribute to create a fair society. As a result, the notion of social cohesion, social inclusion, social equality and social mobility are increasingly being addressed in the discussion of education (Jackson et al, 2007).
On the other hand, inequality is perceived as an unfavorable element for growth and development. It can be a signal of lack income mobility or a reflection of persistent divergence rooted in the particular segments of the society (Norris et al, 2015). OECD studies show that the widespread increase in income inequality raises concerns over its potential impact on societies and economies since it tends to slow down all kinds of productivity. To speak with numbers, the study argues that rising inequality by 3 Gini points, would downgrade economic growth by 0.35 percentage point per year for 25 years; means a cumulated loss in GDP at the end of the period of 8.5 per cent (OECD, 2014). Its impact on social background also reveals that huge gap in privileges undermines various public opportunities for disadvantaged individuals, lowers social mobility, connection, hampers skills development and intellectual life, all the conditions necessary to form effective democratic institutions. Therefore, considering the fact that both education and inequality are critical in determining countries’ political, social, economic and civic institutions, it is plausible to use them as possible determinants for democracy.
2.2- The Great Debate: Democracy
10
should hold to be considered as democratic? Still, among the many functions of a representative democracy, one seems particularly well accepted by the majority; the ability to set a desirable environment which encourages citizens’ right to vote. That is to say, a democratic government should be responsible in holding free, fair and competitive elections. It is a highly important factor as countries where elections do occur but either involved in fraud, scam or monopolized can also be considered as non-democratic (Papaioannou, & Siourounis, 2008).
For policy makers, academicians and researchers, conceptualizing and measuring democracy matter significantly in order to conduct an academic analysis or a comparative research. The challenge they mainly encounter is the fact that abstract concepts, such as democracy, are relatively hard to materialize for quantitative analysis. In this sense, when the definitional consensus gets volatile and interchangeable, accepting a single, universal measure of democracy becomes harder. As a result, the validity of the measurements and their usage to obtain certain outcomes are subject to critics. To elaborate the issue further, consider the below graph.
Graph 1 : Comparison of two democracy indices Data Source: Seva Gunitsky, 2015
11
regarding different perceptions of democracy. Therefore, measurements are subject to changes on how one understands it. Polity IV cares about the constraints on the president and characteristics of governing institutions, therefore when Russian Duma rejected Yeltsin’s nomination in 1998, such rejection is signaled as legislative independence and led to a significant increase in the country’s policy score. Freedom House score on the other hand is more focused on individual rights and freedom of action, therefore, according to this measure, Russia in 1990s, where the death penalties still existed, was not a desirable environment and entitled with a lower score (Gunistsky, 2015)
To give another example, as Seva Gunitsy points out, mandatory voting is also subject to critique. The democracy measure prepared by The Economist (EIU) considers mandatory voting as bad for democracy as it tends to infringe on individual rights and somehow forces its citizens’ to behave in a certain way whereas another democracy measure, The Vanhanen, is based on measuring the percentage of population that votes in elections. Therefore, countries with mandatory voting system tend to rank lower on The Economist index and higher on Vanhanen index. (Gunistsky, 2015)
Above examples clearly show that, when it comes to study a causal inference, the choice of democracy measurement matters. All these disagreements that suggest different outcomes and interpretations actually reflect the fundamental trade-off how one perceives democracy. It should not come as a surprise that different democracy indicators might reveal different sides of the story. Therefore, Pemstein argues that due to the possible wide divergences between democracy measures, studies that use only one index are more likely to encounter problems regarding robustness of the results ( Pemstein et al 2010)
The possible perceptions regarding the characteristic of a democracy in this thesis will be further discussed in the methodology section where the indicators are explained. At this point, the reader should keep in mind that following debates and analysis are constructed on already existed concepts and measurements but not aimed to impose a particular, specific definition. The main attempt however is to analyze the social content of democracy and its underlying socio- economic elements.
2.3- Education and Democracy
12
(Jefferson, 1779. cited from Murtin& Wacziarg 2014). Jefferson’s view is explained and extended by other scholars throughout the history. However, it has been criticized severely as well. Thereby, two broad approaches emerged to analyze the relationship between democracy and development. One approach advocates the need to build democracy and other executive branches to secure property rights and other institutional elements. By doing so, they suggest, already established political institutions give a rise in investment in human and physical capital levels, thus economic growth occurs and well maintained. Starting with Acemoglu (2001, 2002) and Hall & Jones (1999), it has been suggested that institutional improvements in an economy are the key drivers of political development, therefore the causality runs from economic development to democratization.
13
Figure 4 : Education and Democracy
It has been argued that in countries where educated people are dominant, differences and conflicts are more likely to be resolved through negotiation and voting rather than violent quarrels. Education is necessary for legal entities to operate effectively as well as for citizens to participate in politics and be informed about government’s possible exploitation (Glaeser et al, 2004). Educational attainment is the pillar that shapes an individual’s attitudes and behaviors, therefore it is also viewed as the pillar that leads certain collective socio-cultural changes in human actions. Schools are regarded as the main social institutions where humanitarian values are transmitted, therefore the effect of education had been seen as a universal liberalizing effect. By helping the diffusion of modern values such as equality, freedom, humanism and tolerance, education alters people’s sense of understanding the world. Such diffusion of thoughts lays the foundations of a democratic political system and eases social and national integration. A public opinion research was conducted to question people in different countries regarding to their approaches to democratic norms, minorities and one-party regimes. The study reveals that the most essential single element separating those that support democratic elements from those do not has been education (Lipset, 1994). It seems that educated people are more likely to promote fairness, diversity and peaceful thoughts.
14
The study conducted by Elias Papaioannou and Gregorios Siourounis also reveals the fact that political reforms and democratization processes are more likely to occur in more educated countries. The study includes political freedom indicators and electoral archives that are used to identify democratic transitions and reverse transitions and classify countries as democratic, non-democratic or autocracies. The results show that as education generates the necessary network and communication to support the transition, it is one of the key factors that determine the intensity of economic reforms and the speed of the democratic transitions. The impact of education remains significant even when they controlled the other determinants of democracy such as religion, fractionalization, natural resources, openness, and early institutions (Papaioannou, & Siourounis, 2008). The results demonstrate that, based on the education levels in 1975, among the non-democratic countries with more than four years of schooling, all, except Singapore, moved to democratic government. In contrast, among the fourteen non-democratic countries with less than one year of schooling, only three (Mali, Benin and Mozambique) were able to implement democratic reforms. Thus, it becomes obvious that education was a key driving force behind the 3rd Wave (Papaioannou, & Siourounis, 2008).
In sum, education appears as a factor on both sides of democracy, implying that it is both an initial condition of democracy and once established, it sustains the democratic political systems because it increases the overall mark of efficiency embedded in the total system. Still, the stability of a given democratic system is not only based on a system’s efficiency but also its effectiveness. Thus, socio- economic development is a necessary but not a sufficient condition for the establishment and / or maintenance of a democratic institution (Lipset, 1994). Consequently, regarding the discussion above, the first hypothesis is constructed;
Hypothesis 1a: Educational attainment tends to increase the likelihood of having and maintaining democracy.
2.4- Inequality and Democracy
15
consequences on the economic, political, and social stability of a country. Democracies are found to be less stable and lasting when inequality increases and labor receives a lower share within the economy (Wucherpfennig & Deutsch, 2009). The premise that income inequality has negative impact on a country’s level of democracy grounds in the statement that high inequality tends to generate extreme opposing and inevitable class conflicts that is incompatible with stable democracy (Muller, 1995). Even though conflicts are necessary to generate diversity of thoughts, income inequality separates people itself, not the thoughts and it becomes harder to be resolved through disputes.
There are two general approaches regarding the impact of inequality on political democracy. One involves the question of genesis whereas the other considers the question of stability (Muller, 1988) The genesis hypothesis is related with the likelihood of the emergence of democratic political institutions given a relatively egalitarian society. Studies show that an egalitarian income distribution reflects the rise of a strong and autonomous bourgeoisie class which attempts to establish parliamentary democracy in order to expand its economic power into political power. As a result, the level of democracy in a country follows an inverse pattern. Countries with low income inequality are more likely to possess a strong middle class and establish democracy faster than those with high income inequality where the middle class is incapable and powerless to conduct such action (Rubinson, & Quinlan, 1977). A strong middle class ensures that the power is not dominated amongst the elites, but distributed equally between classes. Stability hypothesis,
on the other hand, mostly focuses on the role of income inequality on the likelihood of maintaining the regime given the democratic system that has already been established. Edward Muller’s study shows that, considering the economically developed European states, most pre-World War II breakdowns of democracies occurred in countries with inegalitarian land tenure systems- ‘”lord-peasant”- where the powerful landlords were opposed to subordinate classes to redistribute property and income through electoral process. On the contrary, other states with more egalitarian- “family-farm”- land tenure systems were able to develop more stable democracies (Muller, 1995). In his 1988 paper, Muller found a strong association between income inequality and democratic stability. His results indicate that, out of 33 democracies, 8 of the 10 with a top 20 percent of income share above 50 percent experienced a collapse of democracy, whereas only 1 out of 23 democracies with a top 20 percent of income share below 50 percent remained unstable. In fact, given high levels of income inequality, the mortality of democracy is more likely to occur, almost 80 percent, whereas it was only 4 percent given low income inequality (Muller, 1995).
Thus, in the light of above discussion a complementing hypothesis is constructed;
16 2.5- Importance of Civic Participation
As Lipset argues, for a stable democracy, existence of a strong civil society is as necessary as having electoral rules. If citizens do not belong to a certain group they become atomized and easily dominated by the controllers of the central power ( Lipset, 1994). The reason behind this argument is simple; certain democratic ideologies and values are evolved through conflicts among individuals or groups in societies. People from different religious, economic or professional backgrounds compete with each other and with state in an attempt to attain popularity, power or to carry out their own agendas. Therefore, legitimizing such group conflicts by allowing the rights of other groups to oppose them provides the basis for democracy. It is stated that in autocracies or totalitarian regimes such civil society does not exist. The system does its best to eliminate conflicts between individual and the state to control its sovereignty and leaves no room for competition (Lipset, 1994). As a consequence, a fully operative civil society or in other words, a participant society seems as a fundamental aspect sustain democracy. 2.6- Education and Civic Participation
This thesis takes Glaeser et al 2006 paper as a groundwork to construct the second hypothesis. In his paper, “Why Does Democracy Need Education” Glaeser also finds consistent evidence to support Lipset’s hypothesis. What differentiates his study from others is its theoretical reasoning. The basic assumption in the paper is that education raises the likelihood of having democratic institutions because it leads higher participation in whole range of social activities, especially in politics (Glaeser et al, 2006). Therefore, civic participation is cited as a significant element in explaining the linkage between education and democracy.
17
to communicate successfully, the likelihood to control their inherent anti-social tendencies increases and be more productive participants (Glaeser et al, 2006). With the increased education and hereby increased capabilities and skills, more educated people tend to act as civic leaders just as they are more likely to earn more profit due to higher private returns from political activities (Glaeser et al, 2006).
On the other hand, some studies show that this is not always the case. As pointed out by Richard Brody in his 1998 paper, political participation had failed to increase with rising levels of educational attainment in the United States. He pointed out this phenomenon as the “puzzle of participation in America”. Therefore, if the conventional approach is correct, then USA should experience rising level of political participation as a result of rising education levels. (Brody, 1978 cited from Berinzky & Lenz, 2010). The fact that participation levels have not kept pace with education gains challenges the conventional wisdom. The reason behind is attributed to the unobserved characteristics of individual attainments. Lenz and Berinsky (2011) together with Kam and Palmer (2008) mainly argue that studies that analyze the impact of education of civic participation do not take into account that people who attain higher levels of education differ from those who do not in various unobserved ways. Especially, given the fact that rising participation levels might be actually due to pre-existing characteristics of an individual. Therefore, education should be considered as a proxy for unobserved pre-adult experiences, early life socialization and experiences rather than a cause of political participation.
Similar with Kam and Palmer, Tenn’s empirical study also supports this statement. By using the panel data of Current Population Survey to compare two groups of respondents who are mainly the same except their levels of education, he investigates the impact of one year additional schooling on political participation. This differentiation enables him to measure the marginal impact of education and apparently the results do not seem support the traditional view of the effect of education on propensity to turn out. Therefore, other factors such as family background, culture and traditions rooted in the society and / or individual characteristics, not education per se, seem to lead to increased participation (Tenn, 2007).
18 2.7- Inequality and Civic Participation
As in line with democracy, conventional studies imply that the relationship between inequality and the degree of political participation should also follow an inverse one. The reason is simple: a lower rank in the social class hierarchy reflects an individual’s relative lack of perceived social and economic worth in society (Kraus & Callaghan, 2015). Therefore, due to the consequences of unequal distribution of power, inequality should have a negative effect on political engagement. Most studies found that despite the possibility of gaining so much from political participation, citizens at the bottom of class hierarchy are less likely to participate in politics compared to their relatively upper- class counterparts. It is clear that high inequality rooted in societies tend to limit opportunities for poor, thus at the end, limit the participation. In societies with low levels of differentiation, more people experience that they are more included to the center of society whereas higher inequality tends to make majority of people feel excluded (Wilsonson 2006). Frederick Solt conducted a study to examine the extent of economic inequality in a country on the impact of political engagement of its citizens. He found that greater inequality increases the relative power of the wealthy class to shape the politics in their own favor rather than considering citizens’ objectives or interests. His analysis shows that higher levels of income inequality significantly depress political interest, frequency of political discussion and participation in elections among all but the most affluent citizens. Thus, he concludes that greater inequality yield greater political inequality (Solt, 2008)
On the other hand, as we’ve had in education, there exist some studies that argue that higher inequality does not necessarily imply low level of political participation. Starting with Jones, (2005), some studies found the effect of inequality is rather vague, very small or even showed no impact. This view is explained by using the so called model SES (Socioeconomic Status Model). The model claims that a citizen’s social and political status can be predicted by the education, income or the occupation he/she possess an in return, individuals with higher socio-economic status tend to participate higher in politics. However, according to Persson, this does not imply or give any prediction on the effects of contextual elements such as the level of inequality. He argues that factors such as where one lives or the societal characteristics is considered to be less important than the individual characteristics. Therefore, in the context of political participation, individual characteristics, again, seem to be a more dominant factor in leading participation, regardless of the level of inequality in society (Persson, 2010).
Therefore, in the lights of the discussions above, the second hypothesis is constructed.
19
3.
Data and Methodology
3.1- Measurement Instruments
3.1.1- Dependent Variables
First, quantifying the political indicators is necessary to conduct an econometric analysis. Researches show that there are two general approaches exist to develop quantitative measures of democracy. While first approach considers the characteristic of democracy, the second approach is mostly related to the characteristics of the electoral process in a country (McCulloch, 2014) As discussed earlier, the challenge is that while some political events can be represented in discrete nature such as coups or overthrowns, some concepts are more difficult to quantify. Again, following the studies, it seems better to choose an arbitrary one dimensional indicator. However, for the robustness of results, rather than focusing on one democracy indicator as in most studies, two different indicators and / or measures of democracy, namely Polity IV and Perception of Electoral Integrity (PEI) are used; the former represents the characteristics of the institutions whereas the latter represents the quality of democracy within the country.
Our first measure is a conventional one, due to its comprehensiveness; the combined Polity IV dataset. This dataset is widely used among scholars as it is considered one of the broadest dataset that covers democratic elements. It is constructed by political scientists Jaggers and Marshall in an attempt to measure the limits of executive power. It is indicated that the democracy measures in Polity IV reflects the conditions that are essential and interdependent for democracy. These elements include “The presence of institutions and procedures through which citizens can express effective preference about alternative policies and leaders”, “the existence of institutional constraints of the exercise of power by the executive” and “the guarantee of civil liberties to all citizens in their daily lives and in acts of political participation” (Glaeser, 2004) Therefore, it captures the key characteristics of executive recruitment, constraints on decision-making authority and the degree of political competition (Jaggers & Marshall 2002)
The unit of analysis in the dataset is called ‘polity’ which is defined as a political or governmental organization; a society or institution with an organized government; state; body politic.” (Jaggers & Marshall 2002) Gurr and Eckstein who initially worked on the project, provide a simple and broad definition of all polities as subsets of the class of authority patterns. They state that authority patterns are equivalent of state organizations and are defined as set of relations among hierarchically ordered members of social unit that involves the direction of the unit (Jaggers & Marshall 2002)
20
1800 – 2015 and are used to obtain our main indicator, polity2, which is derived by subtracting the autocracy value from the democracy value. This value gives a single regime score that ranges from + 10 (full democracy) to - 10 ( full autocracy) with higher values indicating higher degrees of political freedom.
One critical issue needs to be addressed regarding the dataset. Marshall and Jaggers indicate that during central authority interruption, collapse or transition periods, democracy and autocracy scores are assigned as Standardized Authority Codes; such as -66, -77 and -88 and converted into conventional polity scores ( -10 to +10 ) accordingly. The rules in converting values are given as follows; the score -88 represents cases of transition periods and prorated across the span of transition. -77 represents cases of interregnum or anarchy and converted to a neutral polity score of 0 (zero). Lastly, -66 represents cases of foreign interruption and treated as ‘system missing’ with no values added. As mentioned earlier, Polity IV defines democracy as the presence of institutions and procedures through which citizens can express effective preference about alternative policies. In the case of foreign interruption, it is clear that this mechanism is no longer effective, so democracy does not exist as well. Therefore, rather than treating the values as system missing, the missing values are assigned as -10 (full autocracy) in an attempt to achieve more accurate results.
It is believed that Polity IV indicator is minimal in the sense that it does not represent the quality of the established democracies. It only focuses whether there exists a democratic system or not. Therefore, to assess the quality of democratic political institutions and its operational success, a second measure is used; called PEI, which includes one important element, elections. It covers a global expert survey called Perceptions of Electoral Integrity (PEI) which is conducted by Pippa Norris and Max Grömping. The aim of the survey is to evaluate electoral integrity which is defined as the necessary international and global standards and norms to conduct appropriate elections. It is stated that these standards are endorsed through conventions, protocols and guidelines by the international community(Pippa, & Grömping, 2017) As discussed earlier, the way the elections are conducted reflects a country’s political success. Therefore, compared to Polity IV which focuses on whether democracy exists in law or not, PEI truly reflects how democracy operates in real life as it captures the behaviors of individuals who actually give meaning to it.
21
higher understanding of the electoral process are expected to be more faithful and confident as they are familiar with electoral processes and rules, whereas individuals with lack of knowledge and weak awareness are more likely to be doubtful and suspicious about fraud or any kind of cheating. Therefore, they argue that these attitudes are mainly shaped and strengthened by education, experience and strength of civic participation (awareness) (Pippa, & Grömping, 2017).
To operationalize this view, they evaluated elections using 49 indicators which are grouped into eleven subcategories, each reflecting the quality of the electoral process. Then, they summarized all 49 indicators to form a 100- point PEI Index. Respondents are included in the survey one month after the date of a national election conducted in their country. The final release of the dataset, PEI 5.0, covers the assessment of 2,709 individual experts, evaluating 241 elections in 158 countries during the period between 2012 - 2016. PEI Index is generated at the individual level and its imputed version is used for this analysis as it is observed for all experts and states (Pippa, & Grömping, 2017). The index also covers indicators that reflect political participation such as voter registration, party registration, media coverage or level of campaign finance. However for avoiding any possible measurement bias, an independent dataset is used to measure civic participation.
3.1.2- Independent Variables
The dataset regarding education contains the well-known Barro & Lee average years of schooling data. (Barro & Lee 2001). It is indicated that the estimates on educational attainment can be used as a plausible proxy to capture the stock of human capital for wide range of countries (Barro & Lee 2001). The dataset estimates educational attainment for 146 countries in 5 year intervals from 1950 to 2010. Average years of schooling at primary, secondary, and tertiary levels are also measured for each country in the world. To make the data consistent with other indicators in terms of time coverage, Barro and Lee’s projections on educational attainment for years 2015-2040 is used as well. The main indicator is chosen as average years of total schooling. By using backward and forward interpolation technique, the data is extended for years between 1950 to 2016.
22
representing perfect equality and 1 representing perfect inequality. The data reflects the Gini coefficient at each percentile on the population, that is to say, 100 different Gini coefficients are given at lowest 1 percentile until top 100th percentile. Since within country comparison of Gini coefficients is not the purpose of this study, the average of all hundred coefficients (or percentiles) are taken to get one specific Gini value for a country for a specific year (Solt, 2016) 3.1.3- Mediating Variable
To capture political participation, data extracted from the International Institute for Democracy and Electoral Assistance‘s (IDEA) “Voter Turnout” indicator is used. It is one of the most comprehensive dataset that covers voter turnout statistics from both presidential and parliamentary elections that had been occurring all over the world, since 1945. In the dataset, the number of Registered Voters (REG) is used as an indicator of political participation which indicates the actual number of people on the voters’ roll. Voter turnout statistics are obtained by dividing total number of votes to registration. Registration is considered as the number of people who were registered for elections and recorded by the national electoral management body, regardless of countries’ specific registration requirements. To conduct a comprehensive analysis, only countries that experience parliamentary elections are selected.
3.1.4- Control Variables
There are several control variables that could have been taken into account. In this study, three control variables were included to investigate the relationship between democracy and development. All control variables are extracted from the World Bank’s World Development Indicators databank.
23
Land Area: In general, studies show that democracy has little relation with the size of the country. Yet, some researchers found that large countries are less likely to be democratic compared to smaller ones. In small cities, the notion or the spirit of public strength is more strongly felt by each citizen (Montesquieu, cited in Dahl and Tufte, 1973) Small cities possess small units which are more intimate and this intimacy enables an effective and reciprocal communication between political leaders and citizens’. Thus, in small countries, political leaders are easily conduct direct observations and communication regarding their citizens’ need (Dahl & Tufte, 1973). For this reason, smaller countries are associated with high levels of democracy. Thus, land area which is measured by sq km is expected to affect democracy negatively as it gets larger.
Trade: Last but not least, to capture the impact of globalization on democracy, trade as a percentage of GDP is used. Following Dani Rodrik’s well known book “ The Globalization Paradox”, globalization can be considered as an influencing element to democracy. The main idea of the book is that there exist a trilemma between globalization, the nation state and democratic politics and only two of the three can be sustained ( Rodrik, 2011) The balance is especially critical for democracy and globalization. Since different societies have different needs, preferences and institutions in terms of ensuring their markets function efficiently, democratic pressures tend to form various institutions across different territories. Hence, the wide range of diversity among nation states constraints the global integration as it raises transaction costs across jurisdictions. As a result, a country which is completely bounded to its democratic preferences cannot be fully globalized (Rodrik, 2011). Following this argument, one should expect that as the level of trade increases within the country, it negatively affects the political structure.
3.2- Imputation of Missing Data
Both Gini and voter turnout data are highly unbalanced and contain lots of missing observations. Yet, due to the limited number of alternative measurements, they are the most broad datasets. Therefore, to construct a balanced sample, the STATA command ( ipolate & epolate) is used to interpolate missing observations by using the backward and forward interpolation technique. To avoid misleading results and for robustness check, 2 separate analysis are conducted ; one with the original data and other with interpolated observations.
24
Figure 5 : Original Data vs Linear Interpolation
3.3-
Methodology & Research ModelRegarding the first hypotheses; the following multiple OLS regression analyses are conducted. Due to data availability, the full sample is composed of 120 countries except some big economies such as France, Korea, China and Saudi Arabia. The full list of countries is provided in the Appendix -1. For the regression analysis which includes Polity IV indicator as the dependent variable, time coverage is selected as 1980 - 2015 and for the latter, that includes PEI as the dependent variable, time coverage is selected as 2012-2016. The base models are shown below;
(1)
(2)
Mediation Analysis
25
mediate the relationship through its indirect effect and the strength of this indirect effect reflects the mediated ratio of the relationship between dependent and independent variable caused by the mediator (Baron & Kenny, 1986). Here in this paper, as discussed earlier, political participation captured by voter turnout ratio is treated as the potential mediator.
To test the mediation effect, the recommendations provided by Baron and Kenny is followed, where 4 regression equations, including the control variables, were estimated;
1- The effect of the independent variable on the dependent variable ( estimation of path c)
2- The effect of the independent variable on the mediator ( estimation of path a)
3- The effect of the mediator on the dependent variable while controlling the independent variable ( estimation of path b)
4- The effect of the independent variable on the dependent variable while controlling the mediator (estimation of path c’)
Paths a, b, c and c’ are represented in Figure 3.
In path analysis, there are commonly three types of effects: total effect, direct effect, and indirect effect. Path c refers to a simple relationship between independent and dependent variable, which is usually referred to as the total effect of independent on dependent variable. Paths a and b reflect the indirect effect whereas path c’ refers to the direct effect. As a result;
Total Effect ( c ) = Direct Effect ( c’ ) + Indirect Effect ( ab )
26
Baron and Kenny recommended to compute Sobel test to identify the significance of the indirect effect ( a x b) . To do so, below formula is used; where a is the regression coefficient for the relationship between the independent variable and the mediator, b is the regression coefficient for the relationship between the mediator and the dependent variable, SEa is the standard error of the relationship between the independent variable and the mediator, and SEb is the standard error of the relationship between the mediator variable and the dependent variable (Sobel, 1982)
Before moving on to the results, an additional concern regarding the issue of endogeneity for the 2nd hypothesis must be addressed. It seems likely that democracy indicators could also drive or have an impact on voter turnout ratios. To deal with this, a large body of research suggests to use lagged explanatory variables. However, Bellemare and Pepinsky’s 2015 paper stands as a critique to this traditional approach. They argue that using lagged identification, meaning lagging independent variables, to solve endogeneity problems, is nothing but an illusion. By using lag identification, one introduces a bias to the estimates because, lagging independent variables simply moves the channel through which endogeneity biases causal estimates, replacing a “selection on observables assumption” with an equally untestable “no dynamics among unobservable” assumption (Bellemare & Pepinsky, 2015). Therefore, in this thesis, lagged explanatory variables are not included in the analysis to avoid any incorrect inferences.
4. Results
4.1- Emprical Results
The data is prepared for panel data regression commands by generating the country variable and declaring the panel structure of the dataset ( by using the xtset command). Then, for the 1st group (which includes Polity IV), to differentiate between random effects model and fixed effects model, Hausman test was conducted. Since the p-value is smaller than 5% , country & year fixed effect estimators were included in the analysis. Further, White Test and Wooldridge test has been conducted to detect potential heteroscedasticity and autocorrelation. The tests are significant at 99 % level of significance, therefore to control heteroscedasticity and autocorrelation, all regressions are run using robust standard errors.
27
4.2- Data Description
4.2.1- Descriptive Statistics
The summary statistics of the variables included in this study are shown in Table 1 and Table 2 and provided in the Appendix-2. Since the time coverage of Polity IV data is longer than PEI, number of observations is higher accordingly. As it can be seen, both for Polity IV and PEI, standard deviations are high and usually different from each other due to the unbalanced nature of the dataset. As discussed earlier, Polity IV score ranges within -10 and 10, whereas PEI scores are represented as percentages. Looking at the mean values, it seems that most countries can be considered as close to non-democratic between the years 1980- 2015 as the mean value of the Polity IV score is around 3 whereas considering PEI, the level of satisfaction of citizens to the electoral process, thereby the way democracy works in their country, is around 57 %. Average years of schooling has its mean value around 8 in both datasets which indicates that the countries are almost well educated.
4.2.2- Correlation Matrix
Tables 3 and 4 show the correlation matrices and provided in the Appendix-3. Both include the dependent, independent, control and mediator variables used in the entire analysis. From the results, it can be seen that most of the signs of correlations are as expected, except trade and land area. Despite of the fact that correlation does not mean causality, if one specifically focus on Gini, average years of schooling and democracy indicators, it can be said that while education has a positive correlation with the democracy indicators, inequality has negative correlation as the theories suggest. For PEI, both the correlation ratios of education and inequality are very high, almost 55 % indicates strong correlation whereas for Polity IV, correlation coefficient of Ginis remain significantly low. Also, urban population seems to have strong, positive correlations with the democracy indices. The highest correlation ratio in the table is around 60 % therefore we can say that our models do not likely to suffer from multicollinearity, however an additional test was conducted to detect potential multicollinearity.
4.2.3- Check for Multicollinearity:
28 0 .0 2 .0 4 .0 6 D e n si ty -30 -20 -10 0 10 20 Residuals Kernel density estimate Normal density
kernel = epanechnikov, bandwidth = 2.3387
Kernel density estimate
most studies consider values between 3-5 more appropriate. As one can see in Table 5, all of the received results are less than 3 so it can be concluded that multicollinearity is not an issue among the explanatory variables used in this thesis.
4.2.4- Check for Normality:
Finally, to check for normality, both the Kernel density estimate and Shapiro-Wilk W are conducted and results are provided below. Even though a high value of W indicates normality, p value rejects the null hypothesis that states residuals are normally distributed. However, as the sample size is sufficiently large, non-normality is not an important issue.
Table 6 : Normality Test
Shapiro-Wilk W test for normal data Variable Obs. W V z Prob
29 4.3- Regression Results
Table 7 reports the results of the first hypotheses (1a & 1b) testing for education and inequality. For both Polity IV and PEI, the results indicate a significant relationship for education. Average years of schooling is positive and statistically significant both for Polity IV and PEI but shows different values for each, again as expected. Considering PEI, the coefficient is higher, indicating that an additional increase in education level is associated with approximately 2.45 units of increase in the public faith in the electoral process. As discussed earlier, an increase in integrity represents the legitimacy of the system. Therefore, one can conclude that education strengthen the perceptions and awareness of how elections work within the country. This effect will in turn, generates more trust in elections and eventually lead more democratic practices to emerge.
Considering Polity IV, the correlation coefficient of education is also significant and positive, indicating that one year additional schooling causes a 0.45 unit increase in a country’s democracy score. Although the overall conclusion is the same regarding the impact of education on democracy, comparing the strength of the coefficients and R squares, one can see that the explanatory power is much higher for PEI. This points out the fact that education is more effective and powerful in operationalizing democratic institutions rather than initiating a democratic law that would upgrade the Polity Iv score.
Turning to inequality, results become more interesting. For Polity IV, the coefficient is insignificant, means that it has no impact in explaining Polity IV scores. Inequality seems non related with the emergence of political institutions. On the other hand, for PEI, inequality is significant and has a negative coefficient, as expected, referring that one unit increase in Gini score leads almost 0.39 units of decrease in citizens’ faith in electoral process. The fact that inequality contributes the dissatisfaction of democratic process comes as no surprise. After all, PEI scores are constructed on individual perspectives and their actions regarding the system. Therefore, a damage in the system, which is represented by increasing inequality, will eventually result in lower trust, thus less effective democratic political institutions.
30
population has a positive relationship with both indicators, indicating that the level of urbanization within the country fosters both the maintenance and the operational structure of the democratic regimes.
Model 2 outlines the same analysis with interpolated Gini coefficients as a robustness check. All of the results are consistent with the estimates of Model 1. Although the coefficients indicate small changes among variables, the overall conclusion is the same. Looking at both Model 1 and Model 2, it can be seen that R square is higher for PEI which indicates the explanatory power of the variables in explaining PEI scores are stronger.
31 Significance levels: *** p<0.01, ** p<0.05, * p<0.1
For columns 2 and 4, robust standard errors are shown in parenthesis. For columns 3 and 5 standard errors are shown in parenthesis.
32
Tables 8 to 11 report the results of the second hypothesis testing for the mediation effect and provided in the Appendix-4. Looking at the tables, for both groups, the mediation effect is insignificant after controlling the independent and the control variables. As previously mentioned, to arrive the decision that mediation exists, coefficients included in steps 2 and 3 must be significant. However, in both analysis, the indirect effect a & b ,especially path a, remains statistically insignificant. Additionally, the p values of the indirect effects (a x b) for every analysis, that are obtained by the Sobel Test for the significance of meditation, reflects a value larger than 0.1. Given the fact that both step 2 and 3 fails to succeed, the estimated results show that the relationship between education and democracy does not seem to be mediated by political participation.
One additional point also needs to be addressed. Comparing the tables 8 to 10, the correlation coefficients are again different and stronger in PEI when adding the variable voter turnout. The outcome once more reflects the differences between two democracy indicators. To clear that issue further, the correlation between PEI and Polity IV are also presented below.
4.4- Correlation Between Two Democracy Indicators
33
5. Conclusion and Limitations
There is a broad literature that covers the relationship between economic development and democracy. In this thesis, Lipset’s modernization hypothesis that covers “development first, democracy later” approach is first discussed and analyzed. According to the theory, education and inequality are the key driving forces for having and maintaining democratic institutions as they change the societal conditions to form an effective environment for democracy. Therefore, the view that increasing education level and low inequality ratios seem to foster democracy construct the first hypothesis. In the meantime, while discussing their effects, Glaeser’s 2006 paper provides some theoretical support to explain the relationship. As also suggested by Glaeser, democracy is likely to be supported and maintained in countries where education level is high. The reason behind is supported by the role of education in promoting civic participation. Education encourages citizens to involve in democracy by raising the benefits of civic participation, thus, increases the likelihood of democracy. This view, which is also applied to inequality, constructed the second hypothesis which states that the relationship between development and democracy is mediated by political participation. To test this hypothesis, Baron and Kenny’s mediation analysis is conducted; by treating education & inequality, democracy indicators and voter turnout as independent, dependent and mediator variables respectively. While analyzing the effects, there remains one important issue that needs to be considered. Studies that include quantitative analysis suffer from the narrowness of the measurements in terms of democracy. It seems that most studies tend to use similar democracy indices and avoid including any alternative measurements. As this approach might affect the robustness of the results and lead possible misinterpretations, two different democracy indicators are used; named Polity IV and PEI. One represents the conventional wisdom on democracy which focuses on the limits of executive power, whereas the other represents the daily life implications and citizens’ attitudes towards its implementation.
34
Therefore, it is suggested that studies that analyze such a relationship should additionally consult on alternative measurements and not limit themselves with conventional indicators.
35
6. Bibliography
Acemoglu, Daron, Simon Johnson, and James A. Robinson. 2001. “The Colonial Origins of Comparative Development: An Empirical Investigation,” American Economic Review 91(5): 1369-401.
Acemoglu, Daron, Simon Johnson, and James A. Robinson. 2002. “Reversal of Fortune: Geography and Development in the Making of the Modern World Income Distribution,” Quarterly Journal of Economics 117(4), 1231-1294
Acemoglu, D., Johnson, S., Robinson, J. A., & Yared, P. (2005). From education to democracy? (No. w11204). National Bureau of Economic Research.
Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable distinction in social psychological research: Conceptual, strategic and statistical considerations. Journal of Personality and Social Psychology, 51, 1173-1182
Barro, R. J., & Lee, J. W. (2001). International data on educational attainment: updates and implications. oxford Economic papers, 53(3), 541-563
Bellemare, M. F., Masaki, T., & Pepinsky, T. B. (2015). Lagged explanatory variables and the estimation of causal effects.
Berinsky, A. J., & Lenz, G. S. (2011). Education and political participation: Exploring the causal link. Political Behavior, 33(3), 357-373.
Brody, R. (1978). The puzzle of political participation in America. In A. King (Ed.), The new American political system. Washington, DC: American Enterprise Institute
Dabla-Norris, M. E., Kochhar, M. K., Suphaphiphat, M. N., Ricka, M. F., & Tsounta, E. (2015). Causes and consequences of income inequality: a global perspective. International Monetary Fund.
36
Dee, T. S. (2004). Are there civic returns to education?. Journal of Public Economics, 88(9), 1697-1720.
Dewey, J. (1916). Democracy and Education: An Introduction to Philisophy of Education. Macmillan
Dutt, P., & Mitra, D. (2008). Inequality and the Instability of Polity and Policy. The Economic Journal, 118(531), 1285-1314.
Glaeser, E. L., La Porta, R., Lopez-de-Silanes, F., & Shleifer, A. (2004). Do institutions cause growth?. Journal of economic Growth, 9(3), 271-303.
Glaeser, E., Ponzetto, G., & Shleifer, A. (2006). Why does democracy need education? (No. w12128). National Bureau of Economic Research
Glaeser, E. L., & Steinberg, B. M. (2017). Transforming cities: does urbanization promote democratic change?. Regional Studies, 51(1), 58-68.
Gunitsky, S. (23 june 2015). How do you measure ‘democracy’? How do you measure ‘democracy’? Retrieved June 10, 2017.
Hanushek, E. A., Woessmann, L., Jamison, E. A., & Jamison, D. T. (2008). Education and economic growth. Education Next, 8(2).
Högström, J. (2013). Does the choice of democracy measure matter? Comparisons between the two leading democracy indices, freedom house and polity IV. Government and Opposition, 48(02), 201-221.
Huntington, S. P. (1993). The third wave: Democratization in the late twentieth century (Vol. 4). University of Oklahoma press
Inglehart, R., & Welzel, C. (march 2009). Development and Democracy: What We Know about Modernization Today. Development and Democracy: What We Know about Modernization Today. Retrieved April 12, 2017.
37
Jefferson, T. (1779). A bill for the more general diffusion of knowledge. Jefferson: Writings, 365.
Kam, C.D. & Palmer, C.L. (2008). Reconsidering the effects of education on political participation. Journal of Politics 70: 612-37.
Kraus, M. W., Anderson, C., & Callaghan, B. (2015). The Inequality of Politics: Social Class Rank and Political Participation.
Lipset, S. M. (1994). The social requisites of democracy revisited: 1993 presidential address. American sociological review, 1-22
Lipset, S. M. (1994). The social requisites of democracy revisited: 1993 presidential address. American sociological review, 1-22.
Marshall, M. G., Jaggers, K., & Gurr, T. R. (2002). Polity IV project. Center for International Development and Conflict Management at the University of Maryland College Park.
Muller, E. N. (1988). Democracy, economic development, and income inequality. American Sociological Review, 50-68.
Muller, E. N. (1995). Income inequality and democratization: reply to Bollen and Jackman. American Sociological Review, 60(6), 990-996.
Murtin, F., & Wacziarg, R. (2014). The democratic transition. Journal of Economic Growth, 19(2), 141-181.
Norris, Pippa, and Max Grömping. 2017. Codebook – The expert survey of Perceptions of Electoral Integrity, Release 5.0, (PEI_5.0). The Electoral Integrity Project, University of Sydney OECD. (december 2014). Does income inequality hurt economic growth? Focus on Inequality and Growth.
Papaioannou, E., & Siourounis, G. (2008). Economic and social factors driving the third wave of democratization. Journal of comparative Economics, 36(3), 365-387.
38
Persson, M. The Effects of Economic and Educational Inequality on Political Participation. Max Roser (2016) – ‘Democracy’. Published online at OurWorldInData.org. Retrieved from: https://ourworldindata.org/democracy/
McCulloch, A. (2014, May 24). How to measure democracy. Retrieved June 10, 2017.
MacKinnon, D. P., Fairchild, A. J., & Fritz, M. S. (2007). Mediation analysis. Annual Review of Psychology, 58, 593-614.
Rodrik, D. (2011). The globalization paradox: democracy and the future of the world economy. WW Norton & Company.
Rubinson, R., & Quinlan, D. (1977). Democracy and social inequality: A reanalysis. American Sociological Review, 611-623.
Schumpeter, J. A. (1942). Capitalism, socialism and democracy. New York: Hamper Brother. Sobel, M. E. (1982). Asymptotic confidence intervals for indirect effects in structural equation models. Sociological Methodology, Vol. 13, pp. 290-312.
Solt, F. (2008). Economic inequality and democratic political engagement. American Journal of Political Science, 52(1), 48-60.
Solt, F. (2016). The standardized world income inequality database. Social science quarterly, 97(5), 1267-1281.
Tenn, S. (2007). The effect of education on voter turnout. Political Analysis, 15(4), 446-464. Zhao, X., Lynch, J. G., & Chen, Q. (2010). Reconsidering Baron and Kenny: Myths and truths about mediation analysis. Journal of consumer research, 37(2), 197-206.
Wilkinson, R. G. 2006. “The Impact of Inequality”, Social Research. 73 (2), 711-732
39
7. APPENDICES
Appendix 1 : List of Countries
Afghanistan Greece Nicaragua Vietnam Albania
Guatemala Niger Zambia Algeria
Guyana Norway Zimbabwe Argentina Haiti Pakistan Armenia Honduras Panama Australia Hungary Paraguay Austria India Peru Bahrain Indonesia Philippines Bangladesh Iran Poland Belgium Iraq Portugal Benin Ireland Romania Bolivia Israel Russia Botswana Italy Rwanda Brazil Jamaica Serbia Bulgaria
Japan Sierra Leone Burundi Jordan Singapore Cambodia Kazakhstan Slovakia Cameroon Kenya Slovenia Canada
Kuwait South Africa Central African Republic Kyrgyzstan Spain Chile
Laos Sri Lanka Colombia
Latvia Sudan Costa Rica Lesotho Swaziland Croatia Lithuania Sweden Cuba Malawi Switzerland Cyprus Malaysia Syria Czech Republic Mali Taiwan Denmark
Mauritania Tajikistan Dominican Republic Mauritius Tanzania Ecuador Mexico Thailand Egypt Moldova Togo El Salvador Mongolia Tunisia Estonia Morocco Turkey Fiji Mozambique Uganda Finland Myanmar Ukraine Gabon
Namibia United Kingdom Gambia
Nepal United States Germany
Netherlands Uruguay Ghana
40 Appendix 2 : Summary Statistics
Table 1 : Summary Statistics of Polity IV
Gini_i and vti represent the data obtained by linear interpolation Variables Number
of Obs.
Mean Std. Dev Min Max
41 Table 2 : Summary Statistics of PEI
Gini_i and vti represent the data obtained by linear interpolation Variable Number of
Obs.
Mean Std. Dev Min Max
42 Appendix 3 : Correlation Matrixes
Table 3 : Correlation Matrix of Polity IV Variables Polity
IV
Gini Gini_i Avs vti vt Urban
Population Land Area (sq km) Trade ( % of GDP) Polity IV 1.0000 Gini -0.198 1.0000 Gini_i -0.157 1.0000 1.0000 Avs 0.577 -0.485 -0.383 1.0000 vti 0.010 -0.140 -0.100 0.107 1.0000 vt 0.046 -0.185 -0.178 0.117 1.0000 1.0000 Urban Population 0.430 -0.305 -0.256 0.666 0.117 0.169 1.0000 Land Area (sq km) 0.093 0.009 -0.012 0.143 -0.109 -0.017 0.188 1.0000 Trade ( % of GDP) 0.051 -0.082 -0.053 0.274 0.106 0.147 0.184 -0.234 1.0000
43 Table 4 : Correlation Matrix of PEI
Variables PEI Gini Gini_i Avs vti vt Urban
Population Land Area (sq km) Trade ( % of GDP) PEI 1.0000 Gini -0.515 1.0000 Gini_i -0.281 1.0000 1.0000 Avs 0.592 -0.659 -0.347 1.0000 vti 0.153 -0.027 -0.155 0.099 1.0000 vt 0.157 -0.145 -0.101 0.114 1.0000 1.0000 Urban Population 0.560 -0.135 -0.127 0.623 0.134 0.141 1.0000 Land Area (sq km) 0.0004 0.027 -0.011 0.138 -0.015 0.014 0.180 1.0000 Trade ( % of GDP) 0.150 -0.285 -0.173 0.342 0.142 0.075 0.145 -0.283 1.0000
44
Appendix 4 : Regression Results for Mediation Analysis
Table 8 : Mediation Results
Significance levels: *** p<0.01, ** p<0.05, * p<0.1 Standard errors are shown in parenthesis.
