Master’s Thesis
Globalization and Income Inequality:
Could people’s attitudes towards
income equality affect this relation?
By
Pauline Franchimont
Supervisor: dr. G.J. de Vries
Co-assessor: dr. T. Kohl
University of Groningen
Faculty of Economics and Business
The Netherlands
Abstract
The relationship between globalization and income inequality is a topic of intense debate among academics and policy makers. This association might be influenced by citizens’ attitudes towards income equality, which are expressed in their voting behaviour for left-socialist political parties. This MSc thesis examined the moderating effect of attitudes on the relationship between globalization and income inequality empirically by using a panel data approach. The findings do not confirm that globalization is related to an increase in income inequality. The effect of attitudes moderates the association between globalization and income inequality, but it is not statistically significant.
3
Table of Contents
1. Introduction ... 4
2. Literature Review ... 6
2.1 Relationship between globalization and income inequality ... 6
2.1.1 Heckscher-Ohlin model and Stolper-Samuelson theorem ... 7
2.1.2 The offshoring model ... 9
2.1.3 The role of the informal sector ... 9
2.1.4 Other factors that are related to income inequality ... 10
2.2 Attitudes towards income equality ... 11
2.3 Hypotheses and Conceptual model ... 13
3. Data and Methods ... 15
3.1 Data ... 15
3.1.2 Independent variable ... 18
3.1.3 Moderator variable ... 18
3.1.4 Control variables ... 19
3.1.5 Descriptive statistics ... 20
3.1.6 Scatterplot of the relationship between Income inequality and Globalization ... 21
3.2 Methods... 22
3.2.1 Econometric model ... 22
4. Results and Discussion ... 24
4.1 Empirical results and Discussion ... 24
4.2 Model diagnostics ... 28
4.3 Robustness tests ... 30
5. Concluding Remarks ... 34
5.1 Conclusions ... 34
5.2 Limitations and Future Research ... 34
4
1. Introduction
The relation between globalization and income inequality is a topic of intense debate among academics and policy makers (Dean & Ritzer, 2012; Guillen, 2001; Ritzer & Atalay, 2010). The following three (theoretical) channels outline this ambiguous association. The first channel is the Heckscher-Ohlin (i.e., H-O) model that shows which industry in a country gains or loses from opening up to trade in the long-run (Feenstra & Taylor, 2014). When combining this framework with the Stolper-Samuelson theorem, it indicates that globalization in advanced nations is related to an increase in income inequality, while in developing countries it is related to a more even income distribution (Feenstra & Taylor, 2014; Harrigan, Reshef & Toubal, 2016; Atkinson & Bourguignon, 2014). Secondly, trade in intermediate inputs (i.e., offshoring) is associated with a rise in income inequality for both advanced and developed countries (Feenstra & Taylor, 2014; Harrigan et al., 2016). The third channel emphasizes the role of the informal sector in nations (Goldberg & Pavcnik, 2004). Due to more globalization, a bigger share of the workforce has a job in the informal sector because many (especially, low(er)-skilled) workers who previously worked in the formal sector have, e.g., been fired or their tasks have been outsourced to the informal sector. In this way, these workers became unemployed or they found work in the informal sector where they earn lower wages than in the formal sector. This contributes to a rise in income inequality because the gap between low(er)-skilled and high(er)-skilled employees widens (Goldberg & Pavcnik, 2004; Goldberg & Pavcnik, 2007).
Although theories suggest that the relation between globalization and income inequality is ambiguous, empirical studies tend to find that globalization is related to an increase in income inequality (Bourguignon, 2015; Piketty & Saez, 2014). At the moment, income inequality within nations is still rising and raising concerns about the increasing concentration of poverty pockets (Bourguignon & Morrisson, 2002). Many empirical studies have found that globalization is related to an increase in income inequality (Milanovic, 2003; Drennan, 2015; Bourguignon, 2015; Beugelsdijk, Brakman, Garretsen & Van Marrewijk, 2013; Piketty & Saez, 2014).
5 citizens of countries vote for these political parties, which in turn represent their ideas in the government. In this way, citizens enforce indirectly their opinions (i.e., the social context of a country) about, for example, the distribution of income (Milanovic, 2013). The fact that globalization takes place within a certain social context of a country is not yet considered by previous studies regarding income inequality and globalization. The relation between globalization and income inequality can be moderated by citizen’s attitudes towards income equality.
Therefore, the research of this MSc thesis concentrates on a conceptual model that analyses the association between globalization and income inequality where the attitudes of people towards income equality is the moderator variable. The proxy for this variable is the shares of votes (in %) for left-socialist political parties in a government/parliament. Individuals who vote for these political parties are typically in favour of a (more) equal income distribution (Brown-Iannuzzi, Lundberg & McKee, 2017). The left-socialist parties represent the views of these voters because they strive strongly (compared to other parties) for income redistribution (i.e., (more) income equality) (Schumpeter, 2013). Hence, it is relevant to study the incorporation of the attitudes of people towards income equality within a country and its effect on the association between income inequality and globalization.
To examine this relationship between income inequality within countries and globalization over time, while testing for the moderating effect of attitudes of individuals towards income equality, the following research question has been developed: Do citizen’s attitudes towards income
equality expressed in their voting behaviour for left-socialist political parties affect the relationship between income inequality and globalization?
6
2. Literature Review
This section outlines and discusses the relevant theory and studies about the relationship between globalization and income inequality, and attitudes towards income equality. Section 2.1 defines the concepts ‘Globalization’ and ‘Income inequality’ to give a good overview of how these concepts are placed in this MSc thesis. In addition, this section also describes the association between globalization and income inequality. In section 2.2, attitudes towards this uneven income distribution are outlined. The hypotheses and according conceptual model are shown in section 2.3.
2.1 Relationship between globalization and income inequality
Globalization can be described as the “growing interdependence of the world economy” (Weisman, 2015: 3). The current form of globalization started in the late 1970s-1980s. During these decades, many policies were established to promote the enlargement of markets, business and welfare. Due to this, investments and trade expanded (Yergin & Stanislaw, 2002). Especially, firms play an important role in this integration of the world economy because they operate, for example, in foreign direct investment (i.e., FDI) and international acquisitions. The rise and significant changes in technology have also caused that new and more efficient manners of production processes occur in the international market. This also resulted in new and more international competition. Although the rules of competition have been redefined for an interdependent world economy, nations and firms still adapt and learn when coping with this competition (Beugelsdijk et al., 2013). When looking at the consequences of globalization, it can be divided into positive and negative outcomes. On the one hand, it resulted in factor mobility, knowledge spillovers and increased welfare. On the other hand, globalization has also been criticised because it harms the environment, it drives firms to actions that do not always comply well with human and worker rights and it increases poverty and income inequality within countries (Elliot, Kar & Richardson, 2004). Besides, it is also important to mention that there are still big differences between countries regarding the amount of goods and services they export and import (i.e., trade).
7 Gini coefficient is one of the most widely used estimates to show this gap and it is indicated with a number between 0 (= complete equality) and 1 (= complete inequality) (OECD, 2017). This coefficient can also be explained graphically by using the Lorenz Curve. This curve shows the cumulative distribution of income against the cumulative percentage of a population (ordered from the lowest to the highest income groups or individuals). The extent of income inequality is determined by the difference between the Lorenz curve and a 45° line. The area between this line and the Lorenz curve is called the Gini coefficient (Rogerson, 2013). In this MSc thesis, this coefficient is used as a proxy to estimate the extent of income inequality within a country in a specific year.
In what follows, three (theoretical) channels have been described by which globalization is related to income inequality, namely: the Heckscher-Ohlin model and Stolper-Samuelson theorem, the offshoring model and the role of the informal sector.
2.1.1 Heckscher-Ohlin model and Stolper-Samuelson theorem
The Heckscher-Ohlin (i.e., H-O) framework is a long-run model that outlines trade patterns by focusing on the impact of different factor endowments (i.e., land, capital and labour). In addition, it determines relative prices and factor returns. The long-run character of this framework is explained by the fact that all factors of production (i.e., land, capital and labour) can move between sectors (e.g., between the agricultural and manufacturing sector) within a country. Moreover, within this theory, it is assumed that nations have identical technologies for their production and also the same consumer preferences. In this way, the H-O model concentrates on the differences in resources between countries that reflect the reason for trade. Furthermore, to make a very clear forecast about which industry in a country gains or loses from trade in the long-run, the model can be simplified to just two factors: Skilled and unskilled labour (Feenstra & Taylor, 2014).
8 they can produce these goods cheaper. Moreover, the relative price of skilled labour and skilled-intensive goods rises when they open up to free trade. Nonetheless, these goods remain still cheap compared to the countries, which are not skill abundant. Due to this, advanced countries specialize in skilled-intensive goods. Hence, the skilled-intensive sectors grow, whereas, e.g., the unskilled-intensive sectors in these countries shrink. According to the Stolper-Samuelson theorem, a higher relative price of skilled-intensive goods results in a relative rise in the wages of skilled labour and a relative decrease in the earnings of unskilled labourers. Hence, the unskilled workers in these skilled labour abundant countries lose, while the skilled workers gain. Additionally, this skilled labour abundant factor realizes benefits in all industries, not only in the skilled-intensive sectors. This means that all unskilled workers in these nations experience a relative fall in their earnings, while the skilled labourers benefit due to a relative rise in their wages (Feenstra & Taylor, 2014). This causes a rise in income inequality in advanced nations (Harrigan et al., 2016).
The developing countries can also be analysed by using the H-O framework and Stolper-Samuelson theorem. These nations are assumed to be more unskilled labour abundant and due to this they export unskilled-intensive goods (and import skilled-intensive goods). They specialize in these unskilled-intensive goods and the relative price of unskilled labour increases too. Therefore, the unskilled-intensive sector enlarges, while the skilled-intensive sector shrinks. A higher relative price for unskilled-intensive goods causes a relative rise in real wages for unskilled workers and a relative decline in the earnings for skilled labour. In this way, the unskilled workers in these countries benefit, whereas the skilled labourers lose (Feenstra & Taylor, 2014). The consequence of the surplus of unskilled labour (compared to skilled labour) in these developing countries leads to a decline in income inequality. The reason for this decrease is that the relative return for unskilled labour has increased because it is a relative abundant factor. This makes the relative difference between the returns for unskilled and skilled labour smaller and it creates a more even distribution of income (Atkinson & Bourguignon, 2014).
9
2.1.2 The offshoring model
Offshoring is trade in intermediate inputs. These intermediate inputs are combined to produce a final good that is sold in or outside the home country. Although this results in cheaper products and services for consumers in a home country, it also alters which jobs are located in a country (Feenstra & Taylor, 2014).
In advanced nations, firms, which offshore their activities, decide to relocate activities that are performed by the least skilled employees and are labour-intensive. These activities are, for example, the assembly and component production of a good. The offshoring of these activities enables cost savings for firms because the wages are lower abroad. Globalization has caused that more activities have been offshored. This also means that the activities that are offshored by firms have a broader range: Besides the least skilled activities, the low(er)-skilled activities have been included too. In developing countries, these broader range of activities are still relatively skilled, while they remain relatively unskilled activities in advanced nations. Hence, the relative demand for high(er)-skilled workers increases in advanced countries. This results in a higher relative wage for these employees. In this way, the income distribution between high(er)- and low(er)-skilled workers becomes (more) uneven in advanced nations (Feenstra & Taylor, 2014).
In developing countries, the offshoring of firms’ activities that were previously located in advanced nations has provided many jobs for the least and the low(er)-skilled employees. Though, the consequence of an increase in offshored activities to these developing nations, is a rise of the relative wage for relatively skilled workers in these countries. This is the result of the fact that from the perspective of developing countries, the offshored activities are relatively (more) skilled. Additionally, the demand for (more) skilled labour increases in these nations. In this way, the income gap between the low(er)- and high(er) skilled workers widens. Hence, this results in an increase of income inequality in developing countries too (Feenstra & Taylor, 2014).
2.1.3 The role of the informal sector
10 competition, firms in the formal sector may reduce or even stop with their compliance with labour standards to lower costs. Hence, firms decrease expenses, replace permanent workers for temporary ones or even fire workers, and outsource activities to the relatively cheaper and smaller firms in the informal sector. Due to this, workers who worked in the formal sector become unemployed and/or find jobs in another sector. In this way, a bigger share of the workforce is employed in the informal sector. Individuals who work in the informal sector perform jobs of low(er) quality (e.g., these jobs do not provide pensions or other worker benefits) and earn lower wages than individuals who work in the formal sector (Goldberg & Pavcnik, 2004). In addition, it is presumed that the informal sector provides employment for a disproportional share of unskilled and low(er)-skilled workers. The result is that these workers earn less than in their initial employment and this widens the gap between the wages of low(er)- and high(er)-skilled workers. The latter contributes to a rise in income inequality (Goldberg & Pavcnik, 2004; Goldberg & Pavcnik, 2007). Moreover, Dix-Carneiro and Kovak (2017) have argued that the benefits and costs of globalization are unevenly dispersed geographically and not only across skills or industries. Hence, regions, which are affected most by the negative effects of globalization encounter a relative bigger increase in unemployment and employment in the informal sector in the medium run. In the long-run, the size of the informal sector increases due to fact that it also absorbs a significant share of the workers who were unemployed due to the negative effects of globalization. Furthermore, this means that income inequality can rise between geographical locations and not just between skills or industries (Dix-Carneiro & Kovak, 2017).
2.1.4 Other factors that are related to income inequality
Income inequality can also be the result of other factors. One of those factors is the existence of corruption in a country. When the latter is present, the government’s role in the resource allocations of their nation is distorted. Individuals who have a better connection or relationship with the government, receive most or the biggest share of resources. Most of these individuals belong already to the group with the high(est) incomes. This results in a more uneven income distribution within a country. Therefore, when the level of corruption rises, income inequality increases too (Gupta, Davoodi & Alonso-Terme, 2002).
11 increased. Although this technological progress raises productivity and overall wealth, not everyone benefits. Three groups of individuals who lose can be described (Brynjolfsson & McAfee, 2012). First of all, the rising use of technology has resulted in skill-biased technological change. This implies a relative higher demand and wages for high(er)-skilled workers compared to low(er)-skilled workers (Brynjolfsson & McAfee, 2012). Another division between groups can be made where only one or a few person(s) or firm(s) earn(s) almost all in an industry (e.g., the market for music or CEOs). The growing use of technology has expanded the scope and size of these sectors, which increased the earnings of these persons or firms. In this way, the income distribution has become more uneven as a small group receives most of the income that can be earned (Brynjolfsson & McAfee, 2012). Lastly, the division between labour and capital is affected by the rising use of technology. Automatization is related to a decrease in the relative significance of labour and to an increase in the importance of capital. So, capital owners receive a bigger share of income from services and goods produced than workers. As the number of capital owners is significantly lower than the number of labourers, the distribution of income becomes more uneven (Brynjolfsson & McAfee, 2012).
2.2 Attitudes towards income equality
12 still (moderately) expensive for individuals who belong to the middle and lower classes of society (Harrigan et al., 2016; Dabla-Norris et al., 2015).
These above described policies to moderate a country’s extent of income inequality are directly imposed by their government/parliament and institutions. However, it is also indirectly enforced by the citizens of a country and their ideas (i.e., the social context of a country) about, e.g., the distribution of income (Milanovic, 2013). It is an indirect influence because citizens have voted for certain political parties, which in turn represent these ideas in the government. The following paragraphs outline the different determinants that influence a person’s opinion about an uneven income distribution.
Firstly, the different views of citizens about the redistribution of income in their country are established by their different opinions about the costs of this redistribution, which are determined by their individual economic mobility experience. This means that when someone agrees that a high income is a result of individual effort, he/she in turn does not agree with the fact that more redistribution is needed. Hence, this person beliefs that his/her high(er) income is a result of his/her effort and that this income is justified due to this effort. The opponents of this view, mostly the persons who have a low(er) income, think that a high income is the result of luck or other factors beyond a person’s control (Piketty, 1999).
13 Lastly, these above described determinants result also, inter alia, in a person’s choice concerning political affiliation and, eventually, in a vote for a certain political party. It appears that the opinion of the individual about income inequality is decisive regarding his/her voting decision, not his/her income (Piketty, 1999). In general, individuals who are in favour of a (more) equal income distribution vote for left-socialist parties (Brown-Iannuzzi et al., 2017). These parties are representatives of these individuals’ views because they strive strongly (compared to other political parties) for income redistribution and hence, income equality (Schumpeter, 2013). Individuals who are less or not in favour of income redistribution in their country tend to vote for right-wing political parties, which represent these ideas (Piketty, 1999).
Hence, the association between income inequality and globalization can be influenced by the opinions of people towards income equality. To use and measure these people’s attitudes towards income equality in this MSc thesis research, a proxy indicating the shares of votes (in %) for left-socialist parties in a government/parliament of a country are utilized.
2.3 Hypotheses and Conceptual model
In the previous parts of this section, the literature has been reviewed and outlined. Consequently, the following two hypotheses have been constructed for this MSc thesis that concentrates on the effect of attitudes towards income equality influencing the relation between globalization and income inequality:
H1: Globalization is related to an increase in income inequality within a country.
H2: When citizens consider income equality important and express this in their voting for left-socialist parties, it is related to a lowering of the effect of globalization on increasing income inequality within a country.
14 the variables. The first association (i.e., H1) focuses on the fact that more globalization is related to a more unequal income distribution within a country. The second relationship (i.e., H2) moderates this effect of globalization on income inequality. Hence, the more important to persons an equal income distribution is, the smaller the effect of globalization on income inequality will be.
Figure 1: Moderation model
Attitudes:
Income equality
Globalization
Inequality
Income
-
+ H1
15
3. Data and Methods
In order to examine the association between globalization and income inequality when attitudes towards income equality is the moderator, the research of this MSc thesis focuses on the following parts of a statistical analysis: Data (3.1) and Methods (3.2). The next subsections outline these steps by using a quantitative research design.
3.1 Data
The data used for this MSc thesis research derives from two data sources. First, the current version of the World Income Inequality Database (WIID3.4) has been utilized. This panel dataset contains 54 variables, 8,817 observations and covers 182 countries all over the world up and including the year 2015. In addition, the WIID3.4 has collected and stored the Gini coefficient (although in a scale from 0-100 instead of the original scale from 0 to 1) and other information regarding income inequality for developed, developing and transition nations (UNU-WIDER, 2017). From this dataset, the Gini coefficient has been used to represent the dependent variable.
Secondly, the Quality of Government institute (QoG) Standard dataset has been employed. It includes approximately 2,500 variables from more than 100 data sources worldwide. Moreover, it represents panel data encompassing observations from 1946 to 2016 (Teorell et al., 2017). From this panel data of the QoG Standard dataset, the following four variables have been used: Trade (in % of GDP), average schooling years (female and male, 25+), GDP growth (in %) and share of votes (in %) for left-socialist parties. These variables represent the independent variable, two control variables and the moderator variable, respectively.
The two above datasets have been merged to create one final dataset, which resulted in an unbalanced panel dataset. In this sample, it means that the number of observations over time is not the same across countries.
The next subsections outline more specific information about the five chosen variables.
3.1.1 Dependent variable
16 of income among people or households within a nation differs from a perfectly equal distribution, ranging from 0 to 100. Hence, if the Gini coefficient is 100 there is complete income inequality, whereas a coefficient of 0 indicates complete income equality. This variable has been chosen because it represents the level of income inequality in a country, i.e., the DV.
Nevertheless, the observations of this Gini coefficient in the WIID3.4 dataset have been derived from multiple data sources. In addition, the measurements of income1 and who receives this income2 differ (UNU-WIDER, 2017). In this way, the Gini coefficient is not consistent and comparable over time. To solve this data comparability problem, the empirical approach of Bensidoun, Jean and Sztulman (2011) has been used. These authors have reconsidered the evidence regarding the influence of international trade on the income distribution, which shows an overlap with this MSc thesis study. In addition, they have concentrated in particular on data consistency of the Gini coefficient and a feasible empirical method. Besides this approach, the user guide of the WIID3.4 dataset has been utilized because this MSc thesis research uses the newest version of the WIID (which includes updates and revisions), while Bensidoun et al. (2011) have utilized an earlier version. The following paragraph describes the five different steps to make the Gini coefficient variable consistent and comparable over time.
First of all, only the Gini coefficients which are labelled as ‘high quality’ have been selected to create reliable observations, which meet the minimal standards3. Secondly, the measurements of income and who receives this income differ per country and according year. Therefore, the
1 The measurement of income has four categories, namely: (1) Consumption (i.e., income spent on consumed
non-monetary items); (2) Gross income (i.e., income before tax deductions); (3) Net income (i.e., income after tax deductions); and (4) Other (i.e., income that is not defined as consumption, gross or net income) (UNU-WIDER, 2017).
2 There are four categories regarding the fact who receives the income, namely: (1) Family; (2) Household; (3)
Person; and (4) Tax Unit (UNU-WIDER, 2017).
3 The three minimal standards of reliability are: “(1) Inequality measures are based directly on household surveys
17 same measurement of income4 and the same recipient unit5 (i.e., who receives this income) have been used for a country and according years (Bensidoun et al., 2011). For example, Canada has the income measurement ‘Income, gross’ and the recipient unit ‘Household’ for the years 1969-2011. Thirdly, when still multiple options can be selected for a specific country and year (e.g., for Angola in the year 2009), the same source where the Gini coefficient of a country has been derived from (e.g., the World Bank) has been utilized because (almost) each Gini coefficient per country has been measured by one and the same source over time. The written comments for each source have also been considered (e.g., in which month the Gini coefficient has been measured). Additionally, if two or more sources have estimated all available Gini coefficients for a country over time, the most recent source have been chosen to control for accuracy and reliability. Fourthly, the same measurement level of income6 has been selected when a source has still more than one Gini coefficient estimation for a country and year. Lastly, three criteria regarding (1) which part of a specific country, (2) which part of this country’s population or (3) which age groups within this country’s population7 have been followed when still multiple Gini coefficients per country and according year can be selected (UNU-WIDER, 2017).
These steps make the Gini coefficient variable consistent and comparable over time. It results in a substantial decrease in the amount of observations that can be used in the empirical analysis. Bensidoun et al. (2011) have also utilized a comparable small sample of Gini coefficients (120 observations, which covers 41 countries) in their empirical approach.
4 When there is a choice in multiple measurements of income for a specific country, the following hierarchy has
been employed: (1) Gross income; (2) Net income; (3) Consumption; and (4) Other (Bensidoun et al., 2011).
5 When there is a choice in multiple recipient units for a specific country, the following hierarchy has been utilized:
(1) Household; (2) Person; (3) Family; and (4) Tax unit. (Bensidoun et al., 2011; UNU-WIDER, 2017).
6 When there is a choice in multiple measurement levels of income, the following hierarchy has been used: (1)
Household per capita; (2) Household adult equivalent; (3) Other; and (4) Without adjustment (UNU-WIDER, 2017).
7 When there is a choice in multiple options for these three criteria, it is preferred to use the following hierarchy
18
3.1.2 Independent variable
The IV Globalization is a continuous variable, which has been derived from the QoG Standard dataset. It is measured by taking the sum of exports and imports of both goods and services as a share (in %) of GDP. Although this proxy can be seen as a standard measurement, which is used in many studies, and has a wide availability in multiple datasets, is has two disadvantages (Rodriguez & Rodrik, 1997; Frankel & Romer, 1999). Firstly, trade (i.e., the sum of exports and imports of both goods and services) does not measure perfectly the economic interactions with other countries. The reason for this is that it does not take into account the specific (geographical) characteristics of a country. For example, the size of a country implies the amount of trade: A smaller country trades more internationally than a bigger country (which has, in contrast, more within-country trade) (Frankel & Romer, 1999). Secondly, more trade of a country is often a consequence of lower policy-induced barriers to international trade from this country’s government or parliament. The latter causes more openness to trade, which results eventually that this country exports and imports more internationally. Hence, it can be said that the lowering of trade barriers (or removing (almost) all trade barriers) reflects the level of globalization within a country better (Rodriguez & Rodrik, 1997). Nonetheless, trade remains the common proxy for globalization because it contains data that is not contrived (such as, binary (0 1) measurements) (Rodriguez & Rodrik, 1997; Frankel & Romer, 1999).
3.1.3 Moderator variable
19
3.1.4 Control variables
Two control variables (CVs) have been used in this research.
The first CV (CV1) reflects the average schooling years of male and female persons aged 25 years or older. This age category has been chosen because at approximately 25 years old a person can achieve his or her highest level of educational attainment. CV1 (Education_years) includes observations from 147 countries over the years 1950-2010. This control variable has been selected because of the supply effect of more educated (i.e., skilled) workers (due to, e.g., the rising use of technology within firms). In this way, the demand for more relatively skilled employees rises, whereas the demand for the relatively unskilled employees decreases (Harrigan et al., 2016). Advanced countries have relatively more skilled workers and developing countries relatively more unskilled workers (Feenstra & Taylor, 2014). This results in an increase of income inequality within advanced countries, while it elicits a more even income distribution within developing countries (Harrigan et al., 2016; Gregorio & Lee, 2002). Hence, the effect of Education_years on Income_Inequality is ambiguous as it depends on the level of development of a country. Observations for educational attainment are only available every five years. Therefore, missing observations have been imputed using linear interpolation. It is expected that the amount of years of education per individual in countries grows each year (Weisman, 2015; Papageorgiou, Lall & Jaumotte, 2008).
20 respectively high Gini coefficients (Kuznets, 1955). Thus, Economic_growth can have a positive or negative effect on Income_Inequality as it depends on a country’s stage of economic development. In this way, CV2 has an ambiguous impact on income inequality.
3.1.5 Descriptive statistics
The variables differ considerably in the amount of observations due to limited data availability and missing observations. Moreover, the five variables are measured at different scales.
Income_Inequality has a scale from 0-100, whereas Globalization, Economic_growth and Attitude show percentage values without an indicated range for these variables. The variable Education_years has been indicated in years.
Furthermore, the initial data sample points out huge outliers. Especially, the variables
Globalization, Economic_growth and Attitude show big differences in minimum and maximum
values. The maximum values are approximately 5 to 24 standard deviations bigger than their mean values. Therefore, both the 1st and 99th percentiles of this sample have been deleted. In this way, the extreme values (i.e., outliers) have been removed from the sample. The final descriptive statistics can be found in Table 1.
Table 1: Descriptive statistics
Variables Obs. Mean St. Dev. Min. Max.
DV (Income Inequality) 104 31.12 6.24 20.2 49.9
IV (Globalization) 104 77.11 47.35 9.22 297.56
CV (Education_years) 104 10.18 1.50 5.92 13.42
CV (Economic_growth) 104 2.40 5.21 -21.40 9.79
MV (Attitude) 104 1.79 4.94 0 29.8
21 which are high-income nations and/or, which have governments/parliaments dominated by right- or center-wing political parties. This may imply for the analysis that the effect of Attitude will (almost) not be found and that this effect is very small too. In addition, the variable Attitude has caused that the final data sample is substantially small due to many missing observations.
3.1.6 Scatterplot of the relationship between Income inequality and Globalization
A scatterplot has been created to show the direction of the relationship between
Income_Inequality and Globalization (see Graph 1). This graph indicates that more
globalization is related to a lower Gini coefficient. This is a somewhat surprising finding given the current body of theory and empirical evidence. It might be found due to the sample that is considered for this analysis. It will turn out to be a relation that is also observed in the econometric analysis.
Graph 1: Scatterplot of Income_Inequality and Globalization
22
3.2 Methods
The five chosen variables in this MSc thesis research derive from a panel data set. Such a data set contains a group of cross-sectional units (in this case, countries), which are analysed over time (in this case, years).
3.2.1 Econometric model
An econometric model has been established to show the relationship between the DV (Income_Inequality), IV (Globalization), CVs (Education_years and Economic_growth) and MV (Attitude). It is a fixed-effects panel model because this research focuses on analysing the impact of variables that varies over time. Moreover, this model investigates the association between the five variables within an entity (in this case, a country). Each country encompasses its own individual characteristics (e.g., its proximity to a sea or amount of neighbouring countries) that can (or cannot) have an influence on the IV, CVs and MV. To ensure that these characteristics do not have an effect on or even bias the variables, the fixed-effects model has been used to delete the effect of these time-invariant characteristics. This has been done by adding slope coefficients for all countries to the econometric model. Only the intercept 𝛽"
differs per country and this intercept is also called the ‘fixed-effect’. In this way, the net effect of the IV, CVs and MV on the DV can be studied.
Moreover, the intercepts can be estimated by using the effect estimator. Within the fixed-effects panel model, it is possible that the standard errors (𝜀%&) correlate within clusters, which
are countries. Therefore, cluster robust standard errors clustered by country have been used. However, the fixed-effects model also has a disadvantage. Due to the fact that it adds dummy variables for each, in this case, country, it uses only the within country variations in the variables to obtain the estimates of the coefficients. Hence, the between country variations are not utilized.
Equation (1) shows the relationship between the DV, IV, MV and two CVs as has been described in H1 and H2:
𝐼𝑁𝐶𝑂𝑀𝐸_𝐼𝑁𝐸𝑄𝑈𝐴𝐿𝐼𝑇𝑌%& = 𝛽"+ 𝛽7𝐺𝐿𝑂𝐵𝐴𝐿𝐼𝑍𝐴𝑇𝐼𝑂𝑁%&+ 𝛽;𝐸𝐷𝑈𝐶𝐴𝑇𝐼𝑂𝑁_𝑌𝐸𝐴𝑅𝑆%&+
𝛽?𝐸𝐶𝑂𝑁𝑂𝑀𝐼𝐶_𝐺𝑅𝑂𝑊𝑇𝐻%&+ 𝛽B𝐴𝑇𝑇𝐼𝑇𝑈𝐷𝐸%&+ 𝛽C(𝐺𝐿𝑂𝐵𝐴𝐿𝐼𝑍𝐴𝑇𝐼𝑂𝑁 𝑥 𝐴𝑇𝑇𝐼𝑇𝑈𝐷𝐸)%&+ 𝜀%&
23 where INCOME INEQUALITY shows the extent of income inequality within countries (i) over time (t), the independent variable is trade in both goods and services, calculated as a share (in %) of GDP (GLOBALIZATION) and the moderator variable indicates the attitudes of people towards income equality (ATTITUDE). In addition, the two control variables are, respectively, the average schooling years (EDUCATION_YEARS) and the GDP growth (in %) (ECONOMIC_GROWTH). The interaction effect (GLOBALIZATION x ATTITUDE) shows the impact of GLOBALIZATION on INCOME_INEQUALITY when it is moderated by ATTITUDE.
It is expected that the coefficient of Globalization is positive and is related to an increase in the extent of Income_Inequality. On the other hand, the coefficient of Attitude is expected to be negatively related to lower the level of Income_Inequality. The effects of Economic_growth and Education_years can be positive or negative due to their ambiguous influence on
24
4. Results and Discussion
In this section, the two hypotheses are tested. Section 4.1 presents and discusses the results. In section 4.2 and 4.3, control diagnostics and robustness tests are outlined.
4.1 Empirical results and Discussion
For this MSc thesis analysis, the fixed-effects model is theoretically preferred (see also Section 3.2.1). Therefore, the regression analyses of the two hypotheses have been performed using the within regression estimator. Nonetheless, the Hausman test8 shows that the random-effects model (i.e., the GLS estimator) cannot be rejected. Additionally, the test of over-identifying restrictions (i.e., orthogonality conditions)9 also indicates that the random-effects model cannot be rejected. However, the findings of the random-effects model are qualitatively similar compared to the results of the fixed-effects model. Hence, the fixed-effects model remains the preferred model. All the results of the regression analyses described in the following paragraphs can be found in Table 2 (see next page).
The first model (Model 1) includes the relationship between Income_Inequality and
Globalization, encompassing the control variables Education_years and Economic_growth.
The estimate of Globalization is significant at the .05 level and is related to a decrease in
Income_Inequality. Moreover, the coefficient of Globalization suggests that when Globalization increases with one standard deviation, it is related to a 0.43 lower standard
deviation of Income_Inequality10. Model 1 is significant at the .01 level and 18.6% of the within variance can be explained by the variables in this model. Consequently, H1 (Globalization is
related to an increase in income inequality within a country) cannot be confirmed.
8 Test results of the Hausman test are: p-value = .2922 and 𝜒; = 3.73.
9 Test results of over-identifying restrictions are: p-value = .3592 and 𝜒; = 3.218.
10 This number has been measured as follows (also for the other variables in this model and the other two models):
(Standard deviation of Globalization x the B-coefficient of Globalization) / Standard deviation of
25
Table 2: Regression analysis of Income_Inequality (DV), Globalization (IV),
Education_years (CV1), Economic_growth (CV2), Attitude (MV) and the Interaction effect
(1) (2) (3)
VARIABLES Model 1 Model 2a Model 2b
Globalization -0.0562** -0.0559** -0.0675** (0.0245) (0.0247) (0.0260) Education_years 1.446*** 1.452*** 1.543*** (0.361) (0.363) (0.367) Economic_growth 0.00558 0.00111 0.000635 (0.0524) (0.0537) (0.0534) Attitude -0.0721 -0.698 (0.170) (0.495) Interaction effect: Globalization x Attitude 0.00579 (0.00430) Constant 20.72*** 20.78*** 21.06*** (2.905) (2.925) (2.916) Observations 104 104 104 R-squared 0.186 0.188 0.208 Number of countries 30 30 30
Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
This is a surprising finding. Based on the studies of Bourguignon (2015) and, Piketty and Saez (2014), it is expected that more trade (due to globalization) is related to an increase in the Gini coefficient within advanced countries. However, the results suggest that globalization is related to a decrease in income inequality. In this way, it supports the fact that globalization can also lower the Gini coefficient within, especially, developing countries, because globalization is, inter alia, related to more economic growth, which implies, e.g., less unemployment. (Milanovic, 2003). Additionally, the H-O model and Stolper-Samuelson theorem also describe that developing countries could experience a decrease in income inequality (Feenstra & Taylor, 2014; Atkinson & Bourguignon, 2014). Nonetheless, the sample on which the regression analysis is based, includes 30 mainly high-income countries11. The found result contrasts with the research by Harrigan et al. (2016), which finds that income inequality increases within
11 These 30 mainly high-income countries are: Australia, Austria, Canada, Croatia, Denmark, Finland, France,
26 advanced countries due to the fact that the wages of skilled workers relatively rise and the earnings of unskilled workers relatively decrease. It should be noted that the sample of this MSc thesis is small and not representative to make conclusions for all countries worldwide. It should also be mentioned that this has consequences for the interpretation (and limitations) of the other two models discussed below.
Education_years is the only significant control variable (at the .01 level) in Model 1 and it is
related to an increase in Income_Inequality. The coefficient of this control variable suggests if
Education_years rises with one standard deviation, Income_Inequality increases with 0.35
standard deviations. This is consistent with the study of Harrigan et al. (2016) that indicates: Income inequality rises in advanced countries due to the increasing demand of skilled (i.e. more educated) employees, which is caused by the growing use of technology within firms. The latter is also known as skill-biased technological change (Harrigan et al., 2016).
The second model (2a) has one extra variable compared to the first model: The moderator variable Attitude. The expectation of the variable Attitude is that it is related to lower
Income_Inequality (see also Section 2.2). This cannot be confirmed because the negative
coefficient of Attitude is insignificant. As in Model 1, both Globalization and Education_years are significant at the .01 level.
The last model (Model 2b) presents if and how the interaction between Globalization and
Attitude influences the effect of Globalization on Income_Inequality. In this way, the second
hypothesis has been tested. Almost the same results as in Model 2a have been found in Model 2b. The differences are best presented in the estimates of Attitude and the interaction effect. The extent of the influence of Attitude on Income_Inequality increased, but remains insignificant. The interaction effect is related to an increase in Income_Inequality, though, it is also not significant. Only the estimates of Globalization and Education_years remain significant in this model. Consequently, their coefficients suggest that when these variables rise with one standard deviation, Income_Inequality alters with, respectively, -.51 and .37 standard deviations. The within R-squared value (.208) has increased compared to both Model 1 and Model 2a, and the whole model is significant at the .01 level. Thus, statistical proof of the decreasing effect of
Globalization on increasing Income_Inequality when it is moderated by Attitude (i.e., the
27
important and express this in their voting for left-socialist parties, it is related to a lowering of the effect of globalization on increasing income inequality within a country) cannot be
confirmed.
The interaction effect of the moderator variable (i.e., Attitude) has been visualized by using a (margins)graph (see Graph 2).
Graph 2: Interaction effect (Globalization x Attitude) on Income_Inequality
Graph 2 shows two outer lines (=green), which indicates the 95% confidence range for the interaction line. This interaction line (=red) presents the marginal effect of Globalization on
Income_Inequality for the full range of possible standardized scores of the moderator variable Attitude. The small dots show all observations for Attitude in the sample of this research. The
interaction effect is only significant when both green lines (representing the confidence interval) are both above or below the y-line=0 (=dark grey line). Hence, this graph only shows a significant impact when the standardized score of the moderator variable Attitude is approximately lower than five (and higher than zero). Though, the overall finding of Graph 2 is that Attitude does not influence the marginal effect of Globalization on Income_Inequality.
-. 2 -. 1 0 .1 .2 Ma rg in a l e ff e ct o f G lo b a liza ti o n o n I n co me _ In e q u a lit y 0 5 10 15 20 25 30
28 The findings in Model 2b and Graph 2 do not confirm the studies of Piketty (1999), Brown-Iannuzzi et al. (2017), Atkinson and Bourguignon (2014), and Schumpeter (2013). These authors argue that when citizens find income equality important, they express this opinion by voting for left-socialist parties in their country because these parties strive for income redistribution.
4.2 Model diagnostics
The following model checks are considered: (1) Normality, (2) heteroscedasticity and (3) serial correlation.
The first model check examines if the residual errors have a normal distribution. The Jarque-Bera test has been used to analyse this distribution. The kurtosis (2.52) and skewness (.43) values of Model 1 show an almost normal distribution. The second model (2b) has slightly lower kurtosis (2.50) and skewness (.41) estimates, which means that this model indicates an almost normal distribution too. Model 2b presents the same values for kurtosis and skewness as Model 2a. Hence, the Jarque-Bera test has determined that all three models have (almost) normal distributions of their residuals.
Secondly, the dataset has been checked concerning heteroscedasticity. If heteroscedasticity is observed, the variance of the error terms differs across observations. Therefore, the three models have been re-estimated using (cluster) robust standard errors. If the models improve statistically, heteroscedasticity is present. The results of the three models with (cluster) robust standard errors12 are presented in Table 3 (see next page). These three models show qualitatively the same findings. However, the standard errors, and therefore, also the significance levels, are somewhat lower.
12 The three models have an asterisk (*) behind their name to differentiate them from the original models in Table
29
Table 3: Heteroscedasticity analysis –
Inclusion of (cluster) robust standard errors in the regression analyses of the three models
(1) (2) (3)
VARIABLES Model 1* Model 2a* Model 2b*
Globalization -0.0562* -0.0559* -0.0675** (0.0324) (0.0326) (0.0318) Education_years 1.446** 1.452** 1.543** (0.605) (0.594) (0.570) Economic_growth 0.00558 0.00111 0.000635 (0.0715) (0.0764) (0.0763) Attitude -0.0721 -0.698 (0.216) (0.727) Interaction effect: Globalization x Attitude 0.00579 (0.00528) Constant 20.72*** 20.78*** 21.06*** (5.304) (5.324) (5.227) Observations 104 104 104 R-squared 0.186 0.188 0.208 Number of countries 30 30 30
Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
Lastly, an analysis has been done concerning serial correlation (or auto-correlation) as this is often observed in panel data. This kind of correlation indicates a relationship between observations of the same variable over specific periods of time. Hence, if serial correlation is found, the observations are not independent from each other. A test proposed by Drukker (2003) for serial correlation has been performed. This test shows that serial correlation is present in this sample13. This is surprising, as one would expect that serial correlation would not be present in a fixed-effects model. However, serial correlation has also been found when the random-effects model has been considered for this sample.
Thus, the results of these model diagnostics suggest that the findings should be considered with caution due to the possibility of heteroscedasticity and serial correlation that may affect the results.
13 The test presents the following: p-value = .0347 and F-statistic = 9.886. This shows the presence of serial
30
4.3 Robustness tests
Besides the model diagnostics, the robustness of the models has been evaluated by adding three potential control variables. All these control variables have been derived from the QoG Standard dataset.
The first potential extra control variable (CV3: Infrastructure) encompasses: The quality of the overall infrastructure in a country (e.g., energy, transport and telephony)14. This variable has been chosen because a better (development of) infrastructure in a country results in, e.g., more and/or better (paved) roads and telephone connections. Due to this, a country establishes a better economic integration, which results in more access to and possibilities of employment for, especially, people who do not live in urban areas. In this way, unemployment decreases and subsequently, the Gini coefficient might decrease in a country (Calderón & Servén, 2004).
Secondly, the continuous control variable that contains the number of internet users (from any location and device) per 100 people in the last 12 months is CV4 (Technology). According to Acemoglu (2003) and, Brynjolfsson and McAfee (2012), income inequality increases in advanced countries due to automatization of production processes, which causes more unemployment for low(er) skilled workers. In contrast, more technology improves, e.g., the access to and quality of education or better access to financial services in emerging and developing countries. In this way, income inequality decreases in these nations (Papageorgiou et al., 2008). Thus, the expected effect of Technology on Income_Inequality is ambiguous.
The last potential control variable (CV5: Corruption) indicates a corruption indicator. A high corruption indicator15 (i.e., high level of corruption) for a country presents a distorted role of the government regarding its resource allocations. This means that the individuals who have a better relationship or connection with the government – and who often also belong already to the high(est) income groups – receive most or the biggest share of resources. Thus, corruption might positively relate to income inequality (Gupta et al., 2002).
14 It is an ordinal variable with a Likert scale where (1) means “Extremely underdeveloped” and (7) “Extensive
and efficient by international standards”. Due to the fact that this variable has a negative influence on
Income_Inequality, the values of the Likert scale have been reversed.
15 The corruption indicator shows values between 0 and 100. This scale represents the following: 0 indicates that
31 The findings of this robustness check are presented in Table 416.
Table 4: Robustness test 1: Adding new CVs (Infrastructure, Technology and Corruption)
(1) (2) (3)
VARIABLES Model 1** Model 2a** Model 2b**
Globalization 0.0438* 0.0141 0.0131 (0.0177) (0.0293) (0.0397) Education_years -4.039* -5.727* -5.653 (1.770) (2.155) (2.944) Economic_growth 0.0427 0.0638 0.0641 (0.0670) (0.0653) (0.0801) Attitude 0.368 0.203 (0.299) (2.936) Interaction effect: Globalization x Attitude 0.000807 (0.0142) Infrastructure -1.176 -0.759 -0.720 (1.002) (1.002) (1.400) Technology 0.00173 0.0126 0.0169 (0.0214) (0.0220) (0.0797) Corruption 0.741** 0.642* 0.619 (0.200) (0.204) (0.480) Constant 55.14** 74.82** 74.60 (17.07) (22.63) (27.95) Observations 25 25 25 R-squared 0.902 0.935 0.935 Number of countries 15 15 15
Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
Model 1** shows that the coefficients of both Globalization and Education_years have changed from positive to negative (or the other way around) compared to Model 1.
Globalization is now related to an increase in Income_Inequality, while Education is related to
a decrease in Income_Inequality. In addition, it is interesting to mention that Globalization shows the expected positive coefficient as described in H1, but is now significant at a higher significance level (.10). The potential control variable Corruption has the lowest significance level (.05) in this model and it is related to an increase in Income_Inequality. In Model 2a**, only the control variables Education_years and Corruption are significant at the .10 level.
32 Moreover, the impact of Education_years on Income_Inequality increased, whereas the effect of Corruption on the dependent variable decreased. When all these three potential control variables are added to Model 2b**, all variables are insignificant. However, it is interesting to note that the inclusion of these three potential control variables results in relatively high within R-squared values for all models in Table 4 compared to the original models. In addition, only 15 countries have been analysed in Model 1**, 2a** and 2b**, which results in a smaller sample than the original models.
Finally, an alternative moderator variable has been considered. In this MSc thesis, the selected moderator variable measured attitudes towards income equality with the following proxy: Shares of votes (in %) for left-socialist political parties in a government/parliament. These votes of citizens are generally more in favour of the redistribution of income (Brown-Iannuzzi et al., 2017; Schumpeter, 2013). Though, these voters have maybe (also) chosen for left-socialist parties because they want, e.g., equalities in justice, more accessible public services or more/better rights for workers through labour unions (Schumpeter, 2013). Hence, it is not completely certain if these votes include also the fact that people find income equality important. Moreover, it can also be that the left-socialist voters and political parties strived too much for equality. This affects a country’s economic growth negatively because policies regarding income redistribution (e.g., a rise in (minimum) wages could make a country less cost competitive compared to other countries) could hinder this growth (Okun, 1975). In addition, the used proxy can be seen as an indirect measurement of attitudes towards income equality because the governments/parliaments decide in the end policies and rules that support or not support a (more) equal income distribution.
Although the used proxy in this MSc thesis research is preferred due the wider availability of data, an alternative proxy has been considered to control for its robustness. This alternative proxy has been derived from the combined longitudinal dataset of the World Values Survey (WVS) and European Values Survey (EVS). This variable shows the extent to which people find income equality important17. Hence, this proxy can be seen as a more direct estimate of people’s opinions towards income equality.
17 This variable has a Likert scale from 1 (“Incomes should be made more equal”) to 10 (“We need larger income
34
5. Concluding Remarks
5.1 Conclusions
This study examined whether the relation between income inequality and globalization is affected by attitudes from people towards income equality. The importance attached to government policies and institutions influencing income equality is represented by the share of votes (in %) for left-socialist parties in a government/parliament of a country. To study this moderation model (see also Figure 1 in Section 2.3), the following research question has been analysed: Do citizen’s attitudes towards income equality expressed in their voting behaviour
for left-socialist political parties affect the relationship between income inequality and globalization?
A panel dataset has been utilized to test the relationship between income inequality and globalization, and the moderation effect of citizen’s attitudes. There are several interesting findings. First of all, it cannot be confirmed that globalization is related to an increase in income inequality. According to Bourguignon (2015) and, Piketty and Saez (2014) more trade (due to globalization) is related to a rise in an uneven income distribution within advanced countries. Nevertheless, the H-O framework and Stolper-Samuelson theorem outline that, especially, the developing nations would encounter a decrease in income inequality (Feenstra & Taylor, 2014; Atkinson & Bourguignon, 2014). Secondly, this research did not find that attitudes towards income equality can lower the effect of globalization on increasing income inequality. In contrast, the opposite effect has been found: Income inequality increases when globalization is moderated by attitudes towards income equality. However, the interpretation of this interaction effect should be considered with care as the estimate is insignificant. This is not consistent with the studies of Piketty (1999), Brown-Iannuzzi et al. (2017), Atkinson and Bourguignon (2014), and Schumpeter (2013). These authors have argued that when citizens find income equality important, they express this opinion by voting for left-socialist parties in their country because these parties strive for income redistribution.
5.2 Limitations and Future Research
There are several limitations that might be addressed in future research.
35 observations was unequal per variable. Moreover, the dataset contained many missing observations for, especially, attitudes towards income equality, which made the sample smaller and biased to particular countries and years (e.g., the sample included Gini observations belonging to earlier years, whereas attitudes towards income equality had observations from more recent years). The limited data availability has created a small sample. This small sample included observations from mainly high-income countries, although some of the observations were from emerging countries (from Eastern Europe). In this way, no consistent time observations for low-income countries were used. Due to these limitations regarding the data, future research should utilize a bigger and more recent dataset. Additionally, future research could find alternative empirical approaches regarding the Gini coefficient, which could also create a dataset that includes both high- and low-income countries.
Secondly, this analysis has considered one specific proxy for the moderator variable (i.e., attitudes towards income equality has been assessed by considering citizens’ votes (in %) for left-socialist parties in governments/parliaments). The choice of this proxy (also due to data availability) and not adding more moderator variables and according proxies can have caused that the expected findings have not been revealed. Moreover, it is possible that this selected proxy does not represent the opinions/feelings about income equality in the best manner. The left-socialist voters and/or political parties (probably) have other views besides their left-social idea regarding income redistribution. Consequently, it is not completely certain that these voters find income equality important. Hence, the results of this variable and interaction effect should be considered with caution. To overcome this limitation, future research should consider alternative proxies for attitudes towards income equality in the empirical analysis.
36
References
Acemoglu, D. (2003). Patterns of skill premia. Review of Economic Studies, 70(2), 199-230.
Atkinson, A.B. & Bourguignon, F. (Eds.). (2014). Handbook of income distribution. Elsevier.
Bensidoun, I., Jean, S. & Sztulman, A. (2011). International trade and income distribution: Reconsidering the evidence. Review of World Economics, 147(4), 593-619.
Beugelsdijk, S., Brakman, S., Garretsen, H. & Marrewijk, C. van (2013). International
economics and business: Nations and firms in the global economy (2nd edition). Cambridge:
Cambridge University Press.
Bourguignon, F. & Morrisson, C. (2002). Inequality Among World Citizens: 1820-1992. The
American Economic Review, 92(4), 727-744.
Bourguignon, F. (2015). The Globalization of Inequality. Princeton: Princeton University Press.
Retrieved from
http://ebookcentral.proquest.com.proxy-ub.rug.nl/lib/rug/detail.action?docID=1903251 on September 16th, 2017.
Brown-Iannuzzi, J. L., Lundberg, K. B. & McKee, S. (2017). Political Action in the Age of High-Economic Inequality: A Multilevel Approach. Social Issues And Policy Review, 11(1), 232-273.
Brynjolfsson, E. & McAfee, A. (2012). Race against the machine: How the digital revolution
is accelerating innovation, driving productivity, and irreversibly transforming employment and the economy. Lexington, Mass.: Digital Frontier Press.
Calderón, C. & Servén, L. (2004). The effects of infrastructure development on growth and
37 Dabla-Norris, E., Kochhar, K., Suphaphiphat, N., Ricka, F., Trounta, E. & International Monetary Fund (Strategy, Policy, and Review Department). (2015). Causes and consequences
of income inequality: A global perspective. Washington, District of Columbia: International
Monetary Fund.
Dean, P. & Ritzer, G. (2012). Globalization. In George Ritzer, ed., Wiley-Blackwell Encyclopedia of Globalization. Oxford: Wiley-Blackwell.
Dix-Carneiro, R. & Kovak, B. (2017). Margins of Labor Market Adjustment to Trade (Working paper No. 23595). National Bureau of Economic Research.
Drennan, M. (2015). Income inequality: Why it matters and why most economists didn't notice. New Haven: Yale University Press.
Drukker, D.M. (2003). Testing for serial correlation in linear panel-data models. Stata Journal,
3(2), 168–177.
Elliot, K.A., Kar, D. & Richardson, J.D. (2004) Assessing globalization’s critics: Talkers are
no good doers?, in Baldwin, R.E. and Winters, L.A. (eds.) Challenges to Globalization:
Analyzing the Economics. University of Chicago Press/National Bureau of Economic Research, 17-62.
Feenstra, R. & Taylor, A. (2014). International economics (3rd edition). New York: Worth.
Frankel, J.A. & Romer, R. (1999). Does trade cause growth? The American Economic Review,
89(3), 379-399.
Goldberg, P.K. & Pavcnik, N. (2004). Trade, inequality, and poverty: What do we know?
Evidence from recent trade liberalization episodes in developing countries (Working paper No.
10593). National Bureau of Economic Research.
38 Guillen, M.F. (2001). Is Globalization Civilizing, Destructive or Feeble? A Critique of Five Key Debates in the Social Science Literature. Annual Review of Sociology, 27, 235-260.
Gupta, S., Davoodi, H. & Alonso-Terme, R. (2002). Does corruption affect income inequality and poverty?. Economics of governance, 3(1), 23-45.
Harrigan, J., Reshef, A. & Toubal, F. (2016). The march of the techies: Technology, trade, and
job polarization in France, 1994-2007 (Working paper No. 22110). National Bureau of
Economic Research.
Milanovic, B. (2003). Can we discern the effect of globalization on income distribution?
Evidence from Household Surveys (Policy Research Working Papers). The World Bank.
Milanovic, B. (2013). The Inequality Possibility Frontier: Extensions and New Applications. (Policy Research Working Paper No. 6449). The World Bank, 1-31.
OECD (2017). Income inequality (indicator). Retrieved from
https://data.oecd.org/inequality/income-inequality.htm on September 14th, 2017.
Okun, A. (1975). Equality and efficiency: The big tradeoff. Washington, D.C.: Brookings Institution.
Papageorgiou, C., Lall, S. & Jaumotte, F. (2008). Rising income inequality. Retrieved from
http://ebookcentral.proquest.com.proxy-ub.rug.nl/lib/rug/detail.action?docID=1605838 on October 9th, 2017.
Piketty, T. (1999). Attitudes toward income inequality in France: do people really disagree? (Working paper No. 9918). Paris: CEPREMAP.
Piketty, T. & Saez, E. (2003). Income inequality in the United States, 1913–1998. The
Quarterly journal of economics, 118(1), 1-41.
39 Ritzer, G. & Atalay, Z. (eds. 2010). Readings in Globalization: Key Concepts and Major
Debates. Oxford: Blackwell.
Rodriguez, F. & Rodrik, D. (1997). Trade policy and economic growth: a skeptic's guide to the cross-national evidence. NBER macroeconomics annual, 15, 261-325.
Rogerson, P.A. (2013). The Gini coefficient of inequality: A new interpretation. Letters in
Spatial and Resource Sciences, 6(3), 109-120.
Schumpeter, J.A. (2013). Capitalism, socialism and democracy. Abingdon: Routledge.
Teorell, J., Dahlberg, S., Holmberg, S., Rothstein, B., Khomenko, A. & Svensson, R. (2017).
The Quality of Government Standard Dataset (version Jan 17). University of Gothenburg: The
Quality of Government Institute. Retrieved from
http://qog.pol.gu.se/data/datadownloads/qogstandarddata on October 6th, 2017.
UNU-WIDER (2017, January). World Income Inequality Database (WIID3.4). Retrieved from
https://www.wider.unu.edu/database/world-income-inequality-database-wiid34 on October 6th, 2017.
Weisman, S.R. (2015). The Great Tradeoff: Confronting Moral Conflicts in the Age of
Globalization. Washington: Peterson Institute for International Economics. Retrieved from
http://ebookcentral.proquest.com.proxy-ub.rug.nl/lib/rug/detail.action?docID=4394674 on September 14th, 2017.
Yergin, D. & Stanislaw, J. (2002). The Commanding Heights: The Battle for the World
