Governance and income inequality

(1)

Department of political science

Governance and income inequality

A panel data analysis on the relationship between the quality of government and

the Gini index

Kevin Ahrlind

(2)

Abstract

This study investigates the relationship between quality of government and income inequality. The purpose of this study is to understand how differences in quality of government across states affect their income inequalities. By utilizing a theoretical framework that incorporates Kuznets inverted U-hypothesis, quality of government and state capacity, this study argues that a high level of quality of government and state capacity are essential for states to handle income inequality. The method used was a regression analysis using panel data that covers the time period from 1946 to 2020. Furthermore, pooled OLS and fixed effects model were used to study the relationship. The results from the pooled OLS showcases that there’s a negative relationship between quality of government and income inequality. However, when controlling for entities by using the fixed effects model the results show a positive relationship. In addition, an F-test was conducted to find the regression model with the best fit where the fixed effects model showcased a superior fit to the given data used in the study. The study suggest that further research should be done on how income inequality and governance differs depending on the historical onset of the countries. Moreover, further research on how decentralization or centralization of power in a country affects income inequality is suggested.

(3)

Acknowledgments

(4)

1. Introduction

Economic growth has been central in making it possible for millions of people to increase their standard of living, simultaneously helping millions of people to es-cape extreme poverty. However, this prosperous growth is not without its adverse effects, as growth has been accelerating in the last decades so has income inequality (Keeley 2015, 11-12). This widening income inequality is the defining challenge of our generation, we are seeing a similar pattern in many of the rich, middle, and de-veloping countries. According to an OECD report released 2012, income inequality has risen globally where the top earners capture large shares of income gains while the rest of society has only seen minimal growth (OECD 2012, 182). Similarly, a UN report points out that income inequality has been growing globally since the end of the 20th century and consequently it has been growing most prevalently in emerging economies (Vieira 2012, 4). Problems with the unequal distribution of income among a country’s citizens are that it hampers economic performance which furthermore creates long-term problems for the state and its population. While eco-nomic growth is central for creating a higher standard of living amongst citizens of a country, it can be argued that the state and its government have to a certain extent the responsibility to increase the utility of the country’s inhabitants, by overseeing and neutralizing problems such as income inequality.

(7)

2015) points out that capital flows have changed amongst workers and owners, fur-thermore the disappearance of 9-to-5 jobs and the increase in short-term work seem to have had an effect. Moreover, state-related causes such as new tax systems seem to play a role in this complex topic (Ibid.). Hence, the studies provide evidence that there are more than a single factor affecting the growing income gap. Similarly, the consequences of income inequality are diverse as the reasons mentioned. It affects human capital, in other words, education and development of skills which are crucial in a well functioning society (Ibid.). In addition, it has an effect on individual level, where greater income inequality causes decreasing social cohesion and participation, two elements that are important for inhabitants of a state (Van de Werfhorst & Salverda 2012). Other studies show that an excessive income gap hampers eco-nomic growth, affecting countries and millions of people in different spheres of life (Dabla-Norris et al. 2015). Hence, it should be in the state’s interest to exercise its power to handle the issue of income inequality and consequently, minimize the consequences on the society and its inhabitants.

1.1 Purpose & research question

(8)

Quality of government represents several conditions which have to be optimized for a government to exercise its power in an optimal way in regards to issues and public affairs (La Porta et al. 1999). Meanwhile, state capacity represents the state’s abil-ity to complete tasks, it should be see as power of how well the state’s agents can get members of society to do things that they wouldn’t do otherwise (Lindavall & Torell 2017). Moreover, this study emphasizes that quality of government and state capacity often times come hand in hand. This study argues that the growing income inequality is a case of the theoretical framework presented above and that without a well-functioning government, i.e that has a high quality of government and proper state capacity, the state cannot counteract the growing income inequality. Thus, the study’s hypothesis is as the following: that there’s a relationship between income inequality and the quality of government, given that quality of government which is intertwined with state capacity increases the states commitment and means to deal with income inequality. Hence, an important assumption of the study is that the variable quality of government represents a well functioning state, but also a state that thrives for finding solutions to deal with issues such as income inequal-ity. Therefore, this study interprets growing income inequality as a problem being rooted in the quality of how a state is governed.

The purpose of this study is to investigate variances of income inequality among states through the time period from 1946 to 2020 and how the governance of these states affect the difference in income inequality. The topic of income inequality is seen as interesting and relevant since it affects individuals in all countries and states throughout the world. Moreover, it’s regarded as interesting from a political science perspective as the income gap differs across all states. And governments are head of these states i.e in charge of the set of tools and means to handle income inequality which may cause long term hazardous effects if not treated. Hence, the topic is fascinating from a power perspective, given differences in income inequality which orchestrate how governments are handling the issue. Thus, the research question of this study is:

• What role does quality of government have on income inequality?

(9)

inequality. The two models were selected since Pooled OLS doesn’t control for entities over time, whereas fixed effect model on the other hand does. In addition, to find which model suits the given data sample used in the study an F-test will be conducted to find the most suitable model of these two.

To summarize, this paper will study the income inequalities across states by utilizing the Gini index as the dependent variable, which measures the distribution of income across a state. The study argues that the growing income inequalities is a problem of the governance. Hence, from a theoretical point of view it can be understood as the problem is rooted in quality of government and state capacity which are conditioned on the Kuznets curve. The aim of the study is to investigate the relationship between quality of government and income inequality by using panel data analysis across a multitude of states between the time period 1946-2020. It should be mentioned that this paper aims to expand the existing research in the field by focusing on how income inequality has grown over the last decades in relation to quality of government across the globe. Most of the research done previously in the field usually covers a certain geographic area or country during a specific set of years. Therefore, this study argues that there’s a knowledge gap in the field in regards to studies of income inequality and governance across vast numbers of countries and time, a gap which this study intends to fill.

1.2 Predisposition

(10)

2. Background

2.1 Factors of income inequality

Income inequality has been on the rise since the last decades, according to a OECD report (Keeley 2015) in the 1980s the richest 10% of the population in the OECD countries earned seven times more than the poorest 10% (Keeley 2015, 3). Likewise in the US the income inequality has grown about 20% from 1980 to 2016 (Horowitz et al. 2020, 14). Similar patterns can be observed in many developing countries since the 1990s during which time inequality has risen in many of the emerging economies. Meanwhile, a certain decline in inequality has been seen in some countries (UN 2020, 21). Hence, a global pattern can be discerned where we see a growing income in-equality across several economies worldwide. However, the factors contributing to the rise in income inequality are many and diverse. One of these factors is glob-alism, which has made the global economy far more integrated with technology, information, trade and investment. Consequently, it makes the low-skilled worker far less attractive on the world market, due to advanced skills and human capital which are required for many of the jobs. Consequently, killing job opportunities of some middle and low-skilled workers (Keeley 2015, 42). A study published by Meshi and Vivarelli (2007) shows that growing income inequality seems to have been significantly affected by globalization, especially trade. Developing countries and middle-income countries often fall prey to growing income inequality (Meshi & Vivarelli 2007, 20-21). Bergh and Nilsson (2010) confirm similar results where economic globalization and trade liberalization has a positive effect on income in-equality, hence increasing the income gap (Bergh & Nilsson 2010, 488-489).

(11)

workers, instead more capital flows toward the owners (Keeley 2015, 47). In addition, there’s increasing evidence that the share of national income going to capital has been rising steadily, while the share that’s going to labor has been falling. In 1990s, 66.1% of the national income was going to labor in the OECD countries, but by the 2000s that number had fallen to 61.77% (Ibid.). As is the case with globalism, there are multiple factors such as trade and technology playing a big part, but it can be simply put as income that once went to the workers go to owners who finance machines and software (ibid.). Another factor behind the increasing income inequality we are seeing is that traditional jobs that last between 9-to-5 are in decline. Also, the numbers of union members have also been declining since 1990s. Hence, short-term work, or self-employment is becoming more widespread and since the mid-1990s, it represents more than half of all new jobs created in the OECD countries (Keeley 2015, 49).

Furthermore, the state has an important role in dealing with income inequality. Tools such as taxes and transfers that used to secure redistribution in society have been changing since the mid-1990s. There has been a decline in spending on unem-ployment benefits in countries, this is mostly due to falling unemunem-ployment, meaning fewer were claiming those benefits. Meanwhile, the rules for claiming benefits were tightened (Keeley 2015, 54). Taxes have also been declining, although the picture is more complex for taxes since lower tax by some experts lead to greater economic growth. New tax systems, such as progressive tax reforms were also introduced during the same period. However, the effect has been unclear in relation to income inequality. According to OECD the falls in top tax rate have been seen in many developed countries in the past few decades where the average top statutory tax rate fell from 66% in 1981 to 41% in 2008 (Keeley 2015, 60). Similarly, tax on property and inheritance has been on the fall in many regimes, allowing not only high earners to develop greater income but also greater wealth and capital. As we see with the factors presented above, the growth in income inequality is a complex issue with several elements and components playing a crucial role in the growing concern about this specific topic.

(12)

able to escape poverty. Therefore, one wonders why the issue of income inequality hasn’t been better met. According to some scholars, income inequality might have a good effect on economic growth since it creates more entrepreneurs that are more willing to take risks (Keeley 2015, 67). They mean that if there’s an overwhelm-ing interest to curb income inequality through state tools such as taxes it might instead harm the incentive to innovate, and so discourage entrepreneurs from taking such risks (Ibid.). Another argument why putting too much emphasis on possible negative outcomes of income inequality is that there might be a trade-off between economic inequality and economic efficiency according to the American economist Arthur Okun (Ibid.). However, these ideas have been increasingly criticized with rising evidence that excessive inequality has a negative effect on economic growth. According to OECD (Keeley 2015) the widening wealth gap between the individuals in societies leads to low earning families investing far less in education and skills, which creates a long-term problem since it hurt economic growth by reducing the numbers of skilled workers and diminishes the numbers of highly productive workers (Keeley 2015, 69). Likewise, a report from the International Monetary Fund, found similar results that a higher net Gini coefficient is associated with lower economic output (Dabla-Norris et al. 2015, 6). Thus, factors of income inequality is a highly complex topic where several factors contributes to its growth.

2.2 Consequences of income inequality

(13)

a greater tendency to endure long periods of unemployment on the labor market (Ibid.). These consequences can be broken down to the fact that students with well-off backgrounds generally have better learning opportunities, since they have a greater possibility in accumulating knowledge outside formal education systems. In addition, educational systems can also further reinforce this effect, since quality of education depend on the socioeconomic standard of the student and its family. The effect has been most prominently seen during the PISA 2012 test, where students from better-off families on average are one year ahead in mathematics competence compared to the students with worse-off backgrounds (Keeley 2015, 72-74).

(14)

might affect the taxpayer’s interest to invest in the economy, subsequently causing lower volumes of investment or even cause greater capital flight that would lower economic growth (Petersen & Schoof 2015, 6). Another outcome of excessive income inequality that might lead to sub-optimal outcomes is when economic power is used to exercise political influence, such as that well-off individuals advocating for a tax reductions. The associated decline in state revenues will further on have an effect on the state’s expenditure on fields such as infrastructure and education. Consequently, in long term dampening economic growth because of under-supplying in the public sector, given a decline in government revenue (Ibid.).

2.3 Measuring income inequality

There are several methods of measuring income inequality. The most common meth-ods are the Theil index, Hoover index and Gini index. However, since the study will only include the Gini index to study income inequality, the following section will only present the fundamental assumption of the Gini index.

(15)

Figure 2.1:

Graphical representation of the Gini coefficient. The X-axis represents the share of population, while the Y-axis represents the share of income accumulated. The curved line between A and B is the Gini coefficient, the closer to the 45 degrees

(16)

3. Previous research

(17)

may have.

Chiung-Ju Huang and Yuan-Hong Ho of Feng China University studies the relation-ship between governance and income inequality in “The impact of governance on income inequality in ten Asian countries” (2018). The authors use ordinary least squares regression to study the impact of governance quality on income inequality in Asian countries. The data used in the study is panel data covering advanced-and developing economies between the time period 1996 to 2015. The advanced economies include countries such as Japan, Singapore and South Korea, while de-veloping countries are represented by countries as China, Indonesia and Thailand (Huang & Ho 2018, 218-219). The dependent variable in the regression is the Gini index followed by several control variables as economic growth, the share of elderly in the population, two types of governance quality variables, one of which is demo-cratic quality and the other technical quality (Ibid.). Furthermore, the authors use a fixed effect and random model to deal with heterogeneity, they also test the cross-sectional dependence, slope homogeneity and panel unit root. The empirical results show that technical quality on income inequality is positive for advanced economies, while democratic quality is non-significantly positive. Hence, improving an advanced country’s political situation, such as accountability and voice will not be effective to reduce income inequality (Huang & Ho 2018, 222). On the other hand, promoting technical quality such as government effectiveness and regulatory quality might increase income inequality. For developing countries, the effect of democratic quality and technical quality is negative on income inequality. Therefore, promoting accountability and political stability but also the effectiveness of government and control of corruption will reduce income inequality (Ibid.). The authors conclude that a high quality of governance can play an important role for improving income equality in developing countries. However, promoting good governance in advanced countries may not be as effective as in developing countries (Ibid.).

(18)

(19)

4. Theoretical framework

4.1 Kuznets inverted U-hypothesis

(20)

the birth rate. Consequently greater parts of the population move into urban areas which in the long term results in the urban population dominating the rural pop-ulation (Kuznets 1955, 19). As a result, the average income continues to grow. In addition, the income gap narrows because the majority of the population now lives in urban areas of the city.

Hence through several factors such as people relocating into urban areas for labor and protective legislation being passed by the state and increasing standard of living the income inequality should decrease. Furthermore, Kuznets points out several factors why some countries don’t reach the last stage of declining income inequality. According to Kuznets, this pattern is often seen in emerging countries, one of the reasons is because the average income is much lower than in countries in the west which makes it difficult for the average individual to accumulate wealth, something the more asset-rich can do (Kuznets, 1955, 23). Another reason is that the industrial structure and opportunities are far more limited compared to the more industrialized countries. The last factor why some countries haven’t been able to narrow income inequality is because of government failure, i.e that they cannot build a strong foundation to protect and bolster the low-income class (Kuznets 1955, 24). However, according to research, the inverted U-pattern hypothesis has fared less well in recent years. The inequality was falling until the middle of the century as the hypothesis predicts but, since the latter end of the 20th century, it has been on the rise, creating a U-shaped growth instead (Keeley 2015, 65). Hence, its validity has been questioned with time by the empirical support it seems to somewhat lack.

4.2 Governance

(21)

(1999) a government can optimize its performance through several different factors which play crucial parts in it being able to in the best way possible handle issues and public administrative affairs.

(22)

which is more related to political interest rather than public-spirited intentions. The last dimension of good government is democracy and political rights. Political and economic freedom generally goes hand in hand and both are crucial of a well-doing government. However, this relationship has been difficult to find in recent data. On the other hand, it can be pointed out that over the long span of history, countries with greater freedom are generally more linked to a higher quality of government (La Porta et al. 1999, 9).

(23)

(24)

4.3 Kuznets inverted U-hypothesis & governance

Given the theoretical onsets presented above, this study argues that Kuznets in-verted U-hypothesis is conditioned by the quality of government but also by the state capacity. Especially, if the quality of government and state capacity is high, the government should be able to deploy countermeasures to handle income inequal-ity. However, if the opposite is true then the government will not be able to deal with income inequality. Specifically, that without a government that has high qual-ity, the countermeasures to deal with income inequality will not be efficient. Thus, as the inverted U-hypothesis states, in the early stages of economic growth, the income gap will grow steadily (Kuznets, 1955). However, after a significant time conditional on quality of government and state capacity, the government should, if quality and capacity is high provide policy and legislation to redistribute the income throughout the economy in a more optimizing manner. Hence, as La Porta et al. (1999) states crucial elements such as, limited interventions, high level of efficiency, public services, government expenditure and freedom must be optimized.

However, if state capacity isn’t optimized in the first case, a high quality of govern-ment cannot be achieved. This is because state capacity as previously govern-mentioned measures the causal effect between policy instruments and the desired outcome (Lindvall & Teorell 2017). Thus, a government that has a high ability to accomplish its task will generally represent a higher level of governance, because without the ability to fulfil the tasks, the crucial elements in dimensions of quality of government wouldn’t be optimized.

(25)

by a transfer system that rewards resources as a trade-off, optimal legislations or policies should be able to be implemented to decrease income inequality. Therefore, the relationship between state capacity can be seen as somewhat more complex than just a causal relation in one direction. A government that can maintain a high level while having a high state capacity can hence through legislations, taxes, and subsidies transfer and redistribute the capital flow within the society. Consequently, increasing the income of the lower percentiles in the state simultaneously boosts the economic output of the state as the U-hypothesis claims (Kuznets 1955).

On the other hand, if the criteria for high quality of government and state capacity cannot be reached according to the standards of La Porta et al. (1999) and Lindvall and Teorell (2017) the inverted U-hypothesis would have a far more difficult time achieving a smaller income gap amongst the percentiles in the society.

(26)

5. Theoretical hypothesis

This study will utilize a statistical hypothesis where a null hypothesis and an alter-native hypothesis will be in use to test if there’s a significant relationship between the variables of interest. In this study, the critical value that will decide if the null hypothesis can be rejected or not will be at a 10% level. Hence, if the critical value standard of this study is met at the population mean µ, it is said to be statistically significantly different from null hypothesis. The null and alternative hypothesis can be written as the following:

H0 = µ = 0 vs H1 = µ 6= 0

(27)

6. Data & method

6.1 Data

The data for this study is collected from the Quality of Government Institute at Gothenburg University. The data set includes 350 variables from 75 different data sources all related to the topic of good governance and quality of government. The specific data set used in this study is a panel data set that includes variables and observations from 1946 to 2020 across a multitude of countries. The units in the data is country-year based. It can further be mentioned that the data set is fairly balanced. However, it has some missing observations across entities during some years, particularly in the early years of the data set. In regards to reliability, the data set has collected its variables through international organizations and universities, thus reliability of the data set should be regarded as high. Given the wide range of countries and years the data set includes, the study can ensure that the results are not just regional tendencies during a specific year.

6.2 Method

This study utilized ordinary least squares regression analysis to study the relation-ship between income inequality and governance. The relation between these vari-ables can be estimated through the ordinary least squares estimator. The estimator chooses a regression coefficient so that the predicted regression line fits the observed data as close as possible. The closeness is measured by the sum of squared mistakes made in predicting the outcome variable, Yi given the control variable, Xi (Stock

(28)

β0 and slope of the regression line β1. The residual or error term, ε represents the

difference between outcome variable, Yi and the OLS predicted outcome value, ˆY

(Stock & Watson 2015, 117).

There are some central assumptions of the linear regression model that need to hold, otherwise the assumptions estimates will not give useful estimates. Since this study will utilize multiple control variables the assumptions presented will cover multiple regression analysis. However, the assumptions are much alike to the assumptions for linear regression with single control variable. The first least-squares assumption is that conditional distribution of the error term, εi given the control variables,

X1, X2...Xk has a mean of zero (Stock & Watson 2015, 199). Simply put, the other

factor in the error term should be unrelated to control variables when they take on a value. The second assumption is that the variables are independently and identically distributed, i.i.d across observation. In other words that the observations have the same probability distribution as the others and all are mutually independent (Stock & Watson 2015, 199). However, since this study is working with panel data, the assumption will be somewhat relaxed. The third assumption is that large outliers are unlikely, this is because the OLS estimator of the regression can be sensitive to large outliers thus affecting the output (Ibid.). And the last assumption, no perfect multicollinearity, that no regressor i.e control variable should be perfectly linear with other ones. The mathematical reason for this problem is that perfect multicollinearity produces division by zero in the OLS formula making it impossible to compute the OLS estimator (Stock & Watson 2015, 200).

The reason why this study will also utilize multiple regression is to control for possible omitted variable bias. The bias occurs when the following conditions are true: (1) when the omitted variable is correlated with the control variable, Xi and

(2) when the omitted variable is determinant of the outcome variable, Yi (Stock &

Watson 2015, 185). Since the control variable Xi, is correlated with the error term,

εithe first least-squares assumption will be violated, because the conditional mean of

εi given Xi is nonzero. In other words, since the least-squares assumption is violated

(29)

other words, a regression with a single regressor is vulnerable to omitted variable bias and multiple regression makes it possible to mitigate such bias in the estimates by including these omitted variables in the regression (Stock & Watson 2015, 206). In the context to the topic of the study, several control variables for the reasons mentioned above.

6.2.1 Logarithmic regressions

The regression model will also include natural logarithm, therefore, the function will be a nonlinear regression function. By using logarithms, changes in variables can be converted into percentage changes (Stock & Watson 2015, 269). Moreover, log-transformed variables is a convenient way of transforming highly skewed variables into more normal distribution and reducing heteroskedasticity (Benoit 2011). The types of nonlinear regression used in this study are log-linear regression, where 1 unit change in Xi is related to a 1% change or 0.01 unit in Yi. And log-log regression

where 1% change or 0.01 unit in Xiis associated with 1% change or 0.01 unit change

in Yi (Stock & Watson 2015, 276). Hence, the study has logarithmized the outcome

variable that represents income inequality, Gini coefficient and the control variable real GDP per capita to reduce skewness and heteroskedasticity of the variables. The logarithmisation of the Gini index might sound unintuitive however, since we are interested in how income inequality is affected by the quality of government, exact value of the Gini index is not sought after since, the paper studies if the relationship is positive, negative or non-significant. Additionally, it reduces heteroskedasticity and skewness as previously mentioned above.

6.2.2 Regression analysis with panel data

(30)

is a pooled ordinary least squares model where observations in the different entities, i and the different time periods, t will not be controlled, hence the structure of the observations given entity and time is irrelevant (Wooldridge 2010, 191-193). Our second panel data analysis model is a fixed effects model, its main characteristic is that it controls for omitted variables in the panel data when the omitted vari-ables vary across the entities but don’t change over time (Stock & Watson 2015, 357). Moreover, compared to the pooled OLS model fixed effects regression has n different intercepts for each entity. These intercepts are represented as binary vari-ables and absorb influences of omitted varivari-ables that differ between entities but are constant over time (ibid.). Hence, the main regressions in the study are the following: Pooled ordinary least squared model:

Ln(Gini)it = β0+ β1QOGit+ β2Ln(GDP/capita)it+ β3Globalizationit+ εit (6.1)

Fixed effects model:

Ln(Gini)it = β0+ β1QOGit+ β2Ln(GDP/capita)it+ β3Globalizationit+ αi+ εit

(6.2) Where the outcome variable is Ln(Gini)it with entity i and time t, and the control

variables is represented with the coefficient β with entity i at time t. αi is the entity

fixed effects, lastly εit represents the error term at entity i, at time t. Since this

(31)

6.2.3 Autocorrelation and standard errors

In some instances, the regressions can be autocorrelated, i.e when the value of the outcome, Yi in one period is correlated with its own value in the next time period

(Stock & Watson 2015, 528). Then the heteroscedasticity robust standard error formula is not valid because it’s derived under the false assumption of no possible autocorrelation. To deal with this issue this study will utilize a clustered standard error, a type of standard error that is heteroskedastic and correlated over time called, heteroskedasticity and autocorrelation consistent standard error, HAC. Clustered standard errors allow for heteroskedasticity and autocorrelation within a certain entity however, it treats the errors as uncorrelated across entities. The clustered standard errors allow for heteroskedasticity and autocorrelation in a way that’s consistent with the second central assumption of fixed effects regression (Stock & Watson 2015, 367). Thus, clustered standard errors will be utilized for the fixed effects model.

6.2.4 F-test

To compare which of the statistical models presented above have the best fit for the given data set an F-test will be conducted. To be able to conduct the F-test the F-statistics will be used to test the joint hypothesis that all the slope coefficients are zero. The null hypothesis and alternative hypothesis are the following:

H0 = β1 = 0...βk = 0 vs H1 = βj 6= 0 atleast j, j = 1...k

(32)

6.3 Control variables

In the following section the control variables included in the regression models will be presented. The two additional control variables except the quality of government variable are selected given the relevance they have to the topic of income inequality.

6.3.1 Quality of government

The first regressor included in the study is the quality of government, measured by International Country Risk Guide. The quality of government variable is a mean value of the ICRGs variables corruption, law and order and bureaucracy quality. The corruption variable is the assessment of corruption within the political sphere, while law and order are assessed separately. Law is an assessment of strength and impartiality of the legal system and order is assessment of popular observance of the law (Dahlberg et al. 2021, 98). Bureaucracy quality represents institutional strength and quality of a government. The ICRG quality of government is measured from scale 0 to 1 where a high value represents better governance (Ibid.).

6.3.2 Real GDP per capita

(33)

6.3.3 Index of globalization

The last control variable included in the regression is an index of globalization col-lected by ETH Zurich. The globalization index is the weighted average of economic, social and political globalization. Economic globalization represents financial flows such as the trade of goods and services. Meanwhile, social globalization represents interpersonal contact, flows of information, internet access and cultural proximity. Political globalization contains variables focusing on memberships of international organisations and treaties. The index of globalization ranges from 0 to 100 where a higher value indicates a higher degree of globalization (Dahlberg et al. 2021, 64-65).

6.4 Descriptive statistics

In this section, all the variables included in this study is presented. Note that logarithmized Gini index and GDP per capita will be included in the table.

Table 6.1: Descriptive statistics

Obs Mean Std.Dev. Min Max

(34)

7. Results

In the following section the results from the panel data regressions are presented. The first table presents the results from the pooled OLS model and the following table presents results from the fixed effects model.

.

Table 7.1: Pooled OLS model

(1) (2) (3) Quality of government -0.563***(0.0261) -0.380***(0.0370) -0.314***(0.036) Ln(GDP/capita) -0.0534***_(0.00778) 0.0390***_(0.0118) Globalization -0.00777***_(0.000761) Constant 3.957***_(0.0162) 4.353***_(0.0599) 3.960***_(0.00695) Standard errors Standard Standard Standard

Observations 1429 1425 1425

R-squared 0.246 0.270 0.320

Adjusted R-squared 0.245 0.269 0.318

Standard errors in parentheses. *p <0.05, ** p <0.01, ***p <0.001

(35)

0.0261) hence, the effect of one additional unit increase in quality of government leads to -56.3% decrease in the Gini coefficient. Moreover the result is significant on a 99% level i.e we can reject the null hypothesis, which states that there’s no relationship between Gini index and quality of government on a 99% level. The constant also tell us that if quality of government would take the value 0, the logarithmized Gini coefficient would take the value 3.957 (SE 0.0162) its also significant on 99% level. Moreover by looking at, the R-squared and the adjusted R-squared value we can interpret how much of the variation in the outcome variable is explained by the control variables (Stock & Watson, 2015, 196-197). Furthermore, adjusted R-squared is generally used for multiple regression compared to the ordinary R-R-squared. The main difference is that when a new variable is added in a model, generally the R-squared value increases. If the R-squared would have been used to analyse the multiple regression an inflated estimate would be given (ibid.). Therefore, by using the adjusted R-squared we can correct, or in other terms deflate the value. In model (1) the R-squared value is 0.246 in other words, the control variable is able to explain 24,6% of the variation in the outcome variable.

In model (2) we control additionally for logarithmized real GDP per capita, by keeping it fixed we can see that the relationship between Gini coefficient and quality of government is somewhat reduced to -0.380 (SE 0.0370). On the other hand, if we keep quality of government fixed we can see that one percent increase in real GDP per capita leads to decrease of the Gini index by -0.0534% (SE 0.00778). According the relationship given we can see that the inverted U-hypothesis (Kuznets 1955) seem to hold given the results from model (2). However, the relationship is fairly weak. Moreover, we find that both results for quality of government and real GDP per capita are significant on a 99% level. Additionally, the constant takes on the value 4.353 (SE 0.0599) and it is significant on a 99% level. Since model (2) is a multiple regression we now interpret the value from the adjusted R-squared. The value given is 0.269 the two control variables are able to explain 26.9% of the variation in the dependent variable.

(36)

(37)

Table 7.2: Fixed effects model (1) (2) (3) Quality of government 0.186**(0.0636) 0.155*(0.0703) 0.152*(0.0697) Ln(GDP/capita) -0.0570*_(0.0229) -0.0483_(0.0352) Globalization -0.0004257_(0.00129) Constant 3.520***_(0.0372) 4.074***_(0.228) 4.022***_(0.282) Standard errors Clustered Clustered Clustered

Observations 1429 1425 1425

R-squared 0.918 0.921 0.921

Adjusted R-squared 0.910 0.914 0.914

Standard errors in parentheses. *p <0.05, ** p <0.01, ***p <0.001

In the table above the regression outputs from the fixed effects model are presented. In the model (1) similarly to the pooled OLS model, a bivariate model between the logarithmized Gini index and quality of government is showcased. The result illustrates the coefficient 0.186 (SE 0.0636). Thus it can be interpreted as, one unit increase in quality of government leads to an increase of the Gini coefficient by 18.6%. Moreover, the result is significant on a 95% level i.e we can reject the null hypothesis on a 95% level. The constant takes on the value 3.520 (SE 0.0372) and its significant at a 99% level. The R-squared value from the bivariate fixed effects regression is 0.918. Hence, 91.8% of the variation in the logarithmized outcome variable is explained by the control variable.

(38)

increases the Gini coefficient by 15,5%. On the other hand, if we keep quality of gov-ernment fixed we get the result -0.0570 (SE 0.0229) for real GDP per capita. Hence, we see that logarithmized real GDP per capita increases by one percent we see a -0.0570% decrease in the logarithmized Gini index. Furthermore, the variables are both significant at a 90% level. The constant takes the value 4.074 (SE 0.228) and is significant on a 99% level, it represents the value logarithmized Gini coefficient takes if all control variables take on the value 0. In addition, the adjusted R-squared gives the value 0.921. Thus 92,1% of the variation in the outcome variable is explained by the two control variables included in the model (2).

(39)

8. Analysis

The pooled OLS regressions showcase a negative relationship between quality of government and income inequality in all three regressions. Where in model (1) the coefficient value presented is -0.563 (SE 0.0261) in model (2) -0.380 (SE 0.0370) lastly in model (3) -0.314 (SE 0.0363). The effect seems to diminish when more additional control variables are included in the model. These results are on par with some of the previous studies made on the topic that there’s a negative relationship between the variables. Hung et al. (2020) conclude that there’s a negative relation between quality on government and income inequality through economic growth. Meanwhile, Huang and Ho (2018) points forth that there’s a similar relationship in developing countries in Asia.

(40)

Figure 8.1:

Graphical comparison of the bivariate regression pooled OLS and fixed effects model. The negative sloped line represents the pooled OLS regression, meanwhile

the positively sloped lines with separate intercepts is the fixed effects model regression.

topic of income inequality and governance and not as concluding results, it should instead illustrate the complexity of the field of the study. Moreover, to answer the theoretical hypothesis presented in the study we can conclude that we can reject the null hypothesis in all regressions in the pooled OLS and the fixed effects model, i.e we find a relationship between quality of government and income inequality given the results presented.

(41)

intercept shows that it has a better fit compared to the latter.

(42)

9. Summary

(43)

all continents. However, comparing the different outcomes of the studies should be taken with a pinch of salt since the studies oftentimes cover different geological areas, time periods and utilizes different methods and variables, such as different variables that measure quality of government differently. In other words, the results should illuminate the complexity of the topic of income inequality and governance rather than rejecting other studies results.

It should be noted that this study has some limitations. Firstly, since the data used is panel data from the period 1946 to 2020, it’s somewhat inconsistent across countries in the early period of the data, since the availability of observations are limited in many of the developing countries from that period. Secondly, since the measurement of quality of government and index of globalization is built on subjective perception, there’s a chance that the data poses a risk of interpretation or even bias, consequently raising questions of the internal validity and reliability of the study. On the other hand, external validity of the study can be considered to have high validity given the observations it includes throughout a broad time span across a multitude of countries from several continents. As previously mentioned, the data set might be somewhat limited in the early years, however, it shouldn’t affect the external validity given the wide data set.

Hence, despite some of the limitations presented above, the study argues that the panel data covers a sufficient time period which perhaps isn’t perfect, although it covers a broad time period across different states. In addition, the variables which are subjectively collected are captured through different indicators giving the variables a broader sense of notion and security from skewed interpretations. Moreover, major international organisations and universities are often in charge of collecting the variables. Therefore, the reliability of the data set and its variables included in the model should be on a high level.

(44)

Bibliography

[1] Andres, Antonio R. & Ramlogan-Dobson, Carlyn (2011) Is Corruption Really Bad for Inequality? Evidence from Latin America, The Journal of Development Studies, 47:7, 959-976.

[2] Bellu, Lorenzo Giovanni & Libertai, Paolo (2006) Inequality analysis the Gini index. EASYPol module 040. Food and agriculture organization of the United Nations.

[3] Benoit, Kenneth (2011) Linear regression models with logarithmic transforma-tions, Methodology Institute, London school of economics.

[4] Bergh, Andreas & Nilsson, Therese (2010) Do liberalization and globalization increase income inequality? European journal of political economy, vol. 26, issue 4, 488-505.

[5] Dahlberg, Stefan; Sundström, Aksel; Holmberg, Sören; Rothstein, Bo; Alvarado Pachon, Natalia; Mert Dalli, Cem (2021). The Quality of Government Basic Dataset, version Jan21. University of Gothenburg: The Quality of Government Institute, http://www.qog.pol.gu.se doi:10.18157/qogbasjan21

[6] Dabla-Norris, Era; Kockhar Kalpana; Suphaphiphat, Nujin; Frantisek, Ricka; Evridiki, Tsounta (2015) Causes and consequences of income inequality: A global perspective. International monetary fund, Staff Discussion Notes No. 15/13.

(45)

[8] Karlsson, Martin; Nilsson Therese; Lyttkens, Carl Hampus; Leeson, Geroge (2009) Income inequality and health: Importance of a cross-country perspective. Social science medicine, vol. 70. 875-885.

[9] Keeley, Brian (2015) How does income inequality affect our lives? in Income Inequality: The Gap between Rich and Poor. OECD Publishing, Paris.

[10] Kuznets, Simon (1955) Economic growth and income inequality, The American Economic Review, Vol. 45, No. 1 (Mar., 1955), pp. 1-28. American economic association.

[11] La Porta, Rafael; Lopez-de-Silanes, Florencio; Shleifer, Andrei; Vishny, Robert (1999) The Quality of Government. Journal of Law Economics and Organisa-tion. 15, no. 1: 222–279.

[12] Lindvall, Johannes & Teorell, Jan (2017) State capacity as power: a conceptual framework. Lund University: The department of political science.

[13] Hung, Nguyen Thanh; Hoang Yen, Nguyen Thi; Minh Duc, Le Doan; Ngoc Thuy, Vo Hoang; Thanh Vu, Nguyen (2020) Relationship between government quality, economic growth and income inequality: Evidence from Vietnam. Co-gent Business Management, 7:1.

[14] OECD (2012) Reducing income inequality while boosting economic growth: can it be done?, Economic policy reforms 2012. going for growth. OECD Publishing, Paris.

[15] Petersen, Thieß Schoof, Ulrich (2015) The impact of income inequality on economic growth, Bertelsmann Stiftung. Impulse 2015/05. UNSPECIFIED. [16] Pickett, Kate E. & Wilkinson, Richard G. (2014) Income inequality and heath:

A causal review. Soc Sci Med. 2015. 316-326.

(46)

[18] Vieira, Sergio (2012) Inequality on the rise? An assessment of current available data on income inequality, at global, international and national levels. United nations, department of economic and social affairs, economic affairs officer. [19] Stock, James H. & Watson, Mark W. (2015). Introduction to Econometrics.

Introduction to Econometrics. Harlow: Pearson Education.

[20] United nations (2020). World social report 2020, inequality in a rapid changing world. Department of economic and social affairs.

[21] Van de Werfhorst, Herman G. & Salverda, Weimer (2012) Consequences of economic inequality: Introduction to a special issue, International Sociological Association Research Committee 28 on Social Stratification and Mobility, 377-387.

[22] Vivarelli, Marco & Meschi, Elena (2007) Globalization and income inequality. IZA DP No. 2958. IZA disscussion paper series. Institute for the study of labor. [23] Wooldridge, Jefferey M. (2010) Econometric analysis of cross section and panel

(47)

Appendix

Figure 1: Gini index histogram

Figure 2:

(48)

Figure 3:

Real GDP per capita histogram

Figure 4:

Governance and income inequality

Department of political science