Populism and neoliberalism : a cross-country analysis of the effect of neoliberal policy on increased support for left and right-wing populism in Europe

(1)

Populism and Neoliberalism

A cross-country analysis of the effect of neoliberal policy on

increased support for left and right-wing populism in Europe

Thomas van der Meiden

Presented for the degree of

Bachelor of Economics

June 26, 2018

Faculty of Economics and Business Roetersstraat 11

1018 WB Amsterdam The Netherlands

Supervisor:

Mr. C.W. (Kees) Haasnoot

Project duration:

April 2018 - June 2018

(2)

Statement of originality

This document is written by Thomas van der Meiden who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

Abstract

The emergence and rise of populist parties all across Europe has increasingly disrupted politics. What caused the rise and emergence? Next to cultural explanations, there are economic factors such as income inequality posed as cause. In this thesis, it is argued that economic inequality is a concept deemed too narrow for explaining the rise of left and right-wing populism. Neoliberalism as a whole causes multiple inequalities and could therefore be able to partly explain populism emergence. The main question is: what is the effect of neoliberalism increased dominance in political practise on support for populist parties? This relation is analysed with panel data for 27 European countries between 1980 to 2015. Election results as dependent variable are predicted by a constructed proxy for neoliberalism. Neoliberalism is found to negatively influence the vote share of moderate parties and positively influence extreme right-wing parties. For left-wing populist parties no strong relation is found.

(4)

List of Figures

1.1 Income inequality . . . 7

2.1 Vote share of moderate parties (4 year average) . . . 15

2.2 Vote share of extreme left and right parties (4 year average) . . . 16

2.3 Moderate parties in six countries . . . 16

2.4 Extreme left and right parties in six countries . . . 17

2.5 Distribution of variables . . . 20

2.6 Distribution of proxy for neoliberalism . . . 20

2.7 Neoliberal policy in six countries . . . 22

2.8 Average neoliberal policy in all countries . . . 23

(6)

List of Tables

2.1 Correlation matrix of proxy components . . . 21

2.2 Summary statistics of control variables . . . 24

3.1 Regression results for moderates . . . 27

3.2 Regression results of extreme left and right . . . 29

3.3 Results for moderates with different border values . . . 31

3.4 Results for extreme left with different border values . . . 32

3.5 Results for extreme right with different border values . . . 33

A.1 Countries in regression models . . . 36

A.2 Parties classified as < 2 . . . 37

A.3 Parties classified as > 8 . . . 38

B.1 Moderates with different border values . . . 41

B.2 Extreme left with different border values . . . 41

(7)

Introduction

The rise of populism with emergence and growing support of right-wing populist parties in mainly Europe reached a climax with Donald Trump’s unexpected win in the 2016 US presidential election. Populists seem to combine a loose mixture of policies (Finchelstein, 2017), but especially populisms appeal to xenophobia and nationalism leads to fierce criticism in public debate. Most scholars agree xenophobia and nationalism are only part of populisms emergence (see for example (Williams, 2017)). A turn to populism for economic have-nots or a cultural backlash are two prominent suggested hypotheses, but results are not decisive (Inglehart & Norris, 2016).

Current debate in political sciences and philosophy cover the rise of populism in western democracies for years now. Abromeit (2017) concludes in a literature review that an interdis-ciplinary approach to populism is necessary. If so, part of this approach should be economic. However, somehow this phenomenon is overlooked by economists. Only since the election of Trump are economists starting to look for economic factors that could explain the rise of pop-ulism (see for example (Autor, Dorn, Hanson, & Majlesi, 2016)).

Growing inequality is in political scientific research often posed as a key economic factor leading to populism. However, empirical research rejects this hypothesis (S. M. Han, 2015). Sociologists and philosophers take a more theoretical approach and find inequality to be one of the core features of the principle of competition (Davies, 2016; Brown, 2015). Davies and Brown argue inequality does not only entail income, but also culture, education, politics and a multitude of other inequalities. Neoliberalism, according to them, stimulates, promotes and celebrates the competition leading to these varieties of inequality. At the same time, Davies and Brown argue, has neoliberalism become increasingly dominant in political practise. Neoliberalism, therefore, has became a suspect in the debate on populism. Furthermore, since this debate is mainly held in political fields of research instead of economics, one could safely say that neoliberal and neoclassical economics is under attack from all sides in the debate on populism.

Neoliberalism was posed firstly and most importantly as alternative to fascism and commu-nism (Hayek, 1945). Historically seen, this seemed successful. But while right-wing populists are being compared to fascists in public debate (Paxton, 2016), communist parties are re-emerging as well, for example in Greece and Spain. Text-book neoliberal policies such as international trade agreements and economic unification exemplified by the EU and Euro are increasingly

(8)

under threat, mainly from populist parties. This poses the question, is the increasing dominance of neoliberalism a factor in the growing support for populist parties? And complementary, is neoliberalism a factor in the re-emergence of the extreme left or right?

To answer these questions, a panel regression is done with parliamentary election results of European countries in the period between 1980 and 2015. The main explanatory variable is a created proxy for neoliberalism consisting of a trade component, financial market freedom and government size. Lastly there are several variables controlling for other explanations of the rising support for populism. Neoliberal policies such as market freedom and a small government are promoted for resulting in economic growth, but at the same time causing the variety of inequalities (Brown, 2015).

The theory section focusses firstly on the debate of populism and suggests antagonism theory as workable channel through which neoliberalism can lead to the emergence of populism. Secondly neoliberalism in theory and practise will be defined, and several criticisms and critiques will be discussed. The methodology focusses on preparing the variables for the analysis and focuses in particular on The Netherlands, Germany, United Kingdom, Spain, France and Greece. These countries are chosen due to their fundamentally differing political and economic situation. The analysis consist of an OLS regression to answer the main question.

This analysis aims to academically distinguish itself in two ways, firstly by analysing populism in data according to the economic methodology and therefore contributing to an interdisciplinary approach on populism. Secondly, next to the overlooking of populism, neoclassical economics as supporting scientific discipline of neoliberalism is increasingly being criticised for lacking general reflective analysis (Earle, Moral, & Ward-Perkins, 2016). This thesis aims to combine these two aspects, by analysing neoliberalism with an economic methodology in the context of populism.

(9)

Chapter 1

Literature review

1.1 Populism

According to Ernesto Laclau (2005), populism is described as a mixture of ideas containing firstly, anti-establishment rhetoric whether against politicians, scientist, banks, multinational corporations, or the state in general. And, secondly that populist leaders claim to speak for a homogeneous population, whether the nativist population, the 99%, or the people in general (Laclau, 2005). While left-wing populism is most often a resurgence of communist and socialist

politics, right-wing populism is a relatively new phenomenon in western democracies.1 _Cas

Mudde (2007) argues that right-wing populism contains, next to speaking for the people and being anti-establishment, a strong leader who appeals to authoritarian values. These general definitions will be used in this thesis.

The question remains, how to interpret right-wing populism? Historian Robert Paxton argues that right-wing populism could be in an early stage of development towards fascism (Paxton, 2007, 2016), while historian Federico Finchelstein found several historical distinctions between populism and fascism and argues that populism operates within democracy and becomes only fascism when it embraces violence and tries to destroy democracy (Finchelstein, 2017). Roger Scruton on the other hand, argues that populism enriches democracy by the new ways it repre-sents people who, untill then, were not represented (Scruton, 2017).

Next to these interpretations of right-wing populism, there are more explanatory approaches. The aspect of xenophobia in populist rhetoric could be explained by Sartre’s notion of Othering

(Sartre, 1995).2 When people, according to Sartre, choose to think emotionally instead of

ra-tionally, it becomes possible to romanticise race, culture, history et cetera and to patronize and

1_{Right-wing populism appeared firstly in South America after the Second World War, but emerged only in the}

late ’90 in western democracies (Finchelstein, 2017).

2_{Although Sartre posed Othering and the accompanying theory in The Jew and the Anti-semite (1944) as an}

explanation of anti-Semitism in Europe around the Second World War, it still provides a useful framework for analysis of right-wing populism in for example on anti-Islam, anti-Mexican, and anti-immigration rhetoric, as well as for the re-emergence of romanticisms of race, culture and history.

(10)

discriminate others. Sartre’s analysis is used to explain Europe’s turn to the right (Wolin, 2011).

Another approach makes use Nietzsche’s notion of ressentiment 3 _{(Nietzsche, 1967).}

Ressenti-ment of disadvantaged social groups and class towards the political establishRessenti-ment and cultural elite is viewed as a possible explanation for the emergence of populism (Berns, 2006).

Ronald Inglehart and Pippa Norris (2016) state that increasing resistance towards progres-sive values are part of the explanation for the emergence of populism. They argue that with the emergence of green and ecologist parties from the 1970’s a counter movement might take place in the form of right-wing populism. An aversion for migrants, refugees, guest-workers and xenophobia are related to this counter movement of mostly older men. Next to these aspects this counter movement includes nationalist elements, anti-EU or anti-UN sentiments and high valuations of authoritarian leadership. Although they think this counter movement is only part of the answer, they still include it in their empirical analysis as the cultural backlash hypothesis. Since they found results supporting this hypothesis, it should be controlled for in an analysis of economic factors in relation to populism.

A more structural analysis of the emergence of populism is taken by political theorist

Chan-tal Mouffe (2015). For Mouffe, as for Carl Schmitt4, politics is fundamentally an antagonism

between friend and foe. According to Schmitt, politics is the continuing of war between par-ties without the actual fighting on a battlefield, but with fighting in a political arena (Schmitt, 2008). He argues that only in this antagonistic, or warlike, conception of politics it is possible to form collective identities. Mouffe argues that Schmitt’s conception of politics presupposes a homogeneous population which could destroy plurality, as has been horrifyingly demonstrated by Nazi-Germany. Therefore, according to her, it is necessary for political parties to acknowledge the legitimacy of the political enemy, by construction an us–them relation instead of a friend– foe relation (Mouffe, 2015, p.27). Put differently, it is necessary in a democracy to accept the differences of people and their opinions instead of trying to extinguish people who are and think differently. In Mouffes conception of politics it becomes possible to agree to disagree, and at the same time to form collective identities.

Mouffe argues that in modern-day liberal democracies there is a lack of this kind of politics. Consensus forming and compromises between classical left and right orientated parties results in an absence of the us–them politics that is necessary to form collective identities (Mouffe,

3_{Nietzsche uses ressentiment in On the genealogy of morals (1887) as historical analysis how we think in terms}

of good and bad. Ressentiment is a form of reproach of some social group towards another based on moral and value judgements.

4_{Carl Schmitt is a jurist and political theorist whose work The concept of the political (1932) has contributed}

(11)

2015, pp.68-69). Differently said, for voters, there is not enough to choose from. This lack of politics in liberal democracy created the opportunity to form new political parties as antagonism against all classical parties altogether (p.71). The reaction of the political centre towards these newly emerged parties was moralistic and full of disapproval, which in turn enforced the us-them antagonism further to the benefit of the populist parties (p.72). This is how, according to Mouffe, populism was able to emerge and become successful in Europe, the United States and in various countries in South America.

Mouffe’s structural analysis of the emergence of populism in modern liberal democracies does not state what values or policy the populist program entails. Populism could incorporate various critiques on the political centre from xenophobic, ressentiment to economic motivations. At the same time antagonism theory is able to provide a rationale for left-wing populism. So, criticism on neoliberal policy that unite people as an antagonism to the political centre might explain the emergence of populism.

1.2 Neoliberalism, criticism and critique

This section firstly poses a workable definition of neoliberalism and exemplifies this with neolib-eral policies. Secondly, criticism such as increasing inequality by neolibneolib-eralism and capitalism will be discussed in the debate on populism. Lastly, it is argued that these criticisms are not sufficient to explain the emergence of populism as antagonisms. Sociological and philosophical critiques on neoliberalism are posed as possible explanations to form antagonisms and therefore create populism.

Neoliberalism

This thesis uses a broad definition of neoliberalism5 posed by philosopher Michel Foucault in

his 1979 lectures Birth of Biopolitics (Foucault, 2013). He states that neoliberalism entails

a rationality which is highly influenced by Friedrich Hayek and the Chicago-School economists (Foucault, 2013). Foucault states that this rationality, firstly, views an individual as ‘entrepreneur of himself’, to increase his human capital, which results in income, and to maximize his utility

5_{Although this definition of neoliberal policy might seem radical, in a debate between Chicago-School}

economist, and economic Nobel-price laureate Gary Becker with Fran¸cois Ewald – who was Foucault’s princi-ple assistant at the time of writing the Birth of Biopolitics and defender of his work after Foucault’s 1984 death – agree on most of Foucault’s analysis of competition, human capital and neoliberalism (Becker, Ewald, & Har-court, 2012). Although Becker might not represent neoliberalism on its own, agreeing indicates some validity to a non-economist defining an economic concept.

(12)

given his income constraints. This individual is free to do as he likes, but increasingly becomes bearer of the (economic) risk and responsibility of his actions (Rose & Miller, 2008). Secondly,

Foucault argues that neoliberalism views government and various institutions6 _{as protectors}

and promoters of market competition between firms and between individuals (Foucault, 2013). Thirdly, he notes, that any form of planning by the state should be avoided, and only the market can direct processes. Fourthly he claims that neoliberal institutions focus on economic growth as the main way to increase welfare for everyone, not by interfering in the market, but by improving the conditions of the market.

Neoliberal policy aims to increase economic growth by a variety of measures and tactics, this thesis focusses on market freedom, the size of government, and trade liberalisation. Hayek argues, against social planning by communism and socialism and to prevent fascism, that the best way to make use of societies division of knowledge among different people is to free the market processes from state control (Hayek, 1945). He argues that the market process is able to combine individuals dispersed bits of knowledge and their interests as if guided by Adam Smith’s invisible hand which will ultimately result in the best outcome for everyone, specifically economic growth. Thus, Hayek’s policy recommendation is twofold. Firstly, liberate markets and secondly reduce the size of the state. Furthermore, policy for trade liberalisation as a specific form of market liberalisation is characteristic for neoliberalism. This is propagated by almost all economists and important supranational institutions such as the WTO, EU and UN.

Next to these three aspects, there are several other aspects and examples of the extensive reach neoliberalism has on political practise. An example would be Buchanan’s (1980) theory of public choice. He criticised government representatives who would have incentives to lower taxes and spend to much for the favour of voters. Such policies can be neutralised in practise by tax and spending limitations (Buchanan & Tollison, 1984). An example of such spending limitations are the the 3% deficit rule and 60% debt rule formulated in the Maastricht Treaty. Policies such as austerity, are captured by the size of government because austerity will result in a declining, or smaller growing government. Therefore by focussing on financial market liberty, trade freedom and size of the state various different neoliberal policies are covered in a general way.

6_{Foucault uses a broad definition of the state.} _{He perceives the state as the totality of institutions that}

govern in various ways with the goal of welfare for the population (Foucault, 2007a). Such institutions are the government, but also schools, universities, hospitals, prisons, ngo’s and supra-national institutions.

(13)

Criticism on neoliberalism

Since Thomas Piketty’s Capital in the Twenty-First Century (2013), economic inequality is a fiercely debated topic (Piketty & Goldhammer, 2014). Piketty argues that if wealth is dis-tributed unequally in society and the gap between the share of return on capital and economic growth is growing, wealth inequality increases (Piketty, 2015). Next to wealth inequality, income inequality is known to have negative effects on economic growth in theoretical models (Alesina & Rodrik, 1994), which is confirmed empirically by a recent IMF study (Dabla-Norris, Kochhar, Suphaphiphat, Ricka, & Tsounta, 2015). In the debate on the rise of populism, economic in-equality is posed as an important explanation expressed in the inin-equality hypothesis (Inglehart & Norris, 2016).

Figure 1.1: Income inequality

Figure 1.1 shows the disposable income development for five countries from 1980 to 2015

measured by the Gini-coefficient (Solt, 2016). Note that income inequality increases in the

Netherlands from 1998 up to 2008. In this period right-wing populist parties List Fortuyn

and the Freedom Party emerged in respectively 2002 and 2006. In France with Front National’s emergence in 1993, there does not seem to be a relation with income inequality, but in the second wave of Front National growing support under Marine Le-Pen’s 2011 leadership could suggest

(14)

some connection. On the other hand, the stark increase of inequality in the United Kingdom due to liberalisations from 1980’s onward did not lead to political polarisation. Furthermore, Germany’s rising inequality also did not lead to populism only until the refugee crisis with the emergence of right-wing populist party the AFD (2013). In Spain the higher and fluctuating inequality does not seem to have caused populism, while in Greece inequality decreases and support for left and right-wing populism increased. Sung Min Han (2015) only found a link between income inequality and populism in some countries and a weak link overall. This result is consistent with other studies by Kyong Joon Han (K. J. Han, 2016) and David Jesuit et al. (2009). They only found a significant result if income inequality is interacted with individual income, or with immigration levels.

Next to inequality there are several economic factors proposed as explanatory for the rise of populism. Increased job-insecurity caused by liberalisation of labour markets gives some result (Jesuit et al., 2009). Manuel Funke et al. find that support for radical right-wing parties increases after financial crises, while support does not increase after economic crises (Funke, Schularick, & Trebesch, 2016). Lastly David Autor et al. found in an unpublished paper strong evidence of increased competition by foreign companies due to trade liberalisation as explanatory for the increased political polarisation in the United States (Autor et al., 2016).

Neoliberalism, or the market economy, are criticised for causing growing inequality, job in-security and causing financial crises and therefore being a factor in the emergence of populism. Some connection between economic factors and political polarisation is found, but the main hypothesis of political research, the inequality hypothesis, is empirically not strongly proved. The inequality hypothesis poses questions to what extend economic factors could explain the emergence of an entirely new form of politics, namely populism. For inequality, job-insecurity, or other economic factors to explain the emergence of populism, thinking in line with Mouffe, antagonisms need to be formed. As will be argued in this thesis, inequality or other economic factors are too narrow concepts to form antagonisms on itself. However, neoliberalism as a whole might be capable of forming antagonisms to the political centre and therefore play a key role in the emergence of populism.

Neoliberal critique

Structural analyses of neoliberalism and it’s supporting academic paradigm neoclassical eco-nomics are in academia only scarcely done within the discipline of ecoeco-nomics (Earle et al., 2016). The following section therefore covers mostly sociological and philosophical critiques on

(15)

neolib-eralism and inequality in specific.

Davies states that neoliberalism is expanding the domain of the market, but at the same time it is expanding the market principles beyond the original domain of markets (Davies, 2016, p.123). He argues that neoliberalism judges individuals as if competing in a market. Examples of this can be found in education (Nadesan, 2006), analyses of criminal behaviour in terms of cost-benefits (Foucault, 2013, pp.323-334), and the perception of the migrant as investor in himself (Foucault, 2013, p.302).

Political theorist Wendy Brown (2015) argues that the central principle of neoliberal ratio-nality is competition. She notes that competition, by nature, is the struggle of being better than your competitor, and that modern day economic thinking applies not only to companies or countries, but to humans in the form of human capital as well. Sociologist Williams Davies (2016) states that if a society celebrates and encourages competitiveness as an ethos, whether in sports, politics, education or the economy, one must not be surprised if results are highly un-equal. So, the neoliberal principle of competition in markets, and the extension of competition to the domain of the individual will produce a multitude of inequalities in all domains of society. This poses the question, how to legitimize an unequal result of competition? Davies notes that liberal politicians such as Bill Clinton and Tony Blair legitimise unequal outcomes, if the entering of competition is relatively equal, then the unequal outcome is meritocratic, thus the result of talent (Davies, 2016, pp.40-47). This conception of justice formulated by Rawls (2009) where upfront equality legitimises unequal outcomes is central in liberal defence of competition. Amartya Sen pointed to the necessity of equality in his criticism on utilitarianism in neoclassical economics (Brue & Grant, 2013, pp.449-451). He argued that maximising individual utility could result in a transfer of income from people with a constricted utility function, by being sick or disabled, to healthy people. So, according to Sen, some equality is necessary in utilitarianism and thus in the neoliberal focus on utility maximization.

Davies, however, argues that when the unequal outcome in one market results in the unequal entering of another market, this claim to meritocracy and legitimacy is not justified (p.40). So when income inequality increases, upfront inequality in various other markets increases and thus the unequal outcome increases even more. Education provides a clear example, if income inequality increases, it could become harder for some people to afford tuition fees for higher education, which in turn results in lower income and thus income inequality increases further. Empirical research by Newman et al. found support for a meritocracy to be decreasing when income inequality increases (Newman, Johnston, & Lown, 2015), which is in line with Davies

(16)

argument.

Next to these varieties of social inequalities, Brown points to a specific form of inequality prop-agated by neoliberalism. She argues that, in practise, neoliberalism transfers political processes to decision making by professionals (Brown, 2015, pp.122-131). She states that deliberation about common values and ends and accountability of politicians on their policies in a democracy are increasingly being replaced by governance. Brown notes that governance, which is a decision making process that is goal orientated, has become the main administrative form of neoliberal-ism (p.122). Brown concludes that replacing politics by technocratic leadership, neoliberalneoliberal-ism is slowly undoing democracy (p.212). This transferring creates political inequality between citizens and businesses. Big business can access process making through lobbying, while for individuals this is hard. An example of this is the technocratic decision making in the European Union surrounded by over ten-thousand lobbyists.

Brown’s argument of the decreasing politics in neoliberalism is parallel to Mouffe’s objection to compromises and consensus forming in liberal democracy. Neoliberal governance reduces pol-itics in its us–them concept and creates the possibility for an antagonism to be formed. Along side this process, is increasing social inequality creating the necessity for a change in policy. Eco-nomic inequality alone might thus be a too narrow concept to capture the variety of inequalities. Economic inequality could decrease due to economic growth, while other forms of inequality increase due to extending competition. This could provide a reason why economic inequality is empirically not strongly connected to populism. If the state, and the traditional parties pro-tect, legitimize and extend competition, which is the cause of the inequalities, individuals could unite against the traditional centre parties. Neoliberalism propagated by centre parties, could therefore be the basis of the antagonism that resulted in the populist parties.

(17)

Chapter 2

Methodology

In this section, the methodology is explained for empirically testing the relation between ne-oliberalism and populism in a cross-country analysis. Firstly, the hypothesis and the empirical model for the ordinary least squares regression are stated. Secondly, the description of variables used in the regression are discussed. In particular the measurement of populism and the proxy created as measurement of the integration of neoliberal policy. The last section describes the data and particularly the development of populism and the proxy for neoliberalism seen in the data.

2.1 Hypotheses and empirical model

To test the relation between neoliberalism and populism a cross-country regression analysis is done with panel data. Neoliberalism was first experimented with in Chile from 1975, but with the elections of Thatcher in 1979 and Reagan in 1981 neoliberal policy became increasingly dominant. The analysed period therefore starts in 1980 and runs to 2015. The cross-country analysis consist of 27 countries. The dataset consists of European countries including the United Kingdom and Turkey. The main reason to analyse only European countries is due to data availability of the European Value Survey, which includes only European countries. A detailed listing of countries is provided in appendix A.1. Two countries are dropped, appendix A.5 provides a graphical overview. Malta has a vote share of 100% over the entire period and has therefore no variation in political composition. Cyprus who has not even 40%, disturbed averages too much. All other countries have variation in political composition and the lowest result for moderates is around 60% of votes. To prevent these relatively small countries from having too much impact on the data, they are dropped.

To formalize the main question, to what extend neoliberalism is related to populism, three hypotheses are formed. The first, and main hypothesis of this thesis aims to answer the question what effect neoliberalism has on the vote share of moderate parties. The second and third

(18)

hypothesis focus on the effect of neoliberalism on extreme left and extreme right parties. (1) Neoliberal policy leads to a decrease of the vote share of moderate parties

H0 : β neoliberalism(i,t)= 0

H1 : β neoliberalism(i,t)< 0

(2) Neoliberal policy leads to an increase of the vote share of extreme left parties

H1 : β neoliberalism(i,t)> 0

(3) Neoliberal policy leads to an increase of the vote share of extreme right parties

H1 : β neoliberalism(i,t)> 0

To answer these hypotheses an OLS regression is done. The dependent variable is the sum of votes by either extreme left-wing parties, moderate parties or extreme right-wing parties. The proxy for neoliberalism is the main variable of interest. The following model is formulated:

X

V oteshare(f,i,t,)= β0+ β1· neoliberalism(i,t)+ β2· voteshare(f,i,t−1)+

β3· culturalbacklash(i,t)+ β4· f ederalism(i,t)+ ε

TheP vote share is the aggregate vote share of a political fraction (f), either extreme left-wing

(f=1), moderate (f=2), or extreme right (f=3), for country (i), in year (t). β0is the constant and

intercepts with the y-axis. Neoliberalism is a proxy which will be extensively discussed in the next section for country (i), in year (t). Vote share of the previous election is used as control variable for country (i), in year (t). The cultural backlash thesis as discussed is formulated as a vector of control variables. Immigration policy for country (i), and year (t), and three time-independent variables depending only on the country (i). These three are attitudes towards authoritarianism, nationalism and international institutions. Lastly, the dummy variable federalism controls for institutional systems for country (i), in year (t). If federalism equals 1, this country has either one or more regions with political autonomy in taxation, law-making and relative policy freedom. Controlling for federalism is in line with Han’s and Han’s analysis of populism (S. M. Han, 2015; K. J. Han, 2016). The control variables are extensively discussed in section 2.4.

(19)

2.2 Populism in data

This section covers the methodology required for an empirical analysis of populism. The first subsection covers the measurement of populism and specific methodological choices for measure-ment. The second describes election results in The Netherlands, Germany, United Kingdom, Spain, France and Greece.

Measurement of populism

For analysis the Parlgov dataset is used (D¨oring & Manow, 2015). This data contains election

results from 1945 until 2015. Election results are used from 1980 onward. The analysis focusses on the lower house, the main reason for this is that the lower house is directly chosen in most countries, while the higher house is indirectly chosen in many countries and might therefore reflect voting differently, and secondly the lower house has more means to influence policy in most countries. Thirdly, in most democracies the administration is accountable to the lower house. This poses a problem, elections are not held regularly since the maximum term in each countries differs and an administration might resign before the term ends. Another option to measure voting behaviour would be to use survey data which is done for example by K.J. Han and S.M. Han. This creates the possibility to analyse on a micro or individual level. However, this can also be problematic because voters change their answer in surveys towards winning parties instead of their actual vote (Weir, 1975). Secondly, people conceal their voting behaviour if society perceives a particular party as unwanted, which is especially the case with populist parties (Knigge, 1998). Therefore, this thesis analyses the actual election results with interrupted time intervals, and consist thus of a macro-analysis.

The second methodological choice is about how to decide which party is conceived as populist. This immediately poses the question where to place each particular party. Mudde’s definition of a right-wing populist party entails nativism and authoritarianism and he classifies parties on these criteria (Mudde, 2007; Mudde & Kaltwasser, 2012). Ignazi, however, characterises right-wing parties as nationalist and therefore classifies parties differently (Ignazi, 2003). But these authors focus mainly on right-wing populism and not on the left-wing variant. Therefore, the

traditional left-right scale is used in this thesis. Holger D¨oring and Philip Manow, who build

the Parlgov dataset, have classified all partaking parties from 0 to 10 in the traditional left-right scale. Where 0 means extreme-left, mostly communist parties, and 10 means extreme-right, which entail populist parties but also some extreme conservative parties. The main problem this

(20)

poses is that some conservative parties could be classified as populist as well and vice-versa. The Dutch SGP party exemplifies this problem. While this party is a Christian reformed party, it is qualified as populist due to it’s extreme conservative ideas. Even when taking Mudde’s difinition in account, the SGP appears as populist party in classifications, see e.g. (Inglehart & Norris, 2016, p.44). Although the definitions of populism from Mudde and Ignazi are more narrow, it does not result in solving many problems in terms of classification. Therefore is decided that to

prevent analysing each party separately, the left-right scale by D¨oring and Manow is used.

This classification, ranging from 0 to 10, classifies each party the first time they participated in a parliamentary election. Every party is time-invariantly scaled based on expert judgements (Castles & Mair, 1984; Huber & Inglehart, 1995; Benoit & Laver, 2006; Bakker et al., 2015). This creates a problem since parties move on the political spectrum over time. Therefore an assumption is made that although parties move, they would stay relatively close to their initial classification. This assumption can be realistically made since it is unlikely moderate parties become populist, it is more likely that a newly emerged party forms an antagonism to the political centre.

A variable is constructed with parties classified below 2 on the left-right scale as extreme-left, between 2 and 8 as moderates, and above 8 as extreme-right. This variable will be the dependent variable in the regression analysis. A detailed list of extreme left-wing parties is found in appendix A.2, and extreme right-wing in A.3. The border values of 2 and 8 are arbitrary, therefore the regression results will be related to results following from different border values, specifically of 1 and 9, 1,5 and 8,5 and 2,5 and 7,5. To describe the dataset, the border values of 2 and 8 will be used for now. Generally accepted right-wing populist parties such as Front national is classified as 9.7, the Dutch Freedom party as 8.8 and the German AfD as 8.7. On the left end below 1.5 mostly communist parties are found, and between 1 and 2 socialist and ecological parties. Examples are Die Linke in Germany as 1.2, Podemos in Spain as 1.3 and Canada’s Green party as 1.9.

The share of votes of each party per classification is aggregated, resulting in an election result for extreme left, moderates and extreme right. Combine the left share, moderate share and right share and most elections reach a total of 100%, some elections only reach 90%. There are two reasons why the aggregate result does not add up to 100%. The first reason is that parties temporarily break up or form in an election and eventually form or separate again, such as Listes Communes in 1986 in France and are therefore not classified. The second reason is one-issue parties who dissolve after a short period and are also not classified. A total of 13 elections are

(21)

dropped with a result lower than 90%, containing 1986 in France, Ireland in 2006, Israel in 1981 and 2015 and eastern European countries mostly around 1990.

Populism in the data

Figure 2.1 shows the average vote share of moderate parties, thus between 2 and 8. Figure 2.2 shows for extreme left, thus below 2, and extreme right, above 8. The bars are four year averages over all countries in the data. The four year average is taken because most countries have a four year term period, so almost every country is at least once in every bar. If a term ends earlier, for example because an administration resigns before end of term, than that country will be overrepresented in the bars, but overall no problems are expected. The share of moderates is declining from 1990 onward by over 6%. Figure 2.2 shows that the extreme left increases almost 2% since 1990, but the decline of moderates is mainly caused by the over 3% increase of the extreme right. The data shows the growing support for populist parties, and decline of moderates.

(22)

Figure 2.2: Vote share of extreme left and right parties (4 year average)

(23)

Figure 2.4: Extreme left and right parties in six countries

Figure 2.3 displays the vote share of moderate parties in six countries and figure 2.4 shows the vote shares of extreme left and right parties, based on the Parlgov database. In The Netherlands, the left-wing parties have been steadily increasing since 1990 and on the right side populist parties in 2002 and from 2006 onward are clearly visible. Moderates lose around 15% of their votes. In Germany, the almost 20% decrease in vote share of moderates since the 1980’s is mainly due to left-wing party Die Linke and in the end the addition of right-wing party the AfD. The almost 100% vote share of moderates in the United Kingdom is misleading. Since the Labour party is not classified as extreme left, and UKIP with 7.8 not as extreme-right, these effects cancel out. This strengthens the necessity to analyse the results eventually with a variety of border values. In Spain, moderate vote share is stable around 95% until extreme left-wing parties such as Podemos reached 20% in 2015. While most countries have had high moderate shares since 1980’s, in France moderates account for around 75%, but this share is not declining much. Note that the election of 1986 is left out due too missing classifications, the line therefore is connected between the 1981 and 1988 elections. Clearly visible in figure 2.4 is right-wing party Front National’s 1988 entrance and the slow decline of extreme left. Greece’s high moderate vote share declines slowly but firmly due to left-wing communist and the right-wing orthodox populist party’s win in 2012, probably connected with the Greek debt-crisis and various social reforming enforced by the EU. Overall, a decline in vote share of the moderate parties is seen in the data. The Parlgov dataset shows the basis for the growing concerns around populism.

(24)

2.3 Neoliberalism in data

In this section the methodology is described of creating a proxy for neoliberalism focussing on market liberty and government size. The first section consist of constructing the proxy and the second section consist of analysing the proxy in specific countries.

A proxy for neoliberalism

Neoliberalism is the main variable of interest and posed as being capable of creating antagonisms in the form of populism. To empirically investigate the possible relation between populism and neoliberalism some form of measurement is required to capture the variety of neoliberal policies. An overall index of neoliberalism does not exist so far, therefore a proxy is created to reflect the state of implemented neoliberal policy in a country. This poses the question, what measurable variable or variables accurately reflect neoliberal policy the most?

In brief, neoliberal policy consist of liberating markets and a decreasing government. Liber-alisation can be achieved in a multitude of markets such as markets for trade, financial markets, the labour market, housing market et cetera. An index for financial markets and trade open-ness as measurement for markets for trade are used as variables in the proxy for neoliberalism. Markets where individuals interact, such as the labour market and housing market are usually very complicated due to regulations and previously implemented policy. It is therefore hard to measure these markets and compare them with other countries. Therefore, a methodological choice is made to focus on financial markets and markets for trade instead of markets where in-dividuals interact. This implies an important assumption, if market liberalisation or regulation is implemented in financial markets or on conditions influencing trade, it is assumed that the direction of development – meaning more or less neoliberal – is taking place in other markets as well.

Financial market development is measured by the Financial Development Index of the IMF (Svirydzenka, 2016). This dataset consist of data from 1980 until 2015 for all countries required for this thesis. The financial development variable consists of the depth, access and efficiency of financial institutions and the depth, access and efficiency for financial markets. These six aspects are combined in one variable which represents the financial development and the financial liberty in a country. To prevent outliers shaping the variable, all countries with financial index of zero are dropped. These are all Easthern European countries before becoming democratic, dropping these results is no problem since eventually they cannot be part of the regression because no

(25)

elections were held. However, to prevent the zero-values having an effect of the constructing of the proxy, they are dropped.

Trade liberalisation is measured by the amount of import and export as percentage of growth domestic product, or trade openness. Data from the Penn World Table 9.0 is used (Feenstra, Inklaar, & Timmer, 2015). This assumes, that the amount of trade is dependent on liberty of international markets. This assumption is regarded valid since tariffs and other protectionist policies hamper the amount of trade. This variable is regarded as typical for neoliberalism because supra-national organisations considered as neoliberal such as the WTO, Worldbank and IMF focus on liberalisation of trade.

The third variable in the proxy for neoliberalism is the size of government expenditure as percentage of GDP. Data is from the Penn World Tables as well (Feenstra et al., 2015). A decrease in the size of the state could occur by waves of privatization or by austerity politics, which are generally promoted by neoliberalism. These two policies, widely promoted during the 2008 financial crisis and superseding Euro-crisis by the IMF and European Union, are consid-ered complementary to the more general variable of government expenditure. The variable for government expenditure is inverse, so an increase in the neoliberal government variable actually is a decrease in government size.

One aspect of the chosen measurements is that policy is reflected and not measured itself. The financial development index is a consequence of policy, trade depends on tariffs and other trade policies and government size can depend on national choices but also be influenced by the Maastricht Treaty. This implies that the proxy is independent of the institution enforcing the policy. Whether it is local, national or an international institution such as the EU, the proxy still reflects the policy. This is important since it entails Foucault’s broad conception of the state. Neoliberalism, as posed by Foucault, as a practise applied not only by governments but also by various institution such as non-governmental organisations and supra-national institutions, is therefore more accurately reflected in the proxy variable.

(26)

Figure 2.5: Distribution of variables

Financial market development, trade openness and size of government are therefore considered as fundamental for neoliberal policy and as accurately reflecting the overall policy direction a specific country is taking. These three variables are standardised with mean of zero and standard deviation of 1. The distributions of the standardised values of financial development, trade openness and government size are shown in figure 2.5. The standardised values of financial development, trade openness and government size are accumulated with equal weight to form the proxy for neoliberalism. This is in line with the OECD (2008) handbook for construction indicators composing of multiple variables with differing scales. The distribution of the new proxy variable is shown in figure 2.6. The proxy is not normally distributed (see appendix B.1) and can take values between -7.5 and +7.5, but all actual values are between -5 and +4.

(27)

The key argument for propagating neoliberal policy is economic growth. Therefore the first way to analyse the validity of the indicator is by relating it to GDP per capita. Table 2.1 shows the correlations between the proxy variable, the components of which it is composed and the real GDP per capita from the Penn World Tables 9.0. Firstly, note that neoliberal policy seems on the basis of the used variables to be strongly related to real GDP per capita. With a correlation of almost 0.74, the proxy variable seems to be in line with the argument for neoliberal policy that it stimulates GDP. Secondly, note that the correlation of the proxy variable with real GDP per capita is higher than the components of the proxy with real GDP. This strengthens the previous statement that neoliberalism, and not one policy aspect in particular, stimulates real GDP. Which indicates that the proxy has added value compared to the components.

Table 2.1: Correlation matrix of proxy components

Neoliberalism Trade Financial Government Real GDP p/c

Neoliberalism 1.0000

Trade openness 0.5670 1.0000

Financial development 0.9301 0.3234 1.0000

Government size 0.6519 0.0423 0.5294 1.0000

Real GDP per capita 0.7356 0.3768 0.6875 0.5196 1.0000

With the key assumption that market liberty in financial development and trade represent market freedom in a multitude of markets on international, national and individual scale, the proxy of neoliberalism is conceived as a relatively accurate measure in relation to the real GDP argument.

Neoliberal policy in practise

In this subsection the proxy for neoliberalism is discussed, mainly for six countries. Analysis consist of two parts, firstly some generally conceived neoliberal policies should be reflected in the neoliberal proxy. Secondly, some trends seen in the proxy are analysed.

(28)

Figure 2.7: Neoliberal policy in six countries

Figure 2.7 shows the proxy for neoliberalism in six countries from 1980 to 2015. The proxy for neoliberalism grows steadily up to the 2008 crisis, which is reflected in all countries, which led to increasing regulation for especially financial markets and requiring states to take over private banks. This increasing regulation and state take-overs are reflected in the proxy. Secondly, The Netherlands is, according to the proxy, the most neoliberal country of the six. This is probably due to high import and export levels over a relatively small economy. In the period of the 1980’s to the mid 1990’s West Germany has seen an increase of neoliberal policy, the peak downward could be the aftermath of reuniting West and East Germany. After this downward peak a steady increase is seen up to the financial crisis. Thatcher’s election in 1979 and her policy conceived as neoliberal is visible in the period up to 1990 in a high increase from -1 to almost +1. After this period the value stabilises until the mid 1990’s reaching ultimately 2 before the financial crisis. Spain is increasing slowly under moderate socialist administrations up to 1996, while in this period Spain’s entering the European Economic community in 1986 is only a little reflected in the proxy. Entering the EEC does not necessarily mean liberalisation and a decreasing government, but the increase from 1986 to 1996 is small. Only until market liberalisation starting in 1996 under the Aznar administration a strong increase is visible. France shows a steady increase of neoliberal policy. The entering of the Euro is against expectations not reflected in the proxy. Lastly, the value for Greece is the lowest of the six countries with two peaks upward. The first, probably due to obliged liberalisations and government contractions

(29)

for allowing entering the zone, and the second peak being the results of being in the Euro-zone. Greece’s low value is in line with criticism from for example the IMF and EU on a variety of social policies for pensions, healthcare and social security combined with many state owned enterprises. The low value therefore is in line with expectations.

Figure 2.8: Average neoliberal policy in all countries

Figure 2.8 shows the average value of the proxy for neoliberalism of all countries in the analysed period. The figure shows a strong increase in all countries up to the 2008 crisis. Three downward peaks are visible, the first could be related to the saving and loan crisis mainly in the US which led to a variety of financial regulations. The second peak after the Dotcom bubble and the last downward peak by increasing regulation, and thus a limiting of market freedom following the 2008 financial crisis. The overall increase is in line with the increasing dominance of neoliberalism on political practise perceived by Brown and Davies. Although, some results in the neoliberalism proxy are unexpected, the proxy is conceived as relatively valid for further analysis in relation to populism.

2.4 Control variables

This section covers the control variables used in the regression analysis to be able to distinguish the effect of neoliberalism on populism from others. Appendix B.2 shows all correlations between

(30)

variables, but no problems of correlation arise. The first control variable is a dummy to control for countries that have an autonomous region and therefore have a federal government. Autonomous regions are defined as having control over taxation, law-making and relative policy-freedom. Autonomous regions increase the likelihood for new parties to get a firm footing on a local level before partaking in country-wide elections (Kitschelt & McGann, 1997). Controlling for countries with an autonomous region is in line with S.M. Han’s and K.J. Han’s analysis (S. M. Han, 2015; K. J. Han, 2016). To construct the dummy for federalism, data is used from the Database of Political Institutions version 5 (Cruz, Keefer, & Scartascini, 2016). The dataset consist of a dummy equalling 1 if a country has an autonomous region and 0 otherwise. Countries with value 1 for federalism are for example Germany, Belgium, Spain and Italy. Summary statistics are shown in table 2.2, note that almost one-third of the countries is federalist.

Table 2.2: Summary statistics of control variables

Observations Mean S.D. Min Max

Immigration policy 899 0.446 0.125 0.290 0.929

Federalism 1307 0.310 0.463 0 1

Authoritarianism 1353 3.179 0.391 2.116 3.833

National institutions 1353 5.065 0.953 2.653 6.303

International institutions 1189 4.790 0.378 4.225 5.905

As described in the theoretical framework Inglehart and Norris (2016) find significant results in their explanation for populism for immigration related issues, valuations of authoritarian leadership, nationalist sentiments and a decrease in trust in global institutions. These issues form the cultural backlash hypothesis which should be controlled for to isolate the effect of neoliberalism. The same approach Inglehart and Norris used is taken. The data used is from the World Value Survey and European Value Survey ranging from 1981 to 2014 (Inglehart et al., 2014; EVS, 2015). The first problem is that survey data is on individual level, but since this analysis is macro based, all value data is combined to country aggregates. In this process a lot of the value from the data is lost. The second problem is that the interviews are not dated on itself, but dated as part of a specific survey wave covering a period. So by taking the averages, a lot of years do not have any data. Therefore all the survey data is taken as average per country as well. This decreases the power of these variables a lot.

With the value survey data, three variables are constructed. Firstly, a valuation of authority in leadership. Secondly, a valuation of national government and political parties, and thirdly a

(31)

measure of valuation of the EU and UN. The relevant questionnaires are provided in appendix A.4. Note that answer 0 is no answer and are therefore dropped, 1 is a high valuation and 5 is a low valuation. These three control variables are the sum of two questions, therefore the value is between 2 and 10. Note that in table 2.2 minimum and maximum values are 2 and 10, by taking the averages the values are only roughly between 2 and 6. This indicates the decrease of power due to the averaging. Secondly, all valuations, but especially authoritarianism, is below six, so in all countries an above average valuation is seen. Thirdly, the standard deviation of valuation of national institutions is higher than other control variables which means countries have a stronger difference of opinion.

Another control variable that is part of the cultural backlash thesis is immigration policy. Although the value surveys contain questions on immigration as well, a separate database is used. The main reason is that data in the value surveys is missing. Therefore the IMPIC database for

immigration policy is used (Helbling, Bjerre, R¨omer, & Zobel, 2017). IMPIC database is a policy

measurement for the openness of a country to immigrants. The database consist of over 50 sub variables of specific policies for application, citizenship, family reunification, schooling et cetera, which are combined to an overall indicator. This indicator takes a value of 0 for completely open to immigrants and 1 for restricted. Countries such as United Kingdom, Ireland and New-Zealand are relatively high indexed between 0.4 and 0.5 which means it is harder to become a citizen of these countries than countries such as Germany and Sweden with an index around 0.3. The downside of this variable is the availability of data until 2010, this restricts the total regression by five years. A second downside, in particularly in relation to populism is that policy might not reflect attitudes of voters very well. To take these downsides in mind, the first model will not include this control variable.

The last control variable is the result of the previous election. The main reason to include this variable is that elections are depending on too many factors to accurately explain why some party won or not. This is exemplified by the incorrect predictions of Hilary Clinton winning the US 2016 elections by several polls (Enns, Lagodny, & Schuldt, 2017). Therefore to increase explanatory power of the regression model, the previous election result is added. This creates a problem, neoliberalism as shown in figure 2.7 and 2.8 is a long-term process. If the previous election results are included, part of the explanatory power of neoliberalism could be included in the previous election result. Thus, by including the previous result, variables only show short-term effects. It is therefore expected that the results will be lower. To cope with these problems, only the third model includes the previous election result.

(32)

Chapter 3

Results

In this chapter, the results of the regression analysis are presented and discussed. This section aims to answer what effect neoliberalism has on election results for moderates, extreme left and right. Firstly, the results of the regression are being discussed based on border values of 2 and 8. Then these results are related to results using different border values.

3.1 Results for moderate parties

Table 3.1 shows four regression models. The first model includes the neoliberalism proxy, the dummy variable for federalism and valuations for authoritarian leadership, national institutions and international institutions. This basic model will be extended by adding the variable for immigration policy firstly, and secondly with the addition of the previous election result. Since panel data is used, every model is tested whether a regression with random effects or fixed effect can be used. The p-value of the Hausman test is shown, if this value is below 0.05, a fixed effect regression is done. This is the case for all regression with the previous result included.

A key problem in interpretation of explanatory variables and election results is causality. For example, does a low and thus open policy for immigration lead to right-wing populism, or does populism lead to a more strict policy towards immigrants. Especially for the control variables this two-way causality problem has to be taken in mind for interpretation. For analysing neoliberalism this two-way causality is less of a problem because neoliberalism is a slowly developing process and is therefore less influenced by a single election result.

The model in the first column explains only 4.1% of total variation of the election results for moderates. The coefficient of the neoliberalism proxy is negative but insignificant. The control variables are all insignificant with low z-values. One reason for the insignificance of the control variables could be that these are all time-invariant. The only control variable with a little higher z-value is the coefficient of authoritarianism. The positive coefficient signifies that a lower valuation of strong leadership has a positive effect on the vote share of moderates.

(33)

Table 3.1: Regression results for moderates

Moderate vote share (1) (2) (3) (4) Regression RE RE RE FE Period 1980 - 2015 1980 - 2010 1980 - 2010 1980 - 2010 Neoliberalism -0.797 -1.306** -0.701* -1.244** (-1.499) (-2.380) (-1.827) (-2.222) Immigration policy -2.975 1.359 -4.340 (-0.383) (0.255) (-0.517) Previous election result 0.711*** 0.407***

(13.520) (5.432) Federalism -0.840 -1.330 0.402 0.758 (-0.308) (-0.495) (0.366) (0.210) Authoritarianism 5.082 6.050 2.215 -(1.041) (0.820) (1.014) National institutions 0.203 0.595 0.385 -(0.116) (0.324) (0.706) International institutions -0.290 3.733 -0.377 -(-0.058) (0.717) (-0.241) Constant 69.524*** 46.795* 16.716** 53.005*** (2.929) (1.798) (2.031) (6.819) Observations 224 182 176 176 Number of countries 27 24 24 24 Hausman test 0.661 0.556 0.000 -R-squared 0.041 0.095 0.574 0.205 z-statistics in parentheses *** p<0.01, ** p<0.05, * p<0.1

Adding immigration to the second model decreases the total analysed period with five years and three countries which are not indexed in the IMPIC database are lost. But, by including immigration, the explanatory power rises to 9.5%. The proxy for neoliberalism is significant in this model at 5% confidence level and a value of -1.3. Neoliberalism has a negative relationship with moderate vote share in this model, and therefore automatically a positive effect on left and right together. The value of -1.3 indicates that an increase of 1 in the proxy of neoliberalism leads to a 1.3% lower vote share for moderates. For example France has implemented policy from 2002 to 2008 which increased the proxy of neoliberalism by 1, the model predicts a decline of

(34)

vote share of moderates in this period by 1.3%. Also, notable is the negative sign of immigration policy which indicates that strict immigration policy could lead to lower moderate vote share, or a lower moderate share to stricter immigration policy.

The third model is random effect but due to the significant p-value of the Hausman test, it should be fixed effect. The main reason model three is included is for the effect of the neoliberal-ism proxy given the inclusion of the previous election result. By including the previous election in the regression, various auto-correlation problems arise. In particular, if neoliberalism has an effect on the vote share, then the long-term effect will be undone by the previous election. The lower but still negative coefficient of neoliberalism indicates that in the short-term an effect is visible on the votes of moderates as well.

The last model includes all variables and is a fixed effect regression. This means the effect of every variable is tested within each country over time. Therefore the control variables of authoritarianism, national and international institutions are dropped because they do not vary over time and therefore have no explanatory power. In the last model, the R-squared increases to 20.5%. This is expected since the last election result is likely to be related to next election. The strong significance of the previous election results confirms this. The proxy for neoliberalism is significant at 5% with a negative effect on moderate vote share. This is consistent with the second model and coefficients are close to each other. Federalism being negative in the first and second model is now positive. Although insignificant the sign of the effect is against Kitschelt’s and McGann’s (1997) findings. Lastly, immigration policy is not significant in all models for moderate vote share.

In relation to the central question and hypothesis of this thesis, it can be concluded with caution that neoliberalism has a negative effect on the vote share of moderate parties. The first model shows a negative, insignificant effect, the second and fourth show a significantly negative effect. To what extend this effect translates in an increase of the extreme left or extreme right is discussed in the following section.

3.2 Results for the extreme left and right

Table 3.2 shows the same models but for the vote share of extreme left and right. The random effect model including all variables is excluded because of significance of the Hausman test. Model five has very low explanatory power of vote share of left-wing parties. This model therefore proofs to be not very useful. The control variables are based on Laclau’s definition of populism and

(35)

the cultural backlash thesis but the constructed variables seem to have little explanatory power. Also when comparing it to models one and eight, which are the same, the explanatory power of model five is low. Lastly, neoliberalism shows to be negative and insignificant.

Model six shows little difference to model five. Neoliberalism is still negative and insignificant, and the addition of immigration policy does not seem to improve the explanatory power of the model much. Immigration is insignificant as well. Model seven entails the same results, only the previous result is significant which is expected but does not say much. It can be concluded that the model is not appropriate for analysing the vote share of the left. It can also be concluded that the increase of neoliberalism does not seem to have effect on the vote share of the extreme left.

Table 3.2: Regression results of extreme left and right

Vote share (5) (6) (7) (8) (9) (10)

Left Left Left Right Right Right

Regression RE RE FE RE RE FE Period 1980 - 2015 1980 - 2010 1980 - 2010 1980 - 2015 1980 - 2010 1980 - 2010 Neoliberalism -0.442 -0.107 0.198 1.230*** 1.400*** 1.370*** (-1.374) (-0.289) (0.609) (3.500) (3.926) (3.441) Immigration policy 1.002 -1.025 -0.175 2.603 (0.192) (-0.210) (-0.035) (0.454) Previous election result 0.578*** 0.229***

(8.892) (2.824) Federalism 1.212 1.277 0.104 -0.634 -0.238 0.675 (0.677) (0.661) (0.049) (-0.345) (-0.133) (0.272) Authoritarianism 2.588 2.570 - -6.035* -8.310* -(0.722) (0.441) (-1.794) (-1.651) National institutions 0.313 0.200 - -0.914 -0.978 -(0.243) (0.137) (-0.759) (-0.780) International institutions 0.157 -1.429 - 0.931 -1.260 -(0.042) (-0.347) (0.268) (-0.355) Constant -5.505 2.229 2.561 26.225 44.455** 2.865 (-0.315) (0.109) (1.107) (1.602) (2.505) (1.049) Observations 224 182 176 224 182 176 Number of countries 27 24 24 27 24 24 Hausman test 0.071 0.090 - 0.173 0.178 -R-squared 0.016 0.034 0.352 0.077 0.143 0.166 z-statistics in parentheses *** p<0.01, ** p<0.05, * p<0.1

(36)

Model eight to ten entail the results for the extreme right. The first difference if comparing the model for left and right is the increased explanatory power expressed in the 7.7% R-squared of model eight. Neoliberalism has a positive value of 1.2 and is significant at 1% . The vote share of the extreme right increases as neoliberalism increases. Federalism, although not significant, has an unexpected negative sign. Valuation of authoritarian leadership is significant at the 10% level for the right, meaning a high valuation of leadership results in more support for the extreme right. Valuation of national institutions, although insignificant, has the expected sign. Meaning a high valuation of national institutions goes together well with higher support for the extreme right. For international institutions, which is also insignificant, the reversed is the case, namely that a low valuation of international institutions goes together with higher support for populism. This result is according to expectations.

Model nine adds immigration policy and against expectations, it is highly insignificant with a z-value of -0.035. Including immigration the explanatory power of the model increases to 14.3% which is still lower than micro-based analysis of survey data for example by S.M. Han who reach a R-squared around 35% (S. M. Han, 2015). Macro analysis such as this thesis is expected to have lower explanatory power, which is confirmed by this result. The main result of model nine is that neoliberalism increases in value and significance if compared to model eight. The coefficient has a value of 1.4. Neoliberalism increasing on average from -0.5 to +1 in 30 years time in the sample, then the vote share of the extreme right increases due to neoliberalism based on this model by around 2% in absolute terms. On average, the increase in support of the extreme right of around 4.5% since 1980 found in this sample could for a large part be caused by the increase of neoliberal policy. Lastly, there is an unexpected change in the sign of the valuation of international institutions.

Model ten includes the previous election result of the extreme right and is a fixed effect regres-sion due to the Hausman test value below 0.05. Firstly, note that the coefficient of neoliberalism does hardly change compared to model eight and nine which confirms that increased neoliberal policy is connected to the growing support of populism. Secondly note that the coefficient of immigration policy suddenly increases to 2.6. Although it is insignificant, the positive sign could indicate some connection between strict immigration policy and populism.

Referring back to the main hypotheses of this thesis, based on the 2 and 8 border values, it can be concluded that the effect of neoliberalism on the vote share of moderate is negative. This decrease is not caused by the increase of the political left. For an increase of the vote share of the extreme left due to neoliberalism there is found insufficient proof since neoliberalism

(37)

proved insignificant in all regressions. One remark has to be made, due to the low R-squared of regressions on the left, the model is inaccurate and could be improved. The decrease of moderates can, however, be significantly contributed to increase of the extreme right due to neoliberalism. The inaccuracy of the model on the left is not found for the right, with a higher R-squared this model seems better fitted for the right.

3.3 Results with different classifications

The previous section discussed regression results based on a classification with border values 2 and 8. Since the choice of these values is arbitrary, the results of the same models are discussed for different border values. Tables below only show the coefficient and significance of neoliberalism for the model without immigration and previous result, and for the model including immigration and without previous result. To analyse robustness of the significance and coefficient of neoliber-alism only the random effect regressions are discussed. The full models and results are presented in appendix B.3.

The model for moderates in table 3.3 shows that neoliberalism has a positive effect if defining moderates between 1 and 9. The positive and significant result in column one is contrary to the other results of other border values. One possible reason is that only some right-wing populist parties are included in the >9 classification. Front national (9.7) is still right-wing but the AfD (8.6) is not any more. The classification of 1 and 9 is therefore hard to theoretically justify which hampers the result.

Table 3.3: Results for moderates with different border values

(1) (2) (3) (4) (5) (6) (7) (8) Border values 1 - 9 1 - 9 1.5 - 8.5 1.5 - 8.5 2 - 8 2 - 8 2.5 - 7.5 2.5 - 7.5 Neoliberalism 0.514** 0.405 -0.303 -0.474 -0.797 -1.306** -1.125* -1.344** (2.274) (1.592) (-0.676) (-1.034) (-1.499) (-2.380) (-1.930) (-2.247) z-statistics in parentheses *** p<0.01, ** p<0.05, * p<0.1

Classifications from 1.5 - 8.5 to 2.5 - 7.5 all show the same result, neoliberalism has a negative relation with moderate vote share. But effect of the coefficient and significance increases if narrowing the moderate classification. Moving the border for left is ideologically not problematic since moving from 0 upward, the influence of communism decreases and shifts to socialism and

Populism and neoliberalism : a cross-country analysis of the effect of neoliberal policy on increased support for left and right-wing populism in Europe