The Rise of Populism Assessing the Economic Hardship Theory after the Global Financial Crisis

(1)

The Rise of Populism

Assessing the Economic Hardship Theory after the Global

Financial Crisis

Master Thesis

MSc International Economics & Business June 2017

Supervisor: Prof. Dr. Jakob de Haan, Rijksuniversiteit Groningen Co-assessor: Dr. Dóra Piroska, Corvinus University of Budapest

Nick Boersma nboersma@live.nl or n.boersma.4@student.rug.nl

(2)

1

Abstract

*

Economic hardship is frequently mentioned as one of the main drivers of the recent rise of populism. This paper will test several hypotheses on socio-economic factors that might influence populism both left- and right-wing populism. Also, the impact of the global financial crisis of 2008 on these factors is taken into account. To do this, a dataset of 29 European countries and 184 elections is compiled for the period 1986-2014. Estimating a panel Tobit model and analysing interactions between the variables, unobserved regional differences in populist performance and clear differences between what influences left- and right-wing populism are shown. Other findings are: First, economic growth negatively impacts both left- and right-wing populist support in a nonlinear way. Second, globalization positively affects right-wing populism, while the different dimensions of globalization show mixed results. Third, increasing unemployment raises the vote share for the populist left. Fourth, evidence is provided that the global financial crisis of 2008 influenced the impact of socio-economic factors on the populist vote share; the impact of unemployment on left-wing populist support became stronger, while the role of globalization became more prominent for right-wing populist support. Finally, no evidence is found for the impact of regional inequality.

Keywords: Populism, Globalization, Economic Hardship

*_{I would like to acknowledge my gratitude to my research supervisor, prof. dr. Jakob de Haan, for providing}

(3)

2

I. Introduction

‘Populism is threatening the safety of Europe’, says Dutch Commander in Chief Tom Middendorp (Nieuwsuur, 2017). With this serious statement, he refers to the anti-EU sentiment many populist parties are propagating, leading to less cohesion between the EU member states and with that, less co-operation.

Populism is an ideology which pits a virtuous and homogenous people against a set of elites and dangerous ‘others’ who are together depicted as depriving the sovereign people of their rights, values, prosperity, identity and voice (Albertazzi and McDonnell, 2013). Throughout Europe, populist parties have gained large influence in modern-day politics, winning seats in legislatures and changing policies of mainstream parties. The recent election of Trump as president of the United States adds North-America to a seemingly growing list of, especially right-wing, populist-led countries. “The forces of openness, tolerance, diversity, multiculturalism, and globalisation have been left reeling at…” making way for “…divisive rhetoric, disregard for facts, promises of simple cures for all ills, nativism, demagoguery, and the power of seductive slogans, which are common features of the new populism” (EEAG, 2017). And although during the 2017 elections Front National’s Marine Le Pen did not become the first female president of France and Geert Wilders’ Party of Freedom did not become the largest party in The Netherlands, both movements have seen a vast increased electoral support in recent years. In Poland and Hungary populist parties have been in power for around a decade, openly disagreeing with EU policy and warding off asylum seekers. The success of UKIP’s initiated Brexit is specifically a milestone in Western European populism, showing the resentment there is against Europe. But what explains this rise of populist parties?

To many, the recent rise of populism is seen as a side effect of the global financial crisis of 2008 (Doležalova, Havlík, Slaný and Vejvodová, 2017; Funke, Schularick and Trebesch 2016; Anduiza and Rico, 2016). After 2008, economic growth in many European countries turned negative and unemployment rose, especially in the Mediterranean countries. This observation would be in line with the economic hardship theory, according to which vulnerable and economically deprived people are more likely to support populist parties. Another factor on which especially right-wing populist parties seem to focus is immigration. While proclaiming nationalism, these parties are mostly against immigrants. The recent wave of refugees might therefore play an important role as well in their electoral success. According to Golder (2003) increasing immigration plays an even larger role when unemployment is higher. But while he finds support for this, others do not (Doležalova et al. 2017).

(5)

4 Apart from the fact that consensus is lacking on the causes of populism, a broader view on populism is also missing in modern research. The focus is mainly on the right-wing parties, but Europe has always had far-left populist parties as well. And somewhat surprisingly, there are similarities between these parties on both ends of the political spectrum (Johansson Heinö, 2016).

This paper is aimed to fill this gap and determine which socio-economic factors influence electoral support for populist parties and in what sense the global financial crisis affected the vote share. It goes beyond previous research since it will include both left- and right-wing populist parties to compare to what extent their vote shares are influenced differently. Operationalizing the economic hardship theory, I will look specifically at economic growth, the level of unemployment, and the mediating effects stemming from social protection benefits. Following Golder (2003), the number of asylum applicants and its interacting effect with unemployment are considered as well. Likewise, the model includes a globalization index that incorporates economic, political and social dimensions. While a major feature of contemporary structural change (Swank et al. 2003), globalization has long been neglected in research. Finally, the effect of regional inequality in income between urban and rural regions is incorporated in the model (Stockemer, 2017; EEAG, 2017).

The dataset used for this research covers 184 elections in 29 European countries for the time span from 1986 to 2014. It therefore allows me to test the salience of traditional explanatory theories and check for the differences between left- and right-wing parties and the effect of the global financial crisis on the rise of populism. The empirical analysis in this paper uses the Tobit model with a maximum likelihood estimator for left-censored dependent observations. This paper contributes to research on populism in the sense that it underlines a clear difference between factors influencing left- and right-wing populism and the heterogeneity in populism between different regions in Europe. The main findings of this study are the following: First, the economic hardship theory is confirmed by in the sense that growth negatively influences the support for populist parties. Second, globalization increases the right-wing populist vote. When disentangling globalization into different dimensions, mixed results are found; both economic and social globalization impact the populist vote share, while surprisingly no evidence is found of a relationship between political globalization and populist support. Third, increasing unemployment has a positive effect specifically on the left-wing vote share. Fourth, this paper provides evidence that the global financial crisis of 2008 has influenced the impact of socio-economic factors on the populist vote share; the impact of unemployment on left-wing populist support became stronger and for right-wing populist support the role of globalization became more prominent. Finally, no evidence is found for the theory that regional inequality impacts the right-wing populist vote share.

(6)

5

II. Literature review

In this chapter, I will discuss previous research on the rise of populism and define what populism is. A clear distinction will be made between left- and right-wing populism. It is important to understand what populism is to examine potential factors that have caused for the rise in its support.

What is populism?

As many argue, a wave of populism is sweeping across Europe and North America (EEAG, 2017). It is said to be one of the main challenges to contemporary democracy and has captured increasing academic and media attention in the last two decades (Kaltwasser and Taggart, 2016). But populist parties have been growing in Europe even before the 21st century. The electoral success of populist parties started already in the 1980s in (mainly Western) Europe (Arzheimer, 2009; Mudde, 2004; Golder, 2003; Ignazi, 1992). We can clearly see this trend in Figure 1 below, which also shows the rising trend of the populist right and the downfall of the populist left until 2011, after which support has been rising again.

Figure 1 – European average of populist vote share (in % of total) in 1980 - 2016

What populism is exactly seems very much debatable, in not only the academic world, but also in the media or politics. The reason is that it can be difficult sometimes to distinguish populist and non-populist parties, ideas and political regimes. Given the chameleonic nature of populism, placing the focus on support for parties adopting diverse ideological stances makes it even more difficult to “separate populism from features that might regularly occur together with it, but are not part of it” (Anduiza and Rico, 2016). Mainstream political parties pursue some populist policies as well. Therefore, populism can be seen as a characteristic of the political process itself, not only of certain political parties or politicians (EEAG, 2017). The word populism goes back to the Latin word populus, which means ‘people’ in the collective sense of the word. In the social sciences the term is arguably among the more elusive. It is Notes: The graph represents the average vote shares for populist parties in the 29 countries used in the dataset for

(7)

6 commonly used to describe a certain political style and a political ideology. In the former sense all parties contain populism to varying degrees – political messages, even opinions, are adapted to what one assumes voters want to hear; the complex is simplified, conflicts of aims hidden. In this matter, populism is a matter of degree, parties and politicians acting more or less populist (Johansson Heinö, 2016).

Mudde (2004) defines populism as an ideology that considers society to be ultimately separated into two homogeneous and antagonistic groups, 'the pure people' versus 'the corrupt elite', and which argues that politics should be an expression of the volonté générale (general will) of the people. A similar definition is given by Albertazzi and McDonnell (2008), who state that populism is ‘an ideology which pits a virtuous and homogenous people against a set of elites and dangerous ‘others who are together depicted as depriving the sovereign people of their rights, values, prosperity, identity and voice’. In this sense, there are certain characteristics that clearly separate populist parties from the non-populist parties. Mudde (2007) states that populist philosophy consists of three core features: The first is that populists are anti-establishment, which means that they emphasize the divide between the ordinary people and the ‘corrupt’ establishment. Secondly, populism features authoritarianism. Populists favour a charismatic leader that reflects the will of the people. Moreover, they prefer direct forms of democracy through polls and referenda, rather than institutional checks and balances. The third feature is nativism. Nativism is an ideology which holds that states should be inhabited exclusively by members of the native group (‘the nation’) and that non-native elements (persons and ideas) are fundamentally threatening the homogenous nation-state.

This is very much in line with the definition of populist parties used by Van Kessel (2013). He states that these parties (i) delineate an exclusive community of ‘ordinary people’, (ii) appeal to those ordinary people, whose interests and opinions should be central in making political decisions and (iii) are fundamentally hostile towards the (political) establishment, which allegedly does not work for the interest of the ordinary people. Since populist parties are often not very specific about their target audience, it is not self-evident who belongs to these ‘ordinary people’. Instead, populist parties are usually clearer about who does not belong to their portrayed community, which means that the community is typically constructed in a negative manner (Albertazzi and McDonnell, 2008). In many cases, specifically for the right-wing populist parties, immigrants and ethnic or cultural minority groups are the usual suspects to be branded as outsiders. But not all populists are xenophobic. The group of ‘others’ could also be the corporate elites, the media (especially in the case of Trump) or intelligentsia whose ideas, values and interests are at odds with those of the ‘silent majority’ (Canovan, 1999).

(8)

7 generally at odd with those analyses of established parties, such as combatting Islam, decreasing diversity, strengthening our borders, getting the European Union off our backs, returning to respect, to the values and norms that used to sustain the fulfilling life and the good society. In this sense populism seems to be politics of hope (Elchardus et al. 2016). Heinisch (2003) adds to these characteristics that populism requires a strong leader to be successful and is flexible in an ideological and programmatic sense in pursuit of voter maximization. Populist parties fill the void that has emerged as a result of the growing discrepancy between popular and constitutional democracy by successfully managing the game of elections. They tend to do this by addressing spectacular issues, using calculated outrage and novel forms of self-presentation. In this way, they draw in also younger, less politically aware and apathetic voters, together with those that have defected from mainstream parties.

Heinisch (2003) also finds that in general, conservative parties tend to be the main beneficiaries from the political fallout of this process, performing a bridging function. Where populist parties seem to lack the ability resolving intra-party disputes and experienced policy makers, conservative politicians could successfully appeal to voters by presenting themselves as effecting political change and reform while offering competent performance and professional political management as well as a more moderate version of the programmatic demands that had fuelled the growth of populist parties in opposition.

Left- and right-wing populism

Populism can be seen as a thin ideology, meaning that it rarely exists on its own; it mostly attaches itself to other ideologies ranging from (neo-) liberalism, the radical right, to socialism (Kaltwasser and Taggart, 2016; Elchardus and Spruyt, 2016; Mudde and Kaltwasser, 2013; Zaslove, 2008; Weyland, 1995). But only few studies examine overall support for populism, not narrowing it down to one side of the political spectrum. Exceptions are Funke et al. (2016) and Johansson Heinö (2016).

(9)

8 language (Johansson Heinö, 2016) and there is a significant overlap between their voter bases (Oesch, 2008).

In explaining the rise of extreme right-wing parties, Golder (2003) states that there are three arguments: (1) the materialist argument, which is the connection between unemployment and immigration, (2) the ideational argument, which comprises the threat to national identity/culture by immigration and (3) the instrumental argument, that states that electoral institutions help extreme right-wing parties to achieve their goals. This last argument is made clear by the example of larger district magnitudes for which it is easier for populist parties to gain a seat during elections and with such a position can exert political influence.

Purely on the right-wing spectrum an important distinction should also be made. For a long time, extreme-right parties existed, but Ignazi (1992) distinguished fascist from new right-wing parties in this perspective. He did this based on spatial, historical-ideological and attitudinal systematic differences. The new-right wing parties Ignazi described we would now call right-wing populist parties (Heinisch, 2003; EEAG, 2017). And as Mudde (2007) states: ‘All populist radical right parties are nationalist, but not all nationalist parties are radical right populist.’

Previous research

Academic research on populism has examined many potential indicators of populism, and at many different levels: individual, regional, national and cross-national. Several authors focus on the attitudes and behaviour of voters for populist parties, while others showed the importance of the characteristics of populist parties themselves, or contextual factors, like economic indicators, but also the media and electoral systems. In this section, I will highlight the main contributions of research on populism so far. This analysis functions as the foundation on which the hypotheses for this paper will be based. While this research will be mainly focused on macro-indicators, this part will give a more comprehensive overview of important findings on factors influencing populist support. Therefore, it also comprises research on electoral systems, media and immigration. Furthermore, since it is important to understand what is driving voters to support populist parties to be able to establish well-founded indicators for further research, the following overview will start with covering this part of the academic literature too, before moving on to research on macro-indicators that influence populist support.

Individual-level: characteristics of voters

Research on support for populist parties and research on populist attitudes address quite different questions, so one cannot serve as a surrogate for the other. By having populism as a defining characteristic of a given party, it is assumed that this component likely contributes to explain its electoral performance. Research on support for the populist radical right, for example, often uses populist attitudes (or some related construct, such as political trust) as an independent variable (Anduiza et al. 2016). Research on populist attitudes wants to find out what drives these people and who are the voters for populist parties? There are several examples of research on this and they make mainly use of surveys.

(10)

9 background characteristics and public opinion polls alongside country characteristics and characteristics of extreme right-wing parties. From a series of multilevel models they conclude that after controlling for individual anti-immigrant attitudes and political dissatisfaction, the number of non- Western residents as well as characteristics of the extreme right parties themselves, as organizational strength, membership activism and charisma of the leader, have a substantial impact on the likelihood of an extreme right vote. On the other hand, they find that the level of unemployment has no significant effect.

More recently, an increasing number of studies examines voter characteristics, such as attitudes. (Akkerman, Mudde and Zaslove, (2014) measured populist attitudes among voters based on previous research by Hawkins, Riding and Mudde (2012) using a survey of over 600 Dutch citizens and find that voters who score high on their populist scale have a significantly higher preference for the Dutch populist parties (Party for Freedom and the Socialist Party). Building on these findings, Spruyt, Keppens and Van Droogenbroeck (2016) extended this research to Flanders (the Dutch speaking part of Belgium) and confirm the findings of Akkerman et al. (2014). Besides, they find that populist attitudes can be distinguished from feelings of lack of external political efficacy. Notable people who are unsatisfied with politics as societal life vote for populist parties. Populism receives strong support from stigmatized groups who face difficulties in finding a positive social identity (which is possibly a result of globalization), but Spruyt et al. (2016) also conclude that populist support comes from especially subjectively experienced economic, cultural and political vulnerability. They state that a stigmatized group (e.g. low educated) will find in the empty signifier, ‘the people’, a means to adopt a group perspective to interpret their social position and maintain their self-respect.

Another survey-based research in Flanders comes from Elchardus and Spruyt (2016). Their approach focuses not so much on the supply side of politics, like populist ideology, populist rhetoric and the societal developments explaining the rise of presumably populist parties, but examines the extent to which (core elements of) populism are accepted among the electorate; populism is measured as an attitude. Their main finding is that the support for populism is based on relative deprivation, especially ‘declinism’, stemming from the feeling that establishment politicians did not offer convincing solutions to their problems.

Another paper based on Akkerman et al. (2014) comes from Anduiza et al. (2016). In their analysis they look at nine European countries to investigate what economic factors influence populist attitudes among voters. This stems from the recurrent impression that populism is enhanced in times of economic hardship. Based on the idea of Moffitt (2015) that crisis does not have to be a precondition of populist support, but populism may act as a trigger to crisis. In this sense, perceptions of crises become more important. From their survey-based research, they indeed conclude that it is not so much the objective economic situation that matters for the development of populist attitudes, but rather the perceptions that there is indeed a critical economic situation. Their research indicates that people with a high need for social acceptance are more likely to support populist parties.

(11)

once-10 predominant sectors reacting to progressive value change. Taking the European Social Survey for 2002-2014 as a basis for their research to examine cross-national evidence at the individual level, the authors find that populist support is greatest among the older generation, men, the less educated, ethnic majority populations (non-minority or what populists would call ‘natives’), and the religious. According to the authors, people in this group suffer most from changes in values in the past few decades and feel they are being marginalized within their own countries. This is contrary to the finding of Heinisch (2003), who states that populism attracts the young, the less political aware and apathetic voters through its novel form. Furthermore, Inglehart et al. (2016) state that not only economic indicators are important in explaining populist support, but also psychological factors. But how novel this finding may seem, they do not underline that according to their own research, economic indicators have a much stronger effect on populist support than psychological factors.

Stockemer (2017) focuses on the negative effects of modernization on certain groups in explaining support for far right-wing parties. Different from most research that focuses either on the country or individual level, he employs data at the NUTS2 level provided by the European Union. The focus is on 160 regions in 17 Western European countries from 1990-2013. Based on previous literature, the ethnic competition, economic hardship, and losers of modernization hypotheses are tested. In practice, this means that the effects of immigration, unemployment, and education and urbanization on far right-wing support are examined. Apart from these (more regular) indicators, the effect of turnout is tested and found not to have an influence on the radical right-wing vote. On the other hand, the ethnic composition hypothesis is confirmed, indicating that regions with more foreigners have a higher vote share for radical right parties. For the economic hardship hypothesis, it is found that only increases in unemployment contribute to votes for the radical right, not the unemployment level as such. The ‘losers of modernization’ hypothesis is confirmed in the sense that rural regions show a higher support for the radical right, but not when it comes to education. Higher aggregate education levels lead to more support, at least in the rural regions. This is in contrast to what the literature expects (Spruyt et al. 2016; Inglehart et al. 2016).

Macro-indicators

(12)

11 immigration. The idea that the level of unemployment positively affects the support for populist parties in combination with high immigration is supported by many other studies (see: Stockemer, 2017; Arzheimer, 2009; Boomgaarden and Vliegenthart, 2007; Bale, 2003). Others conclude that solely unemployment causes an increase in populist support (Jackman and Volpert, 1996), while some find no connection between unemployment and the vote share for populist parties (Knigge, 1998; Lubbers et al. 2002).

In a similar fashion as Golder (2003), Arzheimer (2009) focuses on the extreme right vote in Western Europe. The issue he wants to address is why support for these parties varies so much over time and across countries. For the period 1980-2002 he combines Eurobarometer survey data with country level contextual data on unemployment and immigration and looks at party manifestoes. The research shows that the extreme right parties benefit from high levels of immigration and unemployment, but that the effect is moderated by institutions of the welfare state. The lowest support levels for extreme right parties are predicted for a system with low benefits, low unemployment and minimal immigration. On the other hand, high support for these parties would stem from high unemployment in combination with high immigration or unemployment benefits (but not both). Surprisingly, given the level of variables included in his model, there are still country-specific differences. For example, support for the extreme right is persistently much stronger in some countries where it is expected to be lower, and vice versa there are examples of countries where support is weaker than expected.

Ponticelli and Voth (2011) look at indicators of social unrest (demonstrations, assassinations, riots, general strikes, and attempted revolutions) between 1919 and 2009. They relate these to changes in GDP and other variables (including measures of fiscal austerity, which is their particular focus). Their results clearly show a positive correlation between fiscal retrenchment and instability. Interestingly, they find evidence that growing media penetration does not lead to a stronger effect of cut-backs on the level of unrest, while some authors suggest this might be a potential influence on social unrest (Arzheimer, 2009; Boomgaarden et al. 2007). While their measures of social unrest are not the same as the electoral outcomes of concern to me here, they are likely to be correlated. It is suggestive therefore that in most specifications they find a negative correlation between growth and social unrest even when controlling for other factors. De Bromhead, Eichengreen and O’Rourke (2013) examine political extremism in the interwar period of 1919 to 1939. They establish a database of 171 elections in 28 countries of (mainly) European countries and measure extreme political parties based on their received votes. These parties had fascist or communist ideologies. The authors find that cumulative growth is negatively related to the fascist vote. Besides, other factors contributed to the support of extreme parties, like if the countries had relatively short histories of democracy and whether they had been on the losing side in World War I. De Bromhead et al. (2013) conclude that electoral system characteristics also matter. A higher minimum share of the vote needed for a party to gain parliamentary representation made it more difficult for extremists to gain seats in parliament. A higher threshold also appears to have lowered votes for extremists in the first place.

(13)

12 populism, using Chile during Allende’s Unidad Popular and Peru under Alan Garcia as their cases. They find that support for those regimes come from dissatisfaction with the country’s growth performance, and voters are lured into support by so-called macroeconomic populism. This is an approach to economics that emphasizes growth and income redistribution, while deemphasizing the risks of inflation and deficit finance, external constraints and the reaction of economic agents to aggressive non-market policies. The authors conclude that these populist policies ultimately fail.

Welfare state & globalization

Swank and Betz (2003) were the first to empirically analysed the mediating effects of welfare state institutions on the extreme right (ER) vote. Furthermore, they examine the contribution of globalization on the extreme right support. In their macro model, they regress the electoral returns of the ER in 83 elections that were held between 1981 and 1998 in 16 West European countries on trade openness, capital mobility, and foreign immigration as well as on the level of social protection and a number of other contextual variables. From their findings, they conclude that the number of asylum seekers is positively related to extreme right party’s success, whereas a high level of welfare state protection reduces the appeal of the extreme right. Moreover, they conclude that increasing transnational flows of trade, capital and people has contributed to the electoral success of new far-right parties, but for countries with a system of social protection that is comprehensive, generous and employment-oriented this effect from globalization is strongly diminished. Growth rates on the other hand, as well as long-term unemployment and inflation rates, do not seem to affect support for the far-right.

Effects of crises

It is widely believed that a crisis is a necessary precondition for populism (Laclau, 2005; Weyland, 1995; Roberts; 1995). In a similar fashion, Cordero and Simón (2016) show that the economic crisis affects support for democracy as a regime. Based on European Social Survey data for the Eurozone countries, they find that perceptions of the state of the economy have an impact on both the satisfaction and the support for democracy. This is also a sign of people blaming the political establishment of the crisis and therefore a potential cause for growing support for populist parties. In a paper by EEAG (2017) it is stated that crises will inevitably lead to a debate about the failure of the ruling elites and the fact that the costs of the crisis are not borne by those deemed responsible for it.

(14)

13 financial crises. Compared to other recessions, financial crises have more pronounced political effects.

Electoral systems and the media

Jackman et al. (1996) conducted the first largescale quantitative comparative analysis to examine the determinants of the success of extreme right-wing parties between 1970 and 1990. They estimate a Tobit model and find that votes for extreme right-wing parties are negatively related to the electoral threshold and positively related to the unemployment rate, and that the effective number of parties (the degree of multi-partyism) is positively associated with a higher extreme-right vote. Acknowledging that electoral thresholds and multi-partyism may be interdependent, they also analyse the interaction of the two variables. They find that while electoral thresholds have little impact on the vote for extreme right-wing parties when the effective number of parties is low, higher thresholds have a significant dampening effect on such votes when the effective number of parties is high. Similarly, while the effective number of parties does not have much effect on the extreme right-wing vote in the presence of high effective thresholds, it does have such an effect when electoral thresholds are low.

Boomgaarden et al. (2007) investigated the role of news media content in the rise of anti-immigrant parties in The Netherlands in 1990-2002. Their specific focus is on Pim Fortuyn, who had much political influence during this period before he was assassinated in 2002. While also controlling for unemployment, immigration, and leadership, the authors find a positive and significant impact of immigration-related topics in newspapers on the vote intention for anti-immigrant parties. However, it is very likely that external events have driven both media coverage and anti-immigrant party support, while this is not tested for in their research. Therefore, the role of the media remains unclear.

III. Hypotheses

Summing up the main findings taken from the literature overview on populism above, it becomes clear that: First, economic hardship is of importance in explaining the support for populism. Since crisis is a precondition for populism (Laclau, 2005; Weyland, 1995; Roberts; 1995) and impacts political stability (Funke et al. 2016; Cordero et al. 2016), especially financial crises (Funke et al. 2016), I will further examine this by comparing the support for populist parties before and after the global financial crisis, expecting that:

Hypothesis 1 – The financial crisis of 2008/09 has had a positive impact on the support for populist parties in Europe.

(15)

14 negative correlation (Arzheimer and Carter, 2006; Knigge, 1998). Following Stockemer (2017), it is not necessarily the levels of unemployment that matter, but the increase of unemployment. Therefore, I will test the relationship between unemployment and populism in the following manner:

Hypothesis 2 – An increase in unemployment will have a positive impact on the support for populist parties in Europe.

The relationship between unemployment and support for political extremes is compared to other economic indicators, relatively often examined. Besides there being no consensus on the effect of unemployment, the effect of economic growth on populist support is unclear, since it has been put aside after being examined in earlier studies (like in Swank et al. 2003), but no clear relationship was found, except for Latin-American populism (Dornbusch et al. 1990). De Bromhead et al. (2013) report that there seems to be an effect of cumulative growth performance in previous years on support for political extremes. Therefore, this relationship will be tested for its effect on populism;

Hypothesis 3 – Economic growth in previous years has a negative effect on the support for populist parties in Europe.

Another economic indicator that did not receive sufficient examination, but has become popular during recent years, is regional inequality. The hypothesis is based on the empirical findings from recent elections and referenda, concerning for example the 2017 Dutch parliamentary elections, France’s presidential election in 2017 and the Brexit referendum in 2016. During these events, rural regions showed a relatively large support for populist parties and/or against the establishment, compared to urban areas. Stockemer (2017) comes up with evidence for the hypothesis that rural regions show a higher vote share for right-wing populist parties, but his paper is based on a cross-regional research. In this paper, the focus is on national accounts. Therefore, I will test for the following hypothesis:

Hypothesis 4 – The more profound the difference in income inequality is between rural and urban regions, the larger the support for populist parties in Europe.

(16)

15 Hypothesis 5 – The higher the level of globalization in a European country is, the higher the

support for populist parties.

Arzheimer (2009) examines the effect of different types of welfare state on the support for right-wing populism, similar to Swank et al. (2003). He finds that a system with more social protection moderates the effect of unemployment and globalization. The reasoning behind this is that losers of globalization processes, especially the low skilled workers (EEAG, 2017), would not ‘lose’ as much, since there are higher social benefits to compensate for this. Therefore, these people would be less influenced by populist rhetoric. However, as Ponticelli et al. (2011) and Funke et al. (2016) point out, there is mostly pressure on government spending after a financial crisis, directly influencing the level of social expenditure. The impact of this on populist support is still unmeasured, while Lewis-Beck et al. (2000) conclude that electorates are strongly affected by global economic fluctuations (real and perceived) and that plausible economic indicators (objective or subjective) do account for much of the variance in government support. This brings us to the next hypothesis:

Hypothesis 6 – The higher social protection is in a country, the lower is the support for populist parties.

The resentment of the right-wing populist movements towards immigrants (especially those from Islamic countries) seems to have grown during the recent refugee crisis in Europe. This resentment is in line with one of the main features of populism: nativism (Mudde, 2007). Also, following the materialist argument that these immigrants take away jobs and money (Golder, 2003) and the propagated weakness of the government to solve this issue the following hypothesis is formulated:

Hypothesis 7 – A higher percentage of asylum applicants in a country’s population, the higher the support for populist parties is.

(17)

16 Table 1 - Populist Parties & Elections in sample

Country PartyID (Left/Right) Election

Austria BZÖ (R), FPÖ (R), KPÖ (L), SLP (L) 1986, 1990, 1994, 1995, 1999, 2002, 2006, 2008, 2013

Belgium Front National (R), Partij van de Arbeid (L), Vlaams Belang (R), PVDA (L), LCR (L)

1987, 1991, 1995, 1999, 2003, 2007, 2010, 2014

Bulgaria Ataka (R), BBC (R), IMRO (R), NFSB (R), RZS (R) 2001, 2005, 2009, 2013, 2014

Cyprus AKEL (L), ELAM (R) 2001, 2006, 2011

Czech Republic DSSS (R), KSCM (L), REP (R), USVIT (R) 1998, 2002, 2006, 2010, 2013

Denmark Dansk Folkeparti (R), FRP (R), Enhedslisten (L), DKP (L), Faelles Kurs (L), VS (L)

1987, 1988, 1990, 1994, 1998, 2001, 2005, 2007, 2011

Estonia EIP (R), EKRE (R) 1999, 2003, 2007, 2011

Finland KTP (L), PS (R), SKP (L), SKS (R), STP (L), VP (R), Muutos 2011 (R), DV (L), SKDL (L)

1987, 1991, 1995, 1999, 2003, 2007, 2011

France Front de Gauche (L), Front National (R), LCR (L), Lutte Ouvrière (L), MNR (R), MPF (R), PCF (L)

1986, 1988, 1993, 1997, 2002, 2007, 2012

Germany AfD (R), NPD (R), PDS/Linke (L), REP (R) 1987, 1990, 1994, 1998, 2002, 2005, 2009, 2013

Greece ANEL (R), Antarsia (L), EEK (L), EPEN (R), KKE (L), KKE-ML (L), KOIWNIA (R), LAOS (R), LE (L), MERA (L), M-L KKE (L), AKKE (L), PG (R), SEK (L), Syriza (L)

1989, 1990, 1993, 1996, 2000, 2004, 2007, 2009, 2012

Hungary FIDESZ (R), JOBBIK (R), MIEP (R), MM (L) 2002, 2006, 2010, 2014

Iceland IDF (R), PF (L) 1991, 1995, 1999, 2003, 2007, 2009, 2013

Ireland PBP (L), SP (L), Workers Party (L) 1992, 2007, 2012

Italy FDI (R), Forza Nuova (R), La Destra (R), Lega Nord (R), M5S (L), PCL (L), PdCI (L), PRC (L), Rivoluzione Civile (L), La Sinistra (L), Die Freiheitlichen (R), CasaPount (R), Fiamma Tricolore (R) 1987, 1992, 1994, 1996, 2001, 2006, 2008, 2013 Latvia LSP (L), NA (R), TB/LNNK (R), VL (R) 2002, 2006, 2010, 2011, 2014 Lithuania Frontas (L), JL (R), LCP (R), LSP (L), LTS (R), SPF (L), TT (R) 2000, 2004, 2008, 2012 Luxembourg KPL (L), déi Lénk (L) 1989, 1994, 1999, 2004, 2009, 2014

Netherlands Centrum Democraten (R), LPF (R), NCPN (L), PVV (R), SGP (R), SP (L) 1986, 1989, 1994, 1998, 2002, 2003, 2006, 2010, 2012 Norway FLP (R), FrP (R), NFP (R), NKP (L), RV (L), Rödt (L) 1989, 1993, 1997, 2001, 2005, 2009, 2013

Poland Kukiz 15 (R), KPN (R), LPR (R), PiS (R), PPN (R), PPP (L), SD (R) 1997, 2001, 2005, 2007, 2011 Portugal BE (L), CDU (L), PCP (L), PCTP/MRPP (L), PNR (R), POUS (L), UDP (L) 1987, 1991, 1995, 1999, 2002, 2005, 2009, 2011 Romania PNGCD (R), PP-DD (L), PRM (R), PUNR (R), PSR (L) 1996, 2000, 2004, 2008, 2012 Slovakia KSS (L), LS-HZDS (L), PSNS (R), SNS (R), ZRS (L) 1998, 2002, 2006, 2010, 2012 Slovenia Lipa (R), SNS (R), ZKS (L) 2000, 2004, 2008, 2011, 2014

Spain Izquierda Unida (L), PCPE (L), UCE (L) 1986, 1989, 1993, 1996, 2000, 2004, 2008, 2011

Sweden ND (R), NyD (R), SD (R) 1988, 1991, 1994, 1998, 2002, 2006, 2010, 2014

Switzerland AL (L), FPS (R), LEGA (R), MCR-MCG (R), PdA (L), SD (R), SVP (R), Solidarität (L)

1987, 1991, 1995, 1999, 2003, 2007, 2011

United Kingdom BNP (R), UKIP (R), SML (L) 1987, 1992, 1997, 2001, 2005, 2010

Note: For The Netherlands, SP is classified populist until the 1990 elections. For Hungary, FIDESZ is classified populist

(18)

17

IV. Methodology

To explore the hypotheses listed above about the electoral effects of globalization, economic forces and immigration, this research will focus on outcomes of parliamentary elections between 1986 and 2014 in 29 European countries. The main interest is in the percentage of votes received by left- and right-wing populist parties. Focusing on national parliamentary elections (lower chambers) provides an electoral arena that is relatively consistent in structure and in political importance across nations and time (Swank et al. 2003). The elections and populist parties that were used for this research can be found in Table 1. The selection of parties stems from the report of Johansson Heinö (2016). He bases his selection on primary materials like party programs and secondary materials like academic literature on the parties. A party is found to be populist when it meets the following five criteria: the party (1) states that it represents the people against the elite, (2) has a lack of patience with the rule of law and believes it should be broken for the sake of the majority, (3) pursuits a more powerful national state and is sceptical towards international organisations and treaties, (4) uses revolutionary language, and (5) shows significant overlap of voters on both sides of the political left-right spectrum. This leads to a selection that contains right-wing populists, left-wing populists, left-wing totalitarian parties (communists), right-wing totalitarian parties (fascists, neo-Nazis) and parties that want curtailing of liberal democracy on religious (evangelical) grounds. For some parties the classification changes over the years, as for example Fidesz in Hungary that ‘became’ populist in 2002.

Looking at the period between the late 1980s until up a few years after the financial crisis of 2008 should give a clear image about what the effects of the several economic and social forces are on populist parties, and to what extent these have changed because of the global financial crisis. While former research left out countries that were part of the Soviet-bloc due to theoretical and methodological reasons (e.g. Golder, 2003; Swank et al. 2003; Arzheimer, 2009), this paper will include some of these countries. The goal of this is to see the differences in populist movements in these countries. Moreover, populist parties have gained a large influence in Eastern Europe, with the main examples being Hungary and Poland. Potential selection bias for studies analysing the factors that influence the electoral success of populist parties can be a significant problem. It can be a temptation to leave out countries where this kind of parties simply do not exist. But this would lead to biased and inconsistent estimates, since these countries might have factors that discourage the rise of populist parties (Jackman et al. 1996; Golder, 2003). In this study, observations without data for populist parties will be left out of the process. Countries with populist parties that have no electoral support will not be excluded for that reason. Instead the support will be set at zero.

The hypotheses listed in the previous section will be tested using the model shown in equation 1. The model will be calculated by using both the vote share for left- as for right-wing populist parties in the European nations.

(19)

18 where e denotes a national election year, 𝛽0 is the equation intercept (or constant), GROWTH

through VOTESHARE_𝑒−1 are the explanatory variables, 𝛽₁− 𝛽₇ are parameters relating GROWTH through VOTESHARE_𝑒−1 to VOTESHARE. 𝜀 is an error term. 𝛽₈₋₃₆ are country dummies, to create a so-called ‘fixed effects’ estimator that would otherwise be impossible in a Tobit estimation. Through these dummies I attempt to take into account potential country heterogeneity (Beck, 2001; Greene, 2001; Golder, 2003). The same is true for SOVIET, which is a dummy for ex-Soviet nations. Including the vote share received by populist parties during the previous election (VOTESHARE_𝑒−1) as previously used by Swank et al. (2003), should work as a corrective for serially correlated errors. Besides, lagged vote share provides and indicator of the (potential) instability in populist party support (Beck and Katz, 1996). The dependent variable VOTESHARE in the model measures the percentage of electoral support populist parties received. The explanatory variables (except for GROWTH and CRISIS) are all observations from the year prior to the election (at t - 1). This is due to voters basing their opinion on previous experiences and knowledge.

The hypotheses are operationalized via the explanatory variables in this base model. The economic hardship hypothesis will be measured by the variables GROWTH and UNEMPLOYMENT. GROWTH is the sum of growth (in percent) of GDP per capita during three years prior to the election, following De Bromhead et al. (2013). Their research shows that these three years are the most decisive when it comes to voting. Part of GROWTH is also the square of this sum of three years of growth, since De Bromhead et al. (2013) find evidence that the impact of growth is nonlinear. Therefore, it is important to take both into consideration when analysing the impact of economic growth in the next section. GDP per capita growth is chosen as a measurement instead of GDP growth, since this will have a more direct effect on the voters’ personal experience. Including this measurement hopefully gives a clearer picture of the impact of economic growth, since academics have not found a consensus on this yet. UNEMPLOYMENT is measured as the level of unemployment in percentages (at t – 1). This variable is included following the idea that an increase in unemployment will positively impact the support for populist parties, which is in line with the economic hardship theory.

(20)

19 the following sections should be interpreted as follows: a one percent increase in the percentage of the asylum population in a country, leads to a (𝛽3/100)/1000 change in the vote share of

populist parties in that country.

The GLOBALIZATION-variable takes into account globalization on three different dimensions: economic, social and political. While other research mainly focused on economic globalization by merely taking into account openness to trade and/or capital mobilization, I assume globalization goes beyond that in affecting voters’ preferences. Nowadays, one of the main points of interest to populist parties is not only to propagate against free trade, but also to international institutions like the EU and the IMF (Mudde, 2004). Therefore, I introduce the KOF Globalization Index (from Dreher, 2006) as a standard score index for all three dimensions of globalization into this research. This measurement takes into account the economic flows (trade, FDI and income), restrictions (e.g. import barriers and tariffs), international personal contact (through e.g. telephone traffic, internet usage and tourism), and political globalization (e.g. number of embassies in the country, membership in international organizations and treaties). Therefore, this index also includes the effects of European integration, to which many populist parties object.

This research focuses on the differences of populism on before and after the financial crisis of 2008. To take this into account, a dummy variable (CRISIS) is set at 1 for all the years from 2008 until the most recent observation, and at 0 for all years prior to the crisis.

The use of lagged variables has at least two advantages. First, it makes sense when analysing voters’ preferences, to include factors that they have ‘experienced’. Since elections can also be in the beginning of a year, the occurrence of for example an increase in unemployment has not yet happened in that same year. Second, the use of lagged explanatory variables diminishes the effect of potential endogeneity (reverse causality) (Justesen, 2008).

Extensions

Regarding social protection and the discouraging effect it might have on populist support, I include a measurement of expenditure on social protection benefits for the unemployed (WELFARE). It measures the expenditures in euros per inhabitant at constant 2010 prices, making it comparable over time and across countries. For a better readability the observations are divided by 1000.

(21)

20 It is important to remember that results for interaction models cannot be interpreted as in regular additive models, because these coefficients are conditional. The information that can be gathered from this single coefficient is often severely limited and can be quite misleading when it comes to interpretation (Golder, 2003). Therefore, the full range of conditional standard errors need to be calculated. Otherwise, I would fail to provide an appropriate assessment of the uncertainty concerning the variables included in the interaction terms. In this research, significant interactions will be calculated and visualized to explain their full meaning. After the original model, significant variables will be tested for as interactions with crisis to capture the effect of the global financial crisis on these factors. Significant interactions from this estimation will be sufficiently analysed following Golder (2003) as well.

Other extensions of the base model from equation 1 are tests on the effect of the global financial crisis, the different dimensions of globalization and regional inequality. These will be described in more detail in the Robustness-section of the next chapter. Another part of the robustness checks will comprise the potential heterogeneity that might be held by different regions in their relationship with support for populist parties. More specifically, I will test for the differences when the sample does not include former Soviet nations, GIPS (Greece, Ireland, Portugal and Spain), and nations with a universal welfare state (mainly the Nordics).

Descriptive statistics

An explanation of the data and its sources are described in the Appendix Table 1. The descriptive statistics for all dependent and independent variables are found in the Appendix Table 2. What should be noted before analysing these statistics that observations are only for years of election. Besides, observations are not for every election due to the fact that for most variables data has become more widely available only for more recent years. For some, this is due to the falling apart of the Soviet Union, making measurement for certain factors only available in the course of the 1990s. Moreover, measuring social protection expenditures (captured in WELFARE), data is available for most countries between 1995 and 2015, while for some even less. This leads to less observations (135) than for other variables (184). The same is true for ‘Regional Inequality’. This variable is based on data only available for 2004 until 2015, hence there are only 59 elections for which data is completely provided.

(22)

21 Analysing the logged variable for the level of asylum seekers in a country I observe a relatively large disparity too among the minimum and maximum values. There are especially large differences in the lower fifty percent of the observations. Remember that the variable for asylum applicants is a log-variable, which performed a better fit after several tests. The measurement is the number of asylum applicants divided by the population of the country where they apply. In this way, the outcomes can be compared across countries. A smaller country would be more impacted by a large amount of asylum seekers applying in that country, and vice versa. The globalization indices (total, economic, social and political) show similar trends among each other. Some interesting points can be made. For one, political globalization has a clearly higher mean than the others. This is mainly due to the fact that most countries in the sample are members of the EU and NATO. Second, the minimum for the total level of globalization is higher than the three different dimensions. The total shows the mean of the three dimensions and hence indicates that if one country scores low on one of these dimensions, the other two dimensions are relatively a lot higher.

Descriptive statistics for social protection benefits (WELFARE) show a clear West-East divide; the lower values for this variable are found in Eastern Europe, while the higher levels can be found in Western European nations, especially in Scandinavia, Luxembourg and Switzerland. The minimum, €470, is paid for social protection benefits per inhabitant before the 2004 election in Romania, while the maximum of over €18.000 is paid before the election in Luxembourg in 2014.

The dummy variable for former Soviet nations indicates that 23.3% of the elections in the sample were in these countries. After the global financial crisis, 28.2% of the elections in the sample were held, indicated by the dummy for crisis. Finally, for regional inequality Appendix Table 2 indicates that there is a large variation between the observations for the difference between incomes in rural and urban regions in a country. For some countries, like in Italy, it is as low as 10.5%, while in several East European nations the incomes in urban regions are two-thirds higher on average. In most other countries, the divide is around one-third, which is also indicated by both the mean and the fiftieth percentile.

(23)

22

Estimation

For many elections in the sample the support for populist parties is zero. Out of 184 elections in the sample, this is the case for 42 elections for left-wing populist parties and for 35 right-wing populist parties. To account for this, I will use a Tobit model that utilizes a maximum-likelihood estimator for left-censored variables. It is widely used to study the rising support for left- and right-wing populist parties (as for example Jackman et al. 1996; Golder, 2003; Coffé et al. 2007; Jesuit et al. 2009). The result is a much more realistic model of the process generating censored data and may be interpreted as if from a linear normal regression with no censoring. The estimated coefficients represent the marginal effect of the independent variables on the underlying support for extreme right parties (Golder, 2003). The Tobit model extends the familiar probit procedure for estimating a non-linear probability model of a dichotomous variable in the case of a censored dependent variable and produces consistent and unbiased estimates of the beta’s (Swank et al. 2003). A standard Tobit model looks like this:

𝑦𝑖 = {

𝑦_𝑖∗ 𝑖𝑓 𝑦_𝑖∗> 0 0 𝑖𝑓 𝑦_𝑖∗≤ 0 where 𝑦_𝑖∗ is a latent variable:

𝑦_𝑖∗ = 𝛽𝑥_𝑖 + 𝑢_𝑖, 𝑢_𝑖 ~ 𝑁(0, 𝜎2)

Censoring at a value 𝑦_𝐿 different from zero (left-censoring) is used in this paper:

𝑦_𝑖 = {𝑦𝑖

∗_{𝑖𝑓 𝑦} 𝑖∗> 𝑦𝐿

𝑦_𝐿 𝑖𝑓 𝑦_𝑖∗≤ 𝑦_𝐿

In the tables below, the marginal effects of the independent variables on the latent variable are shown, since these indicate the direct effects of these independent variables. These marginal effects follow from:

(24)

23

V. Results

The results of the statistical model are reported in Table 2 and 3. Table 2 reports the results with the dependent variable being left-wing populist vote share. Table 3 does the same, but for right-wing populist vote share. For readability reasons, the country dummies are not shown. The country dummy variable for the United Kingdom is dropped in each case. Therefore, the United Kingdom acts as the base level against which other country effects can be compared. Even though several of the country dummy variables appear to be insignificant, a log-likelihood test indicates that they are jointly significant and must be included in a correctly specified model (Golder, 2003; Gujarati, 1995). The reported Sigma in the Tables is the estimated standard error of the regression and is analogous to the square root of the residual variance in OLS regression. Robust standard errors are adopted to deal with potential heteroscedasticity of the data.

Among some of the independent variables a strong correlation is present as can be seen in the correlation matrix in Appendix Table 3, which might make it hardly possible to distinguish between the individual effects of these variables on the dependent variable (vote share) due to potential multicollinearity among the covariates. Therefore, from the base model (1) onwards in both Table 2 and 3, I added the interaction variables one by one to assess their effect on the other independent variables and the dependent variable. Model 4 includes the social protection benefits and its potential mediating effects on unemployment, the number of asylum seekers and globalization. These variables are added only then, since their inclusion reduces the number of observations strongly. In both tables, model (5) shows the full model with all the variables from the four previous models included.

(25)

24 Table 2 – Results of Tobit estimation on left-wing populist vote share

(1) (2) (3) (4) (5) VARIABLES Base Unem. x Asylum Unem. x Glob. Welfare Full GDP growth (sum t-1 to t-3) -0.351*** -0.319*** -0.370*** -0.420*** -0.382*** (0.0122) (0.0123) (0.0121) (0.0134) (0.0131) GDP growth² (sum t-1 to t-3) 0.0131*** 0.0122*** 0.0142*** 0.0167*** 0.0171*** (0.000670) (0.000683) (0.000663) (0.000785) (0.000780) Unemployment (t-1) 0.179*** 0.360*** 0.936*** 0.0281 3.310*** (0.0182) (0.0183) (0.0198) (0.0211) (0.0215) Asylum (log, t-1) -0.524*** -1.185*** -0.542*** 1.114*** -0.0324 (0.0556) (0.0614) (0.0554) (0.0840) (0.0853) Globalization (t-1) 0.0122*** 0.00476** 0.0933*** -0.288*** 0.0649*** (0.00221) (0.00219) (0.00229) (0.00262) (0.00256) Unemployment x Asylum 0.0948*** 0.111*** (0.00616) (0.00789) Unem. x Globalization -0.0100*** -0.0412*** (0.000253) (0.000267) Welfare (t-1) -3.715*** -4.068*** (0.0219) (0.0217) Welfare x Unemployment 0.0209*** 0.0685*** (0.00353) (0.00350) Welfare x Asylum -0.104*** -0.0305*** (0.0108) (0.0101) Welfare x Globalization 0.0370*** 0.0373*** (0.000251) (0.000249) Ex-Soviet -21.16*** -20.71*** -21.12*** -19.55*** 1.031*** (0.288) (0.288) (0.288) (0.305) (0.179) Crisis -0.209 -0.0876 -0.208 0.425** -19.09*** (0.146) (0.143) (0.151) (0.185) (0.301)

Left Populist vote (e-1) 0.615*** 0.582*** 0.609*** 0.527*** 0.496***

(0.0133) (0.0130) (0.0135) (0.0124) (0.0116) Sigma 3.314*** 3.273*** 3.293*** 2.940*** 2.814*** (0.0128) (0.0133) (0.0127) (0.00857) (0.00759) Constant -4.069*** -4.881*** -10.21*** 25.49*** -3.128*** (0.183) (0.181) (0.189) (0.220) (0.215) Observations 184 184 184 135 135

Country Dummies YES YES YES YES YES

Log Likelihood -385.7 -383.7 -384.9 -256.4 -252.0

Left-censored obs. 42 42 42 35 35

(26)

25 Table 3 – Results of Tobit estimation on right-wing populist vote share

(1) (2) (3) (4) (5) VARIABLES Base Unem. x Asylum Unem. x Glob. Welfare Full GDP growth (sum t-1 to t-3) -0.199** -0.195* -0.199* -0.0981 -0.0931 (0.101) (0.0995) (0.101) (0.0942) (0.0883) GDP growth² (sum t-1 to t-3) 0.0106** 0.0100** 0.0109** 0.0146*** 0.0123*** (0.00432) (0.00432) (0.00483) (0.00451) (0.00440) Unemployment (t-1) -0.157 -0.0904 0.157 0.545* -0.526 (0.197) (0.228) (1.784) (0.321) (2.192) Asylum (log, t-1) 0.894* 0.395 0.905* 1.174 -0.567 (0.517) (0.680) (0.508) (0.896) (1.293) Globalization (t-1) 0.148*** 0.149*** 0.177 -0.262 -0.467 (0.0537) (0.0539) (0.171) (0.182) (0.369) Unemployment x Asylum 0.0597 0.132* (0.0736) (0.0748) Unem. x Globalization -0.00401 0.0180 (0.0230) (0.0297) Welfare (t-1) -5.206 -7.048* (3.695) (4.079) Welfare x Unemployment -0.0944* -0.121** (0.0486) (0.0559) Welfare x Asylum -0.186 -0.0921 (0.130) (0.134) Welfare x Globalization 0.0679* 0.0904** (0.0390) (0.0445) Ex-Soviet 3.285* 3.265* 3.284* -0.566 3.179** (1.704) (1.702) (1.706) (2.719) (1.365) Crisis 0.666 0.623 0.712 3.159** -1.452 (1.232) (1.220) (1.304) (1.388) (2.938)

Right Populist vote (e-1) 0.349*** 0.344*** 0.350*** 0.218*** 0.193***

(0.0726) (0.0726) (0.0717) (0.0488) (0.0528) Sigma 4.645*** 4.636*** 4.645*** 3.941*** 3.904*** (0.397) (0.390) (0.401) (0.296) (0.291) Constant -9.366** -9.924** -11.69 20.22 35.53 (4.603) (4.707) (13.64) (17.30) (30.62) Observations 184 184 184 135 135

Country Dummies YES YES YES YES YES

Log Likelihood -448.8 -448.5 -448.8 -318.7 -317.4

Left-censored obs. 35 35 35 22 22

(27)

26 as Golder (2003) and Arzheimer (2009) did. Findings for the asylum coefficient are remarkable in the sense that an increase in asylum applicants reduces the left-wing, while it raises the right-wing populist vote share. Even though the effects are very small, since they should be divided by 1000, increases in asylum applicants can be relatively large and sudden in the recent refugee-crisis.

From the results, I can confirm that there is heterogeneity in the influences of the support for right-wing populist parties, possibly due to a country’s years of having a democracy. This follows from the results that for former Soviet nations the vote share is increased by several percentage points and is significant at a 5% confidence level. Furthermore, in both tables I find the corresponding vote share for populist parties in the previous election to be highly significant. This implies that there is a certain consistency in the support received by populist parties. This idea was previously dispatched by Swank et al. (2003) (based on data from before the crisis), but Stockemer (2017) has similar findings. Analysing the results for the hypothesis that globalization breeds populist support, I can tentatively confirm this idea for at least right-wing populism. In both model (1) and (2) from Table 3, I find that a one-point increase on the index for globalization raises the populist vote share slightly. Surprisingly, the only significant effect of globalization on left-wing vote share is negative. But this might be affected by the interactions with social welfare benefits. If not, it might be that there is a link between globalization and the decreasing influence of communist parties across especially Eastern-Europe, or that these parties make way for right-wing populist parties since they are known to advocate more strongly against globalization.

Finally, the effects of social benefits and the presumed mediating effects do not follow the hypotheses entirely. For left-wing populist support, I find that an increase of €1.000 per inhabitant would diminish the vote share strongly (by almost four percentage points). However, the mediating effects of social benefits are only confirmed when it comes to the level of asylum applicants, while globalization and unemployment seem to have a rather small, but increasing effect on the vote share for populist parties. These findings seem to hold in the full model, but then the effect of asylum applicants turns insignificant. For support for right-wing populists, there are similar but less significant effects. Only in the full model (6) in Table 3 I find that social protection benefits lower right-wing support. Again, there is a small mediating effect for unemployment found that can be derived from the interaction variable, but the coefficient for the interaction between welfare and globalization is positive as it was for left-wing support. This would imply that in more globalized countries increasing social benefits would lead to a higher left- and right-wing populist vote share.

Interaction variables

(28)

27 (a) Effect of asylum applicants

(b) Effect of unemployment

Figure 1 – Graphical representation of marginal effects of asylum and unemployment on left-wing populist vote share

-. 5 0 .5 1 Ef fe ct s o n Pr (L e ft -wi n g vo te ) -10 -5 0 5

Asylum seekers (log, t-1)

Average Marginal Effects of unemployment on left-wing populist vote share (95% confidence interval)

-1 0 1 2 Ef fe ct s o n Pr (L e ft -wi n g vo te ) 0 5 10 15 20 25 Unemployment (t-1)

(29)

28 (a) Effect of welfare

(b) Effect of unemployment

Figure 2 – Graphical representation of marginal effects of welfare and unemployment on right-wing populist vote share

-1 5 -1 0 -5 0 5 Ef fe ct s o n Pr (Ri g h t-w in g vo te ) 0 5 10 15 20 25 Unemployment (t-1)

Average Marginal Effects of welfare on right-wing populist vote share (95% confidence interval)

-3 -2 -1 0 1 Ef fe ct s o n Pr (Ri g h t-w in g vo te ) 0 5 10 15 20 Welfare (t-1)

(30)

29 The thick line is the effect, the outside lines the confidence intervals. Note again that results are only significant when both the confidence intervals are above or below y=0 (Golder, 2003). From Figure 1a, I can conclude that at high levels of unemployment increases in asylum population leads to more left-wing populist support. But at levels of unemployment below ten percent, it has the opposite effect and this is mostly the case for European countries (127 out of 184 elections). Interestingly, Figure 1b shows that even for relatively low levels of asylum applicants, an increase in unemployment has an increasing effect on the vote share. These results are more in line with the expectation that a larger asylum population will lead to an increase in the populist vote share.

In Figure 2, the effects of the interaction between welfare and unemployment on the right-wing vote share are shown. Figure 2a indicates that social protection benefits influence the vote share negatively for any level of unemployment, but these results become truly significant at high levels of unemployment of around 18%. There are only seven elections with a higher unemployment level than this. Figure 2b depicts that there is no significant effect between the levels of social protection benefits and the impact of unemployment on the right-wing populist vote share.

Robustness

Since the impact of sample composition is an important consideration in models such as those estimated above (Swank et al. 2003), I will estimate several models in this section that exclude certain regions of Europe. Every group is split into two models, since there are fewer observations for social protection benefits. The outcomes of the re-estimations can be found in Appendix Tables 4 and 5. Due to the significance of the dummy variable for former Soviet states in the previous models, the first model will exclude these nations in its estimation. This is another test to see if and to what extent being a former Soviet nation impacts the vote share for populist parties. When the global financial crisis hit Europe, certain countries suffered more than others and even needed bailing out. These nations were called GIPS (Greece, Ireland, Spain and Portugal). It was also in these countries that the left-wing populist parties (re-)emerged after the crisis. To estimate the impact of these countries on the original model, they are the second group that is omitted in the tables. The final group excludes the Nordic countries, due to their universal welfare state. These countries are Norway, Sweden, Finland, Denmark and Iceland. These re-estimations drastically decrease the number of observations and are therefore a profound test for the model’s sensitivity.

The Rise of Populism Assessing the Economic Hardship Theory after the Global Financial Crisis