Voting radically: Why Greeks vote for the
challengers?
Explaining voting behavior in light of transnationalism and the
multiple crises.
Student: Karmen Misiou Supervisor: Theresa Kuhn Second reader: Wouter van der Brug
June, 2020
ii ABSTRACT
The paper attempts to study party preferences in crisis-ridden Greece. Within this context, it investigates whether voting for challenger- radical parties constitutes a form of protest voting. At the same time, the effect of the economy, immigration, and globalization are also taken into account. The empirical analysis draws upon theories concerning spatial models. In other words, distances between voters and parties in terms of ideological and policy issue parameters are taken into account when studying party choice. Against this backdrop, the paper provides evidence that voting for challenger parties in times of crisis is, to some extent, protest voting. Moreover, focusing on attitudes towards globalization, immigration, and the economy shows different patterns regarding party preferences. While the effect of immigration is similar in casting a ballot for either challenger or mainstream parties, economic voting explains better voting for the latter ones. This is the case since responsibility attribution plays out differently in party preferences. Whereas voting for the radicals is intertwined with attributing responsibility to external actors casting a ballot for their mainstream counterparts implies attributing responsibility to the government. Finally, the globalization dimension tends to be significant when voting for the challengers implying that the so-called losers of globalization are willing to vote only for radical parties.
iii TABLE OF CONTENTS
ABSTRACT………..………...…. ii
ACKNOWLEDGMENTS…...………..………….….…… iv
1.
INTRODUCTION………..………...12.
EXPLAINING VOTING BEHAVIOR IN LIGHT OF NOVEL CONFLICTS………..……….……...33.
NEW SOCIAL CONFLICTS: AN OPPORTUNITY FOR CHALLENGERS’ ELECTORAL ESTABLISHMENT?………..………....……...53.1.ATTITUDES TOWARDS GLOBALIZATION AS A DETERMINANT OF PARTY PREFERENCES……….…………..…...…....8
3.2.ATTITUDES TOWARDS IMMIGRATION AS A DETERMINANT OF PARTY PREFERENCES ………..….…..9
3.3.EVALUATIONS OF THE ECONOMY AS A DETERMINANT OF PARTY PREFERENCES………..……..…..10
4.
DATA………..…….…...125.
RESEARCH DESIGN ………..….……….15 5.1.ROBUSTNESS………..…………..176.
RESULTS ……….………..………....18 6.1.MODEL-SET 1………..………..…18 6.2.MODEL-SET 2……….…………...………...….…20 6.3.MODEL-SET 3………...………...…..…22 6.4.MODEL-SET4 ...………...…...247.
DISCUSSION………..………268.
LIMITATIONS ………....…...………...….309.
CONCLUDING REMARKS ………...…....…...…32 REFERENCES ………...……….…...…………...34 APPENDIX ……….……….………...……...……...40 A.1.: PARTIES………..………….………...…….…40iv
ACKNOWLEDGEMENTS
I would like to express my special thanks to my current supervisor Dr. Theresa Kuhn for guiding me and supervising my work even though the COVID-19 crisis has drastically reshaped the fundamentals of teaching. Moreover, this work would be impossible without the valuable suggestions and help of Dr. Wouter van de Brug who was always available when I needed direction.
I also take this opportunity to express a deep sense of gratitude to the professors and the study advisor of the Graduate School of Social Sciences for providing me with all the help needed despite the significant difficulties due to the COVID-19 crisis.
Special thanks to the European Politics & External Relations master track coordinator Dr. Dimitris Bouris for his precious counseling.
I also wish to express my thankfulness to my family, my parents Dimitrios and Argiro and my sister Rania, for supporting me all these years. Accomplishing my dream of studying at the University of Amsterdam would be impossible if they were not standing by my side.
1
1. Introduction
Voting behavior constitutes a puzzling topic analyzed in light of different theories. The paper addresses the rational choice theory and ideological voting in times of crisis. Within this context, Greece constitutes a compelling case since both the economic and the refugee crises have played a decisive part in reshaping national party politics. The principal question of the paper relates to the significant factors driving party choice. Since the outbreak of the Euro crisis, challenger parties on both sides of the political spectrum have experienced a profound electoral reinforcement confronting their mainstream counterparts. Studying the impact of ideology on casting a ballot for a radical party constitutes the first imperative of the research. In other words, in order to explain what drives party preferences, one should first identify the effect of both ideological and policy issue- related parameters in voting behavior. The paper draws upon the idea that casting a ballot for the radicals constitutes a form of protest voting. At the same time, the economy, immigration, and globalization being at the heart of new social conflicts constitute vital parameters when studying party choice. Against this backdrop, ideological parameters such as the left-right dimension are expected to be less critical when voters decide to cast a ballot for the challengers.
Therefore, the research questions of the paper are the following. “To what extent does casting a ballot for the challenger- radical parties constitute a form of protest voting? What is the effect of the economy, immigration, and globalization in electoral behavior in times of crisis?”.
Globalization has caused emerging conflicts related to integration and immigration (Hooghe & Marks, 2009). At the same time, the eruption of the economic crisis has brought economic explanations back when studying electoral behavior (Lewis-Beck & Nadeau, 2012). Voters who have been affected by globalization and the crises have challenged mainstream politics. The paper draws upon the idea that protest voting in times of crisis constitutes a rational choice aimed at punishing mainstream political actors (Kselman & Niou, 2010). Voters hold incumbents accountable for implemented policies related to the economy and immigration management. Moreover, European integration constitutes a further dimension of conflict (Hooghe & Marks, 2018). Hence,
2
voting for challenger parties is expected to be intertwined with the issues mentioned above. Finally, ideology plays out differently among Western party politics. In Greece, the left-right dimension had constituted a principal factor when studying party preferences before the crises (Dinas, 2008). Thus, studying the explanatory power of the left-right dimension in electoral behavior during the crises could contribute to interpreting ongoing electoral patterns.
The Great Recession has favored the emergence and the reinforcement of radical left-wing parties in Southern Europe (Kriesi, 2016), whereas only after the 2015 refugee crisis has the radical right been strengthened (Hutter & Kriesi, 2019). While ideological and party identification is considered to play a fundamental role in party choice, even in light of globalization, further factors related to voters’ short-term evaluations should also be taken into account. Against this backdrop, studying the effect of economic and issue voting is essential. Thus, many scholars have emphasized the advantages of using individual-level evaluations when studying electoral behavior. Retrospective evaluations about the economy (see Holbrook & Garand, 1996) and about policies on immigration (see Mudde, 1999) as well as attitudes towards globalization (see Hellwig, 2008) play out as considerable indicators in party preferences.
The paper contributes to the scholarship on Greek party politics by providing an alternative empirical approach. Specifically, in order to utilize the effect of the parameters mentioned above in party preferences, the paper proceeds to an analysis where an alternative dependent variable is used. Instead of building models based on actual party choice, the paper has used the probability to vote questions in order to create a new dependent variable where probabilities for all parties are included. In so doing, estimating the effect of the independent variables in probability to vote for all parties is feasible. Moreover, combining variables related to both economic and non-economic factors could contribute to a better interpretation of voters’ rationales. Therefore, the paper attempts to give insights regarding the impact of the economy, immigration, and globalization on voting behavior.
3
2.
Explaining voting behavior in light of novel conflicts
Electoral behavior reflects a strategic process (Black, 1978; Cain,1978; Kselman & Niou, 2010), where voters take into account parties’ probabilities to win the elections. Downs (1957) argues that party preferences apply to rationality in the sense that voters assess parties’ utilities in order to decide which party fits better to their interests. The strategic and rational choice models could also apply to protest voting. Specifically, voters who are dissatisfied with policy outcomes tend to vote for a less-preferred party in order to indicate their disagreement with implemented policies (Kselman & Niou, 2010). In this vein, Treib (2014) claims that Eurosceptic voting in the 2014 European elections has a dual rationale. On the one hand, it is about calling into question EU policies. On the other hand, it functions as a form of protest against governments’ political agendas. On the contrary, Altomonte et al. (2019) interpret protest voting as a process related to emotional parameters and social identities.
It comes as little surprise that the emergence of globalized societies has triggered new reactions and conflicts challenging traditional party politics (see Kriesi et al., 2008; Hooghe & Marks, 2018) and requesting for new interpretations regarding electoral behavior. Traditional party families have faced severe contestation, often leading to their decline. At the same time, new types of parties tend to gain electoral success challenging their mainstream counterparts. The gradual transformation of traditional parties in Europe into state-centered political actors has challenged their legitimacy (Ignazi, 2014: 163). At the same time, while traditional cleavages still play a significant role in party preferences, new determinants are also influential leading parties to adopt a catch-all approach (Mair, 2008:220).
Within this context, concepts such as voter dealignment and realignment have come into existence, attempting to interpret political behavior. Dealignment draws upon the idea that parties do not address specific social groups (Marks et al., 2017:2). Accordingly, realignment highlights that the attenuation of traditional cleavages coincides with the emergence of new ones reflecting ongoing conflicts (Marks et al., 2017:14). Therefore, integration and immigration could explain realignment with challenger parties in light of transnationalism. At the same time, parties’ positions on the GAL/TAN scale are also important when analyzing party preferences (Marks et al., 2017: 14-15).
4
This being said, one should not assume that ideology does not affect party choice in contemporary politics. According to relevant research, party and ideological identification seem to play out differently among Western societies. However, voters’ ideological self-placement tends to explain party choice adequately. According to relevant research, ideological parameters such as the left-right dimension are intertwined with European integration, highlighting the magnitude of the economy in both national and supranational political arenas (Hooghe et al., 2002). Moreover, studying the effect of ideological self-placement in voting behavior in Portugal, Lewis-Beck and Costa Lobo (2011:302) argue that ideology has a more substantial effect than party identification.
5
3. New conflicts: An opportunity for the challengers’ electoral
establishment?
Crises function as turning points bringing controversial issues into the public sphere and causing intense politicization. Against this backdrop, integration might be challenged while new structuring conflicts occur (Hutter & Kriesi, 2019). Unsurprisingly, mainstream political parties across Europe fail to deal with the crises-driven dynamics, since they suffer from lack of either responsiveness or responsibility (Bosco & Verney, 2012:132). Thus, a significant part of European electorates pivot challenger- radical parties. Hobolt and Tilley (2016) argue that voting for challenger parties constitutes a twofold process. On the one hand, voters attempt to punish governments and mainstream political parties. On the other hand, choosing a challenger party relates to attitudes towards EU integration, migration, and the economy (Hobolt & Tilley, 2016:13).
The paper draws upon Hobolt’s and Tilley’s work, whereby integration, immigration, and the economy function as critical factors in voting for challenger parties. Moreover, it adopts the authors’ distinction between three different types of challenger radical parties, namely right-wing, left-wing, and green parties (Hobolt & Tilley, 2016). However, since there are no Green- challenger parties in Greece, only the first two categories apply (see A1). It is worth mentioning that although populism constitutes a well-established theory capable of explaining contemporary party politics, it is not used as a theoretical framework in the paper. This is because particular characteristics regarding party politics in Greece constitute populism, a less sufficient concept when analyzing radicals’ electoral success1 (Pappas, 2013; Vasilopoulou, 2018:2).
As mentioned above, the emergence of new critical divisions between societies has led to the formation of new patterns in voting behavior. It comes as little surprise that the outbreak of the economic and the subsequent refugee crisis has significantly challenged the pre-existing political consensus in Europe. Within this context, Greece constitutes a compelling case since it has tackled both crises facing severe disorders.
1 According to many scholars, even before the crises, mainstream political parties such as the
centered-left PASOK had often adopted a populist political discourse. Thus, populism has not been exclusively entailed in the public discourse of radical parties.
6
Previous research has identified a sharp decline regarding citizens’ support for the EU after 2010 (Braun & Tausendpfund, 2014:237) as well as, a striking reinforcement of left-wing and right-wing radical parties (see Dinas et al., 2013; Ellinas, 2013; Stavrakakis & Katsambekis, 2014; Michailidou, 2016) capitalizing on social resentment and either economic or cultural factors.
The eruption of the economic crisis and the subsequent bailouts of Greece led to growing financial instability and an increase of Euroscepticism having severe implications for mainstream political actors (Vasilopoulou, 2018:6). Within this context, the Greek radical-left gained profound electoral visibility capitalizing on social resentment as well as on economic insecurity (see Moschonas, 2013; Panayiotakis, 2015; Tsatsanis & Teperoglou, 2016). On the other hand, the radical right took advantage of growing opposition to immigration as well as the subsequent refugee crisis by emphasizing its impact on national sovereignty (see Georgiadou, 2013; Ellinas & Lamprianou, 2016; Halikiopoulou & Vasilopoulou, 2016). The dual 2012 parliamentary elections constituted the first crunch in the Greek political system after the restoration of democracy since a profound part of the vote share was in favor of both SYRIZA and Golden Dawn2. The double 2012 election, also known as the electoral earthquake, was followed by the 2015 one where SYRIZA took office, forming a government coalition with the radical right-wing Independent Greeks3.
The sovereign debt crisis has contributed to the introduction of new concepts concerning electoral behavior in Greece. Although ideology and partisanship were prominent in party preferences (Dinas, 2008) austerity, politics put the spotlight on a further dimension related to European integration and globalization. Against this backdrop, Katsanidou and Otjes (2015:6-7) argue that attitudes towards the economy and the EU are intertwined, while non-economic factors relate to the traditional left-right dimension. In this vein, Vasilopoulou (2018:7) claims that challenger parties took advantage of the severe implications of the crisis in order to function as a countervailing force to EU policies. However, challengers’ Euroscepticism entails different
2 The vote share for SYRIZA and Golden Dawn for the June 2012 parliamentary elections was 26.9%
and 6.9%, respectively (Hellenic Parliament, 2020). The profound electoral reinforcement of Golden Dawn was alarming since hate speech and racist violence constitute party members’ common practice.
3 In January 2015, radical left-wing SYRIZA and radical- right-wing Independent Greeks formed a
government coalition representing 36.3% and 4.8% of the vote share, respectively (Hellenic Parliament, 2020).
7
characteristics when one takes into account economic and cultural dimensions, ideology, and time (Pirro, et al., 2018; van Elsas et al., 2016). Gemenis and Dinas (2010:191) test the effect of three parameters related to the economy, cultural factors, and integration in party positions, showing that the third one depicts more sufficiently the division between pro and anti- Western parties.
By and large, the outbreak of the crisis led to a sharp decline of party loyalty as well as to the electoral reinforcement of radical parties. Since the Greek party system was volatile, new political actors capitalized on social resentment, while a profound dealignment from mainstream parties took place. According to Tsatsanis and Teperoglou (2016:5), realignment took shape gradually leading to the electoral reinforcement of both right-wing New Democracy and radical left-wing SYRIZA in terms of a new two-party system. In their previous research, the authors claim that the crisis has contributed to the emergence of long- term electoral patterns. According to their findings, confrontation with mainstream parties has not exclusively derived from their poor economic performance. Rather, it depicts an overall declining trend in terms of trusting the political system (Teperoglou & Tsatsanis, 2014:231). Indeed, Texeira et al. (2014:513) studying the impact of the crisis on democratic legitimacy show that there is a correlation between insufficient economic performance with evaluations of democracy, leading to the electoral reinforcement of anti-systemic parties.
Within this concept, the paper studies party preferences in light of ideological and protest voting. If voting for challenger parties constitutes a form of protest, ideology should play a less critical role. Therefore, the left-right dimension would have a smaller effect in casting a ballot for radical parties compared to mainstream ones. Studying voting for anti-immigrant parties, van der Brug et al. (2000) have introduced two parameters in order to identify protest voting, namely ideology, and party size. However, party size is not used as a factor in the paper since challenger parties have drastically increased their size during the crises4. Therefore, the left-right dimension is introduced in the models functioning as a proxy for protest voting. At the same time, perceptions about the economy, immigration, and European unification are used in
4 Specifically, 121 and 194 out of 300 parliamentary seats were occupied by challenger parties in the
2012 and 2015 elections, respectively (Hellenic Parliament, 2020). It should be mentioned that the first party takes a bonus of 50 seats.
8
order to capture critical divisions between voting for either challenger or mainstream parties. Within this context, the key hypotheses are the following:
H1: Ideology plays a less important role in voting for challenger parties compared to mainstream ones (protest voting hypothesis).
H2: Voters who oppose globalization vote for radical- challenger parties (winners and losers of globalization hypothesis).
H3: Voters who oppose immigration are willing to vote for a challenger party (policy issue voting hypothesis).
H4: Negative economic evaluations drive voting for challenger parties (economic voting hypothesis).
3.1.Attitudes towards globalization as a determinant of party preferences
Globalization has led to the emergence of new conflicts and the development of new voters’ blocs (Kriesi et al., 2008). As Hooghe and Marks (2018) argue, the so-called winners and losers of globalization tend to outweigh former social divisions in terms of establishing a new cleavage. Although arguing in favor of the development of a new transnational cleavage could be exorbitant, it is essential to take emerging conflicts into account when studying voting behavior. Kriesi et al. (2008) argue that the new cleavage constitutes the outcome of new social conflicts. Therefore, the economic crisis has severely challenged consensus in Europe, leading to a growing politicization of European integration (de Wilde & Zürn, 2012) and bringing economic explanations back when interpreting support for the EU. The Union represents a central pillar of globalization; therefore, it could function as a proxy when studying globalization’s impact. In this vein, analyzing attitudes towards European unification could indicate a general trend regarding viewpoints on globalization.
Critical distinctions between cosmopolitanism and communitarianism attitudes (de Wilde & Zürn, 2012; Teney et al., 2013), as well as integration and demarcation (Kriesi et al., 2008), attempt to capture decisive divisions between citizens who either
9
are in favor or oppose globalization. According to Teney et al. (2013), cosmopolitan and communitarian notions are translated into support for and opposition to the EU, respectively, while immigration and EU integration are intertwined. National and regional identities, social deprivation, and negative evaluation of globalization contribute to the development of communitarian approaches to the EU (Teney et al., 2013:2). Moreover, age patterns are prominent when analyzing support for globalization since being exposed to cultural heterogeneity results in the development of pro-globalization and pro-immigration attitudes (Rekker, 2018). In the same vein, educational attainment plays a decisive role in cosmopolitan and integrational attitudes (Verhaegen & Hooghe, 2015).
3.2.Attitudes towards immigration as a determinant of party preferences
Abou-Chadi and Helbling (2017) argue that when focusing on the impact of immigration on voting behavior, one should take into account policy reforms. Issue voting relates to policymaking, while ideology constitutes a further parameter. Their argument is in line with van der Brug et al. (2000) findings of casting a ballot for anti-immigrant parties. Specifically, voters seem to take into account both ideological proximity with parties as well as party size (van der Brug et al. 2000). Moreover, previous findings indicate that attitudes towards foreigners relate to the citizens’ perceptions regarding the number of immigrants living in their country (Sides & Citrin, 2007:491).
According to relevant research, the 2015 refugee crisis has intensified anti-immigrant attitudes and decreased trust towards the EU, while concerns about immigration took shape at both national and European levels (Harteveld et al., 2017:13). On the contrary, Stockemer et al. (2019:18-19) argue that Europeans’ anti-immigration and anti-EU attitudes follow a similar trend since 2012. Focusing on the case of Greece, one could claim that a broad scope of factors led to the reinforcement of anti-immigrant and xenophobic attitudes. Greeks who were exposed to a large number of refugees expressed intense hostility and xenophobia. In other words, regardless of the lack of economic (see Mayda, 2006; Edo et al. 2019; Scheve & Slaughter 2001) or cultural factors (see Betts, 1988; Hainmueller & Hiscox, 2007) the
10
exposure to massive refugee influxes functioned as a threat for natives who have developed long-term anti-refugee and anti-Muslim attitudes (Hangartner et al., 2018:13).
3.3.Evaluations of the economy as a determinant of party preferences
Economic voting draws upon the idea that voters hold governments accountable for their performance in terms of the economy. Hence, they tend to deplore incumbents who fail to perform successfully. The vast majority of scholars agree that when focusing on the economy, sociotropic evaluations are more efficient than egotropic ones (Lewis-Beck & Stegmaier 2007; Evans & Andersen, 2006) while using retrospective measures seems plausible (see Hansford & Gomez, 2015). On the contrary, studying electoral behavior in light of the Great Recession, Lindvall (2014:761) argues that voters take into account both retrospective and prospective economic evaluations. While many scholars argue that evaluating incumbents’ performance is feasible even if voters lack specific knowledge regarding the economy (see Fiorina, 1981); others highlight the impact of inaccurate information on punishing parties in office despite their actual performance (see Holbrook & Garand, 1996). Finally, when studying economic voting, psychological biases (Healy & Malhotra, 2013:288) should also be taken into account.
Furthermore, when one attempts to study the effect of the economy in voting behavior, taking into account country-related characteristics is essential. In other words, governments that are not capable of implementing economic policies independently are less affected by economic voting since responsibility is obscure. Within this context, Fernández-Albertos (2006) claims that economic voting better applies in cases where the economy does not depend on external actors, while left-wing governments that traditionally oppose internationalization tend to be less affected by adverse economic performance.
In the same vein, Bengtsson (2004:762) has shown that economic voting is prominent in countries that voters attribute responsibility to the government, only a few alternative political actors exist, and party systems are somewhat stable. Nadeau et al. (2002) have focused on the role of responsibility showing that in order for retrospective
11
economic voting to affect party choice, voters should be able to attribute responsibility to the government. Moreover, economic voting should be studied in light of partisanship. As Evans and Andersen (2006:195) have mentioned, perceptions about political actors and party loyalty should be conditioned when studying economic voting. However, the impact of partisanship on evaluations about the state of the economy is weaker in countries tackling severe economic instability (Evans & Andersen, 2006).
Hernandez and Kriesi (2016:204) argue that typical forms of economic voting taking place in times of normalcy, do not fully apply during crises. The Euro-crisis has led to the emergence of new dynamics reshaping European political systems. Growing dissensus regarding the fundamentals of Western party systems paved the way for severe contestation of mainstream political actors, contributing to the gradual development of new conflicts and patterns of electoral behavior. In this line, Lewis-Beck and Nadeau (2012) test the effect of economic voting in PIGS countries. According to their findings, economic voting is present and even more prominent in countries that face economic disorders, while governments’ blame attribution constitutes a fundamental parameter when testing its effect (Lewis-Beck & Nadeau, 2012:475-476). On the contrary, when the electorate holds the EU accountable for the economic conditions of their country, the impact of economic voting becomes weaker (Lobo & Lewis-Beck, 2012:526-527).
12
4. Data
The paper draws upon an empirical analysis of survey data. Specifically, three waves of the European Election Studies corresponding to 2009, 2014, and 2019 have been used. Using survey data utilizes a well- established approach to studying electoral behavior based on individual attitudes and subjective evaluations. As such, the paper interprets political behavior in light of individual perceptions about a broad scope of issues. Since EES surveys take place before the European elections, they are considered to be plausible when studying political behavior at a national level. Previous empirical research has highlighted that using these surveys reduces the possibility of bias due to growing salience and polarization regarding specific policy issues in times of parliamentary elections (Van der Eijk et al., 2003:167).
Moreover, the EES surveys include a battery of questions regarding the probability of voting for each of the eligible political parties in the country into account. Instead of using actual party choice, the paper focuses on the propensity to vote questions. There are several reasons why these questions apply better when studying political behavior. Specifically, actual vote questions provide information only for the most-preferred party. Therefore, while one is capable of testing the effect of specific factors in voting for a specific party, it is impossible to measure the impact of individual attitudes on electoral behavior in general. On the contrary, by using the propensity to vote questions testing the effect of self-placement and individual perceptions in the probability to vote for all parties is both feasible and could provide satisfactory explanations. According to the scholarship, voting probabilities reflect actual party preferences sufficiently (Van der Eijk et al., 2003:168). Additionally, the propensity to vote questions could be used in order to build stacked data matrices (for more information about the benefits of this method compared to MNL logits see Van der Eijk et al., 2006). This holds true since the dependent variable includes information for all parties, and it is measured in an 11-point scale where 0 equals “no, not at all probable” and 10 equals “yes, very probable.”
In the same vein, empirical research has confirmed that the probability of voting questions capture current party preferences since voters’ responses are not expected to reflect future dispositions (Van der Eijk et al. 2006: 432-433). The exact phrasing of the question used as the dependent variable is the following: “We have a number of parties in each of which would like to get your vote. How probable is it that you will
13
ever vote for the following parties?” Within this context, van der Brug et al. (2000:98, footnote) argue that using the term ‘ever’ contributes to the depiction of current party preferences.
Furthermore, the paper uses a set of independent variables in order to test their effect on the outcome while controlling for sociodemographic characteristics and trust towards EU institutions. In detail, five questions are used measuring individual attitudes regarding ideology, the economy, globalization, and immigration. In order to test the effect of ideology and attitudes towards globalization in voting behavior, self-placement on the left-right, as well as on the European unification axis, are used. The impact of the remaining factors on party preferences is measured using questions concerning voters’ evaluations. Therefore, questions related to both retrospective and prospective evaluations of the national economy as well as viewpoints on immigration are used. Perceptions about the economy are measured in 5-point scales where one means that the economy has got a lot better during the last 12 months or it is expected to be a lot better in the following year (retrospective and prospective evaluations) and five means that it has got or it will be a lot worse.
Additionally, attitudes towards immigration are measured using two different questions, depending on the dataset into account. Hence, regarding the 2014 and 2019 waves, an 11- point scale about viewpoints concerning policies on immigration is used, where 0 equals to fully favoring and ten equals to fully opposing a restrictive policy, respectively. Finally, testing attitudes towards immigrants in the 2009 voter study takes place using a 5-point scale where one means that the respondent fully agrees and five that (s)he fully disagrees with decreasing immigration in the country. The transformation of the variables is analyzed extensively in the following section.
Moreover, age, gender, educational attainment, social class, and trust in EU institutions are used as control variables, while dummy variables regarding party type and party families as well as year are also introduced to the models. The sociodemographic control variables were also transformed into dummy ones based on the provided categorization. Social class is based on respondents’ subjective evaluations. In all models, respondents who are women have completed up to 15 years of education, belong to the working class, and are older than 55 years old are used as the reference category in order for the data transformation to take place.
14
Finally, in the third set of models, the paper focuses exclusively on economic voting. Against this background, both retrospective and prospective evaluations of the economy have been used in the models. At the same time, three independent variables concerning government’s and external actors’ responsibility as well as trust to the Greek parliament have also been included. Specifically, a new variable has been computed depicting respondents’ viewpoints on the responsibility of the EU, the IMF, and banks concerning the country’s economic conditions. All three questions are measured in 11-point scales where 0 and 10 equal “no responsibility” and “full responsibility,” respectively. The Cronbach’s Alpha for the three items is .735. Government’s responsibility and trust to the parliament are measured using the 11-point and 5-point scales, where 0 and 1 equal to “no responsibility” as well as “yes, totally” and 10 and 5 equal to “full responsibility” as well as “no, not at all,” respectively.
15
5. Research design
The research design of the paper derives from the idea that party preferences are based on relationships between voters and parties in light of ideological or policy issue parameters. The principal concept regarding this approach is that party choice does not relate to individual attitudes per se. Rather, it derives from the proximity between parties and voters. Within this context, the independent variables have been transformed in order to depict distances instead of voters’ characteristics or viewpoints. In order to transform the independent variables, two different procedures have been followed. In cases that data about both voters’ self-placement and parties’ placement are included in the datasets, the absolute difference between the former and the latter has been used. The EES voter studies include questions regarding parties’ positions concerning both ideology and attitudes towards European unification (11-point scales). This being said, only the first battery of questions has been used since a considerable part of the electorate failed to place mainly challenger parties in the European unification scale. Moreover, data regarding parties’ position on the unification scale were not provided in the 2009 dataset.
Therefore, in order to capture distances between voters and parties with regards to the remaining variables, an alternative process applies. In detail, each independent variable has been regressed with each dependent one (probability to vote for each party), saving the regressions’ predicted values. Thereafter, the predicted values were subtracted from the mean predicted value for each party. In so doing, the new variables depict the predicted deviation between voters’ and parties’ positions in the issues concerned. Finally, these new variables were used in order to build stacked data matrices. At the same time, the probability of voting questions is transformed into a new variable, including voters’ probabilities for all parties. It should be mentioned that questions related to voters’ sociodemographic characteristics are not explicable per se. Still, they have been introduced to the models in order to get the net effect of the remaining variables after controlling for sociodemographic characteristics.
According to empirical research, using stacked data matrices has been very satisfactory (see Van der Brug et al., 2000; van der Brug, 2004; van der Brug, 2010; Van der Eijk et al., 2003; Van der Eijk et al., 2006; de Vries & Tillman, 2011; De Angelis & Garzia, 2013). This method has specific advantages, while there are certain
16
limitations as well. Since the unit of analysis is respondent multiplied by the number of parties that (s)he indicates the probability to vote for and, as mentioned above, the dependent variable is the probability to vote for all parties; multicollinearity is to a large extent ruled out (Van der Eijk et al., 2006:442, footnote). Regarding the limitations of the method, it should be mentioned that since the process of transforming the independent variables is based on regressed predicted values, one could not introduce interaction effects in the models (van der Brug et al., 2000:99-100, footnote).
The empirical analysis takes place at different levels. Firstly, merging all three datasets into one, the paper provides an overview of voters’ attitudes and perceptions for the three periods into account. Secondly, focusing exclusively on 2009 and 2014, a similar analysis takes place. The 2019 dataset has been excluded from the second stage of the analysis in order to compare the explanatory power of the independent variables as well as the reliability of the models both before and after the electoral win of SYRIZA in 2015. Thirdly, focusing on 2014, three models have been used in order to study economic voting more extensively. As mentioned above, when studying economic voting, attributing responsibility plays a decisive role. Unfortunately, questions regarding responsibility attribution are only available in the 2014 wave of EES.
Finally, the last set of models test the effect of the independent variables used in the first and second stages in probability to vote for SYRIZA. SYRIZA is considered to be one of the most successful radical left-wing parties in Europe. At the same time, it is the only radical-challenger party that managed to take office and for which there are data available for all three years. Therefore, voting for SYRIZA constitutes a case study in order to capture the impact of the economy, immigration, and globalization on the party’s profound electoral reinforcement. Hence, two models are provided. The first model includes data for the whole period into account, while the second one tests the effect of the independent variables in casting a ballot for SYRIZA before its electoral win.
17 5.1.Robustness
The empirical analysis takes place using multiple regression analysis in order to test the effect of a variety of independent variables in probability to vote. As mentioned above, different models are introduced in order to test the effect of the independent variables in probabilities to vote for different types of parties. However, when running regressions’ diagnostics, heteroscedasticity and as well as regression outliers have been identified in all models. Ordinary least squares (OLS) are sensitive to heteroscedasticity and outliers. Thus, both coefficient estimates and standard errors are biased. Therefore, instead of using regular linear models, an alternative process has been followed. Specifically, robust regressions have been used5.
Using robust regression analysis is beneficial concerning data that include outliers since observations’ weights are readjusted. Hence, these regressions provide robust standard errors and consistent coefficient estimators. Moreover, the readjusted standard errors are larger, taking into account the presence of heteroscedasticity. In other words, the paper proceeds to the analysis using weighted least squares where the weights are allocated using the regression’s MM-estimator.
18
6. Results
6.1. Model-Set 1
As mentioned in the previous section, the data indicate heteroscedasticity, while outliers have been identified. Hence robust regressions have been used. The robust residual standard errors are 1.63, .99, and 2.25 for all parties, challenger, and mainstream parties, respectively. The data do not indicate multicollinearity (VIF scores are about 1). Additionally, the residual errors show a normal pattern, although the dependent variable is negatively skewed, while there is no sign of autocorrelation in the models (Durbin-Watson test scores are 1.94, 1.82, and 1.97 respectively).
Table 1 shows the results regarding the first set of models. With regards to the first model where all parties included the coefficients for all independent variables apart from the Radical Left dummy are statistically significant, while the adjusted R2 is .31. Therefore, the model fit is satisfactory, assuring the reliability of the results. The second model tests the effect of the explanatory variables in voting for challenger parties. In this model, prospective evaluations of the economy are not statistically significant. The same applies to the year dummy corresponding to 2019. It is worth mentioning that the adjusted R2 of this model is considerably lower, dropping at .12. Thus, the reliability of the model is reduced. Finally, in the model concerning voting for mainstream parties, the adjusted R2 is .33, indicating higher reliability of the results. European unification is the only statistically insignificant independent variable for this model.
As is shown in the table, left-right distance constitutes the principle dimension when studying voting probability for all three models while it is significant on a .001 confidence level. However, in the case of challenger parties, the coefficient estimate is far smaller compared to the one regarding mainstream parties peaking at -.1 and -.36, respectively (standardized coefficients). The impact of economic evaluations on the probability of voting for challenger parties is less critical compared to the one related to their mainstream counterpart, while only retrospective ones are statistically significant on a .05 confidence level. Viewpoints on immigration play out as a statistically significant parameter regarding casting a ballot for both mainstream and challenger parties; thought the slope is slightly bigger for the former parties. Unaspiringly, European unification does not constitute a significant parameter when
19
focusing on mainstream parties. Finally, in 2014 voting for a challenger party is .04 more probable on a .001 confidence level. At the same time, casting a ballot for a mainstream one peaks at -.22 and reduces to -.09 in 2014 and 2019 respectively on .001 confidence level.
Table 1. Multiple Regression Analysis of Probability to Vote, estimated for different groups of parties (2009, 2014 & 2019)
(All parties) (Challenger parties) (Mainstream parties) _______________________________________________________________ Intercept 3.82*** (.25) 1.03***(.26) 5.10***(.11) Age .07*** (.09) .02* (.09) .08*** (.1) Sex .06*** (.16) .02** (.10) .07*** (.23) Education .04*** (.09) .01 (.06) .04** (.14) Social class .06*** (.09) .02** (.06) .06*** (.10) EU Trust .06*** (.06) .01 (.05) .08*** (.06) Economy: Retrospective .13*** (.06) .01* (.02) .14*** (.05) Economy: Prospective .09*** (.06) -.00 (.04) .15*** (.05) Immigration .05*** (.07) .02* (.04) .04*** (.11) European Unification .02* (.10) .01* (.06) .01 (.14) Left-right distance -.31*** (.03) -.10*** (.03) -.36*** (.01) Radical Left -.00 (.07) Radical Right -.07*** (.06) 2014 -.06*** (.09) .04*** (.04) -.22*** (.09) 2019 -.05*** (.08) -.01 (.04) -.09*** (.12) Adjusted R2 .31 .12 .33 Number of Observations 14.903 7.923 6.980
Table shows the standardized coefficients (betas) and standard errors in parentheses.
Significance code: ‘***’ significant on a 0.001 confidence level, ‘**’ significant on a 0.01 confidence level, ‘*’ significant on a 0.05 confidence level.
20 6.2.Model-Set 2
As mentioned in the research design section of the paper in all three models, heteroscedasticity and outliers, have been detected. Therefore, robust regressions have been used. There is no sign of multicollinearity since the VIF scores are approximately 1. The robust standard errors for the models are 1.89, 1.47, and 2.29, respectively. Moreover, Durbin-Watson test scores are 1.92 for the first, 1.78 for the second, 1.93 for the third model, not indicating autocorrelation. As was the case for the first set of models, the distribution of errors is somewhat normal.
Table 2 demonstrates the results for voting probability for all, challenger, and mainstream parties in both 2009 and 2014. Again, left-right distance constitutes the most critical parameter regarding electoral behavior. Coefficient estimations for voting for the challenger and mainstream parties are -.26 and -.37, respectively, on a .001 confidence level. Retrospective evaluations about the economy do not play a significant role in casting a ballot for the challengers, while their prospective counterpart is statistically significant at a .1 confidence level. On the contrary, the effect of retrospective and prospective economic evaluations is bigger concerning the third model. The regression coefficients are .16 and .13 for retrospective and prospective evaluations, respectively, while they are significant on a .001 confidence level. The impact of immigration on party preferences is slightly bigger concerning challenger parties, while it is statistically significant in all three models. Finally, casting a ballot for radical right-wing parties is -.05 less probable than voting for any other party on a .001 confidence level, while the year coefficient indicates -.08 probability of voting in 2014 on a .001 confidence level.
The adjusted R2 for the following models is .30, .21, and .33 for all, challenger and mainstream parties, respectively. Excluding the 2019 data, the R2 of the second model increases considerably. Hence, the model becomes more reliable. In the same vein, European Unification continues to be a critical parameter when comparing voting for either challenger or mainstream parties since its coefficients are statistically significant on a .05 confidence level only on the former case.
21
Table 2. Multiple Regression Analysis of Probability to Vote, estimated for different groups of parties (2009 & 2014)
(All parties) (Challenger parties) (Mainstream parties) ___________________________________________________________ Intercept 4.29*** (.14) 2.64***(.33) 5.13***(.11) Age .08*** (.09) .06*** (.15) .09*** (.11) Sex .07*** (.18) .04*** (.25) .08*** (.24) Education .04*** (.10) .02 (.13) .04** (.14) Social class .05*** (.10) .04*** (.12) .03** (.14) EU Trust .04*** (.07) .00 (.12) .05*** (.07) Economy: Retrospective .11*** (.06) -.01 (.13) .16*** (.06) Economy: Prospective .09*** (.06) .02’ (.10) .13*** (.06) Immigration .04*** (.07) .04*** (.08) .02* (.14) European Unification .02** (.12) .03* (.15) .00 (.16) Left-right distance -.36*** (.02) -.26*** (.04) -.37*** (.02) Radical Left .01 (.08) Radical Right -.05*** (.07) Year (2014) -.08*** (.07) .03** (.07) -.22*** (.09) Adjusted R2 .30 .21 .33 Number of observations 11.226 5.724 5.502
Table shows coefficients and standard errors in parentheses.
Significance code: ‘***’ significant on a 0.001 confidence level, ‘**’ significant on a 0.01 confidence level, ‘*’ significant on a 0.05 confidence level, “’” significant on a 0.1 confidence level.
22 6.3.Model-Set 3
Model-Set 3 focuses exclusively on the 2014 data in order to test the effect of retrospective and prospective economic evaluations in party preferences. Since the data are heteroscedastic and there are outliers, robust regressions have been conducted. All three models do not indicate multicollinearity (VIF scores are around one except for the prospective economic evaluations variable, where it is approximately 2). The distribution of residuals is normal for all models, while the robust standard errors are 1.77, 1.82, and 1.71 for model 1, model 2, and model 3, respectively. Additionally, the Durbin- Watson test scores are 1.84 for the first model, 1.77 for the second one, and 1.51 for the third one. Therefore, the model concerning voting for mainstream parties indicates slight autocorrelation.
Table 3 illustrates the results regarding voting probabilities for three different categories of parties in light of economic voting. The adjusted R2 for the three models is .28, .31, and .29 indicating the reliability of the results. In order to measure the effect of retrospective and prospective economic evaluations in party preferences, more sufficiently, three independent variables have been introduced to the models concerning responsibility attribution and trust to the parliament. While the effect of trusting the Greek parliament is similar for all three models attributing responsibility to either the government or external actors follows a different pattern. Specifically, attributing responsibility to the government has a bigger effect on voting for mainstream parties than casting a ballot for their challenger counterparts. The regression coefficients are .18 and .04, respectively. On the contrary, attributing responsibility to external actors is statistically significant when voting for challenger parties on a .05 confidence level while the corresponding regression coefficient is .03.
As was the case with the previous sets of models, retrospective and prospective economic evaluations explain sufficiently voting for mainstream parties, while their explanatory power is reduced regarding challenger parties. At the same time, it is -.8 less probable to vote for the incumbents (New Democracy, PASOK, and DIMAR) on a .001 confidence level. Interestingly, the left-right dimension plays out differently compared to the previous set of models. Focusing on 2014 and using variables that are exclusively related to responsibility attribution and the economy gives a different
23
account of ideological voting. In other words, concerning the economy, the left-right dimension has a tremendous effect in voting for challenger parties compared to their mainstream counterparts (-.43 and -.26 on a .001 confidence level, respectively).
Table 3. Multiple Regression Analysis of Probability to Vote, estimated for different groups of parties (2014)
(All parties) (Challenger parties) (Mainstream parties) _______________________________________________________________ Intercept 3.74***(.20) 4.38***(.30) 2.93***(.19) Age .07***(.12) .09*** (.22) .07*** (.15) Sex .08***(.23) .08*** (.33) .08*** (.32) Education .03* (.14) .05** (.20) .02 (.20) Social class .03***(.12) .05*** (.15) .01 (.26) Trust: Parliament .06***(.08) .06 ** (.19) .06* (.10) Responsibility:
(EU, IMF, banks) -.02 (.19) .03* (.27) -.05* (26) Responsibility: Government .10***(.09) .04* (.14) .18*** (.15) Economy: Retrospective .05**(.10) -.02 (.23) .10** (.11) Economy: Prospective .07***(.09) .05* (.19) .10*** (.11) Left-right distance -.34***(.03) -.43***(.04) -.26*** (.03) Incumbent -.08***(.09) Adjusted R2 .28 .31 .29 Number of observations 6.491 3.317 3.174
Table shows standardized coefficients (betas) and standard errors in parentheses.
Significance code: ‘***’ significant on a 0.001 confidence level, ‘**’ significant on a 0.01 confidence level, ‘*’ significant on a 0.05 confidence level.
24 6.4. Model- Set 4
Model- Set 4 focuses on SYRIZA. While Model 1 provides estimations for all three years into account (2009, 2014, and 2019) Model 2 test the parameters’ effect on the probability of voting for SYRIZA before the party’s electoral success in 2015. As was the case with the previous sets of models, the data are heteroscedastic, while outliers have been identified. Therefore, robust regressions have been used. The VIF scores for the first and the second model are around one, indicating that the data are not multicollinear. The residuals’ distribution is normal, while the robust standard errors of the models are 2.97 and 2.96, respectively. Durbin- Watson test score is 1,94 for both models indicating that there is no autocorrelation in the data.
Table 4 demonstrates the regression coefficients of the probability of voting for SYRIZA. The adjusted R2 is .34 and .30, respectively, indicating a good fit of the models. Thus, one could consider the results reliable. The models provide interesting insights regarding voting for SYRIZA. Specifically, when taking into account all three datasets, retrospective evaluations about the economy have a far more considerable effect (the standardized coefficient is .25) while they are statistically significant on a .001 confidence level. On the contrary, the effect of respondents’ perceptions about the economy in casting a ballot for SYRIZA in 2009 and 2014 follows a different pattern. While retrospective evaluations do not play a role in voting for the party, their prospective counterparts have a smaller but still statistically significant effect.
Accordingly, attitudes towards immigrants follow the same pattern in both models. However, when testing the effect of immigration before the outbreak of the refugee crisis, the regression coefficient is halved. Interestingly, attitudes towards European unification are insignificant in the two models, while the effect of the left-right dimension is larger in the second model (-.44 and -.39, respectively). Finally, compared to 2009, the probability of voting for SYRIZA in 2014 and 2019 is .20 and .13 bigger on a .001 confidence level, respectively.
25
Table 4. Multiple Regression Analysis of Probability to Vote, estimated for SYRIZA
(Model 1) (Model 2) __________________________________________________________ Intercept 4.28*** (.15) 4.46*** (.16) Age .06*** (.19) .06* (.21) Sex .08*** (.27) .06* (.47) Education .04* (.15) .05* (.16) Social class .05** (.29) .05* (.45) EU Trust .00 (.31) .01 (1.39) Economy: Retrospective .25*** (.05) .02 (.21) Economy: Prospective .01 (.17) .06* (.21) Immigration .10*** (.10) .05* (.17) European Unification .01 (.30) .02 (.57) Left-right distance -.39*** (.02) -.44*** (.03) 2014 .20*** (.16) .21*** (.16) 2019 .13*** (.16) Adjusted R2 .34 .30 Number of Observations 2.403 1.666
Table shows the standardized coefficients (betas) and standard errors in parentheses.
Significance code: ‘***’ significant on a 0.001 confidence level, ‘**’ significant on a 0.01 confidence level, ‘*’ significant on a 0.05 confidence level.
26
7. Discussion
The model-sets presented in the previous sections indicate different trends regarding party preferences. When taking into account both economic and cultural factors as well as attitudes towards globalization, the left-right dimension plays out differently in the probabilities to vote for either challenger or mainstream parties. While left-right distance constitutes a significant parameter in voting for both types of parties, its impact is smaller regarding the challenger-radical ones. This holds true to both sets of models. However, the effect is smaller concerning the second set. Therefore, one could argue that before challengers taking office, voting for radical parties was driven by ideology to a larger extent than after their establishment. At any rate, the results given in the first two sets of models confirm that voting for challenger parties relates to ideology to a lesser extent than casting a ballot for their mainstream counterparts. One the contrary, focusing on economic voting provides different results. In the third set of models, the left-right dimension affects voting for challenger parties more drastically compared to the probability of voting for mainstream ones. Finally, when testing the effect of both economic and non-economic parameters in casting a ballot for SYRIZA ideology is prominent in both periods into account. Even though the effect of the left-right dimension slightly reduces when including 2019 data in the analysis, it is the most critical parameter in terms of voting for the party.
Therefore, the first hypothesis is partly confirmed. Indeed, casting a ballot for challenger parties could be protest voting when one takes into consideration non-economic factors as well. At the same time, this holds when studying the effect of ideology in voting for both radical left-wing and right-wing parties. In this vein, the results partly deviate from Katsanidou and Otje’s (2015) work. While the authors suggest that cultural factors relate to the left-right dimension, this is not the case when testing their effect in voting for challenger parties as a whole in comparison with casting a ballot for their mainstream counterparts. Instead, including attitudes towards immigration and globalization reduces the effect of ideology. This is in line with Hobolt’s and Tilley’s (2016) findings. Defectors attempt to punish mainstream parties, whereas they take specific policy issues into account when voting for radical ones. Therefore, casting a ballot for the challengers in light of immigration and globalization could, to some extent, be protest voting. Within this concept, the paper aligns with
27
Treib’s (2014) argument that Eurosceptic voting in the 2014 European election entails both protest and opposition to EU policies.
The most striking result of the analysis relates to attitudes towards globalization. In both sets of models, European unification plays out as a statistically significant parameter concerning the group of challenger parties. On the contrary, European unification does not have any effect on voting for mainstream parties. This is also the case when focusing on casting a ballot for SYRIZA. In other words, attitudes towards globalization and integration constitute a vital factor only when one classifies both radical left-wing and radical-right wing parties together. On the other hand, testing the effect of the European unification dimension in casting a ballot for SYRIZA follows a different trend. Therefore, the second hypothesis is also partly confirmed. Attitudes towards globalization do indeed play a role when focusing on challenger parties as a whole.
Voting for radical- challenger parties seems to entail opposition to integration and globalization. However, this does not hold valid in the case of SYRIZA. This is in line with previous findings highlighting different forms of opposition to globalization and integration. Specifically, while radical-right wing parties reject supranational institutions at a constitutional level, their left-wing counterparts oppose specific policies related to the economy and neo-liberalism leading to the adoption of different patterns related to opposition to the EU (Pirro et al., 2018; van Elsas et al., 2016), and therefore, to anti-globalization attitudes. As mentioned above, challenger parties have been identified as countervailing forces concerning supranational institutions (Gemenis & Dinas, 2010; Vasilopoulou, 2018). Interestingly, with regards to SYRIZA, the European unification dimension does not play any role. This could be the case since the party’s agenda highlights the negative impact of economic globalization while it has gradually readjusted its public discourse after taking office (Vasilopoulou, 2018).
Immigration affects voting for both challenger and mainstream parties in a similar manner. While in the first set of models, attitudes towards immigration tend to have a bigger impact on choosing a mainstream party, the opposite trend holds valid when focusing only in 2009 and 2014. However, the impact of immigration on party preferences is both small and almost similar regarding the two types of parties into account. This is also the case when focusing on voting for SYRIZA. There are two
28
reasons why this could be the case. Firstly, in the challenger parties’ models, both radical left-wing and radical right-wing parties are included. Secondly, the increasing number of immigrants entering Greece, in combination with the severe implications of the refugee crisis, has led to growing awareness regarding immigration. This is in line with Hangartner et al. (2018) findings that the abrupt exposure to refugees has reinforced xenophobia and hostility. At the same time, according to a report on party preferences in Greece, attitudes towards the number of immigrants do not differ among party identifiers (Costa Lobo et al., 2019). Therefore, the third hypothesis is rejected. Immigration constitutes an evenly important parameter regarding voting for both mainstream and radical parties.
Finally, testing the effect of the economy on voting behavior took place using two analyses. In the first two sets of models, both retrospective and prospective economic evaluations have significant explanatory power in interpreting voting for mainstream parties. On the other hand, casting a ballot for the challengers relates to either retrospective or prospective economic evaluations regarding the set into account. However, the difference with regards to regression coefficients is striking, indicating that perceptions about the economy can better explain voting for mainstream parties. On the other hand, voting for SYRIZA is strongly affected by retrospective economic evaluations when taking all three years into account. The opposite holds valid regarding respondents’ evaluations before the electoral win of SYRIZA. Specifically, prospective evaluations of the economy tend to explain better casting a ballot for SYRIZA before 2015.
Since the results are both controversial and obscure, a further set of models has been introduced. Responsibility attribution is considered to be a crucial factor when studying economic voting. Therefore, relevant variables have been introduced in the models. Testing the effect of retrospective and prospective economic evaluations took place in light of attributing responsibility to the government as well as to external actors (the EU, the IMF, and banks), while trust towards the Greek parliament has been taken into account. According to the results, only prospective economic evaluations can explain voting for the radicals. This is in line with Lindvall’s (2014) findings. However, the effect of the economy is smaller when studying voting for challenger parties compared to mainstream ones. Voting for mainstream and challenger parties derives from responsibility attribution to the government and external actors, respectively.
29
Hence, voters who are willing to cast a ballot for a radical-challenger party attribute responsibility to external actors and supranational institutions.
The results are in line with previous findings highlighting the impact of responsibility attribution on economic voting. Specifically, while economic voting is prominent in countries that have tackled severe economic instability and austerity (Lewis-Beck & Nadeau, 2012), attributing responsibility to the government is essential in order for it to take place. In other words, since casting a ballot for challenger parties relates to attributing responsibility to external actors, the impact of economic voting becomes weak. The findings confirm both Nadeau’s (2002) as well as Lobo’s and Lewis-Beck’s (2012) argument. In order for economic voting to be meaningful, voters should attribute responsibility to the government. When voters hold accountable other actors, economic voting is less prominent and obscure. Therefore, the fourth hypothesis should also be rejected. Negative economic evaluations do not necessarily lead to voting for challenger parties. Rather, since they are intertwined with responsibility attribution, they follow different trends depending on perceptions about the role of external actors and the government.
30
8. Limitations
The paper studies the effect of different sets of variables related to the economy, immigration, European unification, and ideology in party preferences. The analysis draws upon the idea that subjective evaluations and individual attitudes are linked to voters’ perceptions with regards to parties’ positions on the topics concerned. Therefore, the analysis is based on voters’ attitudes and viewpoints, while actual data regarding the state of the economy and immigration are not included in the analysis. Moreover, the paper uses data from the European Election Studies. In order for the validity of the results to be confirmed, different surveys should also be used. This holds valid since survey data are often prone to errors. Since the answers are based on participants’ interpretation and understanding, one could argue that they are highly subjective.
Furthermore, the paper attempts to study party preferences using three different EES surveys. However, not all waves include all necessary questions for the analysis. Therefore, in order to study the effect of economic voting in light of responsibility attribution, a further set of models using data from the 2014 wave was introduced. Since questions regarding national and supranational actors’ responsibility are neither included in the 2009 nor the 2019 dataset, the paper could not proceed to an extensive study of these factors in voting behavior throughout the whole period into account. Moreover, the paper has not taken into account the impact of collective identities in party preferences. The relevant scholarship has highlighted the effect of cosmopolitan and communitarian attitudes in viewpoints on globalization. Hence, studying the effect of collective identities in light of party preferences would have constituted an interesting dimension. However, for reasons related to the questions coding, collective identities have not been included as a further parameter in the analysis. Specifically, in the 2019 wave, there are no data regarding national and European identities. Additionally, the questions included in the remaining two datasets differ significantly in terms of both categorization and phrasing.
Previous research has identified endogeneity concerning ideology and party choice (Lewis-Beck & Costa Lobo, 2011). Although the vast majority of research on political behavior relies on the assumption that ideology determines party preferences, one could not rule out the possibility of reverse causation. In other words, voters’ self-placement
31
and parties’ placement on the left-right axis could be affected by party choice and vice versa. The same applies with regards to perceptions about the economy and ideology (Evans & Andersen, 2006). It could be the case that voters’ perceptions concerning the state of the economy are affected by party loyalty. Therefore, one could not rule out the possibility of reverse causation regarding economic evaluations as well. Moreover, when testing the effect of economic voting in party preferences, voters’ information and knowledge regarding the state of the economy could have a critical impact on their choice (Holbrook & Garand, 1996). However, since there are no questions related to knowledge in terms of the economy, this parameter has not been taken into account. All in all, the data used are limited to a short period. Thus, one should be cautious regarding causation and the validity of the results.
