Electoral support for left wing populist parties in Europe: addressing the globalization cleavage

ABSTRACT

Drawing on the European Electoral Study (EES) for the 2014 European Parliamentary elections, we analyse the electoral competition between left wing populist parties and their non-populist counterparts in the nine European countries where this type of competition has occurred. By using the EES, we hold constant the electoral level and the electoral system, the timing of the elections and the question wording. To our knowledge, this is one of the broadest comparative analyses of European left-wing populist parties so far. The focus on the competition between leftist parties draws attention to the key question of the vulnerabilities of the non-populist left vis à vis their populist competitors. We test a battery of bivariate clustered logistic models and find that the losers of globalization theories help account for left wing populist parties’ support. However, the sociodemographic profile of these supporters does not fit the mainstream view on who these losers are. Moreover, we obtain strong support for the globalization cleavage theories: citizens who hold critical attitudes towards the EU and who perceive immigration as a threat to their ‘way of life’ are more prone to support left wing populist parties than non-populist left parties. These findings apply both to the entire sample of countries and to each one of them individually.

KEYWORDS:

• Populism
• left wing populist parties
• left
• new parties
• radical parties
• socialist parties
• European elections
• vote
• electoral behaviour

Acknowledgements

We wish to express our gratitude to Susana Aguilar (UCM) and to the anonimous reviewers for their insightful comments on the first version of this manuscript.

ORCID

Andrés Santana http://orcid.org/0000-0003-2594-1360

José Rama http://orcid.org/0000-0002-3990-1696

Notes

1 Information consulted the 6 February 2018.

2 It is important to clarify that we do not take sides on the debate on whether LWPPs qualify as a ‘party family’, nor do we sustain that they conform a European Party Group (although sometimes a leader of a given LWPP has participated in the campaigns of some other LWPP). Rather, we use the LWPP label to refer to a set of parties that share both a leftist and an anti-elite leaning. They emphasize socio-economic issues (March, 2007, p. 74) and claim that political elites only look after business’ interests while neglecting the common working man’s interests (Mudde, 2007). Although the best electoral results of LWPPs occurred in Southern Europe, they also obtained notable electoral support in Northern, Central and Eastern Europe.

3 In Figure 1A in the Appendix, we replicate our models for two alternative codings of the dependent variable. The first one differentiates between voting for LWPPs and voting for any other non-populist party. The second one focuses on the competition between LWPPs and mainstream center-left parties (PASOK, DIMAR, PSOE, UPyD, Labour Party, PD, PvdA, KKO, SDE, USL, and DS), leaving aside Greens, Communists, and Regionalists. Results are overall consistent for both replications.

4 Thus, some scholars have underlined the ‘chameleonic quality’ of populism (Taggart, 2000).

5 There exists a debate on the proper classification of M5S. In our main analyses, we include M5S as a LWPP, following Segatti and Capuzzi (2016, p. 54), who place M5S in the centre-left of the ideological spectrum (4.5 in a 0–10 scale), and Fella and Ruzza (2013, p. 49), who underline that most M5S voters identify themselves as on the left. Nevertheless, and as Rooduijn (2017) has pointed out, other works posit that M5S is neither a RWPP nor a LWPP. To tell appart the posibility that our findins rest on the inclusion of Italy, we replicate our statistical analysis excluding Italy. In fact, we replicate our analysis excluding one by one each of the countries (Figure 2), showing that the results are not an artifact due to the inclusion of any specific country.

6 This lack of fit may not be as straightforward as it could seem. The assumption that young, educated city dwellers are winners of the globalization may be less appropriate in Mediterranean countries.

7 Lipset and Rokkan (1967) had claimed that European party systems were a result of pre-existing, ‘frozen’ social divides (rural-urban, religious, centre-periphery and social class). Dalton, Flanagan, and Beck (1984) and Franklin, Mackie, and Valen (1992), argued that the initial social divides had lost relevance to explain party choice.

8 It is important to clarify that we do not attribute anti-immigration discourses to LWPPs. Our expectation that citizens with anti-immigration attitudes will be more prone to vote for them (instead of LWPs) stems from the relationship between protectionism and anti-immigration attitudes.

9 Online Access: http://europeanelectionstudies.net/european-election-studies/ees-2014-study.

10 Some parties continue to generate debates on whether they could be considered as LWPPs. Although March (2007) includes the German’s Die Linke as a a populist party, the study of Rooduijn and Pauwels, using both a classical and a computer-based content analysis, finds that ‘Die Linke in Germany is only slightly populist’, and recommends to classify it as a non-populist left party (2011, p. 1277). Its name notwithstanding, researchers agree to place Estonian’s Center Party (CPE, the acronym in Estonian being KESK) on the left side of the ideological spectrum: ‘From 2007 onwards, the KESK is the party most to the left’ (Jahn et al., 2018, p. 68). Likewise, several scholars coincide in considering SMER as both a populist and left party (Kriesi & Pappas, 2015, p. 217; Mesežnikov & Gyárfášová, 2008), though not as a radical left one (Stanley, 2017, p. 147). Finally, although some authors posit that KKE in Greece has assumed a populist discourse, for others, the KKE is a communist party with a clear Leninist-Marxist discourse (Pappas, 2014). This is also the case for the Czech KSCM party: whereas some authors consider it as a populist party despite its clear communist position, for others, it belongs to the traditional (1990s) Czech political formations (Havlík & Voda, 2018, p. 2). Although our main analyses employ the specification of the dependent variable shown in Table 1, two additional specifications have been tested: one that includes Die Linke as a LWPP, and another one that does the same for KKE and KSCM. Figure 2A in the Appendix shows that results remain qualitatively unchanged.

11 Each horizontal line in Figure 1 represents an independent variable of the model, the point standing for the best estimation of its effect upon the dependent variable, and the line, for its 5 per cent confidence interval. If a confidence interval crosses the vertical line drawn at the origin (zero) of the horizontal axis, the effect of the variable is not statistically significant. If it does not cross it and is located at its right, the effect is positive, whereas if it is located at its left, the effect is negative. We also tested a model that included the square of age to capture potential nonlinear effects of this variable (cfr. Evans, 2005) and with nominal versions of dichotomous variables. Results (available upon request) remained qualitatively the same (and Wald tests confirmed that dichotomous specification was preferrable). We also tested for possible problems of multicollinearity. This is not the case: the highest correlation is 0.45 between mistrust in the EU institutions and whether EU membership is regarded as a good thing; if attention is restricted to quantitative variables, the highest figure is 0.26 between the attitudes towards EU economic integration and EU unification; moreover, the mean VIF is 1.19, considerably lower than the levels regarded as problematic.

12 Clustering is employed to correct for the potential within-country correlation of residuals. This enables us to guarantee that the variance-covariance matrix used for the estimation of standard errors is the appropiate one. Given the low number of level-two (country) observations, other approaches, such as random-effects multilevel models, would not be advisable. The weighting variable is based on population and political factors: gender, age, urbanization, region, turnout and partisan vote distribution. More information available at http://europeanelectionstudies.net/wp-content/uploads/2015/07/ZA5161_release_notes.pdf.