Social Stigma and Support for the Populist Radical Right: An Experimental Study

The ‘taboo’ or ‘stigma’ associated with many populist radical right parties (PRRPs) has been argued to be an important constraint on their electoral success. In comparison to mainstream parties, there seems to be a higher barrier keeping voters from supporting PRRPs. However, this mechanism has not been tested directly. We conducted a randomized survey-embedded experiment manipulating the social stigma of a fictitious radical right party in Sweden. We compare three conditions. Two of these contain subtle signals about how other respondents feel about this party. In one condition the fictitious party is supported by many voters (the neutralizing condition) and in the other it is evaluated negatively by the overwhelming majority (the stigma condition). Both experimental groups do not differ significant from the control group in support for this fictitious party. However, the difference between the two experimental groups is borderline significant. This suggests that there is a causal effect of social stigma on support for a RRP, even though the evidence is rather tentative.


Introduction
Over the past decades, Western Europe has seen the surge of populist radical right parties (henceforth PRRPs). We define PRRPs as parties for whom nationalism and anti-immigration are the core political issues, next to their authoritarianism and populism (Mudde 2007). Anti-immigration attitudes are the strongest predictors of support for these parties (e.g., van der Brug et al. 2013), but this is also what makes these parties highly controversial. Partially because of the historical experiences with fascism and institutionalized racism, there is at least to some degree a taboo on supporting these parties. In the 1980s, mainstream parties in Flanders, the Netherlands and By tapping into this social aspect of the vote, our research is important to a broader literature in electoral behaviour. Scholars have noticed for decades that electoral decisions, as well as other political activities, are seldom made in social isolation (Festinger 1962). Rather, such decisions are influenced by the attitudes expressed in voters' social environments: "an opinion, a belief, an attitude is 'correct, ' 'valid,' and 'proper' to the extent that it is anchored in a group of people with similar beliefs, opinions, and attitudes" (Festinger 1962, 272-273). Zuckerman (2005, 4) rightly points out that, while it is "both obvious and well-known" that citizens' social circumstances affect their political thoughts and actions, "relatively few political scientists incorporate these principles into their analyses" . Our study suggests that voters are indeed influenced by the opinions of others.
We will first briefly discuss the literatures on the influence of social stigma and more generally the social aspects of electoral decisions. We then discuss our experimental design and treatments, after which we present the results. In a discussion section, we elaborate more generally upon the implications of our findings.

Theory
Our study builds on a long tradition of understanding voting as a socially informed act. In this view, voting is the result of a complex process which involves all elements of the "life space" of a voter, including her/his individual personality and experiences, intimate and close personal contacts ("primary groups"), as well as cues from the broader society (Zuckerman 2005, 9). The importance of the social logic, especially the role of intimate contacts, has been acknowledged at least since The American Voter (Campbell et al. 1960, 274).
The social context of voting can be expected to be especially relevant for populist radical right voting. In particular, we expect PRRPs' social stigma to constitute a very strong social signal. In line with Goffman (1963, 12), we define stigma as "an attribute that is deeply discrediting" (see also Harteveld et al. 2017). A stigma is thus socially defined: whether something is discrediting depends upon the reaction of others. From the point of view of a voter, a party experiences stigma if it is regarded as unacceptable in the social context in which this voter lives. Although, for a given party, the level of stigma might vary between different subgroups in society, we argue that stigma is often constructed at the level of the polity. Populist radical right parties are a well-documented example. Many of these parties are treated as political outcasts or 'lepers' (van der Brug et al. 2000;Minkenberg 2006;van Spanje & van der Brug 2007). In several cases they even face criminal prosecution, party bans, or cordon sanitaires (e.g., van Spanje 2018). We expect the presence of a social stigma to function as a social norm for many voters: this is 'a party one does simply not vote for' . This 'norm' is the direct consequence of another social norm, which is that one is expected not to discriminate people on the basis of their ethnicity or religion; a norm that PRRPs are often -implicitly or explicitly -suggested to breach (Ivarsflaten et al. 2010).
Social norms are both descriptive and injunctive (Chung & Rimal 2016). The descriptive element pertains to perceptions about what types of attitudes and behaviours most people find acceptable. The injunctive element pertains to the pressure that people experience to conform to these norms. Since voting is a secret act, the pressure to conform will be much less than in overt behaviour (Kuran 1997). However, even if there is no direct social pressure, the descriptive element of social norms could still be important, at least to the extent that people are sensitive to the views and preferences of their fellow citizens. This is also true for the expression of political preferences, of which the development is in varying degrees subject to social scrutiny. Voters continuously generate and update opinions about parties, and the fact that a party is stigmatized might prevent voters from developing favourable predispositions towards such a party. So, the social norm may then be internalized, even if the behaviour is private. After all, incompatible differences between a voters' political preferences and those found in his or her social context can constitute an incongruence that voters might want to avoid or resolve, according to theories of cognitive dissonance (Festinger 1962) or motivated information processing (Meffert et al. 2006;Taber & Lodge 2006). The hypothesis to be tested is: H1: a signal of high social stigma decreases voters' preference for a populist radical right party The social stigma is (inversely) related to, but qualitatively different from, the 'social consensus' information distributed through 'viability heuristics' and the 'bandwagon effect' (Lau & Redlawsk 2006, 234;Schmitt-Beck 2015), which lead voters to support parties that are doing well in the polls. After all, those parties that do not gain support in the voters' environment are not necessarily stigmatized.

Case and Design
To test our hypothesis, we designed a survey embedded experiment in which we experimentally manipulate the extent to which a party is stigmatizedor, more precisely, either stimulates or neutralizes respondents' pre-existing associations with social stigma. The empirical data was collected at three different time points and within two online respondent panels administered at the University of Gothenburg. The first data collection was made within the Citizen Panel wave nine fielded in April 2014. The Citizen Panel is carried out by the Laboratory of Opinion Research (LORE). The following second and third rounds of data collection were included in the Swedish National Elections Studies Campaign Panel during the elections for the European Parliament in June 2014 and the election for the National Parliament in September 2014. 1 The Citizen Panel is a combination of a probability-based and a self-recruited opt-in sample where the latter is stratified by age, gender and education. The Campaign Panel is entirely based on a self-recruited opt-in sample mainly collected from two major Swedish Newspapers, Aftonbladet and Göteborgs Posten, during the 2010 election. We combine the data from the three waves, resulting in 5,606 valid responses.
We test our hypotheses by means of a vignette study in which a cue of social stigma is experimentally manipulated. We present "vignettes" of three fictitious parties, each described by a number of policy proposals. The policy proposals reflect parties of different ideological complexions: moderate left, moderate right, and populist radical right. Respondents indicate whether they feel positive (green button), neutral (orange button) or negative (red button) towards each party by clicking thumb buttons (see questions). For some respondents, these thumbs are accompanied by a number, which is said to reflect other respondents' likes, neutrals and dislikes. The PRRP is presented in three conditions (for translations of the questions, see Table A1 in the appendix): To enhance the credibility of the experiment, the numbers shown to the respondent actually changed depending on the respondent's choice. So, for instance, if a respondent clicked on the green "thumbs up" button in the stigma condition, the number would change from 43 to 44. The respondents are thus presented with varying social signals. The neutralizing message condition, by stressing acceptance by other respondents, decreases as much as possible any pre-existing stigma respondents might associate with regard to PRRPs. The stigma condition, by hinting strong unacceptability, triggers the perception that the party is disliked by most other citizens. Subsequently, we measure respondents' propensities to vote for the parties, as well as other evaluations of each party.
We expect the presence of stigma to deter respondents compared to the control group as well as the neutralizing condition. This is our central expectation. Whether the neutralizing condition will be different from the control group is an open question, as this depends on the extent to which citizens are willing to correct their (likely) pre-existing perceptions of the social acceptance of a populist radical right-type party. By contrast, the stigma condition merely needs to confirm such pre-existing associations. Of course, this reasoning is based on the assumption that there is already a social stigma around the populist radical right in the Swedish context. Several scholars have argued that this is indeed the case, and even more so at the time of data collection. In the Swedish national election 2010 the Sweden Democrats, a PRRP, entered the parliament for the first time, with 5,7 percent of the popular vote. The established political parties reacted by adopting a strategy of cordon sanitaire (Kokkonen 2014). According to the Swedish National Election Study, which was conducted during the election of 2014, the Sweden Democrats were highly disliked (placed 10 on a 0-10 like-dislike scale) by 52 percent of the electorate, which should be compared to six percent on average for the established parties (Oscarsson & Holmberg 2016). 2 In order to further scrutinize this assumption, we conducted the same experiment among an additional group of 2,725 respondents, who evaluated the vignette of a green party instead of a PRRP. Social stigma was manipulated in the same way for this green party as for the PRRP. In the Swedish context, there is no (or very limited) prior social stigma associated with a green party. It is therefore unlikely that the stigma condition will deter support for a green party, but possibly the neutralizing condition will raise support. 3 In order for this experiment to work, the respondents need to be (1) convinced that the numbers truly reflect the other respondents' views, and they need to (2) actually care about what the other respondents -which they don't know personally -think. Regarding the former requirement, analysis of the open-question remarks at the end of the Swedish study suggests that most respondents believed the study; some actually suggested that showing the "likes" and "dislikes" of the other respondents was a bad idea of the researchers, because it might have influenced their answers. Regarding the second requirement, there is evidence that people feel an identification with other members of even arbitrary groups. Pierce et al. (2016) show that social cues deriving from co-participants of social science experiments are indeed taken seriously. On top of this, we would like to add that even if some respondents did not find the thumbs convincing or felt little identification with their co-participants, this would downplay the effects of the thumbs. Any affect we do find is therefore a conservative one.
The dependent variable is a propensity to vote (PTV), which is measured using the following question: "When thinking about an election in which these three parties would exists, how likely is it that you would vote for each © 2019 The Authors. Scandinavian Political Studies published by John Wiley & Sons Ltd on behalf of Nordic Political Science Association of them?" , with answer options ranging from 0 "not at all likely" to 10 "very likely" . This measure allows us to capture more variation in respondents' support for a party than merely asking whether they would vote for the party or not. Table 1 and Figure 1 show the propensity to vote (PTV) for the fictitious populist radical right and green party in the three conditions. These are predicted probabilities derived from an OLS model in which the stimulus dummies are interacted with the type of party (PRRP/Green), and which controls for wave dummies.

Results
The table and figure indeed show lower propensities to vote for the fictitious populist radical right party in the stigma condition than in the control group, which in turn is somewhat lower than the neutralizing group. This is the pattern that one would theoretically expect. Both experimental conditions do not differ significantly from the control group. The difference between the neutralizing and the stigma conditions, although substantively very small (b = −0.18; β = −0.03; Cohen's d = 0.06), is significant at the onetailed 5 percent level (p one-tailed = 0.04). This difference between the neutralizing and stigma conditions re-occurs independently (and of similar magnitude) in each of the three waves, although not significantly so (p onetailed between 0.07 and 0.25; Cohen's d between 0.05 and 0.07). Since this is the only significant difference between the groups and since the difference is only borderline significant, we cannot claim that our experiment provides conclusive evidence of a causal effect of social stigma, even though the results point in that direction. If we focus more on the two experimental groups, the stigma condition differs more from the control group than the neutralizing condition. This seemingly larger effect of the stigma condition suggests that it is easier to prime people with information that confirms their pre-existing perceptions of the world, than with information that disconfirms it. This is further confirmed by the fact that, in the additional condition, respondents did not respond to any cue regarding the Green party, as is visible in the lower half of Table 2. Because Green parties have no (or a very limited) social stigma,   the neutralizing condition did not present new information, while the stigma condition did not resonate with existing beliefs. The absence of an effect among the Green party furthermore shows that we were not merely cueing popularity, but a true stigma, which does have to resonate.

Conclusions
In many West European countries, there is a social stigma surrounding support for PRRPs. This stigma (like any social stigma) has two components: most people feel negatively about supporting a PRRP and most people know that most people think of it negatively. Our experiment intended to prime people to realize that most other people think negatively about support for such a party. The difference between the two experimental groups was borderline significant, but both experimental groups did not differ significantly from the control group. So our experiment only provides tentative evidence of an effect of social stigma. Such an effect of stigma would mean that people are influenced by the attitudes of others. One implication is that populist radical right parties are less successful than they would be if people judged them only on the basis of the substance of the policies. It is important that such an effect seems to exist even though voting is a secret act and the pressure to conform to others' attitudes is much weaker than in overt behaviour. Either voters do not fully trust the secrecy of their ballot, or they respond less rationally to norms than the literature suggests (cf. Kuran 1997).
We see this experiment as one piece of evidence among some recently published papers, demonstrating that voters are deterred from supporting a party when there is a social stigma around it (Harteveld et al. 2017). That study showed that many (but not all) PRRPs were tainted by a social stigma, but also that some Communist parties and even the scandal-ridden Italian Christian democrats were perceived extremely negatively by large segments of the population. The results of our experiment provide tentative support for the existence of a main effect of social stigma.
While the difference between two experimental conditions was found to be borderline significant, both did not differ significantly from the control group. To the extent that we observed an effect of our manipulation, the effect was very small. However, it is important to note that there are good reasons to believe that a stigma around PRRPs exists in Sweden. So, rather than creating a stigma, our experiments mainly primed citizens to be reminded of this stigma. That our subtle manipulations seem to exert an effect is therefore an important finding. While suggestive that the mechanism discussed in the theory exists, an important limitation of our design is that we do not know how the small effect of our subtle cue relates to the total effect of stigma. Subsequent research might aim to develop a more explicit cue that has more impact.
Next to developing more explicit stimuli, we propose several ways to continue this line of research. Most importantly, further research should establish among whom social cues have the strongest effect. First, it is likely that the effect of stigma depends on the strength and content of prior convictions regarding the object the social cue refers. Those who feel most capable of navigating political choices (i.e., those with high efficacy) are less likely to be deterred by social cues about such choices, whereas for low-efficacy voters, "support of others is likely to predict personal relevance and utility" (Messing & Westwood 2012, 1047. Second, ideology probably matters too. Citizens who strongly disagree with a party (in our case, left-wing citizens) are likely to be less affected by cues of either social stigma or its neutralization, as they will not consider a party anyway; at the same time, those who are ideologically closer to a party are more likely to be deterred by stigma, because they actually consider the party. Finally, women may be more deterred by stigma than men: a recent study finds that women are more hesitant than men to vote for strongly disliked parties (Harteveld et al. 2015). Indeed, in a preliminary analysis, reported in Table A3 of the Online Appendix, we found descriptive (but not significant) evidence for moderation by gender, but not ideology or efficacy. However, given the weakness of our cue and the resulting lack of statistical power to robustly establish interactions, it remains for future studies to test stronger stigma stimuli and their interactions with potential moderators. This is important: if some groups of citizens are insulated from supporting controversial parties because of their heightened sensitivity to stigma, this will strongly shape the composition of parties' electorates.
Research on the effect of social stigma is politically important as well. As long as there is a stigma around support for the populist radical right, the party will attract less support than it would if supporting it would be more socially acceptable (see also van der Brug et al. 2005). PRRPs are the first to recognize this themselves: think of Marine le Pen's self-proclaimed goal to de-demonize the Front National's image, which recently even resulted in a name change. When PRRPs grow in size, as in Sweden and in Germany after the refugee crisis of 2015, this will have a spin-off effect as well. It signals to the general public that "there are many people out there" for whom the populist radical right is an acceptable option. This in turn signals that there is not that much of a stigma any more, which increases the group of voters who will be willing to consider such a party. Nothing succeeds like success. We control for the different rounds of data collection with dummies in our models. We also show our results separately for the different rounds in the online Appendix in Figure A1.

2.
This is one of the reasons we use fictitious instead of real parties in our experiment. The fact that the Sweden Democrats were so heavily disliked at the time of the experiment would make it difficult to manipulate the stigma surrounding the party in a convincing manner. 3. Table A2 in the online Appendix shows that the three experimental groups are very equally composed in terms of socio-demographic and ideological factors, which shows that randomization was successful.

Supporting Information
Additional supporting information may be found in the online version of this article at the publisher's web site: Table A1: Translation of conditions Table A2: Comparison of covariates across experimental groups Table A3: Interactions Figure A1: Analysis of PRR party support in each experimental condition (separate waves) with 90% confidence intervals