Introduction
Theoretical and historical accounts regard mainstream politicians as crucial for democratic stability (Levitsky and Ziblatt Reference Levitsky and Ziblatt2019; Ziblatt Reference Ziblatt2017; Linz Reference Linz, Hermet, Rose and Rouquié1978). Whether democratic values, norms, and practices persist, they argue, depends on the extent to which mainstream politicians yield to the rhetoric and policies of their extremist counterparts.
Most recent empirical research, however, has tried to explain norm erosion by looking into radical politicians alone. For example, Bursztyn et al. (Reference Bursztyn, Egorov and Fiorin2020) show that when individuals realize that others around them support Trump, they become more likely to openly express racist attitudes. Similarly, supporters of radical-right parties are more likely to publicly express their political preferences after these parties enter political institutions (Valentim Reference Valentim2021).
There is thus a disconnect between these two strands of literature. Both try to explain the same phenomenon, but they do so by emphasizing actors with different ideologies. Our goal is to bridge this divide.
Unlike previous research, we investigate the role of mainstream politicians in norm erosion and compare it to the norm erosion that is generated by radical-right politicians. Our focus is on norms regarding a crucial democratic value: the protection of minorities (Mendelberg Reference Mendelberg2001) and the related anti-prejudice norm (Blinder et al. Reference Blinder, Ford and Ivarsflaten2013; Alvarez-Benjumea and Winter Reference Alvarez-Benjumea and Winter2020). We aim to check how this norm is affected by anti-immigrant rhetoric – one of the ‘golden issues’ of radical-right parties (Mudde Reference Mudde2010).
There are at least two reasons why mainstream politicians may feel impelled to use this type of rhetoric. First, mainstream politicians may introduce it themselves to try and prevent the rise of the radical right. Second, mainstream politicians may deploy issues that the far right has already put on the agenda and adopt their policy stances (Meguid Reference Meguid2008; Green-Pedersen and Mortensen Reference Green-Pedersen and Mortensen2010; Meyer and Rosenberger Reference Meyer and Rosenberger2015; Gessler and Hunger Reference Gessler and Hunger2022; Rooduijn et al. Reference Rooduijn, De Lange and Van Der Brug2014; Abou-Chadi and Krause Reference Abou-Chadi and Krause2020).
While these dynamics have been widely acknowledged, to our knowledge no work thus far has investigated their implications for norm perceptions. What happens to anti-prejudice norms when mainstream politicians themselves make anti-immigrant statements? How does the effect of those statements compare to similar statements made by radical-right politicians?
Answering these questions empirically is difficult because it requires manipulating the source of a message without altering its content. One possibility would be to present experimental subjects with identical fictional statements, varying their partisan origins. This strategy, however, suffers from ethical concerns, since we would be providing respondents with untrue information. Moreover, as any treatment based on hypothetical quotes, it suffers from questionable internal validity, given that it would be unclear to what extent individuals would believe the quotes they were reading. To tackle these concerns, we design a survey experiment that leverages near-identical anti-immigrant statements made by real, radical- and centre-right politicians in Germany. Our results show that exposure to anti-immigrant cues from mainstream-right politicians leads to a more across-the-board perception of norm erosion than when similar statements are made by radical-right politicians.
Our paper makes two main contributions to previous literature. Public opinion research pays pride of place to the role of elites in shaping the public’s political attitudes (Zaller Reference Zaller1992; Lenz Reference Lenz2012; Popkin Reference Popkin1991). Citizens rely on cues from political actors to translate their general value orientations into support for specific policies (Druckman et al. Reference Druckman, Peterson and Slothuus2013). Our first contribution is to extend this work by shifting the focus from attitudes to norms. In highlighting how politicians can affect perceptions of what views are deemed acceptable, our work hopes to offer a more comprehensive view of how elite competition can impact democratic quality.
By the same token, an important question among scholars of party competition is whether accommodation—that is, the adoption of an exclusionary agenda on migration as a way of halting the electoral success of the radical right—pays off for mainstream parties (Abou-Chadi et al. Reference Abou-Chadi, Cohen and Wagner2022; Hjorth and Larsen Reference Hjorth and Larsen2022; Spoon and Klüver Reference Spoon and Klüver2020; Meguid Reference Meguid2008). While this discussion is crucial, it has thus far mostly focused on electoral outcomes – whether accommodation can help mainstream parties increase their vote share or reduce that of their radical challengers. A second contribution of our study is that it shifts focus to a different implication of this type of strategy: how it impacts established norms.
Research Design
We test the relative effect of statements by radical-right and mainstream politicians using a survey experiment fielded in Germany. We chose this country for two reasons. First, it is often identified as a context with strong and widespread norms against prejudice (Blinder et al. Reference Blinder, Ford and Ivarsflaten2013). Second, studying Germany allows us to take advantage of the rising popularity of the radical right party (the AfD), as well as the news – prior to our fieldwork – of the USA withdrawal from Afghanistan and fears of a second migration crisis. These events led some centre-right politicians (from the CDU) to adopt anti-immigrant rhetoric ahead of the 2021 German federal elections. This provided a realistic context for the anti-immigration quotes from CDU politicians that we presented to respondents.
To build our treatments, we identified anti-immigration quotes from CDU and AfD politicians that are roughly identical in content and intensity. These quotes were first circulated to five independent coders to see if they were able to distinguish them from each other or if the quotes could be attributed to the original speaker. We pre-selected quotes that passed this first test – two from AfD and two from CDU. Figure A.2 in the Online Appendix provides evidence of how coders coded those quotes. Next, we subjected the selected quotes to a more rigorous test. We presented them to 200 independent coders and asked them to: a) guess which party these quotes originated from; and b) rate how positive or negative they were towards immigrants on a scale from 1 to 10. As shown in Table A.3, quotes by the AfD were no more likely to be perceived as coming from the AfD than the CDU. Additionally, the two sets of quotes are statistically indistinguishable in terms of how negatively towards immigrants they were rated. Finally, after having identified the anti-immigration quotes, we also identified one quote from a CDU politician that signals a pro-immigration stance.
Participants were randomly split into one of five experimental conditions – which, for simplicity, we term Radical Right Party (RRP) approve [of anti-immigrant rhetoric]; Mainstream approve; Mainstream approve + RRP approve; Mainstream disapprove + RRP approve; and Control.Footnote 1 Participants in each of these conditions received identical surveys, but they were exposed to different treatment vignettes. In the first two conditions (RRP approve, Mainstream approve), they were exposed to vignettes which consist of anti-immigrant quotes. These quotes were attributed to an unnamed German radical-right or mainstream-right politician, respectively. In the third condition (Mainstream approve + RRP approve), participants were exposed to a similar vignette where anti-immigrant quotes are attributed to both an unnamed German radical-right and mainstream-right politician. The fourth vignette (Mainstream disapprove + RRP approve) contains the pro-immigration quote from a German mainstream-right politician, along with anti-immigration quotes from a radical-right politician. The fifth treatment condition (Control) is a similar vignette on an unrelated issue, food waste reductions. Table 1 summarizes the treatment arms, while Table B.1 in the Online Appendix shows each vignette in detail.
Treatment Conditions, where ‘approving’ refers to an anti-immigrant statement, while disapproving refers to a pro-immigrant statement
Note: (+) indicates approval; (-) indicates disapproval.
Outcomes
We focus on a total of six outcomes, which are of two types. The first group of outcomes measures respondents’ predisposition to express an opinion that is markedly contrary to the anti-prejudice norm. First, we presented respondents with a hypothetical petition to reduce immigration from Afghanistan and asked them if they would be willing to sign it. We label this outcome ‘Sign petition’. Second, we ask the degree of agreement with the following sentence: ‘The enrolment of pupils from Syria and Afghanistan has lowered the quality of German schools.’ In the plots that follow, we label this outcome ‘Agree Sensitive Item’. Both questions can be seen as tapping into ‘declared’ preferences, which are in turn a function of both sincere preferences and perceptions of acceptability. Since we do not measure the former, it is difficult to isolate them from the latter. Furthermore, an observed effect on this measure may merely reflect policy cue-taking in response to the treatment rather than genuine norm erosion.
The second group of outcomes builds explicitly on Bicchieri’s definition of social norms, seen as the combination of empirical expectations (what one thinks others do). Normative expectations (what one thinks others think one should do), and the possibility of sanctions. We started by asking respondents whether they thought it was appropriate to sign that petition (labelled ‘Perceived Appropriateness’). Then, to measure ‘empirical expectations’ (which is how we label our fourth outcome), we asked them to guess the percentage of other respondents in the survey who were willing to sign the petition. To tap into ‘normative expectations’ (the label of our fifth outcome), we asked respondents to guess the average percentage given by the other German respondents in the survey in the above-mentioned question. Finally, we employed a measure to gauge the role of sanctions. Participants were presented with a post on Twitter (currently ‘X’) containing anti-immigrant slurs from an anonymous user (shown in Figure B.2 in the Online Appendix). They were informed that these comments had been flagged as potential violations of Twitter’s hate speech policy and provided the option of reporting similar comments. To avoid deception, the post was adapted from an actual message that was removed by Twitter administrators for violating their hate speech policy. We label this outcome ‘Sanctioning’.
As specified in the pre-analysis plan, our main expectation was that our treatments would increase the personal willingness to violate the anti-prejudice norm, decrease such perceptions about the behaviour and normative beliefs of others, and decrease sanctioning. The only exception is the MRP disapprove + RRP approve treatment, which we expected would have the opposite effect.
As also specified in the pre-analysis plan, we not only compare each treatment arm against the control condition, but we also conduct pairwise comparisons. Concretely, we compare the MRP and the RRP treatment arms against each other. Since the messages are, on average, perceived as identical, this comparison nets out any effects of the message in isolation. However, this comparison bundles all possible differences between the AfD and CDU. In the discussion of our results, we unpack these differences to shed light on the mechanisms driving the effects.Footnote 2 This means we can speak more about the relative norm erosion effects of anti-immigrant statements than their absolute effects. That said, since parties in the real world do take positions on these issues, a direct comparison of the two sources of such messages is probably a more realistic way of assessing their implications.
After exclusions, the final sample contains 2009 respondents. Respondents were excluded if they failed to pass a pre-treatment attention check or if they did not complete the survey in its entirety. We have examined the breakdown of responses that were excluded and found no evidence of differential attrition according to treatment status (see Appendix D). The overall attrition rate is 9 per cent. The sample is nationally representative by age and gender, excluding individuals below the current voting age of 18, as well as those born outside of Germany. The survey was administered only by the German survey company infas, and fielded between January 20th and February 1st, 2022.
Results
We start by looking at our average treatment effects. Figure 1 shows the difference in outcome variables depending on a respondent’s treatment arm. As per pre-registration, the coefficients report results from linear regression models that compare the values of each of our six outcome variables in each treatment arm to those found in the control group. Positive coefficients mean that a given outcome has a higher value in the treatment group than in the control group; negative values mean the opposite. To make comparisons easier, all outcomes are standardized.
Exposure to anti-immigrant statements by mainstream or radical-right politicians and perceptions of norms against xenophobia.
Note: Thin lines represent 95 per cent confidence intervals; thick lines represent 90 per cent confidence intervals.
Contrary to expectation, the Figure suggests that exposure to anti-immigration cues from radical-right politicians (circle-shaped coefficients) does not erode widespread perception of the norm against prejudice. If anything, it generates a backlash effect that strengthens norm perception. Individuals in the RRP approve condition are less willing to sign the petition, less likely to agree with the sensitive item, and view signing the petition as less normatively acceptable.
The conclusion is different when it is mainstream politicians who make the anti-immigrant statement. As shown by the cross-shaped coefficients, in that case, individuals view signing the petition as more appropriate, think that a higher number of others will sign the petition (empirical expectations), and that a higher number of others will deem it acceptable to sign (normative expectations). Importantly, these effects hold only when MRP is alone in endorsing the norm-breaching statement, and evaporate when the RRP joins in approving it. When the two types of actors contradict each other – RRP approving while MRP not – their effects seem to cancel out, tilting somewhat in the RRP direction.Footnote 3
One could argue that one reason for these differences is partisanship. If all that respondents react to is whether the quote comes from their party or not, it could be that the mainstream right simply enjoys a higher pool of partisans. To see if this is the case, we split the sample into two groups, namely left- versus right-wing respondents. We are particularly interested in the first set of respondents, for whom none of the quotes come from their own party. Note that these are exploratory analyses that we had specified in the pre-analyses plan, but regarding which we had no particular expectation; our pre-analysis plan specified hypotheses for the full sample only.
The Figure depicts two facets, each representing the results using a different subsample: centre-right respondents (those who support FPD, CDU/CSU or the AfD) and left-wing respondents (those who support SPD, the Greens, or Die Linke). These analyses provide important insights into the reason why we find different effects across these two treatments. The answer does not seem to lie only in the match between individual and party partisanship. To be sure, we do find that the MRP-approve statements appear to move respondents’ empirical and normative expectations away from the anti-prejudice norm. However, the effects among left-wing respondents – whose views are not represented by any of the parties in our treatment – are even more telling. Exposed to an anti-immigrant statement by a radical-right politician (circle-shaped coefficients), this group becomes increasingly supportive of the established norm and perceives it as stronger. However, this backlash is not found when the statement is made by a mainstream-right politician (cross-shaped coefficients). In response to this treatment, left-wing individuals react in the opposite direction – even if the coefficients fail to reach statistical significance.Footnote 4
That left-wing individuals express backlash against anti-immigrant statements made by radical-right but not mainstream-right politicians may be driven by a pro-social backlash mechanism (Pulejo Reference Pulejo2023). According to this reasoning, left-wing individuals would move in the opposite direction to that of the statement to further distance themselves from the view expressed in the statement that they read. But why is that not the case when it comes to mainstream right politicians? One could argue that, to left-wing individuals, both sources represent a political outgroup, which could make them respond equally. In other words, if partisanship alone dictated these responses, there should not be much of a difference for left-wing individuals. Clearly, however, this is not the case. We cannot but speculate about the reasons underlying these responses. One possible explanation is that, at least in Germany, for left-wing individuals, the mainstream right is simply not as much of an outgroup as the radical right. There is a long history of government coalitions between the mainstream left and mainstream right in Germany, rendering agreements across camps quite common. Indeed, in line with this reasoning, previous research has found that a history of coalitions can reduce animosity among supporters of different parties (Gidron et al. Reference Gidron, Adams and Horne2022).
Another reason could be precisely the key, yet underrated, role of the mainstream right in upholding political norms. What we see here, one might argue, is a modern parallel to Ziblatt’s argument that the mainstream contributed to the first wave of democratization by incorporating institutions and actors that might otherwise resist regime change (Ziblatt Reference Ziblatt2017). In today’s context, the mainstream right seems to be defining the boundaries of what is socially acceptable. The radical right is perceived by many as being beyond the bounds of constitutional legitimacy, whereas the mainstream right is not.
Having looked into the individual effects of each treatment, we now turn to the key question of this paper, as described in the introduction: how does the anti-immigration statement coming from MPR compare to the one coming from RRP? The answer is provided in Figure 3, which compares the MRP approve against the RRP approve conditions. The Figure includes the same facets shown in Figure 2, plus one facet (on the left-hand side) including the whole sample, like in Figure 1. Point estimates denote the difference in each outcome between mainstream and radical-right politicians.
Anti-immigrant statements by radical-right politicians generate backlash on the left, but those by centre-right politicians do not.
Note: Thin lines represent 95 per cent confidence intervals; thick lines represent 90 per cent confidence intervals.
Comparing the norm erosion generated by radical-right and mainstream-right politicians.
Notes: Coefficients denote the difference between outcomes in the MRP approve and the RRP approve conditions. Positive values mean that the effect of the former is larger than that of the latter. Thin lines represent 95 per cent confidence intervals; thick lines represent 90 per cent confidence intervals.
The Figure highlights two main patterns. First, the left-most facet reinforces the point that there are some clear, across-the-board differences in the norm-erosion potential of mainstream and radical-right politicians. Often, our coefficients are positive and significant, suggesting that the anti-immigrant statements made by the former erode norms more than those made by the latter.Footnote 5 Second, this difference depends on the political preferences of the individual. As shown in the following facets, this difference is strongest for left-wing individuals, for whom MRP politicians are more likely to lead towards norm erosion than RRP across all outcomes. The only exception is sanctioning intentions, which are unmoved by any treatment in any subsample.Footnote 6
The Online Appendix provides some robustness checks and additional analyses. Figures E.1, E.2, and E.3 replicate Figures 1, 2, and 3, respectively, after adding controls.Footnote 7 The results remain substantively similar. Figures E.4 to E.12 show that there are no unexpected heterogeneous effects when the samples are split in terms of gender, income and education. Likewise, the findings are generally not sensitive to the removal of each control (Figures E.13 to E.16) or the removal of respondents from any particular state, with the exception of Nordrhein-Westfalen (Figures E.17 to E.32).
Discussion and Conclusion
Our results highlight how mainstream-right politicians play a crucial role in the enforcement of democracy and its associated norms (Ziblatt Reference Ziblatt2017). Their statements are more likely to erode mass-level norms, like those against prejudice, than are statements by radical-right politicians. Mainstream politicians may be tempted to give in to radical-right discourse to mitigate their success, but such accommodative strategies may have detrimental effects for democracy and its associated values.
There are two theoretical reasons that may explain these findings. First, these politicians have wider electoral support. For norms to change in society at large, individuals need to perceive that a tipping point has been reached (Andreoni et al. Reference Andreoni, Nikiforakis and Siegenthaler2021) and that a sufficiently large number of individuals in their reference network oppose the norm in place (Bicchieri Reference Bicchieri2016). We extend this reasoning by showing that larger parties (mainstream parties) may also be more capable of eroding established norms.
A second reason may simply be that, as shown by previous work (for example, Asch Reference Asch1956), norm conformity increases with the social status of the actor issuing a message. A similar mechanism may operate at the party level. Mainstream parties occupy a more prestigious position as guardians of democratic norms because they often cite themselves as sharing values and norms of the establishment (Valentim and Widmann Reference Valentim and Widmann2021) and enjoy more long-term ties with the electorate (Converse and Pierce Reference Converse and Pierce1992). Consequently, they are also better suited to change these norms than parties entering the system as pariahs (Van Spanje and Van Der Brug Reference Van Spanje and Van Der Brug2007). Empirically testing which of these mechanisms is driving our effects is beyond the scope of our study, and a question that subsequent research may try to answer.
Beyond this main conclusion, our results also suggest that the sender of a political message may be more important in determining its effect than the actual content of that message. The results of our experiment suggest that a similar statement can have very different effects depending on the actor who makes it, and how that actor is perceived by voters. In so doing, our results are in line with results in recent research on social media, which has shown that, in that realm, the identity of the sender significantly affects the way other users react to it (Taylor et al. Reference Taylor, Muchnik, Kumar and Aral2022).
One question left open by our findings is what happens when politicians of the left, instead of the right, take on rhetoric associated with the radical right. Our results seem to suggest that a similar effect could be found in those circumstances, since the main driver of the difference that we find between statements by mainstream- and radical-right politicians is precisely the reactions of left-wing individuals. However, an authoritative answer to this question can only be achieved via replication. The current German context, where the Bündnis Sahra Wagenknecht (BSW) has made a breakthrough with an economically left, yet anti-immigration, platform, might provide a useful setting for future research to study this question.
Supplementary Material
For supplementary material accompanying this paper visit https://doi.org/10.1017/S0007123425100586
Data Availability Statement
Replication data for this article can be found in Harvard Dataverse at: https://doi.org/10.7910/DVN/9NUMP8
Financial Support
This research received no specific grant from any funding agency, commercial or not-for-profit sectors.
Competing Interests
None.
Open access
