What explains variation in the overall tone of the political debate? A large body of research has shown that the communicative style with which the political debate is carried has crucial implications for democracy. Negative rhetoric can increase affective polarization (Iyengar et al., 2012), affect voter turnout (Ansolabehere et al., 1994; Lau & Pomper, 2001), and make minorities withdraw from public life (Hobbs & Lajevardi, 2019). Despite these important consequences, we still know very little about what causes variation in the tone of the public debate.

We focus on how the success of radical-right politicians affects the political debate. Radical-right politicians are becoming increasingly successful across the world. Their success has been shown to shift the policy positions of the remaining candidates (Abou-Chadi & Krause, 2018), polarize the electorate (Bischof & Wagner, 2019), put new issues in the political agenda (Hobolt & De Vries, 2015), and normalize previously stigmatized behavior (Bursztyn et al., 2017; Valentim, 2021). One of the most distinctive characteristics of radical-right candidates is their rhetoric, which makes extensive use of negative emotions (Widmann, 2021).

How does this success affect the rhetoric of the remaining politicians? We answer this question with resource to a newly-collected dataset of over 100,000 speeches in German state parliaments. We analyze these speeches using an original dictionary that measures the use of positive and negative emotions, specifically created to capture emotional language in German political text. Importantly, it allows us to measure different discrete emotions, instead of just positive and negative sentiment. We report a set of validation tests suggesting that our dictionary outperforms off-the-shelf dictionaries.

Taking advantage of variation in the timing of elections that granted representation in state parliaments to politicians of the radical-right party Alternative for Germany (AfD), we employ fixed-effects models to estimate the effect of radical-right success on the use of emotions by the remaining politicians. Our findings support a distancing effect: the success of radical-right politicians does not make the remaining politicians use more negative emotions, but it makes them use more positive emotions. This effect is large—between 50% and 100% of the baseline, depending on the outcome.

A set of additional quantitative and qualitative analyses suggest that the increasing use of positive emotions may be part of a strategy by which the remaining politicians build a contrast between themselves and the radical-right, who they call out on the breaching of established norms regulating the political debate. We start with quantitative analyses that test for two observable implications of the distancing mechanism. Supporting this mechanism we find that (1) the effect is stronger on mainstream politicians (as opposed to populist ones), who have a stronger incentive to portray themselves as guardians of norms breached by the radical-right; and that (2) among mainstream politicians, the effect is stronger for those left of center, whose values are more threatened by radical-right success. We also find evidence that the increased use of positive emotions by established politicians effectively counterbalances the increased negativity that the radical-right brings to the political debate, keeping it from becoming more negative overall.

We complement these tests with qualitative analyses of speeches in our data. We find ample evidence of politicians calling out the norm-breaching behavior of the radical right as unacceptable, marking a clear contrast between themselves and the radical-right, and mobilizing voters on the basis of opposition to radical-right rhetoric.

These findings speak to three bodies of literature. In the first place, the paper relates to previous studies that have focused on how radical-right success affects the policy positions of the remaining candidates and parties (e.g. Abou-Chadi & Krause, 2018; Bale, 2003). We add to this literature by looking into the effect of radical-right success on another crucial dimension of the political debate: the tone in which issues are debated.

In the second place, a growing strand of research has investigated the strategic usage of emotional language by political actors (Crabtree et al., 2020; Kosmidis et al., 2019; Müller, 2020). This study contributes to this research field by focusing on an additional factor that explains variation in the use of such language: the success of radical-right candidates.

Finally, a growing literature has focused on how norms regulating the political space can erode (Alvarez-Benjumea & Winter, 2020; Bursztyn et al., 2017; Valentim, 2021). One central question in this literature has been the role of radical-right success in normalizing individual-level behavior associated with the radical right. Our paper adds to this literature by testing whether a similar effect can be found when drawing upon political elites, instead of citizens. Our findings suggest that, unlike citizens, politicians react to such success by holding closer to the norms breached by radical-right politicians.

Radical-Right Success and the Political Debate

We study how radical-right success affects the rhetorical style of the remaining politicians by looking into the case of Germany. The German party system was built upon a mixed member proportional system designed to keep a single dominant party from emerging (Lees, 2013, pp. 65–66). Based on a relatively low level of disproportionality, this system gave path to the emergence of a three-party system with centrist tendencies (Decker, 2014). Such system was composed of the Christian Democratic Party (CDU and its sister party in Bavaria, CSU), the main right-wing party; the Social Democrats (SPD), the main left-wing party; and the liberal party FDP. Following the rise to success of the Greens in the 1980s and, after the unification, the entry of the left populist party The Left (the successor of the ruling party in East Germany), the system ended up solidifying as a five-party system (Weldon & Nüsser, 2010).

Radical-right parties traditionally did not enjoy significant success, failing to cross the electoral threshold of 5 percentage points. This situation changed in 2017, upon the first-time success of the radical-right party Alternative for Germany (AfD). This party had been founded in 2013 as a primarily Eurosceptic party. In the election of September that year, it narrowly missed the electoral threshold in the federal election, thus remaining out of parliament. The party then underwent a shift to the right, placing it at the rightmost position in the German political spectrum (Franzmann, 2019). An analysis of its manifestos and online statements portrays the AfD clearly at the far right of the ideological spectrum (Arzheimer, 2015). Its voters also resemble, in many ways, the supporters of radical-right parties in other countries (Arzheimer & Berning, 2019; Goerres et al., 2018). In 2017, capitalizing on anxiety about migration and dissatisfaction with the German political establishment (Hansen & Olsen, 2019), the AfD managed to acquire an impressive vote share of 12.6, winning 94 out of 709 parliamentary seats and becoming the third most voted party.

To derive expectations as to the effect of the parliamentary success of AfD politicians on the political rhetoric of the remaining politicians, our study builds on the literature on the use of emotional appeals by politicians and the literature on how established politicians react to the success of radical-right contenders. Previous studies on the effects of emotional appeals by politicians have shown such appeals to have a number of important effects (Ansolabehere et al., 1994; Brader, 2005; Brader et al., 2008; Iyengar et al., 2012; Lau & Pomper, 2001; Rico et al., 2017; Valentino et al., 2008; Vasilopoulos et al., 2018, 2019; Valentino et al., 2011). Despite these important insights, this literature has mostly neglected that emotional appeals can also affect other politicians.

The success of radical-right politicians can be expected to significantly alter the use of emotions by established politicians. Previous research has shown that the remaining politicians tend to respond to their success by assuming more rightist policy positions (Abou-Chadi & Krause, 2018). Radical-right politicians such as those belonging to the AfD, however, do not just innovate in the choice of issues they politicize or their position on those issues. Their rhetoric is also distinctive in that it tends to make more use of negative language (see Crabtree et al., 2020 for comparative evidence; Widmann, 2021 for evidence focused specifically on Germany). In facing the electoral success of these radical-right contenders, the remaining politicians can not only move their policy positions closer to those of the radical-right, but also try to emulate their discursive style.

Furthermore, the success of radical-right politicians can also affect the rhetoric of the remaining politicians via a process of social norm change. The political debate is regulated by norms that dictate which sets of behaviors are deemed appropriate (Mayntz, 1992). In observing that mainstream politicians tend to use positive emotions, MPs inclined to use negative emotional appeals in their discourse might refrain from doing so because that is not the modal form of communication in the political debate. Radical-right politicians may change this situation because of their characteristic norm-breaching rhetoric, which previous research has found to normalize stigmatized behavior (Bursztyn et al., 2017; Valentim, 2021). Social norms regulating the political debate are, in many ways, similar to the ones regulating other realms of social life (Bernick & Wiggins, 1983), and hence one might expect radical-right success to erode them as well.

For these two reasons, one might expect radical-right discourse to make the remaining politicians use a more negative rhetoric. This leads us to formulate our first hypothesis:

Imitation hypothesis: The success of radical-right candidates will make other politicians use more negative emotions.

There are, however, reasons to expect the opposite effect as well. Established politicians may feel like radical-right contenders have ownership over the negative rhetoric they employ, making it a risky strategy to imitate it. A more profitable strategy may be for established politicians to reinvent themselves around the kind of rhetoric they have ownership on, such as their commitment to established values and norms (De Vries & Hobolt, 2020, pp. 250–251). One way by which they can do so is to use an increasingly positive style of debating. That would allow them to build a clear contrast with the negative debate of the radical-right, and might signal their commitment to established norms that the radical-right openly breaches.

This distancing strategy may be rational due to its potential effect in containing the electoral wins of the radical-right. Negative emotions have been shown to affect a number of individual-level covariates that can make an individual more likely to cast a vote for the populist right (Rico et al., 2017; Vasilopoulos et al., 2018, 2019). This provides the remaining politicians with an incentive to respond to radical-right success by increasing their use of positive emotions to try and contain the effect that the negative rhetoric of the radical-right may have in fueling further support for it.

Such incentives, however, should not be homogeneous across politicians of the whole ideological spectrum. In the first place, the effect should be stronger on MPs of mainstream parties than on those of populist parties. Populist politicians of both the left and the right tend to use negative emotions as a ways of mobilizing on discontentment with the establishment (Crabtree et al., 2020)—a finding that has been replicated in the specific case of Germany (Widmann, 2021). As a consequence, unlike with mainstream parties, the increase in negativity brought about by the success of radical-right politicians may actually prove profitable to left populist politicians. In the German case, this means that the effect should be weaker on MPs of the left-populist party The Left, as compared to the remaining politicians.

Moreover, among mainstream parties, the effect should depend on the ideology of the politician. Left-wing politicians are the ones whose values are most threatened by radical-right discourse. Indeed, previous research has shown that, when political actors are successful in pushing for changes in the status quo, it is actors at the other end of the ideological spectrum who react by moving even farther in the opposite direction (Bustikova, 2014). Furthermore, the voters who support these politicians are also likely to be the ones who more strongly react by condemning radical-right discourse. As Bischof and Wagner (2019) show, after an initial legitimization effect, radical-right success leads to a backlash effect brought about by left-wing individuals becoming more likely to enforce the norms that the radical-right breaches. Consequently, left-wing mainstream politicians have stronger intrinsic (their own values) and strategic (their voters’ reactions) motivations to distance themselves from the radical-right. This leads us to formulate our second hypothesis:

Distancing hypothesis: The success of radical-right candidates will make other politicians use more positive emotions. This effect should be stronger on politicians of mainstream parties and, among those, on politicians of the center-left.Footnote 1

Data and Empirical Strategy

As mentioned in the theory section, our empirical analyses draw upon the case of Germany. Doing so presents two main advantages. In the first place, it allows us to look into the case of a country that recently has witnessed the first-time success of the radical-right party AfD. In the second place, Germany is a federal state where each of the 16 states has its own parliament. The fact that elections for state parliaments are not synchronized allows us to estimate the effect of the parliamentary presence of radical-right politicians (coded as all politicians elected by the AfD) on the behavior of the remaining state MPs—a design that has been used in previous studies drawing upon different outcomes (De Vries & Hobolt, 2020, pp. 204–220). Parliamentary entry is a standard measure of electoral success that has been used in a number of previous studies (e.g. Abou-Chadi & Krause, 2018; Bischof & Wagner, 2019). Table 1 shows the dates of the state elections included in our sample, along with the vote and seat share of the AfD in each of them.

Table 1 Dates of German state elections included in the analyses
Full size table

The outcome variables in our empirical analyses are measured with resource to an original dataset composed of all speeches in the 16 German state parliaments between January 1, 2014 and June 1, 2019, collected from the official site of each state parliament. We include speeches both from MPs that were re-elected and from MPs who are in parliament only in part of the period covered by the analyses.Footnote 2 We removed speeches with length under 100 words, as these were mainly interjections and questions from the plenary audience rather than actual speeches.Footnote 3 We only included speeches from major parties that are represented at the federal level. In addition, we excluded speeches from members of the state government and focus solely on MPs. This leaves us with a sample of 101,844 speeches.

We rely on parliamentary speeches as a data source because they represent the way politicians actually communicate during the political debate. We believe this approach to represent a better way of measuring the overall tone of the political debate than alternatives such as the use of party manifestos. While a readily available source of data, manifestos present the shortcoming of being released only in election years (Sagarzazu & Klüver, 2017). They also include very low-salience topics, and have been shown to be inaccurate in measuring the ideological placement of parties (Dinas & Gemenis, 2010).

Legislative speeches, on the other hand, have the potential to reach large audiences through mass media. Previous research has shown that speeches are a crucial means by which MPs are granted visibility in the media (Tresch, 2009). In Germany concretely, one of the largest newspapers (Frankfurter Allgemeine Zeitung) covers each plenary session of the German parliament since 1950 with at least one article in average (Proksch & Slapin, 2012). TV programs also regularly pick up on (parts of) speeches in news shows (Salmond, 2014), which allows them to reach the larger television audience. Legislative speeches further have the advantage of providing individual-level data. While party manifestos, press releases, or official media statements only allow for the analysis of party-level variation, legislative speeches can reveal within-party differences in political debate (Hjorth, 2020; Proksch & Slapin, 2012). This allows us to build a panel-like dataset where we observe the emotional tone of the same individual legislator across time—and, crucially, before and after the parliamentary entry of the AfD. Finally, unlike party manifestos, which are written before the elections, parliamentary speeches are more adaptive to contextual changes—in this case, the parliamentary entry of radical-right MPs.

To measure how much these speeches make use of positive and negative emotions, we rely on a dictionary approach. We use a novel emotional dictionary (“ed8”) that has been specifically created to measure discrete emotional language, while taking into consideration German language peculiarities. It should be noted that this dictionary was designed having in mind the specific aim of analyzing political text. The ed8 dictionary is based on the augmented dictionary (Rauh, 2018), a German sentiment dictionary that reliably discriminates positive and negative tone in German political language. However, the augmented dictionary cannot differentiate between discrete emotions—which is important since different emotions (even of the same valence) can differently affect information processing, attitude formation, and the behavior of voters (Rico et al., 2017; Valentino et al., 2008, 2011; Vasilopoulos et al., 2019; Vasilopoulou & Wagner, 2017). We thus extended this dictionary with separate categories that attribute words to eight different emotions.Footnote 4

The emotions under scrutiny include five basic emotions (anger, fear, disgust, sadness, and joy) as defined by Ekman (1999). To these, three additional emotions (pride, enthusiasm, and hope) were added given their central role in political discourse (Brader & Marcus, 2013). The measurement of different discrete emotions is crucial given that they have diverging political consequences, often explained by underlying appraisal patterns. Appraisals are the dimensions by which individuals evaluate events eliciting emotional responses (Lazarus, 1991). While some emotions differ in terms of pleasantness (anger versus joy), others vary in the level of certainty or situational control that is associated with them (enthusiasm versus hope) (Lazarus, 1991; Smith & Ellsworth, 1985; Tiedens & Linton, 2001). Furthermore, emotions can also differ in their temporal perspective, focusing either on the presence/past (joy) or on the future (enthusiasm, hope). These differences in appraisal patterns are reflected in the selection of vocabulary included in the different emotional categories of the ed8 dictionary. For instance, while positive emotions all include positive terms, enthusiasm includes more future-oriented vocabulary. In turn, hope includes positive terms that are associated with more uncertainty, as compared to joy and enthusiasm.

To check whether this approach is accurate, we used a German crowd working platform called Crowdguru, which is similar to Amazon’s Mechanical Turk. Online Appendix B describes the crowd coding process, quality tests, and the comparison to other freely available dictionaries in detail. Overall, the results show that the novel ed8 dictionary applied in this study outperforms widely used off-the-shelf dictionaries in several measures across all comparable emotions. It shows robust F1 scores and substantially improves the accuracy of automated text classification compared to other emotional dictionaries.

Applying the ed8 to the data provides a score measuring how many words associated with each emotions are found in a given parliamentary speech. For example, a document containing four anger words and two words connected to joy, obtains a value of 4 for anger and 2 for joy. To create comparable scores independent of the length of a given document, normalized emotional scores are created, i.e. dividing the emotional scores by the word count of each document. These normalized scores were then standardized to make comparisons easier. These final standardized scores for each of the eight emotional categories constitute the outcome variables in our analyses. Table D1 in the Online Appendix shows how the emotions correlate in the speeches that compose our dataset. Figure D1 in the Online Appendix shows the distribution of these eight outcome variables.Footnote 5 In the case of both positive and negative emotions, we also include a summary index as an outcome, which we obtain from extracting the first principal component of the individual standardized emotion. In so doing, we hope to reduce noise and lower concerns with multiple hypotheses testing.Footnote 6

This data set was then grouped by thirty-day periods, centered around the date of the state election in their country. In this final data set, the unit of observation is the politician*month-to-election (\(n = 37,289\)). This means that our data set gives us information on the average level of emotions used by each MP in their speeches, within a given number of months before or after the state election.Footnote 7

Our empirical analyses rely on models with politician fixed-effects that estimate the effect of the parliamentary presence of radical-right MPs (coded as AfD politicians) on the remaining politicians.Footnote 8 A potential concern with one-way fixed-effects models is that there may be periodic fluctuation in the level of emotionality as a function of the distance to the election. Specifically, one might suspect that campaign periods have a level of emotionality that is different from the remaining periods, as research has shown that politicians change their rhetorical strategy in the weeks ahead of elections (Ceccobelli, 2018). For this reason, we rely on two-way fixed effects models including month-to-election fixed effects that control for common trends in the level of emotionality for each actor as we move closer and farther from the date of the election. The treatment effect is then calculated according to the following equation:

$$\begin{aligned} Y_{it} = \alpha + \gamma _i + \lambda _t + \tau D_{it} + u_{it} \end{aligned}$$

where \(Y_{it}\) is the level of the dependent variable in speech i at time period t; \(\gamma _i\) is a vector of politician-level fixed effects that control for unobserved heterogeneity across MPs; \(\lambda _t\) is a vector of month-to-election fixed effects; and \(\tau\) is the treatment effect.

It should be noted that this design makes two assumptions. In the first place, in controlling for the distance to the election, we assume that all time-variant confounders that may affect the relation we are interested in are a function of the distance to the election. We keep this model as the main one because it is the one we pre-registered. Still, in the Online Appendix (Fig. D2), we relax this assumption by replicating the analyses including controls for the calendar month in which each speech took place. In the second place, we assume that there are no spillover effects. In other words, we expect the success of radical-right politicians in one state not to affect the political rhetoric in other states. To check if this assumption is met, Fig. D12 presents a placebo test that replicates the analyses shown in Fig. 2 but using dates of elections in other states as the date of the treatment. We find no clear evidence of spillovers. The effects are small, mostly non-significant, and often contradictory across models.

Findings

Before we estimate the effect of the presence of radical-right politicians in state parliaments, we check whether there is indeed a descriptive difference in use of emotions between politicians of the radical-right and the remaining ones. Fig. 1 plots the mean score for each emotion in MP speeches of each of the main German parties. As the figure makes clear, radical-right politicians (those who belong to the party AfD) tend to use significantly more negative emotions in their rhetoric than do their counterparts from other parties. When we turn to positive emotions, the pattern is the opposite: radical-right MPs are among the ones using positive emotions less often. The only other party whose MPs use a similarly low amount of positive emotions is the left-populist The Left. This finding is in line with previous research, which has found that populist politicians use positive emotions less often than the mainstream ones (Widmann, 2021), and that the rhetoric of parties far from the ideological center also tends to be less positive (Crabtree et al., 2020).Footnote 9

Fig. 1
figure 1

Use of different emotions in parliamentary speeches, conditional on the party of the MP. Lines represent 95% confidence intervals

Full size image

We now turn to estimating the effect of radical-right success on the rhetoric of the remaining politicians. Figure 2 shows the effect of the parliamentary presence of radical-right MPs on the emotional appeals made by state MPs of other parties. It shows clear support for the distancing hypothesis, to the detriment of the imitation hypothesis. The parliamentary presence of radical-right MPs increases the use of positive emotions by MPs of the remaining parties; but has no effect on their use of negative emotions.

The effects on positive emotions are large in substantive terms—ranging between around .8 standard deviations in the case of pride and around 1.4 standard deviations in the case of the summary measure. When compared to the baseline, the magnitude of the unstandardized effects is also remarkable. Tables D2 and D3 show the unstandardized effects and the summary statistics of the outcome variables in the control group (observations in parliaments without radical-right politicians in parliament). The comparison of the two tables shows that our effects range between around 50% (in the case of pride) and around 100% of the baseline (in the case of hope). As we have noted, previous research has found that mainstream politicians tend to employ a more positive rhetoric than politicians from challenger parties (Crabtree et al., 2020; Widmann, 2021). This might make it harder to find an effect of radical-right success due to ceiling effects. The fact that, despite that, we find such large effects, makes the magnitude of the effects all the more remarkable.

Looking at the effect on each individual emotion also shows interesting patterns. The stronger effects are on hope and enthusiasm. As shown in Fig. 1, these are the emotions that radical-right politicians themselves tend to use less often. The fact that the remaining politicians respond to radical-right success by increasing these emotions suggests that they may do so with the aim of creating a sharp distinction between their rhetoric and that of the radical-right. At the same time, both of these emotions tend to be used for electoral mobilization (Brader, 2005; Just et al., 2007), suggesting that politicians may increase their use of positive emotions as a ways of mobilizing voters against the newly successful radical-right and the negativity of its discourse. We explore this mechanism in further detail in the next section. That the effect is stronger on these two emotions further highlights the importance of looking into different discrete emotions, since differences in their patterns can point towards the mechanism behind one’s empirical findings.

Fig. 2
figure 2

Effect of the parliamentary presence of radical-right MPs on the use of emotional appeals by the remaining MPs

Full size image

The Online Appendix provides a number of other robustness checks and additional analyses. Figures D3 and D4 show little evidence of heterogeneity depending on the government or opposition status of the MP’s party. Figure D5 shows that the main effect can be replicated both in states where AfD speeches are more frequent than the median and in states where they are less frequent than the median. Figure D11 presents a placebo using the election in Hessen in 2013. In this election, the AfD narrowly failed to enter parliament. As such, it represents a good placebo test to check whether the results are indeed driven by the parliamentary entry of radical-right politicians, or if these changes in emotionality happen after each state election. Finally, Fig. D13 shows the results of a robustness check where we replicate the analyses after sequentially removing each state from the sample.

Creating Distance Towards Radical-Right Politicians? Investigating the Mechanism

Quantitative Analyses

In the previous section we found that, while the parliamentary presence of radical-right MPs does not affect the extent to which MPs of the remaining parties use negative emotions in their speeches, it increases their use of positive emotions—especially hope and enthusiasm. As we have argued in the theoretical section, a possible explanation for this finding is that MPs of the remaining parties increase their use of positive emotions to distance themselves from radical-right discourse.

If this is the mechanism in place we should find the effect to be stronger on MP’s of mainstream parties—who have more to lose from an increase in the overall negativity of the political debate than do their counterparts from populist parties, who likely profit from this increased negativity. In the case at hand, this means that the effect should be weaker on MP’s of the populist left party The Left. On the other hand, among mainstream parties, the effect should be stronger on left-wing MPs, because their values are more affected by the divisive negative discourse of the radical-right and because their voters also tend to react to radical-right success by holding closer to the norms breached by the radical-right.

To test whether the effect is stronger on MPs of mainstream parties than on their counterparts from populist parties, we replicate the main analyses shown in Fig. 2 on two sets of subsamples: MPs of the populist party The Left; and MPs of mainstream parties—CDU and CSU, SPD, FDP, and Greens.Footnote 10, Footnote 11 We would expect the effect to be much smaller on MPs of the populist party The Left, the descendant of the ruling party in former East Germany.

The results of these analyses are shown in Fig. 3. Supporting the strategic distancing mechanism, the figure shows a clear difference between MPs from The Left party and those from mainstream parties. The effects are much weaker for MPs of The Left party (upper panel in the figure) than for MPs of mainstream parties (lower panel in the figure).

Fig. 3
figure 3

Effect on radical left party The Left and on mainstream parties

Full size image

We now look at the second empirical implication of the distancing mechanism—that, amid mainstream politicians, the effect should be stronger on those left-of-center. To do so, we replicate the analyses on another two subsamples composed of MPs from left-wing parties and MPs from right-wing parties. Because we are interested in mainstream politicians, we remove the party The Left from these analyses.Footnote 12 Thus, we consider as left-wing MPs those who belong to the Greens and the SPD (Social Democrats). We consider as right-wing MPs those who belong to CDU/CSU (Christian Democrats) and FDP (market liberals). As with the main analyses, we remove MPs from the AfD from these subsample analyses. Figure 4 shows the results of these analyses. We find a pattern similar to the one shown in Fig. 2 in both subsamples. The success of radical-right politicians increases the use of positive emotions by both left- and right-wing MPs. Supporting the distancing mechanism, however, the effects on the use of positive emotions are stronger for left-wing MPs (upper panel on the figure), than for right-wing MPs (lower panel on the figure). For most positive emotions, the coefficient size for left-wing politicians are around twice as large as those for right-wing politicians.

Fig. 4
figure 4

Effect conditional on the ideology of the MPs

Full size image

One question that is raised by these findings is whether this increase in the use of positive emotions is sufficient to make up for the increased negativity that radical-right politicians bring to the debate. To investigate this question, we create a variable that gives us the difference in the values of PCA for negative emotions and those of the PCA for positive emotions. Positive values in this variable mean that the debate has an overall positive tone; negative values mean that the debate has an overall negative tone. Since we are interested in evaluating whether the remaining politicians can counter the negativity brought about by the parliamentary entry of the radical-right, we do not remove AfD politicians from these analyses. We then look at the evolution of this variable before and after the state elections that brought the radical-right to parliament.

Fig. 5
figure 5

Does the overall tone of the debate change? Differences in the use of positive and negative emotions as a function of distance to the state elections

Full size image

The results of this exercise are shown in Fig. 5. As the figure shows, there is no significant discontinuity in the difference between the use of positive and negative emotions after the state elections that bring the radical-right to parliament. These descriptive trends seem to suggest that, in using more positive emotions as a response to the success of the radical-right, the remaining politicians are able to compensate for the negativity that radical-right politicians bring to the debate.

Qualitative Analyses

If the results are indeed driven by a conscious strategy by MPs to distance themselves from the radical-right rhetoric, we should find evidence of that strategy by looking into the speeches in qualitative fashion. To check for this, we draw upon a subsample of the speeches in our data that makes the amount of data manageable. Since our results show an increased use of positive emotions as a consequence of radical-right success, we start with taking the pool of speeches in our data that have a very high level of each of the positive emotions—values above twice the mean of each emotion— from the post-election period, by politicians who are not from the AfD, but which refer to the AfD. From this pool, we randomly choose a subsample of 400. After removing speeches from members of state government, as in our main analyses, we end up with a total of 361 speeches, which we look into in qualitative fashion. We draw upon a sample of highly emotional speeches because we are interested in looking for patterns that may explain why MPs react to radical-right success by increasing their use of positive emotions. This being said, it should be noted that this subsample represents a most-likely-case scenario, whose analyses may be biased in the favor of the hypotheses of the paper.

Anecdotally, in these speeches we do find many examples of MPs calling out the negative, norm-breaching discourse of the radical-right as something new and unacceptable, that was detrimental to healthy democratic debate. This suggests that the remaining politicians are aware of how negative discourse of the radical-right breaches established norms regulating the political debate.

We also find many instances of politicians building a stark contrast between such discourse and that of their own party. In so doing, they often present themselves as the enforcers of the norms that this discourse breaches, claiming that they will fight to keep such norms from eroding. This contrast between the radical-right politicians, breaching established norms, and one’s own party is often accomplished by using emotional appeals.

Finally, reinforcing the idea that the use of positive emotions is strategic, we also find instances of politicians explicitly calling for voters to mobilize against the norm-breaching rhetoric of the radical right—and using positive emotions to do so. They seem to make particular use of enthusiasm and hope to this end. This finding is consistent with the quantitative finding that these two emotions were the ones whose use increases the most after the radical-right enters parliament. As we have mentioned in the results section, this is in line with previous research that has shown that these emotions are frequently used to mobilize voters. Appendix E provides speeches that exemplify all these points.

To check whether these anecdotes represent a larger pattern in our data, we conduct more systematic qualitative analyses. Following a deductive approach, we manually code each speech in this subsample for whether the speaker reacts to radical-right success and rhetoric in the ways outlined in the anecdotes above, and which are relevant to our argument: reacting to radical-right success and rhetoric by calling out their general style of doing politics; by calling out their debating style in parliament specifically; and by building a contrast between the style of radical-right politicians and that of the speaker’s party. Within each of these three categories, we also code the reason why MPs react to the radical-right in this way. These are provided in the Appendix F. To put these values into perspective, we also code reactions that criticize radical-right policy instead of their rhetoric. In so doing, we hope to have a baseline against which to compare our values.

Table 2 Frequency of reactions to the success and rhetoric of AfD politicians
Full size table

Table 2 shows the results of these analyses. The most frequent reaction among these four is to call out the general political style of the AfD as unacceptable—a reaction which we find in almost half the speeches analyzed. This is followed by the building of contrast between AfD politicians and the speaker and her party. As detailed in Appendix F, the main ways by which politicians mark contrast involve the use of positive emotions. Calling out of the debating style of AfD politicians in parliament is a reaction found in 29% of the speeches. It may be argued that it is rather unsurprising that politicians react to the success of a new party by disagreeing with it. However, we find the proportions of each reaction quite impressive, considering that we only find criticism of policy—which should happen quite often in parliamentary debates—in 26% of the speeches analyzed.

Figure D16 in the Online Appendix reports a test for an alternative mechanism: that the results may be driven by radical-right politicians giving more importance to issues which the remaining politicians talk more positive about. Our evidence does not support this alternative mechanism.

Conclusion

The way mainstream politicians react to the strategy of their niche challengers is what makes or breaks the success of such strategy (Meguid, 2005). Existing literature has provided extensive evidence that, when it comes to policy positions, mainstream politicians tend to react to the threat posed by radical-right contenders by coming closer to their position. Our results suggest that this is not the case when it comes to the tone of the discourse. While the rhetoric of radical-right politicians is characterized by extensive use of negative emotions, their success increases the use of positive emotions by the remaining politicians. A mix of quantitative and qualitative analyses suggest this is a means by which politicians create a contrast between the radical-right, breaching established norms, and themselves, enforcers of those norms.

One possible explanation for this difference is that mainstream politicians follow different strategies in different dimensions of their electoral appeal. They may adopt increasingly rightist policy positions as a ways of trying to appeal to radical-right voters; and, at the same time, call out their negative rhetoric as unacceptable to try and keep their traditional constituencies. This can represent a strategy of “parroting the pariah”, whereby established politicians come closer to the positions of extremist contenders but ostracize them at the same time (Van Spanje & de Graaf, 2018).

Our results also run contrary to a body of research showing that radical-right success can normalize behavior and preferences associated with the far right (Bischof & Wagner, 2019; Bursztyn et al., 2017; Valentim, 2021). The likely explanation for this difference lies in the different incentives facing citizens and politicians. For citizens, radical-right success can present an opportunity to publicly reveal previously stigmatized preferences (Valentim, 2021). For politicians, instead, it can present an opportunity to highlight their commitment to democratic norms in an effort to mobilize voters.

On the other hand, our findings are in line with previous research in highlighting how politicians tend to use emotional language strategically (Crabtree et al., 2020; Kosmidis et al., 2019; Müller, 2020). However, our paper adds an important insight to this conclusion. As our findings suggest, the emotional appeals used by different politicians change depending on who their challengers are and how electorally successful they become. Concretely, the success of candidates whose rhetoric is deemed unacceptable increases the incentives to use emotions as a tool to mobilize voters against such rhetoric.

One question raised by this conclusion is to what extent these findings are likely to travel beyond the German case. Democracies tend to create norms against parties with affinities to previous dictatorships (Dinas & Northmore-Ball, 2019). This is even more likely in Germany, given the particularly devastating consequences of the Nazi regime the country witnessed. The creation of strong norms against rhetoric associated with radical-right politicians might force established politicians to respond to their success with clear differentiation. This being said, we believe the findings are not simply a product of the specific characteristics of the German case. Empirical findings from Denmark show a similar strategy with members of parliament distancing themselves rhetorically by portraying cooperation with a radical-right party as morally objectionable (Hjorth, 2020). One can also find even anecdotal evidence of such strategies in other contexts. In Spain, the incumbent party PSOE reacted to the successful radical-right Vox by stressing its leftist position and the party’s role in the Spanish transition to democracy (De Vries & Hobolt, 2020, pp. 250–251). During the US Presidential Campaign of 2016, Hillary Clinton refused time and time again to give in to the negative rhetoric of Donald Trump which, like in our data, she called out as unacceptable.

Overall, our findings highlight how electoral competition can push mainstream politicians to distance themselves from the rhetorical style of the radical-right. As the qualitative examples provided in the Appendix E show, radical-right success often impels the remaining politicians to mobilize voters on an opposition to the norms that are breached by the newly successful contenders. While radical-right rhetoric is in many ways damaging to democracy—in their perpetuation of stereotypes about minority groups, spread of fake news, or negative discourse—our results suggest that their success can create incentives for other politicians to distance themselves from negativity of radical-right rhetoric. In so doing, they may be able to mitigate part of these negative consequences. Given that negative debating has been associated with detrimental outcomes such as affective polarization (Iyengar et al., 2012), this finding carries important implications for the quality of democracy. In increasing the use of positive instead of negative emotions, mainstream politicians do not feed potential loops of negativity that could be very detrimental to the process of public deliberation.