Mainstream partisans’ affective response to (non) cooperation with populist radical right parties

Abstract

What is the role of elite-level inter-party strategies in promoting or reducing negative affective evaluations of citizens? This article explores this puzzle by focusing on the decision of mainstream parties to exclude or cooperate with populist radical right-wing parties (PRRPs) in Western European countries. The analysis is based on CSES cross-national data and the LISS Dutch panel data merged with the record on whether PRRPs are subject to a cordon sanitaire or (in)formally cooperate with mainstream parties in a government coalition. The findings suggest that political exclusion fuels loathing for PRRPs among mainstream party supporters, while cooperation reduces dislike of PRRPs both among supporters of PRRP coalition partners and left-wing voters. Distinguishing between informal and formal coalitions, the cross-national data shows that only formal cooperation has a significant effect. However, the longitudinal Dutch data shows a temporary decrease of dislike also during informal cooperation.

Keywords:

• Negative affect
• populist radical right-wing parties
• cordon sanitaire
• inter-party cooperation

In this article we investigate whether voters’ negative affect towards populist radical right-wing parties is partly driven by the decision by mainstream parties to exclude or (formally or informally) cooperate with them.

Harteveld, Mendoza, and Rooduijn (2022) have shown that in Europe populist right parties (PRRPs) are the ones receiving and radiating the highest level of dislike. Other researches consistently found that PRRPs trigger and receive exceptional levels of animosity from citizens (Gidron, Adams, and Horne 2019; Helbling and Jungkunz 2020; Mudde and Kaltwasser 2018; Reiljan and Ryan 2021).

Recent studies have investigated the conditions under which political elites influence citizens’ party dislike by addressing the role of inter-party cooperation. These studies argue that inter-party cooperation can decrease general negative affect by decreasing out-group animosity (Bassan-Nygate and Weiss 2022; Gidron, Adams, and Horne 2019; Wagner and Praprotnik 2024), but that such an effect may be limited to supporters of the cooperating parties (Harteveld, Mendoza, and Rooduijn 2022).

Since the 1980s, PRRPs increasingly succeeded in entering national European parliaments, reaching an electorate of about 15% (see Rooduijn, et al. 2023). This success forced mainstream parties to respond to populist radical right-wing challengers (Schumacher and van Kersbergen 2016), but they did it in different ways. In Belgium, France, Germany, and Sweden mainstream parties consistently decided to exclude the populist radical right-wing parties (Vlaams Blok (VB), Front National (FN/RN), Alternative for Germany (AfD), and Sweden Democrats (SD))¹ by announcing that they would not cooperate in any form with PRRPs – establishing a so-called cordon sanitaire (see e.g. Heinze 2018; van Spanje and Weber 2019).² When mainstream parties establish cordon sanitaires, they claim PRRPs threaten liberal democracy by rejecting the democratic principle of equality.

In other countries such as Austria, Denmark, Finland, Italy, Netherlands, Norway, Switzerland, and UK despite occasional protests, several mainstream parties pursued a more inclusive stance towards their respective PRRPs (see Table 1 for an overview) by forming a governing coalition (Bale et al. 2010; Döring and Philip 2019).

Table 1. Overview final set of cases for the cross-national analysis.

Country	PRRP	Cordon Sanitaire	In opposition but not ‘excluded’	Informal cooperation	Formal cooperation
Austria	Freedom Party	–	2008 (18.6)	–	2017 (26)
	of Austria (FPÖ)		2013 (20.5)
	Alliance for the Future (BZÖ)	–	2008 (10.7)	–	–
Belgium-Flanders	Flemish Block (VB)	2019 (11.95)	–	–	–
Denmark	Danish People’s Party (DF)		1998 (7.4)	2001 (12.3)	–
			2019 (8.7)	2007 (13.9)
	The New Right (NB)		2019 (2.4)	–	–
Finland	True Finns (PS)		2011 (19.1) 2019 (17.5)	–	2015 (17.7)
France	Front National/ National Rally (FN/ RN)	2002 (16.8)	–	–	–
		2007 (4.3)
		2012 (17.9)
		2017 (21.3)
Germany	Alternative for Germany (AfD)	2017 (12.6)	–	–	–
		2021 (10.3)
Great Britain	Independent Party (UKIP)
Greece	Popular Orthodox Rally (LAOS)				2009 (5.6)
	Greek Solution (EL)	–	2019 (3.7)	–	–
Italy	The League (L)	–	–	–	2018 (17.4)*
	Brothers of Italy (FdI)	–	2018 (4.4)	–
Netherlands	Party for Freedom (PVV)	–	2006 (6)	2010 (15.5)	–
			2017 (13.1)
			2021 (10.8)
	Forum for Democracy (FvD)	–	2017 (1.8) 2021 (5)	–	–
Norway	Progress Party (FrP)	–	1997 (15.3)	2001 (14.7)	2013 (16.3)
			2005 (22.1)		2017 (15.2)
			2009 (22.9)
Sweden	Sweden Democrats	2014 (12.9)	–	–	–
	(SD)	2018 (15.5)
Switzerland	Swiss People’s Party	–	–	–	1999 (22.5)
					2003 (26.7)
					2007 (28.9)
					2011 (26.6)

* Italy is coded as formal cooperation for 2018, despite FdI being in oppisition because Lega was in government.

In this article we try to understand how these party strategies influence voters’ affective evaluations of parties, by investigating whether the degree of formality of cooperation influences citizens’ affect. We examine whether exclusion versus cooperation, the two contrasting strategies of mainstream parties, improve or worsen partisan affect. We pursue this agenda by carrying out two complementary studies. The first study uses cross-sectional post-electoral individual-level data from the Comparative Study of Electoral Systems (CSES 1996-2019) in 13 Western European countries, paired with data (from ParlGov) on whether and when PRRPs faced a cordon sanitaire or participated in government. We also carry out a second study in which we employ the Dutch panel data from LISS (Longitudinal Internet Studies for the Social Sciences). LISS covers the period in which two mainstream right parties (Christian Democratic Appeal (CDA) and People’s Party for Freedom (VVD) formed a minority government support agreement with the PRRP ‘Party of Freedom’ (PVV) after the 2010 Dutch parliamentary election. Using panel data helps identifying whether partisans’ affective attitudes towards PVV change in response to changes in the strategy of mainstream parties.

Our research contributes to the existing literature in four ways. First, we investigate how the elites’ behaviour can drive citizens’ negative affect, hence advancing the literature that studies whether citizens are aware of inter-party behaviour.

Second, building on van Heerden and van der Brug (2017), we are, to the best of our knowledge, the first to consider how different types of government cooperation affect citizens’ evaluations by using observational data at the individual level. In particular, we add to the current body of knowledge on the mechanisms responsible for the exceptional position of PRRPs in Europe (Harteveld, Mendoza, and Rooduijn 2022; Gidron, Adams, and Horne 2019).

Third, we contribute to the literature on party responses to populist right-wing challengers. So far, studies have examined how mainstream parties’ dealing with PRRPs influences PRRPs’ electoral support and ideological stances (Akkerman and Rooduijn 2015; Downs 2002; Down and Kyung 2020; van Spanje and van der Brug 2007; van Spanje and Weber 2019) and whether it affects voters’ attitudes of trust in political institutions or satisfaction with democracy (Harteveld et al. 2021; Mudde and Kaltwasser 2018). We show that mainstream parties’ response to successful PRRPs influences mainstream partisans’ evaluations of PRRPs as well. In a broad sense, we also then contribute to the literature on coalition formation. While there is widespread recognition in coalition research that it makes a difference in many ways whether a governing coalition operates under a formal or informal agreement (Adams, Ezrow, and Wlezien 2016; Fortunato and Stevenson 2013), we are the first to show that such a distinction is also important with respect to negative affect.

Finally, we make an explicit contribution to the research on political exclusion. Previous quantitative studies on the cordon sanitaire are based on an expert survey which, however, ends in 2010 (van Spanje and de Graaf 2018). While an expert survey has advantages over the literature review strategy used here, the information collected in this study allows for up-to-date analyses.

State of the art

A large body of studies showed that voters’ negative affect towards parties can be driven by several factors: electoral competition (Bassan-Nygate and Weiss 2022; Hernández, Anduiza, and Rico 2021; Sheffer 2020), exposure to political campaigns in general, and negative campaigns in particular (Iyengar, Sood, and Lelkes 2012; Sood and Shanto 2016), consumption of partisan media (Lelkes, Sood, and Iyengar 2017; Levendusky 2013), economic inequality (Gidron, Adams, and Horne 2018), majoritarian electoral systems (Gidron, Adams, and Horne 2018) and elite ideological polarisation (Banda and Cluverius 2018; Rogowski and Sutherland 2016).

The current state of research on factors that potentially reduce party negative affect is less extensive, with, to the best of our knowledge, only four studies addressing inter-party cooperation and negative affect. First, Gidron, Adams, and Horne (2019) show that partisans of a governing party display significantly more sympathy towards the other coalition party/ies. Second, Bassan-Nygate and Weiss (2022) find that cross-party cooperation (namely: receiving information about a unity government) can decrease general partisan negative affect by decreasing out-group animosity. Third, Harteveld et al. (2021) find that cross-party cooperation reduces dislike of out-groups, but only among coalition partners, whilst among supporters whose parties remain outside the coalition, they find greater reluctance in the cross-state analysis, but less reluctance in the individual case study. Finally, an experiment conducted by Wagner and Praprotnik (2024) involving three Austrian parties shows that coalition signals can reduce negative out-party affect. Empirical evidence on the effects of inter-party cooperation is therefore still scarce, and entirely missing in a wide comparative perspective.

A study by Meléndez and Kaltwasser (2021) shows that more than half of all European voters state they would never vote for a PRRP, suggesting that most Europeans clearly identify as non-PRRP supporters (Medeiros and Noël 2014). Research has also found that the affective gap between mainstream parties and PRRPs is, unlike the affective gap between different mainstream parties, less triggered by ideological differences on the socio-economic left–right or even cultural dimension (Harteveld 2021; Harteveld et al. 2021; Reiljan and Ryan 2021), but rather by PRRPs’ stances on nativism and populism (Harteveld et al. 2021) which is largely perceived by non-PRRP supporters as a social norm violation (Blinder, Ford, and Ivarsflaten 2013).

Political exclusion (also knowns as cordon sanitaire) implicitly or explicitly indicates that a given party violates such social norms by signalling moral distancing (Han 2020; Hjorth 2020). This is often presented as an act of defending the liberal democratic political system from a party which promotes dangerous, unacceptable, or reprehensible ideologies (Art 2007; Hjorth 2020; van Spanje 2010), hence the name cordon sanitaire, a term originated in disease control, which refers to the establishment of boundaries to stop and limit the spread of a harmful virus (Downs 2012).

In multi-party systems, political exclusion prevents PRRPs from having any coalition potential and, consequently, a de-facto no chance of executive power (Downs 2002; van Spanje and Weber 2019). Mainstream parties can decide to systematically rule out any potential form of political cooperation with an existing PRRP (van Spanje and van der Brug 2009) only at the national or both at national and local level (Akkerman and Rooduijn 2015). Following the ‘first-order election’ argument according to which party behaviour at the national level is more salient to citizens (Reif and Schmitt 1980; van Spanje and van der Brug 2009), we only consider the national level.

The lingering question is whether this inter-party behaviour influences public opinion’s affect. We know that social identity theory states that social groups are naturally structured by leadership (van Knippenberg 2011). Group leaders have a disproportionate influence to effectively mobilise their followers, define the groups’ identity, norms, and shared goals (Hogg and van Knippenberg 2003; Huddy and Yair 2021), and leaders’ interaction with outgroups determines how followers will view and perceive out-groups (Hogg 2001; Huddy and Bankert 2017; Huddy and Yair 2021). Also, voters identify party elites as leadership (Cohen 2003; Hogg and van Knippenberg 2003; Müller and Strøm 2003). Thus, partisans are likely to listen to their own party’s messages and endorse its positions and values (Lenz 2012; Rooduijn, van der Brug, and de Lange 2016; Steenbergen, Edwards, and Vries 2007).

The literature on heuristics and party cue taking provides further support for the importance of elites. Heuristics and cues facilitate the process of party sorting by providing citizens with ready policy positions (Kirkland and Coppock 2018; Popkin 1995). Hence, when a party is labelled as ‘pariah’ voters will make inferences about the party’s ideological position, and its social conformity (Adams, Ezrow, and Wlezien 2016; Bowler and Nicholson 2018; Fortunato and Stevenson 2013), influencing citizens’ attitudes and behaviour (Fortunato and Stevenson 2013; Harteveld et al. 2019a; Hobolt 2006; Singh and Thornton 2016; Spoon and Klüver 2015).

Hypotheses

We identify three main mechanisms by which mainstream partisans could increase disliking of PRRPs through political exclusion. The first one is stigmatisation. A social stigma is an ‘attribute that is deeply discrediting’ (Goffman 1963), and it is often grounded in a social moral concept becoming a social norm itself. As soon as the message that a given party violates social norms is accepted, the party itself becomes a taboo (Harteveld et al. 2019a; Harteveld et al. 2019b; Hjorth 2020). The stigmatisation of a PRRP is likely to occur if mainstream parties agree that its values threaten liberal democratic societies, so that it should be excluded from politics.

Second, a cordon sanitaire fosters in- and out-group thinking by implying that PRRPs are dangerous and illegitimate (Harteveld et al. 2021; Harteveld, Mendoza, and Rooduijn 2022; Iyengar and Westwood 2015; Moffitt 2021), hence a political threat (Harteveld, Mendoza, and Rooduijn 2022; Mudde 2004). According to Social Identity Theory, animosity and loathing towards outgroups are fuelled by perceiving another group as a threat (Brewer 1999; Dunwoody and Dennis 2019; Fiske, Cuddy, and Glick 2007; Huddy, Mason, and Aarøe 2015; Renström, Bäck, and Carroll 2021; Suhay, Bello-Pardo, and Maurer 2018). Thus, a cordon sanitaire could trigger and/or increase voters’ perception of PRRPs as a threat, hence increasing mainstream partisans’ hatred of PRRPs.

Finally, political exclusion could also lead to the emergence of negative party identification, that is, a stable negative identity not necessarily rooted in a positive identification with another party, but simply based on loathing a PRRP (Bankert 2021; Maggiotto and James Piereson 1977; Mayer and Russo 2024).

To summarise, a cordon sanitaire could cause social stigmatisation, increased threat perception, the development of negative identification with PRRPs, or even a combination of all three mechanisms. In short, mainstream partisans are expected to be more opposed to PRRPs when there is a cordon sanitaire.

H1: Mainstream party supporters’ dislike toward PRRPs increases when PRRPs are subject to a cordon sanitaire.

We already mentioned studies exploring ways to depolarise mainstream supporters by reducing their aversion to PRRPs (Bassan-Nygate and Weiss 2022; Gidron, Adams, and Horne 2019; Harteveld, Mendoza, and Rooduijn 2022; Wagner and Praprotnik 2024). A destigmatisation or legitimisation process can happen when mainstream parties are open to (in)formal government cooperation. Cooperation signals the opposite of a stigma (Harteveld et al. 2019a; Harteveld et al. 2019b; Hjorth 2020) by implying that a given PRRP does not fall outside social norms, so it is not necessary to deny to it access to executive power (Harteveld, Mendoza, and Rooduijn 2022). This can be seen as an attempt to normalise or mainstream a PRRP (Art 2007; Grabow and Hartleb 2013a; Moffit 2021).

Empirical research into intergroup relations indicates that cooperation between interdependent groups with different identities can reduce prejudices and improve intergroup relations (Christ and Kauff 2019; Gaertner et al. 1999; Gidron, Adams, and Horne 2019). Two cooperating groups will develop an overarching ‘we’ identity, resulting in a reduction of the salience of group boundaries, and of negative stereotypes and prejudices (Carlin and Love 2018; Christ and Kauff 2019; Gaertner et al. 1999). Thus, government coalitions create beneficial structures and shared goals (Bassan-Nygate and Weiss 2022) which should warm up the relation between coalition parties.

Literature on coalition heuristics supports this reasoning. In coalitions, the need for compromise overshadows distinctive ideological identities and subverts parties’ established party labels (Martin and Vanberg 2008), and voters perceive cabinet parties as more ideologically similar than parties that do not cooperate in government (Adams, Ezrow, and Wlezien 2016; Fortunato and Stevenson 2013). The same effect has been observed in informal coalitions and support arrangements for minority governments (Hjermitslev 2020). There is another element to consider: in Western European countries PRRPs are usually the junior coalition party. Joining a coalition can help their normalisation process not only in the eyes of the voters, but also de facto, as it has been found that 1) the perceived ideological similarity seems to play a significant role in the ideological positioning of junior coalition parties, and that 2) when it comes to ideological placement, junior parties tend to follow the larger Prime Minister party and not the other way around (Fortunato and Adams 2015).

Consequently, by building an (in)formal coalition with PRRPs at the executive level, stigmas may be reduced, and an increased sense of ideological similarity can be developed. Each of these mechanisms should, in turn, lead supporters of cooperating mainstream parties to display more positive affect towards PRRPs by reducing perceptions of threat, dissimilarity and competition. Hence, we expect that:

H2a: Mainstream party supporters’ dislike of PRRPs decreases when their preferred party builds a government coalition with PRRPs.

It is likely that government participation by PRRPs will have different implications for supporters of non-cooperating mainstream parties opposing the government (Gidron, Adams, and Horne 2019; Harteveld, Mendoza, and Rooduijn 2022). Intuitively, the relationship between opposition parties and governing parties is more conflictual, regardless of ideological differences (Weschle 2018). In addition, PRRPs’ involvement in the government will likely shift policy outcomes even further away from the positions of opposition parties (and their supporters). In turn, mainstream opposition party supporters’ aversion to PRRPs should rise if they perceive government participation as having tangible negative consequences for them (Dunwoody and Dennis 2019; Fiske, Cuddy, and Glick 2007; Harteveld, Mendoza, and Rooduijn 2022; Huddy, Mason, and Aarøe 2015; Renström, Bäck, and Carroll 2021; Suhay, Bello-Pardo, and Maurer 2018). Thus, we expect that:

H2b: Mainstream opposition party supporters’ dislike of PRRPs increases when PRRPs participate in government.

We also distinguish between formal from informal cooperation. An informal cooperation leaves to mainstream parties more leeway for differentiation, which helps them protect their public profile from an ideological equalisation with PRRPs (Fortunato and Adams 2015; Spoon and Klüver 2015), and shields them better from criticisms of legitimising and normalising PRRPs. As informal coalitions persist, rewards structures and common goals stay more embedded in competition, resulting in stronger compromise negotiation processes (Hjermitslev 2020). However, for supporters of mainstream coalition partners and mainstream opposition parties, this may hamper the development of a common identity. Hence, our expectations about the effects of government cooperation on mainstream partisans (H2a and H2b) vary in strength depending on whether government cooperation is formal or informal:

H3a: Mainstream party supporters’ dislike of PRRPs decreases more when their preferred party builds a formal versus an informal government coalition with PRRPs.

H3b: Mainstream opposition party supporters’ dislike of PRRPs increases more when PRRPs participate formally versus informally in government.

Research design

In order to test our expectations, we rely on the CSES. We examine thirteen Western European countries where PRRPs are either excluded from mainstream political parties or (formally or informally) participating in government. We analyse these data with a multilevel approach, which comes with some limitations, including the impossibility of observing whether supporters’ affective attitudes reflect parties’ strategic inter-party behaviour. To overcome this shortcoming, we also carry out a second study which employs the Dutch LISS Policy and Values Panel (2007 to early 2014), allowing us to observe whether partisans of mainstream parties change their evaluation after their party’s informal cooperation with a PRRP.

CSES data and methods

In order to identify salient populist radical right-wing parties represented in Western European parliaments, we use the PopuList dataset, featuring 31 European countries since 1989. The PopuList dataset data only includes parties that have won (at least) one seat or 2% of the votes at national parliamentary election, and it takes into account variation over time (Rooduijn et al. 2023).

The dataset is created by merging the CSES Integrated Module Dataset (modules 1 to 4, 1996–2019) with Module 5. The final dataset includes thirteen Western European countries with salient populist radical right-wing parties: Austria, Belgium [Flanders],³ Denmark, Finland, France Germany, Great Britain, Greece, Italy, Netherlands, Norway, Sweden, and Switzerland).

To establish whether PRRPs participated (did not participate) (in)formally in national governments we rely on the Parliament and Government database (ParlGov) (Döring and Philip 2019). We gathered information on whether and when PRRPs faced political exclusion using secondary literature review. Table A1 in Online Appendix A provides an overview of the results of the literature review as well as the used literature sources.

The main focus of our analysis is mainstream party supporters’ attitudes. The overall share of respondents who identify with or support a mainstream party is 58% in the CSES data, and it rises to 75% when considering actual voting behaviour. About 30% of the overall mainstream partisans have experienced the political exclusion of a PRRP, whereas about 33% experienced PRRPs’ formal or informal government participation. Moreover, about 23% of partisans from this last group belong to mainstream parties that cooperated with a PRRP (see Table A3 in Online Appendix A).

As data are sampled from both the micro and the macro level for varying time periods, the regression analysis needs to account for the specificity of such a clustered design. In a nested data structure, that is, individual survey responses (level 1, N = 20515) are nested within country-years (level 2, N = 36) that are nested in countries (level 3, N = 13), the influence of the contextual variables would be greatly biased towards high significance levels if the analysis treats all lower-level observations as independent (Hox, Moerbeek, and van de Schoot 2017; Steenbergen and Jones 2002). Running an intercept-only model the intraclass correlation coefficient (ICC) – the between cluster variation divided by the total variance – is 0.0607at level 2 and 0.0343 at level 3. This means that 6.07% of the variance in mainstream partisans’ affective attitude vis-à-vis PRRPs is explained by between year-within-country variation and 3.43% between-country differences. Thus, we use a mixed linear multi-level models run with robust standard errors with random effects per countries.

LISS data and methods

To complement our CSES analysis we employ the LISS (Longitudinal Internet Studies for the Social Sciences) panel dataset using waves 1–7 (2007–2014). The data are based on a true probability sample of households drawn from the population register and ask respondents aged 16 every year a large variety of questions on inter alia politics and values.⁴

The 7 annual waves of the LISS panel allows us to follow Geert Wilders’ right-wing populist party ‘Party for Freedom’ (PVV) during three elections: 2006 (when the party was launched and gained 9 seats, but was ignored by other parties), 2010 (when it won 24 seats informally participated in government), and 2012 (when it won 15 seats, and it did not participate in government). These data allow us to observe whether Dutch mainstream partisan changes their evaluation of the PVV depending on whether the CDA and the VVD informally cooperate with the PVV prior to each wave.

Unfortunately, a direct measure of party affiliation (as used in for CSES) is not available in the LISS data. To identify mainstream party supporters among the respondents a survey item we then use the following question ‘If parliamentary elections were held today, for which party would you vote?’. This item has already been used by others for this purpose (Harteveld, Mendoza, and Rooduijn 2022; Renström, Bäck, and Carroll 2021). As for Study 2, to include a larger part of the sample, we use as alternative measure of partisanship the respondents’ actual vote choice in the most recent parliamentary election (see Table D2 in Online Appendix D).

Table 2. The effects of a cordon sanitaire and government cooperation with PRRPs on mainstream party supporters’ dislike of PRRPs.

Dislike – Like PRRP	Model 0		Model 1		Model 2
(Scale 0-10)
	b	se	b	se	b	se
Degree of Partisanship
Somewhat close	−0.055	(0.058)	−0.071	(0.064)	−0.074	(0.064)
Very close	−0.174**	(0.056)	−0.185**	(0.066)	−0.192**	(0.067)
Ideological Distance	−0.460***	(0.065)	−0.394***	(0.045)	−0.395***	(0.045)
SWD
Not very satisfied	−0.381*	(0.166)	−0.377*	(0.154)	−0.386**	(0.148)
Fairly satisfied	−0.705**	(0.237)	−0.690**	(0.218)	−0.693**	(0.214)
Very satisfied	−0.611*	(0.286)	−0.614*	(0.256)	−0.629*	(0.252)
In-bloc evaluation	0.019	(0.053)	0.020	(0.051)	0.024	(0.051)
Age in years (Squared)	−0.000	(0.000)	−0.000	(0.000)	−0.000	(0.000)
Female	−0.353***	(0.046)	−0.339***	(0.046)	−0.337***	(0.045)
Level of Education
Primary	−0.112	(0.085)	−0.124	(0.109)	−0.140	(0.119)
Secondary	−0.264*	(0.104)	−0.308*	(0.131)	−0.322*	(0.142)
Post-secondary	−0.387***	(0.114)	−0.455***	(0.128)	−0.479***	(0.133)
University	−0.864***	(0.131)	−0.925***	(0.165)	−0.938***	(0.174)
Income [household quintile]
Second	−0.019	(0.044)	−0.009	(0.048)	−0.013	(0.047)
Third	−0.038	(0.065)	−0.030	(0.069)	−0.030	(0.069)
Fourth	−0.168*	(0.070)	−0.152*	(0.074)	−0.152*	(0.071)
Highest	−0.244**	(0.089)	−0.284***	(0.083)	−0.281***	(0.081)
Winner effect	−0.116	(0.117)	−0.288*	(0.113)	−0.392***	(0.091)
Vote share (PRRP)	0.037*	(0.018)	0.043	(0.022)	0.042	(0.021)
Nr. election (PRRP)	0.006	(0.023)	−0.026	(0.033)	−0.025	(0.034)
Cordon Sanitaire			−0.605*	(0.270)	−0.708**	(0.269)
Cooperation			0.134	(0.183)	−0.252	(0.188)
In-party is coalition partner			0.931***	(0.216)	0.560**	(0.212)
Interaction:
Cooperation # In-party is coalition partner					0.861***	(0.234)
Constant	4.768***	(0.584)	4.652***	(0.732)	4.831***	(0.692)
Random effects	Yes		Yes		Yes
Variance (residual)	6187791		5.2005		5.1784
Log likelihood	−45783.449		−45095.364		−45052.411
AIC	91614.9		90244.73		90160.82
N level-1 (persons)	20311		20075		20075
N level-2 (country-years)	36		36		36
N level-3 (countries)	13		13		13

Multilevel mixed-effects regression model examining the predictors of dislike of mainstream partisans of PRRPs. Random effects are specified for countries. Robust standard errors are reported.

Table 3. The effect of informal cooperation on mainstream partisans’ affective attitude towards the PVV, Dutch panel data analyses.

Dislike – Like of PVV	FE Model 0		FE Model 1		LDV Model
(Scale 0-10)	b	se	b	se	b	se
Dislike – Like of PVV (Lagged)					−0.033**	(0.012)
Interested in politics
fairly	0.051	(0.071)	0.042	(0.070)	0.005	(0.061)
very	0.065	(0.082)	0.055	(0.082)	−0.011	(0.075)
Strong partisanship	−0.037	(0.041)	−0.048	(0.040)	−0.011	(0.045)
Ideological distance on:
Socio-economic left-right dimension	−0.059***	(0.014)	−0.049***	(0.014)	−0.057***	(0.014)
EU integration	−0.026	(0.017)	−0.022	(0.017)	−0.021	(0.017)
Multiculturalism	−0.296***	(0.047)	−0.301***	(0.047)	−0.257***	(0.048)
SWD	0.027*	(0.013)	0.023	(0.013)	0.033**	(0.012)
In-bloc evaluation	0.217***	(0.017)	0.218***	(0.017)	0.183***	(0.017)
Winner effect	0.091**	(0.031)	−0.145***	(0.035)	−0.134***	(0.035)
Level of education
Middle	−0.023	(0.162)	0.010	(0.163)	0.079	(0.176)
High	−0.507***	(0.137)	−0.477***	(0.133)	−0.537**	(0.171)
Age (squared)	−0.000	(0.000)	−0.000*	(0.000)	−0.000	(0.000)
Female	0.033	(0.986)	0.094	−1.137	−0.691***	(0.058)
Income [netto household]
1300 − 1900 Euro	0.104	(0.113)	0.102	(0.111)	0.060	(0.107)
1901 − 2723 Euro	0.117	(0.110)	0.111	(0.108)	−0.006	(0.106)
2724 − 3600 Euro	0.112	(0.115)	0.095	(0.114)	0.006	(0.110)
3601 − 4581 Euro	0.039	(0.119)	0.032	(0.118)	−0.076	(0.114)
more than 4581 Euro	0.055	(0.122)	0.038	(0.120)	−0.078	(0.116)
Electoral success PVV	0.014**	(0.004)	0.013**	(0.004)	0.015**	(0.005)
Informal cooperation	0.469***	(0.026)	0.101**	(0.032)	0.157***	(0.032)
In-party is coalition partner	0.312***	(0.060)	0.069	(0.062)	0.002	(0.063)
Interaction:
Informal cooperation # In-party is coalition partner			0.921***	(0.061)	0.906***	(0.060)
Constant	1.834**	(0.562)	2.059***	(0.618)	2.721***	(0.352)
R2 = within	0.082		0.107		0.109
R2 = between	0.279		0.271		0.147
R2 = overall	0.248		0.245		0.143
sigma u	1.967		1.972		2.027
sigma e	1.178		1.162		1.157
Rho	0.736		0.742		0.754
N (observations)	17035		17035		15880
N (respondents)	5914		5914		4999

Note: OLS estimation. Results show unstandardised coefficients with clustered standard error based on respondents in parentheses. The levels of statistical significance are.

* p < 0.05.

** p < 0.01.

*** p < 0.001.

Results

Study 1: CSES

To decide which countries to include we considered several criteria. First, the comparison had to be plausible. Although a certain level of heterogeneity is inherent in cross-country research, our selection attempts to follow a most similar case approach (Seawright and Gerring 2008). PRRPs from Eastern and Central European countries differ from Western European PRRPs with respect to the core issues they mobilise on (Harteveld et al. 2021), and they also display historically a weaker democratic tradition which presumably shapes the elite party–citizen linkage differently (Grabow and Hartleb 2013b). Therefore, we restricted the analysis to Western European countries in which a salient PRRP is present at the national level, and for which data is available in the CSES data. Unfortunately, this means that Spain could not be included in the analysis. Some countries are included only for one election whereas others are included over multiple elections. Table 1 provides an overview of the countries and time points included in the analysis.

The PopuList dataset builds on peer-reviewed information on party classification and most party classifications are uncontested by scholars, yet some disagreement exists (Mudde 2016). In our selection this applies to the cases of the True Finn party (PS) in Finland and the FrP in Norway until 2013 (Mudde 2016). To control for a possible bias resulting from such borderline cases, the cross-national analysis is carried out also once without the two contested PRRP cases.

Apart from France and UK, all other countries have a proportional electoral system (PR) with varying but relatively low thresholds (from the 0% in Finland to 5% in Germany and Belgium). This a) facilitates the access for political parties like PRRPs and b) increases the need of coalition-building for government formation (Gallagher 2005; Grabow and Hartleb 2013b; Koopmans and Muis 2009; Rydgren 2004). Finally, as Table 1 illustrates, mainstream party strategies vary enough across the cases considered.

The dependent, independent, and control variables

Identifying mainstream party supporters is the first step in studying their attitudes towards PRRPs. We could do this 1) by using the two items from CSES which ask whether respondents feel closer to a party, and if yes which one or, 2) by using the vote choice of respondents for the closest election. This broader measure would allow us to include a larger part of the sample and thus to control for possible bias caused by systematic non-responses in the other item (Knudsen 2020; Westwood et al. 2018; see Table A3 in Online Appendix A).⁵ We opted for the more conservative strategy, but we replicate results using the second operationalisation when relevant.

There is no clear definition of what qualifies as a mainstream party (Moffitt 2021). Some use the degree of establishment of a party, i.e., whether it has been successfully represented in parliament several times, its degree of governability, and/or whether it is ideologically positioned in the centre (Meguid 2005; Pop-Eleches 2010). However, one could argue that there are parties that are ideologically more on the fringe but have been an integral part of the party landscape in parliament for several election years. Hence, we opted to employ a broad definition: Mainstream parties are parties that a) are not PRR parties and b) have a certain degree of establishment, as measured by their representation in the national parliament. We argue that this is a viable choice based on three points: 1) previous studies have shown that PRRPs trigger aversion among supporters of all other parties, regardless of whether they are conservative, left-wing or social democrat (Gidron, Adams, and Horne 2019; Harteveld, Mendoza, and Rooduijn 2022; Helbling and Jungkunz 2020; Mudde and Kaltwasser 2018; Reiljan and Ryan 2021); 2) the term mainstream is normally used simply as an antonym for parties that exhibit some peculiarity or deviate from the ‘norm’ (Moffitt 2021); and 3) for a cordon sanitaire all other parties represented in parliament are relevant. Tables A4 and A2 in Online Appendix A provide an overview of all parties who cooperated with a PRRP. Table A4 shows that in some cases this broad definition includes established parties that belong to the socialist/communist and right-wing party families, party families that are located on the ideological fringe. Therefore, the analysis is repeated to check robustness, considering only parties from the Green, Social Democratic, Christian Democratic, Liberal, and Conservative party families without supporters of socialist/communist and right- wing parties.

Variables

Our dependent variable, negative affect towards parties, is operationalised by measuring respondents’ feelings towards the respective PRRPs by using the classic party dislike-like scale (Reiljan and Ryan 2021).⁶ The PRRP evaluation of respondents is simply the rating assigned to the respective PRR party. In other words, we only rely on the out-group evaluation without considering the in-group positive feeling towards the preferred party.

To test H1 the first main independent variable at the election level is a dummy variable indicating whether a PRRP faces a cordon sanitaire (1) or not (0). The baseline category is broad as it involves PRRPs in opposition and PRRPs in informal/formal government.

To test the effects of inter-party cooperation (H2a) we use a dichotomous variable that captures whether a PRRP is cooperating with the government (1) or not (0). The baseline category ‘no cooperation’ refers to situations in which PRRPs are in parliamentary opposition regardless of whether they are subject to a cordon sanitaire. We expected that there would be opposing effects among mainstream parties’ supporters, depending on whether their party does or does not cooperate with a PRRP. Hence, we created a dummy variable distinguishing between respondents whose preferred party is cooperating with a PRRP (1) or not (0). In order to test H2a–b the two variables are interacted.

We then construct two dummy variables to test whether the degree of formality of a coalition with PRRPs matters. The first indicated whether there is an informal government cooperation (1) or not (0), and the second whether there is formal cooperation (1) or not (0). Also in this case, we do not take into account whether PRRPs are facing a cordon sanitaire when in opposition, and whether they are either informally or formally part of the government. As before, to test H3a–b, these two election level variables are respectively interacted with the variable that distinguishes between respondents whose preferred party is a PRRP coalition partner or opposition party.

The use of such broad baseline categories means that we cannot analyse the effects of inclusion and cordon sanitaire separately. The cordon sanitaire variable and the dichotomous cooperation or informal and formal cooperation variables are included within the same regression models. Using narrow baseline categories would have come at the cost of a reduced sample size,⁷ which we thought was not ideal considering that the total number of countries considered is only thirteen.

Our analysis features several control variables. We include a dummy variable at the election level to discriminate among those who voted for a winner party/presidential (1), and those who voted for losers (0). This allows us to isolate the role of inter-party cooperation on citizens’ antipathy towards PRRPs from possible purely ‘winner’ or ‘loser’ effects. Previous empirical studies show that whether one’s party (presumably) wins or loses an election impacts voters’ political behaviour, party identification, and attitudes (Huddy, Bankert, and Davies 2018; Ridge 2020). For example, a victory for the in-party could boost individuals’ party identification and reinforce it or, in the case of defeat, create a sense of threat. Consequently, out-parties may be evaluated more negatively, for example, in an attempt to protect one’s own identity and status.

Voters’ perceptions of a party and its credibility depend on how long it has been operating, and on its established reputation (de Vries and Hobolt 2012; Kitschelt 2018; Tavits 2008). PRRPs differ regarding their history and electoral success story. We then include a variable capturing whether a party is new to the political scene or has been visibly engaged for years, and one that measures the PRRPs’ vote share. The first variable is constructed by measuring the number of elections in which the PRRPs received at least one seat in the national parliament, or, for France, presented a presidential candidate (Wieringa and Meijers 2022). The latter considers PRRPs’ vote share, as stronger support should indicate a reduced PRRPs’ dislike (Dinas, Riera, and Roussias 2015; Harteveld, Mendoza, and Rooduijn 2022).

We include an individual level control that captures the weighted average like-dislike evaluation of respondents vis-à-vis the other mainstream parties (Reiljan and Ryan 2021). PRRP perception is likely to depend on how other out-parties are rated. Additionally, this in-bloc variable allows researchers to control for the possibility that respondents’ dislike of PRRPs may be influenced by how they use the dislike–like scale.

We also control for individuals’ degree of partisanship, their ideological distance from PRRPs, and satisfaction with the way democracy works (hereinafter SWD). In fact, previous research has shown that these three attitudes have an impact on individuals’ level of affective polarisation. A stronger identification with one’s party and a greater ideological distance from an out-party are associated with a stronger dislike of the latter (Dias and Lelkes 2021; Guedes-Neto 2022; Reiljan and Ryan 2021; Rogowski and Sutherland 2016). SWD is generally considered an indicator support for the political system as well as democratic principles (Dahlberg, Linde, and Holmberg 2015; Guedes-Neto 2022; Linde and Peters 2020). A PRRP’s anti-establishment positions often lead to criticism of the political system and questions of democratic principles. In general, people who have high levels of trust and SWD view PRRPs as threats, resulting in higher likelihood of showing negative affect towards them (Guedes-Neto 2022; Reiljan and Ryan 2021).

We compute the absolute distance between the left–right value from the PRRP and the respondents’ left–right value by using three variables: the respondents’ degree of party identification (closeness to party), the ideological distance between mainstream partisans and PRRPs (left–right self-placement, 11-point scale), and the socio-economic left–right value of PRRPs (11-point scale, from Parlgov). Previous studies have shown that the divide between citizens and PRRPs takes place more on cultural/transnational dimensions and less on the classical socio-economic left–right dimension (Reiljan 2020; Harteveld 2021). Unfortunately, the CSES does not include issue-specific items along all modules.

Descriptive statistics and relationships

Figure 1 shows the average likeability values (from 0 to 10) of mainstream partisans towards their in-party, in-bloc, and PRRP, by country and election year.⁸ The dashed lines indicate the overall mean values, with higher values representing more sympathy. Even though there are some differences across countries, mainstream supporters tend to rate their preferred party very positively, with an overall mean just above 8. In addition, when comparing the variation in average party likeability scores between and within parties to the variation in average PRRP and in-party scores, in-party evaluations turn out to be more consistent and stable.

Figure 1. Average in-party, in-bloc, and PRRP affective evaluations by country and election year.

Mainstream partisans display strikingly negative feelings towards PRRP, with an overall mean of 2.25 out of 10. It also emerges that mainstream partisans evaluate PRRPs differently than other out-parties. In most countries, other mainstream parties together receive an average likability of 4.3. But overall, PRRPs are rated on average 2.05 points more negatively than other mainstream parties together.⁹

Figure 2 shows again the average dislike-like values towards PRRPs by country and election year, along with information on whether 1) mainstream parties enforced a cordon sanitaire, 1) there was (in)formal cooperation with PRRPs, 3) the PRRPs were at the opposition. Figure 2 suggests that, apart from the 2019 election in Flanders, all country means are far below the overall mean in the case of a cordon sanitaire. With mean values all less than two, hostility is highest in such contexts.

Figure 2. Average dislike-like evaluations and mainstream party behaviour towards PRRPs by country and election year.

A relationship between affective evaluations and cooperation is less directly visible. However, it looks that overall, despite some exceptions, when PRRP government participation occurs country means are the highest and above the overall mean.¹⁰

Empirical findings

Table 2 presents the results of the mixed linear multi-level regression models. We test the effects of a cordon sanitaire and government cooperation with PRRPs on mainstream partisans’ affective attitude towards PRRPs. Model 0 shows the results of a basic regression model without the variables of interest, Model 1 the results when all variables are included, and Model 2 the results of the full regression model with a cross-level interaction term to test whether the effect of government cooperation differs between supporters whose preferred party is a PRRP coalition partner and those whose party is in opposition. All control variables show the expected direction and almost all are statistically significant. Results indicate that, as expected, the mainstream partisans disliking PRRPs the most are those very close to a party, more ideologically distant from PRRPs, fairly and very satisfied with the way democracy works, and with higher levels of education. In line with previous findings, all corresponding coefficients are negative and statistically significant (Guedes-Neto 2022; Reiljan and Ryan 2021; Rogowski and Sutherland 2016). Holding all other variables constant, mainstream supporters who feel very close to their preferred party differ significantly on average in their affective attitudes from those who do not feel very close to their preferred party. We also find that ideological differences are more central to affective evaluations of PRRPs than strong party identification. Both these findings are in line with the ones of Reiljan and Ryan (2021). The variable ‘In-bloc evaluation’, which measures the weighted dislike–like attitudes towards all other mainstream parties, is however counterintuitively never significant. The negative significant coefficient of ‘Winner effect’ in Models 1 and 2 suggests that those whose preferred party won elections dislike PRRPs more.

Consistently with previous findings, we find that higher support for PRRPs (measured by PRRPs’ vote share) increases mainstream partisans’ sympathy towards PRRPs (Dinas, Riera, and Roussias 2015; Harteveld, Mendoza, and Rooduijn 2022). The degree of establishment of a PRRP (measured by the number of elections in which PRRPs won seats in parliament) does not show a significant effect.

Consistently with previous findings (Guedes-Neto 2022; Renström, Bäck, and Carroll 2021), we find that that on average women, higher educated respondents, and those with higher income, are significantly more negative towards PRRPs than men and those with a low level of education and income, holding all other variables constant.

Our expectation was that mainstream partisans would become more hostile towards PRRPs when PRRPs are excluded by mainstream parties (H1). Models 1 and 2 in Table 2 show significant negative coefficients for the Cordon Sanitaire variable. Figure 3 plots the marginal effect of a cordon sanitaire based on Model 2 with 95% confidence interval. Where PRRPs face no cordon sanitaire the predicted average dislike–like score towards PRRPs is 2.41 among mainstream party supporters. When a cordon sanitaire is present the predicted average dislike–like value is 1.71. Thus, we can reject the null hypothesis.

Figure 3. Marginal effect of a cordon sanitaire: Predicted dislike vis-à-vis PRRPs among mainstream partisans.

We also expected that that when PRRPs participate in government dislike towards them decreases among partisans whose preferred party is a PRRP’s coalition partner (H2a), and it increases among partisans whose preferred party is in opposition (H2b). To test these hypotheses, we include in Model 2 a cross-level interaction term (favourite party is a coalition partner and government cooperation occurs). The coefficient is positive (0.86) and statistically significant. Figure 4 presents the marginal effects of the interaction term with 95% confidence intervals. Among those whose party is a coalition partner at a given election, the average predicted likability score for PRRPs while in opposition is 2.91 (out of a 0–10 dislike–like scale). This value increases to 3.71 when PRRPs move from opposition to cooperation. This means that holding all other variables constant, mainstream party supporters’ predicted dislike towards PRRPs decreases by 0.8 points, on average, when their preferred party cooperates with a PRRP. Among mainstream partisans supporting a party not cooperating with a PRRP, the predicted likability score for PRRPs is 2.5 when a PRRP is in opposition. When PRRPs cooperates, this value slightly increases to 2.7. Both effects are significant. Hence, as expected and suggested by previous studies (Bassan-Nygate and Weiss 2022; Gidron, Adams, and Horne 2019; Harteveld, Mendoza, and Rooduijn 2022), PRRPs gain some sympathy when involved in government from supporters of cooperating partners. Contrary to what we expected, we find that when a PRRP is in government the affect of the opposing mainstream partisans improves as well – although very slightly. Hence, these results provide support for H2a but not for H2b. We found a similar results for H2b when analysing Dutch case with the LISS data (see Online Appendix 1).

Figure 4. Marginal effects of inter-party cooperation: Predicted dislike vis-à-vis PRRPs among mainstream partisans.

We now test whether forming an either informal or formal inter-party matters (H3a–b). To test this assumption, we replicate the same linear multi-level regression models as in Table 2 (Models 1 and 2), but instead of the dummy cooperation variable we employ two dummy variables, one measuring informal and one formal government cooperation. We obtain nearly identical results for the Cordon Sanitaire and control variables are almost identical to those in Table 2 (see Table A5 in Online Appendix A).

Let’s now turn our attention to the two cross-level interaction terms we introduce: 1) ‘In-party is coalition partner’ and ‘Informal cooperation’, and 2) ‘In-party is coalition partner’ and ‘Formal government cooperation’. Both coefficients are positive as expected. Also, the coefficients indicate that the effect size larger for formal cooperation (0.903) than for informal cooperation (0.623). However, only the interaction effect for formal cooperation is statistically significant (p < 0.05). ¹¹ Thus, H3a–b are only partially supported.

Figure 5 plots the marginal effects of the interaction term including ‘Formal cooperation’ with 95% confidence intervals (for informal cooperation the coefficient is not significant). Among mainstream partisans whose preferred party is not a PRRP coalition partner, the average predicted likability score is 2.20 when PRRPs stay in opposition, and an average predicted score of 2.54 when PRRPs are formally part of a governing coalition. Among partisans whose preferred party is a PRRP coalition partner, the average predicted likeability value for PRRPs is 2.43 when PRRPs stay in oppositions, which raises to 4.02 when PRRPs formally participate in government. As for Figure 4, we can observe an increase in liking among both groups of mainstream partisans. Again, we could not replicate these results on the Dutch case using the LISS data (see Online Appendix 1). However, the effect of a formal cooperation has a much more sizable effect among the partisans supporting a party formally cooperating with a PRRP, and this result is robust. We can also observe that the increase for opposing mainstream partisans is similar in Figures 4 and 5, whilst for the supporters of the parties cooperating the increase is liking is much higher in case of formal cooperation.

Figure 5. Marginal effects of formal government cooperation: Predicted dislike vis-à-vis PRRPs among mainstream partisans.

Study 2: LISS

Variables

Respondents’ affective attitude towards PVV is available in each wave, operationalised as the classic 11-points party dislike–like scale. The dependent variable is thus derived from the scores that respondents gave to the PVV on this scale each year.

The period of the PVV’s informal participation in government is included in the 2010 and 2011 waves. When the data collection for the 2012 wave occurred, the cooperation had ended, and a new government was already in office. Therefore, to test the effect of informal government cooperation we introduce a dummy variable that takes the value 1 for these two panel waves, and also a dummy variable that takes the value 1 for respondents whose preferred party cooperates with the PVV, i.e. either the CDA or VVD.

As for our main analysis of CSES data, several control variables are included in the analysis namely:

• An in-bloc evaluation variable that measures the weighted average dislike-like evaluation of respondents towards the other mainstream parties;

• a dummy variable splitting respondents who voted for a party that ended up in government (1) to control for a ‘winner’ effect;

• a control for the electoral success of the PVV (as in Harteveld et al. 2021), operationalised by using the share of LISS respondents who declared that they would vote for the PVV if elections were held today;

• the PVV’s actual vote share in the parliamentary elections (robustness check).

• a proxy for strong partisanship by constructing a dichotomous variable that takes the value of 1 if respondents rated their in-party as 9 or 10 on the dislike-like scale;¹²

• respondents’ level of political interest.¹³ As previous studies have shown that political interest is strongly related to media attention, i.e., consumption of political information (Strömbäck and Shehata 2010) and that politically interested individuals tend to have a greater dislike of other parties than their own (Reiljan and Ryan 2021), we expect that respondents know how the mainstream parties interact with PRRPs;

• the ideological left–right self-placement on an 11-point scale (recoded it so that higher values indicate a more left-wing position and thus an ideologically more distant position from the right-wing PVV);

• attitudes towards European unification, as in whether should go a step further or has been already gone too far (on a 5-point scale, recoded so that higher values indicate that European unification should go further, indicating a greater ideological distance to the Eurosceptic PVV);

• a cultural dimension index including five items on immigration/integration (5-points scale, with higher scores implying greater ideological distance from the nativist PVV);¹⁴

• satisfaction with democracy (11 points scale);

• respondents’ institutional trust measured via the classical four-items index asking respondents how much they trust, the national parliament, politicians, the media, and the legal system (11-point scale);¹⁵

• socio-demographic variables (income, age, education, and gender).

Empirical analysis

Similarly to our CSES analysis, we predict mainstream partisans’ affective attitude towards the PVV via mainstream partisans’ affective responses. We take into account whether the PVV participated informally in government in any given survey wave, and whether respondents’ preferred party is a PVV coalition partner (H2a–b).

A total of 11095 unique respondents with non-missing observations participated in some or all of the 7 waves (on average they participated in 3.8 waves). Of these, 67% (7464) are mainstream partisans, and 79% of them had non-missing responses to the relevant items for at least three waves (for more details see Table D2, Online Appendix D).

As these data have a short panel structure with the same respondent placing the PVV on the dislike–like scale at different survey waves, we use a respondent fixed effects model. This enables us to estimate the effect of PVV’s informal government participation on each respondents’ attitude towards the PVV over time (i.e., the within-respondent effect), while holding the respondent-level time-invariant factors constant (Somer-Topcu, Tavits, and Baumann 2020). For our models we decided to use fixed effects (versus random) as the Hausmann test was significant.

Figure 6 shows the average affective evaluation of mainstream partisans towards the PVV, their preferred party, and towards the other remaining mainstream parties between 2007 and 2014 in the Netherlands. The dashed black line shows the trend in mainstream partisans’ liking for their preferred party, and higher scores mean more sympathy. On average, mainstream partisans rate their in-party between 7 and 8 on the dislike–like scale. Similarly to CSES results, the average evaluation of the in-party is fairly constant. The two reference lines on the x-axis mark the time of the PVV’s informal participation in government. For the period of informal cooperation, there is no clear change visible in terms of average likeability towards mainstream partisan’s in-party.

Figure 6. Average dislike–like evaluations of mainstream partisans towards the PVV, their preferred party and the other mainstream parties in The Netherlands between 2007 and 2014.

The black dotted line shows the trend of the weighted average dislike–like scores towards other mainstream parties. In the period under consideration, other mainstream parties receive an average value of between 4.77 and 5.32. Accordingly, the ratings fluctuate slightly around the theoretical neutrality point of 5. There seems to be no direct correlation between informal cooperation and in-bloc evaluation either. The in-bloc evaluation remains almost unchanged in the first year of informal cooperation. But in the second year they increase and reach their maximum average value, dropping again in the following year.

The solid black line shows the average trend in the affective evaluation of the PVV. In the first wave of the LISS survey (2007/2008), the average dislike of PVV is the highest (1.87). In 2010/2011 PVV is better liked (2.84). PVV’s likeability score among mainstream partisans has increased practically by one point since it began its informal participation in government.

In the second year of informal cooperation the average score slightly decreases to 2.6. And finally, after the end of the cooperation, at the end of 2012 and the beginning of 2013, the dislike increases again (2.03). There is relatively little variation in the average trend between the three survey waves prior to cooperation and the two waves following cooperation. Hence, the sharp decline in sympathy after the cooperation suggests that the informal cooperation with the PVV did not have a lasting and fundamental legitimising effect. The reasons for this are presumably that the cooperation ended in a dispute and that the cooperation was one-time and short-lived.

Figure 6 shows the average likability rating of the PVV by mainstream partisans, broken down by their individual preferred party in the background. The average ratings of the PVV by partisans of the CDA and the VVD, the two parties that cooperate with the PVV, are drawn with lines in different shades of blue, while the average ratings of supporters of the other parties are marked in different shades of gray. At first glance, it looks that among VVD and CDA supporters, the average sympathy for the PVV increases the most at the beginning of the cooperation and decreases the most after the cooperation. However, it is possible to observe a slight increase in sympathy also at the start of the cooperation for almost all other mainstream partisans, albeit much weaker. Furthermore, the VVD supporters rate the PVV most positively overall, whilst CDA supporters are on average less positive towards the PVV.¹⁶

The 2010/2011 LISS data values are comparable to the post-2010 election CSES values featured in our main analysis for the Netherlands (see Figure 1). The in-party average is almost identical, while the average likeability score for the PVV and for the in-bloc is around 0.5 points lower in the LISS data among mainstream partisans. However, larger discrepancies can be seen when comparing the 2006 post-election CSES averages regarding the PVV evaluation with those of the first wave of LISS (2007/2008). Even though there is a year between the survey periods, according to the 2006 CSES data, PVV is rated much more positively with an average score of 3 (1.13 points higher). A difference of more than one point on a scale of 0–10 is not necessarily small, especially given that the LISS data suggest that the variation in the average PVV rating among mainstream partisans appears to be fairly constant except for the period of cooperation. Figure 2 (obtained from CSES data) does not suggest, opposite to Figure 6 (obtained by LISS data) a positive relationship, as we hypothesised, between cooperation and sympathy towards the PVV. The difference could be due to the fact that the PVV was only founded in 2006 and has never faced a cordon sanitaire. With regard to the in-party and in-bloc ratings, no comparable discrepancy is discernible between the 2006 CSES data and the 2007/2008 LISS data.

Finally, it is important to note that Figure 6 shows no support for the hypothesis that cooperation reduces sympathy among supporters of mainstream parties outside the coalition (H2b). However, this is only a representation of trends in average values. For a more accurate assessment, the results of the panel regressions are presented next.

Results

Models 1 and 2 in Table 3 show the results of the individual-level fixed effects regression models (with and without interaction term), and Model 3 shows the results of the LDV model with interaction term, all with respondent-clustered standard errors. In both the FE and LDV models, most of the control variables are statistically significant (p < 0.05) and largely replicate our main analysis. The results suggest that mainstream partisans with more left-wing socio-economic and ideological positions in favour of multiculturalism are, on average, more negative towards the PVV than partisans with ideological positions closer to those of the PVV. In line with previous findings, the effect size for ideological distance is largest for the issue of multiculturalism, since this is the central issue the PVV mobilises on (Reiljan 2020; Harteveld 2021). The coefficient of ideological distance on the issue of EU integration is also negative, but not statistically significant.

The control variables ‘Interested in politics’ and ‘Strong partisanship’ do not show a statistically significant effect. Thus, as in the cross-national study, strong party identification does not appear to have a significant effect. This does not change when the alternative variable with the distinction ‘no’, ‘weak’, and ‘strong’ partisanship is used (see Table E1, Online Appendix E). However, both variables to account for the degree of partisanship here are only proxy variables.

While the CSES results suggest that supporters who are fairly or very satisfied with democracy in their country rate PRRPs more negatively on average, the results in Table 3 show an opposite effect. Mainstream partisans with higher levels of SWD are more positive towards the PVV on average, holding all other variables constant. However, only in the LDV model does the effect reach a statistical significance level. When using institutional trust instead of SWD, we get a positive effect that is statistically significant (p < 0.001) in both the FE and LDV models (see Table E2 in Online Appendix E). This result is not consistent with the Swedish case analysed by Reiljan and Ryan (2021), as they found that the effect of institutional trust was positive for the affective evaluation of other mainstream parties, but negative for the evaluation of the Swedish PRRP. One explanation for the positive effect in the Dutch case could be that, unlike Social Democrats in Sweden, the PVV was never subject to a cordon sanitaire, hence probably not (equally) perceived as a threat to democracy (Harteveld, Mendoza, and Rooduijn 2022).

Our results indicate that mainstream partisans who are more positive about political parties in general, are also on average more positive about the PVV. The coefficient for the ‘In-bloc’ variable is positive, statistically significant (p < 0.001) and, in contrast to CSES results, in a robust manner. Moreover, the results suggest that a party’s electoral success has a positive impact on its affective evaluation. The control variable for PVV’s electoral success is positively related to a change in PVV evaluation (p < 0.01). However, as shown in previous studies, not in a robust way (Harteveld, Mendoza, and Rooduijn 2022). Using the actual vote share of the PVV to measure support for the PVV does not show a statistically significant effect (see Table E3 in Online Appendix E). All in all, looking at the control variables provides confidence in the overall fit of the models.

So, was informal cooperation with the PVV a factor in mainstream party supporters’ evaluations of the PVV? Our results suggest that the answer is positive, as the coefficient of the interaction term is statistically significant (p < 0.001) and positive. The respective marginal effects of the FE Model 1 with 95% confidence intervals are presented in Figure 7. When the PVV is in opposition, supporters of parties collaborating with PVV show a predicted likeability value of 2.05, and partisans of other mainstream parties a value of 1.98 – really close values. However, when the informal cooperation takes place the PVV’s likeability value for CDA and VVD partisans increases up to 3.07, but for those supporting a party not participating in the coalition it is 2.08 – a very minimal increase.

Figure 7. Marginal effects of informal government cooperation: Predicted dislike towards the PVV.

The results in Table 3 mirror the findings of Harteveld, Mendoza, and Rooduijn (2022) and are comparable to our main findings. They provide strong evidence that mainstream party supporters’ dislike of PRRPs decreases when their preferred party builds a government coalition with PRRPs (H2a), and this is especially true given that this is an informal rather than formal cooperation.

It is important to notice that informal cooperation has not led to a stable reassessment of the PVV, even among CDA and VVD supporters. As the average trends in Figure 6 show, sympathy drops again after the end of cooperation. Interestingly, unlike in the CSES analysis, the results do not support the hypothesis that dislike increases among mainstream party supporters outside of cooperation (H2b). On the contrary, the results tend to suggest that there is no effect, or if anything, a small sympathy bonus. A possible explanation for this could again be that the PVV was never politically excluded. There was no direct break with a cordon sanitaire. The PVV’s participation in government could therefore be perceived by opposition party supporters as less as a violation of a social norm or as a real threat to their political preferences or to the political system. Robustness checks for this analysis are in Online Appendix E.

Conclusion

In this article we investigate whether mainstream parties’ strategies of inclusion versus exclusion of radical right-wing populist parties influences voters’ negative affect towards the latter. We test our hypotheses by using the CSES (cross-sectional, comparative) and LISS (panel, the Netherlands).

The first strategy we consider is the political exclusion of PRRPs by all mainstream parties, also called cordon sanitaire (Hjorth 2020; van Spanje 2010; van Spanje and van der Brug 2009). The second strategy is political cooperation between one or more mainstream parties and PRRPs in the form of a joint governing coalition. When inter-party cooperation occurs, we further distinguish between whether the governing coalition with PRRPs is a formal governing coalition cooperation or an informal one to support a minority government.

Consistent with previous research, we found that when a cordon sanitaire is established, mainstream party supporters will have higher dislike levels towards PRRPs This finding suggests that such isolation strategies may lead to increased dislike among mainstream party supporters, driven by mechanisms such as stigmatisation (Harteveld et al. 2019a; Harteveld et al. 2019b; Hjorth 2020), perception of PRRPs as threats (Brewer 1999; Dunwoody and Dennis 2019; Fiske, Cuddy, and Glick 2007; Huddy, Mason, and Aarøe 2015), and negative identification with PRRPs (Bankert 2021; Caruana, McGregor, and Stephenson 2015; Maggiotto and James Piereson 1977).

Cooperation between mainstream parties and PRRPs can have complex effects on political attitudes. One would expect that cooperation may lead to decreased dislike among supporters of PRRPs’ coalition partners, but it could also lead to increased dislike among supporters of mainstream opposition parties as the PRRPs’ influence is likely to lead to policy outcomes that are more distant from the views of opposition partisans (Fortunato et al. 2021). However, for supporters of mainstream opposition parties, we actually found an opposite (positive) effect, showing that a cooperation between a PRRP and mainstream parties improves affective evaluations towards PRRPs also among opposing party supporters, although this effect was very small. This suggests that cooperation tends to have a legitimisation effect on the PRRPs, supporting the argument that parliamentary representation of the radical right normalises radical right support (Valentim 2021). The unexpected positive effect that we found might be a signal of the fact that mainstream voters perceive that norms have shifted hence they perceive the radical right views as less stigmatised (Valentim 2021).

Regarding the further distinction between formal and informal cooperation, we found, as expected, that the two opposing effects of cooperation we just described are stronger in a formal compared to an informal cooperation, indicating that the structure and visibility of cooperation arrangements can shape their impact on political attitudes. However, robustness checks (see Online Appendix C) warn that some caution should be taken with respect to these results.

Finally, we find the same overall patterns also when analysing the Dutch case. Importantly, the Dutch data show that sympathy towards the PRRP dropped again after the end of cooperation, suggesting that the way in which mainstream parties signal acceptance of a PRRP has tangible consequences on voters’ perceptions.

This research has three main limitations. First, when using the CSES data we could not control for important variables such as respondents’ interest in politics, and ideological distance to PRRPs on nativism. This could however be done for the Dutch case thanks to the LISS data. Second, we could also not disentangle direction, that is, whether the affective attitudes of partisans were an effect or a cause of inter-party behaviour. Also in this case, the longitudinal nature of the LISS data provides some more insights. The third limitation concerns the way in which the dependent variable is measured. Clearly, when using a like–dislike scale to measure affective attitudes one needs to take into account that this may conflate with voter’s policy disagreements, as affective and ideological evaluations are often intertwined (Riera and Garmendia Madariaga 2023). Hence results need be interpreted with some caution particularly in contexts where policy disagreements might be a significant factor.

Consistently from previous results, we show that overall elite-level inter-party behaviour can fuel or reduce negative affect, depending on party strategies. This study shows that PRRPs’ pariah status is a powerful explanation of PRRP exceptionalism.

Acknowledgement

We would like to thank Giovanni D’Agostino who worked in this project as research assistant. We would also like to thank the two anonymous reviewers for their constructive comments.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Notes on contributors

Luana Russo

Luana Russo is Assistant Professor in Political Behaviour and Quantitative Methods at the Faculty of Art and Social Sciences, Maastricht University. Her research interests include political behaviour, affective polarisation, negative partisanship, and elections. [[email protected]]

Paula Schulze Brock

Paula Schulze Brock serves as an expert for the Joint Research Centre (JRC) of the European Commission. She holds a Research Master in European Studies (M.Sc.) and in Political Sciences (M.A.) from Maastricht University and University of Cologne. Her research interests include affective polarisation, right-wing populist parties, and electoral behaviour. [[email protected]]

Notes

1 In Sweden, following the Tidö Agreement in 2022, a minority government was installed with confidence and supply support from the Sweden Democrats. Coalitions in four municipalities date back to 2018.

2 In Belgium, the cordon sanitaire against VB was broken for the first time since 1989 in the last election in 2019. Conversely, in Finland, a cordon sanitaire against the True Finns (PS) emerged for the first time after the last election in 2019. Before that, the PS was even a governing party (see Table 1).

3 In Belgium there are two party systems, and, with the only exception of the Brussels region, voters cast votes only for parties represented in the respective regions (Coffé 2008: 179). As only Flanders has a far-right party only this region is included.

4 With regard to the representativeness of the LISS panel survey an underrepresentation of the elderly has been found, yet in comparison with other online access panels the underrepresentation is lower (Knoef and de Vos 2009: 20–21). For more details on the dataset please consult https://lissdata.nl/.

5 For France, the CSES data covers one parliamentary election in 2007, while data from 2002, 2012 and 2017 refer to the presidential elections. In case of presidential elections, we use respondents’ vote choice for the presidential candidate in the first round to identify mainstream supporters.

6 The wording of the question is: ‘I’d like to know what you think about each of our political parties. After I read the name of a political party, please rate it on a scale from 0 to 10, where 0 means you strongly dislike that party and 10 means that you strongly like that party.’

7 For example, countries and election years where PRRPs exclusively face a cordon sanitaire are not included in the regression models testing inter-party cooperation.

8 Figure 1 replicates the findings on ‘PRRP exceptionalism’ as in Gidron et al. (2019) and Harteveld et al. (2022).

9 There is one exception: Italy. See Tables B18–B55 in Online Appendix B to consult party-by-party dislike-like matrixes for each country and year for a more detailed overview of AP patterns.

10 Table B2–B17 in Online Appendix B provide summary statistics for all variables and separately for each country as well as for all variables when using actual voting behaviour to identify mainstream partisans or when excluding supporters of socialist/communist and right-wing parties. Inspection of the data showed that some mainstream partisans rated the PRRP more positively than their stated preferred party on the dislike–like scale.

11 However, Online Appendix C shows that the result concerning significance of formal and informal cooperation is not significant when 1) analyzing the effects of a cordon sanitaire, government cooperation with PRRPs, and (in)formal government cooperation in separate regression models (see Table C1), and 2) when using respondents’ vote choices for the current election to identify mainstream partisans (see Table C2).

12 Unfortunately, there is no direct measurement for individuals’ degree of partisanship in the LISS dataset. As an alternative measurement we also introduce an ordinal variable that distinguishes among none, weak, or strong partisanship. The ‘none’ category includes all respondents who rated their in-party 6 or less, the ‘weak’ category refers to party ratings of 7 or 8, and the ‘strong’ category again refers to ratings of 9 or 10 on the dislike–like scale.

13 Not available in the CSES.

14 1) It is good if society consists of people from different cultures; 2) Immigrants can retain their own culture; 3) There are too many people of foreign origin/descent in the Netherlands; 4) Legally residing foreigners should be entitled to the same social security as Dutch citizens; 5) It should be made easier to obtain asylum in the Netherlands. Prior to index creation, all items were recoded from 0 to 4 with higher scores indicating agreement with multiculturalism. The reliability of the composed variable is satisfied with Cronbach’s alpha in each wave ranging from 0.77 to 0.79 (Tavakol and Dennick 2011).

15 Previous research suggests that individuals’ institutional trust plays a role in their affective evaluation of parties. In Sweden, for example, supporters of mainstream parties with high institutional trust tend to be more positive toward other mainstream parties while they are significantly more negative toward the Swedish PRRP (Reiljan and Ryan 2021). The Cronbach ‘s alpha in each wave ranged from 0.75 to 0.82.

16 Figure D1, Online Appendix D shows the same as Figure 6 when mainstream partisans are identified based on their actual voting behavior. There are no notable differences between the two identification strategies.