1 Introduction

In recent years, several European mainstream partiesFootnote 1 have turned towards more nationalist policy positions, with the Österreichische Volkspartei (ÖVP) campaigning for tightening the national asylum law in the 2017 Austrian national election, the British Conservative Party leading the United Kingdom out of the European Union (EU), and the Danish Venstre (V) closing the borders to control the inflow of migrants, to name only a few examples. However, during the same time period, other mainstream parties have not changed their positions regarding issues such as European integration or migration, e.g., the Dutch Partij van de Arbeid (PvdA), the Italian Partito Democratico (PD), and the German Sozialdemokratische Partei Deutschlands (SPD).

What makes this observed variance in position changes particularly interesting is that issues such as migration and European integration are not related to the classic socioeconomic left–right dimension. Rather, these issues are embedded in a newly politicised cultural dimension in terms of Green/Alternative/Libertarian (GAL) vs. Traditional/Authoritarian/Nationalist (TAN) issues (see Hooghe et al. 2002). The observed movement of mainstream parties on this GAL–TAN dimension is puzzling because basic premises of Lipset and Rokkan’s (1967) cleavage theory suggest that mainstream parties are rather inflexible when it comes to new issue dimensions (see also Hooghe and Marks 2018). Because of this inflexibility, mainstream parties supposedly depoliticised the GAL–TAN dimension to retain the electoral competition to the socioeconomic dimension (Kriesi and Hutter 2017; Kriesi et al. 2006; Mair 2007).

In light of the observed changes in mainstream party policy positions on the GAL–TAN dimension, this article addresses the research question of why several European mainstream parties have changed their policy positions on the GAL–TAN dimension in recent years. To answer this question, I combine the cleavage theory with considerations on party responsiveness derived from Downsian models of spatial electoral competition (see Downs 1957). The argument I make is twofold. First, I argue that the between-country variation in the position changes of mainstream parties can in part be explained by differences in the success of right-wing populist parties (RWPPs). Second, I argue that the within-country variation is dependent on the ideological proximity of conservative mainstream parties to RWPPs on the GAL-TAN dimension.

In defining RWPPs, I follow the populism definition given by Mudde (2004, p. 543), according to which it is a thin ideology that considers society to be “separated into two homogeneous and antagonistic groups, ‘the pure people’ versus ‘the corrupt elite.’” The defining characteristic of right-wing populist parties is that they take strong conservative and nationalist positions and emphasise traditional cultural and moral values, national self-interest and identity, and authority (Rydgren 2005). This makes RWPPs a new party type because they combine the issues of old radical right-wing parties with new right-wing GAL–TAN positions and a populist rhetoric that centres around the alleged failure of “the elites” to address the policy preference of “the people” on the GAL–TAN dimension (Zaslove 2008). In many countries this is a novel policy position, which has been proven beneficial for RWPPs and malicious for mainstream parties.

Altlhough I focus on RWPPs and their impact on the policy positions of mainstream parties on the GAL–TAN dimension, important insights can already be drawn from previous contributions that have addressed how parties react to radical right-wing parties. A contagious effect of rivalling radical right-wing parties has already been shown by van Spanje (2010) and Abou-Chadi and Krause (2020), who found that mainstream parties react to these parties on migration and European integration issues. Focusing on the influence of Eurosceptic parties, Meijers (2017) as well as Wagner and Meyer (2017) confirm this development. Shifting the focus to Eastern and Central European countries, Heinisch et al. (2021), however, do not confirm this trend. Thus, the rich literature on radical right-wing parties points to a responsiveness of Western European mainstream parties towards radical right-wing parties on single issues.

However, studies dealing with mainstream party responsiveness towards rivalling right-wing populist parties on the GAL–TAN dimension in general are still scarce. The existing studies focus on whether mainstream parties follow the populist anti-elite rhetoric (e.g., Rooduijn et al. 2014), only address single cases (e.g., Arzheimer and Berning 2019) and single issues such as migration (e.g., Pardos-Prado 2015) or other policy dimensions (Krause and Giebler 2020), or do not account for specific party characteristics (e.g., Guiso et al. 2017; Schumacher and van Kersbergen 2016). One of the few comprehensive analyses comes from Mudde (2013), who found no systematic impact of RWPPs on mainstream party GAL–TAN policy positions—a finding he ascribes to the modest success of RWPPs in the late 20th century.

Given the recent electoral success and the observed policy position changes of mainstream parties, I build on the vast field of research contributions on radical right-wing parties and address a still remaining gap in the literature. This is done in three ways, by (1) analysing right-wing populist parties, (2) using the electoral success in terms of vote share, and (3) focusing on the position changes of mainstream parties on the GAL–TAN dimension. In doing so, I contribute to the debate on the impact of RWPPs by drawing on arguments from spatial models of electoral competition and providing evidence that first-and-foremost conservative mainstream parties are responsive to the electoral success of RWPPs on the GAL–TAN dimension.

The paper is organised as follows: I first elaborate on the GAL–TAN dimension and describe why mainstream parties sought to depoliticise it. Subsequently, I discuss the recent electoral success of RWPPs and argue why and how mainstream parties have changed their policy positions on the GAL–TAN dimension. In the empirical section, I estimate econometric models concerned with the responsiveness of 25 mainstream parties on the GAL–TAN dimension towards the electoral success of RWPPs in 11 European countries between 2002 and 2019.

2 The GAL–TAN Dimension in the European Policy Space

The European policy space can be conceptualised as two-dimensional, with a socioeconomic and a cultural dimension facing each other orthogonally whilst simultaneously being interrelated to some degree (see, e.g., Hooghe et al. 2002; König et al. 2017). Whereas the socioeconomic left–right dimension has been rather stable in structuring electoral competition and the formation of parties over the 20th century (Mattila and Raunio 2006), the cultural dimension has been highly flexible over time, with issues such as European integration and migration becoming incorporated into this dimension and making the conflict over transnational issues part of the cultural dimension often termed the GAL–TAN dimension (Hooghe and Marks 2018; de Vries 2018).

In this regard, it is important to note that issues like European integration and migration are not exclusively linked to the GAL–TAN dimension. Rather, these issues can also be linked to the socioeconomic dimension (see Armingeon et al. 2015; Hooghe et al. 2002; Hobolt 2016). This complex relationship of the two dimensions is underscored by the fact that (socio)economic crises can be framed as cultural conflicts and vice versa (see Rodrik 2018). This is, however, less of a problem for the analysis at hand, as I am not interested in the determinants of voters’ policy preferences but rather the policy positions of parties. Keeping this in mind, I follow previous contributions in locating EU and migration issues primarily on the GAL–TAN dimension (see, e.g., Kriesi et al. 2006; Marks and Wilson 2000).

Although the GAL-TAN dimension is similar to the classic socioeconomic dimension, empirical studies have shown that mainstream parties are rather unresponsive towards voters and rivalling parties on the GAL–TAN dimension (see, e.g., Mattila and Raunio 2012; van der Eijk and Franklin 2004). Following a basic premise of the cleavage theory, previous contributions argued that mainstream parties are programmatically inflexible regarding new issue dimensions, as mainstream parties are internally divided about these issues and face the threat of upsetting voters and members who, while sharing the party’s policy position on the socioeconomic dimension, might object to its position on the GAL–TAN dimension (Hooghe and Marks 2018; Hobolt et al. 2009).

However, as described above, several European mainstream parties have changed their policy positions on the GAL–TAN dimension in recent years. To further illustrate the anecdotal evidence cited in the introduction of this paper, Fig. 1 shows a box plot of the average changes of mainstream parties’ positions on the GAL–TAN dimension in absolute terms.

Fig. 1
figure 1

Absolute change in policy positions of mainstream parties: 2002–2019. The figure shows the average change in the policy position of mainstream parties on the socioeconomic left–right (0–10) and the GAL–TAN (0–10) dimensions between 2002 and 2019 in the 11 sample countries. (Data taken from the Chapel Hill Expert Survey Trend File 1999–2014 [Bakker et al. 2015; Polk et al. 2017], as well as the Chapel Hill Expert Survey 2019 [Bakker et al. 2020] (CHES))

Full size image

I use absolute terms to avoid positive and negative changes cancelling each other out. Between 2002 and 2006, mainstream parties on average changed their position on the GAL–TAN dimension (0–10) to a much lesser degree than on the socioeconomic dimension, with an average change of 0.06 points. Looking at the period between 2006 and 2019, this number nearly doubles, with average changes on the GAL–TAN dimension of 0.11 points. From the figure it follows that the median value of policy position changes also increased between 2006 and 2014, with a slight decrease between 2014 and 2019. Moreover, between 2006 and 2014 the policy position changes on the GAL–TAN dimension were greater than on the socioeconomic dimension. The interquartile range of changes on the GAL–TAN dimension also increased between 2006 and 2019. The figure further shows that there are some outliers in the sample.

These findings underpin the observation of increased movement on the GAL–TAN dimension made in the introduction, and they raise the question of how these position changes of mainstream parties on the new GAL–TAN dimension—which, according to the cleavage theory, should hardly exist—can be explained. In this paper, I argue that these changes can be explained through spatial models of electoral competition and the responsiveness of mainstream parties to the electoral success of RWPPs, as well as the ideological proximity to these parties.

3 Mainstream Parties and the Other “Populist Zeitgeist”

In the introduction of this paper, I stated that parties termed “populist” are seemingly on the rise in Europe. But, while being used vividly, “populism” is a rather elusive term that has been defined in various ways (see Hawkins 2009). In this paper, I follow Cas Mudde (2004, p. 543) and define populism as a thin ideology that considers society to be “separated into two homogeneous and antagonistic groups, ‘the pure people’ versus ‘the corrupt elite’, and which argues that politics should be an expression of the volonté générale (general will) of the people.”

In his seminal contribution, Mudde (2004) has proclaimed the emergence of a “populist zeitgeist” by identifying traits of populist anti-elite rhetoric in the political discourse of mainstream parties. He argues that various mainstream opposition parties have challenged governing parties by using an anti-elite rhetoric (see also Mair 2007). In testing these arguments, Rooduijn et al. (2014) found that mainstream parties have not become more populist when faced with successful populist challengers. Although I agree with their finding, I argue that this is only half the picture since RWPPs are not only characterised by their anti-elite rhetoric but also by specific policy positions on the GAL–TAN dimension. Thus, the conclusion that mainstream parties are not chasing the “populist zeitgeist” might be premature, as the “populist zeitgeist” in Europe is not only coined by populism as a thin ideology but also by the host ideology of RWPPs. From this it follows that while mainstream parties may not be responsive towards the populist rhetoric, they might nonetheless chase the other “populist zeitgeist,” which consists of the specific policy positions of European RWPPs on the GAL–TAN dimension.

By definition, right-wing populist parties take strong conservative and nationalist policy positions and emphasise traditional culture, national self-interest and identity, and authority, which are allegedly all under threat by a supposed group of “outsiders” and “the elite” (see Mudde and Kaltwasser 2013; Rydgren 2005; Taggart 2000). Right-wing populist parties combine these positions with newer topics such as climate protection, same-sex marriage, and the protection of a cultural status quo that is supposedly undermined by “the elite” (see Lockwood 2018; Pelinka 2013; Pappas et al. 2009). Thus, they position themselves against cosmopolitan values as well as multiculturalism and demand a return to traditional structures, making the GAL–TAN dimension their primary dimension of electoral competition (see Inglehart and Norris 2016).

Nonetheless, this is not to say that all RWPPs put the same weight on each individual GAL–TAN dimension issue. The salience of single issues in the political discourses of parties and the public can show considerable variation over time (see Ademmer et al. 2018; Braun and Schmitt 2020). But this variance makes it all the more necessary to move beyond the analysis of single issues and focus on the coherent policy positions taken by RWPPs.

Taking a strong coherent position on the GAL–TAN dimension is what distinguishes RWPPs from the old radical right-wing parties (see Kriesi 2010; Zaslove 2008). Unlike these parties, RWPPs move beyond the ownership of single—mostly migration—issues and challenge mainstream parties on the GAL–TAN dimension.Footnote 2 This links RWPPs to specific policy positions and gives merit to shifting the analysis from the single-issue competition of radical right-wing parties to analysis of mainstream party responsiveness towards RWPPs on the GAL–TAN dimension in general. Irrespective of this, it is important to point out that populism and radical right-wing positions are strongly intertwined in Europe. This makes it difficult to isolate the individual effects on mainstream parties and increases the need for a theoretically founded combined consideration of populism and right-wing GAL–TAN positions.

Concerned with the policy positions of radical right-wing parties regarding single issues, Abou-Chadi and Krause (2020), Abou-Chadi (2016), Wagner and Meyer (2017), Han (2015), and van Spanje (2010) already found mainstream parties to be responsive to radical right-wing parties on migration and European integration issues. However, none of these studies assessed the impact of rivalling right-wing populist parties’ success on the positions of mainstream parties on the GAL–TAN dimension.Footnote 3 This is, however, necessary for assessing the unique electoral challenge that RWPPs pose. A notable exception is provided by Guiso et al. (2017), who found that the electoral success of populist parties is correlated with a reduction in distance between populist and nonpopulist parties regarding European integration and immigration issues.

Although making valid arguments and assumptions, Guiso et al. (2017) do not distinguish between left-wing and right-wing populism and the issues salient in the respective party discourses. In the context of electoral competition this distinction is important, as spatial proximity of parties is a crucial factor in the analysis of electoral competition (see Williams and Whitten 2015). In the following, I address this research gap by arguing why and how mainstream parties respond to the success of RWPPs on the GAL–TAN dimension and how this responsiveness is affected by ideological proximity to RWPPs.

But first it is necessary to discuss the importance of looking at right-wing populist parties. One could argue that it is sufficient to analyse radical right-wing parties, as the electoral competition is more about specific policy positions. However, the communication of these positions is a key component in the electoral success of RWPPs. As argued above, RWPPs do not address single issues, but take coherent policy positions on the GAL–TAN dimension. The depoliticisation of the GAL–TAN dimension has made mainstream parties less responsive towards (changing) preferences of the electorate, which has created a potential for its politicisation by populist parties that specifically address the failure of “the elites” to represent the “will of the people.”

The uniqueness of the RWPP strategy can best be exemplified by looking at the issue of EU integration, which is frequently addressed by European RWPPs. Right-wing populist parties address EU integration from a policy position point of view by arguing that it undercuts national sovereignty and leads to the loss of cultural identity. Both of these issues are directly related to the GAL–TAN dimension. By advancing EU integration, RWPPs argue that “the elites” allegedly ignore the policy preferences of the electorate, i.e., the will of the people. Rejecting the EU is further closely connected to the populist rhetoric in itself because the EU can easily be framed as a project of the elites and the establishment that is profitable only to them and not to the “ordinary” people.

Because I argue that populism in itself plays an important role in challenging mainstream parties on the GAL–TAN dimension, it is important to discuss whether the concept of populism is really necessary to explain the mechanism at work. As argued by Rooduijn (2019), the main challenges to current populism research are conceptual overstretching and the detaching of populism research from adjacent fields. I argue that both challenges are met by the mechanism I propose.

First, I attribute the success of RWPPs and the possible response of mainstream parties to this success to the populist rhetoric that specifically picks up “the elite’s” lack of responsiveness to the “will of the people” on the GAL–TAN dimension. Second, I combine the analysis of populism with the adjacent field of electoral competition by arguing that recent policy position changes of mainstream parties on the GAL–TAN dimension can be explained by not only taking into account the policy positions of challenger parties but also by accounting for the rhetoric by which these positions are communicated in the electoral strategy.

Taking coherent right-wing policy positions with a populist rhetoric in combination with the lack of responsiveness of mainstream parties on the GAL–TAN dimension has arguably contributed to the electoral success of RWPPs in recent years. In order to examine this electoral success of RWPPs, Fig. 2 shows the average vote share of RWPPs and mainstream parties between 1960 and 2019 in the 11 European sample countries. The figure shows that between 2002 and 2019, RWPPs have, on average, increased their vote share by 8 percentage points, while mainstream parties have lost around 5 percentage points.

Fig. 2
figure 2

Vote share of right-wing populist parties: 1960–2019. The figure shows the average vote share of right-wing populist parties (solid black line) and the average vote share of mainstream parties (dashed line) in the 11 sample countries between 1960 and 2019, both with 95% confidence intervals. The averages were plotted by using moving averages and fractional-polynomial prediction plots. (Data taken from the ParlGov dataset [Döring and Manow 2019])

Full size image

Spatial models of electoral competition suggest that mainstream parties will overcome their inflexibility and change their positions if rivalling parties become increasingly successful (see, e.g., Budge 1994; Harmel and Janda 1994; Meguid 2005). Against this backdrop, I argue that the recent success of RWPPs will trigger a response by mainstream parties on the GAL–TAN dimension. As the previous section has shown, mainstream parties indeed display increased movement on the GAL–TAN dimension. The question, however, is whether this is the case because mainstream parties position themselves in opposition to the policy positions of rivalling RWPPs or adapt a similar position (see Meguid 2005).

I argue that, for two reasons, mainstream parties are more likely to choose an accommodative strategy against RWPPs. First, the electoral success of RWPPs is built on their issue ownership on the GAL–TAN dimension (Arzheimer and Berning 2019). Hence, not engaging them on the GAL–TAN issues will not decrease their support. Second, as mainstream parties will be concerned with gaining back voters they lost to RWPPs, they will, in line with spatial theory, challenge these rival parties directly (van Spanje and de Graaf 2018). Other contributions have shown that this strategy can prevent niche party success on single issues (Pardos-Prado 2015) or at least impact the socioeconomic policy positions (see Krause and Giebler 2020).

With a focus on recent European elections, Downes and Loveless (2018) were able to show that some nonincumbent mainstream parties electorally outperformed RWPPs by emphasising immigration issues but only in response to the electoral threat posed by RWPPs. This is in line with the consideration that mainstream parties might adapt the policy positions of RWPPs to regain or maintain their vote share when an RWPP becomes successful (see Mudde 2013, 2007). Thus, if an RWPP becomes successful, mainstream parties are expected to shift their policy positions closer towards this party’s position. Based on these considerations, I derived the following hypothesis:

Hypothesis 1

The greater the electoral success of an RWPP, the closer a mainstream party will move towards this party’s position on the GAL–TAN dimension.

As argued above, it is likely that specific party characteristics and the location of mainstream parties in the policy space will affect the responsiveness towards the electoral success of RWPPs. Previous contributions have shown that ideological proximity plays an important role in the responsiveness of parties towards rival parties (see, e.g., Williams and Whitten 2015). This is the case because a party’s vote share predominantly depends on the policy position of the rival party closest to its own position (see Adams 2001). To investigate the proximity in detail, Table 1 shows the average policy positions of mainstream parties by party family and pits them against their distance to rivalling RWPPs on the GAL–TAN dimension.

Table 1 GAL–TAN policy positions of mainstream parties: 2002–2019. (Data taken from the Chapel Hill Expert Survey Trend-File 1999–2014 [Bakker et al. 2015; Polk et al. 2017] and the Chapel Hill Expert Survey 2019 [Bakker et al. 2020] (CHES))
Full size table

Two main findings can be derived from the table. First, the table shows that conservative mainstream parties are—on average—positioned closer to the TAN pole than any other mainstream party. Second, the average policy position distance of conservative mainstream parties to rivalling RWPPs is, with 1.84 points, much closer than the average position distance of liberal or socialist mainstream parties to RWPPs. In fact, in all sample countries with a conservative mainstream party, this conservative party is the closest party to the rivalling RWPP.Footnote 4 Thus, in line with the arguments on spatial electoral competition from Meguid (2005), RWPPs will especially compete with conservative parties in the electoral competition. This leads to an increased responsiveness of conservative mainstream parties (see Mudde 2013). Based on this, I hypothesised the following:

Hypothesis 2

Conservative mainstream parties are more responsive towards the electoral success of RWPPs than mainstream parties of other party families.

In sum, I argue that the observed position changes of mainstream parties on the GAL–TAN dimension can be explained by the electoral success of RWPPs. Based on arguments derived from the spatial theory of electoral competition, I expect mainstream parties to reduce the distance to RWPPs once these parties become successful. Conservative mainstream parties are thereby expected to be especially responsive due to their ideological proximity.

4 Operationalisation and Data

To test the hypotheses, I used a time-series, cross-sectional dataset that includes national mainstream parties between 2002 and 2019 from 11 European sample countries that have at least one RWPP.Footnote 5 This restriction to Western European countries is necessary because the shape of the cultural GAL–TAN dimension arguably differs between Western and Eastern Europe, inducing possible biases (see de Vries 2013). The period of analysis was chosen because of data limitations. The unit of analysis is the change in a mainstream party’s policy position distance to a rivalling RWPP between the respective Chapel Hill Expert Survey (CHES) years.

The employed identification strategy of RWPPs was twofold. First, I based the identification of populist parties on the PopuList by Rooduijn et al. (2019). Second, I used the CHES to identify the GAL–TAN policy position. A populist party is coded as right-wing if its average position on the GAL–TAN dimension is one standard deviation above the average GAL–TAN position of other parties in the respective country. I also cross-checked this coding with the far-right coding by Rooduijn et al. (2019), showing that not all RWPPs in my sample were also radical right-wing parties. This underscores the previous argument that there is a difference in analysing radical right-wing parties and right-wing populist parties. Furthermore, all but three RWPPs identified are also coded as Eurosceptic parties by Rooduijn et al. (2019). Based on this, 26 RWPPs were identified (see online appendix, Table A 1).Footnote 6 Note that not all identified RWPPs have competed in all elections.

The vote share of the RWPPs was taken from the CHES. I calculated the change by subtracting the vote share provided by a previous CHES from the one in the subsequent CHES. The vote share in the CHES always refers to the last national election. If an RWPP competes for the first time, the previous vote share is 0. If no election took place between two CHES waves, the change is also 0. The variables were distributed normally (see online appendix, Fig. A 1).

The coding of mainstream parties was based on the data provided by the CHES (1999–2019) and the ParlGov dataset by Döring and Manow (2019). A party was coded as mainstream if it belongs to either the socialist, conservative, or liberal party family and is frequently the majority party in a government coalition or on average wins a considerable number of votes. The latter was defined as an average vote share of one standard deviation above the country average. Based on this, 25 mainstream parties were identified (see online appendix, Table A 2).Footnote 7 All identified mainstream parties competed in all the elections.

The change in policy-position distance between mainstream parties and RWPPs on the GAL–TAN was measured using the CHES. Using the change in position distance is important because the question is whether changes in the electoral success of RWPPs are correlated with a reduction in the position distance between RWPPs and mainstream parties. Since conservative mainstream parties are positioned closer to RWPPs, only this approach makes it possible to go beyond the mere correlation between party family and policy position distance to RWPPs.

I used the CHES position coding because the GAL–TAN dimension in the CHES is congruent to my conceptualization of the GAL–TAN dimension (see Polk et al. 2017). Other sources do not provide such an indexed GAL–TAN policy position of parties. Although a position scale could be constructed (see Wagner and Meyer 2017), I followed Meijers (2017) and based the coding on the CHES data. I deliberately did not calculate the position changes of mainstream parties in relation to the extreme ends of the scale because that was not in line with models of spatial competition (see Adams 2001). The distance was calculated by subtracting the position of a given mainstream party from that of the rivalling RWPP and calculating the change in distance between t − 1 and t. Positive values indicate that the distance between a mainstream party and an RWPP increases, while negative values indicate that the distance is decreasing. In the cases with more than one RWPP, I used the mean position of the RWPPs to calculate the distance to the RWPP platform. The values were distributed normally (see online appendix, Fig. A 2).

Theoretically, the value of this variable can also become negative when an RWPP moves closer to the mainstream parties’ policy positions. Hence, it is important to note that, while theoretically possible, RWPPs on average do not move closer to mainstream parties but, if anything, on average shift 0.08 points towards the TAN pole of the GAL–TAN dimension.

As control variables, I included the change in the mean policy preference of voters regarding migration and EU integration (see Adams and Somer-Topcu 2009). The preferences were based on the European Social Survey (ESS) data (ESS 2018, 2014, 2010, 2006, 2002).Footnote 8 In the course of the robustness checks I also combined both variables into one measure to approach a singular measure of the GAL–TAN policy preference change of the electorate. In the few cases of missing data, I used linear interpolation. Additionally, I also controlled for the influence of mainstream party vote loss as well as government participation by mainstream parties and RWPPs (see Somer-Topcu 2009). I accounted for the previous vote loss with a variable that gives the vote loss of a mainstream party between t − 1 and t. Furthermore, I included a variable that controls for the first time an RWPP competes in an election. Finally, I controlled for the position shifts of nonmainstream, nonpopulist parties by including the mean policy position shift of these other parties between t − 1 and t.

Because the dependent and independent variable showed a normal distribution, I ran ordinary least squares regressions to assess the correlation between a change in the RWPP vote share and the change in the policy-position distance between an RWPP and a mainstream party. To test the hypotheses, I modelled an interaction term that addressed the expectation that conservative mainstream parties would show higher responsiveness to the electoral success of RWPPs than other mainstream parties.

A common problem in panel data that often leads to misspecification is autocorrelation. This, however, would be an issue only if one would use the absolute policy-position distance on the GAL–TAN dimension at t and not the change between t and t − 1. I employed robust standard errors to account for heteroscedasticity (see White 1980). My baseline model is defined by the following:

$$\Updelta D_{c,t}^{m}=\Updelta VS_{c,t}^{p}\cdot \text{ConsM}_{c,t}+\Updelta VS_{c,t}^{p}+\text{ConsM}_{c,t}+\beta X_{j,t}+\mu _{m}+\gamma _{c}+\lambda _{t}+\varepsilon ,$$

where \(\Updelta D_{c,t}^{m}\) is the change in distance (\(\Updelta D\)) between a mainstream party (m) and an RWPP in a country (c) between t − 1 and t (t). The interaction term of the independent variable is given by \(\Updelta VS_{c,t}^{p}\cdot \text{ConsM}_{c,t}\), with \(\Updelta VS\) being the change in the RWPP (p) vote share (VS) in interaction with conservative mainstream parties (\(\textit{ConsM}\)) responding to this change between t − 1 and t (t) in a country (c). With \(X_{j,t}\) a vector of the described control variables (j) is included. Furthermore, terms denoting mainstream party (\(\mu _{m}\)), country (\(\gamma _{c}\)), and time (\(\lambda _{t}\)) fixed effects were incorporated.

5 Results

The regression analyses in Table 2 focus on the main variables and the interaction term. The control variables are introduced in a second step in Table 3 after the effect of the interaction term is addressed. Model 1 reports the coefficient of the RWPP vote-share change variable without any fixed effects. The coefficient of the variable is, at −0.024, negative but not statistically significant. Models 2–4 also include only the RWPP vote-share change variable but with different combinations of year, party, and country fixed effects. The change in fixed effects does not change the size or nonsignificance of the coefficient.

Table 2 Mainstream party responsiveness to right-wing populist parties with and without interaction term
Full size table
Table 3 Mainstream party responsiveness to right-wing populist parties and control variables
Full size table

The interaction term for conservative mainstream parties is introduced in model 5. The term is, at −0.088, negative and statistically significant. In models 6–8, year, party, and country fixed effects are introduced. The size and statistical significance of the interaction term’s coefficient remain stable across these models. The RWPP vote-share change variable and the dummy variable for conservative mainstream parties remain nonsignificant. From this it follows that the correlation between the RWPP vote-share increase and the decrease in distance to these RWPPs on the GAL–TAN is statistically significant only for conservative mainstream parties. For nonconservative mainstream parties, the coefficient is very small and nonsignificant.

To address the substantial relevance of the estimated coefficients, I calculated the effect size, ceteris paribus, of a 1–standard deviation (6.9 percentage points) increase in an RWPP’s vote share on the distance reduction of a conservative and nonconservative mainstream party. The marginal effects were calculated based on model 8. Using model 8 for the calculations, if the vote share of an RWPP increases by 1 standard deviation, the distance between a conservative mainstream party and an RWPP is reduced by 0.5 standard deviation (0.59 points) on the GAL–TAN dimension. Compared to this, a 1–standard deviation increase in the RWPP vote share is only correlated with a 0.05–standard deviation (0.06 points) decrease in the nonconservative mainstream party distance. This result for conservative mainstream parties is quite substantial because mainstream parties, on average, only changed their position on the GAL–TAN dimension between 2002 and 2019 by 0.10 points, with a standard deviation of 1.18.

To illustrate this finding further, Fig. 3 plots the marginal effects for both groups based on model 7 and model 8. The figure shows the estimated marginal effects between the 1st and 99th percentiles of the distribution of the independent variable. The figure shows that an increase in the RWPP vote share is not correlated with a decrease in the distance between mainstream parties and RWPPs on the GAL–TAN dimension in general, but it is strongly correlated with a decrease in distance between conservative mainstream parties and RWPPs. In sum, these findings do not back the first hypothesis but strongly support the second hypothesis.

Fig. 3
figure 3

Marginal effects: mainstream party responsiveness to right-wing populist parties. (Calculation based on the regression analysis displayed in Table 2)

Full size image

Since the previous regression analyses focused only on the main variables and the interaction term for conservative mainstream parties in isolation, the second set of regression analyses introduced several control variables. The purpose of these additional regression models was to analyse whether the small and nonsignificant effect found for other mainstream parties and the strong and substantial effect found for conservative mainstream parties is robust against different control variables. The results are displayed in Table 3. Note that the number of observations in the analysis is reduced by eight due to missing policy preferences of the electorate.

While all models use year fixed effects, the models switch between party and country fixed effects. Using year and party fixed effects, model 1 introduces the two variables that control for the policy preference changes of the electorate as well as for the policy position shifts of other parties. Model 2 uses the same control variables but changes from party to country fixed effects. The coefficient of the RWPP vote-share change variable is small and nonsignificant in both models, whereas the interaction term for conservative mainstream parties is, at −0.086 and −0.083, statistically significant. Both results support the previous findings.

Models 3 and 4 include the variable that measures vote loss of mainstream parties as well as the variable that controls for government participation. Including both variables does not change statistical (non)significance of the RWPP vote-share change variable or the interaction term but slightly increases the coefficient of the interaction term to −0.087 and −0.085, respectively. In the next two models, the significance of the coefficients is also not affected by the variables that control for government participation of an RWPP or for the first time an RWPP competes in an election. However, the coefficient of the interaction term is again slightly increased to −0.092 and −0.088, respectively. Model 7 and model 8 keep the previous control variables and reintroduce the variables that control for government participation and vote losses by mainstream parties, again marginally increasing the size of the interaction term’s coefficient to −0.093 and −0.090, respectively.

Overall, the coefficient of the RWPP vote-share change variable as well as the interaction term for conservative mainstream parties proved to be robust in size and statistically significant across the models. This finding strongly supports the previous results.

To interpret the coefficients in substantial terms, I again calculated the effect size, ceteris paribus, of a 1–standard deviation (6.9 percentage points) increase in an RWPP’s vote share on the distance between the RWPP and mainstream as well as conservative mainstream parties. I again calculated the marginal effects based on model 8. If the vote share of an RWPP increases by 1 standard deviation, the distance between the conservative mainstream party and the RWPP is reduced by 0.5 standard deviation (0.62 points). Reconsidering that the average distance between RWPPs and conservative mainstream parties on the GAL–TAN dimension is 1.84 points, a 1–standard deviation increase in the RWPP vote share correlates with a distance reduction between an RWPP and a conservative mainstream party of about 35%. In comparison, a 1–standard deviation (6.9 percentage points) increase in an RWPP’s vote share is again only correlated with a distance reduction of 0.05 standard deviation (0.06 points). The weak correlation in the case of mainstream parties in general is again not statistically significant.

To further illustrate the correlation, Fig. 4 plots the marginal effects for both groups based on model 7 and model 8. The figure again shows that an increase in the RWPP vote share is not correlated with a decrease in the distance between mainstream parties and an RWPP on the GAL–TAN dimension in general, but it is strongly correlated with a decrease in distance between conservative mainstream parties and RWPPs.

Fig. 4
figure 4

Marginal effects: mainstream party responsiveness to right-wing populist parties with control variables. (Calculation based on the regression analysis displayed in Table 3)

Full size image

In sum, the findings strongly support the second hypothesis, according to which conservative mainstream parties are expected to be especially responsive to the vote-share increase of RWPPs. Additionally, the lack of robust statistical significance in the case of mainstream parties in general leads me to conclude that the same correlation cannot be assumed for this group of parties. Although the first hypothesis on the expected correlation for mainstream parties in general is not support by the analysis, the second hypothesis on the increased responsiveness of conservative mainstream parties is strongly supported.

6 Robustness of Results

Since the dependent and independent variable are measured in the same time period, the analysis cannot rule out reversed causality. One could argue that the distance reduction by mainstream parties draws attention to RWPPs and thereby increases their vote share. However, this would mean that mainstream parties change their positions without any reason. The only pressure that could trigger this behaviour is changes in the preference of voters. But, as I showed, these changes are not correlated with the dependent or independent variable in the analysis. From this it follows that there is no theoretical reason why mainstream parties should accommodate the positions of RWPPs prior to their electoral success (see Meguid 2005).

Further, it is possible that RWPPs move towards the TAN pole while conservative parties move in the same direction, leading to a distance change of 0. In this scenario, the conservative mainstream party nonetheless accommodated the RWPPs’ policy position. Although theoretically possible, this works against the employed hypotheses tests by—if anything—underestimating the strength of the correlation between vote-share increase and distance reduction.

To empirically rule out the possibility that RWPPs increase their vote share if mainstream parties close the distance to them, I calculated the correlation between mainstream party position distance and RWPP vote share (see online appendix, Table A 3). Because of possible issues of autocorrelation, I additionally calculated Prais–Winsten regressions (Prais and Winsten 1954). The analysis shows that there is no statistically significant correlation between the policy position distance of mainstream parties to RWPPs and the electoral success of RWPPs. Hence, the theoretical arguments and empirical results support the direction of causality I propose.

To further assess the robustness of the results, additional regression models were calculated (see online appendix, Table A 4). Model 1 uses fixed-effects regression controlling for both individual parties and countries in the model. The second and third models are also fixed-effects regression models, with model 2 clustering the standard errors by parties and model 3 clustering the standard errors by countries. The last fixed-effects regression is employed in model 4. The model is additionally jack-knifed.

The remaining four models use different operationalizations of variables and exclude a critical case. Model 5 is not based on coding conservative mainstream parties but rather on the mainstream party closest to the respective RWPP. The model shows a negative and statistically significant coefficient of the interaction term. The size of the coefficient is slightly smaller than in the case of conservative mainstream parties. Hence, my results on party responsiveness are not driven by coding conservative mainstream parties as the closest competitor.

Model 6 uses a different operationalization of the voters’ policy preference change by combining the EU integration and migration preferences into a single measure. The same control variable is also included in model 7, with additional control variables. The last model excludes Italy because the coding of some parties in Italy is problematic, with the FI/PdI party being an RWPP mainstream party, no additional conservative or liberal mainstream party competing in elections, and the M5S party sometimes being coded as RWPP. In sum, the coefficients of the main variables of interest are robust against the different model specifications in every case.

7 Conclusion

In this paper I addressed the research question of why several European mainstream parties changed their policy positions on the GAL–TAN dimension in recent years. In answering this question, I argued that mainstream parties changed their positions on the GAL–TAN dimension in response to the electoral success of RWPPs. This success pressures mainstream parties into closing in on the positions of rivalling RWPPs. Due to the ideological proximity, I argued that first-and-foremost conservative mainstream parties will close the distance to RWPPs. Supported by the empirical evidence, the answer to the research question is that the between-country variation can be explained by the variance in RWPP success, while the within-country variation is explained by the almost exclusive responsiveness of conservative mainstream parties.

One limitation of this finding, however, is the narrowness of the empirical analysis due to the short period of analysis. This bears the risk of missing long-term dynamics and developments, since mainstream parties might only move back to a previous position at the TAN pole from which the party had moved away in previous years. Finally, it is also important to stress that the empirical support of my theoretical arguments stems purely from correlation. The statistical analysis employed is not a causal identification strategy.

Having this in mind, the paper provides several connection points for future research. Because I only focused on RWPPs and the GAL–TAN dimension, the findings only hold in this context. Mainstream parties might react differently to left-wing populist parties or to RWPPs on the socioeconomic dimension (see Schumacher and van Kersbergen 2016). Because of the restriction of the analysis to Western European countries, the external validity of the findings holds only for these countries. Thus, a separate analysis is necessary to encompass the specific patterns of the GAL–TAN dimension in Eastern Europe. Such an analysis could be conducted by using the Manifesto Project dataset (see Volkens et al. 2020), with which the geographic and temporal scope of the analysis can be extended. It is also necessary to further investigate the arguments put forward regarding issue ownership of RWPPs on the GAL–TAN dimension. It is also crucial to assess whether chasing the “populist zeitgeist” is an effective measure against RWPPs. So far, this strategy has produced mixed results and is highly dependent on specific factors (see Meijers and Williams 2020; van Spanje and de Graaf 2018; van Spanje and van der Brug 2009). Thus, the strategy of chasing the “populist zeitgeist” might eventually result only in legitimising the nationalist, illiberal, and xenophobic policy positions of RWPPs (see Meguid 2005).

Despite the remaining questions and discussed limitations, I was able to show that first-and-foremost conservative mainstream parties seemingly chase the other “populist zeitgeist” by shifting their policy positions on the GAL–TAN dimension in response to the electoral success of RWPPs. While the question of whether mainstream parties can win the electoral race by choosing such a strategy has yet to be answered, bearing current developments in mind, one can conclude that chasing the “populist zeitgeist” might not only lead to losing the race against RWPPs but also induce the risk of losing the marathon, which is the defence of social cohesion and liberal democracy in general.