The populist radical right:
game changers?
On the impact of the populist radical right on
party system polarization in Western Europe.
Research Master’s Social Sciences Thesis Author: David van Dosselaar (10069232) Contact: davidvandosselaar@hotmail.com
16-6-2017
Thesis supervisor: Prof. Dr. S.L. de Lange Second reader: Prof. Dr. W. van der Brug
[2] Abstract
This paper examines the influence of the populist radical right on the polarization of party systems. This way, it charts to what extent these parties are responsible for the change of Western European party systems from moderate pluralist into polarized pluralist. It expects to find that the electoral success and right-wing policy positioning of populist radical right parties increase party system polarization on the socio-cultural policy dimension. Polarization is measured with the Esteban-Ray Index and examined using the large cross-sectional time series data of the Comparative Manifesto Project. After estimating several OLS models with panel corrected standard errors, it is discovered that an increase in the vote share and a more right-wing positioning of populist radical right parties have a polarizing effect on party systems. Yet, these effects are weak. Therefore, the conclusion is drawn that the populist radical right did not have a fundamental impact on party systems as they are not responsible for a transformation of party systems from moderate pluralist into polarized pluralist.
Introduction
In most Western European countries, the electoral support for populist radical right (PRR) parties is on the rise (Mudde, 2007; Mudde, 2013; Hainmueller & Hopkins, 2014; Akkerman et al., 2016; Mudde, 2016). Many PRR parties have even regularly crossed the ten per cent mark since the turn of the millennium. Hence, these parties are not a ‘pathology’ (Mudde, 2010), or ‘pariah’ (de Lange, 2012) but a lasting electoral threat for mainstream parties. Though many scholars have investigated the causes for the increasing success of the PRR, few have studied the consequences of this success for society and politics (but see Wolinetz & Zaslove, 2017). Especially understudied is the impact of PRR parties on party systems. Since the PRR has seen its vote share rise the most in the recent decades (Mudde, 2013; Akkerman et al., 2016) compared to for example the radical left, the present study will exclusively focus on the PRR.1
A particular concern of the press is the polarizing impact of the PRR on the party system or even on society as a whole. For example, several Dutch newspapers have written that stronger parties on the flanks have contributed to a weakening of the political centre (e.g. Trouw, 2012; Parool, 2016). A journalist in Trouw has even exclaimed that “The Dutch political landscape is […] starting to look like the Grand Canyon” (2012; author’s translation). Still, research on the effect of successful PRR parties on party system polarization is scarce. Therefore this paper aims to fill this gap by answering the following question: what is the impact of the populist radical right on party system polarization in Western Europe? Ultimately, it will put Mudde’s conclusion to the test, who has stated, based on a holistic investigation, that “irrespective of conceptualization and operationalization, populist radical right parties have not fundamentally changed party systems in Western Europe” (2014: 217).
The present paper will improve upon the literature on polarization in the Western European context in four ways.2 First, there is ambiguity concerning the causal direction of the relationship between party system polarization and the success of the PRR (Pelizzo & Babones, 2007). Most of the research on party system polarization fits in the research tradition that uses polarization as an independent variable. Spies and Franzmann (2011) for instance claim that polarization of certain issues could increase the vote share of PRR parties. On the contrary, Abedi (2002) could not find any evidence that party system polarization increases the likelihood of PRR party success. Hence,
1 A noticeable exception is Syriza in Greece, which has attracted a large number of new voters after the financial crisis. 2 Traditionally, most of the literature on polarization is concentrated on the United States and the increasing gap between
[3]
conclusive evidence regarding the relationship between PRR parties and party system polarization is not offered. Therefore, this paper will study possible causes of polarization related to the emergence of PRR parties, instead of its consequences.
Second, most studies on the political impact of the PRR are only indirectly related to polarization, as they are focussed on the ideological responses of mainstream parties due to the electoral growth of PRR parties. More specifically, most of these studies focus on the verrechtsing (right turn) of mainstream right parties because of PRR party success (Bale, 2003; Meguid, 2008; Van Spanje, 2010; Han, 2015; Abou-Chadi, 2016; Wagner & Meyer, 2016). A few other recent publications study the mainstream left’s response to PRR parties (Bale et al., 2010; Van Spanje, 2010; Yilmaz, 2012; Han, 2015). Still, this (re)positioning of mainstream parties is only one possible element of party system polarization. An important contribution of the present paper is thus to expand the literature by taking the direct effect of PRR party electoral success and policy positioning on party system polarization into account.
Third, most of the literature on Western European party system polarization in relation to the PRR exists of case studies. Recently, some case studies have been published on the effect of the PRR on the polarization of the Swiss (e.g. Afonso & Papadopoulos, 2015; Bornschier, 2015; Kriesi, 2015) and the Dutch (e.g. Oosterwaal & Torenvlied, 2010; Pellikaan et al., 2016) party system. The findings of these publications have hinted at the polarizing impact of the PRR. In a very recent case study, Silva (2017) has even found that PRR parties in the Netherlands have increased the level of polarization within the Dutch population. To improve the generalizability of the findings and to tease out the causal mechanism implied by these case studies, this paper aims to offer the first systematic cross-national analysis of Western European party system polarization in light of the increased success of the PRR.
Fourth, many publications on the relationship between the PRR and polarization do not take time into account. Time is an essential factor when investigating party system polarization, as polarization is a time (or period) dependent phenomenon (Shamir, 1984). Eventually, ignoring the temporal aspect could result in biased findings. The aforementioned publication by Spies and Franzmann (2011) for example is based on cross-sectional data, which thus lacks the more thorough investigation of an analysis over a longer period of time. Some of the case studies on the effect of the PRR on polarization are an improvement over this latter article, as they examine a short period of time (e.g. Kriesi, 2015; Pellikaan et al., 2016). Still, these case studies mostly use descriptive evidence and are thus not able to predict polarization. By employing time series, the present study will try to show that success of PRR parties drives polarization of the party system, regardless of its time dependency.
Furthermore, it is crucial to study the polarization of party systems, as it could influence the functioning of democracy (e.g. Layman et al., 2006; Enyedi, 2016). To name just a few consequences of polarization, a highly polarized party system could lead to unstable coalitions or inability to reach consensus (Indridason, 2011; Dodd, 2015; Kriesi, 2015; Pellikaan et al., 2016), to politically weary and distrusting citizens (King, 1997; Shea, 2003; Kim et al., 2010) and to more cases of argumentum ad hominems inside and outside of parliament, instead of ideologically fused debates (Uslaner, 2000; Sinclair, 2002). Some studies suggest that the PRR plays an important role herein. They show that mainstream right (and sometimes other) parties copy PRR ideology after PRR parties have become electorally relevant (Van Spanje, 2010; Kallis, 2013; Han, 2015; Abou-Chadi, 2016; Wagner & Meyer, 2016). Especially if the mainstream right and mainstream left move in opposite directions because of
[4]
PRR parties, then the political centre would be less occupied. Eventually, this could result in a weakening of party democracy.
This paper will start with a conceptualization of polarization. It will then continue with discussing why party systems could theoretically polarize because of increased success of PRR parties. It is expected that PRR parties have a direct and indirect impact on party system polarization, but only on the socio-cultural dimension. It is hypothesized that their vote share and policy positioning on this dimension influence party system polarization directly. Indirectly, it is expected that mainstream parties adjust their policy positioning in response to PRR parties. This eventually results in more party system polarization, which will be measured with the Esteban and Ray-index. To establish the impact of the PRR on party system polarization, several pooled time series cross-section models with ordinary least squares (panel corrected standard errors) will be estimated. The results show that PRR parties only have a weak impact on party system polarization. Therefore, the conclusion is drawn that the PRR did fundamentally change party systems in Western Europe.
The relationship between party system polarization and success of the PRR
The concept of polarization is used in at least two ways in the political science literature. It either refers to party or political polarization (supply-side) or voter polarization (demand-side) (Pardos-Prado & Dinas, 2010). It is necessary to distinguish between these two types of polarization, as previous research from the American tradition has shown that political polarization causes voters to polarize (Huber & Malhotra, 2015; Mason, 2015; Mummolo & Nall, 2017). At the same time, some European case studies have found that polarization of parties does not necessarily correlate with polarization of voters (Pardos-Prado & Dinas, 2010; Oosterwaal & Torenvlied, 2010). Therefore, as political polarization could drive social polarization or occur without it, it is pertinent to focus on political polarization first. Hereafter, this paper will refer to this type of polarization when using the term.
Surprisingly however, many scholars have operationalized polarization (e.g. Lachat, 2008; Bornschier, 2015), but few have conceptualized this concept. The classical definition that does exist, which is used in the present paper, is by Giovanni Sartori (1976). He defines polarization as an ideological distance, or more precisely as “the overall spread of the ideological spectrum of any given polity” (1976: 126). Similar definitions have been adopted by Dalton (2008), Pardos-Prado and Dinas (2010) and Han (2015). In this definition, polarization is thus seen as an aggregated set of ideological relations between political parties (Sartori, 1976; Dalton, 2008; Schmitt, 2016). If the sum of the ideological differences increase between the parties in the party system increase, then polarization increases (Sartori, 1976; Schmitt, 2016).
Sartori’s definition is used in the present study for three reasons. First, Sartori (1976) implies that polarization is more than a mere ideological dispersion between parties. Polarization does not simply entail a distribution along policy dimensions, but rather “a type of political conflict that spans across issue domains, where the sum is more important than the parts” (Vegetti, 2014: 233). When polarization is high, “cleavages are likely to be very deep, […] consensus is surely low, and […] legitimacy of the political system is widely questioned” (Sartori, 1976: 135). It thus characterizes the ideological interactions between the parties within a party system. Some of these interactions are especially important for the present study, as the ideological responses of mainstream parties to the PRR’s success are some of the hypothesized drivers of polarization. Second, the definition of Sartori is applicable to multi-party systems (Sartori, 1982), but is not necessarily positively or linearly correlated with the number of parties in a party system (Pelizzo & Babones, 2007). According to Sartori (1982), a party system with three relevant parties could be more polarized than a party
[5]
system with ten parties as long as the sum of the differences between those parties is larger. This ensures the possibility to compare between different kinds of party systems. Third, Sartori (1976) argues that polarization is also determined by the electoral strength of parties on the flanks such as PRR parties. Other conceptualizations of polarization (e.g. Taylor & Herman, 1971; Gross & Sigelman, 1984) fail to acknowledge the importance of flank parties, as they conceptualize polarization as the average distance of parties to the ideological centre of a policy dimension. If polarization is a function of the distance of parties to the centre, then there is a direct link between polarization and the number of parties in a system. If there are fewer parties in a party system, it is more likely that these parties converge to the political centre (Downs, 1957). Again, this makes it more difficult to compare between party systems. Using Sartori’s definition solves this issue and incorporates the importance of flank parties such as PRR parties.
Often party systems are classified based on the way in which parties compete in the ideological political space (Mair, 1997). Sartori (1976) identifies two archetypes of party systems: the moderate pluralist and the polarized pluralist variants. In the former, political competition between parties is centripetal (i.e. directed towards the centre) and largely based on cooperation, resulting in ideological convergence (Sartori, 1976; Mair, 2008; Pellikaan et al., 2016). In the latter, competition is centrifugal (i.e. directed towards the flanks) and largely based on conflict (Sartori, 1976; Mair, 2008; Pellikaan et al., 2016). Hence, “party system change occurs when a party system is transformed from one class or type of party system into another” (Mair, 1997: 51-52). Therefore, it can be argued that PRR parties have had an impact on party systems if these systems have transformed from moderate pluralist towards polarized pluralist (Mair, 1997; Pellikaan et al., 2016).
In theory, the PRR could influence party system polarization in four ways. They could have a direct impact on polarization because of their size or policy position and indirectly through their influence on the size or policy position of mainstream parties. Mainstream parties could also have a direct impact on party system polarization regardless of PRR party success. Because of their electoral success, the PRR has facilitated the emergence of a new ideological line of conflict on socio-cultural issues such as immigration on which ideological competition has ensued (e.g. Kriesi et al., 2008). Moreover, with their radical positioning on this dimension, PRR parties are expected to reinforce centrifugal political competition (e.g. Warwick, 2009). Indirectly, it is the (re)positioning of mainstream parties compared to the PRR that is expected to increase the overall ideological spread of party systems (Abou-Chadi, 2016).3 What is more, as a third step in this process, a more right-wing repositioning of the mainstream right should push the PRR further to the flank. Hence, a vicious circle could occur in which PRR party success first directly increases polarization, followed by an indirect effect on mainstream parties. Polarization could then be strengthened by the effect of the mainstream right on PRR parties, who are pushed further to the flanks again.
PRR party size
There are some noticeable case studies that (implicitly) address the first driver of party system polarization: the size of PRR parties. In Western European context, especially studies on Switzerland (Afonso & Papadopoulos, 2015; Bornschier, 2015) and on the Netherlands (Pennings & Keman, 2008;
3 Mainstream party size will not be addressed any further in this study. It is not expected that their size will have an impact
on party system polarization whatsoever, nor on the (re)positioning of PRR parties. Mainstream parties as operationalized here (see methods section) are usually large parties anyway, so their ‘threat’ to other mainstream adversaries and to PRR parties is always present. Differences in their size should therefore not have a major impact. This has been confirmed by empirical tests.
[6]
Oosterwaal & Torenvlied, 2010; Pellikaan et al., 2016) have been published. These studies find that the PRR has an impact on politicization of certain issues or polarization of party systems as a whole. The large increase in size of the Swiss People’s Party (SVP), for example, accelerated the emergence of a new line of political conflict on socio-cultural issues (Bornschier, 2015). In congruence with the bipolar division of Swiss citizens, mainstream right and PRR parties took a right-wing position on this new dimension, while the mainstream and radical left took a left-wing position (Afonso & Papadopoulos, 2015; Bornschier, 2015).
Similarly, successful PRR newcomers such as the Freedom Party (PVV) in the Netherlands had an impact on Dutch politics, as inter-party relationships became more adversarial over time on the socio-cultural dimension (Achterberg, 2006; Pennings & Keman, 2008; Oosterwaal & Torenvlied, 2010). In their study of the fragmentation and polarization of the Dutch party system, Pellikaan and colleagues (2016) conclude that parties located near the centre of the political space diverged under pressure of PRR parties such as the PVV and its predecessor, the List Pim Fortuyn (LPF). What is more, the competition in the Dutch party system changed from centripetal to centrifugal at first, but eventually became even more unstructured (Pellikaan et al., 2016).
There is also some evidence from Central-Eastern Europe (Enyedi, 2016; Taggart & Kaltwasser, 2016). Case studies in this context confirm the previously described tendencies. In Hungary, the populism of Victor Orbán’s Hungarian Civic Alliance party has amplified the already emerging polarization of the party system (Enyedi, 2016). This is confirmed by Taggart and Kaltwasser (2016: 358), who state that “the possibility of cooperation at the domestic level diminishes when populist actors have the strongest electoral position within the government coalition”. Of course, PRR parties have not become the largest in Western Europe yet. Nevertheless their success in Eastern Europe is an indication that size matters for polarization. Therefore, the first hypothesis that this paper will test is:
H1: The higher the vote share of populist radical right parties, the higher the level of polarization of the party system on the socio-cultural dimension.
PRR positioning
At least in Western Europe, the upward trend of PRR success has gone hand in hand with the emergence of the socio-cultural ideological dimension, which now coexists with the traditional economic dimension. Hence, since a few decades, there is spatial competition not only over economic issues (the ‘classic’ Downsian left versus right) but also over socio-cultural issues (Warwick, 2002; Green-Pedersen, 2007; Kriesi et al., 2008; Spies & Franzmann, 2011).4 On this new policy dimension, especially immigration and integration issues are important determinants of parties’ positioning (Benoit & Laver, 2006; Pellikaan et al., 2007; Van der Brug & Van Spanje, 2009). PRR parties almost exclusively compete over socio-cultural issues on which they take a radical right-wing position (e.g. Mudde, 2004; Mudde, 2007; Akkerman et al., 2016). When the present study speaks of ‘taking a more right-wing position’ on the socio-cultural dimension, it means that parties take a more ‘TAN’ (traditional-authoritarian-nationalistic) position. Inversely, more ‘left-wing’ is more ‘GAL’ (green-alternative-libertarian). On the economic dimension, however, PRR parties take many different positions. More specifically, their economic profile ranges from ‘welfare chauvinistic’ (De
4 The socio-cultural dimension is also known as GAL-TAN dimension, or green-alternative-libertarian versus
traditional-authoritarian-nationalistic (e.g. Marks et al., 2006). Other terms that have seen use include the liberal-authoritarian dimension (e.g. Wagner & Meyer, 2016) or simply the cultural dimension (e.g. Pellikaan et al., 2016).
[7]
Koster et al., 2013; Schumacher & Van Kersbergen, 2016) to centrist (De Lange, 2007, Mudde, 2007) and to neoliberal (Rovny, 2013).5 Still, PRR parties do not have the credibility to compete on economic issues that mainstream parties have, which means that they mainly focus on socio-cultural issues (Green-Pedersen, 2007; Meguid, 2008; Wagner & Meyer, 2016).
Therefore, the second (direct) driver of polarization could be the positioning of PRR parties on the socio-cultural dimension. Sartori (1976) already emphasized the importance of flank parties for the polarization of party systems, as they are the main drivers of the direction of interparty competition (Esteban & Ray, 1994; Pelizzo & Babones, 2007; Dalton, 2008; Pellikaan et al., 2016). In the literature, there are two arguments that explain the influence of the positioning of the PRR on the level of polarization. First, PRR parties belong to a group of parties called anti-establishment parties, whose aim it is “to sustain centrifugal competition, [which] increases the polarization of the system, that is, the overall ideological distance” (Capoccia, 2002: 16, emphasis added; Schedler, 1996). Already in the early 1990s it has been noted that the PRR accuses the “established political parties […] of having constructed, to the detriment of the average citizen, an all-encompassing system sustained by interventionism, clientelism, and favouritism” (Betz, 1993: 418). This system is what the PRR acts against, taking a radical position on the socio-cultural dimension. Thus, by advocating a radical position, the PRR feeds on the rise “of a profound and diffuse disaffection and disenchantment with the established political parties throughout Western Europe” (Betz, 1993: 419) among the citizens.
Second, in the coalition democracies of Western Europe, parties near the extremes of any ideological spectrum tend to showcase ‘relative extremism’ (Warwick, 2004). This means that they often place themselves ideologically further away from the centre than their voters do. Regarding the strategic positioning of parties on the flanks, two theories exist. First, there are scholars (e.g. Holmberg, 1989; Iversen, 1994) who argue that this positioning is a (semi-)conscious effort of party leaders to “stake out somewhat different positions in hopes of moulding or influencing [their voters]” (Warwick, 2009: 277). Second, Kedar (2005) states that voters anticipate potential outcomes of elections by voting more extreme than their actual policy position, in order to balance out the political forces on their opposite side. Eventually, this should lead to a higher level of polarization, as the distance between the PRR and all other parties increases. Based on the previous discussions, the following hypothesis can be formulated:
H2: The more right-wing the position populist radical right parties take, the higher the level of polarization of the party system on the socio-cultural dimension.
Mainstream parties’ response
The second mechanism that will be addressed is the indirect effect of PRR parties on party system polarization by means of their influence on mainstream parties. It is possible to at least explore this indirect effect by first looking at the impact of mainstream party positioning on party system polarization, and then at the impact of PRR party electoral success on mainstream party positioning. Especially when the ideological response of the mainstream right is the opposite of the response of the mainstream left because of the electoral success of PRR parties, then this success could generate centrifugal forces. Mainstream parties could naturally also have a standalone impact on polarization,
5 Welfare chauvinism is “the view that access to welfare should be restricted to the ‘deserving’ natives (Schumacher & Van
[8]
for example because of the ideological competition between each other.6 Furthermore, PRR parties might have a larger impact on their direct competitors, the mainstream right (Meguid, 2008). This means that the relative contribution towards polarization of the mainstream right would be larger than of the mainstream left, since their positional change is larger. If the mainstream left takes a more left-wing position and the mainstream right a more right-wing position, then both contribute equally to polarization.
Because of their electoral success, PRR parties have generally provoked ideological responses of mainstream parties on the socio-cultural dimension and much less on the economic dimension (Meguid, 2008; Mudde, 2014; Han, 2015; Abou-Chadi, 2016, Wagner & Meyer, 2016). PRR parties are often considered issue-owners on the socio-cultural dimension, which induces ideological competition of their mainstream adversaries (Meguid, 2008; Bale et al., 2010; Neundorf & Adams, 2016). Moreover, It has been demonstrated extensively that PRR parties ‘steal’ voters from all mainstream parties (e.g. Meguid, 2005; Bale et al., 2010; Han, 2015; Abou-Chadi, 2016), for example because “mainstream parties […] have effectively lost touch with ‘the man on the street’” (Odmalm & Hepburn, 2017: 3). Mainstream parties were at least initially unable to connect with the emerging ‘politics of resentment’ (Betz, 1993) and have lost some of their appeal to the working class (Oesch, 2008). Therefore, mainstream parties need new strategies to deal with those electoral losses, which include changing their ideological positions, addressing new issues or depoliticizing socio-cultural issues (Harmel & Janda, 1994; Pellikaan et al., 2016).7 These strategies have been investigated at the party family level (e.g. Bale et al., 2010; Van Spanje, 2010; Abou-Chadi, 2016), but research on the party system level is scarce.
On the one hand, concerning the response of the mainstream right, some form of agreement exists. In the literature, it is often mentioned that PRR parties were able to fill the gap on the right flank because of mainstream party convergence (Kitschelt & McGann, 1997; Mair, 2008; Abou-Chadi, 2016). However, following the increasingly strong competition of PRR parties from the flank, mainstream right parties took a more right-wing position again (Norris, 2005; Meguid, 2008; Abou-Chadi, 2016). Many empirical studies have shown that this verrechtsing of the mainstream right has indeed occurred on the socio-cultural dimension (Bale, 2003; Meguid, 2008; Han, 2015; Wagner & Meyer, 2016) or on the immigration issue specifically (Van Spanje, 2010; Abou-Chadi, 2016). Since some mainstream right parties could become credible issue-owners (or at least alternatives) on typical PRR issues, sometimes at the expense of PRR electoral performance (Meguid, 2008), this has become a viable strategy.
On the other hand, the response of the mainstream left to PRR parties on the socio-cultural dimension has been rather diffuse. Sometimes they move closer to the PRR, but other times they
6 Yet, concerning the spatial competition between mainstream parties, there are theories related to their positional
behaviour which state that these parties tend to move towards the ideological centre (Schofield & Sened, 2006; Mair, 2008; Warwick, 2009). This stems from the assumption that mainstream parties benefit from the relatively voter-rich ideological political centre (Schofield & Sened, 2006). As credible policymakers, moving towards the ideological centre would offer mainstream parties the possibility to gain as many votes as possible (Schofield & Sened, 2006). These converging tendencies would thus decrease polarization of the party system. This has indeed been visible before PRR parties became electorally relevant (Webb, 2005; Van der Brug & Fennema, 2007; Mair, 2008).
7 This so-called issue salience competition, however, is only indirectly linked to polarization as defined by this paper. Issues
generally important to PRR parties tend to be highly polarized and politicized (e.g. European integration, immigration), as opposed to valence issues such as the environment or prosperity (Abou-Chadi, 2016). These polarized issues will possibly attract more ideological competition, ergo possibly more polarization of the party system as a result. Valence issues, on the other hand, do not provoke ideological conflict. When positional issues important to PRR parties attract more attention of voters, it is more likely that similar parties will move towards the issue-owner (Meguid, 2008; Wagner & Meyer, 2016).
[9]
diverge away from them (Bale, 2010; Immerzeel et al., 2016). In spatial theory, it could be a strategically good choice to move slightly more to the political middle, as the gap widens between them and the mainstream right if the electoral success of the PRR increases (Schofield & Sened, 2006; Meguid, 2008; Warwick, 2009). However, it could also be strategically advantageous if the mainstream left stays left of the centre to really distinguish themselves from the right-wing socio-cultural ideological profile of both the mainstream right and the PRR. Still, a certain reluctance of the mainstream left to move on the socio-cultural dimension has been established in the literature.
Therefore, based on the expected verrechtsing of the mainstream right and reluctance of the mainstream left to move on the socio-cultural dimension due to PRR party success, it is expected that the relative contribution of the mainstream right towards party system polarization on the socio-cultural dimension is larger than of the mainstream left. This contribution is further enlarged because the mainstream right is expected to have a direct impact on PRR party positioning on the socio-cultural dimension (see hypothesis 4 further on).
Inversely, mainstream parties’ response on the economic dimension to the electoral surge of PRR parties has been mixed. In a handful of instances, they copy the welfare chauvinistic stances (Schumacher & Van Kersbergen, 2016) of PRR parties. In most instances, however, mainstream parties only compete with each other on the economic dimension (Webb, 2005; Kriesi et al., 2008; Mair, 2008; Rovny & Edwards, 2012). In the Western European two-dimensional party systems, many PRR parties have even deliberately remained vague on their position on the economic dimension (Rovny, 2013). This implies that a mainstream response to the PRR has not been necessary on this dimension. Still, if mainstream right parties would take a more right-wing position it would increase polarization, as the ideological differences with the mainstream left and in some cases the PRR would increase. In the end, the following hypotheses can be formulated:
H3a: The more right-wing the position mainstream right parties take, the higher polarization of the party system on the socio-cultural dimension and on the economic dimension.
H3b: The more left-wing the position mainstream left parties take, the higher polarization of the party system on the socio-cultural dimension and on the economic dimension.
H3c: The relative contribution of mainstream right parties to the polarization of the party system on the socio-cultural dimension is higher than the contribution of mainstream left parties because they react more strongly to an increase in PRR vote share than the mainstream left.
PRR party repositioning due to the mainstream right
Since the literature shows that mainstream (right) parties (re)position themselves on the socio-cultural dimension after an increase in vote share of PRR parties, it is likely that PRR parties want to somehow (re-)establish their issue-ownership and keep their unique position on this dimension. Being the issue-owner is important for PRR parties, as they are mainly vote-seeking (e.g. Akkerman et al., 2016). If PRR parties are not able to offer a clear policy alternative to voters anymore, they often decrease in size (Meguid, 2008). However, no systematic research exists on the radicalization of PRR parties as a response to mainstream parties’ spatial positioning nor on its effect on party system polarization. Still, some empirical evidence has been found that PRR parties take a more extreme position on socio-cultural issues in some circumstances such as a cordon sanitaire (Van Spanje & Van
[10]
Figure 1: Relationships between party support, positioning (on the socio-cultural dimension) and party system polarization. A solid line shows a direct relationship. A dashed line denotes an indirect effect on
polarization. A dotted line means that it is not a hypothesized effect in this paper.
der Brug, 2007; Akkerman et al., 2016) or after a period in government (Albertazzi & McDonnell, 2005; Akkerman & Rooduijn, 2015; Akkerman et al., 2016). As mainstream right parties have offered credible alternatives to voters on the socio-cultural dimension (Odmalm & Hepburn, 2017), PRR parties could strategically react by taking an even more extreme position on this dimension. Contrarily, it is not expected that mainstream left parties have a significant effect on PRR parties, as they are not the main competitor of PRR parties on the socio-cultural dimension in the first place (Meguid, 2008). Therefore, the final hypothesis that will be tested is:
H4: The more right-wing the position mainstream right parties take the more right-wing position populist radical right parties take on the socio-cultural dimension.
In sum, the causal model that will be tested in this paper is schematically drawn in figure 1.
Operationalization, data and method
The next section will start with the case selection and a classification of the PRR. Subsequently, it will be discussed how the dependent variable of this study, party system polarization, is measured. Then this paper will turn to the data selection, operationalization of independent variables and the methods and data processing.
Party classification
Measuring the effect of PRR party success on party system polarization requires a selection of PRR parties in which their radicalism on the socio-cultural dimension shines through. Therefore, PRR parties are understood to be (ideologically) nativist, authoritarian and populist (cf. Mudde, 2007). Populism is defined as:
“an ideology that considers society to be ultimately 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 volonte générale (general will) of the people” (Mudde, 2004: 543).
Level of party system polarization Populist radical right size Populist radical right positioning Mainstream right positioning Mainstream left positioning Contextual factors (control variables)
[11]
The PRR’s populism is mostly expressed in their appeal to the common, hardworking, man and their antagonism towards political, economic and/or cultural elites (Betz, 1993; Mudde, 2007). Nativism can be defined as:
“an ideology which holds that states should be inhabited exclusively by members of the native group (‘the nation’) and that non-native elements (persons and ideas) are fundamentally threatening to the homogenous nation-state” (Mudde, 2010: 1173).
Nativism is central to the ideology of the PRR (Akkerman et al., 2016). More concretely, immigration and integration have been very important factors in these parties’ manifestos and identities. During the last decade, anti-immigration stances have mainly evolved into anti-Islam attitudes. According to the PRR, the Islam is incompatible with Western norms and values such as equal rights for men and women and freedom of speech (Betz, 2007). Though authoritarianism is not only linked to the PRR, but also to for example conservatism, it is important nevertheless as it unites PRR parties. In this present paper authoritarianism is understood as the belief that society should be strictly ordered (Mudde, 2007). Violations of authority are supposed to be punished heavily. In sum, their relatively extreme spatial positioning regarding socio-cultural issues such as immigration and law and order is what makes PRR parties ‘radical’ (Akkerman et al., 2016).
PRR parties are usually ‘niche parties’. They try to distinguish themselves from mainstream parties by focussing on only a handful of policies (Green-Pedersen, 2007; Meguid, 2008; Akkerman et al., 2016) on which they actually have the credibility and resources to compete. PRR parties mostly campaign on socio-cultural issues, issues that mainstream parties traditionally emphasized less. This however does not mean that they do not pay attention to other issues, such as those on the economic dimension, at all (Mudde, 2007).
Mainstream parties are parties that are not radical, not ‘niche’, nor anti-establishment (in relation to the discussion in the theory section). Mainstream is an umbrella term for two types of parties, those that are ideologically centrist (from centre-left to centre-right) and those parties that are part of the establishment (Akkerman et al., 2016). These established parties are loyal to the political system, they do not want to sustain centrifugal competition nor change the system of government. Their ideology encompasses many different political issues on both the economic and socio-cultural dimension. They mainly compete with each other on the economic dimension and with niche parties on the socio-cultural dimension (Rovny & Edwards, 2012; Wagner & Meyer, 2016).
The present paper relies on the publication of Akkerman, de Lange and Rooduijn (2016) to select PRR and mainstream parties, as its definition of PRR and mainstream is roughly equal to the one employed here (see Table 1). For any PRR party existing prior to 1990 not already included by Akkerman and colleagues, the classification of Mudde (2013) is followed for similar reasons. For Germany, the Alternative for Germany (AfD) has been added as well, even though this party has been omitted by Akkerman, De Lange and Rooduijn (cf. Arzheimer, 2015). Following Meguid (2008), Han (2015) and Wagner and Meyer (2016), only centre-left and centre-right parties that have governed at least once during the period of study are operationalized as mainstream parties. In principle, all parties were kept if possible, since it is essential to include as many parties as possible to approach the party system polarization of the social reality. For instance, if only half of all political parties are added, then a lot of information on inter-party competition is lost. All countries not containing PRR parties are removed from the data. Still, robustness checks were done adding the four remaining Western European countries without PRR parties: Ireland, Luxembourg, Portugal and Spain.
[12]
Table 1: populist radical right parties in Western Europe: votes and (seats)
Country English party name Party abb. ‘80-‘89 ‘90-‘99 ‘00-‘09 ‘10-‘16
Austria+ Alliance for the Future of Austria BZÖ 7.4 (7.7) 3.5 (3.5)
Freedom Party FPÖ 7.4 (8.2) 22.0 (22.8) 12.8 (13.3) 20.5 (20.5)
Belgium National Front FNb* 1.6 (1.3) 2.0 (1.0)
Flemish Bloc/Interest VB 1.5 (0.6) 8.1 (7.7) 11.8 (11.7) 5.8 (4.7) Denmark Danish People’s Party DF 7.4 (7.4) 13.2 (13.5) 16.7 (16.5)
Progress Party FrP 6.6 (6.7) 5.1 (5.1) 0.6 (0)
Finland True Finns PS 8.0 (6.5) 2.4 (1.5) 2.9 (2.0) 18.3 (19.3) France National Front FN 9.8 (3.3) 13.7 (0.1) 7.8 (0) 13.6 (0.3)
National Rep. Movement MNR* 1.1 (0)
Germany Republicans REP* 1.9 (0) 0.6 (0)
Alternative for Germany AfD 4.7 (0)
German People’s Union DVU-NPD*/ ** 1.2 (0) 1.6 (0) 1.3 (0)
Greece Popular Orthodox Rally LAOS 3.9 (4.2) 1.8 (0)
Italy Northern League LN 9.1 (12.2) 5.6 (6.2) 4.1 (2.9)
Netherlands Centre Democrats CD 0.9 (0.7) 1.6 (2)
List Pim Fortuyn LPF 11.4 (11.3)
Party for Freedom PVV 5.9 (6.0) 12.8 (13.0)
Norway+ Progress Party FRP 7.1 (5.7) 10.8 (10.6) 19.9 (20.5) 16.3 (17.2)
Sweden New Democracy NyD 4.0 (2.4)
Sweden Democrats SD*** 3.3 (0) 12.9 (9.9)
Switzerland Swiss People’s Party SVP 11.1 (12.0) 16.4 (16.3) 27.8 (29.3) 28.9 (29.8)
UK British National Party BNP* 1.9 (0)
UK Independence Party UKIP 1.9 (0) 7.9 (0.1)
Average 6.6 (5.5) 7.5 (6.4) 7.2 (6.7) 11.3 (9.2)
+ No elections after 2009 were coded in the CMP. *Not in the CMP data. **Only together as one party in the CHES data. *** For the
years that the Sweden Democrats did not win seats, they are not in the CMP. Sources: Bakker et al. (2015), Volkens et al. (2016), Polk et al., (2017); ParlGov; own calculations
Finally, to be able to show the relative contribution of certain party families to polarization, it is necessary to also briefly define the radical left, since they will be used in one analysis as well.8 For the radical left, the classification of March and Rommerskirchen (2015) is used. The radical left is a rather ideologically coherent party family that is radical in its rejection of contemporary socio-economic structures, institutions and in their advocacy of an intense “transformation of capitalism in order to take power from existing political and economic elites” (March & Rommerskirchen, 2015: 41). Moreover, they are left-wing because they mainly advocate collective social and economic rights and see economic inequality as the disturbing basis of the current socio-political arrangement (March & Rommerskirchen, 2015).
Polarization: a needle in the haystack
The dependent variable in most of the analyses is polarization. Taking the definition of Sartori as a starting point, an adequate measure of polarization should include party positioning on a continuum. It should thus include some sort of measure of positional distance between parties. Repositioning of parties towards each other should decrease polarization and repositioning away from each other should increase polarization. Moreover, the polarization measure should contain some form of weighing according to party size. This way, electorally more relevant parties will count more heavily
8 Some of the analyses will be carried out on the party level instead of election level. In these analyses, Green parties are
not added as they are niche parties that usually only compete on one or two (valence) issues and are not expected to influence the general dimensional position of the party families under scrutiny (see also Meguid, 2008; Abou-Chadi, 2016).
[13]
towards polarization. Finally, the number of parties should not make a decisive difference since Western European party systems differ greatly concerning their fragmentation (Sartori, 1982).
Though there are literally dozens of different polarization indices, none are perfect (see also Schmitt, 2016). A first measure of polarization that has been used in the party systems literature is simply based on the ideological distance between the two most distant parties (e.g. Sørensen, 2014). This measure is suitable to study polarization in two-party systems such as in the United States, but less functional in multiparty systems. It disregards any (re)positioning not happening on the flanks, which could still have a large impact on the ideological differences within a party system.
A second measure of polarization counts the number of votes/seats of flank parties (e.g. King et al., 1990; Warwick, 1992). Though this measure takes weighing according to party size into account, it still has numerous pitfalls. First, an arbitrary ideological position has to be chosen as cut-off point to divide parties into flank parties and centre parties. Second, this index does not perform if there are no (or very small) parties on the flanks as defined by the cut-off point. Third, this measure also disregards the (re)positioning of parties more near the centre of the continuum.
A third group of studies uses the (squared) standard deviation of ideological distance to the centre (e.g. Sigelman & Yough, 1978; Dalton, 2008; Ezrow, 2008). This measure takes the positions of parties into account, but relies too heavily on the number of parties in a system. If the number of parties increases, polarization is very likely to decrease, as the probability is higher that there will be more parties near the ideological centre (Schmitt, 2016). An example illustrates this problem (see Figure 2). The vote-weighted bipolar party system will be more polarized as the mean difference of the two parties to the centre is larger than in the party system with more parties (Figure 2, left and right). However, the ideological differences are larger in the example with five parties, as the two flank parties are clearly alienated from all the others. Yet, as the three centre parties are very close to the ideological centre, the standard deviation measure will qualify this party system as weakly polarized. Put differently, this measure is not suitable as “the presence of a substantial ideologically- based conflict should be indicated by the measurement” (Schmitt, 2016: 10). Therefore, in essence, this index measures the distance of political parties to the centre of the ideological dimension, not the overall ideological spread (for a similar critique, see Evans, 2002).
[14]
Finally, in most studies parties are weighted according to their share of votes or seats (Alvarez & Nagler, 2004; Ezrow & Xezonakis, 2011), since larger parties obviously have a larger impact on party systems. One clear issue concerning the size of parties is its context dependence (Schmitt, 2016). In certain political systems such as that of Germany, there is an electoral threshold. Parties performing below that threshold are usually less relevant for political competition than those gaining enough votes to earn parliamentary representation. Hence, the increase in relevance is not always linear. By adding country fixed effects, this study controls for this context dependency to a certain extent (see methods section further on).
In order to satisfy most of the conditions for an adequate polarization measure, this study relies on the index created by Esteban and Ray (ER) (1994). The ER-index measures the sum of the distances on a continuum between all pairs of ‘groups’ (e.g. households, communities, political parties, coalitions, etc.). It starts with the intuitive assumption that polarization is at its highest when two evenly large groups (e.g. parties) are located on the opposite ends of the ideological continuum (see also Indridason, 2011). Then, Esteban and Ray (1994) base their formula on four axioms, as a
[15]
correct measure of polarization should satisfy all those conditions. The first axiom states that polarization increases when two smaller parties on one of the flanks converge, while a larger party on the other side of the continuum does not move, keeping the distance between the smaller parties and the large party consistent (see Figure 3A). Though the average distance between both groups does not increase, the group on the right now is more cohesive, hence polarization increases. The second axiom states that polarization increases if a small intermediate party moves towards a large, but not the largest, party on a flank, while the largest party is on the other flank (Figure 3B). The third axiom states that when a large group in the centre loses voters to two smaller groups on the flanks, polarization increases (Figure 3C). The fourth axiom states that polarization should at least not decrease if a very small extremist group on one of the flanks of the ideological spectrum ceases to exist and has its members transferred to a large extreme group on the other side of the spectrum (with a third large group nearer to this large extreme group - figure 3D).
The ER-index was originally designed to measure a specific form of social polarization using survey data. Still, it is possible to use it to measure many other kinds of polarization between different kinds of actors, using several types of data (Oosterwaal & Torenvlied, 2010; Rehm & Reilly, 2010). The ER-index is calculated with the following equation:
𝑃 = 𝐾 𝜋𝑖𝛼+1𝜋𝑗|𝑦𝑖 𝑛 𝑗=1 𝑛 𝑖=1 − 𝑦𝑗|
Here, K is an arbitrary scalar to normalize the measure.9 In the present paper, it is set to 1. The parameter π stands for the proportional size of a political party, which ranges from 0 to 1. α is a constant ranging from 0 to 1.6, which could be interpreted as the ‘polarization sensitivity’ (Esteban and Ray, 1994: 834).10 In this paper α is also set to 1, since it is a common choice in the literature (Oosterwaal & Torenvlied, 2010; Rehm & Reilly, 2010). Setting this number too low could inflate the results and show a high level of polarization in many countries, while setting this number to the maximum would lump together all countries as non-polarized (Lindqvist & Östling, 2010). Finally, y is the position of political party i or j on any pre-defined ideological dimension.
The data: CMP versus CHES
To test the hypotheses, two data-sets will be used. First, manifesto data from the Comparative Manifesto Project (CMP) will be used to determine party positions on both policy dimensions (Volkens et al., 2016). Only data from 1980 to 2015 is used, since hardly any PRR party existed prior to the 1980s (Betz, 1993; Wagner & Meyer, 2016). In total, the CMP consists of 926 manifestos of 206 parties divided over 123 election years. All procedures will be repeated with data from the Chapel Hill Expert Survey (CHES) (Bakker et al., 2015; Polk et al., 2017), to be able to cross-validate
9 For the analyses, the height of this number does not matter since it only effects the range of the scale on which P is
measured. A higher number increases the range. Setting this number to 1 does not necessarily mean that A further discussion would go too far in light of the purpose of this paper.
10 It goes too far to discuss the α parameter in detail, but the domain of α is restricted to 1.6 to satisfy the four axioms of
polarization as discussed above. If the parameter is set to zero, then P is the measurement of the Gini-coefficient (Duclos, Esteban & Ray, 2004).
[16]
the results. In the case of the CHES, data is only available from 1999 to 2014. It consists of 464 manifestos of 145 parties divided over 55 election years.11
The CMP is based on the coding of quasi-sentences of party manifestos into pre-determined categories such as the appeal for more law and order or the appeal for more market regulation (Volkens et al., 2016). By quantifying statements in manifestos, it is possible to create measurements of party positions. Prior knowledge of the political parties that are coded and their past policy stances are not necessary to generate the data. The CHES is simply a regularly held expert survey (once every three to four years), in which several topics concerning political parties and elections are addressed (Bakker et al., 2015; Polk et al., 2017). For example, experts are asked to place political parties on the socio-cultural and on the economic dimension.
There are, however, several strengths and weaknesses concerning both the CMP (Dinas & Gemenis, 2010) and the CHES (Steenbergen & Marks, 2007). Starting with the CMP, three major advantages can be addressed. First, the CMP offers very rich longitudinal data as it even contains manifestos dating back to the 1940s. As a rule of thumb, parties that have won a seat in at least one parliamentary election are included (with exceptions). Moreover, all Western European countries are available in the data. Second, the data is relatively impartial, especially compared to expert or voter surveys which are based on their respective perceived position of political parties (Dinas & Gemenis, 2010). The perceptions of experts could be heavily influenced by prior knowledge of and opinions on political parties, which could for instance cause an overestimation of party position stability (Steenbergen & Marks, 2007). Third, the CMP can record large changes in ideological competition between political parties relatively well (McDonald et al., 2007). Where experts judge political parties’ policy positions based on their prior knowledge of the parties’ past, the CMP judges these positions case by case. Hence, it will not underestimate larger positional shifts.
Yet, three shortcomings of the CMP should also be acknowledged. First, several scholars have argued that the CMP is not particularly suited for studies on PRR ideology. Most importantly, PRR parties’ core issue of immigration is only marginally covered (Protsyk & Garaz, 2013; Akkerman & Rooduijn, 2015). Second, the data has not proven to be very reliable, at least not on the level of individual issue categories (Mikhaylov et al., 2012; Gemenis, 2013). For example, it is not uncommon that a coder applies the ‘appeal for more democracy’ category to a quasi-sentence referring to an ‘appeal for more human rights’. Hence, since the CMP only uses one coder per manifesto, it is exceptionally prone to human error. Other researchers, however, claim high inter-coder reliability scores when comparing coded manifestos through time (e.g. Budge, 2001; Klingemann et al., 2006). In this study, these two problems will be alleviated to a certain extent by aggregating several issues categories into broader policy dimensions, i.e. the socio-cultural dimension and the economic dimension (see Wagner & Meyer, 2016). For example, several PRR core issue categories such as more law and order and a positive attitude towards the national way of life can be aggregated on the socio-cultural dimension. This will generate a relatively full measure of PRR ideology. Third, smaller parties are sometimes not included, depending on the country, year and party (Akkerman & Rooduijn, 2015; Volkens et al., 2016). Especially prior to the 1990s, several manifestos are missing. Luckily, the most recent version of the CMP (Volkens et al., 2016) includes most of the PRR parties that were previously missing, such as the Centre Democrats (CD) in the Netherlands (Akkerman & Rooduijn, 2015). Still missing however are the British National Party (BNP), the National Front from
11 The structure of the CHES data is regular (intervals of three or four years), which in some cases means that no new
[17]
Belgium (FNb), the German Republicans (Rep) and the Sweden Democrats (SD) before they were electorally relevant.
Still, the CHES has some advantages as well. First, the inter-expert reliability is generally high, among others because experts are able to make sense of rather complex political and social realities (Hooghe et al., 2010). Second, for most countries and parties, several experts have been questioned (ranging between three and twenty-three), which improves the validity of the measures (Bakker et al., 2015; Polk et al., 2017). Third, almost all parties have been included in the dataset, including all PRR parties. However, Switzerland and Norway are completely omitted from the data. This is a large disadvantages compared to the CMP, as the PRR is a very relevant party family in both countries, regularly obtaining more than fifteen per cent of the votes.
However, even though expert surveys such as the CHES are widely used in the political science literature (e.g. Castles & Mair, 1984; Huber & Inghlehart, 1995; Akkerman et al., 2016), they are also widely criticized. Overall, it is unclear whether experts interpret the survey questions similarly (Budge, 2000). In other words, the CHES might not measure what was intended to be measured. For example, some experts might judge a party’s policy position on the socio-cultural dimension based on different criteria than others (Budge, 2000; Steenbergen & Marks, 2007). Some experts might estimate a party’s position merely based on immigration issues, where others also base their judgement on issues concerning the environment or the European Union. Also, where manifestos unequivocally measure the intentional position of a party, experts might describe the empirical behaviour of political parties (Steenbergen & Marks, 2007). Finally, as has been mentioned earlier, experts’ perceptions could generate an overestimation of party position stability. Experts are prone to base their judgements on a longer period in time, while manifestos are snapshots (Steenbergen & Marks, 2007). This stability has been shown to result in a large cross-national centrist bias (Best et al., 2012; Budge & McDonnell, 2012), meaning that country experts place political parties near the ideological centre. In sum, “experts may bring wildly varying considerations to bear when judging party positions” (Steenbergen & Marks, 2007: 348).
Taking all advantages and disadvantages of the CMP and the CHES into account, the former is chosen as the primary data source. Especially considering the fact that the CHES does not include Norway or Switzerland and that it only contains data from 1999-2014, the CMP seems more suitable for this study. Moreover, most CMP-related problems discussed above are partially solved by aggregating sets of issue categories into broader issue dimensions. Still, several studies have used both CHES and CMP data to cross-validate the findings (e.g. Hooghe et al, 2010; Carter, 2013; Polk et al., 2017). By cross-validating the results, one could almost be certain that valid and reliable measurements will be done. The cross-validation can be found in Appendix A. From now on, unless explicitly stated, every statement about data, measures or results refers to the CMP.
Of some parties the manifestos were not coded in the CMP, so they were removed. Non-traditional manifestos such as party bloc manifestos or internet-based texts were kept, however. Although it is said that the inclusion of these ‘proxy documents’ could cause a bias towards the centre (Gemenis, 2012), they are still included for three reasons. First, as many parties as possible should be included to gain a complete picture of polarization of party systems. Second, the aforementioned bias to the centre would only mean that it is more difficult to find polarization, which is thus an extra trial. Third, other scientists have shown that rather than causing bias, consequently keeping ‘proxy documents’ actually just “tends to generate counterbalancing biases and hence ‘noise’ rather than systematic error” (Budge & McDonnell, 2012: 611). Still, robustness checks removing these manifestos will be carried out.
[18]
Finally, some manifestos did not contain any reference to an issue category on the socio-cultural or economic dimension. As parties usually do not change their ideology enormously between elections and since it is not desirable to drop too many parties, the following strategy is chosen: if a party’s position on the dimension could not be calculated, then the party’s position during the previous elections is taken as a proxy (with a maximum of one election). If that data is missing as well, then a party’s position during the following election is used (with a maximum of one election). If that is missing as well, then the manifesto for that party in that particular election year is dropped.12 Eventually, after dropping all parties that did not have a manifesto at all or did not take a policy position on one of the dimensions, there are still 908 (98.1%) manifestos divided over 203 (98.5%) parties in 123 (100%) elections in the dataset.
Operationalization
Before continuing with the measurement of the independent variables, it is necessary to describe the way in which two parameters of the ER-index are operationalized: weighing parties by their size (π, see ER-index) and party positioning on a policy dimension (y). The size of parties is simply operationalized as the proportion of seats that parties have won after the most recent election. Seats are chosen over votes for two reasons. First, this paper is interested in the polarization of party systems, which means that the proportion of seats is a better expression of the party system than the proportion of votes. The proportion of seats literally shows the parties’ strength in parliament and thus their relevance. Second, choosing seats over votes helps to circumvent a possible tautological relationship between polarization and the electoral strength of (PRR) parties. More concretely, one could state that a party system with a stronger party on the flanks in terms of votes is by definition more polarized. In some countries such as the Netherlands, the proportion of votes and seats is roughly the same. In other countries, however, there could be a relatively large discrepancy between these proportions due to a less proportional system (i.e. with an electoral threshold). By using the proportion of seats to calculate the ER-index, it is more difficult to find a significant relationship between polarization and the size of the PRR. In countries with an electoral threshold, flank parties will often obtain fewer seats than votes. Thus, the radical positioning of these parties will weigh less within the calculation of the ER-index. Consequently, the index will have a lower value than if the proportion of votes would have been used.
In order to measure parties’ positions on both dimensions, the present paper relies on the grouping of CMP categories created by Bakker and Hobolt (2012) into ‘left-wing’ issues and ‘right-wing’ issues.13 Note that this operationalization works both for measuring party positioning within
12 This procedure made it possible to include 50 more manifestos. Only the Swedish ecology party is persistent in its refusal
to take any ‘GALTAN’ policy position.
13 For the socio-cultural dimension: political authority (per305), national way of life: positive (per601), traditional morality:
positive (per603), law and order (per605) social harmony (per606) and multiculturalism: negative (per608) are coded as socio-culturally ‘right-wing’, while freedom and human rights (per201), democracy (per202), anti-growth economy: positive (per 416), environmental protection: positive (per501), culture: positive (per502), national way of life: negative (per602), traditional morality: negative (per: 604), multiculturalism: positive (per607), underprivileged minority groups (per705) and non-economic demographic groups (per706) are coded as socio-culturally ‘left-wing’.
For the economic dimension: free market economy (per401), incentives (per402), protectionism: negative (per407), economic growth: positive (per410), economic orthodoxy (per 414), welfare state limitation (per505), education limitation (per507) and labour groups: negative (per702) are coded as economically right-wing, while market regulation (per403), economic planning (per404), corporatism/mixed economy (per405), protectionism: positive (per406), Keynesian demand management (per409), controlled economy (per412), nationalisation (per413), Marxist analysis: positive (per415),
[19]
the ER-index and as an independent variable by itself. Then, using the logit scale for multi-category scales generated by Lowe and colleagues (2011), a party’s policy position can be calculated for each dimension. Their (interval) scale is calculated using the following equation:
𝜃𝑖𝑛𝑑𝑒𝑥(𝐿) = 𝑙𝑜𝑔 𝑅𝑗 𝑗 𝐿𝑘 𝑘
which stands for the logarithmic of the sum of right-wing coded sentences divided by the sum of left-wing coded sentences. The higher the outcome, the more right-left-wing the policy position is. The logic of the authors
“suggests that from the point of view of a party manifesto writer wanting to communicate a position effectively, it is important to manipulate not so much the absolute quantity of sentences allocated (R + L), but rather their relative balance, or R/L” (Lowe et al., 2011: 131; emphasis in original).
This way of calculating party positions has major advantages and some minor disadvantages. A major advantage is the likelihood that the parties’ policy positions will be (roughly) normally divided due to the logit transformation.14 Moreover, with this index, parties could in theory become infinitely more extreme because this equation does not have a minimum or maximum. This could explicitly show the ideological differences in party systems. An important limitation of this measure is its lack of a substantively meaningful centre, as θ=0 does not necessarily denote a centrist policy position (Lowe et al., 2011). Luckily, as the present study is concerned with polarization (i.e. the increase of the ideological spread in party systems) and not with absolute policy positions per se, this is not a major issue. After calculating all policy positions, the dimensions are normalized onto a 0-10 scale.15 Normalizing the scale makes it possible to compare between dimensions and between data sources, as the CHES also uses a 0-10 scale.
Naturally, the degree of polarization is not only influenced by the vote share of PRR parties or the ideological position taken by mainstream or PRR parties. Other factors might confound these relationships and should thus be added as control variables. First, some studies find a temporary moderation (or radicalization) of PRR parties after these parties have been included in government, at least in some cases (Akkerman et al., 2016). Therefore, it is crucial to take PRR government membership and support into account. Government inclusion is operationalized as a nominal variable with three categories: no inclusion (as reference category), minority government support and government inclusion.
Second, even though Sartori (1982) has argued that party systems with three parties could be more polarized than party systems with ten parties, this does not mean that the number of parties has no impact on polarization at all. For example, polarization could be less likely in a two-party system, as parties usually tend to converge to the median voter – at least if the parties are mainly vote-seeking (Downs, 1957; Alesina, 1988). In (effectively) two-party systems, niche parties (among which PRR parties) have a hard time getting representation. However, only in the United Kingdom and to some extent France, Germany and Spain, there has been a party system with two dominant
equality: positive (per503), welfare state expansion (per504), education expansion (per506), labour groups: positive (per701) are coded as economically left-wing.
14 Which is confirmed by a visual inspection
15 Normalizing the scales does not have a major impact on the findings. It does not change anything regarding significance
