How people organise cultural attitudes: cultural belief systems and the populist radical right

Abstract

Political scientists generally agree that all individuals structure their cultural attitudes in the same unidimensional fashion. However, various populist radical right parties remarkably combine moral progressiveness with conservatism regarding immigration-related issues. This suggests that the structuring of cultural attitudes among the electorate may also be more complex than typically assumed. Applying Correlational Class Analysis to representative survey data, the study uncovers three cultural belief systems. For individuals adhering to an integrated one, all cultural attitudes are interdependent, as typically assumed. However, two alternative belief systems are also uncovered: intermediate and partitioned. In the latter, positions on one cultural attitude (e.g. ethnocentrism) are barely related to positions on others (e.g. rejecting Islam or opposing homosexuality). The existence of multiple cultural belief systems challenges the widely held assumption that all people organise their cultural attitudes similarly. Both political party agendas and individuals’ education level and religion appear key to understanding variation in belief systems.

Keywords:

• Belief systems
• Correlational Class Analysis
• cultural issues
• ethnocentrism
• political attitudes
• populist radical right parties
• rejection of Islam

Political attitudes regarding cultural issues are an important focus in both scholarly and popular writing and feature prominently in contemporary political debates. Contrary to attitudes pertaining to the economic domain, cultural attitudes concern issues of individual freedom and cultural diversity. Economic issues differentiate parties and electorates that oppose redistribution from those supporting it, and consequently pit the economically strong against the economically weak in the so-called ‘democratic class struggle’ (Lipset 1960). Cultural issues, alternatively, differentiate parties and electorates that strive for individual freedom and embrace cultural diversity from those favouring a more rigid and less culturally diverse order.

In line with Lipset’s (1959, 1960) seminal observations, most scholars agree that political attitudes among the public (Achterberg and Houtman 2009; Middendorp 1991), as well as the agendas of political parties (Bornschier 2010; Lefkofridi et al. 2014), are structured according to these two dimensions: (1) redistribution issues, and (2) issues pertaining to cultural order and individual liberty. For attitudes on the latter – which are the focus of this study – this assumes that among the public a progressive stance on one cultural issue goes together with a progressive approach to a broad array of other cultural issues (for an overview of five decades of studies reporting such a cultural attitudinal dimension, see: Achterberg and Houtman 2009). In Lipset’s day, this cultural dimension included ‘civil rights for political dissidents, civil rights for ethnic and racial minorities, internationalist foreign policies, and liberal immigration legislation’ (Lipset 1959: 485). Similarly, in his seminal study, Allport (1979[1954]: 68) considers prejudice a generalised attitude and argues that ‘[p]eople who reject one outgroup will tend to reject other outgroups. If a person is anti-Jewish, he is likely to be anti-Catholic, anti-Negro, anti any outgroup’ (emphasis added). Also contemporary research considers a conservative stance towards ‘people of a different race, people with AIDS, immigrants/foreign workers, homosexuals, people of a different religion, and heavy drinkers’ (Inglehart et al. 2008: 269) to be part of one cultural dimension. In short, ‘cultural issues are considered interchangeable’ (De Koster and Van der Waal 2007: 452) in studies on the ideological profile of citizens of Western countries.

Such a one-dimensional outlook on cultural issues may be applicable to large parts of the electorate, but party-level research by Akkerman (2005) and Betz and Meret (2009) demonstrates that this is strikingly at odds with the agendas of various contemporary populist radical right parties (PRRPs) in Western Europe. Whereas such parties grew out of resistance against multiculturalism and culturally progressive values, these studies show that various Western European PRRPs currently combine their well-known culturally conservative agenda regarding immigration and ethnic minorities with a culturally progressive approach to gender issues and sexual minorities. Consider for instance the Flemish Bloc in Belgium, the Danish People’s Party in Denmark, the Northern League in Italy and the List Pim Fortuyn and later on Wilders’ Party for Freedom in the Netherlands. ‘Following Fortuyn’s example, they started to portray themselves as defenders of liberal values and principles: the separation of church and state, freedom of expression and, above all, gender equality and women’s rights’ (Betz and Meret 2009: 322; also see Akkerman 2005). This raises a salient question about their constituencies: is the well-established conservative stance regarding immigration-related issues among supporters of PRRPs (De Koster et al. 2014; Rydgren 2008) actually accompanied by a conservative approach to other cultural issues, as commonly assumed? We aim to uncover the attitudinal configuration of PRRP supporters regarding these issues. To do so, we draw on Converse’s (1964) notion of belief systems and apply an analytical approach which, contrary to conventional approaches that impose a singular structure of attitudes on every individual in the data, allows us to uncover heterogeneity in the ways individuals structure political attitudes.

Belief systems refer to the way individuals understand and structure political attitudes. Different belief systems – that is, different configurations of political attitudes – may exist among the electorate, but conventional analytical approaches ignore this by assuming that all people organise attitudes similarly. While extant studies distinguish between economic and cultural issues, they assume that all members of the electorate coherently combine either conservative or progressive stances on various cultural matters. Yet the counter-intuitive combination of cultural stances in the agendas of PRRPs suggests that an alternative belief system regarding cultural issues may also characterise their constituencies, or parts of them. If so, this has previously gone unnoticed, despite the overwhelming attention paid to cultural issues in political science and sociology in decades of research. Taking seriously the possibility that different parts of the public have different belief systems, we study the ways individual voters structure their attitudes on these cultural issues without a priori imposing the conventional one-dimensional structure (see Baldassarri and Goldberg 2014; Goldberg 2011).

We focus on the Netherlands, a strategic case for the issue at hand for two reasons. First, it is characterised by a widely observed ‘paradox of tolerance’ (Duyvendak 2004): while it is renowned as one of the most progressive countries regarding gender issues and homosexuality (Inglehart 1997), issues relating to immigration and ethnic minorities are fiercely debated (Vasta 2007). This suggests that at least parts of the electorate do not structure these cultural attitudes one-dimensionally. In the political realm, Geert Wilders’ Party for Freedom (PVV), which has been represented in the Dutch parliament since 2006, is an ideal-typical case of a PRRP that combines both progressive and conservative cultural stances. It has a conservative agenda regarding immigration and ethnic minorities and particularly criticises Islam, while it advocates progressive stances on issues of gender and sexuality (Betz and Meret 2009; Vossen 2011). Second, to our knowledge, there is no other Western European country for which representative data are available which include measures of all relevant cultural issues. Considering the electoral success of PRRPs with a similar agenda as the PVV in various other Western European countries, the scholarly implications of our study go beyond the case of the Netherlands.

Theorising the configurations of cultural attitudes among PRRP constituencies

One cultural dimension among PRRP supporters?

Unsurprisingly, previous research has assumed that supporters of PRRPs hold conservative stances on various cultural issues. This is a reflection of the near consensus in the literature that ‘[p]eople have generalised prejudice levels: those who are more prejudiced toward one outgroup tend to be more prejudiced toward others’ (Son Hing and Zanna 2010: 163). Nevertheless, two sets of research findings challenge this assumption.

First, various PRRPs in Western Europe consistently voice a culturally progressive agenda concerning gender issues and sexual minorities to support their anti-immigrant discourse (Akkerman 2005; Canovan 2005; De Koster et al. 2014; Elchardus and Spruyt 2014), and position themselves as ‘defenders of fundamental liberal values’ (Betz and Meret 2009: 313). This observation at the party level suggests that supporters of PRRPs may not necessarily be conservative regarding all cultural issues. If this is not only true for the agendas of political parties, but also for the attitudes of the public, it would be strikingly at odds with the claims of those like Inglehart and colleagues (2008: 269) that the ‘acceptance or rejection of homosexuals … [is] a particularly sensitive indicator of overall tolerance of outgroups’. Do supporters of PRRPs hold conservative attitudes on all cultural issues, as conventional approaches to studying political attitudes assume? Or is the complexity at the party level mirrored in the attitudes of PRRP constituencies?

Secondly, research suggests that religion and religiously inspired attitudes provide the key to understanding why cultural attitudes are not necessarily structured in the same way by everyone (see Baldassarri and Goldberg 2014). De Koster and Van der Waal (2007), for example, demonstrate that moral conservatism, religious orthodoxy and ethnocentrism cannot be lumped together as a generalised indicator of cultural conservatism. They suggest that the strength of the association between these cultural attitudes varies in the population. Moreover, they show that moral conservatism, which includes conservative stances on gender roles and homosexuality, is religiously inspired, while ethnocentrism is not. Accordingly, conservative stances on one set of cultural issues do not necessarily go together with such stances on other cultural issues. Consequently, it is possible that, like their party elites, the constituencies of PRRPs combine a progressive stance towards homosexuals with a conservative stance towards other outgroups like immigrants and ethnic minorities. In trying to further our understanding of the specific constellations of cultural attitudes among PRRP supporters, we recognise that relations between various cultural attitudes may differ systematically between PRRP constituencies on the one hand and supporters of other parties on the other. In this regard, we draw on Converse’s (1964) notion of belief systems.

Cultural belief systems

Converse (1964: 207) famously conceptualises a belief system as ‘a configuration of ideas and attitudes in which the elements are bound together by some form of constraint or functional interdependence’. Constraints refer to the ways different elements of a belief system – in our case: cultural attitudes – are related to one another. That is, whether or not a position on and change in one attitude (e.g. attitudes regarding homosexuals) is associated with positions on and changes in other attitudes included in the belief system (e.g. attitudes regarding immigrants and ethnic minorities). As Converse stresses, these constraints are social in nature and relate to the ways belief systems are created and diffused.

Inspired by Converse’s concept of ‘belief systems’, we use the term ‘cultural belief systems’ to refer to the configuration of political attitudes towards cultural issues that individuals adhere to. The widely held assumption that different cultural attitudes consistently cluster together ensues from the idea that cultural belief systems are guided by one latent factor which shapes the various, more specific, attitudes. Indeed, it is possible that a single, general orientation guides all specific beliefs, generating consistency within a belief system (Moskowitz and Jenkins 2004). This encompassing ideology/theme – if there is one – can relate to economic, religious, racial, political or social beliefs (Conover and Feldman 1984; Moskowitz and Jenkins 2004; Olson and Carroll 1992). For example, consider Christianity, which can inform conservative stances on various issues like gender roles, homosexuality, abortion and euthanasia. Due to an underlying general Christian orientation, attitudes on these specific issues are functionally interdependent (see De Koster and Van der Waal 2007). In a similar way, other segments in society may have other general orientations, and may therefore experience other constraints on their cultural attitudes and the relationships between them. Consequently, different belief systems consisting of specific constellations of cultural attitudes may exist among different parts of the electorate.

Along with different types of belief system guided by different encompassing ideologies or themes, there may be variation in the extent to which cultural belief systems are integrated. Converse (1964) suggests that the level of integration of a belief system is associated with political competence. Although this seems plausible, Achterberg and Houtman (2009) demonstrate that political competence does not affect the extent of value (in)coherence among Dutch citizens. A similar conclusion is reached by Tetlock (1984), who argues – based on Putnam’s (1971) interviews with members of the British House of Commons – that individuals with ample political competence may adhere to very different political belief systems, and that belief systems substantially differ between party families. These findings downplay the role of political competence in understanding belief systems. Consequently, if PRRP constituencies in the Netherlands demonstrate a less integrated cultural belief system, this does not imply that a lack of political competence underlies it. Alternative interpretations involve processes of framing and agenda-setting of Dutch PRRPs and/or skewed media representations of specific cultural issues. After all, research indicates that political attitudes and sentiments among the public can be inspired by the extent to and the ways in which political parties (Hellwig and Kweon 2016; Minkenberg 2001) and the media (Boomgaarden and Vliegenthart 2009; Van Spanje and De Vreese 2014) emphasise specific issues. We argue that such mechanisms may not only be true for the contents of political attitudes, but also for the ways people structure these attitudes ‒ i.e. for their belief systems.

In order to uncover whether those who prefer Dutch PRRPs have a cultural belief system that differs from those of the electorates of other parties, and whether and how these different belief systems are at odds with the standard assumption that a conservative stance on one cultural issue is accompanied by a conservative stance on other cultural issues, we employ a recently developed statistical technique: Correlational Class Analysis (CCA) (Boutyline 2016a; cf. Goldberg 2011). CCA allows us to uncover segments in the electorate that systematically differ in the way they structure their cultural attitudes. In this way, our analyses advance the literature by exploring whether there are multiple cultural belief systems viable among the electorate and by scrutinising the cultural belief systems of PRRP constituencies.

Potential configurations of the cultural belief systems of PRRP constituencies

In order to study cultural belief systems, we consider different cultural issues that figure prominently on the political agenda in many Western European countries. It is undisputed that issues revolving around immigration are salient in contemporary politics (Van der Brug et al. 2015). We therefore study how measures of ethnocentrism fit within cultural belief systems. Additionally, we include measures of moral progressiveness, namely progressiveness concerning gender roles and homosexuality. A progressive stance on these issues is currently explicitly used by PRRPs in various countries, such as Austria, Belgium, Denmark, Italy and the Netherlands, to defend their anti-immigrant position (Akkerman 2005; Betz and Meret 2009). We aim to explore how these issues relate to ethnocentrism in the cultural belief systems of PRRP voters.

We also include measures for religious orthodoxy and attitudes towards the role of religion in the public domain. This because research strongly suggests that religion plays a pivotal role in the way individuals structure political attitudes in general, and ethnocentrism and moral progressiveness in particular (Baldassarri and Goldberg 2014; De Koster and Van der Waal 2007; Wuthnow 1988). Religious orthodoxy indicates the extent to which one endorses Christianity’s central axioms, while acceptance of religion in the public domain pertains to the extent to which people accept that religiously inspired behaviours are actively conveyed and performed in public life (see Van Bohemen et al. 2012). Finally, we include rejection of Islam, which is a measure indicating the extent to which individuals regard this religion as problematic. This is relevant because issues related to Islam feature prominently in contemporary political debates, and a critique of Islam is a key element in the discourse of many PRRPs (Betz and Meret 2009), including in the Netherlands (Akkerman 2005: 346; De Koster et al. 2014). Moreover, it is unclear, a priori, whether rejection of Islam stems from either a progressive rejection of moral conservatism and/or ethnocentrism (Van Bohemen et al. 2012; cf. Elchardus and Spruyt 2014).

The constellation of these five cultural attitudes may vary, yielding different cultural belief systems among different segments of the public. For example, religion can result in the justification of both tolerance and intolerance towards outgroups (Davis and Robinson 1996). Therefore, religious orthodoxy can relate in different ways to progressive attitudes regarding immigrants and ethnic minorities.

Our empirical analysis aims to inductively uncover different cultural belief systems. Nevertheless, we can tentatively formulate four potential scenarios regarding the cultural belief system that may be upheld by PRRP constituencies. These cannot be tested statistically, but will aid interpretations of the inductively established cultural belief systems. The first scenario assumes a conventionally integrated cultural belief system among PRRP constituencies. This would resonate with the extant literature that explicitly assumes one-dimensionality of cultural attitudes among all segments of the public. Conservatism on one issue would go hand-in-hand with conservatism on other issues. This is plausible if the often-reported ethnocentrism of PRRP constituencies is embedded in a more general orientation. Such a generalised xenophobia – a fear of everything that is alien – not only expresses itself in a fear of foreigners, but also, for example, in a fear of homosexuals (Ignazi 1992; Mudde 1999). If this is correct, the stance towards all cultural attitudes used in the analysis will be intertwined, and a position on one of these issues is constrained by the positions on the others.

The second scenario also assumes that there is an integrated cultural belief system among PRRP constituencies, but with contrasting elements. This scenario is informed by the observations at the party level discussed above: resembling PRRPs in other Western European countries, PRRPs in the Netherlands combine ethnocentrism and a critique of Islam with a strong emphasis on gender equality and gay rights (Akkerman 2005; Betz and Meret 2009; De Koster et al. 2014). If this also characterises the cultural belief system of their constituencies, this would imply a strongly integrated cluster of cultural attitudes, but with negative relationships between ethnocentrism and the rejection of Islam on the one hand and moral conservatism, religious orthodoxy and acceptance of religion in the public domain on the other.

In the third scenario, ethnocentrism and rejection of Islam are clustered together, but are separate from other cultural issues. This reflects the primary focus of PRRPs on immigration. In fact, they are often labelled as ‘movements of exclusion’ (Rydgren 2005) or ‘anti-immigration parties’ (Van der Brug et al. 2000), ‘because their common denominator is that the immigration issue is their unique selling point’ (Van der Brug and Fennema 2007: 474). While issues related to immigration and immigrants seem to be part of a more encompassing, integrated cultural belief system at the party level, their constituencies may not display a similarly integrated profile because of the strong emphasis PRRPs put on issues related to immigration. In particular, if framing and agenda-setting at the party level play a role in shaping the public’s belief systems, the overwhelming focus on opposing immigration at the party level may lead to this third scenario. In addition, the way these issues are depicted by various media – part of the so-called ‘tabloidisation of the political discourse’ (Mudde 2013: 15) – may have a similar effect. Consequently, it is possible that ethnocentrism and rejection of Islam are closely intertwined among PRRP constituencies, but are not constrained by the positions on other cultural issues.

In the fourth and final scenario, all types of cultural attitude form separate clusters. This is inspired by the same reasoning as the third scenario, with the main difference being that ethnocentrism is detached from rejection of Islam. This scenario resonates with results of a large-scale quantitative content analysis of how migration-related issues are portrayed by Dutch media, which demonstrates ‘that the frame that focuses on Islam as a threat has been dominant in the media debate’ (Roggeband and Vliegenthart 2007: 535). Furthermore, this scenario is in line with the fierce criticism of Islam that has characterised Dutch PRRPs since the rise of the late Pim Fortuyn, and has radicalised with his political heir Geert Wilders and his PVV. Wilders’ PVV has been the most prominent PRRP in the Netherlands since 2006, and is known primarily because it vehemently opposes Islam. Wilders’ accompanying critique of progressive elites stems from his assessment that they have enabled a process of ‘Islamification’ (Vossen 2010). This strong, enduring emphasis on Islam is the theme rather than just a theme for Wilders (Vossen 2013: 105), which may lead his supporters to perceive rejection of Islam to be the key cultural issue, unconnected to other cultural matters. In fact, some scholars explicitly argue that ‘it is hard to classify the PVV as racist or even nativist as Wilders has not aimed his barrage at relatively well-integrated ethnic minorities such as the Surinamese, Moluccans, Chinese or Indo-Dutch’ (Vossen 2011: 185). If a similar cultural belief system characterises Dutch PRRP constituencies, this would mean that rejection of Islam is an issue in itself, and is not constrained by either ethnocentrism or other cultural issues.

Empirical strategy for uncovering the configurations of the cultural belief systems of PRRP constituencies

Following up on the analytical approach of Goldberg (2011) and Baldassarri and Goldberg (2014), we uncover different cultural belief systems that exist among the Dutch electorate. Subsequently, we study the association between voting preferences and socio-demographic indicators on the one hand and these belief systems on the other.

First, we use CCA to uncover different cultural belief systems. CCA is a correlational-based extension of Relational Class Analysis (Boutyline 2016a; Goldberg 2011). CCA partitions data in different clusters in such a way that the members of each cluster have similar patterns of association between variables. By applying CCA, we do not cluster individuals with similar cultural attitudes; rather, we cluster individuals whose relationships between their cultural attitudes are similar. This is illustrated in Figure 1. Individuals A, B and C (left-hand panel) all have very different cultural attitudes, but share the same belief system because the relationship between their attitudes is similar. Likewise, individuals D, E and F (right-hand panel) have different attitudes, but each of them shows convergent views regarding women and homosexuals, which diverge from their attitudes towards Turks and Moroccans. Indeed, to share a belief system ‘does not imply having identical attitudes or behaviors; rather, it suggests being in agreement on the structures of relevance and opposition that make actions and symbols meaningful’ (Goldberg 2011: 1402).

Figure 1. Two illustrative belief systems.

CCA does not make any a priori assumptions on the constraints characterising belief systems, nor does it impose one belief system on every individual in the data. This is where CCA deviates from conventional statistical methods. If, for example, we were to apply a factor analysis, we would uncover one structure characterised by one or more latent dimensions underlying the cultural attitudes. In so doing, factor analysis implicitly imposes one singular structure onto the data, assuming that the cultural attitudes of all individuals are structured according to these dimensions. CCA, on the other hand, divides data into groups of respondents who exhibit distinctive patterns across a set of variables. It thus clusters individuals whose attitudes on that set of variables are organised in a similar way (Boutyline 2016a; Goldberg 2011). Accordingly, CCA is perfectly suited to analysing individuals’ cultural belief systems. Our analysis reveals that three distinct cultural belief systems are present among the Dutch public.

Second, we study the relationship between preferences for party families and socio-demographic indicators, and the different cultural belief systems, with specific attention paid to individuals who prefer PRRPs. We reveal a strong association between party-family preferences and cultural belief systems. This association remains intact when controlling for socio-demographic variables, which themselves are also associated with cultural belief systems.

Data and measures

We use data collected by CentERdata in 2012 (Achterberg et al. 2012). This research institute maintains a panel of the Dutch population aged 16 and older while carefully preserving representativeness. Questionnaires are completed online. Respondents lacking internet access were provided with the necessary equipment. A sample size of 1302 was obtained with a response rate of 76.3%. These data allow us to address our research questions, as they cover an exceptionally wide variety of cultural attitudes.

Cultural attitudes. The five cultural attitudes outlined above were measured using five sets of five-point Likert items, ranging from ‘strongly disagree’ to ‘strongly agree’. We treated ‘don’t know’ answers as missing.¹ Listwise deletion resulted in 911 respondents being included in the CCA. To aid the interpretability of the results produced by the CCA – which is a graph-based method – we selected several items from well-established scales previously used in the Dutch context. The items indicating ethnocentrism are derived from a scale developed by Eisinga and Scheepers (1989), which proved reliable in recent research (Van Bohemen et al. 2012; De Koster et al. 2014; Van der Waal and De Koster 2015), just like scales used for borrowing items for moral progressiveness (De Koster et al. 2014; Van Bohemen et al. 2012), religious orthodoxy (De Koster and Van der Waal 2007; Immerzeel et al. 2013; Middendorp 1991), acceptance of religion in the public domain (De Koster et al. 2014; Van Bohemen et al. 2012), and rejection of Islam (Van Bohemen et al. 2012). Figure 2 lists the items used, including the labels representing them in the results section.

Figure 2. Visualisations of the different cultural belief systems.

Party families. We measure party preference using the following question: ‘Which party would you vote for if parliamentary elections were to be held tomorrow?’ To aid the interpretation of our results and comparability with previous research in the Netherlands (see De Koster et al. 2014), we categorised the answers into five party families: PRRP (6.5%), old right (18.3%), old left (32.9%), new left (21.9%) and Christian (20.4%). We coded PVV and the similar Democratisch Politiek Keerpunt / Trots op Nederland as PRRPs. The Conservatives (VVD) represent the old right. The new left is represented by the Liberal Democrats (D66) and the Greens (GroenLinks). The old left party family includes Labour (PvdA) and the Socialists (SP). Finally, the Christian category includes the Christian Democrats (CDA), the smaller Christian Union (CU) and the Reformed Political Party (SGP).

Socio-demographic characteristics. We include gender, age, education level (measured as the number of years formally required to attain one’s highest level of education), income (in €1000s) and religious denomination (none; Protestant; Catholic; other).

Results

Three distinct cultural belief systems

The CCA was performed using the corclass package in R (Boutyline 2016b) and yields three clusters, respectively representing 43.8%, 23.7% and 32.5% of the respondents. Note that looking at the mean scores of the included items (as is usually done in other clustering methods) would not help us to understand the nature of the clusters. This is because CCA clusters can include individuals with very different attitudes (but with the same structuring of these attitudes; see Figure 1). We therefore present each cluster’s correlational pattern of the included items, which we represent as networks to increase interpretability (the correlation matrices are available in the appendix). Nodes correspond to cultural attitudes and lines represent the correlations between them. Solid lines represent positive correlations and dashed lines negative ones. The width and shade of the lines is proportional to the strength of the correlation: wider and darker lines represent stronger correlations between the attitudes they connect. Only significant correlations are shown (p ≤ 0.05). The visualisations – presented in Figure 2 – are obtained using qgraph in R and by applying the Fruchterman‒Reingold algorithm (Epskamp et al. 2012).

Taking an overview of the three cultural belief systems, it is clear that their differences are particularly related to the position of the items regarding rejection of Islam and ethnocentrism.

The first cultural belief system uncovered by the CCA can be termed integrated: all cultural attitudes assessed are unified in a consistent, well-integrated system, as all the items are strongly correlated with one another. Accordingly, there are constraints between all attitudes making up the cultural belief system. As conservatively phrased items correlate positively with other conservative items and negatively with items indicating cultural progressiveness, this first cultural belief system closely matches the current state-of-the-art in the field, which suggests that various types of cultural attitude cluster together coherently.

A second cultural belief system clearly adopts an intermediate position between the first and third systems, as the attitudes regarding Islam take a quasi-isolated position. For those adhering to this cultural belief system, their stance vis-à-vis Islam is less constrained by their take on all the other cultural issues (compared to those adhering to the integrated belief system). Strikingly, it is the cultural issue that is most fiercely debated in both the media and political discourse that is least constrained.

The third cultural belief system can be termed partitioned, as it entails three isolated sets of cultural attitudes, respectively pertaining to: (1) rejection of Islam, (2) ethnocentrism and (3) acceptance of religion in the public domain, moral progressiveness and religious orthodoxy. For people characterised by this belief system, items measuring the latter set of issues combine in a well-integrated cluster, which is in line with studies utilising conventional methods for determining the structuring of the public’s political attitudes. Yet, contrary to the consensus in the literature, for the respondents adhering to this partitioned belief system, attitudes to the third set of issues do not constrain their other cultural attitudes (or vice versa). Indeed, for these individuals, ethnocentrism and stances towards Islam are almost not related. Accordingly, for those adhering to the partitioned belief system (a third of the sample), the widely shared assumption that cultural attitudes are interchangeable does not apply.

Having uncovered three distinct cultural belief systems, we assess whether these are unevenly distributed across supporters of the five party families, with specific attention paid to the PRRP constituencies.

A preference for PRRPs and cultural belief systems

As a first rudimentary step, we present a cross-tabulation of a preference for PRRPs on the one hand and the different belief systems on the other (Table 1). The association is very strong: 83% of those who prefer PRRPs adhere to the partitioned cultural belief system, while only 8.5% of those who prefer PRRPs adhere to the integrated version.

Table 1. Cross-tabulation PRRP preference and cultural belief systems (row percentages).

	Cultural belief system
	Integrated	Intermediate	Partitioned
PRRP preference	8.5%	8.5%	83.0%
Preference for other party^a	46.2%	24.7%	29.0%
Total	43.8%	23.7%	32.5%

Pearson chi²: 58.5585 (df = 2); p < 0.001.

^a all non-PRRP parties combined.

Next, we analyse the association between party-family preference and cultural belief systems, taking the socio-demographic characteristics of individuals into account. To do this, we run a multinomial logistic regression with cultural belief systems as dependent variable. Model 1 in Table 2 includes the effect of the party families, while Model 2 also encompasses the socio-demographic variables.

Table 2. Multinomial logistic regression on the cultural belief systems: logit coefficients.

	Model 1		Model 2
	Integrated (vs. partitioned)	Intermediate (vs. partitioned)	Integrated (vs. partitioned)	Intermediate (vs. partitioned)
Intercept	−2.277***	−2.277***	−5.339***	−1.681
Party family
PRRP	–	–	–	–
Old right	1.766**	1.073	1.646**	0.923
New left	4.107***	3.028***	4.025***	2.796***
Old left	3.216***	2.138***	3.335***	2.041***
Christian	1.584**	2.411***	2.172***	2.017***
Education	–	–	0.188***	−0.001
Income	–	–	0.149	0.085
Age	–	–	0.005	−0.007
Gender (female)	–	–	−0.206	0.368
Religion
None	–	–	–	–
Protestant	–	–	−0.947**	0.824
Catholic	–	–	−0.994***	0.561
Other	–	–	−0.656	0.897

^* p < 0.05

^** p < 0.01

^*** p < 0.001 (n = 726).

Model 1 shows a more nuanced picture of Table 1. Again, it is clear that the partitioned belief system is predominantly present among those who prefer PRRPs over other party families: all the logit coefficients are positive and significant. The integrated belief system is predominantly present among those who prefer leftist (both old and new left) political parties: compared to those who prefer PRRPs, the odds of having an integrated (rather than a partitioned) belief system are 60.8 (e^4.107) times higher for those who prefer new-leftist parties and 24.9 (e^3.216) times higher for those who prefer old-leftist parties. The differences between those who prefer PRRPs and those who prefer old-rightist and Christian parties are smaller, but still substantial ‒ i.e. the odds of having an integrated (rather than a partitioned) belief system are 5.8 (e^1.766) times higher for those who prefer old-rightist parties and 4.9 (e^1.584) times higher for those who prefer Christian parties, as compared to those who prefer PRRPs.

The results of Model 2 clearly show that the association between party-family preference and cultural belief systems is not confounded by the socio-demographic profile of the party-family constituencies: the coefficients of the party families hardly change when socio-demographics are added to the model. The high proportion of those adhering to a partitioned cultural belief system who prefer PRRPs over other party families is thus not the result of, for example, a lower level of education or male gender.

Focusing on the importance of socio-demographics for understanding the social bases of the three cultural belief systems, we see clear differences between the integrated and partitioned cultural belief systems in terms of education level. More highly educated individuals more often uphold an integrated belief system instead of a partitioned one: for every additional year of education, the odds of having an integrated cultural belief system instead of a partitioned one increase by 1.21 (e^0.188). Moreover, Protestants and Catholics have lower odds of having an integrated instead of a partitioned cultural belief system than those not belonging to a religious denomination (respectively, e^‒0.947 = 0.39 and e^‒0.994 = 0.37). Finally, adherence to belief systems is not associated with income, age and gender.

Discussion

Cultural belief systems prove to be very strongly related to party-family preferences. Our finding that a large majority of those who prefer PRRPs have a partitioned belief system, seems indicative of the influence political parties have on the electorate. This because the reverse causality seems highly unlikely: it is implausible that individuals develop party preferences as a result of a certain belief system. Individuals are likely to vote for specific parties because their attitudinal positions match party agendas, not because the structuring of their attitudes resembles that of parties. Given our results, we posit that the agenda-setting of the political party individuals prefer influences their cultural belief system. Our findings resonate with the fourth scenario developed in the theory, which stresses the importance of framing and agenda-setting by political parties and the media: rejection of Islam in particular is less (intermediate cultural belief system), or scarcely to not at all (partitioned one), constrained by other cultural attitudes. In the case of a partitioned belief system, a close connection between rejection of Islam and ethnocentrism is even absent.

More specifically, the finding that rejection of Islam plays a pivotal role in discerning the integrated cultural belief system from the partitioned one aligns with suggestions that PRRPs can weigh on the political field, even if they are located in the opposition (Han 2015; Minkenberg 2001; Mudde 2013). That is, PRRPs may succeed in imposing a certain logic which dictates the salience of issues regarding Islam. If this underlies the constellation of the intermediate and partitioned cultural belief systems, this stresses the agenda-setting abilities of PRRPs, despite their relatively modest electoral success (see Jackman and Volpert 1996: 502–503). After all, not only the large majority of the PRRP electorate, but also 29% of the individuals who reported a preference for a non-PRRP adhere to a partitioned belief system. In addition, the intermediate cultural belief system – which is not strongly associated with a preference for PRRPs – is also characterised by a non-conventional take on issues of Islam. This resonates with issue-competition theory which stresses that political parties emphasise issues to make them dominant in electoral competition and to force other parties to address these issues (Carmines and Stimson 1993).

The tabloidisation of political discourse (Mudde 2013: 15) provides an additional interpretation of the pivotal role of stances towards Islam in discerning the three cultural belief systems and their unequal distribution over party-family constituencies. While criticism of Islam is not the sole element of the political agenda of PRRPs, Geert Wilders’ fiercely voiced critique may be exceptionally visible to the public because of the media attention paid to it. This is not only because the theme features prominently in Wilders’ own media messages (De Landtsheer et al. 2011), but also because Dutch media in general overwhelmingly focus on Islam when reporting on issues of migration and integration (Roggeband and Vliegenthart 2007). This may explain why attitudes towards Islam have a remarkably isolated position in the partitioned cultural belief system, not even closely connected to ethnocentrism. In addition, it may ensue from the fact that a one-sided focus on Islam as a threat is most prominent among popular media (Roggeband and Vliegenthart 2007), which are preferred by supporters of PRRPs (Bos et al. 2014). Future research is needed to study whether, and how, media representations of political issues and patterns of media consumption account for the unequal distribution of cultural belief systems over different parts of the electorate.

Next to the possible influence of political parties and media representations, other factors contribute to variation in belief systems. Especially education and religion – rather than the structural factors of income, age and gender – appear to be influential. The role of education for belief systems suggests that Converse’s (1964) idea that political competence results in more integrated belief systems might be valid. Our analysis does indeed show that a higher level of education, which is closely related to political competence (Delli Carpini and Keeter 1996), is associated with adhering to an integrated cultural belief system.²

The finding that an encompassing Christian worldview does not inspire an integrated cultural belief system might seem counter-intuitive. Yet the specific constellation of the partitioned cultural belief system provides an interpretation. Note that issues sanctified by Christianity – moral conservatism, religious orthodoxy and acceptance of religion in the public domain – are integrated within a sub-cluster in the partitioned cultural belief system, but do not constrain ethnocentrism and rejection of Islam. Such a constellation of cultural attitudes among Catholics and Protestants might indicate ambivalence towards Islam (Wiegers 2012). One the one hand, similar dogmas and institutional accommodation of religious minorities make Christians and Muslims political allies. On the other hand, many, especially orthodox, Christians regard Islam as ‘alien to the Netherlands and the Western culture’ (Wiegers 2012: 25). These cross-pressuring factors may account for the greater affinity with the partitioned belief system (instead of the integrated one) among Christians. However, more research is needed to increase our understanding of the exact mechanisms through which religion influences cultural belief systems.

Conclusion

We studied the structuring of cultural attitudes among the Dutch public, with special attention paid to those who prefer PRRPs. We applied a novel analytical approach, which recognises heterogeneity in the ways individuals understand and organise their cultural attitudes (Baldassarri and Goldberg 2014; Boutyline 2016a; Goldberg 2011). Our analyses inductively uncovered three ‘cultural belief systems’ – i.e. specific constellations of political attitudes towards cultural issues – among the Dutch public: ‘integrated’, ‘intermediate’ and ‘partitioned’.

The integrated cultural belief system is in line with findings based on conventional methods, like factor analysis, which impose a similar structuring of cultural attitudes on all research subjects. More conservative views on a specific cultural issue go together with more conservative stances on other cultural issues. Accordingly, in the integrated cultural belief system, individuals consider the different cultural issues to be logically connected and composing a unity. In contrast, for those with a partitioned cultural belief system, the cultural dimension clearly does not represent a coherent monolithic entity. This remarkable phenomenon was hidden from view by previous studies on the structuring of political attitudes. Because a substantial share of individuals adheres to an integrated belief system, conventional analyses mask differences in the degree to which individuals’ belief systems are integrated.

This finding informs the debate regarding electoral competition in the cultural domain. Does this revolve around single issues (e.g. Green-Pedersen 2007) or around a coherent cultural dimension (e.g. Kitschelt 1995)? Our results indicate that both views are, to some extent, correct: whereas PRRP supporters in the Netherlands exhibit a partitioned cultural belief system reflecting the issue-competition approach, other, especially leftist, electorates have an integrated cultural belief system reflecting the dimensional interpretation. This integrated belief system resonates with the way cultural issues are constrained in most party agendas: in the political agendas of non-PRRPs, progressiveness (conservatism) on one of those issues is typically accompanied by progressiveness (conservatism) on the others (Lefkofridi et al. 2014).

Overall, inspired by Baldassarri and Goldberg’s (2014) work on the American electorate, our findings on the Dutch context highlight the vital importance of acknowledging the existence of different belief systems in contemporary politics. More specifically, our study evidences that two out of three cultural belief systems strongly deviate from the one-dimensional structuring of cultural attitudes that has conventionally been assumed in decades of research (Achterberg and Houtman 2009) and in which ‘cultural issues are considered interchangeable’ (De Koster and Van der Waal 2007: 452). We focused on the Dutch case because it is theoretically salient (given a PRRP agenda that combines conservative and progressive cultural stances) and empirically feasible (given available data on a wide range of relevant cultural attitudes). We hope that our findings inspire studies in other countries when suitable data become available. This would enable the uncovering of variation in cultural belief systems in countries with and without PRRPs in parliament, and cross-national comparisons would aid uncovering the generative mechanisms behind them. The role of PRRP agendas and media representations of Islam for inspiring non-conventional cultural belief systems could, for instance, be more systematically scrutinised. More generally, we are convinced that future research should strive to achieve a more in-depth understanding of the ways individuals perceive, understand and structure political issues, in addition to the common focus on the causes and consequences of people’s positions on these issues. We hope this article serves as an impetus to study the development, distribution, backgrounds and implications of belief systems in political research.

Notes on contributors

Stijn Daenekindt is a Postdoctoral Researcher in the Department of Public Administration and Sociology at Erasmus University Rotterdam and the Department of Sociology at Ghent University. He holds an MSc in Statistics (KU Leuven, 2012) and a PhD in Sociology (Ghent University, 2015). His interests lie in cultural sociology and political sociology. [[email protected]]

Willem de Koster is Associate Professor of Cultural Sociology in the Department of Public Administration and Sociology at Erasmus University Rotterdam. Much of his research addresses the genesis, manifestations, and consequences of cultural conflict in Western countries. For more information, see http://www.willemdekoster.nl. [[email protected]]

Jeroen van der Waal is Associate Professor of Sociology in the Department of Public Administration and Sociology at Erasmus University Rotterdam. Much of his research aims to explain why social stratification is linked to political attitudes and behaviours and health disparities in Western societies. In addition he studies the consequences of globalisation and cultural change for inequalities, political attitudes and behaviours. For more information, see http://www.jeroenvanderwaal.com. [[email protected]]

Supplemental data

The supplemental material for this paper is available online at http://dx.doi.org/10.1080/01402382.2016.1271970

Disclosure statement

No potential conflict of interest was reported by the authors.

Acknowledgements

We are grateful for valuable comments on previous versions of this manuscript from Roy Kemmers, Julian Schaap, Josje ten Kate, Jochem van Noord and the anonymous reviewers.

Notes

1. We refrain from using imputation, as this is fundamentally at variance with the relational nature of CCA by assigning individuals scores based on aggregate statistics (see Baldassarri and Goldberg 2014). Additionally, four respondents were removed from the analysis as they completed the survey in less than 10 minutes.

2. An alternative interpretation of the finding that less educated individuals are more likely to adhere to a partitioned belief system is that they perceive issues regarding ethnocentrism as economic rather than cultural issues. However, this idea is not corroborated by empirical research (Hainmueller and Hopkins 2014; Van der Waal and De Koster 2015). Re-estimating the models in Table 2 is also at odds with this alternative interpretation: including a measure for job insecurity proves that labour market position is unrelated to belief systems.