Introduction
The transformation of Western European political spaces, marked by structural changes, value shifts and extensive electoral realignment, has been under the academic limelight for decades. Most commonly, these transformations are understood to have been fostered by globalization, which foregrounded debates on immigration, European integration and national sovereignty, and led to the mobilization of green and far-right parties (Bornschier, Reference Bornschier2010a). While the perforation of globalization has provoked comparable transformations of national political spaces in Western Europe (e.g. Bornschier and Kriesi, Reference Bornschier, Kriesi and Rydgren2013; Kriesi et al., Reference Kriesi2006; Reference Kriesi2008; Hooghe et al., Reference Hooghe2002), the case of Belgium – where the far right only managed a breakthrough in one of two federal regions – remains a striking anomaly. Using post-electoral survey data from the 2019 national elections, this article seeks to enhance our understanding of the Belgian political space by comparing the demand side of the Flemish political space (where a right-wing populist party has become firmly entrenched) to that of Francophone Belgium (where the far right remained absent thus far).Footnote 1 Specifically, we seek to investigate whether Flanders and Francophone Belgium, in view of their supply-side differences, differ along the two main components of a political space: their political dimensions and their cleavage structure.
We define political dimensions as latent polarities shaped by political attitudes that structure ideologies (cf. Wheatley and Mendez, Reference Wheatley and Mendez2021). Historically, research has revealed a two-dimensional structure in Western Europe, including a cultural dimension (encompassing attitudes to law and order, ethics, authoritarianism and religion) and an economic dimension (encompassing issue positions on the welfare state, inequalities and state intervention in the economy) (Kitschelt, Reference Kitschelt1994; Middendorp, Reference Middendorp1978). Bearing in mind the electoral rise of left-liberal and far-right parties, scholars have extensively studied and reconceptualized this dimensional structure, with some suggesting a possible third materialism-postmaterialism (Swyngedouw, Reference Swyngedouw1995), European (Bakker et al., Reference Bakker2012) or environmentalist (Kenny and Langsæther, Reference Kenny and Langsæther2023) dimension. Others argue that realignment has not fundamentally altered the two-dimensional structure and that new issues such as immigration, European integration and environmentalism are effectively integrated into these established dimensions (Kriesi et al., Reference Kriesi2006; Reference Kriesi2008).
Additionally, we consider cleavages to be more organized and enduring divisions, arising from structurally defined social groups with opposing political interests. Cleavages bear the imprint of a society’s historical experience and, as such, have transformed salient social divisions into stable patterns of voting behaviour (Bartolini and Mair, Reference Bartolini and Mair1990; Lipset and Rokkan, Reference Lipset and Rokkan1967). While it is assumed that traditional class and religious cleavages continue to structure most Western European political spaces, the emergence of left-liberal and far-right parties is associated with a new ‘universalism-particularism’ cleavage that pits winners against losers of globalization (Bornschier, Reference Bornschier2010a; Reference Bornschier2010b; Zollinger, Reference Zollinger2024). Thus, while political dimensions illustrate how voters connect different issues into consistent ideological thinking, cleavages relate more directly to voting behaviour and the routinized socio-structural divisions underpinning it.
The current paper takes into account both concepts to explain the supply-side differences between Flanders and Francophone Belgium from a demand-side perspective. In doing so, we assess whether the distinct electoral trajectories of the two regions are undergirded by ideological or socio-structural differences among voters. Over the past few decades, cultural and political segmentation between linguistic communities has led to the formation of two separate party systems that have undergone distinct political sea changes (De Winter et al., Reference De Winter2006). While Flanders witnessed the emergence of the far-right Vlaams Belang party, which secured almost a fourth of the (Flemish) vote share in the 2024 federal elections, the Francophone party system has retained a more traditional bipolar structure that is characterized by the competition between a prominent socialist (PS) and liberal (MR) party (de Jonge, Reference de Jonge2021). Despite modest scholarship on the asymmetrical supply side in Belgium (e.g. Coffé, Reference Coffé2008; de Jonge, Reference de Jonge2021; van den Berg and Coffé, Reference van den Berg and Coffé2012), there has been no comparative analysis that reveals how cleavages and dimensions are structured in the two linguistic groups. In addressing this dynamic, this article taps into a broader debate on the transformation of the Belgian political space, by inquiring whether the spaces of both regions developed asymmetrically (cf. Billiet et al., Reference Billiet2021; Delespaul and Swyngedouw, Reference Delespaul, Swyngedouw and Abts2023). Ultimately, such comparison may also enhance our understanding of the uneven pace of far-right mobilization in similar Western European countries.
To compare the political dimensions and cleavage structures of Flanders and Francophone Belgium, we adopt a sequential design using multiple correspondence analysis (MCA) and the recently developed method of class-specific multiple correspondence analysis (CSA) (Le Roux, Reference Le Roux2014). The steps involved in this design are threefold: (1) to construct, using an MCA on the full sample of Belgian voters, the dimensions along which political attitudes are ordered in a national space; (2) to verify, through a CSA on the Flemish and Francophone subsample, whether these dimensions reflect similar polarities in both regions; and (3) to project, by plotting distances between social groups in the initial national space, unique ‘region-specific’ cleavages for Flanders and Francophone Belgium. This analysis is based on survey data from the Belgian National Election Study 2019 (BNES; n = 1,150), specifically collected for this research purpose.
In the next section, we discuss the literature on political dimensions and cleavages in Western Europe, their transformation in recent decades and the concrete manifestation of these transformations in Belgium. We then introduce our methodological approach, after which we provide a discussion of the results in three steps. We find a remarkable commonality between the Flemish and the Francophone political space, with a Francophone left-wing welfare consensus emerging as the most pronounced difference between the two communities. Hence, we conclude that the asymmetric political trajectories of Flemish and Francophone electorates are not manifested in a different cleavage structure – nor in a different political dimensionality – which is germane to our understanding of contemporary communitarian conflict in Belgium.
Theoretical background
Political dimensions and cleavages
It has been extensively argued, both theoretically and empirically, that large-scale societal transformations in the early 1980s have changed the structure of Western European political spaces (Bornschier, Reference Bornschier2010a; Kriesi et al., Reference Kriesi2006; Reference Kriesi2008; Oesch and Rennwald, Reference Oesch and Rennwald2018). These changes are commonly attributed to processes of economic and cultural modernization, expressed in increased occupational and spatial mobility, rising levels of education and value changes; all of which have fragmented closed social milieus and disentangled the socio-structural ties that once bound individuals to mass parties (Abts, Reference Abts2012). The detachment of voters from their socio-structural moorings has narrowed the social basis of the old cleavages that embodied them, fundamentally weakening their capacity to structure political behaviour (Franklin et al., Reference Franklin1992). Equally, there is a host of evidence to suggest that this dealignment away from traditional cleavages was followed by a realignment along new structural antagonisms, driven by the mobilization of new-left and far-right parties (e.g. Billiet and Swyngedouw, Reference Billiet and Swyngedouw1995).
The political ramifications of these transformations surfaced more quickly in some countries (e.g. Austria, France, Switzerland) than others (e.g. Germany, Spain, Sweden) and these diverse consequences have been especially apparent in Belgium. Indeed, despite similar institutional settings, party system change has been extensive in Flanders, where far-right and conservative parties grew steadily out of anti-immigration sentiments, but not in Francophone Belgium, where the party system has remained frozen (with the emergence of the left populist PTB in 2014 as a notable exception) (de Jonge, Reference de Jonge2021; De Winter et al., Reference De Winter2006).
The main focus of this contribution lies with the demand side of Belgian politics, and pertains in particular to the commonalities and divergences existing between the political spaces of Flanders and Francophone Belgium. We contend that the asymmetrical developments on the supply side of these political spaces should be anchored in the structure of the demand side as well. Cleavage theory posits that historical social conflicts divide voters along durable issue dimensions, that parties mobilize voters by adopting a clear positioning along these dimensions and that strong alignments between parties and voters are stabilized by collective identities (Bornschier, Reference Bornschier2010a; Hooghe and Marks, Reference Hooghe and Marks2018). Party system change occurs when interactions in the party system no longer reinforce the underlying dimensions of voters, thereby amplifying the mobilization potential of new cleavages (Kriesi et al., Reference Kriesi1995). Therefore, we argue that the differences in the supply side between Flanders and Francophone Belgium should be reflected at the demand side, either through (1) a different composition of voters’ political dimensions, or (2) the position of electorates vis-à-vis salient historical divides between social groups, known as cleavages. Taken together, these two concepts have been applied extensively to analyse the structuration of national political spaces elsewhere in Western Europe (e.g. Bornschier, Reference Bornschier2010a; Kriesi et al., Reference Kriesi2006; Reference Kriesi2008).
First, regarding the dimensionality of political spaces, the common consensus is that Western European electorates are framed along two dimensions (but see, for example, Bakker et al., Reference Bakker2012; Bartolini and Mair, Reference Bartolini and Mair1990; Wheatley and Mendez, Reference Wheatley and Mendez2021). This belief can be traced back to Lipset (Reference Lipset1960) who distinguished between economic and cultural political attitudes that have crystallized into a set of separate political dimensions. The economic dimension is the expression of a ‘democratic class struggle’, tapping into issues of wealth redistribution and economic governance, while the cultural dimension reflects the different social outlooks of ‘tough-’ and ‘tender-minded’ individuals on matters such as civil liberties, religious influence, adherence to democratic ideals and internationalism. This two-dimensional structure has been validated at several points in time – also for the case of Belgium (Swyngedouw, Reference Swyngedouw1995) – although the nature of the second ‘cultural’ dimension tends to vary between different accounts. For example, Inglehart (Reference Inglehart1977) introduced an alternative ‘materialism-postmaterialism’ dimension, while Kitschelt (Reference Kitschelt1994) and Flanagan and Lee (Reference Flanagan and Lee2003) laid bare a ‘libertarian-authoritarian’ dimension – both emphasising conflicts over lifestyles, self-determination and the quality of physical and social life. Other conceptualizations of the cultural dimension, such as Hooghe et al.’s (Reference Hooghe2002) ‘GAL-TAN’ dimension, have demonstrated that debates over European integration and immigration are integrated into this second dimension. The enlarged cultural dimension has since surpassed the economic dimension as the main axis of competition in Western politics (Jackson and Jolly, Reference Jackson and Jolly2021), although – crucially for our research purposes – the composition of these dimensions may vary between (national) political arenas (Wheatley and Mendez, Reference Wheatley and Mendez2021).
Second, while the dimensionality of the demand side matters to demarcate the ideological contents of political competition (Bornschier, Reference Bornschier2010a), the structure of the party system itself is rooted in the configuration of social groups along these dimensions. Lipset and Rokkan (Reference Lipset and Rokkan1967) famously postulated that Western European party systems are organized through long-standing social conflicts, known as cleavages, that sustain stable political behaviour. Although the empirical requirements of a cleavage have been the subject of lengthy academic debate, it is generally believed that it describes a particular division characterized by an overlap of a socio-structural element (i.e. a distinct social group), an empirical element (i.e. shared beliefs that provide a sense of self-consciousness) and an organizational element (i.e. a group or party that represents its interests) (Bartolini and Mair, Reference Bartolini and Mair1990). In a review of Western European party systems in the 1960s, Lipset and Rokkan themselves identify four archetypical cleavages: two of which arising from the expansion of the nation-state and its bureaucracy (i.e. the centre-periphery cleavage and the religious cleavage) and two rooted in the Industrial Revolution (i.e. the class cleavage and the urban-rural cleavage); with the former three cleavages materialising in Belgium. Cross-country variations in specific cleavage structures (i.e. the configurations of these cleavages in a political space), have long been considered to be the main source of supply-side differences in European democracies. The most recurring explanations for these cleavage variations hinge on context-specific factors, such as welfare state types, electoral systems or specific historical trajectories of social groups, that determine the continued relevance of old versus new cleavages (Brooks et al., Reference Brooks2006; Elff, Reference Elff2007).
The transformation of Western European political spaces
Before turning to the discussion of the demand-side factors that underpin supply-side differences in Belgium – that is, the main focus of this paper –, we first discuss the contextual dynamics that brought about political transformations in similar Western European countries. We argue in particular that the absence of the far right in Francophone Belgium bears the signs of an asymmetrical transformation of the Belgian political space, where the Flemish space experienced large-scale transformations while the Francophone space remained frozen by the continued influence of traditional cleavages.
Indeed, the mobilization of the far right has been found to constitute a transformative event, as numerous studies have shown that its mobilization instils higher voter polarization (Bischof and Wagner, Reference Bischof and Wagner2019; Castanho Silva, Reference Castanho Silva2018), fosters the politicization of cultural demarcation issues (Akkerman, Reference Akkerman2015; Minkenberg, Reference Minkenberg2001) and facilitates the formation of new group identities (Bornschier et al., Reference Bornschier2021; Zollinger, Reference Zollinger2024). Overall, the programmatic innovations of these parties reconfigured political spaces towards a new conflict that focalizes on the role of the community in a context of increasing globalization (Abts, Reference Abts2012; Swyngedouw, Reference Swyngedouw1995). At the heart of this conflict lies a question of citizenship, and more specifically, of who belongs to the national community and of the criteria that determine this membership. ‘Universalism’, which is driven by the new left, adheres to Enlightenment values and a notion of universal human rights, while ‘particularism’, which is embodied by the far right, envisions a culturally homogeneous society, advocating cultural monism and the protection of particularistic community values through a right of cultural differentialism (Betz, Reference Betz2004). Hence, the new left and the far right propagate opposite normative ideals that are predominantly linked to the epochal challenges of globalization and denationalization (Bornschier, Reference Bornschier2010b). As a result, it is widely assumed that this universalism-particularism conflict instigated a wholesale transformation of political spaces in Western Europe: on the one hand, by driving a rearticulation of political dimensions, and on the other, by exacerbating the formation of a new ‘universalism-particularism’ cleavage.
As for their dimensionality, the rearticulation of Western European political spaces unfolded in two stages: while the mobilization of the new left in the 1980s restructured the cultural dimension towards a conflict between libertarianism and authoritarianism, concerns about globalization, fuelled by the far right, provoked a transformation of both dimensions along heuristics of integration and demarcation. Most Western European political spaces thus remained defined by an economic and a cultural dimension – the so-called ‘basic structure’ –, but the substantive content of these political dimensions changed profoundly (Kriesi et al., Reference Kriesi2006; Reference Kriesi2008). On the economic dimension, globalization amplified the polarization between those seeking integration into global markets and those advocating protectionist policies. This shift was further exacerbated by the maturation of the welfare state which foregrounded debates over its redistributive design instead of over its expansion (Pierson, Reference Pierson2001). On the cultural dimension, new integration issues pertaining to immigration, multiculturalism and European unification took centrer stage, constituting an enlarged cultural dimension that also covers conflicts over value liberalization (Kitschelt, Reference Kitschelt1994), new politics (Inglehart, Reference Inglehart1977) and welfare deservingness and misuse (Häusermann and Kriesi, Reference Häusermann, Kriesi and Beramendi2015).
Beyond altering the dimensionality of political spaces, the universalism-particularism conflict also seemed to bear the imprint of a new structural divide, distinguishing left-libertarian and far-right electorates. The argument here is that globalization has caused certain structural groups to lose out in terms of life chances and privileges, while others have benefitted from increased mobility and the acceptance of alternative lifestyles (Bornschier, Reference Bornschier2010b). As pointed out by Kriesi et al. (Reference Kriesi2008), winners and losers can be distinguished by their level of education, because those with longer education possess ‘specialised skills which are marketable inside and across the national boundaries’, thereby ‘considerably increasing one’s exit options’ (p. 7). In this connection, the effect of education on voting behaviour has increased in recent decades and is particularly strong with respect to voting for left-libertarian and far-right parties (Marks et al., Reference Marks2023). Moreover, a burgeoning literature demonstrates that the winner-loser dichotomy is rooted in distinctive social identities (e.g. Bornschier et al., Reference Bornschier2021; Steiner et al., Reference Steiner2024; Zollinger, Reference Zollinger2024), promoting the view that a new ‘universalism-particularism’ cleavage has indeed materialized in rearticulated political spaces of Western Europe, complete with socio-structural, empirical and organizational foundations. An extensive analysis by Zollinger (Reference Zollinger2024) aptly shows that this new cleavage is linked both to attitudes on the cultural dimension (such as anti-immigration and Eurosceptic attitudes), but equally to attitudes on the economic dimension (such as support for social investment).
Bearing in mind this literature, the emergence or absence of the far right can be viewed as a harbinger for changes in political demand. Indeed, the Lipset-Rokkan narrative posits that shifts in political spaces occur due to ‘critical junctures’ catalysed by social transformations, such as globalization, societal modernization and deindustrialization, that generate new conflicts determining the mobilization and the policy priorities of parties (De Vries and Marks, Reference De Vries and Marks2012). From this bottom-up perspective, far-right party mobilization hinges on the formation of new structural conflicts and the waning salience of established cleavages, while its non-materialization depends on the conjunction of a higher salience of economic conflicts and lower mass polarization along the cultural dimension (Bornschier, Reference Bornschier2010a). Consequently, one might expect that the political spaces in which the far right failed to gain a foothold (e.g. Francophone Belgium) have remained frozen along older conflicts, with the traditional dimensions and cleavages, in spite of new critical junctures.
The case of Belgium: an asymmetrical transformation?
Belgium’s unique political constellation provides an ideal case to assess these hypotheses. The country can best be described as a ‘federal state without federal parties’: parties operate exclusively in one of two federal entities – Flanders and Francophone Belgium – and primarily cater to the needs of their respective linguistic community. Thus, although Flemings and Francophones are socialized in the same institutional setting, they also exhibit clear cultural differences that have crystallized into two distinct party systems with very different electoral trajectories (De Winter et al., Reference De Winter2006). Flanders, the larger and wealthier region, has seen a clear realignment to the right, highlighted by the upsurge of the centre-right Flemish nationalist N-VA and the far-right Vlaams Belang (previously ‘Vlaams Blok’). The latter has been a significant electoral force since its breakthrough in the early 1990s, becoming the region’s second largest party after the 2019 elections. By comparison, the Francophone party system has been less volatile (Swyngedouw et al., Reference Swyngedouw and Abts2023). Despite the region’s sharp industrial decline, the social-democratic party PS has held onto its dominant position among production workers, while simultaneously retaining the support of socio-cultural professionals and other middle-class segments (Gaasendam, Reference Gaasendam2020). Although the Francophone far right gained parliamentary representation in 1991, it never managed to cement its position and disappeared altogether after the 2019 federal election.
Peculiarly, as Hossay (Reference Hossay and Schain2002) has pointed out, most conventional predictors of far-right support would expect the far right to be stronger in Francophone Belgium: Francophones are on average less educated, face higher unemployment rates and have experienced a higher influx of foreigners (de Jonge, Reference de Jonge2021) (Figure 1). In their socio-political outlook, Francophones are as sceptical of immigration (Billiet et al., Reference Billiet2021) and European integration (Delespaul et al., Reference Delespaul2022) as Flemings, but are more likely to harbour feelings of social resentment or express criticism of the welfare state (Van Hootegem et al., Reference Van Hootegem2021). In a more detailed overview of public opinion data, Billiet et al. (Reference Billiet and Rihoux2015) conclude that the stereotypical image of ‘right-wing Flemings’ and ‘left-wing Francophones’ is largely overstated, given the marginal differences in attitudes between the two parts of the country.
Figure 1. Election results in Flemish (left) and Walloon Region (right).
Multiple studies have thus claimed that supply-side differences in Belgium are reflected in the particular structure of the Flemish and Francophone political spaces. Often, this argument rests on the idea that the traditional class cleavage has remained more salient in Francophone Belgium, preventing political mobilization along new cleavages (Coffé, Reference Coffé2008; Gaasendam, Reference Gaasendam2020). Several comparative studies speak to this claim, showing stronger class voting (De Winter et al., Reference De Winter2006; van den Berg and Coffé, Reference van den Berg and Coffé2012) and higher salience of economic issues in Francophone Belgium (Coffé, Reference Coffé2008). Francophone parties have played a pivotal role in maintaining this status quo: on the one hand, by taking up a more outspoken position on the economic dimension, and on the other, by ostracising the far right through a rigid cordon sanitaire (de Jonge, Reference de Jonge2021).Footnote 2
Data and analytical strategy
To sum up our approach, we empirically explore whether the asymmetric transformation of the Belgian party system – with the uneven breakthrough of the far right as one of the key characteristics – will be reflected in an asymmetrical structure of the political spaces of its two regions. Our analysis aims to lay bare this asymmetry by comparing the Flemish and Francophone spaces along their political dimensions and cleavages – both of which are expected to reflect the main lines of opposition in their respective party systems.
The ensuing analysis uses data from the 2020 wave of the Belgian National Elections Study (BNES), a two-stage random probability survey conducted after the 2019 Belgian federal elections (Abts et al., Reference Abts2021b). Fieldwork took place between December 2019 and October 2020, a period during which interviews had to be interrupted due to COVID-related lockdown measures.Footnote 3 From the initial sample (n = 1,659, 32.8% response rate), only those respondents who completed an additional paper-and-pencil questionnaire are included in the analysis (n = 1,150).
To compare the structure of the Flemish and the Francophone political spaces, we use multiple correspondence analysis (MCA) and class-specific multiple correspondence analysis (CSA). MCA is an exploratory technique that examines relationships among multiple variable categories. It does so by translating patterns of association and opposition into a structured distance matrix, which is then used to extract orthogonal axes or dimensions that cover a portion of the total variance in the data (Atkinson, Reference Atkinson2024; Hjellbrekke, Reference Hjellbrekke2019). While MCA establishes a synoptic ‘space’ of associations, CSA re-estimates this initial space for a specific subset of the data, uncovering the similarities of the subsets’ relational structure in comparison to the full sample. More specifically, CSA generates class-specific subspaces by recalculating axes that better fit the structure of the subset, while preserving the distances of the MCA reference space (see also Le Roux and Rouanet, Reference Le Roux and Rouanet2010; Le Roux, Reference Le Roux2014). Because the subspaces remain nested in the reference space, the dimensions emerging out of the CSA can be directly compared to the dimensions of the MCA in terms of interpretation and explained variance. In our approach, we use MCA to construct the political dimensions of a Belgian reference space and CSA to assess whether these dimensions are equivalent for Flemish (Dutch-speaking) or Francophone (French-speaking) respondents.
We explore this dimensionality through a set of 17 attitudinal variables, which can be organized into three blocks. First, a set of economic variables, including items gauging the support of trade unions, unemployment benefits, a lower retirement age, wealth taxes, economic globalization and welfare state criticism. Collectively, we expect these items to reconstruct an enlarged economic dimension akin to the one identified by Kriesi et al. (Reference Kriesi2006), revolving around the state-market conflict, (the conditionality of) social welfare provisions and labour market protectionism (Gaasendam, Reference Gaasendam2020). Second, a set of cultural variables is included, consisting of constructs measuring opposition to immigration, support of gay rights, freedom of speech and freedom of religion and welfare chauvinist and authoritarian beliefs. These variables are anticipated to constitute a cultural dimension, which weighs values of cultural openness, self-expression and progressivism against cultural protectionism, social conformism and traditionalism (Inglehart and Baker, Reference Inglehart and Baker2000). Third, we insert a set of mixed variables, containing attitudes on the environment, regional autonomy (or ‘communitarianism’), individualism, European integration and the choice between freedom or equality. From the extensive literature on political dimensionality, we expect the variables in this block to either be multidimensional, to not contribute to any dimension, or to crystallize into a distinctively new dimension.
Variables in an MCA contribute to the total variance by their number of categories. Therefore, we factorized every variable into ordinal variables with four categories, with each accounting for approximately 25% of the responses. In addition, the number of categories in the economic and cultural set are deliberately equalized (k = 24) to ensure that both sets of variables contribute equally to the total variance in the space (cf. Hjellbrekke, Reference Hjellbrekke2019, p. 94). We also set missing categories to ‘passive’, meaning they remain in the data but do not contribute to the construction of the space. An overview of the question wording, recoding procedure and measurement of the items can be found in Table B1 of online Appendix B.
Next, after assessing the dimensionality of the Belgian political space, we reconstruct the Flemish and Francophone cleavage structures by charting the correspondence of social structure and voting behaviour along the main political dimensions identified in the previous step. In MCA, this is achieved by including non-contributing points, so-called ‘supplementary categories’, which reflect the mean position of a particular category of respondents in the geometrical space. The purpose of this projection is to reveal the correspondences between different sets of variables; in our case between the political attitudes used to chart the space and the variables measuring social structure and vote choice. Thus, by mapping the interplay between social position, voting preferences and ideological stances, we expose a relational dynamic familiar to the one identified by cleavage theory (cf. Lindell and Ibrahim, Reference Lindell and Ibrahim2021; Westheuser, Reference Westheuser2021).
To visualize these cleavage structures, we incorporate five supplementary variables: vote choice (at the 2019 general election) and four socio-structural variables that are commonly associated with a cleavage. For the Belgian case, these are ‘church attendance’ (for the religious cleavage), ‘income’ (for the class cleavage), ‘(sub-)national identity’ (for the linguistic cleavage) and ‘education level’ (for the universalism-particularism cleavage). Although these objective categories offer no direct proxies for the collective identities fostering cleavage formation (cf. Zuckerman, Reference Zuckerman1975), empirical evidence indicates that voters (still) draw in- and out-group boundaries along these socio-structural components (see Bornschier et al., Reference Bornschier2021; Steiner et al., Reference Steiner2024; Zollinger, Reference Zollinger2024).
To compare the cleavage structures across the two regions, we segment the supplementary categories in Flemish and Francophone categories. While our approach involves charting cleavages along the retained dimensions, comparing the Flemish and Francophone supplementary points is possible only in the Belgian reference space – where respondents belonging to both subsets are structured along the same set of axes – and not in their respective subspaces which have their own unique set of axes. However, a joint projection of the Flemish and Francophone cleavage structures in the reference space is only precise if their class-specific dimensions align closely with the ones of the reference space. Specifically, if the first step of our analysis indicates that MCA- and CSA-axes overlap, this would signal that respondents belonging to the subsets are equally well-represented in the Belgian reference space as in their own class-specific space. In this case, supplementary categories that are specific to Flanders (e.g. the Flemish low-income voters) could be compared to all other Francophone supplementary points within the same (Belgian) space.
Summing up, our approach consists of three steps. First, we estimate a Belgian reference space through an MCA involving all respondents, extracting the political dimensions that best represent the main polarities in the data. Second, we run separate CSAs for the Flemish and Francophone respondents, verifying whether their respective subspaces share the same dimensional structure. Third, if dimensions are found to be overlapping, we project Flemish and Francophone social groups and voting preferences as supplementary points in the reference space to compare regional cleavage structures.
Results
Results I – The Belgian political space
The initial MCA yields a two-dimensional space that accounts for 54.7% of the modified variance rate (i.e. ‘modified rate’). Although this measure of explained variance falls short of the 80% threshold required to adequately describe the structure of the data (as detailed in Le Roux and Rouanet, Reference Le Roux and Rouanet2010, p. 51–52), the third (10.4%) axis is a ‘Guttman-scale’ that reveals the ordered structure of the data by contrasting extreme to moderate responses. Although statistically valid, these axes are a common by-product when using ordinal instead of categorical variables in an MCA and are usually not interpreted for that reason (Atkinson, Reference Atkinson2024; Hjellbrekke, Reference Hjellbrekke2019, but see Lindell and Ibrahim, Reference Lindell and Ibrahim2021). We also do not retain the fourth (4.7%) and fifth (3.7%) axis, reflecting preferences regarding welfare responsibilities and decentralization respectively, due to their low eigenvalues (see Appendix C for an overview of the first 7 axes).
In Table 1, we highlight the categories that have an above-average contribution (i.e. exceeding 1.5%) to axes 1 and 2. These contributions signify the proportion of the axis’ eigenvalue that is explained by a particular category (Hjellbrekke, Reference Hjellbrekke2019, p. 37–39) and thus guide the interpretation of the axes. We have labelled the categories to reflect the respondents’ position on the underlying scale: for instance, the category of ‘Lower retirement age (1/4)’ comprises respondents who are least in favour of lowering the retirement age, while category point ‘Lower retirement age (4/4)’ includes respondents who are most in favour.
Table 1. Relative contribution of categories to axes 1 and 2
n = 1,150, Active variables = 17, Active categories = 65, Passive categories = 17, Share of passive mass = 3%.
Turning to the retained axes, we see that axis 1 represents the strongest polarity, accounting for 35.8% of the modified rate. We label this axis as a cultural dimension, juxtaposing respondents harbouring strong preferences for multiculturalism, self-expression and climate action, to those holding anti-immigrant, authoritarian and materialist views. The sizeable contributions of the immigration (i.e. 21.0% for all four categories combined) and European integration item (10.5%) reveal that globalization conflicts have become firmly embedded in this cultural dimension, while the lower contributions of the authoritarianism (11.6%), gay rights (8.9%) and environment items (7.5%) echo the continued influence of conflicts pertaining to value liberalization and postmaterialism. The contributions of the items on welfare chauvinism (16.7%), the retirement age (6.3%) and welfare state criticism (5.8%) underscore that the cultural dimension has extended its scope to encompass newer debates over welfare deservingness (cf. Häusermann and Kriesi, Reference Häusermann, Kriesi and Beramendi2015). Notably, some ‘old’ cultural issues, including support for free speech (1.8%) and religious liberties (0.8%), have lost their importance in this dimension.
The second axis, which captures 18.9% of the modified rate, is an economic dimension, dividing respondents on issues of trade union intervention (16.0%), fiscal policy (13.5%), unemployment benefits (9.6%) and economic globalization (8.8%). A glance over the contributing categories shows that this axis is still primarily associated with the organization of the welfare state, rather than with potential new economic conflicts emanating from climate change (2.8%) and European integration (0.6%). Looking at the lower order contributions, we find a permeation of cultural items that reveal a convergence of pro-immigration and anti-welfare chauvinistic sentiments with egalitarian views. Also nested along this dimension is the item on communitarianism (4.8%), which is a somewhat surprising finding given the strong association of Flemish nationalism with anti-immigration attitudes found in the past (Billiet et al., Reference Billiet2021).
Both retained dimensions are visualized in Figure 2, which presents a so-called ‘cloud of categories’. This figure shows the mean position of all categories that contribute above average to axis 1 and/or axis 2, highlighting the similarity of these categories in terms of respondent composition. While some of the categories are structured almost entirely along one of the two political dimensions, others, such as those pertaining to retirement age and economic globalization, have outlying positions on both. In sum, our results confirm that the Belgian political space is two-dimensional, that it is organized similarly to the ‘basic structure’ found elsewhere in Western Europe and that both dimensions incorporate emergent globalization conflicts.
Figure 2. Cloud of active categories with above-average contributions.
Results II – The flemish and francophone subspaces
In a second step of the analysis, we use CSA to construct class-specific dimensions for the Flemish (n = 789) and the Francophone (n = 361) subsets. CSA generates unique (region-specific) subspaces, of which the axes can be compared to those of the (Belgian) reference space in which they remain nested. To assess the similarity of both solutions, we consider cosine values, the contributions of categories and the axes’ modified rate (cf. Hjellbrekke, Reference Hjellbrekke2019, p. 102).
Cosine values serve as indicators of correlation, denoting the substantive overlap between the reconstructed axes in the CSA-subspace and the MCA reference space. Values close to 1 or –1 imply a strong correlation, while a value of 0 points to complete orthogonality. From the results in Table 2, we can infer that the Belgian cultural dimension exhibits a near perfect correlation with the first axis of the Flemish (0.95) and the Francophone (0.97) subspace. Likewise, the cosines for the economic dimension show a strong correlation with the subspaces’ second axis (0.82 and 0.78 respectively).
Table 2. CSA: cosine angles, variable contributions and modified rates
Bold = contribution > average contribution of the question (note that this average is higher for variables that have more categories).
In order to pin down the differences between the Flemish and the Francophone political dimensions themselves, we can subsequently compare the contributions of the individual items used to chart the two subspaces. To facilitate these comparisons, Table 2 shows the contribution per variable instead of the contribution per variable category (see Appendix D for an overview hereof). While these results indicate more clearly that the Flemish and Francophone subspaces are structured along a cultural and economic dimension – as was the case for the Belgian space –, there are minor variations in the interpretation of these dimensions. As for the first axis, the Francophone cultural dimension seems more clearly geared towards debates of value liberalization, marked by a higher contribution of the authoritarianism (12.6%) and gay rights items (10.1%). However, the lack of a far-right party and limited politicization of immigration in Francophone Belgium has not decreased the contribution of globalization issues along this dimension compared to its Flemish counterpart. More profound differences emerge when interpreting the economic axes. Here, we see that the Flemish sample is more polarized in its support of trade unions, but also that the individualism and authoritarianism item, two attitudes associated to welfare deservingness (van Oorschot, Reference van Oorschot2006), have become firmly entrenched in the redistributive conflict. The Francophone economic dimension, on the other hand, includes higher contributions from the retirement age and wealth tax item, reconstructing a dimension that is marked by a convergence of conflicts on social protection and redistribution, and less so by aspects of welfare differentiation.
When looking at the modified rate, our results show that the cultural dimension is the stronger polarity in Flanders (39.2% versus 32.2% in the Francophone subspace). For the economic dimension, this value is roughly the same (19.0% versus 17.8%), despite popular belief of a higher economic polarization in Francophone Belgium (see, for example, Coffé, Reference Coffé2008; Keating et al., Reference Keating2003).Footnote 4
Overall, we find that the Flemish and Francophone political dimensions are largely overlapping, regardless of their supply-side differences. Our results suggest that asymmetries in the Belgian party system do not affect the overall structure of its political space(s) profoundly. Moreover, the large correspondence of the Flemish and Francophone political spaces with the Belgian reference space indicates that respondents belonging to the two subspaces are equally well-represented in the reference space as in their own class-specific space – hence, confirming we are able to compare Flemings and Francophones in the same (national) space.
Results III – Regional cleavage structures in the Belgian political space
Finally, we examine the cleavage structures of Flanders and Francophone Belgium, to investigate whether their supply-side differences are reproduced in the configuration of both polities’ cleavages. Drawing from previous accounts of the Belgian political space, we consider four cleavages underlying the political divisions in Belgium: (1) a religious cleavage, between active churchgoers and atheists, (2) a class cleavage, between those in the lowest and highest income quintiles, (3) a linguistic cleavage, between those with a predominant regional identity and those who identify strongest with Belgium and (4) an universalism-particularism cleavage, between ‘winners’ and ‘losers’ of globalization, whose position is best represented by the respondents’ level of education (see, for example, Steiner et al., Reference Steiner2024). To allow for a comparative approach, we plot these groups as supplementary categories in the original Belgian MCA-space but interact them with a binary variable indicating membership of one of the two linguistic groups.
We have visualized the Flemish and Francophone cleavage structures alongside the mean position of the electorates (represented by vote choice) in Figure 3. Both figures are different representations of the same space but have been split into a Flemish (left) and Francophone (right) plane for ease of interpretation. To compare the cleavage structure of the two polities, we (1) situate these divisions in the electoral space by examining the categories’ proximity to a specific electorate and (2) estimate the distance between the cleavage categories in standardized deviations, to measure the polarization they encompass.
Figure 3. Region-specific cleavages and electorates in the Belgian political space for Flanders (left) and Francophone Belgium (right). Note on party ideologies: Christian democratic: CD&V (FLA), cdH (FB); social democratic: sp.a (FLA), PS (FB); liberal: Open Vld (FLA), MR (FB); regionalist: N-VA (FLA); left libertarian/green: Groen (FLA), Ecolo (FB); far right: Vlaams Belang (FLA); populist left: PVDA (FLA), PTB (FB).
At first glance, the two figures seem to differ mainly in the position of category points on the (vertical) economic dimension. Notably, nearly every Francophone point, including the average of the subset (emphasized in red), gravitates closer to the redistributionist pole, while the Flemish mean point leans more strongly toward the pro-market end. Moreover, Francophones are found to be less divided on the economic dimension. Their variance on the second axis is 16.3% lower than that of the Flemish sample, revealing – overall – a greater ‘welfare consensus’ among Francophone voters (result not shown).Footnote 5
We see, by comparison, that the position of the two subsets is nearly indistinguishable on the (horizontal) cultural dimension. Although Flemish voters were traditionally considered to be more conservative, these cultural differences have gradually narrowed in the past three decades (Billiet et al., Reference Billiet2021; Delespaul and Swyngedouw, Reference Delespaul, Swyngedouw and Abts2023).
Table 3 furnishes an overview of the distance between the cleavage categories, encompassing the so-called standardized deviations along the axes (d) and along the entire principal plane (D). In the Flemish political space, the most prominent conflicts emerge around the universalism-particularism, class and linguistic cleavages, followed by the less structuring religious cleavage. From the visual representation of Figure 3, it appears that the divisions based on education and income are overlapping (i.e. correlating), reinforcing a polarity of respondents who possess high cultural and economic capital on one end, and with those with low cultural and economic capital on the other. The ‘low’ capital group aligns closely with the far-right Vlaams Belang (VB) electorate, which is also closest to the mean point for voters exhibiting a strong Flemish identity. Conversely, the ‘high’ capital group is electorally divided, but generally leans towards the centre-right electorates of Open Vld (liberal), CD&V (Christian-democrat) and N-VA (regionalist/conservative). Remarkably, it seems that the Flemish left (sp.a, Groen and PVDA) lacks entrenchment in any of these cleavages but instead draws its support from a heterogenous group of left-wing cosmopolitans outside its traditional working-class constituency.
Table 3. Standardised deviations between supplementary categories
A similar pattern emerges in the Francophone political space. Again, we find that the dominant universalism-particularism cleavage is reinforced by the class and linguistic cleavages, although all of them are cross-cut by the less substantial religious cleavage. The linguistic cleavage is less polarising in comparison to its Flemish counterpart, which is unsurprising given the electoral decline of the Francophone regionalist parties. Substantially, these results underscore that the non-materialization of the far right in Francophone Belgium is not the result of a less divisive universalism-particularism cleavage, nor of a more polarising class cleavage mitigating its upsurge.
It is remarkable, however, that the social-democratic PS-electorate retains a position at the edges of the three most polarising cleavages of Francophone Belgium. With an electorate that is closest to the mean points of low-educated and low-income voters, the party’s position in the Francophone space is somewhat equivalent to that of Vlaams Belang in Flanders. While this speaks to previous accounts portraying PS as a ‘buffer’ against the far right (Coffé, Reference Coffé2008; de Jonge, Reference de Jonge2021), two provisos must be offered in this regard. First, we should stress that PS is not a culturally right-leaning party, and that its electorate also contains clear universalistic contingents (Gaasendam, Reference Gaasendam2020). Second, it is unlikely that the far right would benefit from an equally sized electoral reservoir in Francophone Belgium, as the low-capital groups (low income and low educated) that constitute its core constituency in Flanders have maintained clear redistributionist stances. This is further corroborated by the general welfare consensus among Francophones, which does not seem to dovetail with the producerist welfare agenda of the far right.
Discussion and conclusion
Our analysis demonstrates that the Flemish political space bears close resemblance to that of Francophone Belgium, in spite of their asymmetric electoral trajectories. Taken together, these results suggest that, at least in Belgium, supply-side differences do not appear to be associated with a different cleavage structure, nor with a markedly different dimensionality in the political spaces of its two regions. At the level of the spaces’ dimensionality, it appears that issues of immigration, welfare chauvinism and European integration have become incorporated in the Francophone cultural dimension, even without the agenda-setting capacities of a far-right party. This cultural dimension has surpassed the economic dimension as the main axis of political structuration in both regions. At the level of the spaces’ cleavage structures, our findings demonstrate that the universalism-particularism cleavage is equally polarising in Francophone Belgium, although the social democratic left has likely prevented it from being as highly politicized as it is in Flanders. Remarkably, in both cases, this new cleavage coincided with pre-existing structural divisions; a constellation generally thought to foster ideological polarization.
Overall, these conclusions contradict those observers who claim that the relative strength of traditional cleavages at the demand side ‘may be expected to restrict the possibilities of mobilization on the basis of the new cleavage’ (Kriesi et al., Reference Kriesi1995, p. 4). Given that the political space caters to the universalism-particularism conflict in both parts of the country, the asymmetrical breakthrough of the far right in Belgium is more likely to be associated to differing patterns of party competition, primarily with regard to the behaviour of social democratic parties. It is expected that social democratic parties may act as a buffer against the far right when their position is more in accord with the demands and expectations of the working class (Coffé, Reference Coffé2008; Rydgren and van der Meiden, Reference Rydgren and van der Meiden2019). Specifically, prior research indicates that class voting increases when left-wing parties adhere to traditional welfare policies of social insurance and redistribution as opposed to Third Way policies of social investment and labour market activation (Gingrich and Häusermann, Reference Gingrich and Häusermann2015). For the case of Belgium, Gaasendam (Reference Gaasendam2020) attests that the Francophone social-democratic PS party was capable of maintaining its profile of social insurance, as its left-libertarian competitors embraced a similar stance. This should have reduced the competition over its universalist middle-class pockets, ‘shielding PS from the strategic dilemmas many Third Way social democratic parties with a radical right competitor face’ (p. 143).
However, beyond finding many resemblances between the two spaces, we also revealed some differences. First, we found that Francophone low-income voters have not undergone a substantial ‘embourgeoisement’ but remained stable in their traditional position of working-class authoritarianism (cf. Lipset, Reference Lipset1960). It is generally assumed that electoral competition in this left-authoritarian quadrant is between social democrats and the far right, with the latter gaining the upper hand when the cultural dimension is more salient (Oesch and Rennwald, Reference Oesch and Rennwald2018). Our results appear to support this hypothesis, given that the Francophone cultural dimension explains slightly less variance than that of Flanders (32.3% versus 39.2%). At the same time, however, we find that the economic dimension does not bear a higher importance in Francophone Belgium.
Second, we showed that Francophones adhere to a redistributive consensus on welfare-related matters, which may explain the higher support for left-wing parties in the region. Although far-right and conservative parties mobilize voters primarily on cultural grounds, their agenda of welfare retrenchment and differentiation is likely to be at odds with public opinion in the region. One could posit that the Francophone welfare consensus stems directly from its status as the weaker economic entity in the Belgian federation: Francophone cultural identity is clearly rooted in class consciousness and fears of economic marginalization by the Flemish majority (Van Dam, Reference Van Dam1996). In this connection, social identity theory posits that in-group consensuses develop out of salient collective identities (Turner, Reference Turner1991). Ultimately, these identities may restrict the mobilization potential along the new cleavage, despite its structural underpinnings being present (Kriesi et al., Reference Kriesi1995).
Third, we laid bare minor differences between the two regions in the composition of their political dimensions. Despite anticipating that far-right mobilization would have affected the cultural dimension in Flanders, increasing the salience of immigration and welfare chauvinism, it seems that only the composition of the economic dimension differs between the two regions. In Francophone Belgium, the economic dimension primarily centres on a conflict over welfare expansion versus retrenchment, reflecting the social democratic and liberal welfare agendas of the two main parties. Conversely, the Flemish economic dimension is recalibrated somewhat more to a debate on welfare entitlements. While this difference is marginal – and may be due to a variety of contextual factors –, the heuristics of producerism resonate more with the far right’s broader agenda of welfare populism, which is more politicized in Flanders (Abts et al., Reference Abts2021a).
Finally, these results are consistent with some of our earlier work, which revealed substantively similar Flemish and Francophone political spaces in the period between 1991 and 2014 (Delespaul and Swyngedouw, Reference Delespaul, Swyngedouw and Abts2023). Combined, these analyses confirm that the diverging electoral success of the Belgian far right is not the result of an asymmetrical political transformation of the political space of its two regions. While we should emphasize the limited generalizability of this finding beyond Belgium, this unique constellation may improve our understanding of political transformations in comparable Western European countries. Specifically, one hypothesis emerging from this study would be that the unequal pace of green and far-right mobilization in Western Europe does not necessarily reflect a constellation of ‘laggards’ and ‘leaders’ in terms of political transformations at the demand side (cf. Elff, Reference Elff2007), but is rather contingent on a more intricate interplay of demand- and supply-side factors that defines the salience of (new) cleavages (see also Rydgren and van der Meiden, Reference Rydgren and van der Meiden2019).
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S1755773924000274.
Acknowledgements
We would like to thank Jaak Billiet, Alberto Stefanelli, Cristiano Vezzoni and the anonymous reviewers for their helpful comments. All mistakes remain our own.
Funding
Data collection was made possible with support of the Belgian National Lottery and the KU Leuven Special Research Fund (C1l-19-00569).
Competing interests
The authors report there are no competing interests to declare.
Open access
