The middle-class base of European integration? New class divides and attitudes towards market integration in ten EU countries

ABSTRACT

Highly educated ‘new’ middle classes have been identified as a fundamental constituency that, by supporting technological innovation and liberalisation, bolsters the stability of advanced capitalism. By contrast, less educated production workers are usually described as opponents of such policies. This article investigates the extent to which these class divides translate into different positions on support for EU integration, a key process in the liberalisation of European political economy. In addition to class-based positions, we take into consideration how the subjective perceptions of situations such as unemployment, income loss and job insecurity affect support for EU integration. By relying on an original 2019 survey conducted in ten member states, the analysis confirms that high-skilled middle classes overwhelmingly support European market integration, while ‘old’ middle classes appear more concerned about the welfare losses that integration could imply. Low-skilled service workers emerge as a critical cross-pressured group, taking a mid position between ‘old’ and ‘new’ middle classes.

KEYWORDS:

• European integration
• knowledge economy
• middle class
• public opinion
• employment insecurity
• post-industrial labour markets

1. Introduction

A vast amount of literature has documented the deep socio-political changes associated with the transition from industrial to post-industrial, knowledge-based societies. Deindustrialisation, international economic integration and educational expansion triggered epochal transformations in western European societies and labour markets, with far-reaching consequences for class structures (Oesch, 2006), citizens’ policy preferences (Gingrich & Häusermann, 2015; Ares, 2020) and national party politics (Beramendi et al., 2015; Im et al., 2019). Much research on the matter has stressed the role played by cultural over socio-economic factors in driving the process of change. In western Europe and beyond, groups of ‘winners’ and ‘losers’ of globalisation emerged along divides that transcend traditional class-based economic positions and which are mainly rooted in the growing cultural diversity associated with immigration and in the perceptions of threat that it poses to national identity and traditional values (Inglehart, 2008; Kriesi et al., 2008; Teney et al., 2014). Some authors have even argued that a process of dealignment between social classes and policy preferences has closed the historical period of ‘class voting’ (Franklin et al., 2009; for a contrasting view, see Ares, 2022; Oesch & Rennwald, 2018).

More recently, Iversen and Soskice’s (2019) work on the politics of knowledge-based economies (KBEs) brought the focus back to class dynamics that are endogenous to the post-industrial transition. According to Iversen and Soskice, the key for the ‘exceptional resilience of advanced capitalist democracies’ lies in the growth of a ‘new’ middle class composed of high-skilled workers in the service sector who benefit from economic liberalisation and thus support political agents promoting further progress in that direction. Moreover, in partial contrast with the literature on winners and losers of globalisation, Iversen and Soskice contend that large parts of what they call ‘old’ middle classes (lower-skilled workers who have attained a relatively good social status in the industrial era but who now appear to be the losers in the transition to KBEs) may well align their economic-political preferences with those of the new middle classes, especially when they are confident that education and economic growth will provide their children with better opportunities in the KBE. Only ‘old’ middle-class people who are unable to benefit from the spillovers of KBE-led growth would oppose economic modernisation.

European integration is one of the most emblematic instances of supranational market integration and trade liberalisation in the world (Matthijs, 2017). Furthermore, from the ambitious aim of the Lisbon Agenda of making Europe ‘the most competitive and dynamic knowledge-based economy in the world’ (European Council, 2000) to the Next Generation EU plan launched in the aftermath of the COVID pandemic, the EU has actively promoted economic modernisation and technological progress. Nevertheless, no explicit attempt at testing whether Iversen and Soskice's thesis on the middle-class base of advanced capitalist democracies holds for European integration has been made so far. This article does that by addressing the following question: does the EU rest on a (new) middle-class consensus over the modernisation and liberalisation spirit that is intrinsic in the integration project?

We investigate the relationship between changing European social classes and support for the EU by drawing on Iversen and Soskice’s (2019) conceptualisation of new class divides in KBEs. We complement this perspective with insights from more critical accounts of class restructuring in the EU. On the one hand, recent contributions have highlighted how, to the extent that highly educated middle classes are exposed to precarious employment conditions, they are not necessarily shielded from social risks typical of more vulnerable social strata (Häusermann et al., 2015; Schwander, 2020). On the other hand, another social group that is particularly relevant to understand post-industrial class dynamics consists of low-skill service workers (Kurer & Palier, 2019). The demand for low-skilled service jobs is closely linked to the flourishing of the KBE, as it is largely driven by the growth of new middle classes (see for example Morel [2015] on domestic services). However, contrary to the new middle classes, low-skilled service workers have a low level of marketable skills, which makes them vulnerable to economic change in a way similar to the old middle classes. Hence, this article analyses whether these cross-pressured groups of workers – high-skilled labour market ‘outsiders’ and low-skilled service workers – politically coalesce, on the EU integration debate, around the new middle class (in line with Iversen and Soskice's argument) or instead display distinct preferences.

The empirical analysis is based on original survey data that were collected in summer 2019 in ten EU member states. We first assess whether citizens’ attitudes towards European integration and its possible economic consequences are patterned along structural positions that identify different post-industrial social classes (income, education, occupation, urban or rural area of residence, labour market insier-/outsiderness). In addition to objective structural positions, we consider how the subjective perception of contingent situations, such as unemployment, income loss and having more or less positive expectations for one's own children, may also affect citizens’ support (or concern) for EU market integration.

The article is organised as follows. The next two sections outline the theoretical background and the hypotheses. Section 4 provides a description of the data, variables and method used. The fifth section reports the results from regression analysis. The last section wraps up and concludes by elaborating on the implications of our findings for future research on changing class politics in the EU.

2. New class divides in the EU's knowledge-based economy

Political sociology scholars have long sought to map out the distributional consequences associated with the post-industrial transition. Inspired by classic cleavage theory (Lipset & Rokkan, 1967), authors such as Kriesi et al. (2008) and De Wilde et al. (2019) have argued that the post-industrial transition – like secularisation, urbanisation or industrialisation before it – can be understood as a deep societal transformation generating new structural divisions within societies and party systems.

This account has been recently enriched by Iversen and Soskice (2019). According to them, the transition from industrial towards KBEs was accompanied by a substantial upgrade of the skill levels of large parts of the workforce (Hurley & Fernández-Macías, 2008). However, due to the automation of routine tasks, it also involved a sharp decrease in the demand for low-skilled and unskilled jobs, thereby generating a significant margin of workers with poor employment opportunities. As a result, divergent attitudes in relation to economic modernisation started to develop between different groups of workers depending on their skill levels (i.e. depending on their ability to take a leap into the new KBE). Highly educated people, who are most capable of benefiting from the growth of knowledge-intensive service jobs, emerged as a ‘new’ middle class (Iversen & Soskice, 2019). The latter are sympathetic to processes of socio-economic change, chief among them being the development of international open markets and the liberalisation of the domestic political economy arrangements inherited from the postwar, Fordist period. Mau (2015) even speaks about the middle-class base of neoliberalism. According to Iversen and Soskice (2019), this is because knowledge-intensive industries benefiting from market liberalisation are rooted in urban agglomerates with specific social infrastructures, which make new middle classes’ income prospects secure. By contrast, middle- and low-skilled manual workers, whose routine and cost-sensitive jobs are easier to outsource and automate (Brynjolfsson & McAfee, 2014), form what Iversen and Soskice call the ‘old’ middle class: individuals who used to enjoy a relatively comfortable socio-economic position in the industrial society but who are now increasingly worried about the threats that socio-economic change poses to their well-being. Indeed, old middle classes rely more on national social protection arrangements while benefiting more from a less competitive economic environment (Wren & Rehm, 2013). As a consequence, they tend to be mistrustful of economic modernisation and trade liberalisation.

In the European context, liberalising reforms have been associated with EU integration (Ferrera, 2005; Scharpf, 2010), and important sections of society perceive the EU as a threat to national social protection institutions (Beaudonnet, 2015). A large body of literature highlights how the fiscal and economic imperatives emanating from the model of monetary integration established by the Maastricht Treaty eroded the so-called ‘European Social Model’ (Crespy & Menz, 2015; Pochet & Degryse, 2013). Furthermore, European integration has not only implied an internal (intra-European) process of liberalisation but also an external one: the establishment of the Single European Act in 1987 and the policy of external trade liberalisation pursued since then (De Ville & Orbie, 2014) have transformed the EU single market into one of the most open markets in the world (Rodrik, 2011). Thus, over the last three decades, EU integration has primarily implied the internalisation of globalisation dynamics: while competitive pressures have been continuously sharpened, a wide range of policy tools – such as vertical industrial policy, capital controls and monetary devaluation – that were once used to offer protection to workers and industries have been banned (Scharpf, 2010). The destabilisation of domestic institutional arrangements brought about by these processes has been interpreted as nurturing the emergence of a new cleavage rooted in the diverging material opportunities and life chances linked to EU integration (Bartolini, 2005, p. 395). Postfunctionalist integration theory depicts this cleavage as a conflict between ‘cosmopolitans’ – those who support supranational integration– vis-a-vis ‘communitarians’ – those who prefer to strengthen national borders. In the post-Maastricht context, EU integration taps into various sources of conflicts about national identity, solidarity and sovereignity (Hooghe & Marks, 2009; Kriesi, 2020; Kuhn, 2019)

This article investigates the extent to which the new class divides postulated by Iversen and Soskice (2019) translate into different attitudes towards EU integration and the socioeconomic transformations implied by it. In the first place, we expect new middle classes to be more supportive of the European integration project than the old middle classes. While the removal of trade barriers and the free movement of labour in the EU increase job insecurity for lower-skilled workers, new middle classes likely perceive the implications of market integration as an opportunity rather than a threat to their employment and social prospects. In fact, Iversen and Soskice's account shares expectations with classical political economy models of trade liberalisation. The latter which predict that highly skilled individuals face lower labour market risks when they are exposed to international trade integration, which instead increases the risks among low-skilled individuals (Walter, 2017), with both groups forming their policy preferences accordingly. This resonates with previous research that, following a utilitarian approach to explain attitudes towards the EU, shows that individuals with a low income and, even more relevantly, low education tend to reject European market integration (Baute, 2023; Hakhverdian et al., 2013; Hobolt & de Vries, 2016). Hence, we first test the following hypotheses:

H1a: New middle classes show more support for EU integration than old middle classes.

H1b: New middle classes fear less than old middle classes a loss of jobs and social security as a consequence of EU integration.

As discussed above, Iversen and Soskice maintain that liberalising reforms have been the democratic responses of elected politicians to the demands of the enlarged coalition around the new middle classes. Nonetheless, central to Iversen and Soskice's account is the hypothesis that such a pro-modernisation coalition reaches not only the ‘new’ middle classes – highly educated people who are more likely to reach the upper half of the income distribution – but also individuals from lower classes who ‘see themselves as more closely aligned with the top because […] they can still reasonably expect to see their children do well by acquiring the education needed in the new economy’ (Iversen & Soskice, 2019, p. 222). The two authors call this electoral constituency ‘aspirational voters’: parents who are not themselves highly skilled but who aspire to upward social mobility for their children via higher education. Therefore, we hypothesise that:

H2: Individuals who have positive expectations for their children's future show more positive attitudes toward EU integration. This is especially true for old middle classes, who tend to align with new middle classes’ preferences when they think their children will have better economic opportunities.

3. Beyond the new middle class-consensus: cross-pressured social groups in the KBE

The shared interest between new middle classes and aspirational voters is key to Iversen and Soskice's argument because it provides sufficiently large democratic majorities, ensuring stability for KBEs such as those of the EU (Iversen & Soskice, 2019, pp. 40–41). Nevertheless, their theorisation of the support coalitions underpinning liberalisation in post-industrial contexts can be criticised for neglecting two social groups that might divert from Iversen and Soskice's expectations to the extent that they find themselves in cross-pressured positions in the class structures of KBEs. On the one hand, Palier (2020) has underscored how Iversen and Soskice's account may be too optimistic regarding the political implications of the emergence of groups of low-skilled or unskilled workers in labour-intensive service jobs, i.e. the ‘servants’ who allow the ‘brains’ to focus on their tasks (Morel, 2015). Low-skilled service workers, despite sharing a modest skill endowment with low-skilled industrial workers, are located in economic sectors that are closely linked to the KBE in job-rich urban agglomerates (for example, domestic and personal care, delivery, transport, retail, hospitality); as such, their livelihood depends on the prosperity of higher-skilled service sectors (Iversen & Soskice, 2019). Moreover, low-paid service occupations cannot be delocalised: they need to be situated in the same urban centres as the high-skilled professionals from whom the demand for low-skilled services comes. Based on this, in contrast with authors who emphasise the political disenchantment of low-skilled service workers (Ares, 2022; Häusermann, 2020; Kurer & Palier, 2019; Palier, 2020), Iversen and Soskice (2019) posit that the latter social group tends to coalesce politically around the political preferences of the new middle classes. We put this expectation to the test in the context of EU market integration through the following hypotheses:

H3a: Low-skilled service workers show more support for EU integration than production workers from old middle classes.

H3b: Low-skilled service workers fear less than the old middle classes a loss of jobs and social security as a consequence of EU integration.

In addition to low-skilled service workers, the literature on labour market dualisation points towards a second social group that cuts across the ‘old’ versus ‘new’ middle class dichotomy: high-skilled outsiders. The latter are highly educated people with atypical employment contracts which, being often accompanied by inadequate social protection coverage, leave them in more precarious economic conditions compared to labour market ‘insiders’ with standard employment relations (Schwander & Häusermann, 2013). According to this view, ‘labour market insecurity extends well into the higher-skilled middle class in particular to high-skilled young adults and high-skilled women’ (Schwander, 2020, p. 369). As such, not all high-skilled groups do equally well in KBEs: ‘high-skilled outsiders’, most notably young adults with knowledge-intensive yet precarious jobs, share with low-skilled workers similar levels of labour market vulnerability, especially in southern Europe (Häusermann, 2020; Häusermann et al., 2015; Schwander, 2020). Previous literature has also found that high-skilled outsiders tend to show particular political orientations. Namely, they are more likely to vote for radical left parties than both low-skilled outsiders and high-skilled insiders (Marx & Picot, 2013; Schwander, 2019), thus implying preferences that are not aligned with EU market liberalism. Therefore, we expect high-skilled outsiders to be less enthusiastic about European integration and more concerned about some of its potential harmful economic implications. Thus, we hypothesise the following:

H4a: High-skilled outsiders show less support for EU integration than the rest of the new middle classes.

H4b: High-skilled outsiders fear more than the rest of the new middle classes the loss of jobs and social security associated with EU integration.

Employment precariousness is a difficult concept to grasp. Labour market outsiderness and the type of employment contract do not necessarily translate into increased economic insecurity, which is crucial when considering individuals’ attitudes within the EU's KBE. Labour market outsiderness may be a temporary situation for many high-skilled individuals, especially for young adults who aspire to better and more secure employment positions in later life (Schwander, 2019). Rather than referring to the current contractual positions of different types of workers, recent empirical works have in fact focused on the likely risk of unemployment to which different occupational categories are exposed (Rehm, 2009; Schwander & Häusermann, 2013). In this regard, Schwander and Häusermann (2013, p. 249) stressed the fact that it is not only the current situation of individuals that shapes their political preferences (like, for example, working with an ‘atypical’ employment contract) but also their expectations concerning future labour market risks. This echoes Kurer and Palier’s (2019, p. 4) consideration that ‘we might not observe the strongest political backlash against economic modernisation among the hardest-hit but rather among those who are most concerned about their economic well-being and future prospects in the labor market’ (see also Im et al., 2019). The recognition of the importance of the subjective dimension of income and employment security is increasingly shared in studies on attitudes towards European integration and its various facets (Ferrera & Pellegata, 2018; Teney et al., 2014). Based on these considerations, we propose an additional mechanism: we argue that new middle-class support for the European integration process depends on how they evaluate their future opportunities in work and society. Therefore, we hypothesise the following:

H5a: High-skilled individuals who feel economically insecure show less support for EU integration than the rest of the new middle classes.

H5b: High-skilled individuals who feel economically insecure fear more than the rest of the new middle classes the loss of jobs and social security associated with EU integration.

4. Data and method

The empirical analysis relies on data from the survey ‘Reconciling Economic and Social Europe: Values, Ideas and Politics’ (REScEU) (Donati et al., 2021). The survey was administered by IPSOS using CAWI methodology between 28 June 2019 and 2 August 2020 in ten European countries: Finland, France, Germany, Greece, Hungary, Italy, the Netherlands, Poland, Spain and Sweden. National samples include at least 1,500 respondents aged between 18 and 70 (N = 15,149) and are stratified by gender, age groups, level of education and macro-area of residence (NUTS-1).

The REScEU survey comprises a number of questions on citizens’ feelings about the European integration project and its implications. Two items are particularly suitable for measuring the dependent variables we use to test our hypotheses. The first is a general question about the support for EU integration, which asks respondents to place themselves on an 11-point scale where 0 means integration ‘has already gone too far’ and 10 means it ‘should be pushed further’. The second item directly taps into respondents’ concern over the possible consequences of EU integration for national labour markets and welfare systems: it asks about the extent to which people are ‘currently afraid of the loss of jobs and social security’ that ‘the process of European unification’ could imply for their country. In the latter case, we recoded the four-point Likert scale responses into a binary variable that takes the value 1 for those who are (‘very much’ or ‘somewhat’) afraid of the loss of jobs and social security and 0 for those who are not (very) afraid. Appendix 1 reports the exact wording of the survey items.

On the independent variable side, our main interest lies in identifying different post-industrial social classes, most notably the ‘old’ and ‘new’ middle classes discussed in the previous section. Drawing on Iversen and Soskice (2019), we use the following variables to capture distinct dimensions of ‘social class’: (1) level of education – distinguishing between respondents with up to lower-secondary education, up to post-secondary non-tertiary education and graduates; (2) income, measured through respondents’ self-placement in the national income distribution (on a 0–10 scale) or, in robustness checks, through a categorical variable distinguishing between four statuses: living comfortably on present income, coping on present income, finding it difficult on present income, finding it very difficult on present income; (3) urban-rural divide: we differentiate between respondents living in ‘urban areas’ (towns with more than 100,000 inhabitants) and those living in ‘rural areas’ (fewer than 100,000 inhabitants); (4) occupation: Iversen and Soskice (2019:, pp. 238–239) essentially differentiate between manual workers (routine jobs pressured by technological change), low-skill personal and social services (where the lack of routinization makes automation unlikely) and higher-educated professionals (main beneficiaries of the technological advancements in the KBE). In making this distinction, the two authors implicitly emphasise what Daniel Oesch (2006) called ‘hierarchical division’ in his occupational class schema; that is, they classify jobs according to the level of marketable skills associated with them. Since the main aim of this article is to test whether the class divide theorised by Iversen and Soskice also applies to EU market integration, we proceed in a similar way, though we group occupations by drawing more explicitly on (the most parsimonious version of) Oesch’s (2006) class schema. All details on the operationalisation of occupational class are provided in Appendix 2 and summarised in Table A1. The variable we use in the analyses below includes the following categories: employers and small business owners; higher-grade service (managers, technical and sociocultural [semi-]professionals, skilled service employees); lower-grade service (workers in labour-intensive service jobs, including unskilled and casual service workers); manual workers (skilled and unskilled manual workers in agriculture or manufacturing, casual workers in manual jobs); respondents who have never had an occupation (including full-time students and the inactive).

The REScEU survey is particularly rich in detail on respondents’ lived experiences: a number of questions address contingent situations and expectations that affect individuals and their households. One of these questions is directly inspired by Iversen and Soskice’s (2019) ‘aspirational voter’ argument and asks ‘how likely [respondents] think it is that [their] children will have better opportunities to succeed financially than [they] have had’. We use this item to test H2. We recoded the response categories ‘very likely’ and ‘somewhat likely’ into ‘likely’; and ‘somewhat unlikely’ and ‘very unlikely’ into ‘unlikely’; and we grouped ‘don't knows’ and missing answers (including all respondents without children) into a residual category ‘no opinion/no children’.

In order to test H4, which refers to worker outsiderness, we rely on the type of employment contract – an indicator of the structural (objective) labour market position of the respondents, which is most typically used in the literature (see Rovny & Rovny, 2017 for a review). We differentiate between respondents working with a permanent contract, ‘atypical’ workers (including fixed-term employment, temporary employment, apprenticeships and internships, platform work, occasional work and those working without a regular contract) and a residual category including those out of work.¹ We use, instead, different indicators of both income and employment insecurity to gauge people's subjective perception of economic insecurity, addressed in H5. As for income insecurity, we constructed two indicators of what Ferrera and Pellegata (2018) call ‘intertemporal relative deprivation’, based on respondents’ self-placement in the national income distribution at three points in time: five years prior to the interview, present and prospective (in five years from the interview). Both indicators are dummy variables. The first one tells us whether the respondents feel that their income has worsened compared to the past. The second takes the value 1 if the respondent thinks that their income will deteriorate in the next five years, and 0 otherwise. Last, we measure how familiar respondents are with the risk of unemployment through a question that asks whether in the last two years the respondent or another member of the household has experienced a continuous period of unemployment. We coded 1 if so, and 0 otherwise.

We also include a number of control variables: age, gender, ideological left-right self-placement (we recoded the original 11-point scale variable into the categories ‘centre’, ‘right’, ‘left’ and ‘not located’) and an index of ‘transnationalism’. The latter takes into account the fact that, on top of structural socio-demographic factors, attitudes towards EU integration can be influenced by people's transnational experience (Kyriazi & Visconti, 2021). It is an additive index based on the following dummy variables – whether one: has more than one EU citizenship; has at least a parent/ partner/close relatives or friends who were born in another member state; has friends who are from other member states; often communicates with foreign people living in other EU member states; has a job involving contacts with people from other member states; has lived in other EU member states for at least three months; has visited another member state in the last 12 months. Table A2 in the Appendix provides descriptive statistics of all variables.

We test Hypotheses 1a, 2, 3a and 4a (continuous dependent variable ‘support for EU integration’) through OLS regression models, and Hypotheses 2b, 3b and 4b (binary dependent variable ‘fear of loss of jobs and social security’) through logistic models. In both cases we use the full set of independent variables and control variables, plus country dummies to account for unobserved differences across countries. To these baseline models we add step-wise the variables used to test H2 (aspirations for children) and H4 (employment/income insecurity). The Appendix also provides a number of robustness checks, including the full models rerun with the categorical variant of the ‘income’, with post-stratification weights, and with cluster-robust standard errors to account for heteroskedasticity across countries. In order to test our hypotheses on different EU attitudes across social classes, we take a closer look at predicted values of ‘EU support’ (predicted probabilities for ‘fear of job loss’) for specific theory-relevant class profiles. Since we integrate Palier's considerations on the lower strata of the service middle class in Iversen and Soskice’s (2019) dichotomy between ‘old’ and ‘new’ middle classes (Palier, 2020; see Section 2), we focus on three ideal-typical profiles: (1) the ‘new’ middle class – a highly educated person employed in a high-grade service sector, living in an urban area and with an income self-placement of 6 on the 0–10 scale; (2) the ‘old’ middle class – a less educated person employed in manual work, living in a rural area and with an income self-placement of 4 on the 0–10 scale²; and (3) the ‘low-skilled service class’ – a middle-educated person employed in a lower-grade (i.e. labour-intensive) service job, living in an urban area and with an income self-placement of 5 on the 0–10 scale.

5. Results

Before proceeding to multivariate analyses, we inspect the size of the crucial social groups that form the enlarged ‘aspirational’ middle-class constituency (see Section 2) in the ten European countries included in our sample. Figure 1 shows the population shares taken by ‘new middle class’ workers in higher-grade (i.e. skill-intensive) services, most of whom have tertiary education, lower-grade service workers living in job-rich urban agglomerates, and by people who are not part of the former two groups but who think that their children will climb up the social ladder – ‘aspirational’ middle-class voters in Iversen and Soskice’s (2019) words.

Figure 1. Share of the population taken by three key social groups in ten European countries.

Source: REScEU 2019 survey, weighted data

In all countries, the sum of the new middle class, urban lower-grade service workers and aspirational voters goes beyond 50% of the population (65.7% on average). The thus defined enlarged middle-class constituency is wider in the Netherlands, Sweden and Germany (above 70%). The lowest total share shown by Italy – closely followed by Greece and Spain – seems to be driven by a comparatively small size of the new middle class. Overall, however, the enlarged middle-class coalition appears big enough to provide sufficiently wide electoral constituencies throughout the European countries at stake, especially when bearing in mind that electoral turnouts are generally lower among the low-educated – that is, the part of the population that is largely excluded from the social groups reported in Figure 1. Having ascertained this, we move forward to multivariate analyses on citizens’ attitudes towards the EU.

Figure 2 summarises the results of the two full regression models that we use to assess the impact of socio-demographic characteristics on the support for further EU integration (Figure 2(a), showing linear regression coefficients) and on the likelihood of being afraid of the loss of jobs and social security associated with the integration project (Figure 2(b), log-odds shown). For simplicity of presentation, the estimates for gender, age and for the country dummies are omitted. Tables A2 and A3 in the Appendix report the full model specifications.

Figure 2. Regressions coefficients from (a) OLS regression of ‘support for EU integration’ and (b) logistic regression of ‘fear of loss of jobs and social security’ (log-odds).

Note: age, gender and country dummies’ coefficients are omitted. The full specifications of the underlying models are reported in Tables A3 and A4 (Appendix).

Let us focus first on Figure 2(a) (left-hand graph). The coefficients show that, among the key variables that we chose as indicators of social class, education and income are the most relevant for predicting support for EU integration. Respondents with medium and – even more so – tertiary education show significantly higher support for the EU than less educated people, and a higher (self-defined) income position matches with higher support. The urban-rural divide also seems to matter, as respondents living in big towns are significantly more supportive of EU integration than those living in rural areas (p < .05). Occupational categories, on the other hand, explain less variation. Surprisingly, manual workers do not seem significantly less supportive of EU integration than high-grade service workers (the reference category) – the negative coefficient is only marginally significant, at p < .10. It is rather workers in lower-grade services who show more negative attitudes towards the EU than the higher-grade service class (p < .05). This resonates with the gloomier view of the political disenchantment of low-skill service ‘servants’ (Morel, 2015; Palier, 2020) more than with Iversen and Soskice’s (2019) argument, according to which lower-skilled service workers would coalesce around the new middle class in supporting liberalising political blueprints.

Ideological divides over EU integration are apparent: matching with findings from previous research (e.g. Daniele & Geys, 2015), left-leaning individuals are more supportive of the EU integration project than those who locate themselves at the centre of the left-right continuum, with the opposite being true for right-leaning respondents. As expected, transnational experience is positively correlated with support for EU integration. Moving to life experience-related factors, aspirations for children show significant differences: respondents who think that their children will have better opportunities to succeed financially than they have had are more supportive of EU integration than those who consider it unlikely. As regards the indicators of outsiderness and economic insecurity, only those concerning income insecurity turn out to be significant. As hypothesised, and in line with similar considerations discussed by Kurer and Palier (2019) and Im et al. (2019), respondents with negative expectations about prospective income are less inclined to support EU integration. On the other hand, somewhat counter-intuitively, for those feeling that their income has deteriorated in the last five years, we observe a positive association with support for EU integration. However, in the latter case, the coefficient is only marginally significant in the full model (Figure 1(a)), and not significant when the other indicators of outsiderness and insecurity are left out (see Table A3 in the Appendix).

Figure 3 helps us take a closer look at predicted values of support for EU integration for the specific class profiles we detailed in the previous section. The first three bars from the left directly address H1a and H3a. In line with H1a, we see that an ideal-typical ‘new’ middle-class individual shows a higher degree of support for pushing EU integration further than the ‘old’ middle-class counterpart: the predicted support is respectively 6 and 5 on the 0–10 scale, with a statistically significant difference. The low-skilled service class takes a middle position between the two and shows a significantly higher level of support than the old middle class (H3a seems to be confirmed). The last bar on the right shows the predicted value of EU support for the ‘aspirational’ old middle-class ideal type – that is, old middle-class respondents who think that their children will have better economic opportunities. Comparing the first and last bars in Figure 3 reveals that having positive aspirations for children does not close the gap in EU support between old and new middle classes. This brings to light a more nuanced picture with regard to Iversen and Soskice’s (2019) ‘aspirational voter’ argument (H2): although in general, as we hypothesised, positive expectations for one's own children's future correlate with a degree of higher support for EU integration, this does not make old middle-class attitudes towards EU integration align with those of the ideal-typical new middle class.³

Figure 3. Predicted support for EU integration for four different social class profiles.

Note: Estimates are based on model 3 in Table A3 (Appendix).

H4a and H5a also deserve more attention. As shown in Figure 2, among the indicators used to gauge outsiderness and economic insecurity, only those on income insecurity are significantly correlated with support for EU integration, and income deterioration shows the opposite sign (a weak though positive association with EU support) relative to what we expected. Working with an atypical contract and having experienced unemployment in the household do not correlate with support for EU integration. We also interacted step-wise the various indicators of outsiderness and subjective insecurity with education to check whether the difference in EU support materialised among high-skilled outsiders but did not find any significant interaction. This suggests that there is no specific difference in support for EU integration for high-skilled outsiders; instead, only in respect of the future-income-prospects dimension of outsiderness, all those who have negative expectations regarding their future income (and not only the high-skilled) are less supportive of the integration project than people with higher income security. It is worth noting that, for new middle-class people with negative expectations concerning their income, the predicted value of EU support, although lower than that observed for the new middle-class ideal-typical profile, remains above the average (about 5.7 on a 0–10 scale: see Figure A1 in the Appendix).

We now move on to the determinants of the fear of loss of jobs and social security associated with EU integration. Figure 2(b) (right-hand graph) shows the corresponding effect estimates (the full model estimates are found in Table A4 in the Appendix). No occupational class shows significant differences from the reference category ‘higher-grade service workers’. Once again, education, income and the urban-rural divide are the dimensions of social class that better explain variation of the dependent variable. All coefficients have the expected sign: higher-educated people are less afraid of a possible loss of jobs and social security than less educated respondents; a higher income status matches with a lower likelihood of being afraid, as does living in a urban area. Among control variables, ideology follows the expected pattern (left-of-centre respondents tend to be less afraid of job loss and vice versa), while the index of transnational experience does not reach statistical significance. While the objective indicator of outsiderness (‘employment type’) has no significant effect,⁴ the effect estimates of the indicators of subjective economic insecurity all run in the expected direction. When we hold other factors constant, having perceived a loss of income in the last five years, expecting an income loss in the next five years, and having experienced unemployment in the household match with a higher likelihood of being afraid that EU integration implies a loss of job and social security.

As done for EU support, we refer to more specific estimates for the ideal-typical class profile for assessing H1b and H3b. Figure 4 shows the predicted probabilities of being afraid of the loss of jobs and social security for three selected profiles. The pattern that emerges is clearly in line with the two hypotheses: new middle classes fear less than old middle classes a loss of jobs and social security as a consequence of EU integration. The same is true for the low-skilled service class, although the latter appears significantly more concerned about job and social security loss than the new middle class. Like H4a above, H4b is also rejected: we do not find a robust effect of the objective indicator of outsiderness – employment contract type. However, turning to H5b, income and employment insecurity (more specifically, the subjective perception of current or prospective income deterioration, and living in a household that has experienced unemployment) go together with a higher concern for the negative implications that EU integration can have for national labour markets and social security.

Figure 4. Predicted probabilities of being afraid of the loss of jobs and social security attributed to EU integration for three different social class profiles.

Note: Estimates are based on model 2 in Table A4 (Appendix).

To ascertain the consistency of our results we ran a number of robustness checks, with different specifications of the models: using the categorical variant of ‘income’, with post-stratification weights, and with robust standard errors clustered by country (Tables A5 and A6 in the Appendix). Our findings are robust to these tests. A partial exception regards the estimated effect of ‘expected income deterioration’ on support for EU integration in the model with cluster-robust standard errors, which remains negative but becomes only marginally significant in the full model (p < .05) and barely reaches statistical significance (p < .10) when other indicators of outsiderness and economic insecurity are not included (Table A5). Similarly, when refitting the logistic models of the likelihood of being afraid of job and social security losses, we found that reporting a worse income position with respect to the past remains significant only when other indicators of outsiderness and economic insecurity are omitted and loses significance in the full model (Table A6).

6. Discussion and conclusions

By examining post-industrial class-based attitudes towards European integration, this article has sought to bridge the gap between two bodies of research that have so far been developed in isolation. Research on post-industrial class realignment has invested considerable efforts in studying how the transition to a service/knowledge-based economy created new divisions within the middle classes (Iversen & Soskice, 2019; Schwander, 2020) and has changed the structure of party competition and welfare support (Beramendi et al., 2015; Gingrich & Häusermann, 2015; Oesch, 2008). At the same time, research on public attitudes towards EU integration has underlined how a wide range of individual-level and contextual factors, cultural, ideological and economic, shape support for the EU, either in broad terms (Hobolt & de Vries, 2016) or for specific aspects of the integration process, such as fiscal redistribution (Gerhards et al., 2020) or EU-level social policy developments (Gerhards et al., 2016). Drawing on Iversen and Soskice’s (2019) reconceptualisation of class politics in KBEs, and by relying on original survey data from ten EU member states, we have linked these two literatures and explored whether, and if so how, post-industrial class divides translate into attitudes towards EU market integration.

Our analysis reveals that well-off, highly educated ‘new’ middle classes form the core support base of the European integration project. At the same time, those groups that Iversen and Soskice call ‘old’ middle classes – composed of lower-skilled production workers – show greater reluctance towards EU integration and are more concerned about the loss of national jobs and social security that that could imply. Our data also indicate that a relevant part of the electorate beyond the new middle class, namely those ‘aspirational voters’ (Iversen & Soskice, 2019) who expect that their children will attain a higher level of education and thus be better able to reap the benefits of the integrated KBEs in the EU, tend to support European market integration. ‘Aspirational’ middle classes, however, fall short of reaching the higher levels of support that are observed among the new middle classes.

Based on considerations regarding the cross-pressured position of low-skilled service workers in post-industrial economies, we also empirically tested the attitudes towards the EU of this latter group as compared to those of the new and old middle classes. Our results show that, with respect to support for EU integration, low-skilled service workers take a mid position between the ‘old’ and ‘new’ middle class: they are not as enthusiastic about the EU as the new middle class, but they seem to be more favourable toward it (and, in particular, less afraid of its possible economic consequences) than the old middle class. Moreover, some aspects related to income and employment insecurity correlate negatively with EU support and, even more so, increase the fear of job loss associated with European integration. Surprisingly, the most common indicator of outsiderness used in the literature – being in atypical work – does not seem to have direct implications for attitudes towards EU integration. It is subjective perceptions that matter the most. Having bleak future income prospects lowers the support for the EU. Likewise, income insecurity (measured both in terms of experienced and foreseen income deterioration) and living in a household that has recently experienced unemployment contribute to increasing concerns over the possible negative socio-economic consequences of EU integration.

Our findings suggest important implications for the political stability of the EU. Even if support for European integration remains today majoritarian across member states (Gerhards et al., 2020), the economic and social crises brought about by the Great Recession, by the COVID pandemic and, currently, by the energy crisis following the Russian invasion of Ukraine should keep policymakers attentive to the fact that worsening income and employment prospects can translate into Euroscepticism for the most vulnerable strata of EU societies, with the middle classes being no exception. Cultural factors, such as the fear of the loss of national identity, are certainly one of the main drivers of Euroscepticism (Hobolt & de Vries, 2016; Kuhn, 2019). Nevertheless, to the extent that in times of hardship, economic concerns become increasingly salient and boost Euroscepticism (Kuhn et al., 2016; Singer, 2013), they risk eroding the middle-class consensus on which democratic KBEs are based (Iversen & Soskice, 2019), in the EU and beyond.

Future research should examine in greater depth the issues that this article has left aside, or only partially addressed, due to both space constraints and data limitations. First, we showed that cross-country variation in the size of (new) middle classes is remarkable (Figure 1) – to the extent that it may not secure stable majority constituencies in some EU member states – and surely deserves further investigation. In fact, research on public attitudes towards the EU warns against cross-country generalising assumptions given the importance of national contexts in shaping them (Jayet, 2020, p. 1149). Second, attention should be paid to middle classes’ political preferences other than their attitudes towards EU integration. For example, the positioning of different strata of ‘old’ and ‘new’ middle classes, including lower-grade service workers, on the left-right ideological continuum, as well as their party preferences (e.g. Oesch & Rennwald, 2018), are also clearly linked with the future political sustainability of the EU. Moreover, panel data could serve to better address the causal mechanisms implied in the analyses provided in this article – for example, by assessing whether the worsening of income or employment conditions effectively lowers EU support for the same individuals over time.

Open access

Please notice that the ERC requires us to publish open access.

Acknowledgement

The authors would like to thank Julian Garritzman, Alessandro Pellegata, Maurizio Ferrera and Marcello Natili for their helpful comments on earlier versions of the manuscript, as well as the Editor and three anonymous reviewers of European Politics and Society.

Disclosure statement

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

Additional information

Funding

This work was supported by the European Research Council under the Synergy Grant number 810356 (ERC_SYG_2018), in the scope of the project SOLID – Policy Crisis and Crisis Politics, Sovereignty, Solidarity and Identity in the EU post 2008.

Notes

1 Grouping atypical workers and the unemployed together as outsiders yielded the same results as those shown below.

2 These first two profiles are the same as those used in Iversen and Soskice's empirical analysis (2019: Chapter 5), with the only difference being that we did not include ‘gender’ (i.e. woman for ‘new’ and man for ‘old’ middle class). Here, we do not view gender as a building block of ‘social class’. When refitting marginal prediction by also including gender, however, patterns remain consistent.

3 For a more fine-grained test of H2, we also ran additional models in which we interacted the children-aspiration variable with the three most important variables we used to define social class, namely education, income and occupation. Interactions were not significant, and plotting marginal predictions from either of the three interaction models held the same pattern as that shown in Figure 3.

4 The effect of ‘employment type’ is only significant when we exclude from the model the other indicators of outsiderness and economic insecurity (Model 3 in Table A4 in the Appendix).