Opinion polarization of immigration and EU attitudes between social classes – the limiting role of working class dissensus

ABSTRACT

This study examines polarization in political opinions toward immigration and the European Union between occupational classes, i.e. structural polarization. We maintain that two conditions must hold to indicate structural opinion polarization: high between-class divergence and high within-class consensus. Our main contribution is to study these two conditions systematically for a wide variety of immigration and EU-related topics. Using data from four high-quality German surveys spanning three decades, we document three main findings. First, we find substantial between-class divergence: respondents in typical working class occupations express substantially more unfavorable opinions about immigration and the EU than the upper classes across the majority of survey indicators. Second, however, we also observe considerable opinion heterogeneity within the working class. This lack of within-class consensus limits the potential of mobilizing the working class as a group on the basis of anti-immigration and anti-EU sentiments. Third, while we do not document durable increases in structural opinion polarization over time across most of our opinion indicators, we do draw attention to those individual indicators that show relatively high polarization. Overall, our results suggest limited opinion polarization between occupational classes on immigration and EU issues in Germany.

KEYWORDS:

• Opinion polarization
• globalization cleavage
• attitudes toward immigrants
• opinion trends
• measuring polarization
• class polarization

Introduction

Opinion polarization between groups of individuals in similar socio-structural positions is of long-standing interest to sociologists. Social class was undoubtedly a major socio-structural component of political cleavages during the 19th and 20th centuries, and scholars have argued that it continues to structure politics well into the twenty-first century (Oesch 2008). However, the topics of political contention have evolved in recent decades. According to the landmark work of Hanspeter Kriesi et al. (2008), increasing globalization leads to the rise of a globalization cleavage that divides citizens with positions favoring national closure and immigration restrictions from those who support positions favoring transnational integration, denationalization, and immigration (Kriesi et al. 2008; Azmanova 2011).

Our main goal in this study is to assess the extent to which opinions on immigrants and the EU are polarized between social classes in Germany. The empirical debate on globalization cleavage points to occupational class and educational level as the main structural components (Oesch 2008; Bornschier and Kriesi 2012; Bornschier 2018; Oesch and Rennwald 2018; Langsæther and Stubager 2019; Ares 2022). Indeed, recent events, such as the 2016 Brexit referendum (Hobolt 2016), the MAGA movement in the US, or the success of right-wing populist movements (e.g. the pan-European, anti-Islam, far-right political movement PEGIDA in Germany) have raised questions about the role of the working class in the political space (Westheuser and della Porta 2022).

This leads us to ask the following question: ‘What is the extent and breadth of opinion polarization on immigration and the EU between occupational classes?’

Our main contribution to the literature on attitudinal class divides is the introduction of the idea of within-class consensus, a concept born from a long tradition in sociological research on opinion polarization (DiMaggio et al. 1996). Previous research has documented sizable differences in average opinions between working-class and upper-class individuals (Langsæther and Stubager 2019; Ares 2022). However, we argue that within-class consensus, the attitudinal homogeneity of positions within each class, is of major importance in interpreting differences in average opinion between classes as outright class polarization or a cleavage (DiMaggio et al. 1996, p. 698). Only issues that show divergence between classes and consensus within classes are likely candidates for class polarization.

Furthermore, we overcome several limitations of previous research. First, a sizable proportion of empirical studies on the globalization-related divide in Europe are based on cross-sectional analyses or short time periods (Hobolt 2016; Langsæther and Stubager 2019; Lux et al. 2022; Pless et al. 2023). However, the globalization cleavage is expected to increase over time. Globalization pressures – the main proposed causes of this cleavage – have increased continually in recent decades. Moreover, long-term trend analyses are essential in polarization research because past opinion distributions provide a valuable yardstick for assessing the extent of current polarization (Fiorina and Abrams 2008). Second, studies that do track trends in globalization-related attitudes over long periods (Down and Wilson 2008; Caughey et al. 2019; Claassen and McLaren 2022; Teney and Rupieper 2023), rarely analyze attitudinal differences between occupational classes. However, occupational class, next to educational level, is said to constitute an important structural component of opinion divides on globalization-related issues (Bornschier and Kriesi 2012; Bornschier 2018; see Langsæther and Stubager 2019 or Kitschelt and Rehm 2014 for empirical evidence). Considering occupation as a major structural determinant of opinion polarization enables us to group citizens not only according to their skills level but also according to one of the most important places of socialization in an adult´s life (Oesch 2006; Kitschelt and Rehm 2014). Third, the few studies that focus on class differences and longer time frames (Perrett 2021; Ares 2022) focus on mean differences, ignoring the homogeneity or heterogeneity of opinions within classes.

Empirically, we provide an encompassing descriptive assessment of class-based opinion polarization on immigration and the EU for the German case. Immigration and the EU are two issue domains at the core of the globalization cleavage in Western Europe (e.g. Kriesi et al. 2008; Hooghe and Marks 2018). Germany is the EU member state with the largest population and one of the most globalized EU member states (Gygli et al. 2019). Furthermore, the availability of high-quality survey data in Germany allows us to follow trends over a long period and a broad range of attitudinal indicators. Beyond that, the German case features some particularities that make it an interesting case study. First, it did not have a strong, right-wing, radical, populist party that capitalizes on citizens´ anti-immigrant and anti-EU sentiments until 2013, which is relatively late compared to other European countries. Second, Germany fared relatively well during the global financial crisis and European sovereign debt crisis (Hopkins 2020). Nevertheless, these are not a priori reasons for assuming that there is no class polarization on globalization-related issues. According to Hopkins (2020), Germany – perceived as a creditor country during the financial crisis – should be susceptible to anti-system politics from the right (demonstrated by the sudden and forceful rise of the right-wing populist party AfD), particularly by those who do not perceive themselves as benefitting from immigration and European integration. Moreover, Germans tend to consider immigration and the EU as the most important problems faced by Germany in times of related large-scale shocks – such as the so-called 2015 ‘refugee crisis’ or the EU sovereign debt crisis in 2011–2014 (Teney and Rupieper 2023), despite Germany´s relative economic prosperity and stability during this period.

Our results document a clear and stable attitudinal class hierarchy: The working class is more immigration- and EU-skeptic than the higher service class on almost all issues. However, we also find that responses to most of the items which show large between-class differences are exceptionally diverse within the working class. This suggests limited mobilization potential of the working class as a whole on the basis of these issues. Finally, our selection of multiple indicators and fine-grained analysis enable us to identify those individual issues that might be most prone to cause conflict between classes.

Structural opinion polarization along occupational class lines

Following classical cleavage theory (Lipset and Rokkan 1967), the process of repeated conflicts between social groups consolidates collective identities and solidarity within the respective groups, which structures society. According to Simmel's (1992) conflict theory, social conflicts have not only a structuring but also an integrative function: Conflicts that occur within the common rules and norms require interactions between partners and imply the formation of groups (Bonacker 2005, pp. 9–29). Social conflicts can in turn become rooted in grassroots movements and hierarchical organizations that act as channels for the expression and mobilization of protest (Lipset and Rokkan 1967; Hooghe and Marks 2018). This requires that social groups in different structural positions differ in their politically relevant opinions and their expression. It is this overlap between social structure and expressed political opinion that we call structural opinion polarization.

Structural opinion polarization along occupational class lines is a particularly deep and durable form of societal division. This is because opinion differences do not occur between random individuals in a given society but between groups that share similar material living conditions, social networks, cultural practices, and potentially, identities (Blau 1977; DiMaggio et al. 1996; Baldassarri and Bearman 2007). Occupational class effectively characterizes the life chances, material circumstances (Oesch 2006), and everyday culture (DellaPosta et al. 2015; Fielding-Singh 2017) to which individuals are exposed in modern societies. Furthermore, social class structures social relations in many ways, and social networks are segregated along occupational and educational lines (Blau 1977; Alecu et al. 2022). The emergence of structural opinion polarization along occupational class lines implies that these class-specific similarities in experiencing and understanding the world are increasingly coupled with the expression of distinct political attitudes that diverge from those prevalent in other occupational classes. Even seemingly small cultural affinities within classes (e.g. regarding consumption habits, Fielding-Singh 2017) might contribute to polarization dynamics through feedback loops of social influence and homophily (DellaPosta et al. 2015).

Our research is particularly influenced by the consolidation principle developed by DiMaggio et al. (1996). Consolidation (Blau 1977) describes a lack of intersections between socio-structural parameters and is assumed to increase within-group interaction, decrease out-group interaction, and, most importantly, increase the likelihood of group mobilization. DiMaggio et al. (1996) extend this abstract notion of consolidation to describe the alignment of certain sociodemographic characteristics with certain political attitudes.

DiMaggio et al. (1996, p. 698) offer valuable suggestions on how to operationalize consolidation. First, there must be substantial attitudinal differences between sociodemographic groups. This is what we call between-group ‘divergence’, in line with Bramson et al. (2017). Second, the smaller the attitudinal differences between members within the respective social groups, the greater the likelihood of group mobilization on the basis of the attitudes in question. This is what we call within-group ‘consensus’ (see also Bramson et al. 2017). Within-group consensus captures the idea that political actors can only mobilize large parts of a group if its members are in consensus, even if the divergence between groups are high (DiMaggio et al. 1996).

Figure 1 illustrates that the consensus of responses within groups can change the interpretation of existing differences in average opinions between groups. It shows stylized histograms of responses to three hypothetical political issues (panels 1–3) on a hypothetical 10-point Likert scale. Each panel shows a histogram of two classes. The difference in the average opinions is five scale points in all three panels. Thus, if we were to only consider mean differences, all panels in Figure 1 would suggest the same level of polarization. However, considering the shape of the response distribution within each class, we maintain that group polarization is highest for the first issue and lowest for the third issue. In the first panel, opinion and class membership are aligned. Group-based political mobilization is made easier as both groups hold homogeneous positions. Here, social structure overlaps with opinion. In the second panel, the responses in each class are more spread out. Under these circumstances, group-based political mobilization would require greater effort to mobilize members with different opinions. Finally, in the third panel, the blue group is internally split in its opinion. Thus, even though the blue group holds opinions toward the third issue that are markedly different to those of the yellow group, political mobilization of the blue group would require effort to bridge an internal attitudinal divide.

Figure 1. Hypothetical histograms of responses to three items on separate political issues on a 10-point Likert scale for two groups. For each issue, the mean difference between the two groups remains the same, but the consensus within groups differs.

Our study focuses on the opinion polarization on globalization-related issues between occupational classes. Accordingly, we speak of structural opinion polarization if we observe (1) large differences in globalization-related issue positions between occupational classes (between-group divergence), and (2) high homogeneity in globalization-related issue positions within occupational classes (within-group consensus).

Previous research on opinion trends demonstrates that attitudes toward immigration and immigrants have remained stable or shifted in a pro-immigration direction in Western European countries since the late 1990s (Caughey et al. 2019; Dennison and Geddes 2019; Claassen and McLaren 2022). Against this background of a liberalizing trend, studies document that individuals in lower socioeconomic classes consistently express more restrictionist attitudes toward immigration relative to other classes (Langsæther and Stubager 2019; Lindh and McCall 2020; Ares 2022; Lux and Gülzau 2022) and are generally more conservative (Perrett 2021).

However, the class differences found in these studies are usually moderate, even between the two most widely separated classes in the respective analyses (Lindh and McCall 2020; Lux and Gülzau 2022, for Germany). Therefore, considering within-class consensus in addition to between-class divergence enables a more accurate assessment of structural opinion polarization. Indeed, structural opinion polarization might occur despite a lack of extreme opinion antagonism between occupational classes if these classes are homogeneous in their position.

Regarding trends in positions toward the EU, Down and Wilson (2008) find that the distribution of responses to an EU-support item became more dispersed among the general populations of various countries after the Maastricht Treaty. In Germany, the distribution remained stable after an initial increase in dispersion (Down and Wilson 2008, p. 41). While overall trends in support for the EU have been well documented (De Wilde 2021), we know much less about the differences between occupational classes on EU positions over time. Most studies in this strand of literature focus on education as a stratifying factor (see e.g. Hakhverdian et al. 2013 or Fernández and Eigmüller 2018).

Considering occupational class as a structural dimension of the globalization cleavage and based on findings from previous research, we formulate two hypotheses.

First, as globalization pressures such as immigration and Europeanization grew continuously between the 1990s and the most recent past in Europe (see Gygli et al. 2019), we expect an increase in antagonistic positions on the issues of immigration and the EU between occupational classes in the German population since the 1990s. According to previous research on trends in immigration positions, there could be a universal trend toward more liberal positions in all classes. However, even if such a trend exists, we would expect stable, or increasing, and sizable relative differences in positions between classes on most issues related to immigration or the EU (Hypothesis 1: between-class opinion divergence hypothesis).

Furthermore, under the assumption that the globalization cleavage is structurally rooted in the class structure, opinion consensus within occupational classes in the German population should also have increased since the 1990s (Hypothesis 2: within-class opinion consensus hypothesis). It is important to note that Hypothesis 2 is only interesting for issues where Hypothesis 1 is confirmed. Higher within-class consensus can be interpreted as class polarization only when differences in opinion exist.

Data and methods

Data sets

We use high-quality German datasets that allow us to construct our class variable of interest and include immigrant- and EU-related attitudinal items over a longer period. Our results are based on the General Social Survey of the Social Sciences (ALLBUS), European Social Survey (ESS), German Longitudinal Election Study (GLES), and International Social Survey Program – National Identity Module (ISSP). We restrict the data to those survey waves that include samples from both East and West Germany and detailed information on respondents’ occupation (see below). All analyses apply the survey weights made available by the data providers to account for the sampling design of each respective survey. We calculate the median date of all the interviews conducted within each wave for each survey and then use this date variable as a common time scale to compare trends across surveys.

Compositional changes within the different classes constitute a challenge in assessing class-based opinion polarization over time. For example, the working class has become more ethnically diverse, and the higher participation of women in the labor market has increased the proportion of women in low paid jobs. To make our between-class divergence and within-class consensus estimates comparable over time, we restrict our sample to German citizens. Furthermore, we apply weights obtained via Coarsened Exact Matching to adjust for compositional changes within classes. These weights adjust the distribution of sex, age, and East/West German residence within each class in each individual cross-section to resemble the class-specific distribution in the first survey that includes all relevant class information (the ALLBUS in 1992). We match members of the individual classes in each cross-section with their peers in 1992 based on the variables sex (female, male), age (coarsened into 29 years and below, 30–39, 40–49, 50–59, 60–69, 70 and above), and residence (either East or West Germany). Thus, sex, age, and place of residence are held constant at the levels of the 1992 ALLBUS sample in our analyses.

Class position

We measure occupational class position using the five-class Oesch scheme (Oesch 2006).¹ The scheme differentiates between ‘higher-grade service’ (e.g. managers, journalists, architects, and owners of large businesses), ‘lower-grade service’ (primary education teachers, nursing professionals, small business management), ‘small business owners’ (owning businesses with up to 10 employees, restaurants and the like), ‘skilled workers’ (clerks, secretaries, plumbers, personal care workers), and ‘unskilled workers’ (cashiers, housekeepers, waiters, taxi drivers, manufacturing laborers).

The five-class scheme differentiates jobs mainly according to the vertical economic dimension of inequality. The advantage of this broad categorization is that it follows a clear theoretical logic (vertical inequalities in skill levels and market success) and results in groups with sizable numbers which are demographically relevant. However, more differentiated variants of the Oesch class scheme are often used to account for the horizontal inequalities between occupations (Oesch and Rennwald 2018). In Online Appendix OA3, we show that distinguishing occupations on the basis of horizontal occupational characteristics leads to similar conclusions about the overall patterns of class polarization as the five-class scheme (see also Ares 2022). Thus, we are confident that our coarse categorization allows us to identify the most entrenched cleavages between classes.

We use respondents’ current job as measured by ISCO88 and ISCO08 codes, their type of employment relation (employed, self-employed, working in family business), and the number of employees (if self-employed) to categorize them into one of the five classes. If respondents were retired or did not provide a response when asked about their current job, we use information about their former job to construct the class scheme. If respondents did not have a job or did not provide information about their current and former job, we use their spouse’s job.²

Figure 2 shows how the percentage of individuals in the five classes in each cross-section evolved over time. In line with previous research (Oesch 2015), we find that the proportion of skilled workers declined steadily, whereas the proportion of positions in the lower-grade service class rose. The proportion of unskilled workers and small business owners remained at a steady low level.

Figure 2. Percentage of respondents belonging to the five classes used in our analysis. Pooled data from ALLBUS, ESS, GLES, and ISSP. Curves are fitted with locally weighted regressions.

Opinions on immigrants, immigration, and the EU

As is common practice in the polarization literature (Park 2018), we include all available items that fulfill broad criteria to avoid cherry-picking individual items that fit our hypotheses. We employ the following criteria to select the outcome variables. First, the items must match our thematic focus on immigration- and EU-related opinions. Second, they must have been asked in at least three survey waves to investigate trends. Finally, the items must target immigrants as a general group and not specific ethnic groups.

Our overall pool of items consists of 32 items that meet these criteria. Table A1 in the Appendix summarizes our main items, the respective wording of the survey questions, the response scales, and the years in which the item was featured. Our main analyses will focus on a smaller proportion of these items that show the largest mean differences to keep the analysis concise. We categorize the selected items from the three surveys into different domains to assign short labels to each item. For the immigration-related items, our data contains items that measure sentiments toward active exclusion of immigrants (XEN, xenophobia), effects of immigration on culture (CUL), economic effects of the presence of immigrants (ECO), effects on social welfare (WEL), assimilation (ASS), and immigration policy-related positions (IMP). For EU-related items, our data is composed of items asking about personal attachment to Europe (EUA), attitudes toward European Unification (EUU), and trust in European institutions (EUT). If necessary, items were recoded such that higher values indicate positions which are more anti-immigration or anti-EU.

Statistics of between-class divergence and within-class consensus

We define structural opinion polarization on an individual issue as an empirical state of society where divergence between classes in stated positions toward a certain issue is high, and within-class consensus in stated positions toward this issue is high. We report two statistics that characterize the distributions within classes and distributional differences between classes for each issue.

To measure between-class divergence, we calculate the differences between the class-specific mean responses (Bramson et al. 2017). Mean values are an effective way to investigate larger trends in the central tendency of distributions and provide an initial indication of group polarization if there are sizable and stable (or even diverging) mean differences between groups. Furthermore, differences in group-specific mean values are widely used in the literature and therefore provide a valuable starting point to demonstrate the added value of investigating within-class consensus in addition to between-class divergence.

Within-class consensus is measured by the Van der Eijk agreement index (Van der Eijk 2001) (henceforth, VDE agreement), a measure of polarization designed for ordinal response scales. Consensus occurs when members of a particular class overwhelmingly hold similar positions toward a certain issue; for example, if a large proportion of members of a certain class place themselves on the same response option on a Likert scale. VDE agreement ranges from −1 to 1, where 1 indicates perfect consensus (all members of a group choose the same response option), and −1 indicates a polarized distribution with 50 percent of cases at each of the two extreme ends of the response scale. For example, in panel (1) of Figure 1, both classes have a VDE agreement of 0.83, whereas it is 0.77 in panel (2). In panel (3), the blue class has a VDE agreement value of 0.18. In Online Appendix OA5, we provide a more detailed demonstration of how to calculate VDE agreement. We use the R package agrmt (Ruedin 2021).

A unique feature of the VDE agreement measure is that the midpoint 0 has a sensible interpretation, indicating a flat distribution (Aeppli and Ruedin 2022). However, the interpretation of values in the range between 0 and 1 (or −1 and 0) is not as straightforward when comparing items of different scale lengths, a limitation that VDE agreement shares with other polarization measures (Aeppli and Ruedin 2022). For this reason, we follow the advice of a recent simulation study that strongly encourages researchers to (a) study items individually, and (b) rely on graphical depictions of the overall distributions to bolster the main conclusions (Aeppli and Ruedin 2022). Our results are robust to using the standard deviation as an alternative measure of within-class consensus (see Online Appendix OA4).

Presentation of results and descriptive strategy

Description is an essential but often overlooked task in the social sciences (Gerring 2012) that affords researchers enormous degrees of freedom. Our aim is to provide an encompassing country-specific description that allows us to differentiate trends in opinions toward particular issues. Previous studies aggregate responses to different items from different surveys into a single measure of a latent trait of a higher-level unit of analysis, such as countries or groups within countries (Caughey et al. 2019; Claassen and McLaren 2022), making far-ranging modeling assumptions. These aggregation methods are essential when making concise country comparisons. However, our country-specific analysis allows us to screen all available items and thus identify instances of indicators that deviate from larger societal trends. We view responses to individual items not as measurements of a latent trait but as statements about certain political topics in a specific social situation (i.e. an interview with a stranger in a time-specific societal context).

To focus attention on those issues with the greatest potential to polarize, we rank our findings for the individual items according to their standardized mean differences over time. This ensures that those items with the largest class differences are shown at the top of each figure. The differences over time are calculated in the following manner: For each item, we calculate the differences between higher-grade service class and unskilled workers for each available time point on a standardized scale (Online Appendix OA1 shows these differences over time and 95% confidence intervals), and then average these time-specific differences over time. Higher-grade service class and unskilled workers consistently exhibit the greatest divergence in responses, serving as a valid benchmark to identify the issues that, based solely on mean differences, hold the highest potential for class polarization.

Furthermore, our discussion focuses mainly on the 16 out of 32 items with the largest group differences (see Figures 3–5). Results for the full set of items are shown and discussed in Online Appendix OA2.

Figure 3. Trends in the mean response for each occupational class for the 16 items showing large average class differences over time. Variables were standardized by dividing by scale length. For items marked with *, the response scale was reversed such that higher values imply more negative sentiments. See Table A1 for a description of all items.

Figure 4. Histograms for higher-grade service class (blue) and unskilled workers (yellow). Shown are the response distributions from the time point when standardized between-class differences were largest. To enhance the comparability of the two distributions, histograms for unskilled workers are flipped horizontally. For items marked with *, the response scale was reversed such that higher values imply more negative sentiments.

Figure 5. Time trends in both agreement scores (for each class, left y-axis) and mean differences (red, right y-axis) between higher-grade service class and unskilled workers.

Results

Trends in anti-immigration and anti-EU sentiment

We first investigate the average trends in responses to both immigration- and EU-related items for each class, as shown in Figure 3. The panels in Figure 3 are ordered according to average between-class differences, starting with items that displayed the largest class differences. An initial noteworthy observation about Figure 3 is that increases in negative sentiments toward immigrants or less EU-friendly positions occur to only a limited extent. Higher values on the y-axis in Figure 3 indicate a higher aversion toward immigration or the EU on the respective item-specific scales. On most issues displayed in the respective panels, average positions remained relatively stable across all groups. This attitudinal stability is impressive, because even the virulent 2015 debate on immigration in Germany during the so-called ‘European refugee crisis’ did lead to an anti-immigration backlash in some classes. For some immigration items, we even see shifts towards more immigration-friendly positions. For example, xenophobic statements (XEN1, XEN3, and XEN2) found less and less appeal in all classes over time. Similarly, the idea that immigration is bad for the economy or for native jobs develops in a more immigration-friendly direction (ECO3 and ECO1). These results are in line with previous research (Dennison and Geddes 2019; Claassen and McLaren 2022; Teney and Rupieper 2023).

However, Figure 3 also shows that the higher-grade and lower-grade service classes consistently hold more favorable positions toward immigration and are more attached to the EU than skilled and unskilled workers. Furthermore, class differences in mean positions remain stable over time and are statistically significant (Online Appendix A1). The empirical regularity of this attitudinal hierarchy across the range of political issues relating to immigration and the EU is striking, and also holds when more fine-grained horizontal class differences are considered (see Online Appendix OA3). Another important finding is the lack of systematic variation in between-class differences to responses to two categories of items that either tap into whether immigrants are perceived as a cultural threat or as a labor market threat. Unskilled workers even show more positive immigrant attitudes over time on items measuring the labor market threat dimension (ECO3, XEN1, ECO1, ECO4). This finding contrasts with labor market competition theory (Malhotra et al. 2013; Helbling and Kriesi 2014), according to which we would have expected unskilled workers in Germany to consider immigrants to be more of an economic threat in times of rising immigration. One possible explanation is that the large proportion of immigrants in low-skilled positions in Germany represent not only competition to German unskilled workers, but also contact opportunities.

These results allow us to reiterate that our notion of class divergence is based on relative differences between classes, not the extremity of opinions in these classes. Most panels in Figure 3 show that even among unskilled workers, the mean values indicate either an average neutral response or even a slight disagreement with anti-immigrant statements. For example, obviously xenophobic and segregationist statements, as in panel 3 (XEN3), did not receive average agreement in all classes in the 2010s. However, while average unskilled workers might not actively support xenophobic political campaigns, they might not favor campaigns that are actively against xenophobia, whereas the higher-grade service class expresses strong disagreement with xenophobic statements. For these differences to become politically mobilizable and socially relevant between classes, the respective classes must be in consensus (see next section).

There are individual instances of issues where mean differences between classes decrease (in particular, XEN3, IMP1, and ECO1 in Figure 3). However, these trends toward diminishing differences are an exception. It is undeniable that stable dissimilarities exist between classes. Therefore, we continue under the assumption that Hypothesis 1 (between-class divergence) is confirmed, particularly for the contrast between higher-grade service class and unskilled workers.

In the following, we assess whether investigating within-group consensus adds meaningful nuance to our interpretation of average opinion differences as structural opinion polarization. We adopt a two-step approach. First, we plot and investigate class polarization in those years in which individual indicators showed the largest average response differences between the two most widely separated classes (Figure 4). In the second step, we measure VDE agreement and mean differences over time to identify issues that show polarizing dynamics (Figure 5).

Structural opinion polarization and working-class heterogeneity

Figure 4 shows the distribution of the responses of unskilled workers and the higher-grade service class in the form of histograms. From our previous analysis, we know that these two classes consistently showed the largest differences, and thus present the most likely cases for structural opinion polarization. Each panel shows two histograms for the year in which the responses to an item reveal the largest differences between the higher-grade service class and unskilled workers.

The main result from Figure 4 is that workers have substantially more heterogeneous positions toward most issues than the higher-grade service class. The panels are sorted according to the individual item’s standardized between-group differences, starting with the largest differences. In particular, in the responses to items that exhibit the largest differences – those at the top of Figure 4 (CUL1, ECO3, XEN1, XEN2, XEN3) – we can see that working-class responses span the whole range of possible response options and even show slight trimodal or bimodal patterns. In contrast to the responses of unskilled workers, those of the higher-grade service class are more concentrated at the lower end of the scale on these items, indicating consensual pro-immigration sentiments. Higher dissensus among workers is also visible in ECO1, CUL3, CUL2, and ECO4, and thus holds for nine out of the 16 item-year observations shown in Figure 4. This finding of greater opinion homogeneity among the higher-grade service class is also supported by higher values of the VDE agreement index (see our discussion of Figure 5 below).

Furthermore, Figure 4 depicts instances where both classes show rather low within-group consensus. This applies to the issues of whether foreigners put a strain on the welfare system (WEL1), immigrants should receive local voting rights (ASS3), and foreigners cause problems on the housing market (WEL4). Thus, for most of the single-year snapshots shown in Figure 4, we cannot detect strong between-class polarization because of low within-group consensus in either one or both groups.

Investigating consensus within classes is a sensible approach to narrowing down those issues on which opinions are actually polarized between groups. Indeed, individual response distributions in Figure 4 suggest a higher extent of structural opinion polarization relative to responses to other items. Responses to IMP4 in panel 6 show that unskilled workers were overwhelmingly in favor of immigration restrictions in the aftermath of the 2015 so-called ‘refugee crisis’ (in 2017), whereas higher-grade service-class members tended toward the center of this response scale. We can see similar tendencies of between-class difference and within-class consensus in the subjective attachment to Europe (EUA1) in 2018, and attitudes toward the requirement for foreigners to adopt German customs (ASS1) in 2016.

The main insight from Figure 4 is that mean differences can show only one aspect of group polarization: the aggregate divergence between classes. However, while it is true that the two classes hold markedly different positions on average, we can also find a wide range of positions toward the respective issues within at least one class. In the case of issues where positions diverge most between classes, this is the working class. This has substantial implications for the assessment of class polarization. A globalization cleavage along occupational lines would require that a large fraction of the two most widely separated classes can be mobilized on the basis of their positions on globalization-related issues. However, if one class holds very heterogeneous positions on a variety of issues, mobilization on the basis of these issues that includes the entire socioeconomic stratum is very unlikely. This is contrary to our Hypothesis 2.

Trends in structural opinion polarization and item-specific analysis

Figure 4 provides an impression of the response distribution at one point in time. However, we are also interested in the potential trends toward structural opinion polarization over the last three decades. To this end, Figure 5 shows trends in time-specific measurements of mean differences in red (right y-axis) and VDE agreement scores (left y-axis) for the most divergent classes (blue: higher-grade service, yellow: unskilled workers). Thus, a trend toward structural opinion polarization would be visible as a movement of the red line toward the upper right corner, and a rise or high level of both the blue and yellow lines. Figure 5 illustrates the three dimensions of time, mean differences, and VDE agreement simultaneously. Thus, it accounts for the fact that high VDE agreement values are only indicative of class polarization when group differences are large or are decreasing compared to previous time points. In Online Appendix OA2, we present an alternative way of depicting these three dimensions for all 32 items in a more compressed single graph.

As a first impression, Figure 5 supports the main finding from Figure 4 about working-class dissensus: in 7 out of 16 panels, we can see that the yellow dots are clearly positioned below the blue dots, indicating greater heterogeneity of opinion within the working class (see CUL1, ECO3, XEN1, XEN3, XEN2, ECO1, CUL3). In other instances, both classes show low consensus over time. This applies to WEL1 and ASS3.

We now ask a more specific question about trends: Are there any issues that have become continuously more polarized over the last decades? The trends in responses to the single items shown in Figure 5 can be grouped into five sets according to their patterns.

The first set of globalization-related issues shows a constant division within the unskilled worker class with agreement values close to 0, whereas the higher-grade service class was constantly or grew increasingly homogeneous (higher numbers on the y-axis for the blue dots). The VDE agreement for unskilled workers stays well below (or falls back to) 0.2 for these issues. This can be seen in the questions of whether one feels like a stranger because of immigrants (panel 1, CUL1), foreigners take away jobs (panel 2, ECO3), foreigners should be sent home when jobs are scarce (panel 3, XEN1), the political participation of immigrants (panel 4, XEN2), and, to a lesser extent, the question of whether foreigners enrich culture (panel 9, CUL3).

In the second set, we observe issues where increasing agreement or stable levels of high agreement in both classes are accompanied by a decrease in mean differences. This pattern signals convergence between groups over longer time horizons. For example, group differences on the opinion that foreigners should marry among themselves (panel 5, XEN3) decrease substantially and agreement increases in both classes. Similar patterns can be seen in responses to panel 7 (ECO1) after 2014, panel 11 (WEL4), and panel 13 (IMP1).

The third set of items shows constantly low consensus in both classes over time. This applies to panel 8 (WEL1), which additionally shows declining mean differences, and panel 14 (CUL2), which shows a decline in agreement in both classes and stable group differences. Relatedly, whether immigrants should receive communal voting rights (panel 10, ASS3) is not a consensual opinion in both classes, but the latest measurement in 2016 shows increasing mean differences and increasing agreement in the working class. However, polarization in ASS3 is still limited by the very heterogeneous responses within the higher-grade service class.

The fourth set comprises items measuring the perceived economic impact of immigration. The findings for these items suggest decreasing structural polarization after 2014. Responses to ECO1 in panel 7 exhibited their largest mean differences in 2014 (see also panel 16, ECO4), but the mean differences declined in the following years (see also the decrease from 2006 to 2016 in panel 2, ECO3). This fits into a more general observation regarding positions on the effect of immigration on the economy, which were more structurally polarized between 2000 and 2014 than after 2015 (see Online Appendix OA2), suggesting that the issues at the core of between-class polarization can shift over time. Our data suggests that after 2015, the belief that immigration benefits the economy seems to be more widespread among individuals of all classes than before.

The final set of issues shows an increase in mean differences, accompanied by stable high agreement or growing agreement in both groups during the final measurement periods. These items are candidates for current structural opinion polarization: IMP4 in panel 6 on immigration policy, EUA1 in panel 15 on attachment toward the EU, and ASS1 in panel 12 on assimilation expectations toward immigrants. Interestingly, members of the working class that are more in agreement than their responses to those items than the higher-grade service class (but the higher-grade service class also has relatively high agreement scores). Well before the so-called 2015 ‘refugee crisis’, unskilled workers held the position that the number of immigrants should decrease (IMP3 in 2013, Online Appendix OA2), whereas the higher-grade service class accepted a moderate inflow of immigrants. The ‘crisis’ sharpened this contrast (see IMP4 in 2017 in Figure 4). Similar trends can be observed for items on attachment to the EU and its population (panel 15, EUA1), where the two classes grew increasingly apart from 2000 to 2018. On the issue of whether foreigners should adopt German customs (panel 12, ASS1), the two classes are more aligned because both overwhelmingly place their responses above the mid-point, but higher-grade service class members situate themselves in the middle to upper part of the scale, whereas most unskilled workers choose to respond in consensus at the upper end of the scale.

We present further aspects of our data that informed our main conclusions in the Online Supplementary Material. In Online Appendix OA2, we show the results for all 32 items in one graph, including those with lower group differences. In Online Appendix OA3, we further differentiate occupational classes along a horizontal dimension of work logic, singling out socio-cultural jobs among the upper classes and manual jobs among the working class. The main result of this analysis is that the overall hierarchy along the vertical axis expresses the most pronounced differences in opinions. In Online Appendix OA4, we present results using the standard deviation of responses within classes as an alternative measure of within-group consensus.

Overall, time trends over the last decades either show de-polarizing trends or stable and low within-class consensus in at least one class, most often the working class. There are individual issues that stand out as more polarized between classes than other issues at certain time points. However, even if members of one class agree on a particular issue (e.g. immigration restriction), they are likely divided by other, related issues (e.g. agreement to xenophobic statements).

Discussion and conclusion

This article presents an encompassing study of the structural polarization of immigration- and EU-related opinions for the German case. Structural opinion polarization is the alignment of political preferences with social realities shaped by material circumstances. We argue that opinion divergence between social groups should be interpreted as structural opinion polarization only if it is accompanied by high consensus within the respective groups (see DiMaggio et al. 1996). Existing differences can only be mobilized along group lines if consensus is high within groups.

Our results paint a multifaceted picture of polarization on globalization-related issues between occupational classes. Overall, we provide evidence in line with previous research that there are substantial class differences in opinions toward immigration and the EU. However, our study also cautions against overstating the overall class conflict on globalization-related issues, for three main reasons.

First, across most items related to immigration and the EU, we find evidence of high heterogeneity of opinion in at least one class. Most often, we find working-class dissensus: the unskilled working class holds more heterogeneous opinions on political matters than the higher-grade service class. By contrast, the upper classes are often more unified in their opinions. Interestingly, this pattern is most pronounced for responses to items with the largest between-class divergence. One illustrative example concerns xenophobic statements. Our data suggests that, when confronted with a xenophobic statement in an everyday situation, a person belonging to the higher-grade service class would very likely vehemently disagree. If the person were an unskilled worker, they would disagree, agree, or take a neutral position with almost equal probability. This absence of a class-specific consensus on political issues among workers is in line with previous research, which describes the working class as a ‘demobilized class’ (Westheuser and della Porta 2022), or demonstrates general lower levels of political interest and higher proportions of voter abstention among the working class (Oesch and Rennwald 2018). Under conditions where networks are segregated by class, our finding also imply that individuals in the working class are likely to interact with people with different views in their workplace, family, and wider social circles.

Second, we are unable to detect clear trends toward increasing structural opinion polarization over the last decades for most of the issues. Responses to most items suggest stable or decreasing levels of class polarization. Interestingly, opinions on the economic effects of immigration were most polarized before 2014, but we detect a convergence between the classes after 2015. This implies that structural opinion polarization is sometimes a transitory phenomenon that manifests as a deviation from decade-long time trends in pro-immigration directions (Caughey et al. 2019; Dennison and Geddes 2019; Claassen and McLaren 2022). Convergence in issue positions between classes on economic aspects of immigration might be caused by declining feelings of competition with immigrants in working-class occupations, which might be an effect of increasing contact opportunities with immigrant workers. Experiences of contact and competition could also partially explain lower within-class consensus if only a fraction of members of one class experience contact or if certain occupations experience more competition than others (Malhotra et al. 2013). While future research must substantiate whether the convergence on economic issues is durable, the ability to identify such societal patterns highlights the advantage of our fine-grained, long-term trend analyses.

Finally, the classes in the middle of the status hierarchy (lower-grade service class, small business owners, and skilled workers) occupy a centrist political position between higher-grade service professions and unskilled workers on globalization-related issues. These classes might act as brokers between the two most widely separated classes: an unskilled worker might not have strong personal connections to a university professor but maybe to a trained, skilled office worker, and this office worker might in turn share many of their political views with a teacher (lower-grade service). The presence of such interconnected networks would limit the potential siloed transmission of anti-globalization sentiment among unskilled workers.

However, it should be noted that there are individual issues that signal between-class division. Opinions about the economic consequences of immigration were polarized between classes before 2015, and immigration restriction, subjective EU attachment, and assimilation expectations continue to be the issues on which opinions are most polarized between social classes. For example, a large proportion of unskilled workers stated that ‘immigration opportunities to Germany should be restricted’ during the so-called 2015 ‘refugee crisis’. Whether these differences in opinion manifest in outright political conflict requires further research. It is safe to assume that a lot depends on the strategies of the political actors. Mobilizing unskilled workers by emphasizing immigration restrictions may pull workers away from the upper classes, establishing a stable class-based conflict. However, outright xenophobic rhetoric or disregard for the role of immigrants in the economy will likely divide the working class. When asked whether Germany should allow many, some, a few, or no immigrants from poorer countries, we can see a growing consensus across all classes that ‘some’ immigration should be allowed. This indicates that members of all classes know about the benefits of immigration. However, under specific circumstances of high public attention towards the immigration issue and high inflows of immigrants, unskilled workers might view immigration as too much of a burden. Thus, our research does not suggest issue-encompassing class divisions, but rather transitory and context-dependent polarization on individual issues.

Our study has several limitations that present opportunities for future research. First, our results are very likely to be context dependent. In the introduction, we already noted the specifics of the German case. Future research with a similar polarization operationalization in other national contexts would enable the generalizability of our main findings to be assessed. Second, our study ignores potential structural parameters other than class. In particular, there might be value in researching the particularities of the structural position of the small but relatively stable, radical, anti-immigrant parts of the working class. Are these radical opinions occurring in geographic clusters, or are they present in specific occupations? Third, our focus on individual items does not provide measurements of latent attitudes, but the expression of opinions toward item-specific issues in an interaction with strangers (i.e. a survey interview) within a greater time-specific societal context. This includes the risk of social desirability bias. However, our results do not show that the upper classes started to express more negative opinions on immigration topics with the rise of the AfD. This suggests that the upper classes did not hold back potential negative views on immigration before 2015, which they then could have revealed after the rise of the AfD and the accompanying discursive shift in German public debates on immigration. However, it could still be the case that a proportion of the working class are, if not necessarily hiding their true beliefs, very Uncertain as to what to respond to some of the items in this study. This could partly explain working class opinion heterogeneity.

Despite these shortcomings, our study highlights the importance of the fine-grained assessment of structural opinion polarization trends to enrich both the empirical debate on how to measure opinion polarization and the theoretical debate on the globalization cleavage. We hope that future studies will employ a similar, fine-grained approach to other countries or other issue domains. In this way, an encompassing picture of class polarization can emerge that allows us to assess the likelihood of political conflicts entrenched in the social structure of societies.

Acknowledgments

We would like to thank Giuseppe Pietrantuono, Bujar Aruqaj, Swen Hutter, the participants of the OSI Brown Bag at the Freie Universität Berlin; Steffen Mau, Thomas Lux and Linus Westheuser and the participants of the colloquium organized by the chair of Macrosociology at the HU Berlin; as well as Marc Swyngedouw for their insightful comments on previous versions of the manuscript.

Data availability statement

Replication packages can be found under DOI:10.17605/OSF.IO/5D2EW. Survey data sets are available to researchers after signing a data sharing agreement and downloadable from the following sources (DOIs): 10.4232/1.13748; 10.4232/1.13648; 10.4232/1.13471; and the ESS data portal at https://www.europeansocialsurvey.org/ (last accessed May 26, 2021).

Additional information

Funding

This research was made possible thanks to the generous financial support of the Deutsche Forschungsgemeinschaft (DFG) (project number 440923825).

Notes

1 We primarily relied on the scripts provided on Daniel Oesch’s website (https://people.unil.ch/danieloesch/scripts/) and the oesch ado for Stata (programmed by Simon Kaiser, https://ideas.repec.org/c/boc/bocode/s458490.html).

2 Only a small minority of respondents were classified into a certain class based on their spouse’s job because most respondents had a job at some point in their life. Since the first ISSP-National Identity Module did not ask for past jobs in 1995, we use only the current job for all waves of the ISSP to stay consistent within the ISSP.

Appendix

Table A1. Items used in this study. Column ‘reverse coded?’ shows whether variables were recoded such that higher values imply more anti-immigrant or anti-EU opinions.

Original variable name	Data set	Label used in paper	Item text	Item scale	Reverse coded?	Measurement years
ma02	ALLBUS	Home when jobs scarce (XEN1)	When jobs get scarce, the foreigners living in Germany should be sent home again	1 ‘Completely disagree’ to 7 ‘Completely agree'		1994, 1996, 2000, 2002, 2006, 2010, 2012, 2016
ma03	ALLBUS	Deny im. pol. participation (XEN2)	Foreigners living in Germany should be prohibited from taking part in any kind of political activity	1 ‘Completely disagree’ to 7 ‘Completely agree'		1994, 1996, 2000, 2002, 2006, 2010, 2012, 2016
ma04	ALLBUS	Im. no intermarriage (XEN3)	Foreigners living in Germany should choose to marry people of their own nationality	1 ‘Completely disagree’ to 7 ‘Completely agree'		1994, 1996, 2000, 2002, 2006, 2010, 2012, 2016
ma09	ALLBUS	Feel like stranger (CUL1)	With so many foreigners in Germany, one feels increasingly like a stranger in one's own country	1 ‘Completely disagree’ to 7 ‘Completely agree'		1996, 2006, 2016
imueclt	ESS	Im. undermine/ enrich culture (CUL2)	And, using this card, would you say that Germany's cultural life is generally undermined or enriched by people coming to live here from other countries?	0 ‘Cultural life undermined’ to 10 ‘Cultural life enriched'	yes	2003, 2004, 2006, 2008, 2010, 2012, 2014, 2016, 2018
mp03	ALLBUS	Im. enrich culture (CUL3)	What about the following statements about the foreigners who live in Germany? They enrich the cultural life of Germany.	1 ‘Completely disagree’ to 7 ‘Completely agree'		1996, 2006, 2016
v45	ISSP	Im. bring new ideas (CUL4)	Immigrants enrich Germany through new ideas and cultures	1 ‘Agree fully’, 2 ‘Agree’, 3 ‘Neither’, 4 ‘Do not agree’, 5 ‘Do not agree at all'	yes	1995, 2003, 2013
imwbcnt	ESS	Im. Germany worse/better (CUL5)	Is Germany made a worse or a better place to live by people coming to live here from other countries?	0 ‘Worse place to live’ to 10 ‘Better place to live'	yes	2003, 2004, 2006, 2008, 2010, 2012, 2014, 2016, 2018
imbgeco	ESS	Im. bad/ good for econ. (ECO1)	Would you say it is generally bad or good for Germany's economy that people come to live here from other countries?	0 ‘Bad for the economy’ to 10 ‘Good for the economy'	yes	2003, 2004, 2006, 2008, 2010, 2012, 2014, 2016, 2018
v43	ISSP	Im. good for econ. (ECO2)	Immigrants are generally good for the German economy	1 ‘Agree fully’, 2 ‘Agree’, 3 ‘Neither’, 4 ‘Do not agree’, 5 ‘Do not agree at all’	yes	1995, 2003, 2013
mp06	ALLBUS	Take away jobs (ECO3)	What about the following statements about the foreigners who live in Germany? They take jobs away from Germans	1 ‘Completely disagree’ to 7 ‘Completely agree'		1996, 2006, 2016
v44	ISSP	Take away jobs (ECO4)	Immigrants take away jobs from those who are born in Germany	1 ‘Agree fully’, 2 ‘Agree’, 3 ‘Neither’, 4 ‘Do not agree’, 5 ‘Do not agree at all'	yes	1995, 2003, 2013
mp01	ALLBUS	Do jobs Germans won't (ECO5)	What about the following statements about the foreigners who live in Germany? The foreigners who live in Germany do the work Germans don't want to do.	1 ‘completely disagree’ to 7 ‘completely agree'	yes	1996, 2006, 2016
ma01	ALLBUS	Im. adopt German customs (ASS1)	The foreigners living in Germany should adapt their way of life a little more closely to the German way of life.	1 ‘Completely disagree’ to 7 ‘Completely agree'		1994, 1996, 2000, 2002, 2006, 2010, 2012, 2016
ma05	ALLBUS	Allow dual citizenship (ASS2)	Foreigners living in Germany should be able to acquire German citizenship without having to give up their own citizenship, i.e. DUAL CITIZENSHIP should be possible.	1 ‘Completely disagree’ to 7 ‘Completely agree'	yes	1996, 2006, 2016
ma07	ALLBUS	Im. communal voting rights (ASS3)	All foreigners living in Germany – no matter where they come from – should have the VOTE IN MUNICIPAL (LOCAL) ELECTIONS.	1 ‘Completely disagree’ to 7 ‘Completely agree'	yes	1996, 2006, 2016
impcntr	ESS	Im. from poor countries (IMP1)	To what extent do you think Germany should allow people from the poorer countries outside Europe to come and live here?	1 ‘Allow many to come and live here’ 2 ‘Allow some’ 3 ‘Allow a few’ 4 ‘Allow none'		2003, 2004, 2006, 2008, 2010, 2012, 2014, 2016, 2018
v47	ISSP	Measures against illegal im. (IMP2)	Germany should implement harsher measures to fight illegal immigrants.	1 ‘Agree fully’, 2 ‘Agree’, 3 ‘Neither’, 4 ‘Do not agree’, 5 ‘Do not agree at all’	yes	1995, 2003, 2013
v48	ISSP	Increase/ decrease im. (IMP3)	Do you think that the number of immigrants to Germany should be …	1 ‘should be increased substantially’, 2 ‘increased slightly’, 3 ‘should stay as it is’, 4 ‘reduced slightly’, 5 ‘reduced substantially'		1995, 2003, 2013
v88	gles	Facilitate/ restrict immigr. (IMP4)	And what position do you take on immigration for foreigners? Please use the scale.	1 ‘immigration for foreigners should be easier’ to 11 ‘immigration for foreigners should be more difficult'		2009, 2013, 2017
mp02	ALLBUS	For. strain social welfare (WEL1)	What about the following statements about the foreigners who live in Germany? The foreigners who live in Germany are a burden on the social welfare system.	1 ‘Completely disagree’ to 7 ‘Completely agree'		1996, 2006, 2016
mp05	ALLBUS	For. bolster pensions (WEL2)	What about the following statements about the foreigners who live in Germany? They help to secure old age pensions.	1 ‘Completely disagree’ to 7 ‘Completely agree'		1996, 2006, 2016
ma06	ALLBUS	For. same social benefits (WEL3)	Foreigners living in Germany should be entitled to THE SAME WELFARE BENEFITS AND OTHER SOCIAL ENTITLEMENTS as Germans.	1 ‘Completely disagree’ to 7 ‘Completely agree'	yes	1996, 2006, 2016
mp04	ALLBUS	For. housing problems (WEL4)	What about the following statements about the foreigners who live in Germany? Their presence in Germany leads to problems in the housing market.	1 ‘Completely disagree’ to 7 ‘Completely agree'		1996, 2006, 2016
v42	ISSP	Im. increase crime (CRI1)	How far do you agree or disagree with the following statements? Immigrants increase the crime rate.	1 ‘Agree fully’, 2 ‘Agree’, 3 ‘Neither’, 4 ‘Do not agree’, 5 ‘Do not agree at all’	yes	1995, 2003, 2013
mp07	ALLBUS	For. commit more crimes (CRI2)	What about the following statements about the foreigners who live in Germany? They commit crimes more often than Germans.	1 ‘Completely disagree’ to 7 ‘Completely agree'		1996, 2006, 2016
pn17	ALLBUS	Attachment to EU (EUA1)	Now we would like to know how strongly you identify with your own town (community) and its inhabitants. And what about the European Union and its population?	1 Very strong attachment 2 Pretty strong attachment 3 Only weak attachment 4 No attachment at all		2000, 2008, 2016, 2018
v4	ISSP	Close to Europe (EUA2)	How far do you feel attached to Europe?	1 Very strongly attached to 4 Not at all attached		1995, 2003, 2013
trstep	ESS	Trust European Parliament (EUT1)	Please tell me on a score of 0–10 how much you personally trust each of the institutions I read out. 0 means you do not trust an institution at all, and 10 means you have complete trust. Firstly … … the European Parliament?	0 ‘No trust at all’ to 10 ‘Complete trust'	yes	2003, 2004, 2006, 2008, 2010, 2012, 2014, 2016, 2018
pt19	ALLBUS	Trust EU commission (EUT2)	I am now going to read out a number of public institutions and organizations.	1 ‘Absolutely no trust at all’ to 7 ‘A great deal of trust'	yes	1994, 2000, 2008, 2018
pt20	ALLBUS	Trust EU parliament (EUT3)	I am now going to read out a number of public institutions and organizations.	1 ‘Absolutely no trust at all to’ 7 ‘A great deal of trust'	yes	1994, 2000, 2008, 2018
euftf	ESS	EU unif. further/ too far (EUU1)	Now thinking about the European Union, some say European unification should go further. Others say it has already gone too far. Using this card, what number on the scale best describes your position?	0 ‘Unification already gone too far’ to 10 ‘Unification go further'	yes	2004, 2006, 2008, 2012, 2014, 2016, 2018