Affective attachment to the EU: Questioning the importance of childhood socialization
Zuzana Ringlerova, Department of International Relations and European Studies, Masaryk University, Brno, Czechia
Note: This is a preliminary version of the manuscript. The final version was published in the European Union Politics. When directly quoting the manuscript, please, use the final version.1
Ringlerova Z. Affective attachment to the EU: Questioning the importance of childhood socialization. European Union Politics. 2020;21(4):545-568. doi: 10.1177/1465116520950833
Abstract
In a time of rising Euroscepticism across Europe, diffuse support for the European Union (EU) is an especially important concept as it provides a source of stability for the EU. How important is childhood political socialization for the development of diffuse support? The extant literature emphasizes the role of childhood socialization. However, these studies are based on analyses that cannot fully distinguish between the cohort effect and the life-cycle effect. This study overcomes this limitation by looking at a more suitable case (the European Union) and by using a novel technique that effectively distinguishes the cohort effect from the life-cycle effect. The findings show that individuals who experienced early life political socialization in the EU have equal levels of diffuse support as individuals who grew up outside the EU. I thus argue that diffuse support develops through experience in adult life, and childhood political socialization is not essential for its development. Keywords
Acknowledgements I thank Jay McCann and Andrew L Roberts for giving detailed feedback on this study. I also thank the participants at the European Consortium for Political Research General Conference panel, participants at the research workshops at the Faculty of Social Studies at Masaryk University (MUNI) and the Faculty of Law at MUNI for their helpful feedback on the earlier drafts of this article. I thank the three anonymous reviewers for their careful, constructive, and truly helpful comments.
The author(s) disclosed receipt of the following financial support for the research authorship, and/or publication of this article: The work on this article has been funded by the MSCAfellow@MUNI project (grant No.
CZ.02.2.69/0.0/0.0/17_050/0008496). The author is grateful to the Mnistry of Education, Youth, and Sports of the Czech Republic for supporting this project.
1
Euroscepticism,political socialization,Age-Period-Cohort analysis,political support,childhood socialization
Corresponding author
Zuzana Ringlerova, Department of International Relations and European Studies, Masaryk University, Jostova 10, Brno, 60200, Czechia Email: ringler@fss.muni.cz
Introduction
Euroscepticism among European citizens is one of the most serious challenges facing the European Union (EU). The Brexit referendum, as well as a strong showing of populist Eurosceptic candidates in Dutch, French and Austrian elections, are just a few examples of this Eurosceptic wave. These developments suggest that large numbers of citizens in EU member states feel dissatisfied with the EU. This lack of support has potentially serious consequences for the stability of the EU as a political system. If a political system experiences an economic or other crisis, citizens' support is an important resource the system can draw upon in order to weather the crisis. Given the current wave of rising Euroscepticism, understanding why individuals support the political system of the EU becomes especially important.
In assessing the stability of political systems, one type of mass public political support - diffuse support - is particularly important. Diffuse support is an affective attachment to the political system. Thanks to this quality, diffuse support provides the political system with a source of resilience in times of crisis (Dalton, 2004; Easton, 1965; Norris, 1999). Although there is a fairly good accumulation of knowledge about the existence of diffuse support for political systems (Beaudonnet and Franklin, 2014; Dalton, 2004; Down and Wilson, 2017; Hooghe and Marks, 2004; Norris, 1999; Serricchio et al., 2013), it is not very clear how this source of political systems' stability comes about. The dominant line of the existing research emphasizes early life political socialization as the key factor in the development of diffuse support (Down and Wilson, 2013; Lutz et al., 2006; Mishler and Rose, 2007; Sapiro, 2004).
2
Other works, however, suggest that later life experience is the key determinant of diffuse support in adulthood (Bruter, 2009; Mishler and Rose, 2007). How important, then, is early life political socialization for the development of diffuse support? The goal of this article is to answer this question and improve our understanding of the sources of diffuse support for the EU as well as our understanding of diffuse support as a general concept.
I examine the relationship between early life political socialization and diffuse support by using a novel technique that allows me to overcome a problem that is endemic to all studies of early life socialization's effects on later life attitudes. Studies of childhood socialization's effects struggle to identify whether the observed differences between generations are a result of experience during one's childhood or whether these differences are a result of attitude changes as individuals go through the life cycle. The specificities of the historical development of the EU allow me to overcome this problem and employ an innovative methodology that effectively distinguishes the cohort effect from the life-cycle effect (Dinas and Stoker, 2014). By exploiting the fact that different countries joined European integration at different points in time, this methodology is able to focus directly on the impact of early life political socialization while controlling for the effect of age. Using data from the Eurobarometer studies, I find that early life political socialization into the European supranational political system does not influence the level of diffuse support for the EU in adulthood. In contrast to the prevailing view in the existing literature, I argue that early life political socialization into the political system of the EU does not determine diffuse support in later life and that diffuse support for the EU develops as a result of later life experience.
These findings have several practical implications for the EU. First, the findings imply that relying on early life political socialization as a generator of citizens' lasting affective support for the EU is not enough. The EU needs to provide positive experiences for adult citizens in order to maintain and reinforce citizens' diffuse support. Second, the results of this
3
study imply that gaining a new population either through immigration or through the accession of new EU member states does not necessarily constitute a weak spot in terms of public support for the political system of the EU. Given that early life political socialization is not essential for the development of an affective attachment to the EU, the new members of the population can learn to support their new political system as effectively as if they had grown up in it.
Literature review and hypothesis
Diffuse political support is a category within a broader concept of mass public political support. There are two types of political support: specific and diffuse. Specific support is a 'running-tally' type of attitude that fluctuates according to the current performance of the political system. If citizens are satisfied with the political system's performance, specific support is high. If they are not satisfied, specific support decreases. In contrast, diffuse support is an affective attitude. It is independent of the current performance of the political system. Diffuse support persists even in times when citizens become dissatisfied with the regime's policies (Easton, 1965; Harteveld et al., 2013; Norris, 1999). Although diffuse support is resistant to momentary changes in performance, from a long-term perspective, regime performance may affect diffuse support as well. Similarly, a long period of good regime performance helps generate diffuse support (Mishler and Rose, 2007).
Together, specific and diffuse support make up the overall political support that an individual has for a given political object. In other words, overall political support is a mix of these two types of support. In the case of some political objects (such as incumbent politicians), the specific type of support prevails in the mix. In cases of other political objects (such as the political nation), diffuse support is the prevailing type of support (Norris, 2011). Given these empirical findings, it may be tempting to equate diffuse support to support for the political system. This would be misleading, however. The quality of support (diffuse or specific)
4
is conceptually different from the object towards which support is directed. Keeping this theoretical distinction allows us to study whether a political system enjoys an affective (diffuse) or performance-based (specific) political support. Each of these types of political support provides the system with a different level of stability in times of crisis. Since people tend to view the performance of a political system more negatively during a crisis, a political system enjoying diffuse support has more resilience in times of crisis than a political system that draws its political support on a good performance.
Empirical studies of attitudes towards the EU show that there is both diffuse and specific support for the EU (Beaudonnet and Franklin, 2014; Eichenberg and Dalton, 2007; Gabel and Palmer, 1995; Gabel, 1998; Down and Wilson, 2017; Hobolt and de Vries, 2016; Hooghe and Marks, 2004; Ringlerova, 2015b; Serricchio et al., 2013; Torcal et al., 2012a; Torcal et al., 2012b). Since specific support and diffuse support are two distinct attitudes, they develop in different types of processes. Specific support is an up-to-date barometer reflecting the individual's current satisfaction with how well the political system works. Diffuse support, on the other hand, is not easily swayed by momentary changes in individuals' assessment of the political system's performance. How, then, does affective support for the political system develop?
The existing literature on the origins of diffuse support distinguishes between two views. According to one view, early life political socialization is the critical period for the development of affective support for the political system (Mishler and Rose, 2007). During childhood individuals first encounter the political system and form uncritically positive opinions of the political system and its representatives (Carter and Teten, 2002). Through formal education, children are taught to respect the political system, to be proud of their political community, and to respect the symbols of the political system, such as the flag or the anthem. All these experiences instil in young individuals an affective attachment to the political
5
system that persists into adulthood (Down and Wilson, 2013; Mishler and Rose, 2007; Sapiro, 2004). Support for this line of argument has been found not only in comparative research but also in research focusing on the EU. Studies on generational differences in public support for the EU show that younger generations of Europeans have higher overall support for the EU as well as higher diffuse support for the EU than older generations (Down and Wilson, 2013; Down and Wilson, 2017). Early life socialization is emphasized as the key factor explaining these between-generational differences (Down and Wilson, 2013; Lutz et al., 2006).
How does early life socialization influence diffuse support in later life? Thanks to the specific character of the European political system, one can expect that differences in early life political socialization among European citizens will lead to observable differences in the level of diffuse support for the EU in adulthood. European citizens vary greatly in the level of early life socialization into the political system of the EU. Over time, European integration has been progressing. More power has been moved to the supranational level at the expense of the power of the individual member states. As a result, younger generations of European citizens have grown up in a more closely integrated Europe than older generations. Furthermore, European integration has been expanding geographically as well. Therefore, some EU citizens joined the EU as adults, without experiencing the European political system as children. All this variation in the level of early life political socialization into the political system of the EU provides leverage in the study of the effects of early life political socialization on diffuse support for the EU in adulthood.
There are powerful reasons why early life socialization into the political system of the EU may matter for diffuse support in later life. As Down and Wilson (2013) argue, the type of political system that an individual experiences in early life sets the 'normal' for what a political system is supposed to look like. Therefore, if an individual experiences childhood in a political system with strong national power, such a political system constitutes her baseline experience.
6
For such an individual, moving to a political system that puts more power at the supranational level means entering unfamiliar territory. Such an individual may be more hesitant to express support for the supranational political system since putting more power at the supranational level differs greatly from the familiar picture learned in childhood. In contrast, an individual who grew up in a strongly integrated EU is socialized in a political system where substantial power at the supranational level is the status quo. For such an individual, political power at the supranational level is something familiar rather than something that challenges her ingrained view about what a political system looks like. As a result, it is expected that individuals who grew up in a more integrated Europe will have more diffuse support for the EU than individuals who experienced their childhood in a political system with little or no European integration.
In addition, in contrast to growing up in a political system without much European integration, growing up in a strongly integrated Europe is more likely to lead to lasting diffuse support for the EU. Unlike a weakly integrated Europe, a strongly integrated Europe is more noticeable to its citizens. Symbols of the EU, such as the flag, license plates, ID cards, and passports, are clearly visible and are an integral part of daily life. This more noticeable presence of European integration in daily life allows young individuals to accept the EU as something familiar and normal. Taking the EU as the status quo political system then allows young individuals to develop diffuse support for the political system that persists into adulthood. In contrast, individuals who were raised amidst weaker European integration did not have the same opportunity to develop lasting diffuse support in early life. Therefore, according to this theoretical view, the level of early life socialization into the political system of the EU will influence the level of diffuse support for the EU in adulthood.
The other view of the origins of diffuse support emphasizes later life experience as the most important source of diffuse support. This latter view does not completely deny the importance of early life political socialization but argues that later life political experience
7
outweighs the influence of early life political socialization (Mishler and Rose, 2007). Individuals learn about the political system throughout their adult life from the media and from their personal experiences. The accumulation of positive experience with the political system, as well as exposure to the symbols of the system, create an affective attachment to the political system (Easton, 1965; Mishler and Rose, 2007). In the case of the EU, cross-sectional public opinion studies show that the level of diffuse support is related to benefits flowing from EU membership, to the effects of economic globalization (Isernia et al., 2012), and to trust in national political institutions (Harteveld et al., 2013). Although this suggests that long-term exposure to these influences builds diffuse support over time, the cross-sectional research design is not well equipped for detecting the sources of diffuse support. The results of the cross-sectional studies are, therefore, only suggestive. Nevertheless, there is experimental evidence complementing the cross-sectional studies. An experimental study on European identity, which is one of the forms of diffuse support for the EU, shows that European identity tends to grow after long-term exposure to the symbols of the EU and positive news about the Union (Bruter, 2009).
Although existing studies present valuable insights into the origins of diffuse support for the EU, the question of the extent to which early life political socialization determines diffuse support in later life is not clearly answered. General comparative literature, as well as EU-specific literature, finds that early life political experience is one of the key determinants of diffuse support in later life (Down and Wilson, 2013; Easton, 1965; Lutz et al., 2006; Mishler and Rose, 2007). More recent research on diffuse support for the EU, however, suggests that early life political socialization has very little effect on diffuse support for the EU in later life (Ringlerova, 2015a). In addition, research focusing on the effects of later life socialization suggests that later life socialization has substantial influence, as well (Bruter, 2009; Mishler
8
and Rose, 2007). As highlighted by Down and Wilson (2017), the extent to which early life political socialization influences diffuse support in later life, therefore, remains a puzzle.
In this article, I resolve this puzzle by employing a novel approach to analyzing the relationship between early life political socialization and diffuse support for the EU. In contrast to the existing studies, my analysis allows for a much clearer examination of the effect of early life socialization. In studies of the effects of childhood socialization, it is generally very difficult to determine whether the observed effect results from early life socializing experience or from changing attitudes as individuals become older. By employing a design-based approach to Age-Period-Cohort (APC) analysis (Dinas and Stoker, 2014), the present study distinguishes the effect of childhood socializing experience from the effect of age much more effectively than prior studies.
In sum, the goal of this study is to determine whether early life political socialization matters for diffuse support in later life. Following the prevailing theoretical view in the literature (Down and Wilson, 2013; Lutz et al., 2006), I hypothesize that more childhood socialization into the European political system will lead to more diffuse support for the EU in adulthood: The more socialization influence an individual experiences of the European supranational political system in early life, the more diffuse support for the EU she will have in adulthood.
The next section presents the analysis testing this hypothesis. The analysis has two parts. First, I take a broad cross-sectional view and examine Eurobarometer data from 25 EU member states. Second, I narrow the analysis down to two countries: Denmark and Sweden. Although narrowing the analysis down to just two countries weakens the generalizability of the results in comparison to the broad cross-sectional analysis, it allows for stronger causal leverage in assessing the relationship between early life political socialization and diffuse support in later life.
9
Method, data, and variables
Studies that aim at identifying the effect of early life political socialization on later life attitudes face a methodological problem. The problem stems from the difficulty of distinguishing between the effect of childhood socialization and the effect of age. When researchers find that cohorts differ in their attitudes, it is unclear whether this difference is a result of different early life experience or whether attitudes between generations differ because individuals' attitudes change as people go through the life cycle. This difficulty in identifying the true source of between-generational variation is caused by the so-called APC problem (Yang and Land, 2013). The APC problem arises due to the fact that individuals who are in the same cohort are also of similar age. There is thus a perfect linear relationship between age, period, and cohort (where period is the time at which the measurement of attitudes is taken). If a researcher takes a measurement of attitudes at a certain point in time (period) and finds differences in attitudes between generations, it is difficult to know whether to attribute these differences to the effect of aging or to the between-generational differences in early life experience.
Existing studies of the relationship between early life socialization and diffuse support in later life use two common methods of dealing with the APC problem. The first method breaks up the linearity between age, period, and cohort by measuring one of these quantities in a different time unit. For example, cohort and period are measured in years while age is defined in five-year blocs (Lutz et al., 2006). When individuals from multiple birth years make up one cohort, there is no longer an exact linear relationship between age, period, and cohort. The statistical model is then able to distinguish the effects of all three variables. The second method uses a hierarchical setting to break up the APC linearity (Down and Wilson, 2013; Down and Wilson, 2017; Yang and Land, 2013). The hierarchical (multilevel) approach to APC (HAPC)
10
is used for repeated individual-level cross-sectional data. Like in the first approach, either age or cohort is defined in multi-year blocs. The HAPC model then views individual respondents as clustered both in period and in cohort. The hierarchical model does a better job accounting for the period and cohort clustering than the first type of approach. The HAPC model is, therefore, a superior solution to the APC problem.
Although the above-mentioned approaches to APC are helpful, they both share one serious shortcoming. This shortcoming stems from the inherent limitations of the data. Even these advanced methods cannot fully overcome the fact that the measurements of age and cohort in the data coincide. In other words, the fundamental problem here is a measurement problem: Although we are interested in the effect of socialization experience, we are using age to measure socialization experience. In contrast, this article does not measure early life socialization by focusing on individuals' birth year. Instead, it measures the level of early life socialization by focusing on early life socialization experience. In the case of EU citizens, the level of early life political socialization into the European political system does not fully overlap with age. Therefore, in contrast to the traditional approaches to APC, this study is able to parse out the effect of age from the effect of early life socialization into the European political system.
As recommended by Dinas and Stoker (2014), focusing on the socializing influence itself, rather than on the concept of a cohort, is a better way of overcoming the APC problem. In conventional APC studies, the variable cohort is only a proxy variable for the actual cause -the early life socialization experience. In contrast, in the approach recommended by Dinas and Stoker (2014), researchers do not use this proxy variable. They work directly with the concept of early life experience. This approach overcomes the APC problem by finding - within the same age group - individuals with different levels of exposure to the socializing influence. This
11
allows the analysis to effectively distinguish between the effect of early life socializing experience and the effect of age.
The EU provides a suitable case for employing this approach to APC analysis. The EU is appropriate because of its gradual geographical expansion. While in the early years of the integration process, only six countries participated in European integration, by 2013, there were 28 member states. As a result of this gradual stretching of the EU's borders, European citizens have diverse early life experiences with the EU. Some citizens have lived their entire lives within the European supranational political system; others became part of this political system as adults. Thanks to this diversity, the European citizenry encompasses individuals who are the same age but who have widely different early life experiences with the European supranational political system. Taking this approach to APC analysis allows me to study the relationship between early life socialization into the political system on the one hand, and diffuse support in later life on the other, while effectively distinguishing between the effect of childhood experience and the effect of age.
I apply this approach to data from 25 EU member states.1 The analysis uses data from the 2010-14 Eurobarometer surveys11 and examines to what extent different levels of early life socialization into the political system of the EU are related to diffuse support for the EU in adulthood. The time period following the Great Recession of 2008 is especially suitable for an analysis of diffuse support. As the theoretical section above explains, any empirical measurement of political support is a mix of two types of support: specific and diffuse. Even measurements that are commonly viewed as measurements of diffuse support may contain a performance-dependent (specific) component. In a time of economic downturn, however, when citizens' opinions of the political regime's performance are down, specific support decreases, while diffuse support remains stable. As a result, empirical measurements of diffuse support are less 'contaminated' by specific support than in times when specific support is likely to be
12
high owing to the favourable economic situation. Studying diffuse support during a time of economic crisis, therefore, allows us to obtain more precise measurements of diffuse support.
Diffuse support for the EU (the dependent variable) is defined as an affective attachment to the political system. The variable is measured in two ways in order to achieve greater robustness of the results. First, I use a binary variable indicating whether or not individuals have a European identity.111 Second, I use a binary variable indicating whether respondents do or do not see themselves as European citizens.
Each of the two measurements of diffuse support for the EU captures the concept from a slightly different angle. European identity clearly expresses the concept of diffuse support as an affective attachment. However, it is possible that some respondents understand this question as being about an attachment to Europe, not to the EU. If many respondents understand this survey question as being about Europe rather than about the EU, this measurement of diffuse support will have weak validity. In order to make the results of the study more robust, I employ the second measurement of the dependent variable: the measure of European citizenship. The measure of European citizenship captures a different, more civic, type of attachment to the EU than the measure of European identity. Unlike European identity, however, it is clearly tied to the EU. Both measures thus provide valid measurements of diffuse support for the EU.
The level of socializing influence of the European political system is the key explanatory concept. It is defined as the number of years of a respondent's childhood during which the respondent's country participated in European integration. European integration has been strengthening over time, though. The Maastricht Treaty responsible for the creation of the EU (in 1992) was a major step in the strengthening of European integration. I, therefore, distinguish between early life political socialization that took place before the EU was established and early life political socialization that took place after the EU was established. There are thus two key independent variables: (a) early life socialization in a weakly integrated
13
Europe {Socialization (weak)); and (b) early life socialization in a strongly integrated Europe {Socialization (strong)). Early life socialization in a weakly integrated Europe is defined as the number of years lived between the ages of five and 15 during which the respondent's country participated in European integration up until 1993. Early life socialization in a strongly integrated Europe is defined as the number of years lived between the ages of five and 15 during which the respondent's country was a member of the EU. Both variables range from zero to 10.
In addition to the main independent variables, the analysis controls for a number of control variables that, according to the existing literature, are related to support for the EU (Clark and Hellwig, 2012; Gabel, 1998; Harteveld et al., 2013; Lutz et al., 2006; Nelsen and Guth, 2000). These include education, egocentric and sociotropic evaluations of the national economy, confidence in national political institutions, occupation, age, and gender.lv In addition, I include a control variable for post-communist countries in order to control for another major country-level early life socialization influence that is not related to European integration. Religious denomination has also been identified in the existing literature as a source of diffuse support for the EU (Nelsen et al., 2001; Nelsen et al., 2011). Due to data limitations, however, a control variable for religious denomination cannot be included. Nevertheless, control variables that are included account for major sources of diffuse support and omitting religion from the analysis is not expected to seriously bias the results. More details on how the control variables are defined are provided in the Online appendix.
Analysis and results
Both dependent variables in this analysis are binary. I, therefore, rely on logistic regressions. The nature of the data poses a methodological challenge, however. The data are cross-sectional, and they are clustered in countries and in years of data collection/ It is
14
important to account for both types of clustering or the results of the statistical analysis would be biased (Rabe-Hesketh and Skrondal, 2012). I account for the country-level clustering by estimating two-level random intercept models. Country is the clustering variable in these models. In order to deal with the clustering in years of data collection, I run the model separately for each year. When considering the method of analysis, it is important to note that there is a disagreement among scholars as to whether logit models are the best tool in these cases or whether linear probability models are better. Although I believe that the random effects logit is the most appropriate tool for the present case, I address the methodological concerns raised in this debate by estimating a linear probability model as well. The alternative analysis is presented in the Online appendix. Using the alternative method of analysis does not change the substantive conclusions.
I model diffuse support for the EU as a function of early life socialization into the European political system and a number of control variables. First, the analysis focuses on the measure of diffuse support that captures individuals' feelings of European citizenship. Key results of this multilevel logistic regression are reported in the upper half of Table 1 (full results can be found in the Online appendix). Figures 1 and 2 interpret the results of the model, highlighting aspects relevant to the evaluation of the hypothesis. Figure 1 shows the predicted probability of sharing a sense of European citizenship by the level of early life socialization in a weakly integrated Europe. If the hypothesis were supported by the data, the probability would increase with the number of years of socialization in the European political system. This is not the case, however. The probability of feeling oneself a citizen of the EU slowly decreases as the number of years spent in childhood in a weakly integrated Europe increases. The decrease, however, is not statistically significant. The confidence intervals overlap to a large extent. Therefore, growing up in a weakly integrated Europe does not seem to affect diffuse support in later life.
15
[Figure 1 about here] [Figure 2 about here] [Table 1 about here]
Figure 2 shows the probability of feeling oneself a citizen of the EU, depending on the amount of early life socialization in a strongly integrated EU. In contrast to the expectations of the hypothesis, the probability of sharing a sense of EU citizenship decreases slightly as the amount of early life socialization in the EU increases.vl Nevertheless, the differences in predicted probabilities are not large enough to be statistically significant. Early life socialization in a strongly integrated EU does not influence the level of diffuse support for the EU in later life.
If we measure diffuse support for the EU as European identity, the results are very similar. As in the case of European citizenship, I estimated a series of multilevel logistic regressions. Key results are in the bottom half of Table 1, full results are in the Online appendix. I interpret the model by calculating predicted probabilities of having a European identity, depending on the amount of time of early life socialization spent in a weakly or strongly integrated Europe. As in the case of European citizenship, the results show a negative relationship between early life socialization in a weakly integrated Europe on the one hand, and diffuse support for the EU on the other. This relationship, however, is not statistically significant within the range of values of the independent variable. These results are similar for the variable measuring early life socialization in the strongly integrated EU. Although the negative relationship between socialization and diffuse support seems slightly stronger, it is, again, not statistically significant.
16
In sum, no matter how diffuse support is measured, there is not a systematic relationship between early life socialization in the European political system and diffuse support for the EU in adulthood. Although the relationship between childhood socialization and diffuse support is not statistically significant, it is puzzling that the relationship is negative. It may be the case that individuals who experienced more socialization have higher, even unrealistic, expectations about the EU, and as these expectations are confronted by the reality of the EU's functioning, the resulting disappointment decreases diffuse support for the EU. The following section checks the robustness of these findings by examining the hypothesis from a different angle.
Robustness checks
The previous section concluded that there is no systematic relationship between early life socialization in the European political system and diffuse support for the EU in later life. This finding is based on an analysis of a broad set of 25 EU member states. Although the breadth of the sample is an advantage, there are potential drawbacks as well. A cohort analysis of a broad set of countries may raise objections to the strength of the causal inference resulting from such analysis. It may be argued that individuals who grew up in the EU and individuals who grew up outside the EU differ in other socialization experiences, not just in their early life experience with the European political system.
In order to address this weakness, I narrow the analysis down to two EU member states. This allows me to apply the design-based approach to APC analysis (Dinas and Stoker, 2014). The approach requires the researcher to find two groups of individuals. The groups should be as similar as possible, but they should differ on the value of the key independent variable. Denmark and Sweden, the two countries selected for this analysis, fit these requirements well. The two countries differ in their EU-accession dates. While Denmark joined the EU in 1973, Sweden joined in 1995. There is thus a fairly large group of Danish and Swedish citizens who
17
are of the same generation but who differ in their early life experience with the European political system.
Importantly for the logic of the design-based approach, Denmark and Sweden are very similar on a number of characteristics that may potentially influence the development of diffuse political support (such as the economic situation in the 1973-95 time period, the type of welfare state, the party system, and the tradition of strong social democratic parties). Therefore, if the analysis finds that Danish citizens who grew up between 1973-95 have more diffuse support for the EU than Swedish citizens of the same age, I will be able to conclude with greater confidence that this difference in the level of diffuse support originates in their early life experience with the European political system. If, on the other hand, I find that the generations of Danish and Swedish citizens who grew up between 1973 and 1995 have a comparable level of diffuse support for the EU, I will be more confident in concluding that early life political socialization in the political system is not important for diffuse support in later life.
In this analysis, I use the same Eurobarometer data as in the previous section, but I only include Denmark and Sweden in the analysis. Both dependent variables and all control variablesvu are the same as in the preceding analysis. The cohort variable in this analysis, however, is defined differently. In accordance with the logic of the design-based approach to APC analysis, I divide the sample into twelve cohorts. These cohorts are defined by respondents' ages. The first cohort includes individuals who were 16 years or older when European integration started (individuals born in 1939 or earlier). The second-oldest generation and most of the following generations are defined in five-year blocs (born in 1940-1944, 1945-1949, etc.). Only the youngest generation comprises all individuals born in 1990 or later.
The goal of the analysis is to determine whether being subjected to the socializing influence of European integration in early life results in a higher level of diffuse support in later life. The different EU-accession dates of Denmark and Sweden will provide leverage in
18
finding the answer to this puzzle. Using the language of the design-based approach, Danish citizens who grew up between 1973 and 1995 are the 'treated' group because they were subjected to the socializing influence of European integration. Swedish citizens who experienced their childhood in the same time period are the control group. These Swedish citizens are of the same generation as their Danish counterparts, but they did not live within integrating Europe when they were growing up. If the main hypothesis is supported, we will expect that Danish citizens who experienced their childhood between 1973 and 1995 will have higher diffuse support for the EU than Swedish citizens who grew up during the same time period.
In order to establish whether the 'treated' group differs from the control group, I estimate a logistic regression. Diffuse support for the EU (measured as EU citizenship and European identity) is the dependent variable. Cohort is the key independent variable. The model further includes a dummy variable distinguishing Denmark from Sweden {country) and a set of control variables. In order to observe whether the effect of cohort differs between Denmark and Sweden, I include an interaction term interacting cohort and country. In order to account for clustering in years of data collection, I run a separate model for each year.
First, I will present the results of the logistic regression modelling European citizenship, then results from a logistic regression modelling European identity. Full results of the model are available in the Online appendix. Figures 3 and 4 interpret the results of these estimations. The figures show the differences in the predicted probability of diffuse support for the EU between the treated group and the control group for each of the twelve cohorts. If the hypothesis were supported, we would expect the difference to be significantly larger for the cohorts born approximately between 1965 and 1989 than for the other cohorts. This expectation, however, is largely not supported by the data. With the exception of the cohort born between 1985 and 1989, no cohorts display a significant difference between the 'treated' group and the control
19
group. Even the 1985-1989 cohort displays a significant difference in only the data for 2012 (European citizenship) and 2013 (European identity). This suggests that having grown up within integrating Europe does not lead to more diffuse support for the EU in adulthood.
[Figure 3 about here]
[Figure 4 about here]
Overall, this analysis focused on Sweden and Denmark does not support the main hypothesis.vm It thus provides additional evidence in favour of the conclusion drawn in the previous section: early life political socialization in the political system is not essential for creating a lasting affective attachment to the political system.
Conclusion
Early life political socialization into the political system is often considered important for the development of an individual's lasting affective attachment to a political system. Both in the general comparative literature and literature focused on the EU, early life political socialization is highlighted as an important influence on diffuse support for the political system. In contrast to this literature, the present analysis shows that the level of childhood socialization into the European political system makes no noticeable difference for the level of diffuse support for the EU in later life.
This conclusion is based on a series of empirical analyses of data from the 2010-2014 Eurobarometer surveys. In contrast to previous empirical work on the relationship between early life political socialization and diffuse support, this study is able to effectively distinguish the effect of early life political socialization from the effect of age. The analysis presented here
20
thus allows for a much clearer picture of the degree to which early life political socialization affects diffuse support in later life than previous studies have been able to provide.
Despite the innovative nature of the research presented in this study, the conclusions require a caveat. It may be argued that the measure of early life socialization into the political system of the EU is too crude and that there are great within-country differences in the extent to which individuals are socialized into the political system of the EU. For example, schools may differ in how much European integration is covered in their curriculum. Therefore, the conclusions of this study would have been stronger if a more precise measurement of early life political socialization into the EU were available. Unfortunately, this is not the case and the measurement used in this study constitutes the best one available. Nevertheless, even though the measurement of early life socializing influence of European integration is not completely perfect, it is a valid measurement that gauges the level of exposure to the socializing influence of European integration.
The findings presented in this manuscript have implications for the EU's strategies for its relationship with its citizenry. This article shows that those who grew up within integrating Europe do not have more diffuse support for the EU than those who grew up outside the EU. This implies that diffuse support for the political system of the EU develops as a result of experience that individuals gain in adulthood. Political socialization in early life is thus not enough to provide the political system of the EU with a lasting affective attachment of its citizens. The political system needs to provide positive experiences in order to maintain and reinforce this affective attachment throughout the individuals' lives. These results have practical implications for the EU as a political system. Diffuse support among the EU's citizens is one of the essential ingredients that keep the political system of the EU alive. If the EU wishes to maintain this source of stability, it is not enough to rely on early life political socialization. Generating positive experiences for adult citizens and taking active measures to
21
reinforce the affective attachment of European citizens to the EU is essential for maintaining citizens' diffuse support.
In addition, the findings of this study have implications for the way the integration of citizens in new EU member countries or the integration of migrants from outside the EU can affect public support for the EU. The results of this study suggest that the newness of EU citizens or residents does not necessarily constitute a weak spot in terms of public support for the political system of the EU. Given that early life political socialization is not essential for the development of an affective attachment to the EU, the new members of the population can learn to support their new political system as effectively as if they had grown up in it.
These findings suggest new pathways for future research. First, while the conclusions of this examination hold for the EU, without further research, it is unclear to what extent these conclusions apply to nation states. It is possible that, in contrast to nation states, the EU is not an important enough political object to imprint on children a lasting affective attachment. Future research may, therefore, explore opportunities for employing the design-based approach in the study of this question in the context of nation states. Second, there are still open questions about how exactly diffuse support develops as a result of later life socialization. What types of experience promote the development of this affective attachment to the political system? How does early life socialization interact with later-life learning? In addition, exploring whether there are heterogeneous effects of socialization across socio-economic groups and across countries is another question worth examining. Existing cross-sectional survey datasets, as well as some panel data or customized survey data collected in European countries, may provide fruitful ground for exploring these research questions. In a time of rising Euroscepticism as well as rising disenchantment with national governmental institutions across EU member countries, these are critical research questions that lead to the heart of the future of European democratic societies.
22
Acknowledgements
I thank Jay McCann and Andrew L. Roberts for giving detailed feedback on this manuscript. I also thank the participants at the European Consortium for Political Research General Conference panel, participants at the research workshops at the Faculty of Social Studies at Masaryk University (MUNI) and the Faculty of Law at MUNI for their helpful feedback on the earlier drafts of this manuscript. I thank the three anonymous reviewers for their careful, constructive, and truly helpful comments.
23
LIST OF REFERENCES
Beaudonnet L and Franklin MN. (2014) Is There Diffuse Support for the European Project?
The Eurozone Crisis as a Critical Test. Midwest Political Science Association 2014
Annual Meeting. Chicago. Bruter M. (2009) Time Bomb? The Dynamic Effect of News and Symbols on the Political
Identity of European Citizens. Comparative Political Studies 42: 1498-1536. Carter A and Teten RL. (2002) Assessing Changing Views of the President: Revisiting
Greenstein's Children and Politics. Presidential Studies Quarterly 32: 453-462. Clark N and Hellwig T. (2012) Information Effects and Mass Support for EU Policy Control.
European Union Politics 13: 535-557. Dalton RJ. (2004) Democratic Challenges, Democratic Choices: The Erosion of Political
Support in Advanced Industrial Democracies, New York: Oxford University Press. Dinas E and Stoker L. (2014) Age-Period-Cohort analysis: A design-based approach.
Electoral Studies 33: 28-40. Down I and Wilson CJ. (2013) A rising generation of Europeans? Life-cycle and cohort
effects on support for Europe'. European Journal of Political Research 52: 431-456. Down I and Wilson CJ. (2017) A Rising Generation of Europeans? Revisited. European
Journal of Political Research 56: 199-214. Easton D. (1965)^4 Systems Analysis of Political Life, New York: Wiley. Eichenberg RC and Dalton RJ. (2007) Post-maastricht blues: The transformation of citizen
support for European integration, 1973-2004. Acta Politica 42: 128-152. Gabel M. (1998) Public support for European integration: An empirical test of five theories.
Journal of Politics 60: 333-354. Gabel M and Palmer HD. (1995) Understanding Variation in Public Suport for European
Integration. European Journal of Political Research 27: 3-19. Harteveld E, van der Meer T and De Vries CE. (2013) In Europe we trust? Exploring three
logics of trust in the European Union. European Union Politics 14: 542-565. Hobolt SB and de Vries CE. (2016) Public Support for European Integration. Annual Review
of Political Science 19:413-432. Hooghe L and Marks G. (2004) Does identity or economic rationality drive public opinion on
European integration? Ps-Political Science & Politics 37: 415-420. Isernia P, Fiket I, Serricchio F, et al. (2012) But Still It Does Not Move: Functional and
Identity-Based Determinants of European Identity. In: Sanders D, Magalhaes PC and
Töka G (eds) Citizens and the European Polity Mass Attitudes Towards the European
and National Polities. Oxford: Oxford University Press. Lutz W, Kritzinger S and Skirbekk V. (2006) Population: The demography of growing
European identity. Science 314: 425-425. Mishler W and Rose R. (2007) Generation, age, and time: The dynamics of political learning
during Russia's transformation. American Journal of Political Science 51: 822-834. Nelsen BF and Guth JL. (2000) Exploring the Gender Gap: Women, Men, and Public
Attitudes toward European Integration. European Union Politics 1. Nelsen BF, Guth JL and Fräser CR. (2001) Does Religion Matter? Christianity and Public
Support for the European Union. European Union Politics 2. Nelsen BF, Guth JL and Highsmith B. (2011) Does Religion Still Matter? Religion and
Public Attitudes toward Integration in Europe. Politics and Religion 4: 1-26.
24
Norris P. (1999) Critical citizens: global support for democratic government, New York:
Oxford University Press. Norris P. (2011) Democratic deficit: critical citizens revisited, New York: Cambridge
University Press.
Rabe-Hesketh S and Skrondal A. (2012) MultilevelandLongitudinalModeling Using Stata,
College Station: Stata Press. Ringlerova Z. (2015a) Mass Public Support for the European Union: Diffuse or Specific in
Its Character? Department of Political Science. Purdue University. Ringlerova Z. (2015b) Weathering the Crisis: Evidence of Diffuse Support for the EU from a
Six-Wave Dutch Panel. European Union Politics 16: 558-576. Sapiro V. (2004) Not Your Parents' Political Socialization: Introduction for a New
Generation. Annual Review of Political Science 7: 1-23. Serricchio F, Tsakatika M and Quaglia L. (2013) Euroscepticism and the Global Financial
Crisis. Jems-Journal of Common Market Studies 51: 51-64. Torcal M, Bonet E and Costa Lobo M. (2012a) Institutional Trust and Responsiveness in the
EU. In: Sanders D, Bellucci P, Töka G, et al. (eds) The Europeanization of National
Polities? Citizenship and Support in a Post-Enlargement Union. New York: Oxford
University Press, 91-112. Torcal M, Mufioz J and Bonet E. (2012b) Trust in the European Parliament: From Affective
Heuristics to Rational Cueing. In: Sanders D, Magalhaes PC and Töka G (eds)
Citizens and the European Polity Mass Attitudes Towards the European and National
Polities. Oxford: Oxford University Press. Yang Y and Land KC. (2013) Age-Period-Cohort Analysis: New Models, Methods, and
Empirical Applications Boca Raton, FL: RCR Press.
25
Year: 2010
Number of early life years spent in a weak EU
Year: 2013
123456735
Number el early life years spent in a weak EU
Year: 2011
123456789 10
Number of early life years spent in a weak EU
Year 2014
illUliilli
1234        5        6        7        8       9 10
Number of early lile years spent in a weak EU
Year: 2012
1        23456        789 10
Number of early life years spent in a weak EU
Figure 1. Predicted probability of feeling oneself to be an EU citizen (by childhood socialization in a weakly integrated Europe).
26
Year: 2010
lltllllllll
Number ol early life years spent in a strong EU
Year: 2013
a     9 10
Number ol early life years spent in a strong EU
Year: 2011
mmnf H
1123456789 1C
Number ol early lile years spent in a strong EU
Year: 2014
HHHiiHl
)123456789 10
Number ol early lile years spent in a strong EU
Year: 2012
iHiUHHI
3123456789 10
Number ol early lite years spent in a strong EU
Figure 2. Predicted probability of feeling oneself to be an EU citizen (by childhood socialization in a strongly integrated Europe).
27
° 8-
o fi-O
° 8-
o fi-O
-—1     J      1    ^ —1-Jf"—I     J      I-j~ ^-^~—I-^   I   ^      I-T~—I-^1      I-t~
*    *    *    &   &   *    * ^^^^^^ ^ ^ .-^
1 I     J      I-—I   ^1      I     J      I-h——I-T-
j<>" jS-*"  j<>"   j<>" j<>"
° 8-
o r.
0 b
1
b
I    •■'»8{jj{il[l]    j    l-i i i i j { j ] I
° 8-
o r.
0 b
1
b
——1    t!    1 ^—1—3—I    J i—t~ n—t—i—^ I i—t—i—^1 i—T
j^*" j^*" jS-*" jS-*" ^ j^*" j^*" jS-*" ^ ^ jS? jS? jS? jS? ^ jS? .*?*
Figure 3. Differences in predicted probability of feeling oneself to be an EU citizen.
28
4-2         t   I   T   I   T   T T	
;H i M i 1 1 i	
4-	
11' t m i
^+4
Figure 4. Differences in predicted probability of having a European identity.
I Bulgaria and Romania are excluded from the analysis because they entered the EU only two years before the time period under examination. Croatia is excluded because it entered the EU during the time period examined here.
II Standard Eurobarometer surveys of the following numbers have been used: Eurobarometer 73.4, Eurobarometer 75.3, Eurobarometer 77.3, Eurobarometer 79.3, and Eurobarometer 81.2.
III The question wording for the first measurement is as follows: 'For each of the following statements, please tell me to what extent it corresponds or not to your own opinion. ... You feel you are a citizen of the EU. Yes, definitely/ Yes, to some extenl/ No, not really/ No, definitely not/ DK.' The original four-category ordinal variable was recoded into a binary variable in order to make interpretation more straightforward. 'Yes, definitely' and 'Yes, to some extent' were coded as 1, the rest as 0. The wording of the second measurement of the dependent variable is: 'In the near future, do you see yourself as...? (NATIONALITY) only/ (NATIONALITY) and European/ European and (NATIONALITY)/European only/ DK.' This variable was recoded such that the 'nationality only' category is coded 0, the rest is coded 1. In all variables the 'don't know' answers were coded as missing data.
lv Further information about the expectations for each of these control variables is available in the Online appendix.
v The feeling of citizenship variable is present in five Eurobarometer surveys from 2010 to 2014. Data from these five surveys are thus used for the analysis modelling the citizenship variable. The identity variable is present only in Eurobarometer surveys from the following years: 2010, 2013, and 2014. vl Predicted probabilities for all multilevel models were calculated by using the margins command in Stata 14 with the puO option. This option assumes that the random intercept equals zero. The resulting predicted probability is the population median probability (Rabe-Hesketh and Skrondal, 2012). ™ Only the control variable marking post-communist countries is not included because it is not relevant to the comparison of Denmark and Sweden.
vm In addition to this robustness check, further robustness checks have been performed. These included: relaxing the parallel period and age effects assumption in the model restricted to Sweden and Denmark, alternative specifications of the age interval that defines early life political socialization in the main model, using a logarithmized version of the key independent variable in the main model, using a linear probability model (instead of the random effects logit model), accounting for the possibility that socialization into the EU takes place up to three years prior the actual EU accession, as well as running the main model on Western European countries only. Overall, neither of these robustness analyses supports the hypothesis. Results of these robustness checks are available in the Online appendix.
29