One Size Fits All? Testing the Dimensional Structure of EU Attitudes in 21 Countries

One Size Fits All? Testing the Dimensional Structure of EU Attitudes in 21 Countries Abstract Citizens’ attitudes toward the European Union (EU) are important, changing, and multidimensional. Still comparative studies of the dimensional structure of EU-attitudes are virtually absent. Using an extensive battery of EU attitude-items in a 21-country study, we test the dimensional structure of EU attitudes cross-nationally and assess the variation in this dimensional structure. We find (1) that EU attitudes are indeed multidimensional, also comparatively, (2) that the structure varies, but (3) that the structure is widely applicable especially when the EU is more salient in a country. Surprisingly, the attitudinal structure is not more pronounced in long-standing member states, and the structure is most outspoken in countries experiencing a change in migration. The implications for the study of EU attitudes are discussed. What citizens think and feel about the European Union (EU) matters. It matters in and by itself by lending support to or questioning the legitimacy of particular EU decisions or the entire integration process. It also matters indirectly because EU attitudes are (increasingly) important for understanding citizens’ voting behavior in European Parliament elections (Hobolt & de Vries, 2016), in referendums on EU issues (Schuck & de Vreese, 2008), and even in national elections (de Vries & Hobolt, 2016). Understanding EU attitudes moreover has become more relevant in the light of the past decade’s developments in Europe, such as the euro or the refugee crises, the Brexit or the significant rise of anti-EU parties in many member states, with ramifications for the EU itself and the rest of the world. For too long, however, scholars have had to conceptualize EU attitudes on the basis of available survey data and questions. This de facto had led to a too narrow focus on and an overemphasis of EU public opinion research on questions about general support for a country’s EU membership or the perceived benefits from EU membership, questions regularly asked for instance by the biyearly Eurobarometer surveys (see Haverland, de Ruiter, & van de Walle, 2015 for an overview of variables and topics in the EB). While these two survey questions are undoubtedly important, valuable, and provide a rich, cross-nationally comparative and across time account of public opinion toward the EU, they at the same time limit our understanding of the possible multidimensionality of EU attitudes. Scholarship has emphasized that it is unlikely that EU attitudes are one-dimensional (Bakker & de Vreese, 2016; Hobolt & Brouard, 2010; Maier et al., 2015). The most inclusive and comprehensive account was provided by Boomgaarden, Schuck, Elenbaas, and de Vreese (2011) who proposed and empirically established five dimensions of EU attitudes. The dimensions relate to (1) EU effect, referring to feelings of fear of and threat by the EU, (2) a sense of European identity, (3) the performance and democratic functioning of the EU and its institutions, (4) utilitarian considerations, and (5) a strengthening of the EU. Boomgaarden et al.’s (2011) study was based on a survey in The Netherlands. However, while previous research has shown great variation in the level of EU support cross-nationally (Braun & Tausendpfund, 2014), little is known about variation in the dimensional structure of attitudes, though recent scholarship indicates the existence of different dimensions (Kuhn & Stoeckel, 2014). We therefore test the dimensional structure cross-nationally and investigate what factors systematically affect this structure. These two goals form the backbone of this study: first, we investigate whether the dimensional structure of EU attitudes established in prior research applies cross-nationally. We investigate this using original survey data from 21 countries. It is essential to know about this structure, not only to disentangle attitudes by looking at different levels of support and opposition but also different objects and antecedents of the attitude structure. Second, we also see this as foundational comparative research that can help us understand forward looking on which dimensions citizens across the EU shy away from or embrace the work of the EU. Some dimensions may be more stable than others, some may respond more quickly to external events or other explanatory factors, and some people may hold seemingly opposing attitudes toward the EU. To further understand such dynamics, we first need to establish more firmly the structure of EU attitudes. Theory EU attitudes are at the heart of political, societal, and scientific debates regarding the future of European integration. Today, it is acknowledged that integration efforts hinge on support from EU citizens—who appear increasingly skeptical about and disapproving of the EU (Hobolt, 2014). Unlike in the era of permissive consensus when citizens more or less passively followed suit with political elites in furthering European integration (Hooghe & Marks, 2009; Moravcsik, 1991), public support now is a necessary condition for European integration. However, public support seems to be dwindling, and the EU has become an important political cleavage in Europe today (Kriesi et al., 2008). This development has become remarkably present also in scholarly research following the economic crisis (Braun & Tausendpfund, 2014; Hobolt & Wratil, 2015; Kuhn & Stoeckel, 2014), again emphasizing the importance and relevance of thoroughly understanding what EU attitudes are all about. While, for instance, the euro crisis may have impacted on economic or performance-related aspects of public support for the EU, it may have done little to citizens’ sense of a European identity. But as research still relies on limited conceptualizations of EU attitudes, we often simply cannot address such questions. In social and political debates, EU attitudes are often discussed with reference to “Euroscepticism,” which has become a buzzword. Media, not only in Europe but across the globe, are reporting on the surge of this phenomenon in the wake of the euro and the refugee crises, or the Brexit. EU attitudes are also important in response to larger geopolitical transitions, which can result in more or less concerted EU actions. At the core of the Euroscepticism term is some degree of public aversion toward European integration (Lubbers & Scheepers, 2005, 2010). Understanding the nature and extent of this aversion makes attitudes and opinions toward the EU an important object of study in and by themselves (for example, Gabel, 1998), and they are also important to understand voting behavior in relation to European politics (Hobolt, 2009; Schuck & de Vreese, 2008), and increasingly so in national politics (de Vries, 2007). EU attitudes can concern different objects (Easton, 1975) and be diffuse or specific (Gabel, 1998). Previous research suggests that specific attitudes are more likely to fluctuate (see also Armingeon & Ceka, 2014), and that diffuse support is more stable (Easton, 1975). That said, even diffuse support can move in response to important events or new information. In previous studies of public opinion, the focus has been on general Euroscepticism (Hooghe & Marks, 2007; van Klingeren, Boomgaarden, & de Vreese, 2013), instrumental and political Euroscepticism (Lubbers & Scheepers, 2005, 2010), or EU enlargement support (Hobolt, 2014; Karp & Bowler, 2006). Building on these insights, Boomgaarden and colleagues (2011) investigated the configuration of EU attitudes in one country. Considering the multifaceted nature of the process of European integration and the theoretical notions mentioned above, they found that a more specific and fine-grained conceptualization is needed when studying public opinion toward the EU. The theoretical underpinnings of this multidimensional overview of EU attitudes are explicated in their original article. Specifically—and for the current purpose of sufficient level of detail—they identified the following five dimensions: (1) EU affect. The first dimension refers to feelings of fear of and threat by the EU, recognizing the increasing role played by emotions in explaining political evaluations. (2) European identity. The second dimension refers to identification with the EU, pride in being an EU citizen and feeling close to other Europeans and their culture and history, but also adherence to EU symbols such as the flag. (3) EU performance. The third dimension refers to the performance and democratic functioning of the EU and its institutions. It includes considerations about the transparency of decision-making and appropriate public expenditure. (4) Utilitarian evaluations. The fourth factor consists of the traditional general support measure, the country’s and personal benefit measures, and items that express a post-materialist utilitarian approach to European integration in terms of the EU helping to preserve peace, prosperity, and the environment. (5) Strengthening of the EU. Finally, the fifth factor relates to the future of European integration and to a process of further deepening and widening of the EU, such as support for policy transfer and extended decision-making competencies. The Multidimensional Nature In the original, inventory measures were used from foundational EU-related public opinion research, as well as a number of new areas of research. First, the widely used items from the Eurobarometer surveys on country membership (Anderson, 1998; Carey, 2002; Eichenberg & Dalton, 2009) and country benefit evaluation (Lubbers & Scheepers, 2005; McLaren, 2002) were included as traditional Eurobarometer items tapping the desired speed of integration (Hooghe & Marks, 2005; Sánchez-Cuenca, 2000) and support for policy transfer to the EU level (Dalton & Eichenberg, 1998; Gabel & Anderson, 2002; Lubbers & Scheepers, 2005). Furthermore, personal benefit perceptions of EU membership (Gabel & Palmer, 1995; Schuck & de Vreese, 2006) were included. Second, EU public opinion research was included that had explored the conceptual and empirical boundaries of EU attitudes by taking measures into consideration that potentially tap into a broader range of economic, political, and identity-related attitudes. This concerned items like “The European Union should become one country” (Lubbers, 2008) and “The decision-making power of the EU should be extended” (Hobolt & Brouard, 2010). It also included two identity-focused measures: “I am proud to be a European citizen” (Lubbers, 2008) and “The European Union poses a threat to Dutch identity and culture” (Hobolt & Brouard, 2010; Lubbers, 2008). Based on Lubbers’ (2008) study, it also used an item that measures the extent to which citizens perceive the EU to be “wasting tax money.” Finally, two items that relate to post-materialist utilitarianism were included: “The European Union fosters peace and stability” and “The European Union fosters the preservation of the environment” (Hobolt & Brouard, 2010). Third, an array of new items was developed. This included variations of items gauging general trust (“I trust the European Union”) and general support for enlargement (“The EU should be enlarged with other countries”). Given the importance of performance evaluations (see also a later study by de Vreese, Azrout, & Möller, 2016b), items pertaining to the functioning of the EU were included (“The decision-making process in the European Union is transparent,” “The European Union functions according to democratic principles,” and “The European Union functions well as it is”). Furthermore, several additional identity-related measures were included (“Being a citizen of the European Union means a lot to me,” “I feel close to fellow Europeans,” “The European flag means a lot to me,” and “Europeans share a common tradition, culture and history”), some of which were loosely derived from the work of Bruter (2003). Last but not least, the study included emotional responses to the EU, such as feeling “anger,” “fear,” and/or “disgust” toward the EU, reflecting a set of distinct negative discrete emotions (Huddy, Feldman, & Weber, 2007; Lerner & Keltner, 2001). Building on this framework, we included most of these items in a 21-country two-wave panel survey, all countries being members of the EU. The first aim of the study hence is to replicate, in a cross-national design, the dimensional structure proposed by Boomgaarden et al. (2011). By establishing whether this is the case, we will make an important inroad into future studies of EU attitudes.1 The second aim of this study draws on the comprehensive country sample that allows us to identify factors explaining the applicability of the dimensional EU attitude framework. Explaining Variation in Attitude Dimensional Structure To further unpack the second research aim, we review possible explanations for country-level variation in the fit of the attitude dimensional structure. It is important to stress that we here do not predict the degree of support or aversion for one or more of the dimensions. We indeed know that, for example, economic evaluations more strongly predict utilitarian attitudes toward the EU. Instead, we test whether there are country-level factors that explain how well (or not) the distinction between five attitudinal subdimensions fits in the different EU member states in our sample. The original study was conducted in The Netherlands, one of the original members of the Union, a euro country, and a country with a high trade dependency of the EU. It is also a country that has gone from being a prime example of “permissive consensus” to the EU being an issue of great political contestation (de Vreese et al., 2016b). These features (duration of membership; having the euro; being economically codependent on other EU countries) are subsequently elaborated to explain the country-level structure of EU attitudes. To study the attitude dimensionality across countries, we identified six explanatory factors. These factors are expected to relate to the dimensional structure, and they cover different aspects of what we here dub the “EU salience argument.” For the dimensional structure to become more pronounced—that is to show a better fit with clear distinction between the factors—citizens need to be given reason and opportunity to think about the EU and European integration. Accordingly, the factors we identify here all increase the opportunities for encounters with the EU and make the consequences of European integration more tangible. Hence, they all help to make the EU more salient in peoples’ mind. As a result, citizens are likely to have given their positions regarding the EU more thought, and thus increasingly differentiate between different attitude objects and more specific aspects of EU integration. We here stress that, given the virtual absence of systematic, empirical predecessors to draw from, some of these expectations are therefore tentative in nature. First, we expect that the longer the duration of an EU membership, the longer the time (i.e., scope of opportunity) that citizens have been given to think and develop attitudes about the EU. This should make the dimensional structure more pronounced. That said, other research (Hooghe & Marks, 2009) has shown that some original member countries were also consensus countries, which would possibly make a clear-cut multidimensional less clear-cut, which is why we tentatively formulate the “duration of membership” expectation. Second, as perhaps the most tangible consequence of EU integration, participation in the common currency is expected to be positively related to a more pronounced attitude structure. Again, the argument is based on salience with the currency being a daily reminder of the EU. Third, Anderson and Reichert (1995) remind us that economic benefits from EU membership, both at the country and the individual level, are related to general expressions of EU support. We therefore posit that financial ties to the EU (contributions to the EU and EU expenditure in the country) are likely to increase political and therefore public salience of the EU, thus making the dimensional structure more pronounced. Fourth, and related to the above, the strength of economic ties with the EU (EU trade dependency) increases the mutual dependency within the EU of other member states (see also Kuhn & Stoeckel, 2014). We also expect this trade dependency to contribute to salience and thus make the dimensional structure more fitting. A fifth factor relates to the presence of immigrants in the form of EU citizens from other countries. Immigration attitudes matter for EU attitudes (de Vreese & Boomgaarden, 2005; de Vreese, 2017; Kentmen-Cin & Erisen, 2017), so EU immigration arguably makes the EU more salient, and is thus expected to positively relate to the fit of the dimensional structure. Sixth, and finally, visibility of the EU in the media varies significantly (Schuck et al., 2011). A higher degree of media visibility is likely to increase the salience of the EU to citizens, thus making for a positive relationship between media visibility and the dimensional structure. In sum, these explanations have in common that they represent different ways in which the EU can be more or less salient, and the underlying assumption is that, ceteris paribus, the more salient the EU is in a given country, the more pronounced the EU attitudinal structure will be. Method We conducted a two-wave online panel survey in 21 EU member states, with the first wave taking place 3 weeks before the 2009 European Parliament Elections and the second wave directly after Election Day. Owing to financial and practical constraints (panel quality and the distribution of internet penetration in each country), we were unable to collect data in all EU member states, but our country sample ensured high cross-country variation on a number of aspects, taking into consideration the size of member states, countries from North, South, East, and West, and long-term and new members to the EU. The countries included were Austria, Belgium, Bulgaria, Czech Republic, Denmark, Finland, France, Germany, Greece, Hungary, Italy, Ireland, Latvia, Lithuania, The Netherlands, Poland, Portugal, Slovakia, Spain, Sweden, and the U.K. From the TNS databases (a research institute that complies with ESOMAR guidelines for survey research) and those of TNS partner institutes, national samples were drawn, with quotas enforced on age, gender, and education to ensure representativeness. A total of 32,412 respondents participated in the first wave. After deleting 16 respondents of whom we did not have complete data, the final sample size in the first wave was 32,396, which was an average response rate (AAPOR RR2) of 23% (with a minimum of 13% in Denmark and a maximum of 46% in Lithuania; see Supplementary AppendixTable A1 for country-specific details). All respondents were contacted to also participate in the second wave, of which 22,806 did. Owing to panel attrition, the second wave was on average slightly older and higher educated. However, as our analyses are based on correlations, and we do not model over time, variation is more important than representativeness. Also, in Wave 2, we do find sufficient variation (see also Supplementary AppendixTable A2). After removing incomplete data, the final sample size in Wave 2 was 22,795 (average recontact rate of 70%, with a minimum of 58% in Latvia and a maximum of 78% in Lithuania). The questionnaire was developed in English and translated by TNS (who also translates the Eurobarometer surveys) into different languages. As an additional check, all translated questionnaires were translated back into English. Irregularities and problems arising from this process were resolved by deliberation. Operationalization of EU Attitudes We asked our respondents to which degree they agreed or disagreed with a set of statements regarding the EU. The items stem from the dimensions identified by Boomgaarden et al. (2011). They are literal reproductions of the items. The overall framework for these was drawn from Easton (1975) with the distinction between different modes and objects of political support. The new items pertained to, for example, the functioning of the EU. Importantly, items tapping emotional responses to the EU were included, asking respondents to what extent they felt “anger,” “fear,” and/or “disgust” toward the EU, reflecting a set of related but conceptually distinct negative discrete emotions. We included multiple items for all five dimensions and reduced the number of items with eight for space reasons resulting in 17 items. All items were used in both waves of the survey. The item wordings, listed per dimension, are as shown in Table 1 (descriptive statistics per item can be found in Supplementary AppendixTable A2). Table 1 Measured Items per Dimension Performance  perf1:  How satisfied or dissatisfied are you with the way democracy works in the EU?  perf2:  The EU functions well as it is  perf3:  The EU functions according to democratic principles  perf4:  The decision-making process in the EU is transparent  Utilitarian  util1:  (Country’s) membership of the EU is a good thing  util2:  (Country) has on balance benefited from being a member of the EU  util3:  The EU fosters peace and stability  Identity  id1:  I am proud to be a European citizen  id2:  Being a citizen of the EU means a lot to me  id3:  The European flag means a lot to me  Affective  aff1:  I am angry about the EU  aff2:  I feel threatened by the EU  aff3:  I am disgusted with the EU  aff4:  I am afraid of the EU  Strengthening  str1:  The EU should become one country  str2:  In general, are you against or in favor of efforts being made to unify Europe?  str3:  What speed of building Europe would you like?  Performance  perf1:  How satisfied or dissatisfied are you with the way democracy works in the EU?  perf2:  The EU functions well as it is  perf3:  The EU functions according to democratic principles  perf4:  The decision-making process in the EU is transparent  Utilitarian  util1:  (Country’s) membership of the EU is a good thing  util2:  (Country) has on balance benefited from being a member of the EU  util3:  The EU fosters peace and stability  Identity  id1:  I am proud to be a European citizen  id2:  Being a citizen of the EU means a lot to me  id3:  The European flag means a lot to me  Affective  aff1:  I am angry about the EU  aff2:  I feel threatened by the EU  aff3:  I am disgusted with the EU  aff4:  I am afraid of the EU  Strengthening  str1:  The EU should become one country  str2:  In general, are you against or in favor of efforts being made to unify Europe?  str3:  What speed of building Europe would you like?  Table 1 Measured Items per Dimension Performance  perf1:  How satisfied or dissatisfied are you with the way democracy works in the EU?  perf2:  The EU functions well as it is  perf3:  The EU functions according to democratic principles  perf4:  The decision-making process in the EU is transparent  Utilitarian  util1:  (Country’s) membership of the EU is a good thing  util2:  (Country) has on balance benefited from being a member of the EU  util3:  The EU fosters peace and stability  Identity  id1:  I am proud to be a European citizen  id2:  Being a citizen of the EU means a lot to me  id3:  The European flag means a lot to me  Affective  aff1:  I am angry about the EU  aff2:  I feel threatened by the EU  aff3:  I am disgusted with the EU  aff4:  I am afraid of the EU  Strengthening  str1:  The EU should become one country  str2:  In general, are you against or in favor of efforts being made to unify Europe?  str3:  What speed of building Europe would you like?  Performance  perf1:  How satisfied or dissatisfied are you with the way democracy works in the EU?  perf2:  The EU functions well as it is  perf3:  The EU functions according to democratic principles  perf4:  The decision-making process in the EU is transparent  Utilitarian  util1:  (Country’s) membership of the EU is a good thing  util2:  (Country) has on balance benefited from being a member of the EU  util3:  The EU fosters peace and stability  Identity  id1:  I am proud to be a European citizen  id2:  Being a citizen of the EU means a lot to me  id3:  The European flag means a lot to me  Affective  aff1:  I am angry about the EU  aff2:  I feel threatened by the EU  aff3:  I am disgusted with the EU  aff4:  I am afraid of the EU  Strengthening  str1:  The EU should become one country  str2:  In general, are you against or in favor of efforts being made to unify Europe?  str3:  What speed of building Europe would you like?  Explanatory Factors for EU Attitude Structures Length EU Membership We operationalized the length of EU membership in different ways. First, we use year of entry, with Belgium, France, Germany, Italy, and The Netherlands as founding members since 1958 and Bulgaria joining in 2007 (see Supplementary AppendixTable A3 for the descriptives, also of all predictors mentioned below). The second operationalization is by looking at enlargement rounds, with the founding members in the first round and Bulgaria joining in the seventh round. Finally, we constructed a set of dichotomies, whether member states were part of the original EU6, the EU12, or the EU15. Eurozone Member The second (set of) predictor(s) is whether a member state is also member of the Eurozone. As Slovakia only became part of the Eurozone in 2009, a Euro effect may not have established yet. We thus constructed two versions of this variable: first with Slovakia a member of the Eurozone and second with Slovakia not a member of the Eurozone. EU Contributions and Expenditure As financial relations to the EU, we focus on countries’ contributions to the EU, EU expenditure in each country, and the net difference. The 2009 data were derived from Eurostat’s online database. We focus on the absolute numbers in billions of Euros, as well as per capita. Economic Ties/EU Trade Relations We operationalize EU trade relations by looking at the total amount of import to and export from other EU countries, as well as the total amount of trade (import plus export) with other EU countries. Again making use of Eurostat’s online database, we assess the absolute amount of trade and the share of a country’s international trade that is with other EU countries, and for both, we assess the change between 2008 and 2009. EU (and non-EU) Immigration Similar as EU contributions and expenditure, we retrieved immigration numbers from the Eurostat online database. We again focus on the absolute number of immigrants and the number relative to the general population. But also we focus on whether the number of immigrants increased or decreased between 2008 and 2009. Finally, to test whether it is immigration from EU countries or immigration from outside the EU that matters, we also distinguish between immigrants from other EU countries and immigrants from non-EU countries. EU Media Visibility To assess the visibility of the EU in the national media, we used the media data set of the 2009 European Election Study (details about the media study can be found in: Schuck, Xezonakis, Elenbaas, Banducci, & de Vreese, 2011). Three national newspapers and two national television news shows were analyzed in 3 weeks before the 2009 European Parliament Election. Including tabloid and quality newspapers as well as newscasts from public broadcasting and commercial stations, the outlets are a good sample of the news media in each country. Using this data set, we operationalize EU visibility in several ways. First, the absolute number of stories about the EU, with an EU story defined as mentioning the EU, its institutions, or the EP elections at least once. Second, the relative number of EU stories, defined as the proportion of EU stories in television news (as all news stories in television news were coded) and EU stories on the front page and a random page coded in each newspaper. Third, because of different formats (with newspapers more in-depth news but less audience, compared with less in-depth television news but a much larger reach), a distinction between newspaper visibility and television visibility. For both the absolute and relative attention to the EU were considered. Data Analysis We use structural equation modeling to perform a confirmatory factor analysis to test the dimensionality of EU attitudes. The basic measurement structure of the model consists of five latent factors (corresponding to the five dimensions of EU attitudes discussed above) and has the items load on these in line with the Boomgaarden et al. (2011) results. We run the model for all countries (as a single group), for each country separately, and as a multiple group SEM. We assess model fit using model chi-square (although given our large sample size, both overall and per country, a significant chi-square does not necessarily indicate inadequate fit, see Kenny, n.d.), TLI, RMSEA, and SRMR. Also, to achieve a final model, we test whether certain items have consistent (across countries) high residual correlations. We first will try to achieve a fitting model using data from Wave 1, and will consecutively test whether the pattern holds when applying to Wave 2.2 After estimating a cross-country model, we turn to explaining why in some countries the model fits better compared with other countries, using the TLI, RMSEA, and SRMR values from the country models. As we have only 21 cases, we can only test a limited number of predictors in one multiple regression model. So, we start by looking at bivariate correlations to filter out those predictors that are at least correlated with model fit. In the next step, we modeled pairs and trios of those predictors with significant correlations with model fit to discern patterns of true and spurious predictions. Given the small N, we do not predict the chi-square value, as this requires adding sample size to the model (as the chi-square value depends heavily on the sample size) for which we do not have the space. Results Factor Analysis We start with the CFA model (Table 2). When looking at the model with all respondents from the 21 countries together, we find a significant chi-square value. This is, however, not surprising given the large sample (Kenny, n.d.). The TLI indicates a good fit, but the RMSEA indicates a mediocre fit (MacCullum, Browne, & Sugaware, 1996). This is a pattern returning in the measurement models in most individual countries. Table 2 Fit Indices of the Five-Dimensional CFA Model Model  Country  Wave 1   Wave 2       N  χ2df= 67  TLI  RMSEA ( ε^) [90% CI]  SRMR  N  χ2df= 67  TLI  RMSEA ( ε^) [90% CI]  SRMR  Untrimmed model  All, one group  32,396  12,560.141a  .952  .059 [.059, .059]  .042  22,795  9,527.980a  .951  .061 [.061, .063]  .041  All, separate groups  32,396  17,983.106b  .937  .015 [.014, .015]  .044  22,795  14,541.821b  .935  .015 [.015, .016]  .046  Trimmed model  All, one group  32,396  4,652.970  .976  .046 [.045, .045]  .025  22,795  3,495.632  .976  .047 [.046, .046]  .024  All, separate groups  32,396  8,205.220c  .964  .012 [.012, .012]  .027  22,795  6,862.689c  .961  .013 [.013, .013]  .027  Trimmed model  Austria  1,427  310.422  .974  .050 [.045, .056]  .028  1,001  233.192  .976  .050 [.043, .057]  .026  Belgium  2,805  715.058  .961  .059 [.055, .063]  .034  2,000  563.600  .961  .061 [.056, .066]  .035  Bulgaria  1,509  265.409  .977  .044 [.039, .050]  .026  1,016  222.375  .972  .048 [.041, .055]  .029  Czech Republic  1,488  259.393  .975  .044 [.038, .050]  .026  1,013  272.073  .965  .055 [.048, .062]  .029  Denmark  1,421  302.091  .971  .050 [.044, .055]  .031  1,017  305.666  .963  .059 [.053, .066]  .033  Finland  1,440  426.350  .958  .061 [.056, .067]  .038  1,018  347.030  .957  .064 [.058, .071]  .039  France  1,417  389.908  .963  .058 [.053, .064]  .033  1,015  349.019  .956  .064 [.058, .071]  .034  Germany  1,408  369.161  .963  .057 [.051, .062]  .031  1,004  344.682  .958  .064 [.058, .071]  .033  Greece  1,419  407.138  .954  .060 [.054, .065]  .039  1,018  336.694  .967  .063 [.056, .070]  .042  Hungary  1,414  380.170  .961  .058 [.052, .063]  .036  1,019  318.651  .959  .061 [.054, .068]  .039  Ireland  1,417  359.453  .958  .056 [.050, .061]  .039  1,004  309.320  .955  .060 [.053, .067]  .043  Italy  1,420  505.805  .948  .068 [.062, .074]  .045  1,002  412.274  .947  .072 [.065, .078]  .047  Latvia  2,245  309.915  .980  .040 [.036, .045]  .024  1,303  255.961  .977  .047 [.041, .053]  .025  Lithuania  1,393  284.148  .973  .048 [.043, .054]  .031  1,093  239.642  .973  .049 [.042, .055]  .029  The Netherlands  1,408  463.433  .952  .065 [.059, .070]  .042  1,025  386.240  .950  .068 [.062, .075]  .041  Poland  1,409  317.063  .970  .051 [.046, .057]  .037  1,011  310.863  .963  .060 [.053, .067]  .036  Portugal  1,422  502.592  .936  .068 [.062, .073]  .050  1,012  427.685  .931  .073 [.066, .080]  .052  Slovakia  1,628  364.888  .969  .052 [.047, .058]  .032  1,189  252.261  .969  .048 [.042, .055]  .031  Spain  1,457  630.042  .926  .076 [.071, .081]  .050  1,020  449.145  .933  .075 [.068, .081]  .052  Sweden  1,428  327.400  .970  .052 [.047, .058]  .031  1,008  255.420  .973  .053 [.046, .060]  .027  The U.K.  1,421  315.305  .974  .051[.045, .057]  .027  1,007  270.827  .973  .055 [.048, .062]  .027  Model  Country  Wave 1   Wave 2       N  χ2df= 67  TLI  RMSEA ( ε^) [90% CI]  SRMR  N  χ2df= 67  TLI  RMSEA ( ε^) [90% CI]  SRMR  Untrimmed model  All, one group  32,396  12,560.141a  .952  .059 [.059, .059]  .042  22,795  9,527.980a  .951  .061 [.061, .063]  .041  All, separate groups  32,396  17,983.106b  .937  .015 [.014, .015]  .044  22,795  14,541.821b  .935  .015 [.015, .016]  .046  Trimmed model  All, one group  32,396  4,652.970  .976  .046 [.045, .045]  .025  22,795  3,495.632  .976  .047 [.046, .046]  .024  All, separate groups  32,396  8,205.220c  .964  .012 [.012, .012]  .027  22,795  6,862.689c  .961  .013 [.013, .013]  .027  Trimmed model  Austria  1,427  310.422  .974  .050 [.045, .056]  .028  1,001  233.192  .976  .050 [.043, .057]  .026  Belgium  2,805  715.058  .961  .059 [.055, .063]  .034  2,000  563.600  .961  .061 [.056, .066]  .035  Bulgaria  1,509  265.409  .977  .044 [.039, .050]  .026  1,016  222.375  .972  .048 [.041, .055]  .029  Czech Republic  1,488  259.393  .975  .044 [.038, .050]  .026  1,013  272.073  .965  .055 [.048, .062]  .029  Denmark  1,421  302.091  .971  .050 [.044, .055]  .031  1,017  305.666  .963  .059 [.053, .066]  .033  Finland  1,440  426.350  .958  .061 [.056, .067]  .038  1,018  347.030  .957  .064 [.058, .071]  .039  France  1,417  389.908  .963  .058 [.053, .064]  .033  1,015  349.019  .956  .064 [.058, .071]  .034  Germany  1,408  369.161  .963  .057 [.051, .062]  .031  1,004  344.682  .958  .064 [.058, .071]  .033  Greece  1,419  407.138  .954  .060 [.054, .065]  .039  1,018  336.694  .967  .063 [.056, .070]  .042  Hungary  1,414  380.170  .961  .058 [.052, .063]  .036  1,019  318.651  .959  .061 [.054, .068]  .039  Ireland  1,417  359.453  .958  .056 [.050, .061]  .039  1,004  309.320  .955  .060 [.053, .067]  .043  Italy  1,420  505.805  .948  .068 [.062, .074]  .045  1,002  412.274  .947  .072 [.065, .078]  .047  Latvia  2,245  309.915  .980  .040 [.036, .045]  .024  1,303  255.961  .977  .047 [.041, .053]  .025  Lithuania  1,393  284.148  .973  .048 [.043, .054]  .031  1,093  239.642  .973  .049 [.042, .055]  .029  The Netherlands  1,408  463.433  .952  .065 [.059, .070]  .042  1,025  386.240  .950  .068 [.062, .075]  .041  Poland  1,409  317.063  .970  .051 [.046, .057]  .037  1,011  310.863  .963  .060 [.053, .067]  .036  Portugal  1,422  502.592  .936  .068 [.062, .073]  .050  1,012  427.685  .931  .073 [.066, .080]  .052  Slovakia  1,628  364.888  .969  .052 [.047, .058]  .032  1,189  252.261  .969  .048 [.042, .055]  .031  Spain  1,457  630.042  .926  .076 [.071, .081]  .050  1,020  449.145  .933  .075 [.068, .081]  .052  Sweden  1,428  327.400  .970  .052 [.047, .058]  .031  1,008  255.420  .973  .053 [.046, .060]  .027  The U.K.  1,421  315.305  .974  .051[.045, .057]  .027  1,007  270.827  .973  .055 [.048, .062]  .027  Notes:adf = 109. bdf = 2,289. cdf = 1,407. Table 2 Fit Indices of the Five-Dimensional CFA Model Model  Country  Wave 1   Wave 2       N  χ2df= 67  TLI  RMSEA ( ε^) [90% CI]  SRMR  N  χ2df= 67  TLI  RMSEA ( ε^) [90% CI]  SRMR  Untrimmed model  All, one group  32,396  12,560.141a  .952  .059 [.059, .059]  .042  22,795  9,527.980a  .951  .061 [.061, .063]  .041  All, separate groups  32,396  17,983.106b  .937  .015 [.014, .015]  .044  22,795  14,541.821b  .935  .015 [.015, .016]  .046  Trimmed model  All, one group  32,396  4,652.970  .976  .046 [.045, .045]  .025  22,795  3,495.632  .976  .047 [.046, .046]  .024  All, separate groups  32,396  8,205.220c  .964  .012 [.012, .012]  .027  22,795  6,862.689c  .961  .013 [.013, .013]  .027  Trimmed model  Austria  1,427  310.422  .974  .050 [.045, .056]  .028  1,001  233.192  .976  .050 [.043, .057]  .026  Belgium  2,805  715.058  .961  .059 [.055, .063]  .034  2,000  563.600  .961  .061 [.056, .066]  .035  Bulgaria  1,509  265.409  .977  .044 [.039, .050]  .026  1,016  222.375  .972  .048 [.041, .055]  .029  Czech Republic  1,488  259.393  .975  .044 [.038, .050]  .026  1,013  272.073  .965  .055 [.048, .062]  .029  Denmark  1,421  302.091  .971  .050 [.044, .055]  .031  1,017  305.666  .963  .059 [.053, .066]  .033  Finland  1,440  426.350  .958  .061 [.056, .067]  .038  1,018  347.030  .957  .064 [.058, .071]  .039  France  1,417  389.908  .963  .058 [.053, .064]  .033  1,015  349.019  .956  .064 [.058, .071]  .034  Germany  1,408  369.161  .963  .057 [.051, .062]  .031  1,004  344.682  .958  .064 [.058, .071]  .033  Greece  1,419  407.138  .954  .060 [.054, .065]  .039  1,018  336.694  .967  .063 [.056, .070]  .042  Hungary  1,414  380.170  .961  .058 [.052, .063]  .036  1,019  318.651  .959  .061 [.054, .068]  .039  Ireland  1,417  359.453  .958  .056 [.050, .061]  .039  1,004  309.320  .955  .060 [.053, .067]  .043  Italy  1,420  505.805  .948  .068 [.062, .074]  .045  1,002  412.274  .947  .072 [.065, .078]  .047  Latvia  2,245  309.915  .980  .040 [.036, .045]  .024  1,303  255.961  .977  .047 [.041, .053]  .025  Lithuania  1,393  284.148  .973  .048 [.043, .054]  .031  1,093  239.642  .973  .049 [.042, .055]  .029  The Netherlands  1,408  463.433  .952  .065 [.059, .070]  .042  1,025  386.240  .950  .068 [.062, .075]  .041  Poland  1,409  317.063  .970  .051 [.046, .057]  .037  1,011  310.863  .963  .060 [.053, .067]  .036  Portugal  1,422  502.592  .936  .068 [.062, .073]  .050  1,012  427.685  .931  .073 [.066, .080]  .052  Slovakia  1,628  364.888  .969  .052 [.047, .058]  .032  1,189  252.261  .969  .048 [.042, .055]  .031  Spain  1,457  630.042  .926  .076 [.071, .081]  .050  1,020  449.145  .933  .075 [.068, .081]  .052  Sweden  1,428  327.400  .970  .052 [.047, .058]  .031  1,008  255.420  .973  .053 [.046, .060]  .027  The U.K.  1,421  315.305  .974  .051[.045, .057]  .027  1,007  270.827  .973  .055 [.048, .062]  .027  Model  Country  Wave 1   Wave 2       N  χ2df= 67  TLI  RMSEA ( ε^) [90% CI]  SRMR  N  χ2df= 67  TLI  RMSEA ( ε^) [90% CI]  SRMR  Untrimmed model  All, one group  32,396  12,560.141a  .952  .059 [.059, .059]  .042  22,795  9,527.980a  .951  .061 [.061, .063]  .041  All, separate groups  32,396  17,983.106b  .937  .015 [.014, .015]  .044  22,795  14,541.821b  .935  .015 [.015, .016]  .046  Trimmed model  All, one group  32,396  4,652.970  .976  .046 [.045, .045]  .025  22,795  3,495.632  .976  .047 [.046, .046]  .024  All, separate groups  32,396  8,205.220c  .964  .012 [.012, .012]  .027  22,795  6,862.689c  .961  .013 [.013, .013]  .027  Trimmed model  Austria  1,427  310.422  .974  .050 [.045, .056]  .028  1,001  233.192  .976  .050 [.043, .057]  .026  Belgium  2,805  715.058  .961  .059 [.055, .063]  .034  2,000  563.600  .961  .061 [.056, .066]  .035  Bulgaria  1,509  265.409  .977  .044 [.039, .050]  .026  1,016  222.375  .972  .048 [.041, .055]  .029  Czech Republic  1,488  259.393  .975  .044 [.038, .050]  .026  1,013  272.073  .965  .055 [.048, .062]  .029  Denmark  1,421  302.091  .971  .050 [.044, .055]  .031  1,017  305.666  .963  .059 [.053, .066]  .033  Finland  1,440  426.350  .958  .061 [.056, .067]  .038  1,018  347.030  .957  .064 [.058, .071]  .039  France  1,417  389.908  .963  .058 [.053, .064]  .033  1,015  349.019  .956  .064 [.058, .071]  .034  Germany  1,408  369.161  .963  .057 [.051, .062]  .031  1,004  344.682  .958  .064 [.058, .071]  .033  Greece  1,419  407.138  .954  .060 [.054, .065]  .039  1,018  336.694  .967  .063 [.056, .070]  .042  Hungary  1,414  380.170  .961  .058 [.052, .063]  .036  1,019  318.651  .959  .061 [.054, .068]  .039  Ireland  1,417  359.453  .958  .056 [.050, .061]  .039  1,004  309.320  .955  .060 [.053, .067]  .043  Italy  1,420  505.805  .948  .068 [.062, .074]  .045  1,002  412.274  .947  .072 [.065, .078]  .047  Latvia  2,245  309.915  .980  .040 [.036, .045]  .024  1,303  255.961  .977  .047 [.041, .053]  .025  Lithuania  1,393  284.148  .973  .048 [.043, .054]  .031  1,093  239.642  .973  .049 [.042, .055]  .029  The Netherlands  1,408  463.433  .952  .065 [.059, .070]  .042  1,025  386.240  .950  .068 [.062, .075]  .041  Poland  1,409  317.063  .970  .051 [.046, .057]  .037  1,011  310.863  .963  .060 [.053, .067]  .036  Portugal  1,422  502.592  .936  .068 [.062, .073]  .050  1,012  427.685  .931  .073 [.066, .080]  .052  Slovakia  1,628  364.888  .969  .052 [.047, .058]  .032  1,189  252.261  .969  .048 [.042, .055]  .031  Spain  1,457  630.042  .926  .076 [.071, .081]  .050  1,020  449.145  .933  .075 [.068, .081]  .052  Sweden  1,428  327.400  .970  .052 [.047, .058]  .031  1,008  255.420  .973  .053 [.046, .060]  .027  The U.K.  1,421  315.305  .974  .051[.045, .057]  .027  1,007  270.827  .973  .055 [.048, .062]  .027  Notes:adf = 109. bdf = 2,289. cdf = 1,407. Looking at the residual correlations between the indicators, we find that three indicators return in more than half of the countries as problematic (with high residual correlations with different indicators in different countries). This is the first strengthening indicator (“The EU should become one country”), the fourth performance indicator (“The decision making process in the EU is transparent”), and the fourth negative affection indicator (“I am afraid of the EU”). As we are searching for a measurement model that works in all countries (and because the high residual correlations with the problematic indicators are not with the same indicators in each country), adding error correlations would not be helpful. So, we opted to drop these indicators. Dropping these items leads to good model fit: TLI = .976, ε^ = .046 (see Table 2). Figure 1 shows the model when analyzing all respondents as one group. Low convergent validity may indicate the model has too few factors. But with all (but one) standardized factor loadings exceeding .70 (loading PERF1 = .66), it is indicated that the majority of the variance in the indicators is predicted by the factors. Per factor, we also assess the average variance extracted, which are well above the threshold of 50% for all factors in all countries (AVEperformance = 74%; AVEutilitarian = 82%; AVEidentity = 84%; AVEaffections = 81%; AVEstrengthening = 79%). Looking at the individual countries, we see the loading on PERF1 consistently below, but always close to .70. In one case, a strengthening item (STR3) is much lower in one country (Portugal). Although this is low, we observe no pattern of high residual correlations with other indicators with relative low loading, so we interpret this as not problematic for the number of factors. Testing the AVE per factor, we find that also in each country, all are above the threshold of 50%, indicating that the factors explain more than half of the variation in the items. We thus have good convergent validity and are confident we do not have a too little number of factors. Figure 1 View largeDownload slide Measurement model of five dimensions of EU attitudes (respondents of all 21 countries in one group). The entries at the double-headed arrows represent the correlations between the factors (before the slash Wave 1; after the slash Wave 2); the entries at the arrows toward the indicators represent standardized factor loadings. The lighter grey indicators have been dropped from the model because of reducing fit in more than half of the countries in the study Figure 1 View largeDownload slide Measurement model of five dimensions of EU attitudes (respondents of all 21 countries in one group). The entries at the double-headed arrows represent the correlations between the factors (before the slash Wave 1; after the slash Wave 2); the entries at the arrows toward the indicators represent standardized factor loadings. The lighter grey indicators have been dropped from the model because of reducing fit in more than half of the countries in the study In terms of discriminant validity, we find that all correlations between the factors are <.80, which is a rough indication that the factors are empirically distinguishable. We find the strongest correlation between the performance factor and the identity factor (r = .78), and test whether we can merge these factors (i.e., constrain their correlation to be 1). This leads to a substantial and significant loss of model fit (χ2df=1 = 8,077.14, p < .001). Testing whether the correlations between other pairs of factors can be constrained to 1 leads similarly to significant loss of model fit, implying that as one group we cannot merge the factors. Looking at the individual countries, we find that in five countries, eight correlations exceed .80. We also tested whether merging the factors was possible, but in each case, this significantly reduced model fit (indicating that the factors are empirically truly different).3 Applying the trimmed model (i.e., the model excluding the three problematic indicators) to Wave 2 leads to a model with good fit, both for the one-group model as for the multiple group model (see Table 2). We thus find that the five-dimensional model holds in all 21 countries in Wave 2, although in some countries the model certainly fits better than in others.4 We see a particularly well-fitting model for instance in Latvia, Bulgaria, and Austria, while the worst model fits appear in Spain, Portugal, Italy, and The Netherlands. In the next step, we seek to understand whether these differences are systematic patterns that can be explained by the country characteristics specified above. Predicting Model fit Turning to predicting model fit with contextual variables, we first look at the bivariate correlations (see Table 3). With regard to length of EU membership, we find a rather consistent pattern of longer members having worse fit. Higher values of chi-square and RMSEA ( ε^) indicate worse fit, so the negative correlations with year of entry in both Wave 1 (with χ2: r = −.52; with ε^: r = −.51) and Wave 2 (with χ2: r = −.40; with ε^: r = −.40) indicate that newer members have better fit. For TLI, higher values indicate better fit, and the positive coefficient (although not significant) points in the same direction. Focusing on the membership dichotomies of EU12 and EU15, all correlations (with the different indices and in both waves) are significant. But as being a member of EU12 or EU15 is coded as 1 (and not a EU12 or EU15 member 0), the positive correlation with the chi-square and the RMSEA and the negative correlation with the TLI indicate again that older members have worse fit. These findings run counter to our expectations formulated above; it appears that longer membership does not lead to more clearly distinct attitude dimensions. Table 3 Correlations Between Predictors and Model Fit Predictors [of model fit]  Wave 1   Wave 2     χ2  TLI  ε^  SRMR  χ2  TLI  ε^  SRMR  Year of entry  −.52*  .38+  −.51*  −.32  −.62**  .41+  −.58**  −.34  Enlargement round  −.50*  .37+  −.50*  −.32  −.60**  .39+  −.57**  −.34  Member EU6  .42+  −.21  .37+  .17  .50*  −.29  .43+  .16  Member EU12  .61**  −.58**  .62**  .50*  .71***  −.58**  .69***  .56**  Member EU15  .59**  −.51*  .61**  .43*  .61**  −.42+  .59**  .40+  Member Eurozone (including Slovakia)  .69***  −.64**  .69***  .58**  .60**  −.54*  .56**  .57**  Member Eurozone (excluding Slovakia)  .72***  −.68***  .73***  .61**  .72***  −.60**  .69***  .62**  Country’s contribution to EU  .32  −.22  .35  .11  .42+  −.28  .43+  .12  EU expenditure in country  .50*  −.43*  .49*  .37+  .58**  −.46*  .58**  .38+  Contribution–expenditure difference  .05  −.14  −.01  .25  −.01  −.08  −.04  .24  Country’s contribution to EU per capita  .06  −.10  −.02  .14  −.02  −.01  −.07  .19  EU expenditure in country per capita  .34  −.20  .33  .15  .37  −.18  .35  .12  Contribution–expenditure difference per capita  −.15  .03  −.19  .02  −.23  .09  −.23  .08  Absolute number EU citizens in country  .40+  −.33  .40+  .14  .42+  −.31  .41+  .16  Absolute number non-EU citizens in country  .39+  −.35  .41+  .18  .44*  −.34  .46*  .21  Relative number EU citizens in country  .33  −.25  .26  .17  .27  −.19  .21  .20  Relative number non-EU citizens in country  −.14  .01  −.11  −.14  −.08  .10  −.06  −.10  Increase in EU citizens in country  −.43+  .50*  −.42+  −.60**  −.42+  .51*  −.43+  −.53*  Increase in non-EU citizens in country  .06  −.02  .00  −.08  −.04  −.07  −.08  −.04  Absolute EU stories (all media)  .29  −.34  .28  .30  .26  −.13  .25  .34  Relative EU stories (all media)  −.08  −.02  −.04  .07  −.08  .17  −.05  .07  Absolute EU stories (in newspapers)  .05  −.06  .08  .07  .10  .09  .11  .10  Relative EU stories (in newspapers)  −.13  .09  −.10  −.13  −.14  .23  −.11  −.12  Absolute EU stories (in TV news)  .39+  −.46*  .36  .39+  .32  −.24  .29  .43+  Relative EU stories (in TV news)  .02  −.14  .06  .22  .03  .04  .06  .22  Absolute import from EU countries  .31  −.17  .31  .07  .41+  −.25  .39+  .06  Absolute export to EU countries  .32  −.17  .31  .04  .41+  −.24  .39+  .05  Absolute trade with EU countries  .32  −.17  .31  .06  .41+  −.25  .39+  .06  Proportion of total import from EU countries  −.26  .18  −.29  −.18  −.22  .05  −.23  −.16  Proportion of total export to EU countries  −.04  .03  −.10  .01  −.06  −.10  −.09  .00  Proportion trade with EU countries of total trade  −.16  .11  −.21  −.08  −.15  −.04  −.17  −.08  Absolute change in import from EU countries  −.28  .13  −.29  −.03  −.38  .20  −.36  −.02  Absolute change in export to EU countries  −.37  .20  −.35  −.08  −.44+  .26  −.42+  −.07  Absolute change in trade with EU countries  −.32  .16  −.32  −.05  −.41+  .23  −.39+  −.04  Change in proportion of import from EU countries  .39+  −.41+  .38+  .37+  .34  −.39+  .33  .37  Change in proportion of export to EU countries  −.24  .19  −.27  −.13  −.32  .13  −.33  −.14  Change in proportion trade with EU countries  .18  −.23  .15  .24  .09  −.22  .08  .24  Predictors [of model fit]  Wave 1   Wave 2     χ2  TLI  ε^  SRMR  χ2  TLI  ε^  SRMR  Year of entry  −.52*  .38+  −.51*  −.32  −.62**  .41+  −.58**  −.34  Enlargement round  −.50*  .37+  −.50*  −.32  −.60**  .39+  −.57**  −.34  Member EU6  .42+  −.21  .37+  .17  .50*  −.29  .43+  .16  Member EU12  .61**  −.58**  .62**  .50*  .71***  −.58**  .69***  .56**  Member EU15  .59**  −.51*  .61**  .43*  .61**  −.42+  .59**  .40+  Member Eurozone (including Slovakia)  .69***  −.64**  .69***  .58**  .60**  −.54*  .56**  .57**  Member Eurozone (excluding Slovakia)  .72***  −.68***  .73***  .61**  .72***  −.60**  .69***  .62**  Country’s contribution to EU  .32  −.22  .35  .11  .42+  −.28  .43+  .12  EU expenditure in country  .50*  −.43*  .49*  .37+  .58**  −.46*  .58**  .38+  Contribution–expenditure difference  .05  −.14  −.01  .25  −.01  −.08  −.04  .24  Country’s contribution to EU per capita  .06  −.10  −.02  .14  −.02  −.01  −.07  .19  EU expenditure in country per capita  .34  −.20  .33  .15  .37  −.18  .35  .12  Contribution–expenditure difference per capita  −.15  .03  −.19  .02  −.23  .09  −.23  .08  Absolute number EU citizens in country  .40+  −.33  .40+  .14  .42+  −.31  .41+  .16  Absolute number non-EU citizens in country  .39+  −.35  .41+  .18  .44*  −.34  .46*  .21  Relative number EU citizens in country  .33  −.25  .26  .17  .27  −.19  .21  .20  Relative number non-EU citizens in country  −.14  .01  −.11  −.14  −.08  .10  −.06  −.10  Increase in EU citizens in country  −.43+  .50*  −.42+  −.60**  −.42+  .51*  −.43+  −.53*  Increase in non-EU citizens in country  .06  −.02  .00  −.08  −.04  −.07  −.08  −.04  Absolute EU stories (all media)  .29  −.34  .28  .30  .26  −.13  .25  .34  Relative EU stories (all media)  −.08  −.02  −.04  .07  −.08  .17  −.05  .07  Absolute EU stories (in newspapers)  .05  −.06  .08  .07  .10  .09  .11  .10  Relative EU stories (in newspapers)  −.13  .09  −.10  −.13  −.14  .23  −.11  −.12  Absolute EU stories (in TV news)  .39+  −.46*  .36  .39+  .32  −.24  .29  .43+  Relative EU stories (in TV news)  .02  −.14  .06  .22  .03  .04  .06  .22  Absolute import from EU countries  .31  −.17  .31  .07  .41+  −.25  .39+  .06  Absolute export to EU countries  .32  −.17  .31  .04  .41+  −.24  .39+  .05  Absolute trade with EU countries  .32  −.17  .31  .06  .41+  −.25  .39+  .06  Proportion of total import from EU countries  −.26  .18  −.29  −.18  −.22  .05  −.23  −.16  Proportion of total export to EU countries  −.04  .03  −.10  .01  −.06  −.10  −.09  .00  Proportion trade with EU countries of total trade  −.16  .11  −.21  −.08  −.15  −.04  −.17  −.08  Absolute change in import from EU countries  −.28  .13  −.29  −.03  −.38  .20  −.36  −.02  Absolute change in export to EU countries  −.37  .20  −.35  −.08  −.44+  .26  −.42+  −.07  Absolute change in trade with EU countries  −.32  .16  −.32  −.05  −.41+  .23  −.39+  −.04  Change in proportion of import from EU countries  .39+  −.41+  .38+  .37+  .34  −.39+  .33  .37  Change in proportion of export to EU countries  −.24  .19  −.27  −.13  −.32  .13  −.33  −.14  Change in proportion trade with EU countries  .18  −.23  .15  .24  .09  −.22  .08  .24  Notes: Entries are Pearson correlations. For the correlations with chi-square, we control for sample size. *** p < .001, **p < .01, p < .05, +p < .1 (two-tailed). Table 3 Correlations Between Predictors and Model Fit Predictors [of model fit]  Wave 1   Wave 2     χ2  TLI  ε^  SRMR  χ2  TLI  ε^  SRMR  Year of entry  −.52*  .38+  −.51*  −.32  −.62**  .41+  −.58**  −.34  Enlargement round  −.50*  .37+  −.50*  −.32  −.60**  .39+  −.57**  −.34  Member EU6  .42+  −.21  .37+  .17  .50*  −.29  .43+  .16  Member EU12  .61**  −.58**  .62**  .50*  .71***  −.58**  .69***  .56**  Member EU15  .59**  −.51*  .61**  .43*  .61**  −.42+  .59**  .40+  Member Eurozone (including Slovakia)  .69***  −.64**  .69***  .58**  .60**  −.54*  .56**  .57**  Member Eurozone (excluding Slovakia)  .72***  −.68***  .73***  .61**  .72***  −.60**  .69***  .62**  Country’s contribution to EU  .32  −.22  .35  .11  .42+  −.28  .43+  .12  EU expenditure in country  .50*  −.43*  .49*  .37+  .58**  −.46*  .58**  .38+  Contribution–expenditure difference  .05  −.14  −.01  .25  −.01  −.08  −.04  .24  Country’s contribution to EU per capita  .06  −.10  −.02  .14  −.02  −.01  −.07  .19  EU expenditure in country per capita  .34  −.20  .33  .15  .37  −.18  .35  .12  Contribution–expenditure difference per capita  −.15  .03  −.19  .02  −.23  .09  −.23  .08  Absolute number EU citizens in country  .40+  −.33  .40+  .14  .42+  −.31  .41+  .16  Absolute number non-EU citizens in country  .39+  −.35  .41+  .18  .44*  −.34  .46*  .21  Relative number EU citizens in country  .33  −.25  .26  .17  .27  −.19  .21  .20  Relative number non-EU citizens in country  −.14  .01  −.11  −.14  −.08  .10  −.06  −.10  Increase in EU citizens in country  −.43+  .50*  −.42+  −.60**  −.42+  .51*  −.43+  −.53*  Increase in non-EU citizens in country  .06  −.02  .00  −.08  −.04  −.07  −.08  −.04  Absolute EU stories (all media)  .29  −.34  .28  .30  .26  −.13  .25  .34  Relative EU stories (all media)  −.08  −.02  −.04  .07  −.08  .17  −.05  .07  Absolute EU stories (in newspapers)  .05  −.06  .08  .07  .10  .09  .11  .10  Relative EU stories (in newspapers)  −.13  .09  −.10  −.13  −.14  .23  −.11  −.12  Absolute EU stories (in TV news)  .39+  −.46*  .36  .39+  .32  −.24  .29  .43+  Relative EU stories (in TV news)  .02  −.14  .06  .22  .03  .04  .06  .22  Absolute import from EU countries  .31  −.17  .31  .07  .41+  −.25  .39+  .06  Absolute export to EU countries  .32  −.17  .31  .04  .41+  −.24  .39+  .05  Absolute trade with EU countries  .32  −.17  .31  .06  .41+  −.25  .39+  .06  Proportion of total import from EU countries  −.26  .18  −.29  −.18  −.22  .05  −.23  −.16  Proportion of total export to EU countries  −.04  .03  −.10  .01  −.06  −.10  −.09  .00  Proportion trade with EU countries of total trade  −.16  .11  −.21  −.08  −.15  −.04  −.17  −.08  Absolute change in import from EU countries  −.28  .13  −.29  −.03  −.38  .20  −.36  −.02  Absolute change in export to EU countries  −.37  .20  −.35  −.08  −.44+  .26  −.42+  −.07  Absolute change in trade with EU countries  −.32  .16  −.32  −.05  −.41+  .23  −.39+  −.04  Change in proportion of import from EU countries  .39+  −.41+  .38+  .37+  .34  −.39+  .33  .37  Change in proportion of export to EU countries  −.24  .19  −.27  −.13  −.32  .13  −.33  −.14  Change in proportion trade with EU countries  .18  −.23  .15  .24  .09  −.22  .08  .24  Predictors [of model fit]  Wave 1   Wave 2     χ2  TLI  ε^  SRMR  χ2  TLI  ε^  SRMR  Year of entry  −.52*  .38+  −.51*  −.32  −.62**  .41+  −.58**  −.34  Enlargement round  −.50*  .37+  −.50*  −.32  −.60**  .39+  −.57**  −.34  Member EU6  .42+  −.21  .37+  .17  .50*  −.29  .43+  .16  Member EU12  .61**  −.58**  .62**  .50*  .71***  −.58**  .69***  .56**  Member EU15  .59**  −.51*  .61**  .43*  .61**  −.42+  .59**  .40+  Member Eurozone (including Slovakia)  .69***  −.64**  .69***  .58**  .60**  −.54*  .56**  .57**  Member Eurozone (excluding Slovakia)  .72***  −.68***  .73***  .61**  .72***  −.60**  .69***  .62**  Country’s contribution to EU  .32  −.22  .35  .11  .42+  −.28  .43+  .12  EU expenditure in country  .50*  −.43*  .49*  .37+  .58**  −.46*  .58**  .38+  Contribution–expenditure difference  .05  −.14  −.01  .25  −.01  −.08  −.04  .24  Country’s contribution to EU per capita  .06  −.10  −.02  .14  −.02  −.01  −.07  .19  EU expenditure in country per capita  .34  −.20  .33  .15  .37  −.18  .35  .12  Contribution–expenditure difference per capita  −.15  .03  −.19  .02  −.23  .09  −.23  .08  Absolute number EU citizens in country  .40+  −.33  .40+  .14  .42+  −.31  .41+  .16  Absolute number non-EU citizens in country  .39+  −.35  .41+  .18  .44*  −.34  .46*  .21  Relative number EU citizens in country  .33  −.25  .26  .17  .27  −.19  .21  .20  Relative number non-EU citizens in country  −.14  .01  −.11  −.14  −.08  .10  −.06  −.10  Increase in EU citizens in country  −.43+  .50*  −.42+  −.60**  −.42+  .51*  −.43+  −.53*  Increase in non-EU citizens in country  .06  −.02  .00  −.08  −.04  −.07  −.08  −.04  Absolute EU stories (all media)  .29  −.34  .28  .30  .26  −.13  .25  .34  Relative EU stories (all media)  −.08  −.02  −.04  .07  −.08  .17  −.05  .07  Absolute EU stories (in newspapers)  .05  −.06  .08  .07  .10  .09  .11  .10  Relative EU stories (in newspapers)  −.13  .09  −.10  −.13  −.14  .23  −.11  −.12  Absolute EU stories (in TV news)  .39+  −.46*  .36  .39+  .32  −.24  .29  .43+  Relative EU stories (in TV news)  .02  −.14  .06  .22  .03  .04  .06  .22  Absolute import from EU countries  .31  −.17  .31  .07  .41+  −.25  .39+  .06  Absolute export to EU countries  .32  −.17  .31  .04  .41+  −.24  .39+  .05  Absolute trade with EU countries  .32  −.17  .31  .06  .41+  −.25  .39+  .06  Proportion of total import from EU countries  −.26  .18  −.29  −.18  −.22  .05  −.23  −.16  Proportion of total export to EU countries  −.04  .03  −.10  .01  −.06  −.10  −.09  .00  Proportion trade with EU countries of total trade  −.16  .11  −.21  −.08  −.15  −.04  −.17  −.08  Absolute change in import from EU countries  −.28  .13  −.29  −.03  −.38  .20  −.36  −.02  Absolute change in export to EU countries  −.37  .20  −.35  −.08  −.44+  .26  −.42+  −.07  Absolute change in trade with EU countries  −.32  .16  −.32  −.05  −.41+  .23  −.39+  −.04  Change in proportion of import from EU countries  .39+  −.41+  .38+  .37+  .34  −.39+  .33  .37  Change in proportion of export to EU countries  −.24  .19  −.27  −.13  −.32  .13  −.33  −.14  Change in proportion trade with EU countries  .18  −.23  .15  .24  .09  −.22  .08  .24  Notes: Entries are Pearson correlations. For the correlations with chi-square, we control for sample size. *** p < .001, **p < .01, p < .05, +p < .1 (two-tailed). With regard to participating in the common currency, we again find a pattern that the predictor matters for the fit of the dimensional structure but not in the expected direction (see Table 3). For all fit indices and in both waves, we find significant correlations that imply that in countries that participate in the common currency show worse model fit compared with countries not part of the Eurozone. Eurozone membership does thus not lead to a crystallization of attitudes along the five dimensions. Turning to the countries’ financial relationships with the EU, we find that most predictors are not significantly correlated with the fit indices, implying the financial relations hardly matter. We do find a pattern of significant correlations across waves and fit indices for the contributions of the EU in each country, with the coefficients suggesting that the dimensional structure is more pronounced in countries in which the EU invests less. Also, this finding runs counter our provisional expectation in that more financial attachment to the EU does not translate into a clearer manifestation of the five attitude dimensions. Also, the economic ties between EU member states do not seem to matter much for the dimensional structure. Only countries that saw a growth of EU import show worse model fit, again not in line with what we had expected. Overall, it appears so far that the salience of the EU because of membership length or economic ties with the EU rather leads to the five attitude dimensions becoming less distinct and more interchangeable. With regard to the presence of EU (and non-EU) nationals from other countries, we find that the absolute number of EU citizens (and non-EU citizens) is marginally correlated with the chi-square and the RMSEA in Wave 1. In Wave 2, however, these correlations are not significant anymore. The change in the number of EU citizens in one’s country is significantly correlated with all fit indices and in both waves, implying that in countries in which the number of foreign EU nationals increases, we find a clearer representation of the five attitude dimensions. This finding is consistent with our expectation, and it is noteworthy here that we do not find a similar correlation for the growth of the number of non-EU nationals in each country. The last predictive factor that we take into account is media visibility of the EU. We only find a pattern of (marginally) significant correlations (in both waves and for all fit indices) with the absolute number of EU stories in the media, in which more EU visibility, and this higher attention to the EU, leads to a worse fitting model. Again, our expectation that salience leads to attitude crystallization along the five dimensions is not supported, rather to the contrary. These contextual indicators are, however, correlated, so that we may be witnessing spurious correlations with model fit here. Therefore, we aim to test these relationships with multiple regression. However, because of the small sample size (N = 21), we cannot simply add all seven significant correlates as predictors of model fit. We thus followed a strategy of testing pairs and trios of predictors. This showed a pattern of Eurozone membership and the increase in EU citizens as significant predictors (see Table 4), while other predictors in combination with these two turn insignificant (see Supplementary Appendix A6 for the models including the other predictors). Table 4 Multiple Regression Analysis Predicting Model Fit Wave  Predictors  TLI  RMSEA  SRMR  Wave 1  Member Eurozone  −0.62***(4.35)  0.68***(4.81)  0.56***(4.15)  Increase in EU citizens  0.48**(3.36)  −0.40*(2.84)  −0.58***(4.32)  R2  .63  .64  .67  Adjusted R2  .59  .60  .64  Wave 2  Member Eurozone  −0.53**(3.31)  0.55**(3.24)  0.55**(3.61)  Increase in EU citizens  0.50**(3.12)  −0.41*(2.42)  −0.51**(3.34)  R2  .54  .48  .58  Adjusted R2  .49  .43  .53  Wave  Predictors  TLI  RMSEA  SRMR  Wave 1  Member Eurozone  −0.62***(4.35)  0.68***(4.81)  0.56***(4.15)  Increase in EU citizens  0.48**(3.36)  −0.40*(2.84)  −0.58***(4.32)  R2  .63  .64  .67  Adjusted R2  .59  .60  .64  Wave 2  Member Eurozone  −0.53**(3.31)  0.55**(3.24)  0.55**(3.61)  Increase in EU citizens  0.50**(3.12)  −0.41*(2.42)  −0.51**(3.34)  R2  .54  .48  .58  Adjusted R2  .49  .43  .53  Note: Entries are standardized ordinary least squares regression coefficients, with absolute t-values in parentheses. N = 21. ***p < .001; **p < .01; *p < .05; +p < .1 (two-tailed). Table 4 Multiple Regression Analysis Predicting Model Fit Wave  Predictors  TLI  RMSEA  SRMR  Wave 1  Member Eurozone  −0.62***(4.35)  0.68***(4.81)  0.56***(4.15)  Increase in EU citizens  0.48**(3.36)  −0.40*(2.84)  −0.58***(4.32)  R2  .63  .64  .67  Adjusted R2  .59  .60  .64  Wave 2  Member Eurozone  −0.53**(3.31)  0.55**(3.24)  0.55**(3.61)  Increase in EU citizens  0.50**(3.12)  −0.41*(2.42)  −0.51**(3.34)  R2  .54  .48  .58  Adjusted R2  .49  .43  .53  Wave  Predictors  TLI  RMSEA  SRMR  Wave 1  Member Eurozone  −0.62***(4.35)  0.68***(4.81)  0.56***(4.15)  Increase in EU citizens  0.48**(3.36)  −0.40*(2.84)  −0.58***(4.32)  R2  .63  .64  .67  Adjusted R2  .59  .60  .64  Wave 2  Member Eurozone  −0.53**(3.31)  0.55**(3.24)  0.55**(3.61)  Increase in EU citizens  0.50**(3.12)  −0.41*(2.42)  −0.51**(3.34)  R2  .54  .48  .58  Adjusted R2  .49  .43  .53  Note: Entries are standardized ordinary least squares regression coefficients, with absolute t-values in parentheses. N = 21. ***p < .001; **p < .01; *p < .05; +p < .1 (two-tailed). For example, predicting the TLI value in Wave 1, we find significant coefficients for both membership of the Eurozone (b* = −0.62, t = 4.35, p < .001) and the increase of EI citizens in the country (b* = 0.48, t = 3.36, p = .004). The negative sign for Eurozone membership implies that members have a lower TLI value, meaning a worse fit and a less pronounced dimensional structure; the positive sign for EU citizens implies a better fit when the number of EU citizens increases in the country, and this pattern is consistent among waves and fit indices.5 Also, adding any of the other significant correlates with the fit indices to the model shows a pattern of keeping Eurozone membership and the increase in EU citizens’ significant predictors, while the added predictor is not significant (see Supplementary Appendix Table A6). Although there is an exception, with being a member of the EU12 also a significant predictor in one model (predicting RMSEA in Wave 2) and this predictor turning Eurozone membership coefficient marginally significant in two models (TLI and RMSEA in Wave 2), the general pattern remains that participating in the common currency reduces model fit, which is the opposite of what we expected, and that the growth of the number of EU nationals in each country predicts a better fitting model (i.e., a more pronounces dimensional structure), which is consistent with our expectation. Discussion Given the (increasing) political implications of EU attitudes and their importance in the public debate about European integration, this study set out to further understanding the structure of EU attitudes. Particularly, it focused on whether the multidimensional structure established in prior research would hold across different countries, and what might explain any cross-country variation in this structure. Doing so, we answered the call that “EU-wide surveys […] that incorporate the different dimensions of EU attitudes […] are needed, in order to test the cross-national validity of our findings and potentially to consider how contexts affect different attitude structures” (Boomgaarden et al. 2011, p. 261). Our results suggest that indeed, generally speaking, we can confirm the multidimensional structure of five attitude dimensions also cross-nationally. By and large, we find that across the 21 EU member states, people do consider different aspects of the EU and European integration when responding to attitude questions, and hence show variability in terms of their attitudes, distinguishing between emotional responses, utilitarian considerations, matters of identity, performance, and further strengthening of the EU. Our results, however, also suggest that the clarity with which EU attitudes fall into the different dimension does vary between countries. For some, such as Latvia, Bulgaria, and Austria, we find strong and clear evidence of a good fit of the five dimensions of EU attitudes. In other countries, such as Spain, Portugal, Italy, and The Netherlands, the five dimensions are still empirically distinguishable, but the indices suggest a less well-fitting structure. The original single-country study included mostly items that were not pertinent to The Netherlands specifically, and this cross-national study confirms that these measures can be used cross-nationally. Overall, this has important implications for research on public opinion toward the EU and European integration. Not only scholars but also public commentators need to acknowledge that opposition against the EU (or support for that matter) on certain aspects does not necessarily translate into opposition (or support) for other aspects. One can strongly identify with the EU and Europe, but at the same time oppose further steps toward a strengthening of the integration process. One can support the EU from a utilitarian perspective but still disapprove of the performance of its institutions. It will be important for future research and our understanding of public opinion toward the EU to further investigate which aspects and which dimensions of EU attitudes are more or less salient when making political decisions, both on the level of EU politics but also in national situations. While the Brexit referendum for instance is a prime example of the relevance of public opinion toward the EU, it would be important to understand what aspects of EU attitudes were more or less important in opposing membership. The same goes for the domestic ramifications of EU attitudes (Abbarno & Zapryonava, 2013; de Vries, 2007). National and EU politicians will be in a better position to campaign for or against Europe when better understanding the salience but also the possible contrasts of different EU attitudes dimensions. Similarly, some attitudes might be easier to move or quicker in responding to external events or crises than others. Further differentiation between EU attitudes is thus rather likely here to stay if we seek a fuller understanding of how Europe’s citizens view the EU. An important novelty of our study lies in the explanatory account of cross-national applicability of the established dimensionality of EU attitudes. Going beyond anecdotal evidence, our aim was to systematically explain why the structure fits more or less well in different member states. We found that in particular two factors matter for our understanding of country differences. First, and against our tentative expectation, we found that length of membership led to a less clear manifestation of the five dimensions of EU attitudes. One explanation is that rather than making the EU consistently more salient over time, it could be that EU membership wears off and provides fewer inclinations for citizens to differentiate clearly different aspects of the EU and the integration process. This would suggest a centrifugal process in which different EU attitudes do converge and yield an underlying, organizing principle. An alternative explanation is that EU attitudes in long-standing member countries become even more diverse, as citizens increase their awareness of the EU and its development. This could lead to even more nuanced attitudes than even the current five-dimensional structure was able to assess. Future research would be well advised to investigate how time affects and interacts with the nature of such attitudes. Second, migration mattered in that an increase in EU migrants in a given country was related to more crystalized EU attitudes along the five dimensions. This could be related to salience, in that more EU migrants make the free movement of people, and thus the integration process more salient to the citizens. Alternatively however, this could suggest that particular aspects of the integration process become salient and that would lead to differentiation of attitudes. It remains to be seen, however, whether this cross-national variation is stable (see also below). A clear limitation that we acknowledge is that our data were collected around the 2009 European election, and one could criticize that given the euro crisis came into full swing only afterward, and also the refugee crisis or the Brexit has hit the EU only recently, our data would be outdated. While it is reasonable to speculate that such severe events would affect the structure of EU attitudes, there is little empirical evidence to support such speculation. First, most of the EU-related crisis literature (Braun & Tausendpfund, 2014; Hobolt & Wratil, 2015; Kuhn & Stoeckel, 2014) suggested a drop in support but not necessarily a change in the dimensional structure. It would be interesting to see, however, whether this drop would relate more to certain dimensions than to others. Tentatively, one could assume that performance indicators would be particularly prone to such crises. Second, we have evidence from another replication of the item battery in The Netherlands in 2014 and 2016, suggesting that the five-dimensional structure still hold up (de Vreese, Azrout & Möller, 2016a, 2016b; Marquart, Goldberg, van Elsas, Brosius & de Vreese, 2017). This makes us confident that our cross-national data are still relevant. It remains of course to be seen whether the cross-country picture we draw above still holds. In particular, some countries that were more affected by recent events than others, such as Greece or the U.K., could show changes in the dimensionality. We also propose to further elaborate on the role of the media in creating contextual variance (see Hallin and Mancini (2017) for an updated reflection on comparing media systems). Such variance would not only be in the visibility of the EU itself, as modeled in this study, but also as platform for covering issues pertaining to the economy and immigration, thereby being an important source of information for citizens when forming EU attitudes (see also de Vreese & Boomgaarden, 2005). Such an expansion of the inquiry would also be wise to include attention to negative information and its persistency, even after possibly being corrected (Thorson, 2016). Finally, we believe it would be important to develop the multidimensional nature even more, with the EU engraining in more aspects of political and social life or in international affairs. But both trust (Easton, 1975; Harteveld et al., 2015) and the very core of EU attitudes might also experience a resurgence: it still really matters if citizens are “for or against membership of the EU.” Supplementary Data Supplementary Data are available at IJPOR online. Claes H. de Vreese is a Professor of Political Communication, Amsterdam School of Communication Research, ASCoR, University of Amsterdam. Rachid Azrout is a Lecturer in Political Communication at the Department of Communication Science, University of Amsterdam. Hajo G. Boomgaarden, Professor of Social Science Methods, Department of Communication, University of Vienna. Footnotes 1When performing multiple group CFA with SEM, the equivalence between the groups can bet tested at different levels. We test here the first/basic level, which tests whether between groups the same factor structure exists. At the next levels, the equivalence of factor loadings, correlations between factors and error disturbances can be evaluated. We stick to the first level, as this shows whether the same factors emerge from the same items. Would we test also the equivalence of factor loadings, we find that constraining the factor loadings across groups leads to a significant decrease of model fit (χ2df=280 = 3,681.720, p < .001 in Wave 1; χ2df=280 = 2,701.594, p < .001 in Wave 2), implying that factor loadings are indeed not the same across countries. 2At the face level, these dimensions work empirically also in a cross-national setting. Van Spanje and de Vreese (2014) included these five dimensions in a study of campaign effects on voting behavior in European Parliament election. However, they included these dimensions as control variables, as the study was mostly concerned with testing media effects during the campaign. Given this different focus, that study was therefore unable to unpack the structure and composition of the dimensions in more detail. 3We found high correlations (>.80) between the utilitarian dimension and strengthening dimension in France (r = .85), the U.K. (r = .85), and Sweden (r = .82). Merging the dimension in these countries significantly decreased model fit (France: χ2df = 1 = 183.841, p < .001; the U.K.: χ2df = 1 = 197.201, p <.001; Sweden: χ2df = 1 = 209.100, p <.001). With regard to the correlation between the performance dimension and the identity dimension, we found high correlations in the U.K. (r = .83), Sweden (r = .82), Bulgaria (r =.81), Poland (r = .81), and Denmark (r = .81). Again, merging the dimension significantly decreased model fit (the U.K.: χ2df = 1 = 338.532, p <.001; Sweden: χ2df = 1 = 243.755, p < .001; Bulgaria: χ2df = 1 = 271.114, p <.001; Poland: χ2df = 1 = 305.021, p <.001; Denmark: χ2df = 1 = 232.684, p <.001). So although we find a high correlation in these countries, we items do measure separate constructs. 4Based on the one-group factor solution, we constructed individual scores. The aim of this article, however, is not to compare the level of support on these different dimensions across time (Waves 1 and 2) or space (different countries) but merely to show the dimensional structure fits across time and space (and predict why it fits even better in some countries), and we show these the descriptive statistics only in Supplementary Appendix Tables A4 and A5). 5The flipping of the sign between TLI on the one hand and RMSEA and SRMR on the other is to be expected, as higher values of TLI indicate better fit, while for RMSEA and SRMR better fit is represented by lower values. References Abbarno A. J., Zapryanova G. M. ( 2013). Indirect effects of Eurosceptic messages on citizen attitudes toward domestic politics. JCMS: Journal of Common Market Studies , 51, 581– 597. doi:10.1111/jcms.12025 Google Scholar CrossRef Search ADS   Anderson C. J. ( 1998). When in doubt, use proxies: Attitudes toward domestic politics and support for European integration. Comparative Political Studies , 31, 569– 601. doi:10.1177/0010414098031005002 Google Scholar CrossRef Search ADS   Anderson C. J., Reichert S. ( 1995). Economic benefits and support for membership in the E.U.: A cross-national analysis. Journal of Public Policy , 15( 3), 231– 249. doi:10.1017/S0143814X00010035 Google Scholar CrossRef Search ADS   Armingeon K., Ceka B. ( 2014). The loss of trust in the European Union during the great recession since 2007: The role of heuristics from the national political system. European Union Politics , 15( 1), 82– 107. doi:10.1177/1465116513495595 Google Scholar CrossRef Search ADS   Bakker B. N., de Vreese C. H. ( 2016). Personality and European Union attitudes: Relationships across European Union attitude dimensions. European Union Politics , 17( 1), 25– 45. doi:10.1177/1465116515595885 Google Scholar CrossRef Search ADS   Boomgaarden H. G., Schuck A. R. T., Elenbaas M., de Vreese C. H. ( 2011). Mapping EU attitudes: Conceptual and empirical dimensions of Euroscepticism and EU support. European Union Politics , 12, 241– 266. doi:10.1177/1465116510395411 Google Scholar CrossRef Search ADS   Braun D., Tausendpfund M. ( 2014). The impact of the Euro crisis on citizens’ support for the European Union. Journal of European Integration , 36, 231– 245. doi:10.1080/07036337.2014.885751 Google Scholar CrossRef Search ADS   Bruter M. ( 2003). Winning hearts and minds for Europe: The impact of news and symbols on civic and cultural European identity. Comparative Political Studies , 36, 1148– 1179. doi:10.1177/0010414003257609 Google Scholar CrossRef Search ADS   Carey S. ( 2002). The dynamics of public opinion toward European integration, 1973-93. European Journal of International , 3( 4), http://journals.sagepub.com/doi/abs/10.1177/1465116502003004001 Dalton R. J., Eichenberg R. ( 1998). Citizen support for policy integration. In Sandholtz W., Sweet A. S. (Eds.), European integration and supranational governance  (pp. 250– 282). Oxford: Oxford University Press. doi:10.1093/0198294646.001.0001 Google Scholar CrossRef Search ADS   de Vreese C. H. ( 2017). How changing conditions make us reconsider the relationship between immigration attitudes, religion, and EU attitudes. European Union Politics , 18 ( 1), 137– 142. http://dx.doi.org/10.1177/1465116516680763 Google Scholar CrossRef Search ADS   de Vreese C. H., Azrout R., Möller J. ( 2016a). Cross road elections: Change in EU performance evaluations during the European Parliament Elections 2014. Politics and Governance , 4( 1), 69– 82. doi:10.17645/pag.v4i1.462 Google Scholar CrossRef Search ADS   de Vreese C. H., Azrout R., Möller J. ( 2016b). Netherlands: 2014 EP voting patterns: from Euphile to euroskeptic. In Hassing Nielsen J., Franklin M. (eds) The Eurosceptic 2014 European parliament elections  (pp 149– 169). London: Palgrave. de Vreese C. H., Boomgaarden H. G. ( 2005). Projecting EU referendums: Fear of immigration and support for European integration. European Union Politics , 6( 1), 59– 82. doi:10.1177/1465116505049608 Google Scholar CrossRef Search ADS   de Vries C. E. ( 2007). Sleeping giant: Fact or fairytale? How European integration affects national elections. European Union Politics , 8, 363– 385. doi:10.1177/1465116507079546 Google Scholar CrossRef Search ADS   de Vries C. E., Hobolt S. B. ( 2016). EU issue voting in national and European Parliamentary elections. In van der Brug W., de Vreese C. H. (Eds.), (Un)intended Consequences of EU Parliamentary Elections  (pp. 101– 124). Oxford: Oxford University Press. Google Scholar CrossRef Search ADS   Easton D. ( 1975). A re-assessment of the concept of political support. British Journal of Political Science , 5, 435– 457. doi:10.1017/S0007123400008309 Google Scholar CrossRef Search ADS   Eichenberg R. C., Dalton R. J. ( 2009). Europeans and the European community: The dynamics of public support for European integration. International Organization , 47, 507– 534. doi:10.1017/S0020818300028083 Google Scholar CrossRef Search ADS   Gabel M. ( 1998). Public support for European integration: An empirical test of five theories. The Journal of Politics , 60, 333– 354. Retrieved from http://www.jstor.org/stable/2647912 http://dx.doi.org/10.2307/2647912 Google Scholar CrossRef Search ADS   Gabel M., Anderson C. J. ( 2002). The structure of citizen attitudes and the European political space. Comparative Political Studies , 35, 893– 913. doi:10.1177/0010414002035008002 Google Scholar CrossRef Search ADS   Gabel M. J., Palmer H. D. ( 1995). Understanding variation in public support for European integration. European Journal of Political Research , 27( 1), 3– 19. doi:10.1111/j.1475-6765.1995.tb00627.x Google Scholar CrossRef Search ADS   Hallin D. C., Mancini P. ( 2017). Ten years after Comparing Media Systems: What have we learned? Political Communication , 34, 155– 171. Google Scholar CrossRef Search ADS   Harteveld E., van der Meer T., de Vries C. E. ( 2015). In Europe we trust? Exploring three logics of trust in the European Union. European Union Politics , 14, 542– 565. doi:10.1177/1465116513491018 Google Scholar CrossRef Search ADS   Haverland M., de Ruiter M., van de Walle S. ( 2015). Producing salience or keeping silence? An exploration of topics and non-topics of Special Eurobarometers. LEQS Paper No. 88. Retrieved from European Institute website: http://www.lse.ac.uk/europeanInstitute/LEQS%20Discussion%20Paper%20Series/LEQSPaper88.pdf Hobolt S. B. ( 2009). Europe in question: Referendums on European integration . Oxford: Oxford University Press. Google Scholar CrossRef Search ADS   Hobolt S. B. ( 2014). Ever closer or ever wider? Public attitudes towards further enlargement and integration in the European Union. Journal of European Public Policy , 21, 664– 680. doi:10.1080/13501763.2014.897746 Google Scholar CrossRef Search ADS   Hobolt S. B., Brouard S. ( 2010). Contesting the European Union? Why the Dutch and the French rejected the European constitution. Political Research Quarterly , 64, 309– 322. doi:10.1177/1065912909355713 Google Scholar CrossRef Search ADS   Hobolt S., de Vries C. ( 2016). Turning against the Union? The impact of the crisis on the eurosceptic vote in the 2014 European Parliament elections. Electoral Studies , 44, 504– 514. doi:10.1016/j.electstud.2016.05.006 Google Scholar CrossRef Search ADS   Hobolt S. B., Wratil C. ( 2015). Public opinion and the crisis: The dynamics of support for the euro. Journal of European Public Policy , 22, 238– 256. doi:10.1080/13501763.2014.994022 Google Scholar CrossRef Search ADS   Hooghe L., Marks G. ( 2005). Calculation, community and cues public opinion on European integration. European Union Politics , 6, 419– 443. doi:10.1177/1465116505057816 Google Scholar CrossRef Search ADS   Hooghe L., Marks G. ( 2007). Sources of euroscepticism. Acta Politica , 42, 119– 127. doi:10.1057/palgrave.ap.5500192 Google Scholar CrossRef Search ADS   Hooghe L., Marks G. ( 2009). A postfunctionalist theory of European integration: From permissive consensus to constraining dissensus. British Journal of Political Science , 39( 1), 1– 23. doi:10.1017/S0007123408000409 Google Scholar CrossRef Search ADS   Huddy L., Feldman S., Weber C. ( 2007). The political consequences of perceived threat and felt insecurity. The ANNALS of the American Academy of Political and Social Science , 614( 1), 131– 153. doi:10.1177/0002716207305951 Google Scholar CrossRef Search ADS   Karp J. A., Bowler S. ( 2006). Broadening and deepening or broadening versus deepening: The question of enlargement and Europe's ‘hesitant Europeans’. European Journal of Political Research , 45( 3), 369– 390. doi:10.1111/j.1475-6765.2006.00302.x Google Scholar CrossRef Search ADS   Kentmen-Cin C, Erisen C. ( 2017) Anti-immigration attitudes and the opposition to European integration: A critical assessment. European Union Politics , 18( 1), 3– 25. http://dx.doi.org/10.1177/1465116516680762 Google Scholar CrossRef Search ADS   Kriesi H., Grande E., Lachat R., Dolezal M., Bornschier S., Frey T. ( 2008). West European politics in the age of globalization  (pp. 154– 182). Cambridge: Cambridge University Press. Google Scholar CrossRef Search ADS   Kuhn T., Stoeckel F. ( 2014). When European integration becomes costly: The euro crisis and public support for European economic governance. Journal of European Public Policy , 21, 624– 641. doi:10.1080/13501763.2013.867892 Google Scholar CrossRef Search ADS   Lerner J. S., Keltner D. ( 2001). Fear, anger, and risk. Journal of Personality and Social Psychology , 81( 1), 146– 159. doi:10.1037/0022-3514.81.1.146 Google Scholar CrossRef Search ADS PubMed  Lubbers M. ( 2008). Regarding the Dutch ‘nee’ to the European constitution: A test of the identity, utilitarian and political approaches to voting ‘no’. European Union Politics , 9( 1), 59– 86. doi:10.1177/1465116507085957 Google Scholar CrossRef Search ADS   Lubbers M., Scheepers P. ( 2005). Political versus instrumental euro-scepticism: Mapping scepticism in European countries and regions. European Union Politics , 6, 223– 242. doi:10.1177/1465116505051984 Google Scholar CrossRef Search ADS   Lubbers M., Scheepers P. ( 2010). Divergent trends of euroscepticism in countries and regions of the European Union. European Journal of Political Research , 49, 787– 817. doi:10.1111/j.1475-6765.2010.01915.x Google Scholar CrossRef Search ADS   Maier M., Maier J., Baumert A., Jahn N., Krause S., Adam S. ( 2015). Measuring citizens’ implicit and explicit attitudes towards the European Union. European Union Politics , 16, 369– 385. doi:10.1177/1465116515577454 Google Scholar CrossRef Search ADS   McLaren L. M. ( 2002). Public support for the European Union: Cost/benefit analysis or perceived cultural threat? Journal of Politics , 64, 551– 566. doi:10.1111/1468-2508.00139 Google Scholar CrossRef Search ADS   Marquart F., Goldberg A., van Elsas E., Brosius A., de Vreese C.H. ( 2017). Knowing is not loving: media effects on knowledge about and attitudes toward the EU. Under review. Moravcsik A. ( 1991). Negotiating the Single European Act: national interests and conventional statecraft in the European Community. International organization , 45( 1), 19– 56. doi:10.1017/S0020818300001387 Google Scholar CrossRef Search ADS   Sánchez-Cuenca I. ( 2000). The political basis of support for European integration. European Union Politics , 1, 147– 171. doi:10.1177/1465116500001002001 Google Scholar CrossRef Search ADS   Schuck A. R. T., de Vreese C. H. ( 2006). Between risk and opportunity: News framing and its effects on public support for EU enlargement. European Journal of Communication , 21( 1), 5– 32. doi:10.1177/0267323106060987 Google Scholar CrossRef Search ADS   Schuck A. R. T., de Vreese C. H. ( 2008). The Dutch no to the EU constitution: Assessing the role of EU skepticism and the campaign. Journal of Elections, Public Opinion & Parties , 18( 1), 101– 128. doi:10.1080/17457280701858656 Google Scholar CrossRef Search ADS   Schuck A. R. T., Xezonakis G., Elenbaas M., Banducci S. A., de Vreese C. H. ( 2011). Party contestation and Europe on the news agenda: The 2009 European parliamentary elections. Electoral Studies , 30( 1), 41– 52. doi:10.1016/j.electstud.2010.09.021 Google Scholar CrossRef Search ADS   Thorson E. ( 2016). Belief echoes. The persistent effects of corrected misinformation. Political Communication , 33, 460– 480. doi: org/10.1080/10584609.2015.1102187 http://dx.doi.org/10.1080/10584609.2015.1102187 Google Scholar CrossRef Search ADS   van Klingeren M., Boomgaarden H. G., de Vreese C. H. ( 2013). Going soft or staying soft: Have identity factors become more important than economic rationale when explaining Euroscepticism? Journal of European Integration , 35, 689– 704. doi:10.1080/07036337.2012.719506 Google Scholar CrossRef Search ADS   © The Author(s) 2018. Published by Oxford University Press on behalf of The World Association for Public Opinion Research. All rights reserved. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png International Journal of Public Opinion Research Oxford University Press

One Size Fits All? Testing the Dimensional Structure of EU Attitudes in 21 Countries

Loading next page...
 
/lp/ou_press/one-size-fits-all-testing-the-dimensional-structure-of-eu-attitudes-in-SYvj09q10Y
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press on behalf of The World Association for Public Opinion Research. All rights reserved.
ISSN
0954-2892
eISSN
1471-6909
D.O.I.
10.1093/ijpor/edy003
Publisher site
See Article on Publisher Site

Abstract

Abstract Citizens’ attitudes toward the European Union (EU) are important, changing, and multidimensional. Still comparative studies of the dimensional structure of EU-attitudes are virtually absent. Using an extensive battery of EU attitude-items in a 21-country study, we test the dimensional structure of EU attitudes cross-nationally and assess the variation in this dimensional structure. We find (1) that EU attitudes are indeed multidimensional, also comparatively, (2) that the structure varies, but (3) that the structure is widely applicable especially when the EU is more salient in a country. Surprisingly, the attitudinal structure is not more pronounced in long-standing member states, and the structure is most outspoken in countries experiencing a change in migration. The implications for the study of EU attitudes are discussed. What citizens think and feel about the European Union (EU) matters. It matters in and by itself by lending support to or questioning the legitimacy of particular EU decisions or the entire integration process. It also matters indirectly because EU attitudes are (increasingly) important for understanding citizens’ voting behavior in European Parliament elections (Hobolt & de Vries, 2016), in referendums on EU issues (Schuck & de Vreese, 2008), and even in national elections (de Vries & Hobolt, 2016). Understanding EU attitudes moreover has become more relevant in the light of the past decade’s developments in Europe, such as the euro or the refugee crises, the Brexit or the significant rise of anti-EU parties in many member states, with ramifications for the EU itself and the rest of the world. For too long, however, scholars have had to conceptualize EU attitudes on the basis of available survey data and questions. This de facto had led to a too narrow focus on and an overemphasis of EU public opinion research on questions about general support for a country’s EU membership or the perceived benefits from EU membership, questions regularly asked for instance by the biyearly Eurobarometer surveys (see Haverland, de Ruiter, & van de Walle, 2015 for an overview of variables and topics in the EB). While these two survey questions are undoubtedly important, valuable, and provide a rich, cross-nationally comparative and across time account of public opinion toward the EU, they at the same time limit our understanding of the possible multidimensionality of EU attitudes. Scholarship has emphasized that it is unlikely that EU attitudes are one-dimensional (Bakker & de Vreese, 2016; Hobolt & Brouard, 2010; Maier et al., 2015). The most inclusive and comprehensive account was provided by Boomgaarden, Schuck, Elenbaas, and de Vreese (2011) who proposed and empirically established five dimensions of EU attitudes. The dimensions relate to (1) EU effect, referring to feelings of fear of and threat by the EU, (2) a sense of European identity, (3) the performance and democratic functioning of the EU and its institutions, (4) utilitarian considerations, and (5) a strengthening of the EU. Boomgaarden et al.’s (2011) study was based on a survey in The Netherlands. However, while previous research has shown great variation in the level of EU support cross-nationally (Braun & Tausendpfund, 2014), little is known about variation in the dimensional structure of attitudes, though recent scholarship indicates the existence of different dimensions (Kuhn & Stoeckel, 2014). We therefore test the dimensional structure cross-nationally and investigate what factors systematically affect this structure. These two goals form the backbone of this study: first, we investigate whether the dimensional structure of EU attitudes established in prior research applies cross-nationally. We investigate this using original survey data from 21 countries. It is essential to know about this structure, not only to disentangle attitudes by looking at different levels of support and opposition but also different objects and antecedents of the attitude structure. Second, we also see this as foundational comparative research that can help us understand forward looking on which dimensions citizens across the EU shy away from or embrace the work of the EU. Some dimensions may be more stable than others, some may respond more quickly to external events or other explanatory factors, and some people may hold seemingly opposing attitudes toward the EU. To further understand such dynamics, we first need to establish more firmly the structure of EU attitudes. Theory EU attitudes are at the heart of political, societal, and scientific debates regarding the future of European integration. Today, it is acknowledged that integration efforts hinge on support from EU citizens—who appear increasingly skeptical about and disapproving of the EU (Hobolt, 2014). Unlike in the era of permissive consensus when citizens more or less passively followed suit with political elites in furthering European integration (Hooghe & Marks, 2009; Moravcsik, 1991), public support now is a necessary condition for European integration. However, public support seems to be dwindling, and the EU has become an important political cleavage in Europe today (Kriesi et al., 2008). This development has become remarkably present also in scholarly research following the economic crisis (Braun & Tausendpfund, 2014; Hobolt & Wratil, 2015; Kuhn & Stoeckel, 2014), again emphasizing the importance and relevance of thoroughly understanding what EU attitudes are all about. While, for instance, the euro crisis may have impacted on economic or performance-related aspects of public support for the EU, it may have done little to citizens’ sense of a European identity. But as research still relies on limited conceptualizations of EU attitudes, we often simply cannot address such questions. In social and political debates, EU attitudes are often discussed with reference to “Euroscepticism,” which has become a buzzword. Media, not only in Europe but across the globe, are reporting on the surge of this phenomenon in the wake of the euro and the refugee crises, or the Brexit. EU attitudes are also important in response to larger geopolitical transitions, which can result in more or less concerted EU actions. At the core of the Euroscepticism term is some degree of public aversion toward European integration (Lubbers & Scheepers, 2005, 2010). Understanding the nature and extent of this aversion makes attitudes and opinions toward the EU an important object of study in and by themselves (for example, Gabel, 1998), and they are also important to understand voting behavior in relation to European politics (Hobolt, 2009; Schuck & de Vreese, 2008), and increasingly so in national politics (de Vries, 2007). EU attitudes can concern different objects (Easton, 1975) and be diffuse or specific (Gabel, 1998). Previous research suggests that specific attitudes are more likely to fluctuate (see also Armingeon & Ceka, 2014), and that diffuse support is more stable (Easton, 1975). That said, even diffuse support can move in response to important events or new information. In previous studies of public opinion, the focus has been on general Euroscepticism (Hooghe & Marks, 2007; van Klingeren, Boomgaarden, & de Vreese, 2013), instrumental and political Euroscepticism (Lubbers & Scheepers, 2005, 2010), or EU enlargement support (Hobolt, 2014; Karp & Bowler, 2006). Building on these insights, Boomgaarden and colleagues (2011) investigated the configuration of EU attitudes in one country. Considering the multifaceted nature of the process of European integration and the theoretical notions mentioned above, they found that a more specific and fine-grained conceptualization is needed when studying public opinion toward the EU. The theoretical underpinnings of this multidimensional overview of EU attitudes are explicated in their original article. Specifically—and for the current purpose of sufficient level of detail—they identified the following five dimensions: (1) EU affect. The first dimension refers to feelings of fear of and threat by the EU, recognizing the increasing role played by emotions in explaining political evaluations. (2) European identity. The second dimension refers to identification with the EU, pride in being an EU citizen and feeling close to other Europeans and their culture and history, but also adherence to EU symbols such as the flag. (3) EU performance. The third dimension refers to the performance and democratic functioning of the EU and its institutions. It includes considerations about the transparency of decision-making and appropriate public expenditure. (4) Utilitarian evaluations. The fourth factor consists of the traditional general support measure, the country’s and personal benefit measures, and items that express a post-materialist utilitarian approach to European integration in terms of the EU helping to preserve peace, prosperity, and the environment. (5) Strengthening of the EU. Finally, the fifth factor relates to the future of European integration and to a process of further deepening and widening of the EU, such as support for policy transfer and extended decision-making competencies. The Multidimensional Nature In the original, inventory measures were used from foundational EU-related public opinion research, as well as a number of new areas of research. First, the widely used items from the Eurobarometer surveys on country membership (Anderson, 1998; Carey, 2002; Eichenberg & Dalton, 2009) and country benefit evaluation (Lubbers & Scheepers, 2005; McLaren, 2002) were included as traditional Eurobarometer items tapping the desired speed of integration (Hooghe & Marks, 2005; Sánchez-Cuenca, 2000) and support for policy transfer to the EU level (Dalton & Eichenberg, 1998; Gabel & Anderson, 2002; Lubbers & Scheepers, 2005). Furthermore, personal benefit perceptions of EU membership (Gabel & Palmer, 1995; Schuck & de Vreese, 2006) were included. Second, EU public opinion research was included that had explored the conceptual and empirical boundaries of EU attitudes by taking measures into consideration that potentially tap into a broader range of economic, political, and identity-related attitudes. This concerned items like “The European Union should become one country” (Lubbers, 2008) and “The decision-making power of the EU should be extended” (Hobolt & Brouard, 2010). It also included two identity-focused measures: “I am proud to be a European citizen” (Lubbers, 2008) and “The European Union poses a threat to Dutch identity and culture” (Hobolt & Brouard, 2010; Lubbers, 2008). Based on Lubbers’ (2008) study, it also used an item that measures the extent to which citizens perceive the EU to be “wasting tax money.” Finally, two items that relate to post-materialist utilitarianism were included: “The European Union fosters peace and stability” and “The European Union fosters the preservation of the environment” (Hobolt & Brouard, 2010). Third, an array of new items was developed. This included variations of items gauging general trust (“I trust the European Union”) and general support for enlargement (“The EU should be enlarged with other countries”). Given the importance of performance evaluations (see also a later study by de Vreese, Azrout, & Möller, 2016b), items pertaining to the functioning of the EU were included (“The decision-making process in the European Union is transparent,” “The European Union functions according to democratic principles,” and “The European Union functions well as it is”). Furthermore, several additional identity-related measures were included (“Being a citizen of the European Union means a lot to me,” “I feel close to fellow Europeans,” “The European flag means a lot to me,” and “Europeans share a common tradition, culture and history”), some of which were loosely derived from the work of Bruter (2003). Last but not least, the study included emotional responses to the EU, such as feeling “anger,” “fear,” and/or “disgust” toward the EU, reflecting a set of distinct negative discrete emotions (Huddy, Feldman, & Weber, 2007; Lerner & Keltner, 2001). Building on this framework, we included most of these items in a 21-country two-wave panel survey, all countries being members of the EU. The first aim of the study hence is to replicate, in a cross-national design, the dimensional structure proposed by Boomgaarden et al. (2011). By establishing whether this is the case, we will make an important inroad into future studies of EU attitudes.1 The second aim of this study draws on the comprehensive country sample that allows us to identify factors explaining the applicability of the dimensional EU attitude framework. Explaining Variation in Attitude Dimensional Structure To further unpack the second research aim, we review possible explanations for country-level variation in the fit of the attitude dimensional structure. It is important to stress that we here do not predict the degree of support or aversion for one or more of the dimensions. We indeed know that, for example, economic evaluations more strongly predict utilitarian attitudes toward the EU. Instead, we test whether there are country-level factors that explain how well (or not) the distinction between five attitudinal subdimensions fits in the different EU member states in our sample. The original study was conducted in The Netherlands, one of the original members of the Union, a euro country, and a country with a high trade dependency of the EU. It is also a country that has gone from being a prime example of “permissive consensus” to the EU being an issue of great political contestation (de Vreese et al., 2016b). These features (duration of membership; having the euro; being economically codependent on other EU countries) are subsequently elaborated to explain the country-level structure of EU attitudes. To study the attitude dimensionality across countries, we identified six explanatory factors. These factors are expected to relate to the dimensional structure, and they cover different aspects of what we here dub the “EU salience argument.” For the dimensional structure to become more pronounced—that is to show a better fit with clear distinction between the factors—citizens need to be given reason and opportunity to think about the EU and European integration. Accordingly, the factors we identify here all increase the opportunities for encounters with the EU and make the consequences of European integration more tangible. Hence, they all help to make the EU more salient in peoples’ mind. As a result, citizens are likely to have given their positions regarding the EU more thought, and thus increasingly differentiate between different attitude objects and more specific aspects of EU integration. We here stress that, given the virtual absence of systematic, empirical predecessors to draw from, some of these expectations are therefore tentative in nature. First, we expect that the longer the duration of an EU membership, the longer the time (i.e., scope of opportunity) that citizens have been given to think and develop attitudes about the EU. This should make the dimensional structure more pronounced. That said, other research (Hooghe & Marks, 2009) has shown that some original member countries were also consensus countries, which would possibly make a clear-cut multidimensional less clear-cut, which is why we tentatively formulate the “duration of membership” expectation. Second, as perhaps the most tangible consequence of EU integration, participation in the common currency is expected to be positively related to a more pronounced attitude structure. Again, the argument is based on salience with the currency being a daily reminder of the EU. Third, Anderson and Reichert (1995) remind us that economic benefits from EU membership, both at the country and the individual level, are related to general expressions of EU support. We therefore posit that financial ties to the EU (contributions to the EU and EU expenditure in the country) are likely to increase political and therefore public salience of the EU, thus making the dimensional structure more pronounced. Fourth, and related to the above, the strength of economic ties with the EU (EU trade dependency) increases the mutual dependency within the EU of other member states (see also Kuhn & Stoeckel, 2014). We also expect this trade dependency to contribute to salience and thus make the dimensional structure more fitting. A fifth factor relates to the presence of immigrants in the form of EU citizens from other countries. Immigration attitudes matter for EU attitudes (de Vreese & Boomgaarden, 2005; de Vreese, 2017; Kentmen-Cin & Erisen, 2017), so EU immigration arguably makes the EU more salient, and is thus expected to positively relate to the fit of the dimensional structure. Sixth, and finally, visibility of the EU in the media varies significantly (Schuck et al., 2011). A higher degree of media visibility is likely to increase the salience of the EU to citizens, thus making for a positive relationship between media visibility and the dimensional structure. In sum, these explanations have in common that they represent different ways in which the EU can be more or less salient, and the underlying assumption is that, ceteris paribus, the more salient the EU is in a given country, the more pronounced the EU attitudinal structure will be. Method We conducted a two-wave online panel survey in 21 EU member states, with the first wave taking place 3 weeks before the 2009 European Parliament Elections and the second wave directly after Election Day. Owing to financial and practical constraints (panel quality and the distribution of internet penetration in each country), we were unable to collect data in all EU member states, but our country sample ensured high cross-country variation on a number of aspects, taking into consideration the size of member states, countries from North, South, East, and West, and long-term and new members to the EU. The countries included were Austria, Belgium, Bulgaria, Czech Republic, Denmark, Finland, France, Germany, Greece, Hungary, Italy, Ireland, Latvia, Lithuania, The Netherlands, Poland, Portugal, Slovakia, Spain, Sweden, and the U.K. From the TNS databases (a research institute that complies with ESOMAR guidelines for survey research) and those of TNS partner institutes, national samples were drawn, with quotas enforced on age, gender, and education to ensure representativeness. A total of 32,412 respondents participated in the first wave. After deleting 16 respondents of whom we did not have complete data, the final sample size in the first wave was 32,396, which was an average response rate (AAPOR RR2) of 23% (with a minimum of 13% in Denmark and a maximum of 46% in Lithuania; see Supplementary AppendixTable A1 for country-specific details). All respondents were contacted to also participate in the second wave, of which 22,806 did. Owing to panel attrition, the second wave was on average slightly older and higher educated. However, as our analyses are based on correlations, and we do not model over time, variation is more important than representativeness. Also, in Wave 2, we do find sufficient variation (see also Supplementary AppendixTable A2). After removing incomplete data, the final sample size in Wave 2 was 22,795 (average recontact rate of 70%, with a minimum of 58% in Latvia and a maximum of 78% in Lithuania). The questionnaire was developed in English and translated by TNS (who also translates the Eurobarometer surveys) into different languages. As an additional check, all translated questionnaires were translated back into English. Irregularities and problems arising from this process were resolved by deliberation. Operationalization of EU Attitudes We asked our respondents to which degree they agreed or disagreed with a set of statements regarding the EU. The items stem from the dimensions identified by Boomgaarden et al. (2011). They are literal reproductions of the items. The overall framework for these was drawn from Easton (1975) with the distinction between different modes and objects of political support. The new items pertained to, for example, the functioning of the EU. Importantly, items tapping emotional responses to the EU were included, asking respondents to what extent they felt “anger,” “fear,” and/or “disgust” toward the EU, reflecting a set of related but conceptually distinct negative discrete emotions. We included multiple items for all five dimensions and reduced the number of items with eight for space reasons resulting in 17 items. All items were used in both waves of the survey. The item wordings, listed per dimension, are as shown in Table 1 (descriptive statistics per item can be found in Supplementary AppendixTable A2). Table 1 Measured Items per Dimension Performance  perf1:  How satisfied or dissatisfied are you with the way democracy works in the EU?  perf2:  The EU functions well as it is  perf3:  The EU functions according to democratic principles  perf4:  The decision-making process in the EU is transparent  Utilitarian  util1:  (Country’s) membership of the EU is a good thing  util2:  (Country) has on balance benefited from being a member of the EU  util3:  The EU fosters peace and stability  Identity  id1:  I am proud to be a European citizen  id2:  Being a citizen of the EU means a lot to me  id3:  The European flag means a lot to me  Affective  aff1:  I am angry about the EU  aff2:  I feel threatened by the EU  aff3:  I am disgusted with the EU  aff4:  I am afraid of the EU  Strengthening  str1:  The EU should become one country  str2:  In general, are you against or in favor of efforts being made to unify Europe?  str3:  What speed of building Europe would you like?  Performance  perf1:  How satisfied or dissatisfied are you with the way democracy works in the EU?  perf2:  The EU functions well as it is  perf3:  The EU functions according to democratic principles  perf4:  The decision-making process in the EU is transparent  Utilitarian  util1:  (Country’s) membership of the EU is a good thing  util2:  (Country) has on balance benefited from being a member of the EU  util3:  The EU fosters peace and stability  Identity  id1:  I am proud to be a European citizen  id2:  Being a citizen of the EU means a lot to me  id3:  The European flag means a lot to me  Affective  aff1:  I am angry about the EU  aff2:  I feel threatened by the EU  aff3:  I am disgusted with the EU  aff4:  I am afraid of the EU  Strengthening  str1:  The EU should become one country  str2:  In general, are you against or in favor of efforts being made to unify Europe?  str3:  What speed of building Europe would you like?  Table 1 Measured Items per Dimension Performance  perf1:  How satisfied or dissatisfied are you with the way democracy works in the EU?  perf2:  The EU functions well as it is  perf3:  The EU functions according to democratic principles  perf4:  The decision-making process in the EU is transparent  Utilitarian  util1:  (Country’s) membership of the EU is a good thing  util2:  (Country) has on balance benefited from being a member of the EU  util3:  The EU fosters peace and stability  Identity  id1:  I am proud to be a European citizen  id2:  Being a citizen of the EU means a lot to me  id3:  The European flag means a lot to me  Affective  aff1:  I am angry about the EU  aff2:  I feel threatened by the EU  aff3:  I am disgusted with the EU  aff4:  I am afraid of the EU  Strengthening  str1:  The EU should become one country  str2:  In general, are you against or in favor of efforts being made to unify Europe?  str3:  What speed of building Europe would you like?  Performance  perf1:  How satisfied or dissatisfied are you with the way democracy works in the EU?  perf2:  The EU functions well as it is  perf3:  The EU functions according to democratic principles  perf4:  The decision-making process in the EU is transparent  Utilitarian  util1:  (Country’s) membership of the EU is a good thing  util2:  (Country) has on balance benefited from being a member of the EU  util3:  The EU fosters peace and stability  Identity  id1:  I am proud to be a European citizen  id2:  Being a citizen of the EU means a lot to me  id3:  The European flag means a lot to me  Affective  aff1:  I am angry about the EU  aff2:  I feel threatened by the EU  aff3:  I am disgusted with the EU  aff4:  I am afraid of the EU  Strengthening  str1:  The EU should become one country  str2:  In general, are you against or in favor of efforts being made to unify Europe?  str3:  What speed of building Europe would you like?  Explanatory Factors for EU Attitude Structures Length EU Membership We operationalized the length of EU membership in different ways. First, we use year of entry, with Belgium, France, Germany, Italy, and The Netherlands as founding members since 1958 and Bulgaria joining in 2007 (see Supplementary AppendixTable A3 for the descriptives, also of all predictors mentioned below). The second operationalization is by looking at enlargement rounds, with the founding members in the first round and Bulgaria joining in the seventh round. Finally, we constructed a set of dichotomies, whether member states were part of the original EU6, the EU12, or the EU15. Eurozone Member The second (set of) predictor(s) is whether a member state is also member of the Eurozone. As Slovakia only became part of the Eurozone in 2009, a Euro effect may not have established yet. We thus constructed two versions of this variable: first with Slovakia a member of the Eurozone and second with Slovakia not a member of the Eurozone. EU Contributions and Expenditure As financial relations to the EU, we focus on countries’ contributions to the EU, EU expenditure in each country, and the net difference. The 2009 data were derived from Eurostat’s online database. We focus on the absolute numbers in billions of Euros, as well as per capita. Economic Ties/EU Trade Relations We operationalize EU trade relations by looking at the total amount of import to and export from other EU countries, as well as the total amount of trade (import plus export) with other EU countries. Again making use of Eurostat’s online database, we assess the absolute amount of trade and the share of a country’s international trade that is with other EU countries, and for both, we assess the change between 2008 and 2009. EU (and non-EU) Immigration Similar as EU contributions and expenditure, we retrieved immigration numbers from the Eurostat online database. We again focus on the absolute number of immigrants and the number relative to the general population. But also we focus on whether the number of immigrants increased or decreased between 2008 and 2009. Finally, to test whether it is immigration from EU countries or immigration from outside the EU that matters, we also distinguish between immigrants from other EU countries and immigrants from non-EU countries. EU Media Visibility To assess the visibility of the EU in the national media, we used the media data set of the 2009 European Election Study (details about the media study can be found in: Schuck, Xezonakis, Elenbaas, Banducci, & de Vreese, 2011). Three national newspapers and two national television news shows were analyzed in 3 weeks before the 2009 European Parliament Election. Including tabloid and quality newspapers as well as newscasts from public broadcasting and commercial stations, the outlets are a good sample of the news media in each country. Using this data set, we operationalize EU visibility in several ways. First, the absolute number of stories about the EU, with an EU story defined as mentioning the EU, its institutions, or the EP elections at least once. Second, the relative number of EU stories, defined as the proportion of EU stories in television news (as all news stories in television news were coded) and EU stories on the front page and a random page coded in each newspaper. Third, because of different formats (with newspapers more in-depth news but less audience, compared with less in-depth television news but a much larger reach), a distinction between newspaper visibility and television visibility. For both the absolute and relative attention to the EU were considered. Data Analysis We use structural equation modeling to perform a confirmatory factor analysis to test the dimensionality of EU attitudes. The basic measurement structure of the model consists of five latent factors (corresponding to the five dimensions of EU attitudes discussed above) and has the items load on these in line with the Boomgaarden et al. (2011) results. We run the model for all countries (as a single group), for each country separately, and as a multiple group SEM. We assess model fit using model chi-square (although given our large sample size, both overall and per country, a significant chi-square does not necessarily indicate inadequate fit, see Kenny, n.d.), TLI, RMSEA, and SRMR. Also, to achieve a final model, we test whether certain items have consistent (across countries) high residual correlations. We first will try to achieve a fitting model using data from Wave 1, and will consecutively test whether the pattern holds when applying to Wave 2.2 After estimating a cross-country model, we turn to explaining why in some countries the model fits better compared with other countries, using the TLI, RMSEA, and SRMR values from the country models. As we have only 21 cases, we can only test a limited number of predictors in one multiple regression model. So, we start by looking at bivariate correlations to filter out those predictors that are at least correlated with model fit. In the next step, we modeled pairs and trios of those predictors with significant correlations with model fit to discern patterns of true and spurious predictions. Given the small N, we do not predict the chi-square value, as this requires adding sample size to the model (as the chi-square value depends heavily on the sample size) for which we do not have the space. Results Factor Analysis We start with the CFA model (Table 2). When looking at the model with all respondents from the 21 countries together, we find a significant chi-square value. This is, however, not surprising given the large sample (Kenny, n.d.). The TLI indicates a good fit, but the RMSEA indicates a mediocre fit (MacCullum, Browne, & Sugaware, 1996). This is a pattern returning in the measurement models in most individual countries. Table 2 Fit Indices of the Five-Dimensional CFA Model Model  Country  Wave 1   Wave 2       N  χ2df= 67  TLI  RMSEA ( ε^) [90% CI]  SRMR  N  χ2df= 67  TLI  RMSEA ( ε^) [90% CI]  SRMR  Untrimmed model  All, one group  32,396  12,560.141a  .952  .059 [.059, .059]  .042  22,795  9,527.980a  .951  .061 [.061, .063]  .041  All, separate groups  32,396  17,983.106b  .937  .015 [.014, .015]  .044  22,795  14,541.821b  .935  .015 [.015, .016]  .046  Trimmed model  All, one group  32,396  4,652.970  .976  .046 [.045, .045]  .025  22,795  3,495.632  .976  .047 [.046, .046]  .024  All, separate groups  32,396  8,205.220c  .964  .012 [.012, .012]  .027  22,795  6,862.689c  .961  .013 [.013, .013]  .027  Trimmed model  Austria  1,427  310.422  .974  .050 [.045, .056]  .028  1,001  233.192  .976  .050 [.043, .057]  .026  Belgium  2,805  715.058  .961  .059 [.055, .063]  .034  2,000  563.600  .961  .061 [.056, .066]  .035  Bulgaria  1,509  265.409  .977  .044 [.039, .050]  .026  1,016  222.375  .972  .048 [.041, .055]  .029  Czech Republic  1,488  259.393  .975  .044 [.038, .050]  .026  1,013  272.073  .965  .055 [.048, .062]  .029  Denmark  1,421  302.091  .971  .050 [.044, .055]  .031  1,017  305.666  .963  .059 [.053, .066]  .033  Finland  1,440  426.350  .958  .061 [.056, .067]  .038  1,018  347.030  .957  .064 [.058, .071]  .039  France  1,417  389.908  .963  .058 [.053, .064]  .033  1,015  349.019  .956  .064 [.058, .071]  .034  Germany  1,408  369.161  .963  .057 [.051, .062]  .031  1,004  344.682  .958  .064 [.058, .071]  .033  Greece  1,419  407.138  .954  .060 [.054, .065]  .039  1,018  336.694  .967  .063 [.056, .070]  .042  Hungary  1,414  380.170  .961  .058 [.052, .063]  .036  1,019  318.651  .959  .061 [.054, .068]  .039  Ireland  1,417  359.453  .958  .056 [.050, .061]  .039  1,004  309.320  .955  .060 [.053, .067]  .043  Italy  1,420  505.805  .948  .068 [.062, .074]  .045  1,002  412.274  .947  .072 [.065, .078]  .047  Latvia  2,245  309.915  .980  .040 [.036, .045]  .024  1,303  255.961  .977  .047 [.041, .053]  .025  Lithuania  1,393  284.148  .973  .048 [.043, .054]  .031  1,093  239.642  .973  .049 [.042, .055]  .029  The Netherlands  1,408  463.433  .952  .065 [.059, .070]  .042  1,025  386.240  .950  .068 [.062, .075]  .041  Poland  1,409  317.063  .970  .051 [.046, .057]  .037  1,011  310.863  .963  .060 [.053, .067]  .036  Portugal  1,422  502.592  .936  .068 [.062, .073]  .050  1,012  427.685  .931  .073 [.066, .080]  .052  Slovakia  1,628  364.888  .969  .052 [.047, .058]  .032  1,189  252.261  .969  .048 [.042, .055]  .031  Spain  1,457  630.042  .926  .076 [.071, .081]  .050  1,020  449.145  .933  .075 [.068, .081]  .052  Sweden  1,428  327.400  .970  .052 [.047, .058]  .031  1,008  255.420  .973  .053 [.046, .060]  .027  The U.K.  1,421  315.305  .974  .051[.045, .057]  .027  1,007  270.827  .973  .055 [.048, .062]  .027  Model  Country  Wave 1   Wave 2       N  χ2df= 67  TLI  RMSEA ( ε^) [90% CI]  SRMR  N  χ2df= 67  TLI  RMSEA ( ε^) [90% CI]  SRMR  Untrimmed model  All, one group  32,396  12,560.141a  .952  .059 [.059, .059]  .042  22,795  9,527.980a  .951  .061 [.061, .063]  .041  All, separate groups  32,396  17,983.106b  .937  .015 [.014, .015]  .044  22,795  14,541.821b  .935  .015 [.015, .016]  .046  Trimmed model  All, one group  32,396  4,652.970  .976  .046 [.045, .045]  .025  22,795  3,495.632  .976  .047 [.046, .046]  .024  All, separate groups  32,396  8,205.220c  .964  .012 [.012, .012]  .027  22,795  6,862.689c  .961  .013 [.013, .013]  .027  Trimmed model  Austria  1,427  310.422  .974  .050 [.045, .056]  .028  1,001  233.192  .976  .050 [.043, .057]  .026  Belgium  2,805  715.058  .961  .059 [.055, .063]  .034  2,000  563.600  .961  .061 [.056, .066]  .035  Bulgaria  1,509  265.409  .977  .044 [.039, .050]  .026  1,016  222.375  .972  .048 [.041, .055]  .029  Czech Republic  1,488  259.393  .975  .044 [.038, .050]  .026  1,013  272.073  .965  .055 [.048, .062]  .029  Denmark  1,421  302.091  .971  .050 [.044, .055]  .031  1,017  305.666  .963  .059 [.053, .066]  .033  Finland  1,440  426.350  .958  .061 [.056, .067]  .038  1,018  347.030  .957  .064 [.058, .071]  .039  France  1,417  389.908  .963  .058 [.053, .064]  .033  1,015  349.019  .956  .064 [.058, .071]  .034  Germany  1,408  369.161  .963  .057 [.051, .062]  .031  1,004  344.682  .958  .064 [.058, .071]  .033  Greece  1,419  407.138  .954  .060 [.054, .065]  .039  1,018  336.694  .967  .063 [.056, .070]  .042  Hungary  1,414  380.170  .961  .058 [.052, .063]  .036  1,019  318.651  .959  .061 [.054, .068]  .039  Ireland  1,417  359.453  .958  .056 [.050, .061]  .039  1,004  309.320  .955  .060 [.053, .067]  .043  Italy  1,420  505.805  .948  .068 [.062, .074]  .045  1,002  412.274  .947  .072 [.065, .078]  .047  Latvia  2,245  309.915  .980  .040 [.036, .045]  .024  1,303  255.961  .977  .047 [.041, .053]  .025  Lithuania  1,393  284.148  .973  .048 [.043, .054]  .031  1,093  239.642  .973  .049 [.042, .055]  .029  The Netherlands  1,408  463.433  .952  .065 [.059, .070]  .042  1,025  386.240  .950  .068 [.062, .075]  .041  Poland  1,409  317.063  .970  .051 [.046, .057]  .037  1,011  310.863  .963  .060 [.053, .067]  .036  Portugal  1,422  502.592  .936  .068 [.062, .073]  .050  1,012  427.685  .931  .073 [.066, .080]  .052  Slovakia  1,628  364.888  .969  .052 [.047, .058]  .032  1,189  252.261  .969  .048 [.042, .055]  .031  Spain  1,457  630.042  .926  .076 [.071, .081]  .050  1,020  449.145  .933  .075 [.068, .081]  .052  Sweden  1,428  327.400  .970  .052 [.047, .058]  .031  1,008  255.420  .973  .053 [.046, .060]  .027  The U.K.  1,421  315.305  .974  .051[.045, .057]  .027  1,007  270.827  .973  .055 [.048, .062]  .027  Notes:adf = 109. bdf = 2,289. cdf = 1,407. Table 2 Fit Indices of the Five-Dimensional CFA Model Model  Country  Wave 1   Wave 2       N  χ2df= 67  TLI  RMSEA ( ε^) [90% CI]  SRMR  N  χ2df= 67  TLI  RMSEA ( ε^) [90% CI]  SRMR  Untrimmed model  All, one group  32,396  12,560.141a  .952  .059 [.059, .059]  .042  22,795  9,527.980a  .951  .061 [.061, .063]  .041  All, separate groups  32,396  17,983.106b  .937  .015 [.014, .015]  .044  22,795  14,541.821b  .935  .015 [.015, .016]  .046  Trimmed model  All, one group  32,396  4,652.970  .976  .046 [.045, .045]  .025  22,795  3,495.632  .976  .047 [.046, .046]  .024  All, separate groups  32,396  8,205.220c  .964  .012 [.012, .012]  .027  22,795  6,862.689c  .961  .013 [.013, .013]  .027  Trimmed model  Austria  1,427  310.422  .974  .050 [.045, .056]  .028  1,001  233.192  .976  .050 [.043, .057]  .026  Belgium  2,805  715.058  .961  .059 [.055, .063]  .034  2,000  563.600  .961  .061 [.056, .066]  .035  Bulgaria  1,509  265.409  .977  .044 [.039, .050]  .026  1,016  222.375  .972  .048 [.041, .055]  .029  Czech Republic  1,488  259.393  .975  .044 [.038, .050]  .026  1,013  272.073  .965  .055 [.048, .062]  .029  Denmark  1,421  302.091  .971  .050 [.044, .055]  .031  1,017  305.666  .963  .059 [.053, .066]  .033  Finland  1,440  426.350  .958  .061 [.056, .067]  .038  1,018  347.030  .957  .064 [.058, .071]  .039  France  1,417  389.908  .963  .058 [.053, .064]  .033  1,015  349.019  .956  .064 [.058, .071]  .034  Germany  1,408  369.161  .963  .057 [.051, .062]  .031  1,004  344.682  .958  .064 [.058, .071]  .033  Greece  1,419  407.138  .954  .060 [.054, .065]  .039  1,018  336.694  .967  .063 [.056, .070]  .042  Hungary  1,414  380.170  .961  .058 [.052, .063]  .036  1,019  318.651  .959  .061 [.054, .068]  .039  Ireland  1,417  359.453  .958  .056 [.050, .061]  .039  1,004  309.320  .955  .060 [.053, .067]  .043  Italy  1,420  505.805  .948  .068 [.062, .074]  .045  1,002  412.274  .947  .072 [.065, .078]  .047  Latvia  2,245  309.915  .980  .040 [.036, .045]  .024  1,303  255.961  .977  .047 [.041, .053]  .025  Lithuania  1,393  284.148  .973  .048 [.043, .054]  .031  1,093  239.642  .973  .049 [.042, .055]  .029  The Netherlands  1,408  463.433  .952  .065 [.059, .070]  .042  1,025  386.240  .950  .068 [.062, .075]  .041  Poland  1,409  317.063  .970  .051 [.046, .057]  .037  1,011  310.863  .963  .060 [.053, .067]  .036  Portugal  1,422  502.592  .936  .068 [.062, .073]  .050  1,012  427.685  .931  .073 [.066, .080]  .052  Slovakia  1,628  364.888  .969  .052 [.047, .058]  .032  1,189  252.261  .969  .048 [.042, .055]  .031  Spain  1,457  630.042  .926  .076 [.071, .081]  .050  1,020  449.145  .933  .075 [.068, .081]  .052  Sweden  1,428  327.400  .970  .052 [.047, .058]  .031  1,008  255.420  .973  .053 [.046, .060]  .027  The U.K.  1,421  315.305  .974  .051[.045, .057]  .027  1,007  270.827  .973  .055 [.048, .062]  .027  Model  Country  Wave 1   Wave 2       N  χ2df= 67  TLI  RMSEA ( ε^) [90% CI]  SRMR  N  χ2df= 67  TLI  RMSEA ( ε^) [90% CI]  SRMR  Untrimmed model  All, one group  32,396  12,560.141a  .952  .059 [.059, .059]  .042  22,795  9,527.980a  .951  .061 [.061, .063]  .041  All, separate groups  32,396  17,983.106b  .937  .015 [.014, .015]  .044  22,795  14,541.821b  .935  .015 [.015, .016]  .046  Trimmed model  All, one group  32,396  4,652.970  .976  .046 [.045, .045]  .025  22,795  3,495.632  .976  .047 [.046, .046]  .024  All, separate groups  32,396  8,205.220c  .964  .012 [.012, .012]  .027  22,795  6,862.689c  .961  .013 [.013, .013]  .027  Trimmed model  Austria  1,427  310.422  .974  .050 [.045, .056]  .028  1,001  233.192  .976  .050 [.043, .057]  .026  Belgium  2,805  715.058  .961  .059 [.055, .063]  .034  2,000  563.600  .961  .061 [.056, .066]  .035  Bulgaria  1,509  265.409  .977  .044 [.039, .050]  .026  1,016  222.375  .972  .048 [.041, .055]  .029  Czech Republic  1,488  259.393  .975  .044 [.038, .050]  .026  1,013  272.073  .965  .055 [.048, .062]  .029  Denmark  1,421  302.091  .971  .050 [.044, .055]  .031  1,017  305.666  .963  .059 [.053, .066]  .033  Finland  1,440  426.350  .958  .061 [.056, .067]  .038  1,018  347.030  .957  .064 [.058, .071]  .039  France  1,417  389.908  .963  .058 [.053, .064]  .033  1,015  349.019  .956  .064 [.058, .071]  .034  Germany  1,408  369.161  .963  .057 [.051, .062]  .031  1,004  344.682  .958  .064 [.058, .071]  .033  Greece  1,419  407.138  .954  .060 [.054, .065]  .039  1,018  336.694  .967  .063 [.056, .070]  .042  Hungary  1,414  380.170  .961  .058 [.052, .063]  .036  1,019  318.651  .959  .061 [.054, .068]  .039  Ireland  1,417  359.453  .958  .056 [.050, .061]  .039  1,004  309.320  .955  .060 [.053, .067]  .043  Italy  1,420  505.805  .948  .068 [.062, .074]  .045  1,002  412.274  .947  .072 [.065, .078]  .047  Latvia  2,245  309.915  .980  .040 [.036, .045]  .024  1,303  255.961  .977  .047 [.041, .053]  .025  Lithuania  1,393  284.148  .973  .048 [.043, .054]  .031  1,093  239.642  .973  .049 [.042, .055]  .029  The Netherlands  1,408  463.433  .952  .065 [.059, .070]  .042  1,025  386.240  .950  .068 [.062, .075]  .041  Poland  1,409  317.063  .970  .051 [.046, .057]  .037  1,011  310.863  .963  .060 [.053, .067]  .036  Portugal  1,422  502.592  .936  .068 [.062, .073]  .050  1,012  427.685  .931  .073 [.066, .080]  .052  Slovakia  1,628  364.888  .969  .052 [.047, .058]  .032  1,189  252.261  .969  .048 [.042, .055]  .031  Spain  1,457  630.042  .926  .076 [.071, .081]  .050  1,020  449.145  .933  .075 [.068, .081]  .052  Sweden  1,428  327.400  .970  .052 [.047, .058]  .031  1,008  255.420  .973  .053 [.046, .060]  .027  The U.K.  1,421  315.305  .974  .051[.045, .057]  .027  1,007  270.827  .973  .055 [.048, .062]  .027  Notes:adf = 109. bdf = 2,289. cdf = 1,407. Looking at the residual correlations between the indicators, we find that three indicators return in more than half of the countries as problematic (with high residual correlations with different indicators in different countries). This is the first strengthening indicator (“The EU should become one country”), the fourth performance indicator (“The decision making process in the EU is transparent”), and the fourth negative affection indicator (“I am afraid of the EU”). As we are searching for a measurement model that works in all countries (and because the high residual correlations with the problematic indicators are not with the same indicators in each country), adding error correlations would not be helpful. So, we opted to drop these indicators. Dropping these items leads to good model fit: TLI = .976, ε^ = .046 (see Table 2). Figure 1 shows the model when analyzing all respondents as one group. Low convergent validity may indicate the model has too few factors. But with all (but one) standardized factor loadings exceeding .70 (loading PERF1 = .66), it is indicated that the majority of the variance in the indicators is predicted by the factors. Per factor, we also assess the average variance extracted, which are well above the threshold of 50% for all factors in all countries (AVEperformance = 74%; AVEutilitarian = 82%; AVEidentity = 84%; AVEaffections = 81%; AVEstrengthening = 79%). Looking at the individual countries, we see the loading on PERF1 consistently below, but always close to .70. In one case, a strengthening item (STR3) is much lower in one country (Portugal). Although this is low, we observe no pattern of high residual correlations with other indicators with relative low loading, so we interpret this as not problematic for the number of factors. Testing the AVE per factor, we find that also in each country, all are above the threshold of 50%, indicating that the factors explain more than half of the variation in the items. We thus have good convergent validity and are confident we do not have a too little number of factors. Figure 1 View largeDownload slide Measurement model of five dimensions of EU attitudes (respondents of all 21 countries in one group). The entries at the double-headed arrows represent the correlations between the factors (before the slash Wave 1; after the slash Wave 2); the entries at the arrows toward the indicators represent standardized factor loadings. The lighter grey indicators have been dropped from the model because of reducing fit in more than half of the countries in the study Figure 1 View largeDownload slide Measurement model of five dimensions of EU attitudes (respondents of all 21 countries in one group). The entries at the double-headed arrows represent the correlations between the factors (before the slash Wave 1; after the slash Wave 2); the entries at the arrows toward the indicators represent standardized factor loadings. The lighter grey indicators have been dropped from the model because of reducing fit in more than half of the countries in the study In terms of discriminant validity, we find that all correlations between the factors are <.80, which is a rough indication that the factors are empirically distinguishable. We find the strongest correlation between the performance factor and the identity factor (r = .78), and test whether we can merge these factors (i.e., constrain their correlation to be 1). This leads to a substantial and significant loss of model fit (χ2df=1 = 8,077.14, p < .001). Testing whether the correlations between other pairs of factors can be constrained to 1 leads similarly to significant loss of model fit, implying that as one group we cannot merge the factors. Looking at the individual countries, we find that in five countries, eight correlations exceed .80. We also tested whether merging the factors was possible, but in each case, this significantly reduced model fit (indicating that the factors are empirically truly different).3 Applying the trimmed model (i.e., the model excluding the three problematic indicators) to Wave 2 leads to a model with good fit, both for the one-group model as for the multiple group model (see Table 2). We thus find that the five-dimensional model holds in all 21 countries in Wave 2, although in some countries the model certainly fits better than in others.4 We see a particularly well-fitting model for instance in Latvia, Bulgaria, and Austria, while the worst model fits appear in Spain, Portugal, Italy, and The Netherlands. In the next step, we seek to understand whether these differences are systematic patterns that can be explained by the country characteristics specified above. Predicting Model fit Turning to predicting model fit with contextual variables, we first look at the bivariate correlations (see Table 3). With regard to length of EU membership, we find a rather consistent pattern of longer members having worse fit. Higher values of chi-square and RMSEA ( ε^) indicate worse fit, so the negative correlations with year of entry in both Wave 1 (with χ2: r = −.52; with ε^: r = −.51) and Wave 2 (with χ2: r = −.40; with ε^: r = −.40) indicate that newer members have better fit. For TLI, higher values indicate better fit, and the positive coefficient (although not significant) points in the same direction. Focusing on the membership dichotomies of EU12 and EU15, all correlations (with the different indices and in both waves) are significant. But as being a member of EU12 or EU15 is coded as 1 (and not a EU12 or EU15 member 0), the positive correlation with the chi-square and the RMSEA and the negative correlation with the TLI indicate again that older members have worse fit. These findings run counter to our expectations formulated above; it appears that longer membership does not lead to more clearly distinct attitude dimensions. Table 3 Correlations Between Predictors and Model Fit Predictors [of model fit]  Wave 1   Wave 2     χ2  TLI  ε^  SRMR  χ2  TLI  ε^  SRMR  Year of entry  −.52*  .38+  −.51*  −.32  −.62**  .41+  −.58**  −.34  Enlargement round  −.50*  .37+  −.50*  −.32  −.60**  .39+  −.57**  −.34  Member EU6  .42+  −.21  .37+  .17  .50*  −.29  .43+  .16  Member EU12  .61**  −.58**  .62**  .50*  .71***  −.58**  .69***  .56**  Member EU15  .59**  −.51*  .61**  .43*  .61**  −.42+  .59**  .40+  Member Eurozone (including Slovakia)  .69***  −.64**  .69***  .58**  .60**  −.54*  .56**  .57**  Member Eurozone (excluding Slovakia)  .72***  −.68***  .73***  .61**  .72***  −.60**  .69***  .62**  Country’s contribution to EU  .32  −.22  .35  .11  .42+  −.28  .43+  .12  EU expenditure in country  .50*  −.43*  .49*  .37+  .58**  −.46*  .58**  .38+  Contribution–expenditure difference  .05  −.14  −.01  .25  −.01  −.08  −.04  .24  Country’s contribution to EU per capita  .06  −.10  −.02  .14  −.02  −.01  −.07  .19  EU expenditure in country per capita  .34  −.20  .33  .15  .37  −.18  .35  .12  Contribution–expenditure difference per capita  −.15  .03  −.19  .02  −.23  .09  −.23  .08  Absolute number EU citizens in country  .40+  −.33  .40+  .14  .42+  −.31  .41+  .16  Absolute number non-EU citizens in country  .39+  −.35  .41+  .18  .44*  −.34  .46*  .21  Relative number EU citizens in country  .33  −.25  .26  .17  .27  −.19  .21  .20  Relative number non-EU citizens in country  −.14  .01  −.11  −.14  −.08  .10  −.06  −.10  Increase in EU citizens in country  −.43+  .50*  −.42+  −.60**  −.42+  .51*  −.43+  −.53*  Increase in non-EU citizens in country  .06  −.02  .00  −.08  −.04  −.07  −.08  −.04  Absolute EU stories (all media)  .29  −.34  .28  .30  .26  −.13  .25  .34  Relative EU stories (all media)  −.08  −.02  −.04  .07  −.08  .17  −.05  .07  Absolute EU stories (in newspapers)  .05  −.06  .08  .07  .10  .09  .11  .10  Relative EU stories (in newspapers)  −.13  .09  −.10  −.13  −.14  .23  −.11  −.12  Absolute EU stories (in TV news)  .39+  −.46*  .36  .39+  .32  −.24  .29  .43+  Relative EU stories (in TV news)  .02  −.14  .06  .22  .03  .04  .06  .22  Absolute import from EU countries  .31  −.17  .31  .07  .41+  −.25  .39+  .06  Absolute export to EU countries  .32  −.17  .31  .04  .41+  −.24  .39+  .05  Absolute trade with EU countries  .32  −.17  .31  .06  .41+  −.25  .39+  .06  Proportion of total import from EU countries  −.26  .18  −.29  −.18  −.22  .05  −.23  −.16  Proportion of total export to EU countries  −.04  .03  −.10  .01  −.06  −.10  −.09  .00  Proportion trade with EU countries of total trade  −.16  .11  −.21  −.08  −.15  −.04  −.17  −.08  Absolute change in import from EU countries  −.28  .13  −.29  −.03  −.38  .20  −.36  −.02  Absolute change in export to EU countries  −.37  .20  −.35  −.08  −.44+  .26  −.42+  −.07  Absolute change in trade with EU countries  −.32  .16  −.32  −.05  −.41+  .23  −.39+  −.04  Change in proportion of import from EU countries  .39+  −.41+  .38+  .37+  .34  −.39+  .33  .37  Change in proportion of export to EU countries  −.24  .19  −.27  −.13  −.32  .13  −.33  −.14  Change in proportion trade with EU countries  .18  −.23  .15  .24  .09  −.22  .08  .24  Predictors [of model fit]  Wave 1   Wave 2     χ2  TLI  ε^  SRMR  χ2  TLI  ε^  SRMR  Year of entry  −.52*  .38+  −.51*  −.32  −.62**  .41+  −.58**  −.34  Enlargement round  −.50*  .37+  −.50*  −.32  −.60**  .39+  −.57**  −.34  Member EU6  .42+  −.21  .37+  .17  .50*  −.29  .43+  .16  Member EU12  .61**  −.58**  .62**  .50*  .71***  −.58**  .69***  .56**  Member EU15  .59**  −.51*  .61**  .43*  .61**  −.42+  .59**  .40+  Member Eurozone (including Slovakia)  .69***  −.64**  .69***  .58**  .60**  −.54*  .56**  .57**  Member Eurozone (excluding Slovakia)  .72***  −.68***  .73***  .61**  .72***  −.60**  .69***  .62**  Country’s contribution to EU  .32  −.22  .35  .11  .42+  −.28  .43+  .12  EU expenditure in country  .50*  −.43*  .49*  .37+  .58**  −.46*  .58**  .38+  Contribution–expenditure difference  .05  −.14  −.01  .25  −.01  −.08  −.04  .24  Country’s contribution to EU per capita  .06  −.10  −.02  .14  −.02  −.01  −.07  .19  EU expenditure in country per capita  .34  −.20  .33  .15  .37  −.18  .35  .12  Contribution–expenditure difference per capita  −.15  .03  −.19  .02  −.23  .09  −.23  .08  Absolute number EU citizens in country  .40+  −.33  .40+  .14  .42+  −.31  .41+  .16  Absolute number non-EU citizens in country  .39+  −.35  .41+  .18  .44*  −.34  .46*  .21  Relative number EU citizens in country  .33  −.25  .26  .17  .27  −.19  .21  .20  Relative number non-EU citizens in country  −.14  .01  −.11  −.14  −.08  .10  −.06  −.10  Increase in EU citizens in country  −.43+  .50*  −.42+  −.60**  −.42+  .51*  −.43+  −.53*  Increase in non-EU citizens in country  .06  −.02  .00  −.08  −.04  −.07  −.08  −.04  Absolute EU stories (all media)  .29  −.34  .28  .30  .26  −.13  .25  .34  Relative EU stories (all media)  −.08  −.02  −.04  .07  −.08  .17  −.05  .07  Absolute EU stories (in newspapers)  .05  −.06  .08  .07  .10  .09  .11  .10  Relative EU stories (in newspapers)  −.13  .09  −.10  −.13  −.14  .23  −.11  −.12  Absolute EU stories (in TV news)  .39+  −.46*  .36  .39+  .32  −.24  .29  .43+  Relative EU stories (in TV news)  .02  −.14  .06  .22  .03  .04  .06  .22  Absolute import from EU countries  .31  −.17  .31  .07  .41+  −.25  .39+  .06  Absolute export to EU countries  .32  −.17  .31  .04  .41+  −.24  .39+  .05  Absolute trade with EU countries  .32  −.17  .31  .06  .41+  −.25  .39+  .06  Proportion of total import from EU countries  −.26  .18  −.29  −.18  −.22  .05  −.23  −.16  Proportion of total export to EU countries  −.04  .03  −.10  .01  −.06  −.10  −.09  .00  Proportion trade with EU countries of total trade  −.16  .11  −.21  −.08  −.15  −.04  −.17  −.08  Absolute change in import from EU countries  −.28  .13  −.29  −.03  −.38  .20  −.36  −.02  Absolute change in export to EU countries  −.37  .20  −.35  −.08  −.44+  .26  −.42+  −.07  Absolute change in trade with EU countries  −.32  .16  −.32  −.05  −.41+  .23  −.39+  −.04  Change in proportion of import from EU countries  .39+  −.41+  .38+  .37+  .34  −.39+  .33  .37  Change in proportion of export to EU countries  −.24  .19  −.27  −.13  −.32  .13  −.33  −.14  Change in proportion trade with EU countries  .18  −.23  .15  .24  .09  −.22  .08  .24  Notes: Entries are Pearson correlations. For the correlations with chi-square, we control for sample size. *** p < .001, **p < .01, p < .05, +p < .1 (two-tailed). Table 3 Correlations Between Predictors and Model Fit Predictors [of model fit]  Wave 1   Wave 2     χ2  TLI  ε^  SRMR  χ2  TLI  ε^  SRMR  Year of entry  −.52*  .38+  −.51*  −.32  −.62**  .41+  −.58**  −.34  Enlargement round  −.50*  .37+  −.50*  −.32  −.60**  .39+  −.57**  −.34  Member EU6  .42+  −.21  .37+  .17  .50*  −.29  .43+  .16  Member EU12  .61**  −.58**  .62**  .50*  .71***  −.58**  .69***  .56**  Member EU15  .59**  −.51*  .61**  .43*  .61**  −.42+  .59**  .40+  Member Eurozone (including Slovakia)  .69***  −.64**  .69***  .58**  .60**  −.54*  .56**  .57**  Member Eurozone (excluding Slovakia)  .72***  −.68***  .73***  .61**  .72***  −.60**  .69***  .62**  Country’s contribution to EU  .32  −.22  .35  .11  .42+  −.28  .43+  .12  EU expenditure in country  .50*  −.43*  .49*  .37+  .58**  −.46*  .58**  .38+  Contribution–expenditure difference  .05  −.14  −.01  .25  −.01  −.08  −.04  .24  Country’s contribution to EU per capita  .06  −.10  −.02  .14  −.02  −.01  −.07  .19  EU expenditure in country per capita  .34  −.20  .33  .15  .37  −.18  .35  .12  Contribution–expenditure difference per capita  −.15  .03  −.19  .02  −.23  .09  −.23  .08  Absolute number EU citizens in country  .40+  −.33  .40+  .14  .42+  −.31  .41+  .16  Absolute number non-EU citizens in country  .39+  −.35  .41+  .18  .44*  −.34  .46*  .21  Relative number EU citizens in country  .33  −.25  .26  .17  .27  −.19  .21  .20  Relative number non-EU citizens in country  −.14  .01  −.11  −.14  −.08  .10  −.06  −.10  Increase in EU citizens in country  −.43+  .50*  −.42+  −.60**  −.42+  .51*  −.43+  −.53*  Increase in non-EU citizens in country  .06  −.02  .00  −.08  −.04  −.07  −.08  −.04  Absolute EU stories (all media)  .29  −.34  .28  .30  .26  −.13  .25  .34  Relative EU stories (all media)  −.08  −.02  −.04  .07  −.08  .17  −.05  .07  Absolute EU stories (in newspapers)  .05  −.06  .08  .07  .10  .09  .11  .10  Relative EU stories (in newspapers)  −.13  .09  −.10  −.13  −.14  .23  −.11  −.12  Absolute EU stories (in TV news)  .39+  −.46*  .36  .39+  .32  −.24  .29  .43+  Relative EU stories (in TV news)  .02  −.14  .06  .22  .03  .04  .06  .22  Absolute import from EU countries  .31  −.17  .31  .07  .41+  −.25  .39+  .06  Absolute export to EU countries  .32  −.17  .31  .04  .41+  −.24  .39+  .05  Absolute trade with EU countries  .32  −.17  .31  .06  .41+  −.25  .39+  .06  Proportion of total import from EU countries  −.26  .18  −.29  −.18  −.22  .05  −.23  −.16  Proportion of total export to EU countries  −.04  .03  −.10  .01  −.06  −.10  −.09  .00  Proportion trade with EU countries of total trade  −.16  .11  −.21  −.08  −.15  −.04  −.17  −.08  Absolute change in import from EU countries  −.28  .13  −.29  −.03  −.38  .20  −.36  −.02  Absolute change in export to EU countries  −.37  .20  −.35  −.08  −.44+  .26  −.42+  −.07  Absolute change in trade with EU countries  −.32  .16  −.32  −.05  −.41+  .23  −.39+  −.04  Change in proportion of import from EU countries  .39+  −.41+  .38+  .37+  .34  −.39+  .33  .37  Change in proportion of export to EU countries  −.24  .19  −.27  −.13  −.32  .13  −.33  −.14  Change in proportion trade with EU countries  .18  −.23  .15  .24  .09  −.22  .08  .24  Predictors [of model fit]  Wave 1   Wave 2     χ2  TLI  ε^  SRMR  χ2  TLI  ε^  SRMR  Year of entry  −.52*  .38+  −.51*  −.32  −.62**  .41+  −.58**  −.34  Enlargement round  −.50*  .37+  −.50*  −.32  −.60**  .39+  −.57**  −.34  Member EU6  .42+  −.21  .37+  .17  .50*  −.29  .43+  .16  Member EU12  .61**  −.58**  .62**  .50*  .71***  −.58**  .69***  .56**  Member EU15  .59**  −.51*  .61**  .43*  .61**  −.42+  .59**  .40+  Member Eurozone (including Slovakia)  .69***  −.64**  .69***  .58**  .60**  −.54*  .56**  .57**  Member Eurozone (excluding Slovakia)  .72***  −.68***  .73***  .61**  .72***  −.60**  .69***  .62**  Country’s contribution to EU  .32  −.22  .35  .11  .42+  −.28  .43+  .12  EU expenditure in country  .50*  −.43*  .49*  .37+  .58**  −.46*  .58**  .38+  Contribution–expenditure difference  .05  −.14  −.01  .25  −.01  −.08  −.04  .24  Country’s contribution to EU per capita  .06  −.10  −.02  .14  −.02  −.01  −.07  .19  EU expenditure in country per capita  .34  −.20  .33  .15  .37  −.18  .35  .12  Contribution–expenditure difference per capita  −.15  .03  −.19  .02  −.23  .09  −.23  .08  Absolute number EU citizens in country  .40+  −.33  .40+  .14  .42+  −.31  .41+  .16  Absolute number non-EU citizens in country  .39+  −.35  .41+  .18  .44*  −.34  .46*  .21  Relative number EU citizens in country  .33  −.25  .26  .17  .27  −.19  .21  .20  Relative number non-EU citizens in country  −.14  .01  −.11  −.14  −.08  .10  −.06  −.10  Increase in EU citizens in country  −.43+  .50*  −.42+  −.60**  −.42+  .51*  −.43+  −.53*  Increase in non-EU citizens in country  .06  −.02  .00  −.08  −.04  −.07  −.08  −.04  Absolute EU stories (all media)  .29  −.34  .28  .30  .26  −.13  .25  .34  Relative EU stories (all media)  −.08  −.02  −.04  .07  −.08  .17  −.05  .07  Absolute EU stories (in newspapers)  .05  −.06  .08  .07  .10  .09  .11  .10  Relative EU stories (in newspapers)  −.13  .09  −.10  −.13  −.14  .23  −.11  −.12  Absolute EU stories (in TV news)  .39+  −.46*  .36  .39+  .32  −.24  .29  .43+  Relative EU stories (in TV news)  .02  −.14  .06  .22  .03  .04  .06  .22  Absolute import from EU countries  .31  −.17  .31  .07  .41+  −.25  .39+  .06  Absolute export to EU countries  .32  −.17  .31  .04  .41+  −.24  .39+  .05  Absolute trade with EU countries  .32  −.17  .31  .06  .41+  −.25  .39+  .06  Proportion of total import from EU countries  −.26  .18  −.29  −.18  −.22  .05  −.23  −.16  Proportion of total export to EU countries  −.04  .03  −.10  .01  −.06  −.10  −.09  .00  Proportion trade with EU countries of total trade  −.16  .11  −.21  −.08  −.15  −.04  −.17  −.08  Absolute change in import from EU countries  −.28  .13  −.29  −.03  −.38  .20  −.36  −.02  Absolute change in export to EU countries  −.37  .20  −.35  −.08  −.44+  .26  −.42+  −.07  Absolute change in trade with EU countries  −.32  .16  −.32  −.05  −.41+  .23  −.39+  −.04  Change in proportion of import from EU countries  .39+  −.41+  .38+  .37+  .34  −.39+  .33  .37  Change in proportion of export to EU countries  −.24  .19  −.27  −.13  −.32  .13  −.33  −.14  Change in proportion trade with EU countries  .18  −.23  .15  .24  .09  −.22  .08  .24  Notes: Entries are Pearson correlations. For the correlations with chi-square, we control for sample size. *** p < .001, **p < .01, p < .05, +p < .1 (two-tailed). With regard to participating in the common currency, we again find a pattern that the predictor matters for the fit of the dimensional structure but not in the expected direction (see Table 3). For all fit indices and in both waves, we find significant correlations that imply that in countries that participate in the common currency show worse model fit compared with countries not part of the Eurozone. Eurozone membership does thus not lead to a crystallization of attitudes along the five dimensions. Turning to the countries’ financial relationships with the EU, we find that most predictors are not significantly correlated with the fit indices, implying the financial relations hardly matter. We do find a pattern of significant correlations across waves and fit indices for the contributions of the EU in each country, with the coefficients suggesting that the dimensional structure is more pronounced in countries in which the EU invests less. Also, this finding runs counter our provisional expectation in that more financial attachment to the EU does not translate into a clearer manifestation of the five attitude dimensions. Also, the economic ties between EU member states do not seem to matter much for the dimensional structure. Only countries that saw a growth of EU import show worse model fit, again not in line with what we had expected. Overall, it appears so far that the salience of the EU because of membership length or economic ties with the EU rather leads to the five attitude dimensions becoming less distinct and more interchangeable. With regard to the presence of EU (and non-EU) nationals from other countries, we find that the absolute number of EU citizens (and non-EU citizens) is marginally correlated with the chi-square and the RMSEA in Wave 1. In Wave 2, however, these correlations are not significant anymore. The change in the number of EU citizens in one’s country is significantly correlated with all fit indices and in both waves, implying that in countries in which the number of foreign EU nationals increases, we find a clearer representation of the five attitude dimensions. This finding is consistent with our expectation, and it is noteworthy here that we do not find a similar correlation for the growth of the number of non-EU nationals in each country. The last predictive factor that we take into account is media visibility of the EU. We only find a pattern of (marginally) significant correlations (in both waves and for all fit indices) with the absolute number of EU stories in the media, in which more EU visibility, and this higher attention to the EU, leads to a worse fitting model. Again, our expectation that salience leads to attitude crystallization along the five dimensions is not supported, rather to the contrary. These contextual indicators are, however, correlated, so that we may be witnessing spurious correlations with model fit here. Therefore, we aim to test these relationships with multiple regression. However, because of the small sample size (N = 21), we cannot simply add all seven significant correlates as predictors of model fit. We thus followed a strategy of testing pairs and trios of predictors. This showed a pattern of Eurozone membership and the increase in EU citizens as significant predictors (see Table 4), while other predictors in combination with these two turn insignificant (see Supplementary Appendix A6 for the models including the other predictors). Table 4 Multiple Regression Analysis Predicting Model Fit Wave  Predictors  TLI  RMSEA  SRMR  Wave 1  Member Eurozone  −0.62***(4.35)  0.68***(4.81)  0.56***(4.15)  Increase in EU citizens  0.48**(3.36)  −0.40*(2.84)  −0.58***(4.32)  R2  .63  .64  .67  Adjusted R2  .59  .60  .64  Wave 2  Member Eurozone  −0.53**(3.31)  0.55**(3.24)  0.55**(3.61)  Increase in EU citizens  0.50**(3.12)  −0.41*(2.42)  −0.51**(3.34)  R2  .54  .48  .58  Adjusted R2  .49  .43  .53  Wave  Predictors  TLI  RMSEA  SRMR  Wave 1  Member Eurozone  −0.62***(4.35)  0.68***(4.81)  0.56***(4.15)  Increase in EU citizens  0.48**(3.36)  −0.40*(2.84)  −0.58***(4.32)  R2  .63  .64  .67  Adjusted R2  .59  .60  .64  Wave 2  Member Eurozone  −0.53**(3.31)  0.55**(3.24)  0.55**(3.61)  Increase in EU citizens  0.50**(3.12)  −0.41*(2.42)  −0.51**(3.34)  R2  .54  .48  .58  Adjusted R2  .49  .43  .53  Note: Entries are standardized ordinary least squares regression coefficients, with absolute t-values in parentheses. N = 21. ***p < .001; **p < .01; *p < .05; +p < .1 (two-tailed). Table 4 Multiple Regression Analysis Predicting Model Fit Wave  Predictors  TLI  RMSEA  SRMR  Wave 1  Member Eurozone  −0.62***(4.35)  0.68***(4.81)  0.56***(4.15)  Increase in EU citizens  0.48**(3.36)  −0.40*(2.84)  −0.58***(4.32)  R2  .63  .64  .67  Adjusted R2  .59  .60  .64  Wave 2  Member Eurozone  −0.53**(3.31)  0.55**(3.24)  0.55**(3.61)  Increase in EU citizens  0.50**(3.12)  −0.41*(2.42)  −0.51**(3.34)  R2  .54  .48  .58  Adjusted R2  .49  .43  .53  Wave  Predictors  TLI  RMSEA  SRMR  Wave 1  Member Eurozone  −0.62***(4.35)  0.68***(4.81)  0.56***(4.15)  Increase in EU citizens  0.48**(3.36)  −0.40*(2.84)  −0.58***(4.32)  R2  .63  .64  .67  Adjusted R2  .59  .60  .64  Wave 2  Member Eurozone  −0.53**(3.31)  0.55**(3.24)  0.55**(3.61)  Increase in EU citizens  0.50**(3.12)  −0.41*(2.42)  −0.51**(3.34)  R2  .54  .48  .58  Adjusted R2  .49  .43  .53  Note: Entries are standardized ordinary least squares regression coefficients, with absolute t-values in parentheses. N = 21. ***p < .001; **p < .01; *p < .05; +p < .1 (two-tailed). For example, predicting the TLI value in Wave 1, we find significant coefficients for both membership of the Eurozone (b* = −0.62, t = 4.35, p < .001) and the increase of EI citizens in the country (b* = 0.48, t = 3.36, p = .004). The negative sign for Eurozone membership implies that members have a lower TLI value, meaning a worse fit and a less pronounced dimensional structure; the positive sign for EU citizens implies a better fit when the number of EU citizens increases in the country, and this pattern is consistent among waves and fit indices.5 Also, adding any of the other significant correlates with the fit indices to the model shows a pattern of keeping Eurozone membership and the increase in EU citizens’ significant predictors, while the added predictor is not significant (see Supplementary Appendix Table A6). Although there is an exception, with being a member of the EU12 also a significant predictor in one model (predicting RMSEA in Wave 2) and this predictor turning Eurozone membership coefficient marginally significant in two models (TLI and RMSEA in Wave 2), the general pattern remains that participating in the common currency reduces model fit, which is the opposite of what we expected, and that the growth of the number of EU nationals in each country predicts a better fitting model (i.e., a more pronounces dimensional structure), which is consistent with our expectation. Discussion Given the (increasing) political implications of EU attitudes and their importance in the public debate about European integration, this study set out to further understanding the structure of EU attitudes. Particularly, it focused on whether the multidimensional structure established in prior research would hold across different countries, and what might explain any cross-country variation in this structure. Doing so, we answered the call that “EU-wide surveys […] that incorporate the different dimensions of EU attitudes […] are needed, in order to test the cross-national validity of our findings and potentially to consider how contexts affect different attitude structures” (Boomgaarden et al. 2011, p. 261). Our results suggest that indeed, generally speaking, we can confirm the multidimensional structure of five attitude dimensions also cross-nationally. By and large, we find that across the 21 EU member states, people do consider different aspects of the EU and European integration when responding to attitude questions, and hence show variability in terms of their attitudes, distinguishing between emotional responses, utilitarian considerations, matters of identity, performance, and further strengthening of the EU. Our results, however, also suggest that the clarity with which EU attitudes fall into the different dimension does vary between countries. For some, such as Latvia, Bulgaria, and Austria, we find strong and clear evidence of a good fit of the five dimensions of EU attitudes. In other countries, such as Spain, Portugal, Italy, and The Netherlands, the five dimensions are still empirically distinguishable, but the indices suggest a less well-fitting structure. The original single-country study included mostly items that were not pertinent to The Netherlands specifically, and this cross-national study confirms that these measures can be used cross-nationally. Overall, this has important implications for research on public opinion toward the EU and European integration. Not only scholars but also public commentators need to acknowledge that opposition against the EU (or support for that matter) on certain aspects does not necessarily translate into opposition (or support) for other aspects. One can strongly identify with the EU and Europe, but at the same time oppose further steps toward a strengthening of the integration process. One can support the EU from a utilitarian perspective but still disapprove of the performance of its institutions. It will be important for future research and our understanding of public opinion toward the EU to further investigate which aspects and which dimensions of EU attitudes are more or less salient when making political decisions, both on the level of EU politics but also in national situations. While the Brexit referendum for instance is a prime example of the relevance of public opinion toward the EU, it would be important to understand what aspects of EU attitudes were more or less important in opposing membership. The same goes for the domestic ramifications of EU attitudes (Abbarno & Zapryonava, 2013; de Vries, 2007). National and EU politicians will be in a better position to campaign for or against Europe when better understanding the salience but also the possible contrasts of different EU attitudes dimensions. Similarly, some attitudes might be easier to move or quicker in responding to external events or crises than others. Further differentiation between EU attitudes is thus rather likely here to stay if we seek a fuller understanding of how Europe’s citizens view the EU. An important novelty of our study lies in the explanatory account of cross-national applicability of the established dimensionality of EU attitudes. Going beyond anecdotal evidence, our aim was to systematically explain why the structure fits more or less well in different member states. We found that in particular two factors matter for our understanding of country differences. First, and against our tentative expectation, we found that length of membership led to a less clear manifestation of the five dimensions of EU attitudes. One explanation is that rather than making the EU consistently more salient over time, it could be that EU membership wears off and provides fewer inclinations for citizens to differentiate clearly different aspects of the EU and the integration process. This would suggest a centrifugal process in which different EU attitudes do converge and yield an underlying, organizing principle. An alternative explanation is that EU attitudes in long-standing member countries become even more diverse, as citizens increase their awareness of the EU and its development. This could lead to even more nuanced attitudes than even the current five-dimensional structure was able to assess. Future research would be well advised to investigate how time affects and interacts with the nature of such attitudes. Second, migration mattered in that an increase in EU migrants in a given country was related to more crystalized EU attitudes along the five dimensions. This could be related to salience, in that more EU migrants make the free movement of people, and thus the integration process more salient to the citizens. Alternatively however, this could suggest that particular aspects of the integration process become salient and that would lead to differentiation of attitudes. It remains to be seen, however, whether this cross-national variation is stable (see also below). A clear limitation that we acknowledge is that our data were collected around the 2009 European election, and one could criticize that given the euro crisis came into full swing only afterward, and also the refugee crisis or the Brexit has hit the EU only recently, our data would be outdated. While it is reasonable to speculate that such severe events would affect the structure of EU attitudes, there is little empirical evidence to support such speculation. First, most of the EU-related crisis literature (Braun & Tausendpfund, 2014; Hobolt & Wratil, 2015; Kuhn & Stoeckel, 2014) suggested a drop in support but not necessarily a change in the dimensional structure. It would be interesting to see, however, whether this drop would relate more to certain dimensions than to others. Tentatively, one could assume that performance indicators would be particularly prone to such crises. Second, we have evidence from another replication of the item battery in The Netherlands in 2014 and 2016, suggesting that the five-dimensional structure still hold up (de Vreese, Azrout & Möller, 2016a, 2016b; Marquart, Goldberg, van Elsas, Brosius & de Vreese, 2017). This makes us confident that our cross-national data are still relevant. It remains of course to be seen whether the cross-country picture we draw above still holds. In particular, some countries that were more affected by recent events than others, such as Greece or the U.K., could show changes in the dimensionality. We also propose to further elaborate on the role of the media in creating contextual variance (see Hallin and Mancini (2017) for an updated reflection on comparing media systems). Such variance would not only be in the visibility of the EU itself, as modeled in this study, but also as platform for covering issues pertaining to the economy and immigration, thereby being an important source of information for citizens when forming EU attitudes (see also de Vreese & Boomgaarden, 2005). Such an expansion of the inquiry would also be wise to include attention to negative information and its persistency, even after possibly being corrected (Thorson, 2016). Finally, we believe it would be important to develop the multidimensional nature even more, with the EU engraining in more aspects of political and social life or in international affairs. But both trust (Easton, 1975; Harteveld et al., 2015) and the very core of EU attitudes might also experience a resurgence: it still really matters if citizens are “for or against membership of the EU.” Supplementary Data Supplementary Data are available at IJPOR online. Claes H. de Vreese is a Professor of Political Communication, Amsterdam School of Communication Research, ASCoR, University of Amsterdam. Rachid Azrout is a Lecturer in Political Communication at the Department of Communication Science, University of Amsterdam. Hajo G. Boomgaarden, Professor of Social Science Methods, Department of Communication, University of Vienna. Footnotes 1When performing multiple group CFA with SEM, the equivalence between the groups can bet tested at different levels. We test here the first/basic level, which tests whether between groups the same factor structure exists. At the next levels, the equivalence of factor loadings, correlations between factors and error disturbances can be evaluated. We stick to the first level, as this shows whether the same factors emerge from the same items. Would we test also the equivalence of factor loadings, we find that constraining the factor loadings across groups leads to a significant decrease of model fit (χ2df=280 = 3,681.720, p < .001 in Wave 1; χ2df=280 = 2,701.594, p < .001 in Wave 2), implying that factor loadings are indeed not the same across countries. 2At the face level, these dimensions work empirically also in a cross-national setting. Van Spanje and de Vreese (2014) included these five dimensions in a study of campaign effects on voting behavior in European Parliament election. However, they included these dimensions as control variables, as the study was mostly concerned with testing media effects during the campaign. Given this different focus, that study was therefore unable to unpack the structure and composition of the dimensions in more detail. 3We found high correlations (>.80) between the utilitarian dimension and strengthening dimension in France (r = .85), the U.K. (r = .85), and Sweden (r = .82). Merging the dimension in these countries significantly decreased model fit (France: χ2df = 1 = 183.841, p < .001; the U.K.: χ2df = 1 = 197.201, p <.001; Sweden: χ2df = 1 = 209.100, p <.001). With regard to the correlation between the performance dimension and the identity dimension, we found high correlations in the U.K. (r = .83), Sweden (r = .82), Bulgaria (r =.81), Poland (r = .81), and Denmark (r = .81). Again, merging the dimension significantly decreased model fit (the U.K.: χ2df = 1 = 338.532, p <.001; Sweden: χ2df = 1 = 243.755, p < .001; Bulgaria: χ2df = 1 = 271.114, p <.001; Poland: χ2df = 1 = 305.021, p <.001; Denmark: χ2df = 1 = 232.684, p <.001). So although we find a high correlation in these countries, we items do measure separate constructs. 4Based on the one-group factor solution, we constructed individual scores. The aim of this article, however, is not to compare the level of support on these different dimensions across time (Waves 1 and 2) or space (different countries) but merely to show the dimensional structure fits across time and space (and predict why it fits even better in some countries), and we show these the descriptive statistics only in Supplementary Appendix Tables A4 and A5). 5The flipping of the sign between TLI on the one hand and RMSEA and SRMR on the other is to be expected, as higher values of TLI indicate better fit, while for RMSEA and SRMR better fit is represented by lower values. References Abbarno A. J., Zapryanova G. M. ( 2013). Indirect effects of Eurosceptic messages on citizen attitudes toward domestic politics. JCMS: Journal of Common Market Studies , 51, 581– 597. doi:10.1111/jcms.12025 Google Scholar CrossRef Search ADS   Anderson C. J. ( 1998). When in doubt, use proxies: Attitudes toward domestic politics and support for European integration. Comparative Political Studies , 31, 569– 601. doi:10.1177/0010414098031005002 Google Scholar CrossRef Search ADS   Anderson C. J., Reichert S. ( 1995). Economic benefits and support for membership in the E.U.: A cross-national analysis. Journal of Public Policy , 15( 3), 231– 249. doi:10.1017/S0143814X00010035 Google Scholar CrossRef Search ADS   Armingeon K., Ceka B. ( 2014). The loss of trust in the European Union during the great recession since 2007: The role of heuristics from the national political system. European Union Politics , 15( 1), 82– 107. doi:10.1177/1465116513495595 Google Scholar CrossRef Search ADS   Bakker B. N., de Vreese C. H. ( 2016). Personality and European Union attitudes: Relationships across European Union attitude dimensions. European Union Politics , 17( 1), 25– 45. doi:10.1177/1465116515595885 Google Scholar CrossRef Search ADS   Boomgaarden H. G., Schuck A. R. T., Elenbaas M., de Vreese C. H. ( 2011). Mapping EU attitudes: Conceptual and empirical dimensions of Euroscepticism and EU support. European Union Politics , 12, 241– 266. doi:10.1177/1465116510395411 Google Scholar CrossRef Search ADS   Braun D., Tausendpfund M. ( 2014). The impact of the Euro crisis on citizens’ support for the European Union. Journal of European Integration , 36, 231– 245. doi:10.1080/07036337.2014.885751 Google Scholar CrossRef Search ADS   Bruter M. ( 2003). Winning hearts and minds for Europe: The impact of news and symbols on civic and cultural European identity. Comparative Political Studies , 36, 1148– 1179. doi:10.1177/0010414003257609 Google Scholar CrossRef Search ADS   Carey S. ( 2002). The dynamics of public opinion toward European integration, 1973-93. European Journal of International , 3( 4), http://journals.sagepub.com/doi/abs/10.1177/1465116502003004001 Dalton R. J., Eichenberg R. ( 1998). Citizen support for policy integration. In Sandholtz W., Sweet A. S. (Eds.), European integration and supranational governance  (pp. 250– 282). Oxford: Oxford University Press. doi:10.1093/0198294646.001.0001 Google Scholar CrossRef Search ADS   de Vreese C. H. ( 2017). How changing conditions make us reconsider the relationship between immigration attitudes, religion, and EU attitudes. European Union Politics , 18 ( 1), 137– 142. http://dx.doi.org/10.1177/1465116516680763 Google Scholar CrossRef Search ADS   de Vreese C. H., Azrout R., Möller J. ( 2016a). Cross road elections: Change in EU performance evaluations during the European Parliament Elections 2014. Politics and Governance , 4( 1), 69– 82. doi:10.17645/pag.v4i1.462 Google Scholar CrossRef Search ADS   de Vreese C. H., Azrout R., Möller J. ( 2016b). Netherlands: 2014 EP voting patterns: from Euphile to euroskeptic. In Hassing Nielsen J., Franklin M. (eds) The Eurosceptic 2014 European parliament elections  (pp 149– 169). London: Palgrave. de Vreese C. H., Boomgaarden H. G. ( 2005). Projecting EU referendums: Fear of immigration and support for European integration. European Union Politics , 6( 1), 59– 82. doi:10.1177/1465116505049608 Google Scholar CrossRef Search ADS   de Vries C. E. ( 2007). Sleeping giant: Fact or fairytale? How European integration affects national elections. European Union Politics , 8, 363– 385. doi:10.1177/1465116507079546 Google Scholar CrossRef Search ADS   de Vries C. E., Hobolt S. B. ( 2016). EU issue voting in national and European Parliamentary elections. In van der Brug W., de Vreese C. H. (Eds.), (Un)intended Consequences of EU Parliamentary Elections  (pp. 101– 124). Oxford: Oxford University Press. Google Scholar CrossRef Search ADS   Easton D. ( 1975). A re-assessment of the concept of political support. British Journal of Political Science , 5, 435– 457. doi:10.1017/S0007123400008309 Google Scholar CrossRef Search ADS   Eichenberg R. C., Dalton R. J. ( 2009). Europeans and the European community: The dynamics of public support for European integration. International Organization , 47, 507– 534. doi:10.1017/S0020818300028083 Google Scholar CrossRef Search ADS   Gabel M. ( 1998). Public support for European integration: An empirical test of five theories. The Journal of Politics , 60, 333– 354. Retrieved from http://www.jstor.org/stable/2647912 http://dx.doi.org/10.2307/2647912 Google Scholar CrossRef Search ADS   Gabel M., Anderson C. J. ( 2002). The structure of citizen attitudes and the European political space. Comparative Political Studies , 35, 893– 913. doi:10.1177/0010414002035008002 Google Scholar CrossRef Search ADS   Gabel M. J., Palmer H. D. ( 1995). Understanding variation in public support for European integration. European Journal of Political Research , 27( 1), 3– 19. doi:10.1111/j.1475-6765.1995.tb00627.x Google Scholar CrossRef Search ADS   Hallin D. C., Mancini P. ( 2017). Ten years after Comparing Media Systems: What have we learned? Political Communication , 34, 155– 171. Google Scholar CrossRef Search ADS   Harteveld E., van der Meer T., de Vries C. E. ( 2015). In Europe we trust? Exploring three logics of trust in the European Union. European Union Politics , 14, 542– 565. doi:10.1177/1465116513491018 Google Scholar CrossRef Search ADS   Haverland M., de Ruiter M., van de Walle S. ( 2015). Producing salience or keeping silence? An exploration of topics and non-topics of Special Eurobarometers. LEQS Paper No. 88. Retrieved from European Institute website: http://www.lse.ac.uk/europeanInstitute/LEQS%20Discussion%20Paper%20Series/LEQSPaper88.pdf Hobolt S. B. ( 2009). Europe in question: Referendums on European integration . Oxford: Oxford University Press. Google Scholar CrossRef Search ADS   Hobolt S. B. ( 2014). Ever closer or ever wider? Public attitudes towards further enlargement and integration in the European Union. Journal of European Public Policy , 21, 664– 680. doi:10.1080/13501763.2014.897746 Google Scholar CrossRef Search ADS   Hobolt S. B., Brouard S. ( 2010). Contesting the European Union? Why the Dutch and the French rejected the European constitution. Political Research Quarterly , 64, 309– 322. doi:10.1177/1065912909355713 Google Scholar CrossRef Search ADS   Hobolt S., de Vries C. ( 2016). Turning against the Union? The impact of the crisis on the eurosceptic vote in the 2014 European Parliament elections. Electoral Studies , 44, 504– 514. doi:10.1016/j.electstud.2016.05.006 Google Scholar CrossRef Search ADS   Hobolt S. B., Wratil C. ( 2015). Public opinion and the crisis: The dynamics of support for the euro. Journal of European Public Policy , 22, 238– 256. doi:10.1080/13501763.2014.994022 Google Scholar CrossRef Search ADS   Hooghe L., Marks G. ( 2005). Calculation, community and cues public opinion on European integration. European Union Politics , 6, 419– 443. doi:10.1177/1465116505057816 Google Scholar CrossRef Search ADS   Hooghe L., Marks G. ( 2007). Sources of euroscepticism. Acta Politica , 42, 119– 127. doi:10.1057/palgrave.ap.5500192 Google Scholar CrossRef Search ADS   Hooghe L., Marks G. ( 2009). A postfunctionalist theory of European integration: From permissive consensus to constraining dissensus. British Journal of Political Science , 39( 1), 1– 23. doi:10.1017/S0007123408000409 Google Scholar CrossRef Search ADS   Huddy L., Feldman S., Weber C. ( 2007). The political consequences of perceived threat and felt insecurity. The ANNALS of the American Academy of Political and Social Science , 614( 1), 131– 153. doi:10.1177/0002716207305951 Google Scholar CrossRef Search ADS   Karp J. A., Bowler S. ( 2006). Broadening and deepening or broadening versus deepening: The question of enlargement and Europe's ‘hesitant Europeans’. European Journal of Political Research , 45( 3), 369– 390. doi:10.1111/j.1475-6765.2006.00302.x Google Scholar CrossRef Search ADS   Kentmen-Cin C, Erisen C. ( 2017) Anti-immigration attitudes and the opposition to European integration: A critical assessment. European Union Politics , 18( 1), 3– 25. http://dx.doi.org/10.1177/1465116516680762 Google Scholar CrossRef Search ADS   Kriesi H., Grande E., Lachat R., Dolezal M., Bornschier S., Frey T. ( 2008). West European politics in the age of globalization  (pp. 154– 182). Cambridge: Cambridge University Press. Google Scholar CrossRef Search ADS   Kuhn T., Stoeckel F. ( 2014). When European integration becomes costly: The euro crisis and public support for European economic governance. Journal of European Public Policy , 21, 624– 641. doi:10.1080/13501763.2013.867892 Google Scholar CrossRef Search ADS   Lerner J. S., Keltner D. ( 2001). Fear, anger, and risk. Journal of Personality and Social Psychology , 81( 1), 146– 159. doi:10.1037/0022-3514.81.1.146 Google Scholar CrossRef Search ADS PubMed  Lubbers M. ( 2008). Regarding the Dutch ‘nee’ to the European constitution: A test of the identity, utilitarian and political approaches to voting ‘no’. European Union Politics , 9( 1), 59– 86. doi:10.1177/1465116507085957 Google Scholar CrossRef Search ADS   Lubbers M., Scheepers P. ( 2005). Political versus instrumental euro-scepticism: Mapping scepticism in European countries and regions. European Union Politics , 6, 223– 242. doi:10.1177/1465116505051984 Google Scholar CrossRef Search ADS   Lubbers M., Scheepers P. ( 2010). Divergent trends of euroscepticism in countries and regions of the European Union. European Journal of Political Research , 49, 787– 817. doi:10.1111/j.1475-6765.2010.01915.x Google Scholar CrossRef Search ADS   Maier M., Maier J., Baumert A., Jahn N., Krause S., Adam S. ( 2015). Measuring citizens’ implicit and explicit attitudes towards the European Union. European Union Politics , 16, 369– 385. doi:10.1177/1465116515577454 Google Scholar CrossRef Search ADS   McLaren L. M. ( 2002). Public support for the European Union: Cost/benefit analysis or perceived cultural threat? Journal of Politics , 64, 551– 566. doi:10.1111/1468-2508.00139 Google Scholar CrossRef Search ADS   Marquart F., Goldberg A., van Elsas E., Brosius A., de Vreese C.H. ( 2017). Knowing is not loving: media effects on knowledge about and attitudes toward the EU. Under review. Moravcsik A. ( 1991). Negotiating the Single European Act: national interests and conventional statecraft in the European Community. International organization , 45( 1), 19– 56. doi:10.1017/S0020818300001387 Google Scholar CrossRef Search ADS   Sánchez-Cuenca I. ( 2000). The political basis of support for European integration. European Union Politics , 1, 147– 171. doi:10.1177/1465116500001002001 Google Scholar CrossRef Search ADS   Schuck A. R. T., de Vreese C. H. ( 2006). Between risk and opportunity: News framing and its effects on public support for EU enlargement. European Journal of Communication , 21( 1), 5– 32. doi:10.1177/0267323106060987 Google Scholar CrossRef Search ADS   Schuck A. R. T., de Vreese C. H. ( 2008). The Dutch no to the EU constitution: Assessing the role of EU skepticism and the campaign. Journal of Elections, Public Opinion & Parties , 18( 1), 101– 128. doi:10.1080/17457280701858656 Google Scholar CrossRef Search ADS   Schuck A. R. T., Xezonakis G., Elenbaas M., Banducci S. A., de Vreese C. H. ( 2011). Party contestation and Europe on the news agenda: The 2009 European parliamentary elections. Electoral Studies , 30( 1), 41– 52. doi:10.1016/j.electstud.2010.09.021 Google Scholar CrossRef Search ADS   Thorson E. ( 2016). Belief echoes. The persistent effects of corrected misinformation. Political Communication , 33, 460– 480. doi: org/10.1080/10584609.2015.1102187 http://dx.doi.org/10.1080/10584609.2015.1102187 Google Scholar CrossRef Search ADS   van Klingeren M., Boomgaarden H. G., de Vreese C. H. ( 2013). Going soft or staying soft: Have identity factors become more important than economic rationale when explaining Euroscepticism? Journal of European Integration , 35, 689– 704. doi:10.1080/07036337.2012.719506 Google Scholar CrossRef Search ADS   © The Author(s) 2018. Published by Oxford University Press on behalf of The World Association for Public Opinion Research. All rights reserved.

Journal

International Journal of Public Opinion ResearchOxford University Press

Published: Feb 22, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off