1
Do political extremists oppose all aspects of European integration?
The moderating role of media use in the Netherlands.
Mark Goedknegt s2376121 First reader: Dr. D.D. Toshkov Second reader: Dr. B. J. Carroll
Leiden University Public Administration - IEG
A thesis submitted for the degree of Master of Science
2
Acknowledgements
I want to thank several people who have helped me during this thesis project. Without them I would not have been able to finish the master’s program in Public Administration. First, to Dr. D.D. Toshkov for his insights and valuable feedback which motivated me to keep going. I very much appreciate you helping me and I look back on a satisfying collaboration. And to Dr. B. J. Carroll for acting as a second reader, thank you. I also want to thank my parents Nienke and Jan-Willem and my brother Bart who were there every step of the way. And to my dear friends (you know who you are) for not calling me when I had to concentrate, and for calling me when I didn’t want to concentrate. It was not easy finishing this degree alongside a master’s program at Rotterdam School of Management, but I’m glad I did it anyways. I look forward to life after graduate school, wherever that may be, and I hope the research contributes to a better understanding of European integration. Enjoy the read.
3
Summary
What is the impact of media use on the relationship between self-reported political ideology and support for European integration in the Netherlands? In order to answer this question, I conduct 4 binary logistic regressions based upon the Eurobarometer survey 89.1. This multidimensional approach to the concept of European integration produces some interesting results. In the Netherlands, citizens who position themselves on the extreme right of the political spectrum trust European institutions, support European policy proposals, and identify with the European Union significantly less than political moderates. Citizens on the extreme left, however, also distrust the EU representative bodies, net of control variables such as gender, age and education. There is no evidence for significant moderation effects of media use, however. Yet, use of new media sources, such as social networks and websites, is associated with lower support for European integration. These findings contribute to the understanding of public opinion on European integration, and they set the stage for further analyses of the role of media use.
4
List of Abbreviations
EU
European Union
EC
European Commission
EP
European Parliament
PCA
Principal Component Analysis
FvD
Forum voor Democratie
5 Table of Content
1. Introduction………
1.1 Why this research?... 1.2 Academic relevance……….. 1.3 Societal relevance……….. 1.4 Eurobarometer 89.1………... 1.5 Outline………... 8 8 10 11 11 11 2. Literature………. 2.1 European integration……….. 2.2 Media impact……… 2.3 Traditional and new media……… 2.4 Media use in the Netherlands……….. 2.5 Political self-placement……….. 2.6 Media use and political self-placement………...
12 12 14 16 16 17 18 3. Theory……….. 3.1 European integration……….. 3.2 The dimensions of European integration……… 3.3 Political self-placement and one’s attitude towards European integration……… 3.4 Media use and one’s attitude towards European integration……….. 3.5 The moderating role of media use………..
19 19 20 21 23 24
6
4. Research Design……….…………..
4.1 Data collection and source……… 4.2 Population of interest……….. 4.3 Why a Principal Component Analysis……….. 4.4 Results of Principal Component Analysis………. 4.5 Operationalization of variables………..
4.5.1 Don’t know category………..………
4.5.2 Dependent variables……….……….
4.5.3 Reliability scales dependent variables……….. 4.5.4 Independent variable……….. 4.5.5 Moderating variable………... 4.5.6 Interaction variable………. 4.5.7 Control variables………. 4.6 Statistical analyses………... 26 26 26 27 28 29 29 30 32 33 26 33 34 36 5. Results……….
5.1 Dependent variable: EU_Representation……… 5.2 Dependent variable: EU_Policy……… 5.3 Dependent variable: EU_Identity……….. 5.4 Dependent variable: Preferred_Speed of European integration………. 5.5 Hypotheses……… 36 36 40 43 46 49 6. Conclusion………..
6.1 The research question……… 6.2 Limitations………. 6.3 Future Research……….. 50 50 52 53
7 7. Appendix and tables………. Table 1 PCA……….. Table 2 Cronbach & Omega……….. Table 3 Descriptives……… Table 4 EU_Representation regression……… Table 5 EU_Policy regression………... Table 6 EU_Identity regression………. Table 7 EU preferred_speed regression……….………. Table 8……….….. Table 9………... Table 10………. Figure 1 Causal flows……….. Figure 2 Plot EU_integration……….. Figures 3 Model EU_integration……….… Figures 4 Plot EU_Policy……… Figure 5 Model EU_Policy………... Figure 6 Plot EU_Identity………. Figure 7 Model EU_Identity………. Figure 8 Plot EU_preferred_speed……… Figure 9 Model EU_preferred_speed ………... Figures 10 till 20..…….……… ………... Formula 1 Interaction………..…. Advisory Report………...……… Appendix 1 Survey………...……… Appendix 2 Robustness check EU_Representation……….. Appendix 3 Robustness check EU_Policy……….. Appendix 4 Robustness check EU_Identity………. Appendix 5 Robustness check EU_preferred_speed………. Appendix 6 till 16 frequencies……… Appendix 17 Expenditure political parties media……… Syntax……….……… 55 29 32 35 39 43 45 47 39 49 49 25 38 38 41 41 44 44 48 48 62 34 55 56 57 57 58 58 59 62 67 8. References………... 63
8
1. Introduction
1.1 Why this research?
The rise of nationalistic politics within the European Union (Bieber, 2018) and the increase of discontent among Europeans (Galston, 2018) forces one to wonder whether liberal democracy and the European dream (Van Middelaar, 2013) will remain a viable vision in the future. What is the optimal trajectory and what do Europeans think about current policies? Other influences such as geographical expansions, technological developments, migration flows, and economic crises increase the problematic need for Europe to continuously reinvent itself. After all, these influences determine the direction and the degree of European integration, meaning the process of merging states within Europe (Van Middelaar, 2013), as well as people’s outlook on their European membership.
And within this perspective, the role of the media as a source of support or opposition to European integration has become fundamental (Conti and Memoli, 2016). Not only is it increasingly acting as a cue to underlying values and interests of Europeans; it’s also increasingly being used as a source for political information (Hooghe and Marks, 2005, 2008; Conti and Memoli, 2016). This new trend, influencing the public’s opinion, has great potential to steer the course of European integration. But it also has great potential to have a negative impact on the relation between Europeans and their attitude toward the EU. As such, this empirical research examines citizens’ attitudes in the Netherlands in 2018 with the help of the most recent Eurobarometer survey to answer the following research question: What is the impact of media use on the relation between political self-placement and European support within the Netherlands?
Despite its increasing influence on European public opinion, the media’s absence from the evolutionary trend of European support and opposition, and scholarly debate, has led to unanswered questions about its true impact on individuals. Will the use of internet with its low start-up costs and anonymity, increase one’s likelihood to support political extremism (Wodak,
9 KhosraviNik, and Mral, 2013)? And will that result in a progressive or conservative political ideology? As a result of these unanswered questions, the European Commission (EC) is still devoting itself to provide and gather information via different media sources. Not only to improve its legitimacy amongst the public, but also to gather valuable data on public opinion which is of pivotal importance for the EU’s future (Hooghe and Marks, 2009).
Even so, the EC’s ability to be covered by journalists and editors remains a twenty first century obstacle (Bijsmans and Altides, 2007). Thus bridging the gap with the public remains difficult as well. And it’s not only traditional media sources (television, radio and press) the EC is trying to reach. Given that Europeans’ weekly use of ‘new media’ as a source of information (the internet and social networks) has increased by 8 percent between 2011-2014, and global access to the internet has increased from 1 billion users in 2005 to 3.2 billion in 2015; the future of the EC’s information provision is shifting (Conti and Memoli, 2016). More attention must be given to online media sources. Thus, further exploring the influence of the media on one’s attitude towards European integration is of pivotal importance not only for scholars but also for the future of European integration.
Research after the 1980s explore the factors that influence one’s attitude towards European integration vigorously (Gabel, 1998; Pierson, 1996; Hooghe, Marks, and Wilson 2002; Hooghe and Marks, 2005, 2009; Hobolt and Wratil, 2015; Schimmelfennig, Leuffen, and Berthold, 2015; Katsourides, 2016; Conti and Memoli, 2016). One of the research results is that citizens who support far left or far right parties are in general less supportive of European integration (Gabel, 1998). Another contemporary result is that citizens who favor new media as a source for political information are in general more inclined to oppose European integration (Conti and Memoli, 2016). However, these two results have yet to be combined in order to explore how the use of media moderates the relation between political self-placement and one’s attitude towards European integration. A remarkable scholarly shortcoming, given that McCombs and Shaw (1972) mentioned almost 5 decades ago that one interprets media differently based on one’s individual characteristics, one of them being political self-placement.
10 As such, the EC and political extreme media sources, try to influence citizens’ attitudes towards European integration ever still, be it to stay on or divert from its intended path.
1.2 Academic relevance
In the Netherlands, where the use of new media is the absolute highest in all of Europe, citizens are becoming more Eurosceptic where once they were one of the biggest supporters (Conti and Memoli, 2016). For example, a political victory for the Dutch Eurosceptic party ‘Forum voor Democratie’ is reason to believe that support for European integration is losing momentum in the country (Conti and Memoli, 2016, p.56; Kranenburg, 2019) and reason to believe that the media does indeed have a profound impact on citizens’ attitudes towards European integration. After all, this party was one of the most active campaigners on social networks (Van Weezel and Dirks, 2018). Thus, to explain this unexpected political shift, one must understand which type of media use, leads to a particular stand on European integration.
Apart from this political occurrence, the direct influence of media use in interaction with political self-placement and one’s attitude towards European integration, has hardly been researched, with exception of some scholars (De Vreese, 2007; Conti and Memoli, 2016). Those acting on behalf of the exception (Hooghe and Marks, 2009) do argue, for example, that cues have the potential to replace one’s lack of knowledge about European political matters. Cues, either from one’s position on the left/right political scale or instead originating from the media could, in their opinion, influence one’s attitude (p.10). There is actually evidence for both the media and political ideology to influence one’s attitude (Gabel, 1998; Conti and Memoli, 2016). But as mentioned, the two have been researched separately.
This research, in contrast, expects that political ideology and media use stand in an underexplored relation to one another, interacting as variables possibly even strengthening each other’s influence. Hence, this research will explore rigorously how media use influences a Dutch citizen’s attitude towards European integration, whilst focusing on one’s political affiliation and confounding for the most relevant pre-existing characteristics.
11 1.3 Societal relevance
Alongside this academic gap, the Dutch and European society can also benefit from a more robust theory on public opinion since public opinion is becoming more dynamic (Hobolt and Wratil, 2015), constrained by dissensus (Hooghe and Marks, 2009), and polarized via social media acting as a source of political information (Toshkov, Mazepus, and Dimitrova, 2018). In other words, Europeans are increasingly involved in politics. Whilst social media increases Europeans’ awareness of supranational policies. Furthermore, the support for European integration has declined whilst the rise of new media outlets has increased (Conti and Memoli, 2016). Society thus needs to become aware of the possible effects their media use has on their attitude toward the EU.
1.4 Eurobarometer 89.1
In order to answer this question, this research makes use of the standard Eurobarometer Survey 89.1 which is part of a series of cross-national longitudinal studies collected by TNS opinion on behalf of the European Commission. The period of analysis is 2018, thereby exploring a more recent trend of Dutch attitudes towards European integration than Conti and Memoli (2016). Trends within the current and potential member-states have been analyzed since 1972, resulting in one of the largest datasets on European public opinion in the world. Annual interviews are conducted with new respondents, twice a year, and face-to-face. With a size of roughly 1000 respondents per country its external validity is optimal. As such, this research aims for the most reliable research results.
1.5 Outline
The outline of this paper is structured as follows: First, I review the literature to explore what is already known, agreed upon, and where disagreements are evident in relation to the research question. After this, the theory describes the main concepts and implications upon which I formulate hypotheses. Then, the data design is explained and operationalized in such
12 a manner that statistical analysis can be conducted, empirically. The results from the logistic regressions will answer the hypotheses after which the conclusion gives answer to the research question. In the next segment, the most relevant existing literature will be described.
2. Literature
2.1 European integration
In the past, researchers have rigorously explored citizens’ attitudes towards European integration. Fueled by a change of “permissive consensus” (Lindberg and Scheingold, 1970) to “constraining dissensus” (Hooghe and Marks, 2009), meaning that citizens are getting more involved and pronounced, the EU’s need to explore the public’s opinion and to gain public support has grown. Reasons for this increased need for support lie in people’s perception of their nation within the EU. They often view their nation to be an island of certainty in a sea of uncertainty (Van Middelaar, 2013, p.310). In other words, citizens no longer trust the political elite, the European Commissioners, to decide what happens for their own country. And thus they oppose integrating national cultures, policies, and even laws to a supranational level (Hooghe and Marks, 2009). As a result, researchers have searched for and have given numerous macro-, meso-, and micro-level reasons that shape one’s political ideology (Hooghe and Marks, 2009). These reasons are necessary for the political elite to understand in order to gain back the trust the citizens’ once had in EU decision-makers.
One of those reasons is that first of all, European integration issues are increasingly being politicized (Hooghe and Marks, 2008). National parties target European issues to empathize with citizens’ uncertainties, who as result follow the cues from that national party to shape a particular attitude towards European integration (Hooghe and Marks, 2008, p.18). Often this politicization is fueled by new media sources such as the internet and social networks, thereby enabling national parties to reach a large group of citizens with relative ease (Conti and Memoli, 2016). Consequently, the ability to influence citizens’ attitudes on European integration has increased.
13 Second, the economic utility of a nation’s European membership, and one’s level of identity with his or her home country, significantly determine whether or not one adopts a supportive attitude towards European integration (Risse, 2010; Hooghe and Marks, 2008). The more one views the EU to be economically fruitful, and the less one identifies with his or her home country, the more European support can be expected (Risse, 2010; Hooghe and Marks, 2008). Hooghe and Marks (2009) theorize that these determinants underly one’s preference on supranational policies. They label it as a post functionalistic theory, meaning that even though institutionalized cooperation among member-states is solving transnational problems; one’s level of national identity and perception of the economic utility largely determine whether one favors European integration in the end.
Third, other individual and political values have also been assigned to influence one’s attitude significantly. For example, post materialists (often labelled as liberals) tend to favor European support, in contrast to materialists who focus more on the economic utility (Gabel, 1998). Thus, aggregated by politicization, pre-existing characteristics have been proven to shape one’s attitude towards European integration ever since the late 1980s, and with increasing influence.
Even so, the history of European integration shows that one’s support is flexible, unpredictable, and thus not easy to pre-determine. Take for example the global financial crisis and the subsequent euro area crisis (Hobolt, Wratil, 2014). It resulted in many Europeans acknowledging the importance of Europe’s economic usefulness, given that it was a collaborative effort of member-states and supranational institutions that proved to be the answer to the crisis (Hobolt, Wratil, 2014). Yet, several years later, the European immigration crisis changed that assumption. Where the economic utility of European membership largely determined one’s support during the Euro crisis, now during the immigration crisis, one’s identity with his or her home country, largely predicted one’s support for European integration (Leruth, Startin, and Usherwood, 2017). People’s opinion on European integration thus vary under different circumstances making prediction of individual support difficult.
14 A cause for concern given that the European Commission’s reasons to integrate during Europe’s founding years in the 1950s, via for example the Coal and Steel Treaty of Paris in 1951, were to avoid future global wars and to strengthen economic prosperity among all member states (Van Middelaar, 2016). Yet, they did this at times of permissive consensus, and thus the foundation of European integration was never agreed upon by the public (Hooghe and Marks, 2008) but agreed upon by the political elite. Now that the public opinion is more important than ever, further deepening European integration is subject to scrutiny and opposition that are fueled by the media. The media being a source of information with insidious impact on the freedom and decision-making power of those at the top of the political pyramid (Conti and Memoli, 2016). As such, one’s attitude towards European integration is a relative new subject for the European citizen as well as for the European Commission, decreasing the speed of European integration, which historically was less influenceable under permissive consensus (Hooghe and Marks, 2008).
The future of European integration is therefore a continuous reinvention of Europe through the passage of time (Van Middelaar, 2013). It remains to be seen, though, whether the influence of the media will take complete control of its course. The politicization of European matters, together with the spread of technology makes it, as mentioned, difficult for European institutions to remain legitimate and authoritative in the eyes of citizens. Technological disruption is, as a matter of fact, increasingly capable of disrupting institutional set-ups and the way of life as we know it (Harari, 2018). Being aware of how technology, in the form of media information provision, is shaping the future of the European Union is therefore of pivotal importance. How, then, is the media impacting European integration according to existing research?
2.2 Media impact
Existing research argues that the influence of media use on one’s attitude towards Europe is subject to a wide array of pre-existing motivations that are increasingly being
15 exposed to the influence of the media; from which citizens gather the appropriate level of knowledge to form an attitude towards European integration (Conti and Memoli, 2016). In view of the fact that the media landscape as an information environment is such a polarized source of political information and debates on European policies, citizens are often confused and incentivized to oppose European integration (Hooghe and Marks, 2005; Azrout, Van Spanje, and De Vreese, 2012; Toshkov, Kortenska, Dimitrova and Fagan, 2014). For example, Katsourides (2016) researched why in 10 years’ time the Cyprus citizens changed their attitude from viewing the EU as a panacea to a sudden cause for concern. Results revealed that the dynamic between print media and the public had reinforced Euroscepticism because a critical national lens was used exclusively to evaluate matters of international scale. Thereby a consistently visible one-sided negative news coverage changed the public’s opinion (De Vreese and Boomgaarden, 2006). Hence, the media’s influence on one’s attitude towards European integration is significant; capable of cueing and framing information with insidious impact on the future of European integration.
Furthermore, when categorized as new and traditional media (Conti and Memoli, 2016); media can have a positive or a negative impact on one’s attitude towards European integration. Even though the fundamental task of the media is to make politics transparent, it often relies on insufficient information (e.g. behind closed doors meeting of the European Central Bank) or an overflow of information; resulting in the media selecting which information is made available to the public (Trenz, 2008). This way, the media emphasize what information citizens gather via priming and from which perspective it’s given via framing (Semetko, De Vreese, and Peter, 2000).
As such, citizens who gather their political information primarily from new media sources (e.g. internet and social networks) are faced with an overflow of polarized information resulting in more Eurosceptic attitudes (Conti and Memoli, 2016). Those that gather their political information from traditional media sources (e.g. television, radio, and press) are, according to research so far, positively impacted by the news resulting in a general support
16 for European integration (Conti and Memoli, 2016). Thus, the fact that new media is increasing is a challenge to promote European integration.
2.3 Traditional and New Media
Reasons for the new media to be more capable of polarizing and to increase opposition to EU integration lie in part within the history of both new and traditional media. On the one hand traditional media sources such as the television, written press and radio, have been around for quite some time. These sources have narrated the process of European politics prior to occurrences such as the Maastricht Treaty in 1992 which gave room for these media sources to control citizens’ attitudes accordingly. Thus, generations are used to these media sources and prefer them to more modern social media sources.
New media, on the other hand, has been catapulted by the 300 percent increase of internet users between 2005-2015 (Conti and Memoli, 2016) reaching far more citizens than traditional media is capable of. This forces the EC to deal with a challenging information environment now and in the future, which as of yet the institution has insufficient control over. This lack of control is evident by the research results of Fletcher et al., (2018) who revealed the following on online disinformation by new media sources: “In France, one false news outlet generated an average of over 11 million interactions per month—five times greater than more established news brands
” (p.2). Thus, Europeans who use new media are exposed to an
information environment that is at risk of being less reliant than traditional media sources.
2.4 Media use in the Netherlands
In this European context, the use of new media in the Netherlands is the highest of all member-states (Conti and Memoli, 2016). One can therefore assume that a Dutch citizen opposes European integration based on his or her media use alone. Meaning when all other variables are held constant. Yet, Conti and Memoli (2016) argue that the public’s attitude is in
17 fact supportive of European integration when other micro-level variables are accounted for. Other researchers agree with the general supportive attitude of Dutch citizens (Toshkov et al., 2014). An interesting finding, given that in 2005 the Dutch voted against a constitution for Europe and the ‘Forum voor Democratie’ party has gained momentum on the far-right political landscape (Adam and Maier, 2011). Being aware of the fact that the Netherlands has strict regulations on cross-ownership of press and broadcasting which makes mass-scale traditional media biases and manipulation unlikely (Semetko et al., 2000); it’s nonetheless reason for concern that so many Dutch citizens use new media as their source for political information. Will this lead to a Eurosceptic country in the future?
Moreover, Machill, Beiler and Fischer (2006) argue that the media landscape of the Netherlands is Europeanized, in other words, the Dutch media devotes sufficient amount of coverage to EU matters and are not using a national lens exclusively. But these are controlled information environments with strict regulation in contrast to the internet and social networks in which fake news and easy access increase the odds of one taking on a Eurosceptic attitude. In fact, these ‘new’ media sources are often times beyond the control of European legislation (Conti and Memoli, 2016). Thus, the influence of new media in the Netherlands is reason to believe that the unexpected victory of the far-right party ‘Forum voor Democratie’ is the result of this consistently high and increasing use of new media (Kranenburg, 2019), and reason to believe that more uncontrolled and unexpected political results are still to come. How about the political self-placement as an influencer on one’s attitude?
2.5 Political self-placement
In light of the far-right victory of ‘Forum voor Democratie’ in the Netherlands, research argues that apart from the influence of the media on public opinion, one’s political self-placement is (among others) a significant influencer of one’s attitude towards European integration (Hooghe, Marks, and Wilson, 2002; Van Elsas, Hakhverdian, Van der Brug, 2016). Scholars argue that the most Eurosceptic citizens also place themselves in the radical left and
18 radical right corners of European politics (Lubbers and Scheepers, 2010). For example, and as mentioned, the Dutch voted against a constitution for Europe in 2005. At the time this was advocated by radical socialist parties on the left and radical nationalistic parties on the right (Adam and Maier, 2011; Elsas et al., 2016). Such parties have been identified as having strong effects on the attitude of citizens (Carey and Burton, 2004). Thus scholars argue that not only the use of media, but also political self-placement, influences one’s attitude towards European integration significantly.
Furthermore, reasons for citizens to oppose European integration differ between left and right. Existing research has focused on party ideologies and motivations to be Eurosceptic and has often assumed that these would run parallel to the citizens that vote for them (Van Elsas et al., 2016). However, empirical results reveal that radical left citizens are less supportive of the current European integration course than the radical right, thereby mainly opposing the current welfare state arrangements and liberalization of the market (Van Elsas et al., 2016). When it comes down to the EU’s future course, radical right are more skeptical than the radical left, often arguing that national sovereignty and cultural homogeneity is threatened by supranational governance (Van Elsas et al., 2016). The latter is gaining ground in the Netherlands; in part because austerity programs and migration flows have fueled the lack of trust citizens have in the EU to handle future crises effectively (Brack and Startin, 2015). Thus, even though motivations for radical citizens to oppose European integration differ, they have increased in scope not only in the Netherlands but for Europe as well.
2.6 Media use and political self-placement
There’s a lack of empirical evidence on what the impact is of one’s attitude towards European integration when media use and political self-placement interact. Research on cues from the media (Conti and Memoli, 2016) and cues from political parties (De Vries and Edwards, 2009) have often been researched separately with the purpose of determining one’s attitude towards European integration. Even though there’s a general agreement among
19 scholars that individuals react and interpret media differently based on individual characteristics and pre-existing motivations (McCombs and Shaw, 1972); the interaction of these two variables have been neglected nonetheless (Inglehart, 1970; Hooghe and Marks, 2004; Azrout and Van Spanje, 2012; Conti and Memoli, 2016). What’s evident, though, is that extreme right and extreme left citizens are more active on the web than moderate citizens (Conti and Memoli 2016). Together with the fact that the EU has been increasingly politicized in the media in general (Conti and Memoli, 2016) and one is left with an unavoidable assumption: the young generations are less capable of forming independent rational attitudes towards European integration because of their relatively high use of new media sources.
However, despite the lack of evidence there have been researchers that do explore political ideology and media use simultaneously. For example, Azrout, Van Spanje, and De Vreese (2012) argue that the relation between anti-immigration feelings and one’s attitude towards Turkey’s EU membership, is strengthened by the use of Eurosceptic traditional media outlets. However, the dependent variable of Azrout et al. (2012) isn’t focused nor detailed enough to evaluate one’s attitude towards European integration accurately and holistically. As such, this research will try to contribute the scholarly debate by interacting political self-placement and media use on one’s attitude towards European integration in the Netherlands, thereby including relevant control variables and conceptualizing a more detailed dependent variable. After this, the theory describes the main concepts and implications upon which relevant hypotheses are formulated.
3. Theory
3.1 European integration
In what is considered by many to be an undisputed theory, one’s attitude vis-à-vis European integration remains a contested concept due to a lack of overall agreement among scholars and different research methods being used. This contestation over the most appropriate concept and research method is problematic. For example, Carey (2002) defines
20 and conceptualizes European integration as the ‘degree to which one perceives European membership to be a good thing’, where in contrast, De Vreese, and Boomgaarden (2005) conceptualize and define European integration against five general attitude and opinion items, such as ‘European integration is being pushed too fast’, and ‘I would be willing to make a sacrifice to help a less strong country’. As a result, the concept of European integration is producing research based on vastly different foundations.
As such I agree with the recent definition of Vasilopoulou (2017) who mentions the following: “one’s attitude towards European integration describes the level of opposition to the process of EU integration and/or various aspects of it” (p.23). It’s indeed a multidimensional
concept and it can be specified into factors or applied as a whole. The underlying theoretical assumption of this research is that one’s media use as well as one’s political self-placement influence different factors of the concept. For example, radical left citizens oppose European integration because they view it as a capitalist project; whereas radical right citizens do so to defend national sovereignty (Hooghe and Marks, 2007). Therefore I expect that the variation in answers within this Eurobarometer survey run parallel to the variation in factors measuring one’s attitude. A one dimensional definition alone is therefore no longer adequate in exploring public opinion on EU integration. Hence, this research assumes that one can support one factor of European integration whilst opposing another. To test this assumption this research conducts a principal component analysis of the Eurobarometer survey to determine if indeed Dutch citizens distinguish different dimensions of the concept, after which relevant variables will be analyzed (see research design p.27).
3.2 The dimensions of European integration
Despite the disagreement within existing research, the need to capture more specific aspects of the integration process (Conti and Memoli, 2016) is becoming evident. Conti and Memoli (2016) suggest that the concept consists of the following factors: “(1) feelings of identification to and attachment for a political community (identity), (2) sentiments towards
21 representative institutions at the EU level (representation) and (3) judgements on governance of collective goods and interests by the EU (policy).” (p.31). In other words, how one feels about European integration and/or aspects of it, can be categorized into these three factors. Due to a lack of consistency in the Eurobarometer surveys, Conti and Memoli (2016) are not able to include the factor of identity. As such, this research will include identity to strengthen their intended approach and to increase the potential of a statistically robust multidimensional concept measuring one’s attitude towards European integration.
It is important to include identity as a dimension of European integration given that the theory of post functionalism emphasizes that national identity can constrain the process of European integration (Hooghe and Marks, 2008). It is, in other words, likely that a Dutch citizen will oppose European integration if he or she mostly or even exclusively identifies as a Dutch citizen instead of a European citizen. The concept of collective identity (Kuhn, 2019) has therefore gained ground in the literature, given that a collective feeling of European identity has become part of a successful European integration trajectory. The factor identity will thus contribute to a more robust multidimensional concept of ‘one’s attitude towards European integration’. See the results section to view the results of the conceptualization analysis.
3.3 Political self-placement and one’s attitude toward European integration.
First of all, existing research suggests that the more politically extreme a citizen feels to be, the more he or she will be skeptical of European integration (Hooghe, Marks and Wilson, 2002; Van Elsas, Hakhverdian, Van der Brug, 2016; Adam and Maier, 2011). Moderates are in general much more inclined to support the status-quo and they have a bigger chance to identify themselves as a European citizen (Hooghe et al., 2002). Not surprising given that citizens with either an extreme left or an extreme right oriented political ideology aren’t represented by their governments and thus seek to restructure the political arena. Yet, despite the logic of it, the situation remains problematic given that, for example, the identification with Europe has decreased significantly in the Netherlands with 5 percent points, to just 42 percent
22 of citizens identifying themselves as a European citizen in 2017 (Eurobarometer, 2017). So, it’s expected that Euroscepticism will increase when the support for extreme parties within the Netherlands continues.
Second, the main goal of European integration is to create and maintain a decades old market-liberal project (Van Middelaar, 2013). Those placing themselves at the far-ends of the political spectrum, often reject the European policies just as much as they oppose the current dominant construction of the EU’s ideology (Hooghe et al., 2002). They favor taking back control of their national sovereignty. Furthermore, the success of EU proposals to increase financial integration, to maintain a single currency, and to increase the amount of member states; mostly depends on the radical right attitudes towards these capitalist projects who oppose them the most (Hooghe and marks, 2005). Together with the radical left who oppose the lack of transparency of supranational governance and one can safely say that the political extreme have been found to be the most skeptical of Brussel’s trustworthiness (Koopmans, 2007, p.184; European Commission, 2017, p.21). This in contrast to moderate citizens who generally express support for European policy and trust in European institutions (Casey, 2002; Azrout, Van Spanje, and De Vreese, 2013). In short, the lack of active representation, the history of a moderate outlook on supranational governance, and the lack of trust in Brussel among the extremists, has led many scholars to label political extreme citizens as being generally Eurosceptic. It remains to be seen, though, whether this applies to a multidimensional European Integration variable as well. On the basis of these findings, this report expects that:
(H1a) = A left extremist Dutch citizen has less chance of supporting all dimensions of European integration than a moderate.
(H1b) = A right extremist Dutch citizen has less chance of supporting all dimensions of European integration than a moderate.
23 3.4 Media use and one’s attitude toward European integration
Previous research has rigorously focused upon social constructivist theories (Risse, 2010). According to these theories, the building of common knowledge and one’s view of reality are socially constructed by external phenomena. A classical research is that of Inglehart (1970) and the cognitive mobilization theory. In definition Inglehart (1970) argues that “mass media’ and the increase of exposure to information increases one’s capacity to interpret and receive messages about remote institutions such as the EU” (p.47). He concludes that rising levels of communications would increase information and therefore positive awareness of European institutions. However, critical contemporary research argues that the exponential increase of global internet use along with the increase of citizens that gather information via new media (internet and social networks) has decreased support for integration in the twenty first century (Conti and Memoli, 2016). Thus, media use has the potential to increase support but also great potential to segregate the European citizens into Eurosceptic and Euro supportive groups.
Furthermore, new media (internet and social networks) has decreased citizens’ trust in European institutions along with support for EU policy (Lilleker et al., 2011; Jungherr, 2014; Conti and Memoli, 2016). First because radical left citizens and right citizens are the most active on the web thereby providing the most information and provoking the most interaction (Lilleker et al., 2011). Second, because radical politicians are a lot more active in new media sources, thereby more often framing information with Eurosceptic ideologies (Jungherr, 2014). Third, because the younger generations are more inclined to access new media sources than older generations; thereby automatically being subject to a more polarized media environment and forming one’s attitude in times of easy access to new media sources (Conti and Memoli, 2016). For example, the recent victory of the Eurosceptic party “Forum voor Democratie” is significantly due to the use of new media by the campaign team (Van Weezen and Dirks, 2018). In short, new media sources have increased the capability of Eurosceptic ideologies to be spread, and for anti-European sentiments to be increased.
24 These expectations are easier to understand when one explores framing-, agenda-setting-, and cue theories that have been developed within the field of public opinion. Cue theory is based on Inglehart’s (1970) cognitive mobilization theory and states that political cues – originating from political elites’ ideologies or media sources – can influence how one shapes an attitude (Hooghe and Marks, 2005). They act as a means to an end; filtering one’s objective attitude before one shapes an opinion. Moreover, Maier and Rittberger (2008) argue that framing – as a theory – means that media can characterize an issue, an event, or an actor; thereby influencing one’s understanding of the subject (p.247). At last, agenda-setting theory of Maxwell, McCombs and Shaw (1972) argues that “the mass media set the agenda for each political campaign, influencing the salience of attitudes toward the political issues” (p.177). Thus, it appears that media significantly shapes the information environment citizens find themselves in. In accordance with these facts, this research expects that the use of new media (internet/ websites and social networks) decreases one’s likelihood of supporting European integration. The following hypothesis will be tested:
(H2) = When one favors new media for European political matters, one’s chance of supporting European integration declines.
3.5 The moderating role of media use
At last, this research will formulate a hypothesis on the moderating role of media use. Since this micro-level tripartition has been neglected by research so far, expectations must rely on careful evaluation of the above mentioned binary relations. Research has argued, though, that the effect of media can be undermined or strengthened by other individual characteristics (Azrout, Van Spanje, and De Vreese, 2012). As such, reliable expectations can still be formulated. For example, extreme politicians and citizens are more active on new media sources, be it to share or receive information (Lilleker et al., 2011; Jungherr, 2014). This research therefore expects the following hypothesis:
25 (H3a) = the chance of a political left extremist to support all dimensions of European integration decreases when new media sources are used for European political information.
(H3b) = the chance of a political right extremist to support all dimensions of European integration decreases when new media sources are used for European political information.
The hypotheses of this research can be displayed in a causal diagram to simplify and illustrate their causal flows. To rule out association instead of causation the research makes use of a large N sample (Eurobarometer 89.1) from which respondents are sampled in a stratified manner (Toshkov, 2016). Also, those variables that prove to be confounding political self-placement and one’s attitude towards European integration will be included. As a result, this research guarantees a strong external and internal validity. Figure 1 displays the causal flows of this research.
Figure 1. Causal flows.
26
4. Research design
4.1 Data collection and source
The unit of analysis are all Dutch citizens being represented by those who are respondents within the Eurobarometer survey 89.1. Within this survey, a Dutch ‘sub-population’ is extracted from the rest of the European respondents. The population of interest was questioned between March 13th 2018 and March 26th 2018 (European Commission, Brussels, 2018). Respondents are interviewed face-to-face and – for the purpose of survey 89.1 – questioned on topics relating to: Attitudes towards the EU, Europe in 2020, the European economy, the European citizenship, the EU budget, the future of the EU, and media use (European Commission, Brussels, 2018). Therefore the survey is suitable for the purpose of this research. The Eurobarometer survey itself has been conducted ever since 1972 on behalf of the EC and is operationalized by TNS opinion. Thereby hypotheses that are tested with this data are reliable representations of the actual European population.
4.2 Population of interest
At the time of questioning in March 2018, the total population of the Netherlands was 17.192.048 (CBS Statline, 2019). With the use of the SELECT IF function the survey has been reduced to a sample of Dutch respondents only (n=1054). As mentioned, the Dutch citizens are analyzed because within the Netherlands the use of new media is the highest and support for extreme political ideologies is gaining ground (Conti and Memoli, 2016). This offers potential to evaluate the interaction effect of political ideology and media use, on one’s attitude towards European integration. The sample group has not been set at a European level since this is beyond the reach of this report and a detailed exploration favors a national scope.
However, it will be interesting for future research to observe cross-national differences. Thereby one can examine whether, besides micro-level characteristics, country-level determinants influence one’s attitude on European integration, and to what extent.
27 4.3 Why a Principal Component Analysis
For the sake of argumentation and exploration, a Principal Component Analysis (PCA) is operationalized to examine whether in today’s Europe and with respect to the whole European population, a multidimensional model is empirically valid to measure citizens’ attitudes towards European integration. This could prove that the concept of ‘European Integration’ needs to be made more multidimensional. Furthermore, one needs to make sure the Dutch respondents even distinguish different dimensions of the concept during their interview, given that theory alone isn’t decisive enough to prove that this could indeed be the case (Conti and Memoli, 2016). For clarification, in case a respondent would not be able to distinguish between being in favor of a ‘common European trade policy’ and ‘one’s level of attachment to Europe’ (the former measures support for European policy and the latter measures one’s level of European identity); a multidimensional concept can’t be used for the purpose of identifying one’s overall attitude. In that case one would not measure separate components of one multidimensional concept. Thus, a PCA is relevant to examine whether the multidimensional approach of this research is valid.
Besides the need for a PCA, this research uses its results to conduct a valid computation of variables that fall under each factor. This because using factor scores obliges the researcher to use linear regression. And linear regression assumes the data to be normally distributed which even after log-transformation is not the case. Second, categorical variables are better suited for logistic regression. All variables to be used are categorical, and thus a PCA would not only take away all of those defined categories, it would also overcomplicate the interpretation. Even so, the PCA can prove the existence of Conti and Memoli’s (2016) multidimensionality with regard to European integration, and thus this research follows that intended scientific approach.
28 4.4 Results of Principal Component Analysis
Table 1 displays which factors (rotation varimax) explain the most variance. This is operationalized on the assumption of an eigenvalue of more than 1, thereby determining the number of factors to validate (Cliff, 1988). The analyses show that a PCA is useful (KMO > 0.5) and the variables themselves are suitable for structure detection (Bartlett < 0.001). The variables are standardized so that former variations in scales don’t bias the results of the PCA. Table 1 displays that the factor Representation corresponds to: trust in the European Commission and trust in the European Parliament. Other relevant variables have been excluded to the point where the rotated PCA analysis displays sufficient similarities among the variables within the factor. As for Identity, this is represented by: attachment to European Union, and attachment to Europe. The same rationality has been used for exclusion of other relevant variables. Policy is measured by seven EU proposals displaying a supranational governance initiative: a single currency, a common foreign policy, a common defence policy, a common trade policy, a common migration policy, a common energy policy and a digital single market.
The results of the PCA reveal that there’s a sufficient amount of clustering and indeed a identification of three different components by the respondents. Besides this, it’s an interesting observation that the European policy proposals ‘future enlargement’ and ‘free movement of citizens’ do not fall under the same factor. One can therefore assume that Dutch citizens distinguish these variables from the others. As such, they have not been included within the analysis.
Alongside this multidimensionality, a one dimensional approach is also included to explore differences within the Dutch society when respondents are asked on factors of European support on the one hand, and a general support for Europe on the other, namely ‘preferred speed of European integration’. As such, the multidimensional logistic regression results can be compared to the one dimensional results. This enables the research to uncover
29 differences when the multidimensional approach instead of the one dimensional approach is used.
Table 1. Principal Component Analysis (Rotated Varimax)
Component
1 2 3 4
EU proposals: Single Currency 0,518 0,162 -0,113 0,066
EU proposals: Common Foreign Policy 0,557 0,186 -0,020 -0,004 EU proposals: Common Defense Policy 0,473 -0,056 0,045 0,094
EU proposals: Common Trade Policy 0,559 0,034 0,035 0,135
EU proposals: Common Migration Policy 0,475 0,003 0,074 0,099 EU proposals: Common Energy Policy 0,656 0,157 -0,071 -0,088 EU proposals: Common Digital Single Market 0,491 0,063 -0,053 -0,053
Attachment to: European Union 0,177 0,255 -0,084 0,841
Attachment to: Europe 0,018 0,091 0,003 0,890
European Parliament - Trust 0,074 0,859 0,015 0,138
European Commission - Trust 0,143 0,859 0,037 0,082
KMO .693
Bartlett’s test of Sphericity .000
* See appendix for related survey questions.
4.5 Operationalization of variables 4.5.1 Don’t know category
Before going over the operationalization of the variables, it’s important to understand the following: those respondents who ‘do not know’ whether they trust European institutions, whether they identify with the EU, or whether they agree with its policies, are not be excluded from the analyses. The answer category is often given to reduce the amount of missing data,
30 but there are scholars wo oppose its use in surveys (De Leeuw, Hox, and Boeve, 2016). In contrast, Luskin and Bullock (2011) argue that the use of ‘don’t know’ answers should not be discouraged. Thus, it’s a controversial debate, but for the purpose of answering the research question it is the best course of action to.
Within this research, those who fall into the don’t know category, need to be merged with those who don’t trust, don’t identify with the EU, or oppose European policies. This is based on the rationale that only explicit support counts and a cross tab analysis reveals that the variables political interest and education don’t skew the data in either direction of support or opposition for those who fill in ‘don’t know’. Furthermore, one must assume the worst: those that don’t know can’t be treated as being in support of the European institutions, at least not to a satisfactory level. Finally, setting don’t know as user missing would decrease the amount of respondents too significantly (30 percent). Thus, for the purpose of this research, they need to be interpreted at the same level as those who oppose.
4.5.2 Dependent variables
The binary dependent variable ‘EU_Representation’ measures whether or not one trusts the EC and European Parliament (EP). The two variables are first recoded separately where 0 = ‘do not trust’, 1 = ‘trust’, and 2 = ‘don’t know’. Then, on the basis of these 3 categories, 2 dummy variables are created whereby 0 = don’t know or do not trust, and 1= trust. They are then combined into one binary variable: EU_Representation. In order to operationalize this, I use the AUTOsum function in Excel and the ADD VARIABLE function in SPSS. Only respondents who trust both institutions are coded as 1. As a result, 509 Dutch citizens don’t trust or don’t know whether they trust the European institutions, whereas 545 do trust the European institutions. The variable has a mean of .520, a SD of .500 and 1052 valid cases within the survey.
Second, the binary variable ‘EU_ Policy’ measuring whether or not one supports European policy is computed from the seven variables measuring one’s attitude towards a
31 European policy proposal. The same rationale applies: only explicit support counts. As such, the variables are first recoded separately. The values are coded where 0 = no, 1 = yes, 2 = don’t know. In contrast to EU_Representation, this variable has 17 missing values set as system missing, given that they refused to answer or are not a Dutch citizen; thus inapplicable for the research. A valid reasoning given that the pattern analyses reveals that the ‘no missing pattern’ is the most observable within the data-set (93.74%) and the little MCAR-test was non-significant meaning the data is missing completely at random (.358). One can observe this in figure 10 (see appendix 18).
Then, the seven dummy variables are combined. In case a respondent agrees with more than 5 proposals, he or she is in support (value 1). In case he or she agrees with 4 or less proposals, he or she is against (value 0). This, because the AUTOaverage function within Excel reveals the mean to be 5.3 and a cutoff point has to be made. Ones again, only explicit and significant identification with the EU counts. As a result, the variable displays 250 respondents opposing further European policy, and 787 being in support of it. The variable has a mean of .760 with a SD of .428.
Third, the nominal variable ‘EU_Identity’ measures whether one identities with the European identity. The variables ‘attachment to the European Union’ and ‘Attachment to Europe’ are first recoded separately where 0 = no, 1 = yes and 2 = don’t know. Those that identify with both variables (identify with Europe and European Union) are the only respondents being given a value of 1. The binary variable ‘EU_Identity thus reveals 601 respondents identifying themselves with the European Union, whereas 453 do not. The variable has a mean of .43 and a SD of .495.
Fourth, the nominal variable ‘Preferred_speed’ measures one’s evaluation of the preferred speed of European integration. Its Likert-scale is made dichotomous where 0 stance for ‘standstill/ lower moderate speed or don’t know’ and 1 for ‘upper moderate/fast speed’. Those who are inapplicable are set as system missing (n=1). The frequency table displays 433 respondents favoring a standstill/ lower moderate future speed of European
32 integration, whereas 620 respondents favoring a faster speed of European integration. The mean is .588 with a SD of .492.
4.5.3 Reliability of dependent variables’ scales
To assess the internal consistency of the above mentioned variables, a Cronbach’s Alpha analysis is conducted alongside the Omega analysis of Hayes and Coutts (2019) using the macro extension within SPSS. This, to not only determine whether a multidimensional model is empirically valid to measure citizens’ attitudes towards the EU, but also to determine whether the variables measure the same construct. See table 2 for the overview of these analyses. Note, the Cronbach’s Alpha of EU_Policy is ‘questionable’ because of the binary nature of the 7 variables included. Even so, the factor analysis and the fact that they are part of the same question within the survey, allows them to be operationalized as they are presented.
Table 2. Internal consistency tests for the dependent variables. Variable
Test EU_Policy EU_Representation EU_Identity
Cronbach's Alpha 0.645 0.769 0.702
Omega 0.652 na na
* na = not applicable 4.5.4 Independent variable
The nominal variable ‘political_extremism’ has been recoded to the point where scores 1 through 3 = 1 (left extremists), 4 through 7 = 0 (Moderates), and 8 through 10 = 2 (right extremists). Moderates are the reference category as the research’ objective is to measure the absolute difference between moderates and extremists. Within the survey, it is measured via one’s political self-placement according to a 10-point scale. Out of the 1054 respondents, 42 are set as missing values since they gave inapplicable answers. With a minimum of 0.42, a maximum and a SD of .65, the variable displays most Dutch citizens to be
33 moderates. One can therefore interpret that the Dutch society is shaped by a U-curve (Hooghe and Marks, 2002) meaning a majority of citizens see themselves as moderates when asked about their political ideology.
4.5.5 Moderating variable
The questions of the survey asking about the respondent’s preferred source of media for European political matters has been chosen as the moderator: Source_EU_news (Conti and Memoli, 2016). After setting those who don’t know or refused to answer as user missing (n=18), a dummy variable is created as follows: the value 1 = new media (websites and social networks) and 0 = traditional media (tv, radio, and written press). By setting the reference category to be traditional media one can research the difference in chance of those preferring new media sources to those preferring traditional sources. The frequency table displays 825 respondents favoring the traditional media for European political information, whereas 211 prefer new media sources. Respondents could only fill out 1 source of preference during the questionnaire. With a mean of .204 and a SD of .403, Dutch citizens prefer traditional media sources over new media for gathering European political information, in contrast to the generally high use of new media sources among Dutch citizens according to Conti and Memoli (2016). It seems that the European Commission is therefore not yet sufficient enough in information provision via new media sources. See Advisory Report in the Appendix.
4.5.6 Interaction variable
To measure the moderation effect of different media sources and their effect on the relation between political extremism and one’s attitude towards European integration, an interaction variable has been computed via the following formula 1:
34 Formula 1. Interaction formula
Interaction = political_extremism * Source_EU_news.
4.5.7 Control variables
First, the control variables age, gender, education, and occupation_status, are included since these could bias the regression results and cause a Type 1 error (Inglehart, 1970; Gabel, 1998; Oesch, 2008; Conti and Memoli, 2016). The Eurobarometer gathers respondents randomly so the power of randomization should eliminate the necessity for other control variables (Toshkov, 2016). However, out of research interests, these are included within the model. Age is a categorical variable where 1 = 15-24 years, 2 = 25-39 years, 3 = 40-54 years and 4 = 55 years or older. Alongside the fact that the variable has a mean of 3.42 and a standard deviation of .855, the frequency results displays that 61.5 percent of the respondents are 55 years or older. As figure 11 displays, this associates with traditional media, meaning that 581 respondent of 55 years or older use traditional media outlets to gather information thereby explaining why in general traditional media is preferred.
Second, the variable gender has been coded where 0 = man and 1 = woman. There are 506 men and 548 women within the survey. Moreover, education as a categorical variable measures at what age one stopped full-time education. After recoding the variable to let it increase when one stopped at a later age with his or her education; the minimal value 1 = 15 years or younger, whereas the maximum value 4 = still studying. With a mean of 2.540 and a standard deviation of .696 this variable displays that most respondents (51.9 percent) were 20 years old or older when they stopped full-time education. A surprising result is that 7.3 percent, or 77 respondents, of the total sample group stopped at the age of 15 years or younger with full-time education; displaying a relatively high level of education within the Netherlands as of March 2018.
35 Finally, the variable occupation as a dichotomous variable contains 2 values. Respondents who answered to be self-employed or employed = 0 whereas those who are not working = 1. With a mean of .496 and a standard deviation of .0.500 it is evident that a significant proportion of the sample group does not work: 523 of the total 1054. Given the fact that 61.5 percent of the respondents are 55 years or older, one should be wary that conclusions based on education are significantly skewed because of this overrepresentation. Table 3 provides an overview of the descriptive statistics. See appendix 6 till 16 for the frequencies.
Table 3. Descriptive Statistics
N Minimum Maximum Mean
Std. Deviation
I support European policy 1037 0 1 0.76 0.43
I trust these European institutions 1054 0 1 0.52 0.50
I identify with the EU 1054 0 1 0.43 0.50
Interaction 1041 0 2 0.08 .35
Preferred speed European Integration 1053 0 1 0.59 0.49
What source for EU pol. matters. 1036 0 1 0.20 0.40
Political_extremism 1012 0 2 .42 .65
Working or not working 1054 0 1 0.50 0.50
Age when full-time education stopped 1047 1 4 2.54 0.70
Man or woman 1054 0 1 0.52 0.50
Age March 2018 1054 1 4 3.42 0.86
36 4.6 Statistical analyses
To test whether or not the hypotheses can be met, 4 binary logistic regressions are conducted in which for every logistic regression a different dependent variable is used. Multicollinearity is ruled out given that the VIF scores are < 4.0. The logistic regressions are divided into three models, in which the first model includes ‘Political_extremism’, the second model adds to this ‘Source_EU_news’ as a moderating variable, after which in model three the interaction variable and control variables are included. This way, one can determine whether or not the influence of political extremism on the dependent variable depends on the type of media source one gets European political information from or on control variables.
And above all else, the different binary logistic regressions can be compared to examine whether a Dutch citizen could oppose one aspect of European integration, whilst having support for another dimension. This would argue that not only a one dimensional approach is biased, but also that the future course of European integration might be more difficult to predict, to examine. and to develop.
5. Results
5.1 Dependent variable: EU_Representation
Model 1 of the first binary logistic regression (Table 4) with the dependent variable ‘EU_Representation’ displays that left extremists have a 60 percent (p < 0.001) higher chance of trusting the European institutions, in contrast to extreme right respondents who have 53 percent (p < 0.001) less chance of trusting them in comparison to moderates, and holding all other variables constant.
Furthermore, model 2 includes the moderating variable and displays that using new media increases one’s chance of trusting the European institutions by 10 percent, holding all other variables constant. However, this result is non-significant. The estimates of the independent variables are unchanged.
37 Model 3 includes the interaction variables. As a result, the odds of trusting the European institutions slightly increase for the extreme right, who now have 51 percent less trust than the moderates (p < 0.05). The estimates for the extreme left remain the same. The moderating variable displays that using new media decreases one’s odds of trusting the European institutions by 18 percent compared to traditional media, however this is a non-significant result.
The two interactions are non-significant, meaning that the odds of trusting the European institutions don’t differ significantly per type of media used. The plot (figure 2) does display, though, that using new media slightly increases the odds of the political right extremists to trust the European institutions, and slightly decreases it for left extremists; in comparison to traditional media. However, future research needs to redefine another way of measuring media use to be more certain of its potential impact.
Finally, the odds to trust are 96 percent greater for those between the ages of 40 and 54 in reference to those aged between 15 and 24 years (p < 0.05). Those who have had 16 to 19 years of full-time education have 363 percent more chance to trust than those with 15 years or less (p < 0.01). Those with more than 20 years of education have 70 percent more chance to trust (p < 0.05) than those who have had 15 or less, and those who still study have 120 percent more chance to trust (p < 0.01). Those that don’t work have 24 percent less chance to trust than those that do work (p < 0.01). See figure 3 for an model overview of the results.
To be certain that the independent variable is robust when being included in the same equation as EU_Representation, a second binary regression is operated (see Appendix 2). Here it is evident that indeed the left extremists are more pro-Europe, with right extremists being the most critical. For example, in comparison to those placing themselves in the first category (1), respondent in the second category have 122.2 percent more chance of trusting the EU institutions, in contrast to those placing themselves in category 10 who have 69.2 percent less chance to trust. Thus, the methodological approach is robust.
38 Figure 2. Interaction effect Source_EU_news & Political_extremism ‘EU_Representation’
39 Table 4. Logistic Regression EU_Representation
Model 1 Estimates (st. error)
Model 2 Estimates (st. error)
Model 3
(Interaction effect) Estimates (st. error)
Constant 0.57 (0.46) 1.04 (0.08) 0.43 (0.95) Political ideology *** *** *** Extreme Left 1.60 (0.00)*** 1.60 (1.16)*** 1.60 (0.18)*** Extreme Right 0.47 (0.00)*** 0.47(0.24)*** 0.49 (0.29)** Source_EU_News 1.10 (0.16) 0.82 (0.22) Interaction
Extreme Left * Media 0.67 (0.41)
Extreme Right * Media 1.34 (0.56)
Age (15-24 years) *** 25-39 years 0.87 (0.27) 40-54 years 1.96 (0.28)** 55 years or older 4.78 (0.99) Gender (man) 0.95 (0.14) Education (≤ 15 years) *** 16-19 years 4.63 (0.93)* ≥ 20 years 1.70 (0.92)** still studying 2.20 (0.92)* Occupation 0.76 (0.16)* Chi-Square: 24.09*** Chi-Square: 24.46*** Chi-Square: 85.48***
Dependent variable: I trust these European institutions N = 1037. * significant at 10%, **significant at 5%, *** significant at 1%
