Knowing is not loving: media effects on knowledge about and attitudes toward the EU

ABSTRACT

Media provide the public with information related to the European Union which may alter individuals’ perceptions, ultimately resulting in changes in performance evaluations of the EU. Knowledge gains may be an important mediator in this process. We present data from a study in the context of the 2016 Bratislava summit in which the Heads of the Union’s governments discussed the outcome of the Brexit vote and the EU’s future. A panel survey assessed the relationship between exposure to media content, event-related knowledge gains, and changes in attitudes towards the European Union. Our results show that when attending to news about the summit, citizens attain event-related knowledge which negatively affects EU performance evaluations. We discuss our findings in light of the role media play in informing the European citizenry.

KEYWORDS:

• Media effects
• European Union
• knowledge
• public opinion

Summits of the European Council are a recurring part of the EU agenda. The informal meeting that was held in Bratislava in the fall of 2016 was a special case, being the first to take place after the United Kingdom voted to leave the Union in a referendum unprecedented in European history. In Bratislava, the Union’s Heads of Government had the opportunity to discuss the potential consequences of the Brexit vote. Following the referendum, media and politicians across Europe reacted with doomsday forecasts, painting a dark picture of the Union’s future. In times like these, EU Council meetings hold the potential to emit important messages to the European citizenry: These messages might reassure the population about the Union’s stability and tone down pessimistic views, but they may also fuel negative beliefs. Against this background, we investigate whether media reports on the EU-27 summit in Bratislava are capable of informing citizens about its outcomes and altering existing opinions toward the European Union. Specifically, we examine whether citizens learn from the news coverage, and whether the acquired knowledge helps them in (re)evaluating the EU’s performance. These questions are particularly relevant when addressing what some have termed a crisis of legitimacy for the European Union (e.g., Hobolt and Tilley 2014).

We argue that events like EU Council meetings may lead to heightened visibility of the EU in national news, making the Union more salient to its citizens. One possible outcome of this heightened visibility may be, first, for individuals to reconsider their attitudes toward the Union’s performance, since summits can be understood as incisive moments in EU history that ‘serve as a catalyst for changes in or development of opinions’ (Semetko, van der Brug, and Valkenburg 2003, 625). Second, and central to this study, we argue that knowledge about the outcomes of such meetings (i.e., event-specific information) which citizens obtain via the media affects their evaluation of the EU’s performance. After all, ‘EU summits are central events in European politics [and] precisely the sort of political event that ought to inform the public about key manifestations of EU performance’ (Elenbaas et al. 2012, 729, emphasis added; see also de Vreese, Azrout, and Moeller 2016). We focus our investigation on citizens’ judgment of the efficiency of the EU’s decision-making process and its democratic performance, and test whether knowledge gains affect European Union performance evaluations. We use two-wave panel data from the Netherlands to test whether a) citizens are exposed to information about the summit, b) whether this information leads to an increase in knowledge about the meeting, and c) whether this knowledge in turn affects performance evaluations of the EU. Our study further contributes to the extant literature by providing insights into the mediating role of event-specific political knowledge as obtained via the mass media, and we discuss our findings in light of ongoing criticisms about the European Union’s information deficit.

Media coverage as a relevant source of information about the EU

Politics in general and supranational politics in particular are fairly distant to citizens’ everyday experiences, which makes it necessary for individuals to rely on information provided by the mass media (Azrout, van Spanje, and de Vreese 2012; Maier and Rittberger 2008). As such, the communication or information deficit of the EU (Clark 2014; Meyer 1999) has often been held accountable for a lack of information and resulting low levels of interest and knowledge of the European citizenry: If individuals only learn about the Union via the news, and if this news is both infrequent and, if existent, mostly negative in nature, people are likely to form EU-related attitudes on the basis of either their knowledge about national politics or to develop negative evaluations in line with the media coverage. Accordingly, media exposure directly or indirectly (i.e., via knowledge gains) affects EU support. Notably, in all member states, EU-related media coverage has increased over the past years, which makes it more likely that citizens are exposed to information about the Union (e.g., Boomgaarden et al. 2013; Sifft et al. 2007). Still, critics evaluate information shortfalls as a serious threat to EU legitimacy, especially if they coincide with a citizenry that lacks the motivation and/or ability to make use of the available information (Clark 2014; but see Tillman 2012for an opposing argument). Yet, research clearly demonstrates that media can provide citizens with substantive information (Barabas and Jerit, 2009; Delli Carpini and Keeter 1996): De Vreese and Boomgaarden (2006) discuss how information-rich environments are better able to educate citizens, and show how news media may affect attitudes toward the European Union, specifically altering individuals’ perceptions of the EU’s performance and efficiency (see also de Vreese, Azrout, and Moeller 2016). Along similar lines, Semetko, van der Brug, and Valkenburg (2003) found media reporting to significantly alter individuals’ supranational-idealistic attitudes toward the EU (i.e., the way citizens are, for example, willing ‘to make personal sacrifices to help other European peoples’, Semetko, van der Brug, and Valkenburg 2003, 628).

Importantly, research distinguishes between forms of general political knowledge and information that is specific to certain events or policies (e.g., Gilens 2001). The level of the former, once established during adolescence and early adulthood (Moeller and de Vreese 2015), is relatively stable, a correlate of education, and consists of facts about politics and ‘political systems that are filed in an individual’s long-term memory and can be recalled to identify, interpret, and understand political events’ (Clark and Hellwig 2012, 542; Delli Carpini and Keeter 1996). In the case of the European Union, general political knowledge relates to, for example, citizens’ awareness of the number of EU Commissioners, the name of the current President of the Commission, or the number of member states (e.g., de Vreese and Boomgaarden 2006; Karp, Banducci, and Bowler 2003). While these facts form an important background for understanding the workings of the EU, they do not directly bear on the performance of the Union, and we argue that individuals are less likely to form attitudes on their basis. Instead, higher levels of general political knowledge (or political sophistication) may increase the likelihood that people a) get exposed to political information in the media in the first place, and b) are able to incorporate new information into their existing knowledge (de Vreese and Boomgaarden 2006). Contrary to that, event- or policy-specific knowledge results from learning about ongoing events through interpersonal talk or the media. Scholars have urged to differentiate the two concepts in measurements and theories, since ‘many people who are fully informed in terms of general political knowledge are nonetheless ignorant of policy-specific information that would alter their political judgments’ (Gilens 2001, 380; see also Shehata et al. 2015). Hence, a sole reliance on measures of general knowledge is insufficient when addressing potential attitude changes as a result of exposure to information in the media. Even more so, specific knowledge bears greater effects on citizens’ opinions about political entities than general information (Barabas and Jerit, 2009; Gilens 2001). For these reasons, in order to measure knowledge gains, scholars need to address new political information, that is, knowledge about current events or changes in policies, which adds to the general level of knowledge established earlier in life.

In the process under investigation here, we posit that citizens need to learn about an event (the 2016 EU Bratislava summit) in order for it to bear any effect on their attitudes toward the European Union. Therefore, we acknowledge that the media play a ‘double-barrelled role’ (de Vreese and Boomgaarden 2006, 334): Information provided by the media may, on the one hand, influence event-specific knowledge and, on the other hand, affect attitudes directly (Semetko, van der Brug, and Valkenburg 2003). Yet the ‘assertion that higher levels of media coverage increase knowledge is neither obvious nor is it a foregone conclusion’ (Barabas and Jerit, 2009, 73) – in addition to mediated information, citizens also learn about the European Union via, for example, interpersonal communication (de Vreese and Boomgaarden 2006). This influence of media exposure on knowledge gains thus needs to be established first; subsequently, we focus on the indirect effect of media exposure on attitude change via event-specific knowledge. Importantly, in providing data from a panel survey, our study sheds light on the first step in this mediated relationship, that is, the causal impact of news exposure on knowledge gains, in a realistic situation and outside the forced-exposure designs of experiments. We hence start with the core focus on the effect of media exposure on specific knowledge about the Bratislava summit:

H1: Exposure to news reports about a key EU event increases individuals’ event-related knowledge.

EU performance evaluations and the influence of event-specific knowledge

In a second step, we assume that individuals’ knowledge about specific events and policy issues may affect their attitudes toward the respective object – in this case, the performance of the EU (Gilens 2001). The literature on the antecedents of EU support, or ‘Euroscepticism’, is vast and growing, and scholarly consensus is arising that any study seeking to explain EU attitudes should clearly define and delimit its dependent variable (Hobolt and de Vries 2016). Citizens’ attitudes towards the European Union are best understood as multidimensional (Boomgaarden et al. 2011; Hobolt and Brouard 2011), and different EU attitude dimensions can have different antecedents, and may be more or less volatile over time (Boomgaarden et al. 2011). A central distinction is based on Easton’s framework of diffuse versus specific political support (1965), which was first applied to EU support by Kopecký and Mudde (2002). Whereas diffuse support refers to a basic attachment to the political community and regime, and is thus less sensitive to short-term influence, specific support is based on evaluations of the current functioning of the regime, authorities, and specific political actors. Boomgaarden et al. (2011) have built on this distinction by additionally differentiating between various objects of support, and conceptualize public opinion towards the European Union in five dimensions: citizens’ evaluation of the EU’s performance, their European identity, preferences regarding the strengthening of the Union, utilitarian considerations, and individuals’ feelings of negative affect toward the EU.

We believe that among those five dimensions, performance evaluations are most likely to be affected by event-specific media coverage: This dimension entails attitudes towards the functioning of the EU, that is, regarding the political, democratic, and financial performance of institutions such as the Council or European Parliament (e.g., ‘The European Union functions well as it is’, ‘I am satisfied with the way democracy works in the EU’). As such, procedural performance evaluations are one important aspect of Easton’s (1965) notion of regime support. They also relate to the perceived transparency in the political decision-making process and are conceptualized as rather specific support that may undergo considerable changes over time and under the influence of particular events, policy-outcomes, and economic evaluations (Boomgaarden et al. 2011). Performance evaluations thus assess the output of the EU and its legitimacy (de Vreese, Azrout, and Moeller 2016). The Bratislava summit, similar to other events on the level of the European Union, potentially increases media coverage, which makes it easier for citizens to observe the EU ‘in action’. Based on existing work (de Vreese, Azrout, and Moeller 2016; Elenbaas et al. 2012; Karp, Banducci, and Bowler 2003), we therefore believe performance evaluations to be particularly susceptible to information presented in the media, especially if this information leads to an increase in knowledge about specific EU-related events. In line with this argument, it can be assumed that EU performance evaluations fluctuate in line with actual performance of the Union (de Vreese, Azrout, and Moeller 2016) – provided that individuals do, in fact, know about this performance.

Gilens (2001) shows that factual knowledge about government spending in the United States has a positive influence on respective policy-related attitudes, highlighting the influence of political ignorance on public opinion. In the context of EU attitudes, competing theories regarding the precise mechanism through which knowledge affects support for the Union exist, which also imply different directions of influence (i.e., positive or negative): First, according to the concept of cognitive mobilization (Inglehart 1970), added knowledge leads citizens to become more familiar with a remote institution, making this institution less threatening and more transparent. This may lower scepticism and allow citizens to take a more positive stance on EU performance (e.g., Karp, Banducci, and Bowler 2003). In its most basic form, cognitive mobilization is message-independent, that is, ‘all information about [EU] integration promotes support’ (Gabel 1998, 335). Indeed, Inglehart (1970, 54) finds that ‘cognitive mobilization [is] the most powerful predictor of pro-European attitudes’. Second, and in contrast to cognitive mobilization, the democratic deficit hypothesis states that citizens’ satisfaction with the way the EU works should decrease with growing information about what is actually happening in the political decision-making process, as political sophistication leads individuals to be able to hold a more critical view on the Union’s performance (e.g., Clark 2014; Karp, Banducci, and Bowler 2003; Norris 2011). The increase in democratic aspiration (Norris 2011) allows citizens to have higher demands in democracy since they are better able to evaluate it on the basis of the information and knowledge available to them. Accordingly, the concepts of cognitive mobilization and democratic deficit lead to diverging assumptions about the influence of knowledge on EU performance evaluations.

As one example, Karp and colleagues (2003) provide evidence for lower support levels among more knowledgeable citizens: The more individuals know about the European Union, the less satisfied they are with the way democracy works in the EU. Interestingly, a reverse relationship can be observed for national evaluations, where knowledge positively correlates with satisfaction. Thus, these findings lend support to the democratic deficit hypothesis on the EU level, while providing some insights into cognitive mobilization for the assessment of national governments. The authors conclude that, in contrast to national institutions’ evaluations, ‘[d]issatisfaction with the EU is influenced by a lack of confidence in EU institutions and the perception that the EP [European Parliament] is weak’ (Karp, Banducci, and Bowler 2003, 289). One possible explanation for this lack of confidence and lower performance perceptions may be found in diverging levels of media coverage. However, Karp and colleagues relied on basic knowledge facts about the European Union, asking, for example, whether subjects were able to identify the current president of the EU Commission. Hence, their assessment of knowledge is less related to performance evaluations, and can be better understood as a proxy of citizens’ overall knowledge about the Union and its institutions (see also Clark and Hellwig 2012). In addition, while analysing valuable data from a cross-national survey, the authors could not provide evidence of causal influences by assessing panel effects over time. Hence it is not clear whether specific events may directly affect citizens’ evaluations of the EU.

A more nuanced differentiation in this regard was rendered by Elenbaas and colleagues (2012), who investigated news coverage of a particular EU summit and subsequent alterations in Dutch citizens’ knowledge about and attitudes toward the Union. The authors found information acquisition to positively affect EU utilitarian performance judgements (e.g., ‘I personally benefit from Dutch membership of the EU’), but to not have a significant impact on evaluations of EU democratic practices (e.g., ‘The decision-making process in the EU is transparent’). Even though the latter became slightly more negative, the information effects were not significant, and we can thus only speculate about the negative influence of news coverage on evaluations of the Union’s work and proceedings. Furthermore, the authors did not directly model information acquisition as an outcome of media exposure.

In sum, there are strong reasons to expect an effect of political knowledge on EU support, but the direction of this effect is less self-evident. Furthermore, our study emphasizes event-specific knowledge, whereas extant research focuses on general political knowledge. Given the ambiguous theoretical mechanisms as well as the lack of research on the effects of event-specific knowledge, we assume increased levels of knowledge about the Bratislava summit to alter citizens’ attitudes toward the EU’s performance and democratic principles, but do not specify the direction of this effect in our hypothesis:

H2: Event-related knowledge affects performance-related attitudes toward the European Union.

Our two hypotheses are interrelated: If exposure to news is conducive to knowledge gain (H1) and specific knowledge affects attitudes (H2), we are essentially proposing an indirect relationship (Barabas and Jerit, 2009; Elenbaas et al. 2012; Gilens 2001), which is subsumed in our last assumption:

H3: Event-related knowledge mediates the relationship between media exposure and performance-related EU attitudes.

Method

In order to test our hypotheses, we collected original Dutch survey data around the EU summit which was held in Bratislava on September 16–17, 2016. Though not legally binding due to the UK not being present at the summit, recommendations and plans made during the meeting may well have important consequences for the future of the EU, if only as a symbolic event. Among other topics, the summit provided an opportunity for the European leaders to discuss the ongoing ‘refugee crisis’ and cushion the aftermath of the Brexit vote. The ‘roadmap’ which was developed as a result of the meeting also includes proposals for security policies inside and outside of the EU. Through the media coverage of the summit, European citizens thus had the chance to evaluate the performance of the EU in relation to very specific issues (Brexit, security), despite the high complexity of the EU-topic in general.

We collected two wave panel survey data, adding these waves to an ongoing panel survey of the Dutch voting population before the 2014 European Parliamentary elections. The original sample was drawn from the TNS NIPO Netherlands database, which includes 200 000 individuals that were recruited through multiple strategies, including telephone, face-to-face, and online recruitment. The subsequent survey in this study was conducted using Computer Assisted Web Interviewing (CAWI). Quotas on age (W1: M = 48.7), gender (50.6% female), and education (34% bachelor’s degree or higher) were enforced in sampling from the database. The panel started with its first wave in December 2013/January 2014 and collected data from 2 189 respondents. Afterwards followed several waves up to and including the European Parliament elections in May 2014 and pre- and post-waves around the April 2016 Dutch Ukraine-EU Association Agreement referendum. To these panel waves we added questions specific to the summit in Bratislava in two more waves (one before [August 31–September 7], one after the summit [September 19–26]). Response rates in the panel varied between 83.1% (wave 2) and 96.1% (wave 7), with the final post-summit wave (wave 9) reaching an overall retention of 37.9% of the initial sample (n = 829). Our post-summit wave is similar to the original first wave of the panel in the distribution of age, gender, and education.

In addition, we used a keyword-based search to assess the visibility of the summit in Dutch news outlets to ensure that media coverage did, indeed, provide information on the outcomes of the Council meeting. This scan of the media environment offers us an important, albeit basic, estimate on the extent to which the summit was covered in the Netherlands, and thus enables a better interpretation of the media effects under investigation. Since we claim that event-related knowledge is a result of exposure to media content about the summit, we have to ensure that facts on the meeting’s outcomes are actually provided by the coverage. Hence, this additional media data backs our theoretical assumptions and strengthens the model’s reasoning.

Measures

Independent variable

Our dichotomous media exposure variables asked respondents, in the second, post-summit wave, whether or not they had heard about the Bratislava summit through TV news (40.6% did), via newspapers (14.7%), on the radio (10.4%), or online (7.3%). Roughly 55% (n = 458) of our sample reported having heard about the EU meeting through at least one of these outlets. 233 individuals (28.1%) said they exclusively learned about it through television, while few people relied solely on radio (4.8%), newspapers (4.6%), and online news (3.5%). In order to check for actual media information, we computed search queries on LexisNexis (hardcopy newspapers) and websites for TV- and online news in the Netherlands, using terms related to the summit as well as their respective combinations (top, Bratislava, Slowakije, Slowaakse, EU, Europese Unie, Tusk, or Juncker). No content data was available for radio coverage, but given its relevance as an information source, it was included in the score of the independent variable (see below).

Since we were interested in citizens’ knowledge gains with regard to the outcomes of the Bratislava summit, we had to ensure that only those media reports published on the days including and following the meeting were taken into account. We hence specified our search period as ranging from Friday, September 16, to Tuesday, September 20 of 2016. During these five days, we identified a total of 23 Dutch news items covering the Bratislava summit. The media outlets include hardcopy newspapers (Telegraaf, 1 item; NRC Handelsblad, 3 items: Algemeen Dagblad, 3 items; de Volkskrant, 2 items), online newspapers (telegraaf.nl, 1; nrc.nl, 3; ad.nl, 3; volkskrant.nl, 2), and news items from the public TV channel NPO (Nederlandse Publieke Omroep, 5 items). On the one hand, the selection of these outlets was driven by practical reasoning (i.e., availability through LexisNexis and/or online archives); on the other hand, we made sure to include nationwide broadsheet and tabloid newspapers with high circulation numbers that cover a broad range of political stances as well as the most important news shows from the public broadcaster NPO (e.g., NOS Journaal, Nieuwsuur, Goedemorgen Nederland). Originally, we also included the free newspaper Metro which has the highest nationwide circulation numbers; however, no articles addressing the Bratislava summit were published during our period of investigation. The dichotomous indicators for exposure to content about the Council meeting through these outlets (plus radio) were merged and included as one general quasi-experimental predictor of media exposure in our model, measuring whether or not respondents were exposed to the EU summit through the news.

Mediator

In addition to our main independent variable, we measured individuals’ knowledge about the summit. Event-related knowledge was assessed by asking respondents four knowledge questions that were directly linked to the outcome of the meeting in the post-summit wave and adding these to a sum score (KR-20 = .63; M = 1.1, SD = 1.12, see Appendix for exact question wording). These questions tap into specific event-related knowledge as opposed to general, EU-related facts (Clark 2014), and we consider it very likely that this information would be provided by way of media.

Dependent variable

We assessed respondents’ performance evaluation of the European Union both before (t₁) and after the summit (t₂). Here, we relied on the respective EU attitude dimension’s index as suggested by Boomgaarden and colleagues (2011; four items: ‘The European Union functions well as it is’, ‘The EU functions according to democratic principles’, ‘The decision-making process in the EU is transparent’, ‘I am satisfied with the way democracy works in the EU’, index mean-centred ranging from −3 to +3, t_1: M = −1.02, SD = 1.27, α = .88; t_2: M = −0.96, SD = 1.27, α  = .89), based on a principal component analysis of the first, pre-summit wave. EU performance evaluations at t₂ serve as our dependent variable.

Controls and analysis

In order to test our assumptions, we estimated a process model (Hayes 2013) in which we assumed exposure to media reports about the summit to influence European Union performance evaluations at t₂ through event-specific knowledge (see Figure 1). We simultaneously controlled for sociodemographics (age: M = 53.86, SD = 16.88, ranging from 20 to 90 years; gender: 50.6% female; education: three levels, 37.3% bachelor degree or higher), performance evaluations at t₁, and general political interest, which we derived from an earlier wave of the panel (two items measuring interest in politics in general and on the level of the EU in particular on a seven-point scale, respectively; α = .84, M = 3.47, SD = 1.49). In addition, we took into account the interview date on which respondents were questioned in the post-summit wave, since one may argue that the number of days (M = 3.32, SD = 1.80) that passed in between the end of the summit and the interview date may affect the likelihood of individuals being exposed to media content about the meeting, which subsequently could influence knowledge gains, memory, and attitude changes as well. Overall, the described procedure allows us to estimate individual changes in attitudes as an outcome of knowledge gain through media reporting. Given the amount of restrictions and controls included in the model, our analysis offers a conservative test of the hypothesized relationships. Most notably, explained variance in changes in performance attitudes is, to a large extent, explained by prior attitudes in the pre-summit wave (i.e., the autoregressive effect). Hence, results and effect sizes reported below need to be interpreted with caution, but also against the background of this conservative estimation. Finally, it is important to note that even though we treat exposure to media content about the summit as a quasi-experimental condition in our model, the data is, in fact, observational, and exposure was self-reported by our respondents, not manipulated.

Figure 1. Conceptual process model.

Results

In a first step, we assess the direct effects as outlined in the conceptual model. As expected (H1), media exposure through newspapers, websites, radio, and/or TV positively and significantly predicts knowledge about the Bratislava summit (b = .889, p = .000), along with the control variables age (b = .012, p = .000), education (b = .221, p = .000), and political interest (b = .144, p = .000; see Table 1). Interview date negatively affects knowledge gains (b = −.052, p = .006), which is to say that the more time passed between the summit itself and individuals being questioned about what they learned about its outcomes, the less knowledge gain they reported (a possible explanation for which might be fading memory over time). Attitudes toward the EU’s performance at t₁ also influence knowledge, albeit negatively (b = −.064, p = .017), while gender has no effect in this context.

Table 1. Regression results for event-related knowledge and EU performance evaluations (n = 829).

	Event-related knowledge			Performance t2
	b	SE	p	b	SE	p
Constant	−1.013	.228	.000	−.293	.192	.128
Media exposure	.889	.072	.000	−.024	.065	.708
Performance t1	−.064	.027	.017	.774	.022	.000
Gender	.080	.068	.239	.028	.056	.616
Age	.012	.002	.000	−.001	.002	.634
Education	.221	.050	.000	.037	.042	.376
Political interest	.144	.025	.000	.008	.021	.722
Interview date in days	−.052	.019	.006	.025	.016	.109
Event-related knowledge	–	–	–	−.054	.029	.064
	R^2 = .332 F(7, 821) = 58.403 p = .000			R^2 = .611 F(8, 820) = 161.037 p = .000

Note. b represent unstandardized coefficients.

When it comes to our main dependent variable, attitudes at t₂, we find no effect of media exposure about the Bratislava summit on performance evaluations (b = −.024, p = .708). Performance attitudes at t₁ are the strongest predictor of the same dimension in the post-summit wave (b = .774, p = .000). None of the other predictors exerts a significant effect in this equation (see Table 1). In analysing results for H2, event-related knowledge lowers performance evaluations of the European Union, but this effect shortly fails statistically significance at the conventional .05-mark (b = −.054, p = .064). Even though media exposure significantly and positively affects event-related knowledge, and we find knowledge to be a non-significant predictor for performance evaluations to change, we need to take into consideration the indirect effect as well (the mediation hypothesis, H3). Bias-corrected bootstrap confidence-intervals (95%, 10,000 samples) include zero for the indirect path (CI [−.0963, .0008]), again failing to reach statistical significance. Therefore, we cannot demonstrate event-related knowledge to function as a mediator between media exposure and EU performance evaluations, and reject H3.

Conventional regression analyses support the mediation results: In a model predicting EU performance evaluations at t₂ through media exposure, our control variables, and performance attitudes at t₁, media exposure in itself does not significantly affect attitudes (ß = −.023, p = .328; detailed tables not depicted here). Adding knowledge about the summit in a second step further decreases this influence (ß = −.002, p = .954), and increases the explained variance of the model (adj. R² = .607 compared to adj. R² = .606, p = .052). Yet it should be noted that in terms of its substance, this increase is very marginal overall. Again, we cannot claim mediation of media exposure on performance evaluations through event-related knowledge.

Discussion

This study aimed at providing insights into the stability of European Union performance evaluations, and we tested whether such attitudes change under the influence of event-related knowledge provided by the media immediately after a potentially decisive EU meeting. In general, we consider this a rigorous test of our theoretical propositions about the impact of the media on short-term public opinion dynamics. More specifically, in our case, we argue that the Bratislava summit of the ‘remaining’ 27 heads of EU-states functions as an important symbolic event in the aftermath of the United Kingdom’s vote to leave the European Union. The information the European citizenry receives about such an event could ideally reassure them that all may not be as bad as expected, but may also worsen performance evaluations of the EU.

Our findings reveal that some people in the Netherlands were not only exposed to information about the summit, but that they also learned event-related facts via the media. Importantly, media exposure in itself did not bear any effect on EU performance evaluations, but only affected knowledge gains. Even though both the resulting direct effect on EU performance evaluations and the full indirect path fail to reach statistical significance, our results hint at a relevant media influence under conditions of a comparably small EU event. These findings are in line with extant literature and point to an important distinction with regard to the differentiation between general and event-specific political knowledge: As Elenbaas and colleagues (2012) argued, ‘new facts about the EU should be […] more consequential for judgements about EU performance […] when the facts bear directly upon those judgements’ (p. 731, original emphasis). We argued that it is necessary to investigate the effect of event-specific knowledge on political attitudes in contrast to only considering general knowledge effects, and indeed find that event-specific facts work in addition to general education levels (see below).

In this context, we find it noteworthy that learning about the Bratislava 27-summit actually shows a tendency to decrease individuals’ evaluations of general EU performance. This result apparently stands in contrast to a recent study by de Vreese and colleagues (2016), who found performance evaluations of the European Union to improve as a result of media exposure. The authors attribute this positive effect of an overall negative media content to citizens being able to observe the ‘EU in action’ (De Vreese, Azrout, and Moeller 2016, 79). In the case of the Bratislava meeting, it may be argued that the summit raised high expectations in Dutch citizens, which were afterwards met with dissatisfaction in light of the minimal outcomes of the summit. Indeed, such overall disappointment was observable in the coverage as well – even international media outlets evaluated the summit as resulting in a ‘loss of faith in the E.U.’ (The New York Times 2016). While we can only speculate about this potential disillusion-effect here, it tallies with the democratic deficit hypothesis (Norris 2011). Performance-related EU attitudes seem to be easily affected by new incoming information, and are prone to change under the influence of short-term events – an explanation that is in line with the underlying theory of the respective attitude dimension (e.g., Boomgaarden et al. 2011; de Vreese, Azrout, and Moeller 2016). Therefore, at least in the context under investigation here and with regard to what the European Union actually does, ‘to know it’ certainly does not mean ‘to love it’ (Karp, Banducci, and Bowler 2003).

Some limitations are worth considering for future studies: First, we cannot provide information about the possible moderating influence of individuals’ general political knowledge (e.g., Clark 2014; Delli Carpini and Keeter 1996; de Vreese and Boomgaarden 2006; Zaller 1992). However, we account for individuals’ level of education, and general political knowledge can be understood to be dependent on educational levels, and to be less susceptible to short-term influences (e.g., Fraile and Iyengar 2014). Indeed, we see that higher educated people are more likely to have been exposed to media content about the Bratislava summit in the first place, but that education does not affect attitudes. Along similar lines, general political interest positively predicts knowledge gain, but does not affect performance evaluations either. Clearly, general political knowledge and event-specific learning capture different concepts (Barabas and Jerit, 2009; Gilens 2001). We thus urge future research to take a closer look at the (possibly moderating) impact of general political knowledge when it comes to assessing what and under which circumstances people learn from media reports.

Second, we do not directly include content data of the media reports on the summit in our analysis, since the low number of news items covering the Bratislava meeting does not allow us to draw inferences about content features (e.g., valence or positioning). Furthermore, to assess the degree of a positive or negatively framed coverage would require us to compare the Bratislava summit coverage with news items on a similar, but differently valenced event. We therefore believe that any interpretation could be potentially misleading in this regard, which is why we relied on a dichotomous and quasi-experimental predictor of media exposure in our models only. We strongly advise future studies to take into consideration media content of the news as well, given that these are available in sufficient numbers for content coding (Barabas and Jerit, 2009; Bruter 2003). In light of existing research that finds knowledge gains to be especially pronounced after exposure to hard- as opposed to soft-news sources (e.g., Fraile and Iyengar 2014), and broadsheet newspapers to report on EU-related issues more often than tabloid outlets (Boomgaarden et al. 2013), we deem it worthwhile for future studies to look into these aspects in more detail. Specifically, such efforts should focus on additional media outlets, content features, and the role of social media and interpersonal communication as well. In addition, since we focused on one specific event, we assume that differences between newspaper, online, and TV news are marginal. Indeed, inspection of the content finds repeating information of the same key facts and interpretations throughout all outlets under investigation, and differences in the amount of exposure may not affect knowledge differences overall. We deem it likely that other events such as EU Parliamentary elections or important referendums may result in greater media coverage both in terms of quantity and diversity, but these questions remain open for further investigation.

Third, a generalization of our results needs to be seen through the prism of the specificities of this particular EU summit and our respective data-collection. Even though we model media exposure as a quasi-experimental predictor in our analysis, the measurement remains observational and relies on respondents’ self-reported exposure. We can thus not exclude the possibility that media exposure, as the first step in our analysis, is influenced by other relevant variables outside our observation in the first place. Furthermore, our case study is situated in the context of a real-world event, which, albeit being important to the EU at large (and allowing for timely data collection and interpretation), only held minor direct implications for Dutch citizens. At the same time, the Netherlands can be considered an exemplary Western EU country for testing our assumptions, and we find it unlikely that investigations in a different member state would have yielded deviating results. We deem it likely that the Bratislava Council meeting was not particularly influential, in that no ‘grand’ decisions were made, and that it resulted in no new policy outcomes that would have warranted stronger attitude change. Given that the summit was thus less spectacular than (could have been) expected and received relatively little coverage in the news, it is not surprising that some of our investigated relationships fall just short of conventional levels of statistical significance. While caution is thus advised in evaluating the relevance of the meeting and our comparably small sample, we still deem it noteworthy that the summit did, in fact, alter people’s evaluations of the Union, at least to some extent. As such, the communication deficit (Meyer 1999) that is often hinted at in research about the European Union’s legitimacy cannot be detected in the case of this particular event. Furthermore, our findings highlight that ample media coverage alone is not sufficient in improving public opinion of the European Union’s performance (see also Goldberg and de Vreese 2018), but that knowledge gains aid in building an informed and politically sophisticated citizenry. From a normative democratic perspective, this is indeed a positive outcome, even if EU issues usually are of little relevance to the Union’s citizens. Political events similar to such a Council meeting, but with higher direct impact on individuals’ lives, may produce stronger effects. Hence, more work is needed to analyse the effects of specific events and the accumulated impact of a consistent coverage over time, which may eventually lead to a more general conclusion about the exact relationships between media exposure, knowledge, and attitudes toward the EU. Such a conclusion would also benefit from a cross-country analysis that includes a variety of European countries.

Finally, we cannot make any claims about the longevity of these effects. Research further investigating the stability (or fluctuation) of EU performance evaluations as affected by information in and knowledge gained through the media should thus consider whether or not changes are short-termed. Boomgaarden et al. (2011) established that this dimension of European Union attitudes relates to transparencies in political decision-making processes, and we assumed that performance evaluations are more volatile and susceptible to short-term influences compared to, for example, feelings of European identity. Yet the design of our study does not allow us to test this longevity, and we cannot claim that the negative influence noticeable in our models remains stable over time. Whether or not coverage of single events alone is thus powerful enough to lastingly affect evaluations and long-term public opinions or divides remains a question for future research. Ideally, studies should look into the reinforcing process of media exposure and effects in this context (Slater 2007), since it could be assumed that attitudes, media selectivity, and knowledge-related outcomes are mutually influential (Moeller and de Vreese 2015).

Taken together, our findings hint at the media’s capacity to contribute to an increase in awareness for European issues in the Netherlands. The fact that public information led to lower performance-related EU evaluations here does not diminish its influence: An informed citizenry should be considered a necessary, albeit not sufficient criterion for a well-functioning democracy (Inglehart 1970), even more so if the knowledge obtained through news media allows individuals to take a more critical position toward the political system.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This research is funded by a grant from the European Research Council (ERC) as part of the Horizon 2020 program, grant number [647316].

Appendix

Event-related knowledge t₂: Closed multiple-choice questions ‘How many different countries participated in the EU summit?’ 26 – 27 – 28 – 29 different countries – Don’t know.

‘Which country did not participate in the Bratislava summit?’ Great Britain – The Netherlands – Lithuania – Spain – Don’t know.

‘Who was the most critical of the proceedings right after the summit?’ Angela Merkel (Germany) – Viktor Orbán (Hungary) – Francois Hollande (France) – Matteo Renzi (Italy) – Don’t know; multiple answers possible, answers were coded as correct if either Orbán and/or Renzi were named.

‘Which topic was not discussed at the EU summit in Bratislava?’ Brexit – Turkish EU membership – Security inside and outside of the European Union – Distribution of refugees among EU countries – Don’t know.