EU Issue Voting in Simultaneous Elections: The Case of Belgium

Abstract

In terms of electoral rules, Belgium is an odd case out among the countries investigated in this volume on several levels. Elections for the regional, national and European level are held on the same day, and all voters are obligated to turn out to vote. This combination of circumstances makes Belgium a least-likely case to find evidence for EU voting on the national level: voters can express their European interests with a vote for that level specifically at the same moment, and the forced and less interested part of the electorate is less likely to express strong issue preferences with their vote. This study first shows that, overall, the European issue is not very salient in Belgium, and that there is a generally positive consensus among the parties and in the media. However, still, there are some effects of the EU issue on national voting behaviour. While these are limited overall, also other well-known determinants of the vote do not show strong effects. Explaining voting behaviour in the complex Belgian case remains challenging, and the EU issue seems to provide one piece of the puzzle.

Introduction

On 24 May 2019, Belgium held a major Election Day in which representatives for three different governmental levels were chosen. In this “mother of all elections”, voters could cast three different votes: for the regional, the national, and finally the European level—in that order. With this system of simultaneous elections, the European level could be expected to be omnipresent in the campaign and therefore also influence the vote at the regional and national levels. On the other hand, this also means that the European issue faced fierce competition for attention during the campaign, which focused mostly on regional and national issues (Pilet, 2020). The electoral results showed substantial losses by governing parties, which were gained by the opposition (Pilet, 2020). Especially radical (populist) parties performed particularly well (Gallina et al., 2020; Goovaerts et al., 2020). In general, the country appeared divided, with the most support for (centre-)right parties in Flanders, and (centre-)left in Wallonia (Hooghe & Stiers, 2022).

The purpose of this chapter is to investigate whether people voted for the national level with European issues in mind. As explained in the theoretical framework of the book (Chapter 1), there are important implications of this kind of EU issue voting in national elections for the legitimacy of the European institutions. As the national governments largely decide on their representation on the European level, from a normative point of view it is important to investigate whether people vote for the national level with the European issue in mind. As a net contributor, the investigation of the Belgian case moreover provides a useful point of comparison for the other countries discussed in the subsequent chapters to test whether the extent of EU issue voting diverges or converges.

Even though it is a central part of the EU, Belgium very much presents a least-likely case to find evidence of EU issue voting, as the European issue is likely to be overshadowed by the issues that are important for the “first-order” regional and national elections (Reif & Schmitt, 1980), and as the compulsory voting system in place is likely to draw non-interested voters to the ballot box, engaging less in ideological issue voting (Dassonneville et al., 2019; Selb & Lachat, 2009). The findings in Chapter 5 already revealed Belgium to be an odd case out. While there was evidence for EU issue voting and its expected asymmetric nature, voters only seem to punish parties that are more positive on the EU than they are themselves—i.e., the reversed asymmetry than what was found in the other countries. This chapter will use individual-level survey data to contrast the EU issue with other determinants of the vote to test whether and which voters vote for the national level with the European issue in mind.

EU Issue Voting in Context: Belgium as a Least-Likely Case Study

The European Union (EU) is an important issue in Belgium, one of its founding members. Not in the least, Brussels, Belgium’s capital, houses the most important institutions of the European Union and is the location for its summits. This means that Belgians feel the presence of the EU very concretely. Media report regularly on European decision-making, as it is happening in the centre of the country, and reporting that “Brussels has decided to…”, means something more concrete in the minds of the people than when the same news is brought to people in countries further away. However, many people also feel the impact of the EU institutions on their daily lives, e.g., when demonstrations obstruct the traffic in Brussels. Belgian politicians have also held important positions at the European level, with two of the three permanent Presidents of the Council so far. Clearly, Belgians have good reasons to pay attention to the EU, and to take considerations about this level into account when they take part in the political process, for instance when they turn out to vote. There are indeed good reasons to believe that Belgian voters should be likely to take the EU issue into account, as in general Belgians have been found to be quite positive and supportive of the EU (McLaren, 2004). Furthermore, while the economic crisis of 2009 decreased public support for the EU in many countries, there was only a limited loss in support for the EU between 2007 and 2011 in Belgium compared to other countries (Armingeon & Ceka, 2014; Braun & Tausendpfund, 2014).

However, even though the European Union is very much present in the Belgian political landscape, when it comes to explaining voting behaviour, several characteristics of the Belgian national elections make it a least-likely case to find the effects of the European issue on the national vote choice (de Vries, 2007). More specifically, the system of simultaneous elections together with compulsory voting, makes it a difficult context to find the effects of opinions on the EU on the vote at the national level.

First, as mentioned above, one peculiar characteristic of the Belgian elections of 2019 is that elections for three different levels of government were held on the same day: the regional, national, and European elections. This system has been in place since the election of 2014 and was introduced to increase the stability of the federation by making it more likely that the coalitions at different levels of government are the same (Deschouwer & Reuchamps, 2013). The only elections that follow a different schedule are the local elections, held every six years. Simultaneous elections also make for simultaneous campaigns, which means that the attention of parties, media, and voters, is divided over issues of the different electoral levels. In principle, it could be thought that these simultaneous elections make it more likely to find EU issue voting in national elections: as the European elections are held together with the national elections, this means that the EU issue should also be discussed at length during the (national) campaign. Especially compared to national elections in other countries, where European elections follow a different schedule, this means that the European level should be at least somewhat salient in the months before Election Day.

However, there is reason to believe that simultaneous elections actually decrease the attention on the EU issue in the campaign overall. As previous chapters already discussed, European elections are often considered “second-order national elections”, as they are decided mainly on issues that are at play in the first-order (i.e., national) electoral arena (Reif & Schmitt, 1980; Schmitt, 2005). This means that other issues on the national level might overshadow the EU issue. This effect can be expected to be even stronger when the elections for these levels are held on the same day. Furthermore, and importantly, given the Belgian federal structure, the regional elections can also be considered first-order elections (Deschouwer, 2012; Schakel & Jeffery, 2013), so regional issues also compete with the EU issue for the necessary attention. Instead of the EU issue being more prominent overall because of the European elections taking place, it thus competes for attention against two “first-order” electoral arenas, which further limits the attention for this level specifically when voters turn out to vote for all three levels. Therefore, it can be expected that the simultaneous elections decrease the importance of the EU issue overall, and the fact that the European elections are held together with the national elections does not increase the salience of the EU issue for the electoral contest on the national level.

Besides the fact that the EU issue likely receives less attention, simultaneous elections can also decrease the importance of the EU issue in the vote for the national level specifically: voters who wish to express their opinions on the EU issue can do so in the election for that level while expressing their interests on regional and national issues with the votes for those levels respectively. This might decrease EU issue voting in the national election. A finding that further corroborates this argument is that a substantial amount of voters split their ticket (i.e., vote for different parties for the different levels). In their investigation of the simultaneous elections of 2014, Willocq and Kelbel (2018), for instance, find that about a third of the voters voted for at least two different parties on the three levels, and the largest group of the ticket-splitters (14.8%) votes for the same party on the regional and national level while supporting a different party on the European level. Van Aelst and Lefevere (2012) confirm for the 2009 elections that people split their tickets based on Euro-specific considerations. While these studies thus imply that voters do take the European issue into account when casting a vote at the European level, their findings further indicate that Belgium is a least-likely case to find the EU to influence voting behaviour for the national level specifically. If people split their ticket between the national and European levels with European issues in mind, they are also more likely to cast a vote for the national level with specific national issues in mind instead of using the same—possibly European—considerations for the vote on every level.

Finally, an important characteristic of the Belgian electoral system that makes it a hard test for EU issue voting is that all voters are obligated to turn out to vote on Election Day. This system of compulsory voting has ensured that turnout rates are stable at around 90% for each electoral level. This is a stark contrast with the lowering turnout rates in other countries (Hooghe & Kern, 2017). This decline has also been observed for the European elections: even though the 2019 election saw an increase in turnout compared to the previous elections, in general, there is a downward trend in participation that is expected to continue in the decades ahead (Bhatti & Hansen, 2012). As a consequence of these high rates, however, Belgium draws voters to the ballot box that would otherwise not turn out because they are not interested in politics (Hooghe & Pelleriaux, 1998; Selb & Lachat, 2009). While there is some evidence that voters who are forced to turn out inform themselves at least to some extent (Shineman, 2018), other studies have found that these “reluctant” voters cast votes that are less in line with their own ideological opinions and interests because they are less interested in politics (Dassonneville et al., 2019; Hooghe & Stiers, 2017; Selb & Lachat, 2009). In terms of the research question under investigation, if the compulsion to turn out to vote draws a larger proportion of voters casting less congruent or even random votes to the ballot box, this makes it harder to find evidence for EU issue voting. Therefore, as will be further explained below, voters’ levels of political information will be taken into account to empirically investigate EU issue voting.

Politicisation of EU in Media and Parliamentary Debates

Before testing EU issue voting at the individual level using survey data, it is useful to examine how salient the European level is in general in Belgian politics, and in what tone the EU is usually discussed. To do so, I look at salience levels and tone both in the country’s media and in parliamentary debates.

First, I investigate the salience and tone in two main newspapers—one of each main region respectively: De Standaard (Flanders) and Le Soir (Wallonia). The data span the last five national elections, between 2003 and 2019.¹ Salience was measured as the percentage of articles that referred to the EU. The tone is measured as a combination of a measure of the sentiment of the title of the article and a measure of the average sentiments of the EU sentences in the article itself. Positive values represent a positive tone, and values below zero represent a negative perception (see also Chapter 3). The results are displayed in Fig. 7.1.

Fig. 7.1

EU salience and tone in Belgian media

The results in Fig. 7.1 show that, in all election years, there is more attention for the EU in the Flemish newspaper (i.e., De Standaard) than in the Walloon newspaper (Le Soir). However, there is substantial variation in EU salience between the years. In Flanders, EU salience is highest in 2014, while it was highest in Wallonia in 2010. In general, there was an upward trend in salience starting from 2003 on. 2019 breaks this trend, however, with the lowest EU salience in the series. This could partly be explained by the fact that this year the regional and national elections were held on the same day as the European election and received the most media attention. However, interestingly, also in 2014 these elections were held on the same day, and this year we note the highest salience numbers. Looking at the tone, a downward trend can be observed. While the worst tone was observed in 2010 in both regions, the last two elections under investigation scored lower than the first two. Furthermore, while the Flemish newspaper scores highest in the whole series, in 2007, in all other years, the Walloon paper is somewhat more positive overall.

Second, I look at the salience and tone of the interventions by politicians of different parties during parliamentary debates. Here as well the salience measure refers to the percentage of interventions referring to the EU, while tone reflects the average tone of all sentences about the EU in a given speech. The results are displayed in Fig. 7.2. The top part of Fig. 7.2 shows the salience for Flemish parties and the bottom part of the Walloon parties.²

Fig. 7.2

Salience in parliamentary debates in Flanders (top figure) and Wallonia (bottom figure)

Looking at the salience of the EU in Flanders, what stands out is the very high salience for the nationalist party N-VA and liberal party Open VLD in 2007. This year, national elections were held in Belgium, and the outgoing prime minister, Guy Verhofstadt, would hold the first position on the European list and won more than half a million preference votes in the European election two years later. It is likely that his European ambitions, which he pursued ever since, also led him to attach importance to the European issue—something that the ideologically close N-VA had to follow in reaction to this increased attention. Do note, however, that these numbers are based on a rather low number of speeches, so they might be somewhat inflated. What also stands out is the recent high attention for the EU of the two radical parties. First, the radical-left PTB-PVDA scores are constantly high. Second, since 2015, the radical-right Vlaams Belang (VB) has the most attention for the European issues of all Flemish parties—although this is mostly because of its criticism of the European project (see below).

The bottom part of Fig. 7.2 depicts the same information but for the Walloon parties. A first observation is that the EU is generally less salient—as can be seen in the lower range of the y-axis. Among those parties, however, it is especially the ecologists that talk most about the EU, although this decreases somewhat in the last 10 years. The liberal MR, on the other hand, strongly increased in attention to the EU in the last years. PTB-PVDA, which scored lowest among the Flemish parties, scores relatively high among the Walloon parties, which again shows the large difference in salience for the EU overall.

Next, I look at the tone with which the EU is discussed during these parliamentary interventions. The results are displayed in Fig. 7.3.

Fig. 7.3

Tone in parliamentary debates in Flanders (top figure) and Wallonia (bottom figure)

The top graph in Fig. 7.3 shows that Flemish parties are generally positive about the EU. The highest score (i.e., most positive) is reached by the green party in 2004, although this party scores rather average in the other years. Again, the radical parties stand out to some extent, with Vlaams Belang scoring the lowest in some recent years, and PTB-PVDA being very critical in 2017, although they score just above zero in 2019. The Walloon parties started out very positive two decades ago, but more recently they converged as well around a slightly positive tone. Especially the socialist PS stabilised on positive scores after a more volatile period in which they were also critical. Mostly the MR seems to be positive in recent years, while also here the low scores of the PTB-PVDA stand out. Overall, however, Belgian political parties are generally positive about the EU, and only the radical parties have been most strongly negative in the last years.

Modelling EU Issue Voting in 2019

The results above imply that, overall, there is a positive consensus about the European level among most Belgian political parties, but generally the European level is not very salient in national elections. Does this mean individual voters do not take into account the EU issue when they vote? As the main test of EU issue voting in the Belgian national elections of 2019, individual-level survey data are used. More specifically, I use the data of the post-electoral wave of the MAPLE project, gathered in an online survey among Belgian respondents after the elections of 24 May 2019. Sampling was based on gender, age, educational level, and NUTS1 regions.

The dependent variable in the models indicates the party voted for in the election at the national level. Respondents could indicate their vote choice, and the vote for the major parties will be modelled.³ The main independent variable measures the respondent’s opinion about further European integration: ‘Some people believe that the process of European integration should move forward to the creation of the United States of Europe. Others believe that the European Union should be dissolved in order to return to a situation in which states are fully sovereign. At which point of the following scale would you place yourself?’ Respondents could indicate their opinion on a scale ranging from 0 (“The EU should be dissolved”) to 10 (“The EU should move towards the United States of Europe”).

Other possible determinants of the vote are included as well. First, a question asking the respondent’s opinion about immigration is included, as this was a highly salient issue in Belgian politics in the years before the elections—and even led the largest Flemish party, N-VA, to quit the government some months before Election Day (Pilet, 2020). In the survey, respondents could indicate their opinion on the statement that Belgium should implement a more restrictive immigration policy on a scale from 1 (“strongly disagree”) to 5 (“strongly agree). The values on this scale have been reversed so that higher values indicate a more pro-immigration opinion. As a more traditional issue, people’s opinions on the statement that “women should have the right to access to an abortion in all circumstances”, measured on the same scale, is included as well. As radical parties performed well in the 2019 elections, two measures of support for the political system are included as well. The first is a measure of political trust, calculated as the mean indicated trust in the country’s National Government and National Parliament on a scale from 0 (“no trust at all”) to 10 (“complete trust”). The answers to these two questions are taken together as they correlate substantially (Pearson’s r = 0.87). A second measure of system support is satisfaction with the way democracy works in Belgium.

Besides these more specific issues, a measure of ideology is included as well. To this end, I include the respondent’s self-placement on the general ideological continuum in which 0 refers to a left position and 10 to a right position. Previous research has shown that people can interpret this scale meaningfully and that it leads to a useful summary of the ideological stances of voters vis-à-vis parties (Dalton et al., 2011; van der Eijk et al., 2005). Finally, in the first model, two more strong valence measures will be included: a retrospective sociotropic evaluation of the economy, which has been shown to influence voting for an incumbent party (Lewis-Beck & Stegmaier, 2000), and a rating of how much the voter likes the leader of the party they voted for—on a scale from 0 (“strongly dislike”) to 10 (“strongly like”).

Besides these main variables of interest, some control variables are included in the models as well. First, I control for the standard socio-demographics sex, age, and educational level. Sex distinguishes male and female respondents, with female respondents as the reference category. Age denotes the age of the respondent in the election year. Educational level is measured on a 9-point scale starting from early childhood education to doctoral or equivalent education. Besides these characteristics, I control for trade union membership status (0 = not a member; 1 = member) and religiosity (self-declared religiosity on a scale from 1 (“I have no religious believes”) to 4 (“I am very religious”)).

As the dependent variable consists of several categories without a logical ordering (i.e., party choice), I estimate multinomial logistic regression models to investigate the impact of the different issues and control variables on the vote. To take into account the different scales on which the variables are measured, all continuous measures have been rescaled to range between 0 and 1. Given that almost all parties operate exclusively in one of the two main regions, there are basically two different party systems, split along the language border (Deschouwer, 2012). Therefore, the analyses will be presented for Flanders and Wallonia separately. In Flanders, the Christian-Democratic CD&V will be included as a baseline category as it was one of the two Flemish incumbent parties (together with the liberal party Open VLD) and it has the most centrist ideological position. For the analysis in Wallonia, the single Walloon incumbent party MR (liberals) will be included as the baseline category.

In the first set of analyses, I will look at the direct effects of the variables of interest on the vote. In a second analysis, I take into account the challenging setting Belgium offers for voters who want to express their issue stances with their vote. As explained above, the compulsory voting system in Belgium also draws the less politically interested to the ballot box. Previous research has shown that less sophisticated voters tend to be less consistent in their attitudes and are less able to connect their political opinions with a vote that expresses these interests (Converse, 1964; Lau et al., 2014; Zaller, 1992), and this has also been found to matter in the elections under investigation specifically (Gallina et al., 2020). Therefore, I include an interaction of the measure on the EU issue with political sophistication, to test whether the more sophisticated voters take into account the EU issue to a larger extent when they vote. Sophistication is measured by an indicator of media use, measuring how many times per week the respondent uses the newspaper for political information (ranging from 0 “never” to 8 “7 days per week”).

Before turning to the results, one important note that needs to be made is that the number of observations in the dataset is low. While there is some attrition between the first survey wave—held before the election—and the second wave after the election (of main interest here), the number of observations for some parties drops more given that the respondents are divided over the two main regions, and then the many parties within those regions. This results in some cells that are too small to make reliable estimates. Therefore, two sets of results will be discussed: one focusing on the vote choice as reported after the election, and one focusing on the vote intention as reported in the first survey wave, gathered some months before the elections on a larger sample. With this dual strategy, the first wave is used in its advantage of containing more observations, and the second as it is a post-electoral study. In this way, the conclusions from both sets of analyses reinforce each other. In terms of reporting the results in more detail, the focus here will be on the post-electoral study, while the results using the vote intention are reported in Appendix 2.

Results

Was the EU an important issue when voters cast their votes for the national level in 2019? To investigate this question, I estimate multinomial models explaining the vote. As the coefficients of multinomial models are complex to interpret and depend on the chosen baseline outcome, I present the results by plotting the average marginal effects of the main variables of interest in Fig. 7.4 (Flanders) and Fig. 7.5 (Wallonia). The full tables with results are included in Appendix 1.

Fig. 7.4

Explaining the vote in Flanders (Note Average marginal effects and 95% confidence intervals, based on the results reported in Model 1 in Table 7.1)

Fig. 7.5

Explaining the vote in Wallonia (Note Average marginal effects and 95% confidence intervals, based on the results reported in Model 1 in Table 7.2)

The results in Fig. 7.4 show that the different variables of interest help to explain the vote choice between Flemish parties. First, looking at the European issue, this indicator only shows modest effects on the vote. When examining the recalled vote after the election, there is a significant positive effect for the Christian-Democrat CD&V, where voters are 14 percentage points more likely to vote for this party when they believe that European integration should go further than when they oppose to the EU. This result is not confirmed when looking at the intention to vote before the campaign, but in this model, there is a small positive effect of the liberal Open VLD, and a negative effect of almost 11 percentage points for radical-right Vlaams Belang. This result could be expected, as Open VLD is strongly present at the European level by means of Guy Verhofstadt, while Vlaams Belang is openly opposed to further European integration (Almeida, 2010; Marks et al., 2007).

The other variables of interest also show some interesting patterns. Positive attitudes about immigration positively predict support for the radical-left PVDA, and negative for the Christian Democrats (vote) and nationalist party N-VA and Vlaams Belang (vote intention). Being in favour of abortion is negatively associated with support for the Christian Democrats, which is in line with its customary call for traditional family values. As was expected, the measures of support for the system—political trust and satisfaction with democracy—are strongly negatively related to supporting Vlaams Belang, while they are positively related to supporting the traditional parties. Finally, the general left–right continuum performs well in the model, with most left-wing voters supporting the radical left and social democrats, and the most right-wing voters finding their way to the nationalists and radical right. While the Christian Democrats are commonly assumed to hold the centre position, in this case, their voters score somewhat more on the left side of the political spectrum.

In Fig. 7.5, the same analysis is repeated, but this time for Walloon parties. Also here, we find some significant effects of the EU issue on the vote. The largest effect is the negative association between favouring more European integration and the vote for the socialist party PS. Voters supporting a fully integrated “United States of Europe” are 33 percentage points less likely to vote for the PS than voters who want to dissolve the EU. This is rather surprising, as the party does not have a strong anti-EU narrative. Possibly, voters remembered the PS’s strong stand against the CETA agreement between the EU and Canada from some years before. Contrarily, there is a sizeable positive effect for supporting Christian Democrats cdH (vote) and the green party ECOLO (vote intention).

The other issues, abortion and immigration, add less to explaining the vote than they did in Flanders. The only significant association is the positive effect of positive perceptions of immigration on the intention to vote for ECOLO. Political trust and satisfaction with democracy also only help explaining the intention to vote—the former positively for liberal incumbent MR and negatively for its main challenger PS; the latter negatively for radical-left PTB. Finally, as in Flanders, the general left–right continuum is an important determinant of the vote, with the expected negative associations with the radical left and socialists, positive for the liberals, with a moderate position for the Christian Democrats. Somewhat contrary to its commonly attributed left-wing position, there is no significant association with supporting the ecologist party.

The results above show mixed results regarding the EU issue: while there are some significant effects for some parties, these are not consistent and substantially rather small. However, it is possible that there is individual-level heterogeneity. More specifically, I test whether more sophisticated voters are better able to express their issue opinions with their vote, by including an interaction between the EU issue and media use. The results are summarised in Table 7.3 in Appendix 1 for Flanders and Table 7.4 for Wallonia. The replication using the vote intention from the first survey wave, is reported in Appendix 2, and leads to the same conclusions as the analyses reported here.

The results do not show strong support for a moderation effect to be at play. Only in Flanders, there is a significant negative interaction between the EU issue and media use for voting for Vlaams Belang compared to the baseline CD&V. This indicates that voters who are critical of the EU become less likely to vote for Vlaams Belang at higher levels of political sophistication—and indeed supports the assertion that higher sophisticates express their issue opinions more strongly with their vote. To get a full overview of the interaction effects, Figs. 7.6 and 7.7 display the average marginal effects of the EU issue at different levels of media use for the different parties respectively.

Fig. 7.6

Average marginal effects of the EU issue at different levels of media use: Flanders (Note Average marginal effects and 95% confidence intervals, based on the results reported in Table 7.3)

Fig. 7.7

Average marginal effects of the EU issue at different levels of media use: Wallonia (Note Average marginal effects and 95% confidence intervals, based on the results reported in Table 7.4)

The results in Fig. 7.6 show a more nuanced picture of the effects of the EU issue on the vote. For Christian Democrat voters, there is a positive effect of favouring further European integration on the vote, but only from a certain level of political sophistication. The same counts for Vlaams Belang, where the negative effect manifests itself among its more sophisticated voters. These results are in line with those presented in Fig. 7.4 but show that the main mechanism is strongest for voters at higher levels of sophistication. These results provide further evidence that the Belgian system of compulsory voting might inhibit EU issue voting in national elections.

Figure 7.7 shows the same results for the Walloon parties. Also here there is a positive effect of the EU issue on voting for the Christian Democrats, although it only manifests itself at lower levels of sophistication—where the number of observations is higher and the uncertainty around the estimates is generally lower, also for the other parties. The results also show that the surprising negative association between the EU issue and voting for the PS is limited to its least sophisticated voters. At higher levels of sophistication, the association is not significant.

Conclusion

At first sight, Belgium seems to provide an ideal setting to find strong effects of the European issue on political behaviour. With the main European institutions in the centre of the country, the European Union is omnipresent in Belgium. Yet, however, when looking at media and Parliamentary data, while media and parties are generally positive about the EU, it has a very low salience. Furthermore, in 2019 specifically, its salience was substantially lower compared to the previous elections. Low partisan conflict and low salience are considered factors that strongly inhibit EU issue voting (de Vries, 2007). Furthermore, Belgium’s simultaneous elections create a challenging setting to find evidence for EU issue voting in national elections, as voters can express their specific European concerns with another vote almost instantaneously. Previous research has also indicated that those voters who split their ticket between the different levels mostly do so between the national and European levels, casting the latter vote with a European issue in mind. This context, together with the Belgian system of compulsory voting also drawing the least politically interested voters to the voting booth, makes Belgium a least-likely case to find evidence of EU issue voting in national elections. This would lead to a rather pessimistic normative evaluation of the state of democracy for the European representation in Belgium: as the representatives on the national level play an important role on the European level, it is important that people choose which party to support in the national election with the European issue in mind.

To some extent, the results corroborate that Belgium is a hard case to test the EU issue voting on the national level, as there is only limited evidence for the EU issue to be at play in voting behaviour for the national level. In Flanders, the strongest and most consistent finding is the negative association between pro-European opinions and support for the radical-right party Vlaams Belang. This result could be expected, given the party’s open objection against further European integration (Marks et al., 2007), which increases the politicisation and thus salience of this issue for this party (Hutter & Kriesi, 2019). This association is especially prominent among voters with higher levels of political sophistication, who are better able to bring their vote in line with their political opinions. In Wallonia, pro-European voters seem to be more likely to vote for the Christian Democrats.

Importantly, it needs to be noted that there are no consistently strong effects of the other issues included in the models either. Besides the general left–right ideological continuum and, to some extent, measures of support for the system, the models have low explanatory power when it comes to explaining the vote choice. This shows the complexity of the Belgian party system, which is structured on many different cleavages and has a complex history. Hence, when putting the results of the EU issue in the context of the findings of the other issues included in the models, the conclusion needs to be that its effects are comparable to those of more traditional issues on the national level: it helps to explain the vote for some but not all parties, varying in size. This provides further evidence for the claim that the politicisation of the European issue is indeed growing to be a new structuring conflict next to existing cleavages (Hutter & Kriesi, 2019). There is most consistent evidence for a positive association with support for the Christian-Democratic parties. These parties traditionally hold the most moderate ideological position overall in Belgium, and it is possible that this pushes voters to consider a wider range of issues than for other parties campaigning strongly on one or a small set of traditional national issues. Overall, however, it seems like Belgian voters support different parties for a variety of reasons, and there is no one specific issue that primarily drives voters’ decisions.

Appendices

Appendix 1: Full Tables of Models Presented in the Text

See Tables 7.1, 7.2, 7.3, and 7.4.

Table 7.1 Multinomial regression models explaining the vote in Flanders

Table 7.2 Multinomial regression models explaining the vote in Wallonia

Table 7.3 Multinomial regression models explaining the vote in Flanders including interaction effect

Table 7.4 Multinomial regression models explaining the vote in Wallonia including interaction effect

Appendix 2: Replication of the Results Using Vote Intentions (Survey Wave 1)

See Figs. 7.8, 7.9, 7.10 and 7.11.

Fig. 7.8

Explaining the vote intention in Flanders (Note Average marginal effects and 95% confidence intervals)

Fig. 7.9

Explaining the vote intention in Wallonia (Note Average marginal effects and 95% confidence intervals)

Fig. 7.10

Average marginal effects of the EU issue at different levels of media use: Flanders—vote intention (Note Average marginal effects and 95% confidence intervals)

Fig. 7.11

Average marginal effects of the EU issue at different levels of media use: Wallonia—vote intention (Note Average marginal effects and 95% confidence intervals)