Ambivalent Europeans? A study about the effects of cognitive and affective factors on ambivalent attitudes towards the European Union.

11 June 2019

Ambivalent Europeans?

A study about the effects of cognitive and affective factors

on ambivalent attitudes towards the European Union.

Name: Lejla Selimovic

Student number: S1396595

Master: Public Administration

Track: Public Management and Leadership

Supervisor: Dr. D. D. Toshkov

Second reader: Dr. B. J. Carroll

Table of contents

1. Introduction

Citizens’ support for the European Union is more relevant then ever. The times of permissive consensus have been replaced by constraining dissensus (Hooghe & Marks, 2009, p. 5). The public questions more often the European integration project through elections, referendum campaigns and the media (De Vries, 2013, p. 435). Does this mean that citizens are either pro or anti the European Union? Scholars indeed viewed public support for a long time as a fixed attitude. However, over the last decade, several scholars have challenged this approach and argued that attitudes are rather inherently variable or even ambivalent than fixed (Stöckel, 2013; De Vries & Steenbergen, 2013; De Vries, 2013).

Moreover, public support in the context of the European Union can be divided into regime support and policy support. Regime support refers to citizens’ support for the constitutional settlement of the EU. This type of support is crucial for the legitimacy of an organization. Regime support can be measured by looking at the support for membership of the EU or by looking at the amount of trust that citizens have in EU institutions. On the other hand, policy support refers to citizens’ support for the content of specific decisions and actions of EU actors. In other words, this type of support is focused on the amount of citizens’ support for decisions within a specific policy area (Hobolt & De Vries, 2016, p. 416.). Connecting this to the approach of ambivalent attitudes, it is imaginable that citizens do support the EU in general but disagree on the way certain policies are handled.

Current studies most often use regime support or policy support in order to measure general public support on European integration (Hobolt & De Vries, 2016, p. 426). However, no attention has been given to ambivalence between regime and policy support. Therefore, this research focuses on which factors influence the occurrence of ambivalent attitudes between regime and policy support. A distinction is made between cognitive factors, such as knowledge, and affective factors, which are factors based on the feelings that someone has towards an object (Stöckel, 2013). Given this, I will research the following question:

What are the effects of cognitive and affective factors on ambivalent attitudes towards the European Union?

This question is relevant for the science of public administration. Current studies have already proven that ambivalent attitudes occur more often (De Vries, 2013). As a matter of fact, research has shown that regime support is not always accompanied by policy support (Hobolt & De Vries, 2016, p. 419) and that cognitive and affective factors influence ambivalence between how people view (positively/negatively) the EU and regime support (Stöckel, 2013). However, we do not know whether these factors also influence ambivalence between regime and policy support. So, this study contributes to the state of the art by connecting the existing theories regarding ambivalence in order to test whether they also apply to the occurrence of ambivalent attitudes between regime and policy support.

Moreover, this study has societal relevance. Research from the US has proven that ambivalent citizens are more driven by salient considerations and are more sensitive to persuasion (Lavine et. al, 1998). Eventually, this leads to less predictability of public support (Steenbergen & Brewer, 2004). Predictability of public support is important for EU decision-making. The first step in increasing predictability is to know which factors influence the occurrence of ambivalence. Therefore, this study is also useful for the society.

The Standard Eurobarometer survey will be used to test the effect of cognitive and affective factors on ambivalent attitudes. Scholars often use this survey to research public support towards the EU (Hobolt & De Vries, 2016, p. 416). The dataset includes a representative sample of approximately 1000 respondents per country. This research will only focus on the 28 EU Member States. I will perform a logistic regression to test the effect of cognitive and affective factors on ambivalent attitudes.

The next chapter will elaborate more on the current knowledge that we have regarding public support and ambivalence. In chapter 3, I will focus on the cognitive and affective factors and I will formulate the hypotheses. This is followed by chapter 4, in which the research method, the used data and operationalization of the variables are explained. Further, in chapter 5, I will present the results and analyze them. The final chapter will summarize the findings and discuss the limitations of this research.

2. Literature review

Public support for European integration has proven to be more important in the last few decades. There is a noticeable shift from permissive consensus, which refers to the period wherein the public blindly accepted EU actions, to constraining dissensus, where the public voices their objections when necessary (Hooghe & Marks, 2009, p. 5). This shift can also be seen in the focus of the academic literature on European integration. The early studies focused mainly on support for the EU (Inglehart, 1970; Gabel & Palmer, 1995; Sánchez-Cuenca, 2000), whereas in the last decade scholars have given more attention for Euroskepticism (De Wilde & Trenz, 2012; Hooghe 2007; Hakhverdian et. al, 2013). This division insinuates that citizens are clear-cut pro or anti European integration. However, some recent studies have disputed such a clear-cut division and suggest that public support includes differential degrees of ambivalence (Stöckel, 2013; De Vries & Steenbergen, 2013; De Vries, 2013) and that a multidimensional approach is necessary (Boomgaarden et. al, 2011). This literature review will first focus on the different approaches of public support, followed by an explanation of ambivalent attitudes and their consequences.

2.1 Public support on European integration

How can public support be defined? Easton (1975) mentions that public support ‘refers to the way in which a person evaluatively orients himself to some [political] object through either his attitudes or his behavior’ (p. 436). Moreover, he divides public support into diffuse support, which refers to ‘what the object is or represents, not what it does’ (Easton, 1975, p. 444) and specific support, which refers to the support for specific decisions and actions taken by actors. While reviewing the literature on public support for European integration, I noticed that scholars conceptualize public support in different ways, but it can often be traced back to Easton’s definition and distinction of public support. These different conceptualizations of public support can be divided in three groups.

The first group of scholars focuses on diffuse support for European integration (Macarena, Ceka & Kriesi, 2017; Zuzana, 2015). Another word for this that is used within the public support for European integration literature is regime support. Regime support refers to citizens’ support for the constitutional settlement of the EU.

This type of support is crucial for the legitimacy of an organization. Scholars measure regime support by looking at the support for membership of the EU or by looking at the amount of trust that citizens have in EU institutions (Hobolt & De Vries, 2016, p. 416; Bolstad, 2014; Kritzinger, 2003).

Moreover, there is a group of scholars that is more focused on policy support, which is quite similar to Easton’s specific support. Policy support refers to public support for the content of specific decisions and actions of EU actors. In other words, this type of support is focused on the amount of citizens’ support for decisions within a specific policy area (Hobolt & De Vries, 2016, p. 416.), such as support for the euro (Hobolt & Wratil, 2015) or the European economic governance policy (Kuhn & Stöckel, 2014). However, it is quite difficult to disentangle support for national policies and European policies, due to citizens’ limited knowledge of EU policies (Hobolt & De Vries, 2015, p. 416).

Furthermore, some scholars argue that a multidimensional approach is more appropriate when researching public support for European integration. For example, Boomgaarden al. (2011) are clear advocates of this approach because they suggest five dimensions of public attitudes. Moreover, Hobolt and Brouard (2011) emphasize the multidimensionality of public attitudes towards European integration in their research on why French and Dutch voters rejected the European Constitution.

2.2 Ambivalent Europeans

Survey responses have been used for almost eight decades to research public opinion. For a long time these responses have been seen as fixed attitudes with random measurement error due to the imprecision of survey items. In other words, the main assumption was that an opinion is a fixed point on a continuum and that all variation within responses was due to measurement errors (Achen, 1975; De Vries & Steenbergen, 2013, p. 123; De Vries, 2013, p. 442).

However, many public opinion scholars have shifted away from this view and argue that public attitudes are inherently variable. Several researchers from the US have shown that individuals often have ambivalent considerations towards a political object. First, Cacioppo et al. (1997) introduced the evaluative space model. With this model they argue that individuals are ambivalent and that an attitude includes several points that characterize the intensity of positive and negative attitudes. This rejects the

assumption that a complete attitude is based on one single point. Secondly, Zaller and Feldman (1992) conceptualized attitudes as distributions of considerations in which the considerations vary in their evaluative implications, with the possibility of considerations having opposing evaluative implications. This again refers to ambivalence. Closely connected to the view of Zaller and Feldman is the theory of Alvarez and Brehm (2002), who argue that citizens’ predispositions can be conflicting or reconciled. Moreover, they mention that a lack of relevant information provokes uncertainty, which in turn increases conflicting (in other words, ambivalent) predispositions.

Nowadays, several scholars also argue that support for the European integration is not fixed and clear-cut pro or anti, but rather, ambivalent (De Vries & Steenbergen, 2013; De Vries 2013; Rose & Borz, 2017). In other words, scholars argue that public support can simultaneously endorse positive and negative evaluations on European integration (Stöckel, 2012, p. 23-24). For example, citizens can be positive about the existence of the European Union, which refers to regime support, but they may object certain policies, which refers to policy support. Despite this shift in the view of public support, little empirical research has been done on ambivalent support in the European context. But what do we already know on this subject? De Vries (2013) has, for example, shown that ambivalence is higher in Western Europe than in Central or Eastern Europe, which is likely the result of the greater experience of positive and negative effects of European integration. Moreover, Stöckel proved that it is important to make a distinction between indifferent and ambivalent support towards the European Union (Stöckel, 2012, p. 34).

2.3 Consequences

What are the consequences of ambivalent attitudes? Due to the little knowledge that we have on ambivalence towards European integration, we have to look at the knowledge retrieved from studies in the US to answer this question. First of all, citizens with ambivalent attitudes are easier to persuade and they are more driven by salient considerations (Lavine et al., 1998), resulting in a decrease in the predictability of public support (Steenberger & Brewer, 2004). Moreover, research has proven that ambivalence results in less stable support over time, less certainty (Zaller, 1992), and increases the importance of available information (Stöckel, 2012, p. 25).

This study contributes to the current literature by researching what factors influence ambivalence between regime and policy support. The state of the art literature has already shown that the public support on European integration is ambivalent by using the heteroskedasticity in a regression model as an indicator for ambivalence (De Vries, 2013; De Vries & Steenbergen 2013). Moreover, ambivalent attitudes towards the EU are researched by measuring what the EU means to respondents. If the respondent assigned simultaneously positive (e.g. peace, social protection and freedom to travel) and negative (bureaucracy, more crime and waste of money) meanings to the EU, then this respondent was considered ambivalent (Stöckel, 2012, p. 30). However, more research is necessary to fully understand how ambivalent attitudes occur, especially because ambivalent attitudes are on the rise (De Vries, 2013). We already know that high levels of regime support are not always accompanied by support for further integration or support for a specific policy (Hobolt & De Vries, 2016, p. 419). Therefore, this research focuses specifically on the factors that influence ambivalence between regime and policy support on European integration.

3. Theoretical Framework

As mentioned before, ambivalent support refers to having simultaneously positive and negative attitudes towards an object (Stöckel, 2012, p. 23-24). This study focuses specifically on ambivalence between regime support and policy support. Therefore, an ambivalent citizen is someone who supports the existence of the EU (regime support) on one hand, and feels negatively about further EU integration or a specific policy (policy support) on the other hand. Moreover, the converse is also possible, thus, a citizen who does not support the regime but does support specific policies is also considered as ambivalent. However, I want to make the important note that the hypotheses in this thesis are based on the assumption that regime support is generally higher than policy support (Hobolt & De Vries, 2016, p. 419).

The literature provides different factors that influence the occurrence of ambivalent attitudes. This theoretical framework divides the factors into cognitive and affective factors. This division builds on the approaches of Stöckel (2012) and De Vries and Steenbergen (2013). The reason for this is that social psychology has shown that attitudes are formed by cognitive assessments and affective sentiments. First, I will discuss the underlying mechanisms of the cognitive factors, followed by a description and explanation of the affective factors.

3.1 Cognitive factors

Being politically informed is considered as a cognitive factor that influences ambivalent attitudes. Citizens use the information and knowledge that is available to them in order to make up their mind about the EU. However, there are conflicting explanations of the role of political information on the occurrence of ambivalent attitudes. Some scholars argue that a lack of information provokes uncertainty, which can increase ambivalence (Alvarez & Brehm, 2002). While other scholars argue that political sophistication, which refers to how much a certain citizen pays attention to politics and to which extent they understand the political information, decreases ambivalent attitudes (Zaller, 1992). Moreover, De Vries (2013, p. 452) argues that paying attention to political news decreases the level of ambivalent support towards European integration, while Stöckel (2013, p. 36) argues the exact opposite. In addition, De Vries and Steenbergen (2013) mention that political knowledge is not relevant at all for response variation. These differences could be attributable to the

different research designs. I would like to test the effect of political knowledge on ambivalence between regime and policy support. Hereby, I would like to follow De Vries (2013) and focus on two sources, namely the level of attention to the news and the political knowledge a certain citizen has. I expect that having more knowledge and following the news leads to an increase in ambivalence between regime and policy support. Well-informed citizens will often receive ambivalent information, which could lead to having ambivalent attitudes. To be more specific, I think that a well-informed citizen can have better reasons why they do not support a specific policy area while they do support the regime. Therefore, I propose the following hypotheses:

H1: Political knowledge increases ambivalence between regime and policy support. H2: Following the news media increases ambivalence between regime and policy support.

3.2 Affective factors

Affective factors influence the level of ambivalence towards the EU. These factors are easier to measure because they are closely linked to a citizens’ subjective opinion, while measuring someone’s knowledge is more difficult (Lavine et al., 1998). Affective factors refer to whether a citizen feels attached to the EU (Stöckel, 2012, p. 29), and identifies with the EU (De Vries & Steenbergen, 2013, p. 128). Citizens can, for example, identify themselves with the EU, with their own country, but also with both. Previous research has shown that dual identities - so identifying with both - provide conflicting cues. On one hand, citizens want to preserve their national autonomy and they view European integration as a threat, but on the other, they also identify with the EU (Cinnirella, 1997). However, some evidence suggests that it works the other way around. Having dual identities does not necessarily mean that these are conflicting identities (Brewer, 2001), especially if European integration is not seen as a threat to the member state (Cinnirella, 1977). De Vries and Steenbergen (2013, p. 136) have even shown that having a dual identity results in a reduction in response variation.

Moreover, the level of attachment towards the EU is an affective factor that has proven to influence the occurrence of ambivalent attitudes. Attached citizens are

citizens that have strong feelings and that care about an object. These feelings are likely to decrease ambivalent attitudes. Evidence has shown that feeling attached to the EU decreases the level of ambivalence (Stöckel, 2012, p. 39). Moreover, I am interested in whether it also decreases ambivalence between regime and policy support.

I expect that having affective factors will decrease ambivalence between regime and policy support. Citizens with dual identities and who feel attached to the EU will be more likely to support regime and policy support and, therefore, be less ambivalent.

H3: Feeling attached to the EU decreases ambivalence between regime and policy support.

H4: Having dual identities decreases ambivalence between regime and policy support.

3.3 Cognitive vs. affective factors

Furthermore, affective factors are more central in belief systems than cognitive factors (Lavine et al., 1998; Brandt et al., 2018). In other words, feeling attached to the EU is superior compared to, for example, political knowledge when forming a political opinion. Hypotheses 1 and 2 measure cognitive factors, while hypotheses 3 and 4 refer to affective factors. I expect that, in general, the factors discussed in hypotheses 3 and 4 better explain ambivalence than the factors discussed in hypotheses 1 and 2. After all, citizens with no knowledge form their opinion based on the affective factors. Moreover, I expect that strong affective factors can frustrate the ambivalent affect of conflicting cognitive factors. Therefore, I propose the following:

H5: Affective factors are better able to explain ambivalence between regime and policy support than cognitive factors.

4. Methods, data and operationalization

The data that will be used to test the hypotheses are retrieved from the Standard Eurobarometer wave 86.2. Scholars who are interested in public support on European integration often use the Eurobarometer surveys (Hobolt & De Vries, 2016, p. 416). The data from Eurobarometer 86.2 was collected in November 2016 by face-to-face interviews and covers the population of all 28 Member States. Moreover, the data includes a representative sample of approximately 1000 respondents per Member State. I choose this dataset because it includes questions that operationalize all the variables that are mentioned in the hypotheses, while more recent Eurobarometer surveys do not include questions about the usage of news media.

4.1 Dependent variable

Ambivalence has been measured in several ways. Within psychology, ambivalent attitudes are usually accounted for by using two measures to see whether an individual has simultaneously positive and negative attitudes towards an object (see for example, Cacioppo et al., 1997). Moreover, scholars that focus on ambivalent attitudes towards the politics of the US use items from the American National Election Study (ANES). The ANES taps the number of positive and negative things a respondent mentions about a candidate (Stöckel, 2012, p. 30). However, the Eurobarometer survey lacks similar items. Therefore, studies that already have been conducted on ambivalent attitudes towards the EU use creative ways to measure ambivalence. For example, Steenbergen and De Vries (2012) used the heteroskedasticity in a regression model as an indicator of ambivalence. However, this is not the most ideal way to measure ambivalence. Heteroskedasticity can also occur for reasons other than ambivalence, such as an omitted variable. Moreover, heteroskedasticity does not refer to individual observations but to a model as a whole. Therefore, it is not possible to say something about which respondents are ambivalent or how many Europeans are ambivalent. Stöckel (2012) circumvents these limitations by measuring what the EU means to respondents. If the respondent assigned simultaneously positive (e.g. peace, social protection and freedom to travel) and negative (bureaucracy, more crime and waste of money) meanings to the EU then this respondent was considered ambivalent (Stöckel, 2012, p. 30).

The dependent variable in this study is the ambivalence between regime and policy support. Regime support is measured by the question of whether the respondents trust the European Union. This is an often-used operationalization of regime support (Hobolt & De Vries, 2016, p. 416). If a respondent trusts the European Union, then this means that the respondent supports the regime. Moreover, policy support is measured by asking the respondents: What is your opinion on each of the following statements? Please tell me for each statement, whether you are for it or against it.

1. A European economic and monetary union with one single currency, the euro. 2. A common foreign policy of the 28 Member States of the EU.

3. Further enlargement of the EU to include other countries in future years. 4. A common defence and security policy among EU Member States. 5. A free trade and investment agreement between the EU and the USA. 6. A common European policy on migration.

7. A common energy policy among EU Member States. 8. A digital single market within the EU.

9. The free movement of EU citizens who can live, work, study and do business anywhere in the EU.

The policy support is measured on a 9-point scale index. The score of respondents on this scale are measured by summing up all the statements that the respondent supports. I maintained a strict view regarding whether respondents support policies. In other words, if respondents choose not to answer the question, I assumed that they do not support it.

I have created four dummy variables that show the ambivalence of a respondent. If a respondent has a positive attitude towards the regime and supports the policies, then this respondent is considered as consistent. If a respondent has a positive attitude towards the regime, but a negative attitude towards the policies, then this respondent is considered as ambivalent. The four dummy variables vary on what the threshold on the policy index is. I will test the hypothesis and base the analysis mainly on the dummy variable at which the respondents scored for all the statements. In the first dummy variable, the respondents have to be for all the statements. In the second dummy variable, it is sufficient to have a threshold higher than or equal to 8

on the policy-index. The third dummy variable has a threshold of at least 7 statements, and the fourth, of 6 statements. This way I can also analyze whether there is a difference and how large the difference is in results regarding several thresholds on the policy-index.

4.2 Independent variables

The first hypothesis focuses on political knowledge. I use objective knowledge to measure the political knowledge of respondents (Stöckel, 2012, p. 32). Objective knowledge is measured by testing the knowledge of respondents about the EU by asking whether the following three statements are true or false.

1. The EU currently consists of 28 Member States.

2. The members of the European Parliament are directly elected by the citizens of each Member State.

3. Switzerland is a Member State of the EU.

Respondents who provide the correct answer to all the three statements are considered as respondents who have political knowledge.

Moreover, the second hypothesis states the expectation that following the news media increases ambivalence between regime and policy support. Whether a respondent follows the news on European political matters is operationalized by the question: Where do you get most of your news on European political matters? The answer options are: television, the written press, radio, websites, online and social networks. Additionally, the respondent could also mention spontaneously that they do not look for news on European political matters. I will use this question and make a dummy variable in which 0 refers to the answer that the respondent does not look for news on European political matters and 1 refers to one of the answer options.

Furthermore, hypotheses 3 and 4 are focused on affective factors that could influence the level of ambivalence between regime and policy support. I expect having dual identities to decrease ambivalence. Whether a respondent has dual identities is measured by the question: Do you see yourself as …?

1. Only own nationality

2. Own nationality and European 3. European and own nationality 4. Only European

The respondents that choose answer option two or three are the respondents that have dual identities.

The second affective factor that I expect to decrease ambivalence is the level of attachment to the EU. This is operationalized by the question: Please tell me how attached you feel to the European Union. The answer options are: very attached, fairly attached, not very attached or not at all attached.

4.3 Control variables

In order to ensure that I am measuring the effect of cognitive and affective factors, it requires isolating their effect by including control variables or, in other words, by including confounding factors in the analysis. I follow Stöckel (2012, p. 33) in the selection of control variables because his research is also focused on ambivalent attitudes towards the European Union. Therefore, I will add age and gender as control variables. These are the standard control variables that are used in the literature regarding European support. Moreover, I will include the countries as a categorical control variable in order to see whether this has an influence on the effect of cognitive and affective factors on ambivalent attitudes.

4.4 Method

The dependent variable is a binary variable. Therefore, I will test the hypotheses by doing a multiple logistic regression analysis in SPSS. I will run 4 analyses, since I will have four different dependent variables that vary on what the threshold on the policy index is. In model 1, I will only test the effect of the cognitive and affective factors. Moreover, in model 2, I will add the control variables that are focused on the individual (age & gender). In the last model, I will also include the country as a control variable in order to see whether this influences the effect of the cognitive and affective factors.

Moreover, in order to compare the effect of the cognitive and affective factors (hypothesis 5), I need to make sure that a change of 1 unit means the same thing on all the variables. However, the level of attachment has 4-answer options, which makes it impossible to test hypothesis 5. In order to be able to make such a comparison, I will recode the level of attachment into a binary variable and run another logistic

regression. After this, I will be able to compare the absolute values of the coefficients, and thus, test hypothesis 5.

5. Analysis

In this section the results of the research will be analyzed. First, I will start by discussing the descriptive results. I will focus on the frequencies of cases within the dependent variable and on the distribution of cases regarding the specific policies. Moreover, I will show which countries have the most ambivalent citizens regarding the regime and policy support. After this, I will discuss the results of the logistic regression and argue whether the expectations comply with the results. Furthermore, it is good to keep in mind that the analysis is based on the dependent variable with a threshold of 9 on the policy-index. I will mention it explicitly when I am discussing results with another threshold on the policy-index.

5.1 Descriptive results

As mentioned previously, I operationalized the dependent variable in a strict way. Only the respondents that support the regime and are for all of the 9 policies are considered as consistent. Moreover, this also works the other way around. One could argue that it is too strict to consider a respondent as ambivalent if he only does not support one policy. However, ambivalent support refers to having simultaneously positive and negative attitudes towards an object (Stöckel, 2012, p. 23-24). In my view, this is already the case of you are against one policy. Moreover, I believe that this is the best way to measure the effect of the independent variables on ambivalent attitudes.

Despite this choice, it is good to look at what this means for the distribution of the four possible categories within the dependent variable: respondents who support the regime and all policies, respondents who support neither and respondents who only support one of the two. Table 1 shows how the numbers of respondents within the dependent variable are distributed over these four categories. We can see that there is large difference in how the dependent variable is distributed if we look at the dependent variable where the policy-index has a threshold of 9 compared to the variable where the policy index has a threshold of 7 or higher.

Policy index threshold = 9 ≥ 7 Consistent

attitude

Regime support & policy support 2.781 _7.182

No support 12.525 _8.842

Ambivalent Regime support but no policy support 8.255 _5.008

But how is this distribution in the 28 Member States? Table 2 shows us the percentages of the four categories per country. Regarding the consistent attitudes, we can see that Lithuania, Romania and Portugal are the three countries that have relatively the largest population of citizens that support the regime and all the policies. The opposite are Denmark, Sweden and Austria. These countries have relatively the lowest percentage of citizens that support both: regime and all policies. Moreover, it is interesting to take a closer look at the ambivalent attitudes. We can see that Malta, Finland and Denmark have high percentages of citizens that do support the regime but are against for at least one policy area. More remarkable are Greece, Spain and Cyprus. In these countries at least 10% of the citizens is for all the policy areas but they do not support the regime in general.

Table 2: Percentage per country on the four possible categories

Country Regime support & _{policy support support} No Regime support but _{no policy support} Policy support but _{no regime support}

France 4% 69% 24% 3% Belgium 7% 50% 40% 3% The Netherlands 7% 55% 35% 3% Germany - West 7% 55% 35% 2% Germany - East 5% 68% 24% 3% Italy 12% 56% 23% 8% Luxembourg 6% 43% 45% 5% Denmark 3% 49% 46% 2% Ireland 15% 42% 36% 7% Great Britain 6% 64% 28% 3% Northern Ireland 6% 66% 26% 2% Greece 9% 65% 10% 15% Spain 13% 48% 24% 15% Portugal 20% 39% 33% 8% Finland 6% 46% 46% 2% Sweden 3% 55% 40% 1% Austria 3% 60% 36% 1% Cyprus (Republic) 11% 58% 21% 10% Czech Republic 5% 69% 24% 2% Estonia 10% 42% 44% 4% Hungary 16% 48% 30% 5% Latvia 14% 44% 37% 5% Lithuania 28% 28% 37% 7% Malta 19% 30% 48% 3% Poland 13% 44% 38% 4% Slovakia 16% 49% 30% 5% Slovenia 11% 51% 28% 9% Bulgaria 19% 39% 39% 2% Romania 23% 34% 34% 8% Croatia 13% 45% 33% 9%

Graph 1 shows that the citizens from Greece (74%), the Czech Republic (74%), East Germany (73%) and France (73%) have the most consistent attitudes with respect to regime and policy support. After a glimpse at table 2, the statement can be made that the consistency is mainly the result of the large amounts of citizens that fall into the category: no support. Opposite to this are Malta, Luxembourg, Denmark and

Estonia who have the most ambivalent citizens. It is notable that Malta and Luxembourg score 51% on ambivalent attitudes, meaning that they have even more ambivalent than consistent citizens, regarding attitudes towards regime and policy support.

Moreover, table 3 gives insights on how the attitudes of respondents are distributed regarding the specific policies. There are a few interesting results that table 3 provides. First, ambivalent attitudes occur the most on the policies regarding the future enlargement, the USA trade agreement, the energy and the free movement of citizens. A lot of people do not support the EU in general, but are for the free movement of citizens who can live, work, study and do business anywhere in the EU. Further, 8.991 respondents do not trust the EU, but they are for a common European energy policy. Another interesting result is that a large amount of respondents supports the EU in general but is against including more countries in the EU in the future. The same applies for the USA trade agreement: 3.368 respondents support the regime of the EU but are against the trade agreement.

The policies

Consistent attitude Ambivalent attitude Regime support & policy support No support Regime support but no policy support Policy support but no regime support 1. Single currency 8.265 6.706 2.771 7.144 2. Foreign policy 8.665 5.916 2.371 7.934 3. Future enlargement 6.024 9.408 5.012 4.442 4. Defence policy 9.361 4.590 1.675 4.590

5. USA trade agreement 7.668 7.019 3.368 6.831 6. Migration policy 8.496 5.698 2.540 8.152

7. Energy policy 9.190 4.859 1.846 8.991

8. Digital single market 8.249 6.370 2.787 7.480 9. Free movement of citizens 10.128 3.130 908 10.720 Table 3: Frequencies dependent variable per policy

5.2 Empirical results

Before testing the hypotheses, I will discuss the general results of the different models. A logistic regression analysis has been performed in order to test whether political knowledge, following the news, EU attachment and dual identities influence ambivalence. This model is statistically significant, X2(4) = 1.657, p < ,001. Moreover, if I add the control variables (age and gender) the model stays statistically significant, X2(6) = 1.713, p < ,001. Furthermore, model 3, in which the countries are added as a control variable, is also statistically significant, X2_{(7) = 1.733, p < ,001.} Beside this, the logistic regression results show that between approximately 7% (Cox & Snell R2_{) and 9% (Nagelkerke R}2_{) of the variance in consistent vs. ambivalent} attitudes is explained by the independent variables. The percentage of accuracy in classification is 62%, meaning that the model was able to predict in 62% of the cases whether a citizen has an ambivalent or consistent attitude. The reason that the model only predicts such a small variance is probably due to the size of the model. Whether a citizen is consistent or ambivalent can be influenced by many more factors than those that are included in this research.

Table 4 shows the coefficients, significance and standard errors of the variables. Moreover, it shows the difference in effect of the variables on the dependent variable when changing the threshold on the policy-index. Several conclusions have to be made regarding the effect of the model on ambivalence when changing the threshold on the policy-index. First of all, the lower the threshold on the policy-index, the less the model is able to predict ambivalence. For example, a threshold of 6 or higher on the policy-index only explains 0,1% (Nagelkerke R2) of the variance in the dependent variable. However, the threshold on the policy-index does not affect the significance of the models. It does not matter whether the threshold on the policy-index is higher or equal to 6, 7, 8 or 9. Another noticeable finding is that the significance of the variables, e.g. following news, age and gender, decreases when the threshold on the policy-index decreases. Also, the effect of the variables decreases when the threshold on the policy-index decreases. Thus, the effect of the independent variables is the strongest when maintaining a strict view on ambivalence.

Table 4: Summary results logistic regression on ambivalence ***p <,001, **p <,01, *p <,05 Model 1 Model 2 Model 3

Variables B SE B SE B SE

Dependent variable with policy index = 9 Political knowledge ,041 ,028 ,061* ,028 ,051 ,028 Following news ,249*** ,050 ,281*** ,051 ,292*** ,051 Attachment EU -,530*** ,018 -,531*** ,018 -,529*** ,018 Dual identities ,320*** ,031 ,297*** ,031 ,308*** ,031 Age -,004*** ,001 -,004*** ,001 Gender ,129*** ,028 ,123*** ,028 Country ,007*** ,001 Constant ,392 ,073 ,395 ,094 ,269 ,099

Dependent variable with policy index ≥ 8

Political knowledge -,022 ,028 -,003 ,028 -,004 ,028 Following news ,175*** ,049 ,205*** ,050 ,206*** ,050 Attachment EU -,353*** ,017 -,354*** ,017 -,353*** ,017 Dual identities ,207*** ,031 ,183*** ,031 ,184*** ,031 Age -,004*** ,001 -,004*** ,001 Gender ,109*** ,027 ,108*** ,027 Country ,000 ,001 Constant ,014 ,072 ,053 ,094 ,045 ,098

Dependent variable with policy index ≥ 7

Political knowledge -,040 ,028 -,032 ,028 -,020 ,028 Following news ,155** ,048 ,166** ,048 ,155** ,049 Attachment EU -,138*** ,017 -,138*** ,017 -,141*** ,017 Dual identities ,160*** ,030 ,155*** ,031 ,144*** ,031 Age -,001 ,001 -,001 ,001 Gender ,066* ,027 ,072** ,027 Country -,007*** ,001 Constant -,473 ,071 -,525 ,093 -,387 ,097

Dependent variable with policy index ≥ 6

Political knowledge -,034 ,027 -,036 ,027 -,017 ,028 Following news ,162** ,047 ,161** ,047 ,143** ,047 Attachment EU ,048** ,017 ,048** ,017 ,043* ,017 Dual identities ,083** ,030 ,084** ,030 ,067* ,030 Age ,000 ,001 ,000 ,001 Gender -,006 ,027 ,003 ,027 Country -,012*** ,001 Constant -,831 ,071 -,834 ,092 -,614 ,096

5.4 Effect cognitive factors

Cognitive factors have proven to be influential in the formation of attitudes. However, there are conflicting explanations on the role of cognitive factors (Alvarez & Brehm, 2002; De Vries, 2013, p. 452, Stöckel, 2013, p. 36). What is the effect of cognitive factors regarding ambivalence between regime and policy support? The results in table 4 show that political knowledge has a small positive effect on the ambivalence between regime and policy support. However, the statistical analysis does not provide evidence that the effect of political knowledge is different from zero. Moreover, it is interesting that it does become statistically significant when adding the control variables age and gender and it loses is significance when adding the countries to the model.

The second hypothesis focuses on the effect of following the news media regarding the EU on ambivalent attitudes. I expected that following the news would increase ambivalence. The evidence from the statistical models is consistent with the expectation that following the news on EU will have a positive effect on ambivalence between regime and policy support.

5.5 Effect affective factors

The other independent variables are considered in the literature as affective factors. One of them is the level of attachment of a citizen towards the European Union. The expectation was that a stronger attachment to the EU would decrease ambivalence. Attachment is connected to the feelings of citizens towards an object and has proven to decrease ambivalence (Stöckel, 2012, p. 39). The results from table 4 show that the model is compatible with the expectation that the level of attachment decreases ambivalence.

The second affective factor is whether citizens have dual identities or not. Based on the literature, I expected that having dual identities would decrease the ambivalence between regime and policy support. However, the statistical results are not consistent with this expectation. Assuming this model and this dataset, the results show rather that dual identities increase ambivalence between regime and policy support.

5.6 Cognitive factors vs. affective factors

The last hypothesis focused on the difference in cognitive and affective factors. I expected that affective factors are better able to explain ambivalence between regime and policy support than cognitive factors. The reason for this is that citizens with no knowledge at all form their opinions based on the feelings that they have towards an object. Table 5 shows the results of the logistic regression with the level of attachment as a binary variable. A comparison between the absolute coefficients of the cognitive factors, political knowledge (,052) and following the news (,286), on one hand and the affective factors, level of attachment (-,831) and dual identities (,356) on the other, shows that this expectation is compatible with hypothesis 5.

Table 5: Summary results logistic regression on ambivalence ***p <,001, **p <,01, *p <,05 Model 1 Model 2 Model 3 Variables B SE B SE B SE

Dependent variable with policy index = 9 and binary attachment EU

Political knowledge ,052 ,028 ,073** ,028 ,062* ,028 Following news ,286*** ,050 ,318*** ,050 ,330*** ,051 Attachment EU -,831*** ,029 -,835*** ,029 -,832*** ,029 Dual identities ,356*** ,031 ,331*** ,031 ,342*** ,031 Age -,005*** ,001 -,004*** ,001 Gender ,131*** ,028 ,125*** ,028 Country ,007*** ,001 Constant -,584 ,055 -,578 ,081 -,707 ,086

6. Conclusion

In sum, I researched what the effects of cognitive and affective factors are on ambivalence between regime and policy support. In order to answer this question, I performed a logistic regression by using the Standard Eurobarometer 86.2 wave. The expectation was that the cognitive factors, political knowledge and following the news media regarding the EU, would increase ambivalent attitudes. Moreover, I expected that the affective factors, level of attachment towards the EU and dual identities, would decrease ambivalence between regime and policy support. Furthermore, the last expectation was that affective factors are better able to explain ambivalent attitudes compared to cognitive factors. An answer on the research question can be formulated now. Regarding the cognitive factors only following the news media indeed increases ambivalence. Moreover, both the affective factors influenced ambivalent attitudes. The higher the level of attachment the more ambivalent attitudes would occur. Further, contrary to the expectation, the results showed that dual identities moderately increased ambivalence.

It is interesting to link these results to current theories. One thing is certain: the state of the art literature does not agree on what the effect is of cognitive and affective factors on ambivalent attitudes. One could even say that there are many conflicting argumentations (except for the level of attachment variable). First of all, scholars do not agree on the role of political knowledge regarding ambivalent attitudes. There are scholars who argue that political knowledge decreases ambivalent attitudes (Zaller, 1992), while others argue that political knowledge is not relevant at all (De Vries & Steenbergen, 2013). The conclusion can be made that, regarding ambivalence between regime and policy support, political knowledge is not relevant at all. Moreover, scholars also did not agree on the effect of following the news media on ambivalent attitudes. Some researchers argued that following the political news decreases ambivalent support (De Vries, 2013, p. 452), while others argued the exact opposite (Stöckel, 2013, p. 36). My results have shown that following the news media decreases ambivalent attitudes between regime and policy support. It is difficult to say what the reason for these conflicting explanations is. However, I believe that this could be due to the different operationalizations and designs of research. Furthermore, the analysis has proven that dual identities increase ambivalence between regime and policy support. This probably has to do with the fact that dual identities provide

conflicting cues. Citizens want to preserve their national sovereignty, but they also identify with the EU. In other words, people see the EU as a threat to national autonomy and they simultaneously identify with the EU (Cinnirella, 1997). This study has enriched the current literature by proving that a large amount of the citizens is indeed ambivalent and that ambivalence also occurs regarding regime vs. policy support. Moreover, the influence of the cognitive and affective factors on ambivalent attitudes is tested in a different way compared to how it was researched in the existing studies.

There are a few clear limitations of this research. First of all, due to time limitations and data availability, only two cognitive and two affective factors have been included in the analysis. At this point I cannot argue with certainty that affective factors are better able to explain ambivalent attitudes compared to cognitive factors, due to the limited amount of factors. Thus, it would have been better for the validity of the research if more cognitive as affective factors were included in the analysis. Moreover, the operationalizations of political knowledge and following the news media were not ideal. A respondent was considered to have political knowledge if the respondent answered 3 knowledge questions correctly. This is a very limited way of testing someone’s EU knowledge, since the possibility exists that the respondent coincidently knew these answers. Further, regarding the following the EU news media operationalization, there was not a formulated answer option that referred to the fact that a respondent did not follow the news on EU matters. The dataset only included this as an answer if the respondent mentioned it spontaneously.

More research is necessary for a better understanding of ambivalent attitudes towards the EU. It would be interesting to further investigate what the consequences of ambivalence are. The current knowledge that we have regarding the consequences of ambivalence is based on US research. Our over-seas colleagues have argued that ambivalent citizens are more driven by salient considerations through, for example, party cues or the media (Lavine et al., 1998; Zaller, 1992). At the moment, we see more contestation regarding European integration during party competition. It could be expected that citizens are more sensitive to the information that is provided to them by the parties (De Vries, 2007). However, the first step would be to provide better data, since this is necessary to avoid the limitations experienced during this research.

Reference list

Achen, C.H. (1975). Mass Political Attitudes and the Survey Response. American

Political Science Review, 69(4): 1218–31.

Alvarez, R.M. and Brehm, J. (1995). American Ambivalence towards Abortion Policy: Development of a Heteroskedastic Probit Model of Competing Values.

American Journal of Political Science, 39(4): 1055–82.

Boomgaarden, H.G., Schuck, A.R.T., Elenbaas, M. and De Vreese, C.H. (2011). Mapping EU attitudes: conceptual and empirical dimensions of Euroscepticism and EU support. European Union Politics 12(2): 241–66.

Bølstad, J. (2015). Dynamics of European Integration: Public Opinion in the Core and Periphery. European Union Politics, 16(1): 23-44.

Brewer, M. B. (2001). The many faces of social identity: Implications for political psychology. Political Psychology, 22(1): 115–125.

Cacioppo, J.T., Gardner, W.L. and Berntson, G.G. (1997). Beyond Bipolar Conceptualizations and Measures: The Case of Attitudes and Evaluative Space.

Personality and Social Psychology Review, 1(1): 3–25.

Cinnirella, M. (1997). Towards a European identity? Interactions between the national and European social identities manifested by university students in Britain and Italy. British Journal of Social Psychology, 36(1): 19–31.

De Vries, C. (2013). Ambivalent Europeans? Public Support for European Integration in East and West. Government and Opposition, 48(3): 434-461.

De Vries, C., Steenbergen, M.R. (2013). Variable opinions: the predictability of support for unification in mass European publics. Journal of Political

Marketing, 12(1):121–41.

De Wilde, P., Trenz, H.J. (2012). Denouncing European integration: Euroscepticism as policy contestation. European Journal of Social Theory, 15(4): 537–554. Easton, D. (1975). A re-assessment of the concept of political support. British Journal

of Political Science, 5(4): 435–57.

Hakhverdian, A., Van Elsas, E., Van Der Brug, W., and Kuhn, T. (2013_. Euroscepticism and education: a longitudinal study of 12 EU member states.

Hobolt, S. and Brouard S. (2011). Contesting the European Union? Why the Dutch and the French Rejected the European Constitution. Political Research

Quarterly 64(2): 309-22.

Hobolt, S. and Wratil, C. (2015). Public Opinion and the Crisis: The Dynamics of Support for the Euro. Journal of European Public Policy, 22(2): 1-19.

Hobolt, S., and De Vries, C. (2016). Public Support for European Integration. Annual

Review of Political Science 19(1): 413-32.

Hooghe, L. and Marks, G. (2009). A postfunctional theory of European integration: from permissive consensus to constraining dissensus. British Journal of

Political Science, 39(1): 1–23.

Hooghe, L. (2007). What drives Euroskepticism? Party-public cueing, ideology and strategic opportunity. European Union Politics, 8(1): 5–12.

Gabel, M.J., Palmer, H.D. (1995). Understanding variation in public support for European integration. European Journal of Political Research, 27(1): 3–19. Inglehart, R. (1970). Cognitive mobilization and European identity. Comparative

Political Studies, 3(1): 45–70.

Kuhn, T., and Stöckel, F. (2014). When European Integration Becomes Costly: The Euro Crisis and Public Support for European Economic Governance. Journal of

European Public Policy, 21(4): 1-18.

Kritzinger, S. (2003). The Influence of the Nation-State on Individual Support for the European Union. European Union Politics, 4(2): 219-41.

Lavine, H., Huff, J.W., Wagner, S. and Sweeney, D. (1998). The Moderating Influence of Attitude Strength on the Susceptibility to Context Effects in Attitude Surveys. Journal of Personality and Social Psychology, 75(2): 359–73. Macarena A., Ceka, B., & Hanspeter, K. (2017). Diffuse support for the European

Union: spillover effects of the politicization of the European integration process at the domestic level. Journal of European Public Policy, 24(8): 1091-1115. Steenbergen, M. and Brewer, P.R. (2004). The not-so ambivalent public: Policy

attitudes in the political culture of ambivalence. In: Sniderman P and Saris WE (eds) Studies in Public Opinion: Attitudes, Nonattitudes, Measurement Error,

and Change. Princeton, NJ: Princeton University Press, 93–129.

Stöckel, F. 2013. Ambivalent or indifferent? Reconsidering the structure of EU public opinion. European Union Politics, 14(1):23–45.

Ringlerova, Z. (2015). Weathering the Crisis: Evidence of Diffuse Support for the EU from a Six-wave Dutch Panel. European Union Politics, 16(4): 558-76.

Rose, R., and Borz, G. (2016). Static and Dynamic Views of European Integration. JCMS: Journal of Common Market Studies, 54(2): 370-87.

Down, I, and Wilson, C. J. (2017). A Rising Generation of Europeans? Revisited.

European Journal of Political Research, 56(1): 199-214.

Zaller, J. (1992). The Nature and Origin of Mass Opinion. New York, US: Cambridge University Press.

Zaller, J. and Feldman, S. (1992). A Simple Theory of the Survey Response: Answering Questions versus Revealing Preferences. American Journal of