Illiberal democratic attitudes and support for the EU

Are those who support the core values of liberal democracy also more likely to support the European Union? In this article, we study the relationship between EU support and support for the principles of liberal democracy among citizens in the 28 EU member states, using data from the European Election Studies 2019. Our findings demonstrate that supporters of liberal principles of democracy tend to be more supportive of the EU, while supporters of more direct forms of citizen influence are more Eurosceptic. We argue that this may be in part due to the design of the EU with strong institutional checks-and-balances, but a weak link to citizens. Attitudes towards liberal democracy are less structured than previous research suggests. Yet, the structuration of attitudes towards liberal democracy and the association between these attitudes and EU support is stronger in contexts where the role of the institutions of liberal democracy is more contested. This reconfirms that elite cues are essential for the formation of structured mass attitudes.


Introduction
Do Europeans support liberal democratic institutions and norms, and if so, is this support more widespread among those who support the European Union (EU) than among Eurosceptics? Liberal democracy is founded on two pillars: the electoral pillar based on citizen representation and the principle of majority rule, and the constitutional pillar that consists of institutional checks and balances to limit executive power and protect minorities (Mair, 2002;Mény and Surel, 2002;Sartori, 1995). Ideally, there is a balance between the two pillars, but there is also an inherent tension between the principles of majority rule on one hand, and the principle of checks and balances and protection of minority rights on the other hand.
While the EU has been criticized for its poorly developed electoral institutions (e.g. Andersen and Burns, 1996;Follesdal and Hix, 2006;Raunio, 1999), others have argued that its institutions perform well when they are evaluated against the principles of liberal democracy (e.g. Moravcsik, 2002;Zweifel, 2002). These principles include institutional checks and balances on executive power, respect for the rule of law, human rights, and civil liberties. Since the EU is seen to perform well in terms of the constitutional pillar of democracy, but weakly with regard to electoral representation of citizens, support for European integration can be expected to be associated with support for specific models of democracy. We would thus expect that individuals who value the idea that public policies should always reflect the preferences of a majority of the citizens are expected to be very critical of the EU. In contrast, people who value compromise, consensus building, protection of minorities, and constraints on executive power are likely to be more in favour of the EU.
While much research exists on the determinants of support for European unification, we know less about the support for values of liberal democracy and how these are related to support for European integration. This is the focus of our study, which addresses the following questions: To what extent are citizens in European democracies supportive of liberal democracy? Are such illiberal attitudes related to Euroscepticism? And are these attitudes more strongly related in contexts where the role of the institutions of liberal democracy is more contested?
Research on public support for specific liberal democratic values is scarce and existing survey research is based on different types of conceptualizations of liberal democratic attitudes. Schedler and Sarsfield (2007) define liberal democratic attitudes by the combination of support for democracy and various liberal values, such as tolerance towards homosexuals. We conceptualize liberal democratic values as the attitude that executive power needs to be constrained; that the power of the executive needs to be limited, even if the executive represents the majority of the citizens. This conceptualization is very similar to that of Ferrin and Kriesi (2016), whose operationalization relies on survey questions that ask respondents how important certain democratic principles are. Since overwhelming majorities of citizens consider 'equality before the law' and 'freedom of the press' highly important for a democracy, their study concludes that support for the basic principles of liberal democracy is widespread. Such conclusions may, however, be too optimistic, because these items do not tap into the core of liberal democracy, which is putting institutional constraints on executive power.
When asking questions on the importance of various aspects of democracy, Ferrin and Kriesi (2016) included items referring to the popular and the constitutional pillars of democracy. They conclude that these items measure one latent attitude dimension. So, people who consider elections important are also the ones who say that they find equality before the law important. In our study, we propose a different operationalization of liberal democratic attitudes and we do not expect to find support for the one pillar to form a scale with support for the other pillar. There are two reasons for this.
First, there is an inherent tension between the popular and the constitutional pillars of liberal democracy. The popular pillar ensures that governmental power rests upon an electoral mandate, while the constitutional pillar introduces checks and balances, which limit the power of elected politicians and thereby safeguard citizens from an abuse of power and from 'majority tyranny' (e.g. Dahl, 1956). The constitutional pillar consists of a set of institutions and regulations that are able to overrule majority decisions, if these decisions violate rights of individuals. So, support for the principles of liberal democracy implies that one accepts the fact that rights of minorities or individuals can sometimes prevail over majority opinions. The inherent tension between the two pillars makes it unlikely that support for one of the pillars would be strongly related to support for the other pillar. Second, liberal democracy is a highly abstract concept. This makes it implausible that attitudes are so neatly structured. We will return to this below.
We argue that any valid operationalization of support for liberal democratic principles should seek to capture the different dimensions of democracy with majority rule on one hand and the constitutional protection of basic rights and the limitation of executive power on the other hand. For this purpose, we have developed a short battery of survey items that explicitly tap into this tension between majority rule and protection of individual rights. These questions were included in the most recent European Election Studies (EES) (2019; Schmitt et al., 2019). We show that, when it comes to constraining the power of the executive, support for liberal democratic values is not as widespread as some of the more optimistic studies concluded. As expected, support for liberal democratic values is positively correlated to support for European unification. However, support for principles of liberal democracy does not exist in a vacuum. How citizens respond to these items is context-dependent. In countries that score relatively low on the Varieties of Democracy Project (V-Dem) indicators of liberal democracy, we tend to find most structuration in attitudes towards liberal democracy -that is, citizens organize their views on different aspects of liberal democracy in a relatively more consistent and structured manner -and we find that these attitudes are most strongly related to EU support. Our interpretation is that most citizens do not form structured attitudes towards complex concepts such as liberal democracy, unless the institutions of liberal democracy have become the object of political contestation.
Our findings contribute to three strands of literature. First, we contribute to the literature that investigates support for liberal democracy. We show that there is less support for the principles of liberal democracy than the recent study by Ferrin and Kriesi (2016) suggested. Second, we contribute to research on support for European integration. Most research focuses on the question whether that support is mainly driven by utilitarian considerations or by cultural attitudes and value orientations (e.g. Gabel, 1998;Inglehart, 2003). We demonstrate that in addition, support for the EU is related to preferences for a specific model of democracy. Finally, our article speaks to the literature on public opinion formation, by demonstrating that structured attitudes are most likely to develop around issues that are politicized. This finding speaks to a related classic argument of V.O. Key (1966), that the behaviour of voters is but an 'echo chamber' that reflects (clarity of) the alternatives presented to them. While V.O. Key referred to party/candidate choice in elections, the same can be said about public opinion more generally.
Our article is structured as follows. We start with a more detailed outline of liberal democratic values and then discuss how this might be related to support for European unification. Also, we will argue why we expect these relationships to become stronger in the face of politicization. After that, we will discuss our data and operationalization, followed by a 'Results' section. In the final section, we will reflect upon our findings.

Conceptualizing liberal democracy
Liberal democracy hinges on widespread public support for the rules and values guiding political processes and institutions. If citizens value democratic procedures and institutions, these become legitimate and self-perpetuating; if they reject them, democracy becomes unstable (Weber, 1946). Contemporary liberal democracy is upheld by two pillars: the 'popular pillar', based on the principle of majority rule, and the 'constitutional pillar', based on the principle of individual and minority rights and the rule of law (Mair, 2002;Mény, 2003;Rawls, 1971;Sartori, 1995). The more 'minimal' definitions of democracy emphasize just the role of elections. A well-known example comes from Schumpeter (1949Schumpeter ( [1976: 269), who defined democracy as 'an institutional arrangement for arriving at political decisions in which individuals acquire the power to decide by means of a competitive struggle for the people's vote'. In this way of thinking, free and fair elections are not just at the heart of democracy, they are its sole defining characteristics. However, it is important to emphasize that even minimal definitions of democracy imply several other criteria that must be realized, such as freedom of speech, freedom of the press, and freedom of association, including organized political opposition and protest. Moreover, these definitions demand that elections are competitive, so that voters have an actual choice (APSA, 1950;Mair, 2008;Powell, 2019). In addition, Schumpeter emphasized the importance of rotation of office. Elections are a method which guarantees that there is at least the option of a peaceful transition of power at regular intervals. So, free, fair and competitive elections are a central component of (representative) democracy.
However, one of the main problems that democratic theorists have been struggling with, ever since Madison, is the possibility of a 'majority tyranny' (Dahl, 1956;Mill, 1859). An electoral majority could use its democratic power to challenge some 'fundamental rights' of minorities. Defining what those fundamental rights are, is of course by itself a political question that cannot be settled without dispute. If five friends go out for dinner and four of them decide by a majority vote that number five has to pay the bill, most of us would agree that this is not fair. At the same time, most citizens (including those who are better off) consider it fair that higher income groups pay more taxes than lower income groups. While it is thus not easy to define which types of decisions violate fundamental rights of some individuals, the risk of majority tyranny is inherent to majoritarian decision making.
An additional problem is that, even when there are free and fair elections, politicians may abuse their power. They may try to make it harder for opposition parties to organize and win the next election, they may be corrupt and accept bribes from big companies in exchange for contracts, or they may engage in activities of patronage or clientilism (i.e. reward their supporters with all kinds of goods or services or withhold these goods and services from opponents). It is a basic principle of liberal democracy that society should be protected from such abuse of power. Even when political leaders enjoy huge electoral support, their power should be constrained. The 'constitutional pillar' thus consists of institutional arrangements that limit the power of elected politicians and thereby safeguard citizens from an abuse of power and 'majority tyranny'. In the literature, this is generally referred to as 'checks and balances' (Dahl, 1956).
These checks and balances are not just essential to defend individuals and minority groups from majority tyranny. They are also necessary to protect the popular pillar of democracy. The proper functioning of elections as a mechanism for democratic representation and accountability is undermined if incumbents make it hard for the opposition to organize, if there are constraints upon the possibilities for interest organizations to emerge, or when there are restrictions on the media. Yet, particularly when it comes to the protection of citizens against majority tyranny, we must realize that there is always a tension between the two pillars of liberal democracy. The popular pillar is at the base of political representation and if that functions well, public policy would be in line with the will of the majority of the citizens. The constitutional pillar puts constraints on the exercise of power. These kinds of checks and balances that could overturn majoritarian decisions, distinguish liberal democracy from a more basic understanding of electoral democracy. The V-Dem conceptualizes liberal democracy as a democracy that has all features of an electoral democracy 'plus three additional components: the rule of law ensuring respect for civil liberties, judicial constraints on the executive branch, and legislative checks and oversight of the executive' (Mechkova et al., 2017). Countries that are electoral democracies, but that do not meet the three criteria of liberal democracies, are sometimes referred to as 'illiberal democracies'. This term describes countries that hold regular elections, but where there are only weak constraints on the powers of the executive, and where freedom of the press is not always respected.
The purpose of our study is not to classify countries as liberal or illiberal democracies, but to examine support for the principles of liberal democracy among the public, and the association between such attitudes and support for the EU. 'Illiberal democratic attitudes' manifest themselves by a rejection of the legitimacy of institutions, such as constitutional courts, that impose constraints on the power of the executive, and possibly limit the exercise of the 'will of the people' (Van Hauwaert and Van Kessel, 2018;Mudde and Kaltwasser, 2012;Müller, 2016). Such attitudes may conceivably be widespread also in countries that fulfil all the criteria of liberal democracies.
We are not the first to study liberal democratic attitudes. As already referred to, a recent volume edited by Ferrin and Kriesi (2016) studies cross-country differences in support for democracy, by means of a battery of questions about the relative importance of various democratic norms. Liberal democratic values are measured by items asking about the importance of 'equality before the law', 'checks of government by the courts', 'minority protection', 'freedom of the media' and 'freedom of expression'. These questions clearly ask about the fundamental political liberties and institutions that safeguard citizens from the abuse of executive power. However, as these items ask about the importance of each of these norms in turn, they do not consider the inherent tension between support for the popular pillar of democracy and support for the 'constitutional pillar' as it is identified in the literature. So, while we think this is an important contribution, we also think it is important to focus on the extent to which citizens support the principles and institutions of liberal democracy that have been designed to impose limitations on the power of the executive. Support for these principles implies that one should accept the fact that these institutions have a legal obligation to overrule government decisions, even when these reflect the opinions of a majority of the citizens, if need arises. So, the first two questions that we explore in our article are the following: RQ1. How much support is there for various components of liberal democracy?

RQ2. Is there a single underlying dimension of liberal democratic attitudes that is similar across different countries?
Liberal democratic attitudes and support for the EU Building on our measurement of liberal democratic attitudes, we examine the association between such attitudes and support for the EU. The values of liberal democracy are at the heart of the EU. Article 2 of the Treaty on the EU starts as follows: 'The Union is founded on the values of respect for human dignity, freedom, democracy, equality, the rule of law and respect for human rights, including the rights of persons belonging to minorities'. While the EU aspires to be one of the core defenders and promoters of democratic institutions in Europe and beyond, the EU itself is sometimes also said to suffer from a 'democratic deficit'. So, this does raise the question whether the EU is indeed committed to democracy and how we would expect liberal democratic attitudes to be linked to support for European unification.
Those who argue that there is a 'democratic deficit' in the EU, focus mainly on the popular (electoral) pillar. Critics point to the 'second-order' nature of EP elections, which lack a genuine European dimension that can link voter preferences to policy outcomes at the European level (Reif and Schmitt, 1980;Van Der Eijk and Franklin, 1996;Van Der Brug and Van Der Eijk, 2007;Hix and Marsh, 2007). Moreover, the link between European Parliament (EP) elections and the EU executive, the EU Commission, is still weak, despite the introduction of the Spitzenkandidaten process in recent EP elections (Hobolt, 2014). This means that unlike most national parliamentary elections, European elections do not produce a clear mandate for a European government and a set of public policies. Without a distinct European government, with a clear government-opposition dynamic, voters cannot use their vote in EP elections to hold their representatives accountable for policies at EP-elections (Hobolt and Tilley, 2014). Moreover, since the policies agreed upon are outcomes of long negotiations between national governments (represented in the Council), national parliaments have little room to influence those decisions. The legislative powers of the EP, while increasingly similar to those of the EU's other legislative body, the Council of the EU, do not extend critical areas of decision making -particularly in times of crises (like the Euro crisis or the current COVID-19 pandemic). So, effectively, policy making processes in the EU are quite insulated from popular influence and control (e.g. Follesdal and Hix, 2006;Mair and Thomassen, 2010;Raunio, 1999).
Scholars, and commentators, who defend the EU and challenge the notion of a democratic deficit do not speak about the popular pillar of democracy, but refer mostly to the second (the constitutional) pillar. Moravcsik (2002: 605) argues as follows: 'The classic justification for democracy is to check and channel the arbitrary and potentially corrupt power of the state'. Moravcsik argues that the EU is democratically legitimate when its performance is judged by the standards of liberal democracy. In his own words, 'The EU's ability to act, even in those areas where it enjoys clear competence, is constrained by institutional checks and balances, notably the separation of powers, a multi-level structure of decision-making and a plural executive' (Moravcsik, 2002: 609). Moreover, policy making in the EU is constrained by rules and regulations, which are safeguarded by the European Court, consisting of appointed judges (Grabbe and Lehne, 2016). So, while the popular representation of European citizens in EU decision-making has come under criticism, few will challenge the notion that the EU is built on the principles of liberal democracy.
By stressing the popular pillar of democracy, 'illiberal' democrats want the 'will of the people' to be reflected in public policy. In other words, decisions should reflect the will of a majority of the population. In contrast, by placing emphasis on the constitutional pillar, 'liberal democrats' argue that executive power needs to be constrained and that there are fundamental rights of minorities and individuals that need protection by institutional checks and balances. It follows, therefore, those individuals who support the constitutional pillar of liberal democracy are likely to be more enthusiastic about the EU than those who support the popular pillar of liberal democracy. After all, the EU's values are ones at the heart of the constitutional pillar of liberal democracy: checks-andbalances, equality, rule of law, respect for human rights. Illiberal democrats are critical of the democratic legitimacy of institutions, such as the European Court of Justice, that limit the 'popular will of sovereign nations'. It is inevitable that rulings of European courts and decisions of a transnational union like the EU will sometimes overrule national laws and practices, even if these are supported by a majority of the population. And illiberal democrats may find it difficult to accept such outcomes. We expect therefore that H1. Individuals who are more supportive of liberal democratic institutions and norms are more supportive of the EU, while individuals who are less supportive of liberal democratic institutions are less supportive of the EU.
We cannot make any claims about the causal direction of this relationship, that is, whether liberal democratic attitudes influence EU attitude, or whether EU attitudes drive support for liberal democracy, or both. 1 It is reasonable, however, to expect that the strength of the relationship between these two sets of attitudes is conditioned by the politicization of liberal democratic institutions in the national context, as we discuss in the next section.

The role of domestic politicization
In public opinion surveys, an overwhelming majority of the citizens in all European countries express diffuse support for democracy. People consider free and fair elections important, as well as equality before the law and independent judges. However, liberal democratic values are more abstract and complex than that. As we have seen in the previous sections, liberal democracy involves a delicate balance between majoritarian principles, constraints of executive power, and minority rights. It seems very unrealistic to expect that the average citizen would have clearly elaborated opinions on these matters, and that these opinions would be coherently structured by underlying attitudes (Converse, 1970;Zaller, 1992). After all, it seems highly plausible that most citizens do not think much about the constitutional design of democratic institutions. However, when these institutions are being criticized by politicians, and the institutional design of the country becomes object of political contestation, it becomes more likely that people form more structured attitudes. There are two reasons why we would expect this. First, because the issue becomes more salient, so that people pay more attention and think more about these matters. Second, because political elites will provide more cues to citizens as part of the increased politicization of the design of the political system. When citizens organize their views on different aspects of liberal democracy in a shared and systematic way, we refer to this as structuration of these attitudes (see also Hernández, 2019).
The expectation that we find more structuration in public opinion when issues are politicized is supported by literature in the field of issue voting as well as public opinion formation. In his seminal work, Converse (1964) had argued that most American citizens lacked the knowledge and interest to form structured attitudes on political issues and to link those attitudes to positions of political candidates. He later referred to this as 'nonattitudes' (Converse, 1970). Partially in response to Converse, several scholars argued that the extent of structuration in public opinion as well as the extent of issue voting depends upon the supply side as well (e.g. Carmines and Stimson, 1986;De Vries et al., 2011;Granberg and Holmberg, 1986;Key, 1966;Kinder, 2003). The more issues are politicized in the public debate, the more citizens feel that the topic matters and the more elite cues they receive. This increases the likelihood that people form attitudes that are clearly structured.
If we translate these insights to the relationship between liberal democratic attitudes and support for European integration, we should thus think of the relevant cues. Recent examples of democratic backsliding in Europe are especially the cases of Hungary and Poland, where radical right-wing politicians have questioned the authority of courts and where they have tried to lay off some judges who were deemed too critical of the government. In these circumstances, we expect people to form more structured opinions on these issues. Moreover, the governments of these countries were fiercely criticized, not only by the liberal opposition in their country, but also by prominent members of the EP and the European Commission. So, in these contexts, we would expect attitudes on liberal democracy to become more strongly linked to attitudes towards European Unification. Our second set of hypotheses is, therefore, the following: H2. Attitudes towards liberal democracy become more structured when liberal democratic institutions are more contested.
H3. The positive relationship between support for liberal democracy and support for the EU (H1) is stronger when liberal democratic institutions are more contested.

Data, measurement, and methodology
To examine our research questions and test the hypotheses, we analyse data from the EES 2019 (see Schmitt et al., 2019). The EES 2019 is a cross-national post-election survey conducted in all 28 member states with over 26,500 respondents. It was designed to examine the determinants of electoral behaviour in the 2019 EP elections and importantly it also contained a battery of questions on support for liberal democratic values. The survey was conducted by Gallup International, mostly online. Respondents were selected randomly from access panel databases using stratification variables, with the exception of Malta and Cyprus where a multi-stage Random Digit Dialling approach was used. In all countries, the samples were stratified by gender, age, region, and type of locality. The sample size is roughly 1000 interviews in each EU member state (except Cyprus, Luxembourg, and Malta where the sample size is 500).
Unsurprisingly, the samples largely reflect the population in terms of the variables on which the respondents were selected. Yet, in terms of party choice, the samples are also highly representative of the population, and not further off from the population than previous rounds of the EES, which were based on random probability samples. Still, one might expect specific groups of voters (e.g. the politically interested) to be overrepresented in this sample. Since these people are less likely to hold illiberal democratic attitudes, we may overestimate the support for liberal democracy to a certain degree. This, however, does not constitute a real handicap as relationships observed in multivariate models are less likely to be affected by sampling bias than in more descriptive statistics.

Individual level variables
We measure support for liberal democracy (i.e. support for the 'constitutional' pillar) by means of four statements that focus on constraining the power of democratically elected governments. Respondents were asked to indicate their agreement with these statements by means of 5-point Likert-type scales. Support for an independent judiciary is measured by the item 'Independent judges should be able to overrule the decisions of democratically elected governments if these decisions violate the rights of individuals or groups in society'. Support for freedom of speech is measured by the statement 'Governments should be allowed to prohibit a peaceful protest if many citizens will be offended by its slogans'. One issue with this item could be that people think about the kind of demonstrations that are sometimes prohibited, so that their responses to this statement may be affected by their sympathy for those demonstrations. In Western Europe, it is quite common for demonstrations of the far right being prohibited. Yet, this is not because the slogans are offensive, but because of threats to public order. In contrast, it is a common practice in illiberal 'democracies' like Turkey and Russia to prohibit gay right parades, because they are perceived as 'offensive'. Our survey item is framed in a way that seeks to tap into a fundamental principle of liberal democracy, namely to allow peaceful protest. Support for freedom of the press is measured by the proposition 'In order to secure the independence of the public media, elected politicians should have no influence on the content of their broadcasting'. Finally, the principle that the power of the executive needs to be constrained by checks and balances is measured in a quite direct way with the item 'Having a strong leader in government is good for [COUNTRY] even if the leader bends the rules to get things done'. In the 'Results' section, we present Mokken-scale analyses to assess whether these items form a single scale.
The EES 2019 only contains one item that taps into the popular pillar of democracy. The item wording is the following: 'The people should decide, not politicians'. This reflects an emphasis on 'the will of the people' as paramount, rather than politicians and even representative democracy. It, thus, also captures a populist idea of democracy to some degree.
We measure support for the EU by means of an item that asks whether EU membership is 'a good thing' (coded 1), 'neither good nor bad', or 'a bad thing' (both coded 0). Our models also control for satisfaction with democracy at the national level, satisfaction with the current state of the economy, support for privacy rights, support for same-sex marriage, support for immigration and political interest. In terms of socio-demographic variables, the models include subjective social class, employment status (unemployed vs other), religiosity, gender, age, level of education, location (urban vs rural), and subjective wealth. All individual level predictors are rescaled to take values between 0 and 1 (see Supplemental Appendix A for a full description of the control variables).

Country level variables
To test H3, we need a measure of the extent to which liberal democratic institutions are contested in the member states of the EU. As no direct measures of such contestation exist, we have to rely on proxy measures. In this article, we present the results of two of these proxies, which are most closely related to our theoretical concept. The first proxy measure is whether there are populist parties in office, as these parties are the main actors questioning the legitimacy of these institutions. Of course, they may criticize these institutions when they are in the opposition, in which case their criticism tends to focus on 'the elite', which includes mainstream parties. Yet, when in government they take concrete steps to attack the main institutions of liberal democracy, by trying to limit the independent role of courts, by attacking journalists or by trying to limit the right to protest and/or to organize oppositional activities (e.g. Mudde and Kaltwasser, 2012). So, this leads to visible contestation on the institutions of liberal democracy. 2 Populist parties are identified using of the classification provided by the PopuList (Rooduijn et al., 2019), 3 while government parties were identified using the data provided by the Parlgov (Döring and Manow, 2019).
The second set of indictors is different V-Dem measures of liberal democracy (Coppedge et al., 2020). V-Dem is an ambitious international effort to produce comprehensive and detailed indicators of democracy for all countries. It is based on a survey among country experts (mainly political scientists) who assess the extent to which countries fulfil certain criteria. If a country scores low on these indicators, democratic institutions are weaker. It is not a direct measure of contestation, but in countries with a relatively low score, people have less reason to take democracy for granted and we assume that there is more concern about these issues, thus leading to more structured opinions. Rather than using aggregate indicators of democracy, we identify the sub-facets that are the closest to our individual level measures of liberal democratic attitudes. 4 We have also explored two alternative measures, based on anti-elite rhetoric of parties (based on the Chapel Hill Expert Survey and the Global Party Survey). While these variables are correlated with the other two indicators that we use, 5 we think these variables are more distant proxy measures of political contestation over the institutions of liberal democracy. 6 We present the results based on these measures in Supplemental Appendix F.

Results
We start our analyses by presenting descriptive statistics of the overall level of support for democracy and for liberal democratic values across the 28 EU countries. The majority of respondents support an independent judiciary and a public media system free from the influence of politicians. Support for an independent media is well above 80% in most Central and Eastern European (CEE) countries and reaches the highest level in Hungary; we find high levels also in some other countries like Austria and Cyprus. The lowest levels are found in the United Kingdom and Finland where support for independent media is around 60%. Support for an independent judiciary reaches an average of around 60% across the EU; it is again generally higher in CEE countries, while it drops below 50% in the three Nordic countries and Italy.
The picture looks different when investigating the other two aspects of liberal democracy covered in our study. In a substantial number of EU member countries only 35% of our respondents reject the idea that 'it would be good for the country to have a strong leader in office who would sometimes bend the rules to get things done'. This proportion is even below 30% in well-established democracies such as the Netherlands and the United Kingdom. Finally, only around 40% of the respondents disagree with the idea that governments should be allowed to ban a peaceful demonstration if the slogans would be offensive for many citizens. The right to organize a peaceful demonstration is a fundamental part of liberal democracy. In addition, while support for that right is above 50% in a number of countries, it is obviously not a right that is particular widely supported. This percentage drops at around 31% in Italy and Lithuania, and barely reaches 35% in the Netherlands and in Poland.

Is there an underlying attitude?
Given the fact that the items are distributed differently (see Figure 1), measurement models based on Item Response Theory (IRT) are the only appropriate measurement models for testing the construct validity of the scale (e.g. Van Der Eijk and Rose, 2015;Van Schuur, 2003). We applied Mokken Scaling, based on IRT, to assess whether the items form a coherent scale.
The most important diagnostic to assess whether a set of items forms a cumulative scale is the H-value. According to Mokken (1971), H-values below 0.3 indicate that the items do not form a scale. Based on this criterion, the results presented in Table 1 show clearly that our four items do not form a scale in any of the 28 EU countries. In a pooled analysis across all 28 countries, only two items (the independent media and independent judiciary items) form a unitary scale. However, even this pattern is not consistently identified at the country level as the two items only form a scale in half of the EU member countries. This seems to be the case especially in CEE countries where issues related to freedom of the press and judicial freedom are often politicized due to governments that violate these principles. This suggests that the formation of a coherent attitude towards the principles of liberal democracy is most likely to develop when these principles are jeopardized, as predicted by H2. Figure 2 demonstrates that this is indeed the case. At the country level, we see a strong negative correlation between the average of press and judicial freedom as measured by V-Dem (Coppedge et al., 2020) and the scalability of the two items on the other. 7 The items are most likely to form a scale in countries where violation of one or both these principles of liberal democracy were recorded. Furthermore, we also find a strong relationship between these H-values and the presence of populist parties in a government. To be specific, the strength of the correlation is 0.47 in the case of the 'Judiciary-Media' scale and respectively 0.45 for the 'Protest-Leader' scale, both are statistically significant at p < 0.05. This suggests that, as expected by H2, the politicization of these issues acts like a catalyst and increases the formation of coherent attitudes when it comes to media freedom and an independent judiciary. We also checked whether the two other items form a Mokken scale (see Table 1). While the freedom to protest    (Mokken, 1971). and opposition to a strong leader do not form a scale when we analyse the pooled data, there are some countries where the two items form a coherent scale. Nevertheless, in this case, we did not detect a relationship between scalability of these items and the V-Dem scores associated to these items. Given that there is no coherent structure in our four items capturing liberal democratic values, we continue to analyse our respondents' attitudes on these items separately.

Are liberal democratic values related to EU support?
As our data have a multilevel structure with individuals nested with countries, we estimated multilevel logistic regressions with random intercepts at the country level using the lme4 package in R. In our model specifications, the country level predictors were centred around their grand mean, while individual level predictors are all rescaled on a 0-1 scale. As a result, the regression coefficients are roughly comparable (see Table 2). Table 2 offers overall support for our hypothesis that liberal attitudes and EU support are correlated (H1). The results show positive and statically significant relationships between the four liberal democracy items and support for the EU. While the magnitude of the relationships between these items and EU support is not particularly large, three of the effects are comparable in size to those of other attitudes associated with EU support (e.g. economic satisfaction and support for immigration) and also with some of the well-established socio-demographic predictors of EU support (e.g. education, wealth, social class). The only exception is the item on the right to protest, where the relationship is weaker and also is no longer statistically significant when we simultaneously include all items (see Model 6, Table 2). 8 In the case of three out of the four variables, those who hold liberal democratic attitudes are two times more likely to support the EU in comparison to those who hold illiberal values. This translates to roughly a 16 percentage point difference in supporting the EU between the two groups (see Supplemental Appendix D). As expected, support for the popular pillar of democracy is negatively related to support for the EU (see       Model 5), and this relationship is of similar magnitude to the ones discussed earlier (i.e. support for the popular pillar is associated with a 17 percentage points decrease in support for the EU). The results presented in Table 3 provide ample support for H3. 9 In particular, when populist parties are in government, the relationship between Euroscepticism and liberal democratic attitudes tends to be stronger. This is the case for three out of the four liberal democracy items (the exception is support for an independent judiciary). When populists are in power, people who want a strong leader in power, those who support the idea of the politician controlling media content, and those who think governments can ban the right to protest are particularly likely to oppose the EU, while people who object to authoritarian leaders, support and independent media and support the right to protest are more likely to support the EU.
In the case of the V-Dem scales, we also find that all interactions are in the theoretically expected directions. The interactions are negative, meaning that the relationships are strongest when liberal democracy is weakest. In this case also, three of the four interactions turn out to be statistically significant.
To get a better grasp of these effects, we plot the marginal effects and predicted values (Figure 3(a)) and the difference in the probability to support the EU between those who fully support liberal democratic views and those who fully disagree with them ( Figure  3(b)). We can note that, in both, the case of opposition to a strong leader and support for the right to protest, those who hold liberal democratic attitudes are approximately two times likely to support the EU when populists are in government, and that this relationship only reaches statistical significance when populists are in power (see Figure 3). In both cases, we can further observe that the difference in the probability to support the EU between the two groups is about 15 percentage points higher when there is a populist party in government compared to countries where there are no populist parties in government. The presence of populist parties also moderates the relationship between support for the EU and support for independent media, but in this case, the moderating role of populist in power is weaker (see Figure 3(a) and (b)). Surprisingly, however, the interaction with the independence of courts is not statistically significant, even though the topic was politicized by populists in office.
When we use the V-DEM indicators to test H3, the picture is similar. The important difference is that the interaction with the independent judiciary is now statically significant, while the interaction with support for the right to protest is not statistically significant.
In all three cases, as expected, there is a strong positive relationship between liberal democratic attitudes and support for the EU in countries where the role of liberal democratic institutions is weakened, while in countries where we do not note substantial infringement on liberal democratic institutions this relationship is weaker. In fact, the relationship fails to reach statistical significance for the countries where liberal democratic institutions function at their best (see Figures 4 and 5).

Conclusion
In this article, we have examined support for liberal democracy in Europe and the relationship between EU support and support for the principles of liberal democracy among citizens in the 28 EU member states. The essence of liberal democracy is that the power of the executive needs to be constrained by checks and balances, even if the executive has a clear electoral mandate. The EU is often criticized for its democratic deficit, by which these critics mean that the electoral pillar is poorly developed. EU policies are weakly linked to citizens' preferences and elections offer no possibilities to hold anyone accountable for these policies (e.g. Andersen and Burns, 1996;Follesdal and Hix, 2006;Raunio, 1999). Those who defend the EU argue that the EU performs well when evaluated by the standards of liberal democracy, particularly its built-in checks and balances on executive power (e.g. Moravcsik, 2002). To be sure, the EU has been criticized for its inability to sanction cases of democratic backsliding in Poland and Hungary (e.g. Kelemen, 2020). Yet, both the European Commission and the EP attempted to initiate sanctions against Poland and Hungary for not respecting these liberal democratic institutions. These initiatives were highly visible.
Across the board, we found that those who support the liberal democratic model tend to be more favourable towards the EU, while Eurosceptics are more in favour of popular democracy. This relationship is particularly strong when populist parties are in office and when liberal democratic institutions are weaker. We think this is because populists tend to challenge the institutions of liberal democracy, while they tend to support more direct forms of democracy. This politicizes the liberal democratic institutions, and at the same time, it creates a clash with European institutions.
As far as we know, we are the first to measure support for liberal democracy by means of a battery of survey questions that focus explicitly on constraining the power of the executive, thus tapping into the tension between liberal democracy and majoritarian principles. Using this approach, we uncover lower levels of support for liberal democracy than some previous work. For example, the ESS (round 6) indicated that more than 90% of European citizens say that 'equality before the law' is important (see, for example, Ferring and Kriesi). However, in our study, we find that less than 40% of the European citizens disagree with the statement that it would be good to have 'a strong leader in office, even if s/he bends the rules to get things done'. While the liberal attitudes are more widely supported than the illiberal ones, support is much lower than previous research would suggest. Another important finding is that liberal democratic attitudes are much less structured than the items that measure 'importance of values'.
To some extent, cross-country differences in the formation of attitudes can be explained by whether the institutions of liberal democracy are being jeopardized. In countries that score lower on the V-Dem indicators of freedom of the press and freedom of the judiciary, the items show more structuration. It thus seems that in other countries, people just take things more for granted. Even if people express opinions on these items (opinions that turn out to be related in a meaningful way to other attitudes), citizens combine their opinions on liberal democracy in a more idiosyncratic way.
These results are in line with much existing research showing that consistent attitudes are most likely to form in contexts in which elites provide cues to citizens in the context of polarized debates (Carmines and Stimson, 1986;Granberg and Holmberg, 1986;Key, 1960;Kinder, 2003). From a normative point of view, some scholars and pundits will probably be worried about the consequences of the contestation of institutions that are designed to constrain the power of the executive. Yet, a more positive side effect might be that citizens do form more consistent attitudes on these matters. Given the fact that there is widespread support for EU membership also in countries where the institutions of liberal democracy are being challenged, the EU should feel confident to defend those institutions. At the very least, this is what the citizens in these countries expect.
Finally, from the perspective of liberal democracy, should we conclude that the glass is half empty or half full? On the more positive side, the proportion of people supporting liberal democracy tends to be greater than the proportion opposing it. Yet, there is cause for concern. There is clearly less support for the principles of liberal democracy than some previous research suggests (Ferrin and Kriesi, 2016). And these attitudes are not clearly structured, so that even if many people support one of the principles of liberal democracy, the same people may oppose one of the other principles. Worryingly, this may suggest that support for liberal democratic institutions and norms may be highly malleable and context-dependent.

Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article. They made use of the European Elections Studies 2019, which was funded by the Volkswagen Foundation.

Supplementary Information
Additional supplementary information may be found with the online version of this article.