Liberal, Republican, Conservative and Social-Democratic Mindsets? Exploring the existence of Citizenship Regimes in Civic Attitudes

This paper explores whether civic attitudes cluster in ways that correspond to distinct citizenship regimes. Drawing on political philosophy and citizenship literature, it identifies a liberal, a republican, a conservative, a social-democratic and a post-communist regime. These regimes are said to prevail in particular European regions and to show a certain level of stability. Using European Values Study / World Values Survey data, the paper explores whether socio-political attitudes ‘fit’ the theoretical regimes in terms of substance, country membership and durability and whether distinct European citizenship regimes can also be identified at the global level. It finds fairly strong evidence for a social-democratic, a liberal and a post-communist regime of civic attitudes at the European level, but could not find much support for a specific republican or conservative regime. The regimes identified at the European level disappear at the global level. At that level, a group of western countries appears, which distinguishes itself from other countries by showing relatively high scores on a range of citizenship indicators. Thus, it depends on the vantage point whether qualitatively different clusters of civic attitudes emerge that correspond to distinct European citizenship regimes.


Introduction
The scholarly literature on political philosophy and citizenship has identified different regimes defining state-citizen relationships. These regimes have not only been proposed as normative ones, i.e. as regimes that prescribe certain roles, behaviours and mentalities for citizens and specify rules for access to the community of citizens (e.g. Etzioni 1993; Honohan, 2002;Castles & Miller, 2003;Walzer, 2012;Fawcett, 2014), but also as empirical phenomena that prevail in particular countries and regions. Thus, regimes often mentioned include a liberal, a republican, a communitarian/conservative and a social-democratic one, which are presented as relatively enduring moral and legal frameworks exerting a disproportionate influence on policy in the English-speaking countries, France and Southern Europe, the German-speaking countries, and Scandinavia, respectively (Janoski, 1998;Preuss et al., 2003;Green & Janmaat, 2011). Scholars often draw on these regimes to explain differences between European countries in immigration, integration and naturalization policies Brubaker, 1992;Favell, 1998;Koopmans et al. 2005), citizenship education and educational policies regarding the incorporation of immigrants (Mannitz, 2004;Osler & Starkey, 2009;Qureshi & Janmaat, 2014) and policies on welfare and social benefits (Esping Andersen, 1990).
Few studies have examined whether these regimes are also reflected in people's civic attitudes and behaviours. Green & Janmaat (2011) have done so, they have explored such attitudes as part of more encompassing 'regimes of social cohesion' characterising different European societies (see also Dragolov et al., 2016). Hoskins et al., (2015) focused on young people in their study of whether liberal, republican and critical/cosmopolitan regimes of citizenship shape civic competences. Levanon & Lewin-Epstein (2010) explored whether the citizenship regimes developed by Castles & Miller (2003) are reflected in people's opinions on the criteria for granting citizenship. However, none of these studies explored whether the clusters of civic dispositions that they found are lasting. Nor have they examined whether these clusters can also be identified at the global level amidst other distinct and geographically concentrated bundles of civic dispositions. Addressing these omissions, we will explore whether there is evidence of lasting clusters of civic dispositions in Europe that correspond to the regimes of citizenship noted above and whether such clusters can still be observed from a global vantage point.
It is important to examine these questions as the dominant approaches in the field of civic culture have tended to understand civic values, such as tolerance, trust, equal treatment, public spiritedness, and political and civic engagement, as a coherent, one-dimensional set of qualities. Writing within the modernist paradigm, Inglehart & Welzel (2005) and Welzel (2013), for instance, have argued that such attitudes and dispositions in their aggregate form a consistent syndrome of "self-expression" or "emancipative" values at the societal level, which they see as the product of socio-economic development. They contend that these values, in turn, help to bring about and sustain a responsive and participatory democracy. Others contest the claim that civic culture is shaped by socio-economic condition and see it as a more exogenous force with deep roots in the past (e.g. Putnam 1993). Scholars examining civic culture from an institutional perspective turn the causal arrow around and see civic culture as little more than the product of democracy (e.g. Schmitter & Karl 1991;Rose, 1997;Jackman & Miller, 2005). Yet, all these perspectives do not challenge the proposition that civic culture is composed of a coherent set of dispositions. Indeed, Putnam (1993; considers these dispositions to cluster both within individuals and -in an aggregated form -at the level of societies. According to Rice & Feldman (1997), despite disagreement on exactly what dispositions make up this 'syndrome' of civic culture, there is a fair amount of consensus on its key ingredients, which include the ones mentioned above.
However, if civic attitudes and dispositions cluster in different ways at the societal level and comprise enduring and qualitatively distinct "regimes" of civic culture characterising specific European regions, and if these regimes are even discernable at the global level, then the phenomenon of civic culture assumes a high degree of complexity. Such complexity will invalidate approaches seeking to measure the concept as a coherent one-dimensional phenomenon across societies and to establish straightforward linear relationships of the phenomenon with other social conditions and outcomes, such as socio-economic development and degree of liberal democracy. In other words, such complexity would render civic culture an unpredictable, path dependent phenomenon. Interventions aimed at fostering civic values would not have the same effects everywhere because of the qualitative differences between the various civic regimes, complicating any efforts to promote such values.
The earlier mentioned studies of Green &Janmaat (2011) andHoskins et al., (2015) offer evidence that civic dispositions indeed cluster in different ways in their aggregate across countries. Thus, the former found that tolerence, support for equal treatment and civic participation are all relatively high in the United States and Canada, while Portugal, France and Spain combine equally high levels of tolerance and support for civic equality with low levels of civic participation. In similar vein, the latter find high levels of support for social justice values to coincide with high levels of civic knowledge and skills in the Scandinavian countries and high levels of support for civic duties to combine with a high appreciation for political participation in Italy and Greece. These findings only further strengthen the case for assessing whether civic attitudes are clustered in ways that correspond to citizenship regimes.
Another reason to examine whether such regimes are reflected in peoples' attitudes is the existence of 'top-down' approaches in political science, which argue that public opinion is significantly shaped by politics, the media and similar elite activities (Leruth & Taylor-Gooby, 2019). One of such approaches is Jacob and Shapiro's (2000: xiii) theory of "crafted talk". According to these scholars, politicians use public opinion polls to develop a media and communications strategy -"crafted talk" -aimed at bringing public opinion in line with their own policy agendas. If politicians and other opinion leaders are successful in shaping people's values and preferences and if they are drawing on the distinct citizenship regimes noted above -such as the French president Emmanuel Macron invoking a republican spirit and the values of the French Revolution in his speeches (Baruch, 2019) or the remark of the British Prime Minister Boris Johnson that the corona measures may have been less effective in the United Kingdom because of the "freedom-loving Brits" (Peck 2020) -then one might expect these regimes to also be reflected in people's civic attitudes and behaviors. Again, the aforementioned studies found evidence for this. While Green &Janmaat (2011) andDragolov et al., (2016) uncovered a liberal, English-speaking cluster of countries combining high levels of civic participation with strong preferences for freedom and merit and a cohesive Scandinavian group of countries with high level of trust and strong beliefs in civic equality, Hoskins et al., (2015) found the civic attitudes among young people in Southern Europe to be in agreement with a republican ethos of citizenship. In addition, Levanon & Lewin-Epstein (2010) found public opinion on the rules governing access to citizenship to reflect a pluralist regime in English-speaking settler states and an assimilationist regime in France and Britain.
We wish to emphasize that the objective of the current paper is not to demonstrate that there are lasting clusters of civic dispositions reflecting the aforementioned citizen-ship regimes. It addresses this question in a more explorative fashion in order to assess the limitatons of approaches considering civic culture as a coherent macro-level phenomenon that is conceptually equivalent across countries. Another crucial point to make is that the proposition about people's attitudes reflecting distinct citizenship regimes concerns the way these attitudes are clustered in their aggregate. In other words, the argument is that distinct combinations of attitudes should be observable at the national level, such as -say -relatively high levels of appreciation for individual libertly combined with relatively low levels of support for state intervention in countries with a liberal regime of citizenship. Regime theory does not make any claims about how these attitudes cluster within individuals (Green & Janmaat, 2011). This feature has implications for the modelling strategies, as explained further below.
The next section discusses the different citizenship regimes in greater detail and draws hypotheses from these regimes regarding clusters of civic dispositions in distinct European regions. Subsequently we discuss the European Values Study (EVS) and World Values Survey (WVS), as the data sources used for this paper, and the survey items selected as indicators for the different citizenship regimes. The findings section reviews the evidence for citizenship regimes as reflected in people's civic attitudes at the European and global level. The last section discusses the implications of the findings for modernization theory and the conceptualisation of civic culture.

Different regimes of citizenship
As noted above, the literature on citizenship and traditions of political thought identifies a liberal, a republican and a conservative/communitarian regime (or theory) of citizenship, which are seen to predominate in the English-speaking countries, France and Southern Europe, and the German-speaking states and the Low Countries, respectively. While in the liberal regime the emphasis is on individual rights and freedoms which need to be protected from a "tyrannical" state, the state is viewed as an essential enabler of individual interests and liberties in the republican regime (Green & Janmaat, 2011). In the conservative regime the obligations of citizens towards the community are pivotal (Janoski, 1998).
In the liberal regime individual rights and freedoms are central. Citizens are encouraged to adopt a critical attitude towards authority, particularly that of the state, as such authority can constrain individual autonomy (Abowitz & Harnish, 2006). In view of this critical attitude towards the state, the state is not entrusted with major responsibilities and is expected to do little more than to protect rights and property. There is no moral obligation on citizens to contribute to the national community or the state (Hoskins et al., 2015). Consequently, although political participation is encouraged, people do not consider it their duty to vote in national elections or to be engaged with societal affairs (Delli Carpini & Keeter, 1996). They do, however, organise themselves at the local level to protect their interests, keep the government in check, or contribute to the greater good through charities and other activities. Thus, marginal involvement at the national level is matched by a vibrant civil society at the grassroots level which is believed to foster virtues like trust and cooperation (de Tocqueville 1969;Putnam, 2000;Green & Janmaat, 2011). A 'thin' socialization is preferred focussing on basic political knowledge and tolerance to ensure peaceful relations between different groups (Crick & Heater, 1977;Abowitz & Harnish, 2006). If this regime is reflected in civic attitudes and dispositions, we would expect to find relatively high levels of civic participation and ethnic tolerance in combination with relatively shallow support for state involvement in society and low levels of political participation and public spiritedness in English-speaking countries (see Table 1, which offers an overview of the hypothesized civic attitudes clusters).
By contrast, a commitment to the political community of the nation is at least as important as individual rights in the republican regime of citizenship, the roots of which go back to the French Revolution (Abowitz & Harnish, 2006). Hence political participation, particularly at the national level, is not so much seen as a right that people can choose to exercise or not but as a key civic duty. Citizens need to be actively engaged in public affairs in order to prevent public government from becoming corrupt and autocratic (Lovett, 2010;Hoskins et al., 2015). Public education should not only foster political literacy and tolerance but also public spiritedness, patriotism, civic equality and a commitment to act for the common good in order to foster engagement with and loyalty to the nation as political community (Hohohan 2002;Abowitz & Harnish 2006). Civil society is not appreciated, as in the liberal regime, but distrusted because organizations mediating between the state and individual citizens are seen by the political elite as threatening the unity of the state and undermining the principle of political equality (Preuss et al., 2003). Vertical state-citizen relations are more important than horizontal bonds between citizens resulting in relatively high levels confidencce in state institutions and low levels of social trust. Immigrants are welcomed and incorporated in society provided they endorse the values of the republic and keep their cultural and religious traditions private (ibid. ;Favell 1998;Mannitz, 2004). Consequently, we expect France and the countries of Southern Europe to show relatively high levels of political participation, engagement with pubic affairs, patriotism, tolerance and insitutional trust and high levels of support for the principle of equality before the law. This is combined with relatively low levels of civic participation and social trust. (see Table 1) In the conservative regime intermediary organisations are not considered a threat to social cohesion but as the glue that holds society together, particularly those with deep historical roots such as churches and professional organisations tracing their origins back to guilds. They are seen as legitimate channels of representation and play a key role in political decision-making through their semi-public status and formal consultation rights (Preuss et al., 2003). They also reflect uneven and hierarchical forms of representation, however (Janoski, 1998). The conservative regime embraces a communitarian conception of the citizen: socialization into the values and identity of the community provides individuals with a sense of direction and purpose in life and a basic sense of trust (Etzioni, 1993). This conception has been associated with Christian Democracy, which has been the dominant political force in Germany, Austria and the Low Countries in the second half of the 20th Century. Because of a long tradition of state involvement in social affairs through corporatist arrangements with unions and employer organisations (Esping Andersen, 1990), support for a strong role of the state in society is relatively high (Green & Janmaat, 2011).
Scholars have further claimed that a thick ethno-cultural understanding of nationhood prevails in these countries (notably in Germany and Austria), which has been associated with exclusionary attitudes and policies towards immigrants and foreigners (Kohn, 1944;Brubaker, 1992). More generally the norm of civic equality is said to be not well developed, not only because of exclusionary identities and intermediary organisations but also because of the prevalence of traditional attitudes on gender roles (Esping Andersen, 1990). Exclu-sionary understandings of the nation do not translate into a virulent nationalism, however, due to the trauma of defeat or occupation in World War II. These events have had a lasting effect on the population in terms of discrediting nationalism and related phenomena such as national pride and patriotism (Dogan, 1994). We therefore anticipate the German and Dutch speaking countries to combine relatively high levels of civic participation and strong support for state involvement with relatively shallow support for civic equality and relatively low levels of ethno-racial tolerance and patriotism (see Table 1).
Although these three regimes have been identified by many scholars, there is no consensus on them. For instance, Janoski (1998) does not mention a republican regime but does discern a distinct form of citizenship which he considers to prevail in the Scandinavian countries. In his view, this form reflects a social-democratic ideology, emphasizing inclusion, equality and solidarity and calling for state intervention in society to counter inequality. Societies where this ideology is a dominant factor in the political landscape are marked by egalitarian beliefs, social trust, trust in the state, and mass participation in both politics and civil society, which is a characterization confirmed by Green & Janmaat (2011) in their study of different regimes of social cohesion. Delhey & Newton (2005) also found levels of social trust to be exceptionally high in the Nordics. Consequently, we would expect these countries to show relatively high scores on tolerance, civic equality, public engagement, support for state intervention, social and institutional trust, and civic and political participation (see Table 1).
More controversial still is the existence of a separate post-communist regime of citizenship prevailing in Eastern Europe. Political philosophers and education scholars have certainly not identified it a distinct normative regime and one that should govern state-citizen relationships and inform citizenship education. Nonetheless, one could argue the case for an empirical manifestation of such a regime in view of the many scholars claiming that a strong, ethnic form of nationalism has come to prevail in Eastern Europe after the collapse of communism and one that shapes both state policies and solidarities among citizens (e.g. Linz and Stepan 1996;Kolstoe 2000;Oltay 2017). Governments of both new and existing states have seized on this nationalism to justify state independence, pursue nation-building projects and base their citizenship and naturalization policies on jus sanguinis (ethnic descent) (Kolstoe 2000; Ragazzi 2014). Others discern a rising tide of illiberal ideology in Eastern Europe espousing a coctail of family values, christian heritage, patriotism, xenophobia and loyalty to authority (Szelényi and Csillag 2015;Pogani 2018). Political leaders such as Victor Orban, Jaroslaw Kaszynski and Vladimir Putin, invoke this ideology to dismiss human rights, the rule of law, and cosmopolitanism as Western values foreign to the region and to weaken the democratic checks and balances of the political system (Hanley and Vachudova 2018;Przybylski 2018). Scholars have argued that the roots of ethnic nationalism and illiberalism lie in the communist period (Zakaria, 1997;Pogani, 2018), with communist rule seen as having destroyed civil society and as having created an atomistic society composed of individuals distrustful of both state institutions and fellow citizens Uslaner 2003). In this environment, Schoepflin (2000) contends, ethnic and family loyalties prevailed as these were the only bonds people could rely upon to counter feelings of alienation and isolation. This regime leads us to hypothesize that the post-communist states in Central and Eastern Europe combine relatively low levels of tolerance, civic equality, public engagement, social and institutional trust, and civic and political participation with relatively high levels of patriotism (see Table 1).

Date sources
We decided to use survey data from Waves 2 (1990), 3 (1999), 4 (2008) and 5 (2017) of the European Values Study (EVS) to explore the durability of the hypothesized clusters. 1 We further complemented the Wave 5 EVS data with data from Wave 7 (2017-2020) of the World Values Suvery (WVS) to assess whether any clusters identified at the European level can also be observed at the global level. 2 The advantages of the combined EVS/WVS survey are that it (1) covers an extensive period of time (27 years) allowing us to assess the durability of the hypothesized clusters, (2) includes appropriate indicators for all the theorised components of the citizenship regimes, and (3) covers a sufficiently wide collection of countries at both the European and world level. Other possible surveys, such as the European Social Survey, are not fulfilling one or more of these criteria. The EVS/WVS includes nationally representative samples of the adult population ranging between 700 and 3500 respondents per country/wave combination, with the vast majority of the samples varying between 1000 and 1500 respondents. As noted before, in view of the claims by regimes theorists we are interested in the prevalence of civic dispositions in a country and in how these dispositions combine at that level. We therefore aggregated individual-level survey data to the country level to create countrylevel indicators of the 10 components shown in Table 1. These indicators either represent means or percentages. In this approach we follow Inglehart and Welzel who have argued that individual civic dispositions only show a robust coherence in their aggregate and that only country-level stocks of civic dispositions can meaningfully be related to other societallevel phenomena such as socio-economic development and democracy (Inglehart & Welzel, 2005;Welzel & Inglehart, 2016). Different from these authors, however, we do not assume these stocks to form a coherent one-dimensional syndrome on which countries can be ranked but want to explore the possibility that they consist of a unique combination visible in a distinct group of countries (see further below, where we explain the consequences of this proposition for the methods of analysis). Our reliance on aggregated data means that the degree to which the national EVS/WVS samples are representative is an important issue. The EVS/WVS includes a weight (g_weight in the EVS; w_weight in the WVS) that corrects for biases in the sample on age, gender, educational level and region and regional biases in the samples (EVS 2020). Consequently we applied this weight in calculating the aggregate indicators. The database with aggregated data that we subsequently created includes data on 28 European countries at four points in time 3 and data on 53 countries at a single point in time. We will use the former for the analysis of cluster durability at the European level and the latter for the analysis at the global level. All variables are standardised to a mean of zero and a standard deviation of one. We used the SPSS default option of data imputation, which produces five datasets and a pooled one with imputed data, to address missing values on individual indicators and prevent data loss. This was only necessary for the 1990 and 1999 waves as we had complete records for the 2008 and 2017 waves.

Measuring the components of the postulated regimes of civic dispositions
Being aware of the problem of measurement validity, we sought to select survey items that represent the components of the theorised clusters (see Table 1) as closely as possible semantically. As we further wanted to have some minimal reassurance that the measures of these components are broadly conceptually equivalent across national contexts, we tried to find at least two strongly interrelated items to represent each of the components. These items should be signficantly correlated to each other in the vast majority of countries. The online Appendix A, which provides an overview of these items and their correlations for all the components, shows that this criterion is met. Ideally, each concept is measured by at least 4 items and multiple group confirmatory factor analysis is used to assess conceptual equavalence across countries (Aleman and Woods 2016). However, adhering to these standards would reduce the number of components that can be explored to just one (political participation) due to lack of appropriate items. The compromise that we adopted allows us to address the hypothesized substance of the regimes comprehensively while simultaneously ensuring a basic level of cross-national comparability of the indicators developed. For each component that is tapped with several items (i.e. all but one -see further below), we calculated a synthetic index representing the mean of the aggregated indicators derived from these items. Before calculating this mean the indicators were standardised to ensure they have equal weight.
The component of political participation was tapped with an index combining five indicators. Four of these concern forms of political participation queried in the EVS: (1) having signed a petition, (2) having taken part in a boycott, (3) having attended a lawful demonstration, and (4) having joined a strike. These four items were strongly intercorrelated everywhere with Cronbach Alpha reliability values ranging between 0.60 and 0.91 for all 53 countries (see Appendix A). The four indicators represent the percentages of people confirming taking part in these activities. Unfortunately, voting, as the most common form of political participation, was not asked in all waves. We relied on data from the Institute for Democracy and Electoral Assistance (IDEA) (https://www.idea.int/) on turnout in parliamentary elections to capture this political activity. The index of political participation was created by first creating a sub-index of alternative political participation representing the average of the four EVS indicators and then by averaging this sub-index with the national turnout data on voting. The index thus gives voting a 50% and each of the EVS indicators a 12.5% weight in the overall measure. Voting and the alternative participation sub-index were positively correlated at the national level in all EVS waves. We excluded countries with compulsory voting systems, such as Belgium, Australia and many Latin American countries, and countries with pronounced authoritarian regimes making political participation practically impossible, such as China and Zimbabwe, from the sample. This reduced the number of countries to 53 globally.
We created an index expressing the percentage of people saying that politics is very or quite important in their lives and the percentage of people saying they are very or somewhat interested in politics to measure public spiritedness / engagement. We assumed here that involvement with and interest in politics indicates an engagement with societal affairs more generally. The EVS/WVS includes a battery of items asking about belonging to various organisations, which can be used as indicators for civic participation. From this battery we selected belonging to a church, a cultural organisation, a sports club and a union as relevant indicators, not only because membership of these organisations is relatively widespread but also because the first three of these organisations are primarily local phenomena (certainly sports clubs). This local character expresses the grass roots nature of civic participation as theorised in the liberal and the conservative regime well. The index tapping civic participation thus represents the percentages of people stating belonging to these four organisations.
The concept of civic equality was measured with an index capturing the percentages of people disagreeing with two statements on unequal treatment of women and immigrants in times of crisis ("when jobs are scarce, men/the native population have more right to a job than women/immigrants"). We tapped ethnic/racial tolerance with an index representing the percentages of people not mentioning "people of a different race" and "immigrants/ foreign workers" as "people one one would not like to have as neighbours". Patriotism was captured with an index expressing the percentage of people stating they are very or quite proud of being [nationality] and the percentage of people answering the question whether they would be willing to fight for their country in times of war in the affirmative.
State involvement was measured with an index based on an item on the role of the state in providing for people, with response categories ranging between 1 = "individuals should look more after themselves" and 10 = "the state should take more responsibility to ensure everyone is provided for", and an item about income equality, with categories ranging between 1 = "there should be greater incentives for individual effort" and 10 = "incomes should be made more equal". The index represents the mean of the national averages of these items. Higher values denote greater support for state involvement. We considered these items to be appropriate indicators of state involvement as they concern intervention in social affairs, which is postulated to have broad public support in the countries associated with a socialdemocratic, conservative and republican regime of citizenship.
Institutional trust was based on three items tapping confidence in the police, the justice system and parliament, as three key institutions of state authority. According to Sønderskov & Dinesen (2016) confidence in these institutions reflects a belief in the fairness and effectiveness of these institutions and notably in their ability to enforce the law. The three items were strongly intercorrelated everywhere, with Cronbach alphas varying between 0.53 and 0.86 across countries (see Appendix A). The index represents the average of the percentages expressing "quite a lot" and "a great deal" of confidence in these institutions.
Finally, social trust was tapped with a single item, namely the percentage of people saying that "most people can be trusted", which is one of the most often used indicators of social trust (Delhey & Newton, 2005). It is further said to be a key measure of trust in unknown others (Uslaner, 2002), which is considered to be a vital component of bridging social capital (Putnam, 2000). We could not find an accompanying item to strengthen the measurement of this component as other items tapping social trust varied across waves. However, the item showed strong positive correlations across the board with trust in other people in the country (in the four EVS waves) and with trust in people of a different nationality (in the WVS), affirming its ability to tap trust in strangers.

Analytic strategy
In order to assess whether the theorised citizenship regimes are reflected in people's civic attitudes we need to examine the hypothesized substance of these regimes, their expected country membership, and the stability of the regimes on these two dimensions. We explore the substance by first assigning countries to the regime they are expected to be part of and then by calculating the regime means on the components. We standardised the components to an international mean of zero and a standard deviation of 1 to enable an easy identification of relatively high and low regime scores. The results can then be matched to the proposed scores on the components shown in Table 1. We also use this approach to assess the stability of regimes in terms of their hypothesized substance by comparing the regime means on the components across the four EVS waves.
A disadvantage of this approach is that it is merely descriptive and does not offer statistical criteria to determine the substantive profiles of regimes. However, more advanced analyses to explore the dimensionality of the components, such as confirmatory factor analysis (CFA) or item response theory (IRT) cannot be applied. Not only does the low number of observations rule this out, also the nature of the conjectured regimes. These regimes are hypothesized as distinct combinations of components characterizing only some observations (i.e. countries), not as latent dimensions among linearly related variables applying to all observations, which CFA and IRT are designed to identify. To illustrate this with a concrete example, regime theory would predict that high levels of civic participation go together with low levels of political participation in the liberal regime countries but not necessarily that the opposite applies (i.e. low levels of civic participation and high levels of political participation) in many other countries. For this reason the straightforward descriptive approach of comparing regime means on citizenship components is the most appropriate for addressing the propositions of regime theory regarding the content of civic attitude regimes.
Ideally, the anticipated country membership of a regime is investigated with latent profile analysis (LPA) as this type of cluster analysis permits the identification of an optimal solution in terms of the number of clusters generated. One can assess the optimal solution using two criteria: the change in loglikelihood and the Bayesian Information Criterion (BIC) (Weller et al., 2020;Schmidt et al., 2021). The former identifies the best solution as the model with a loglikelihood that represents a significant improvement over the next best model and that in turn is not succeeded by a model with a signficantly better loglikelihood. The BIC takes both the number of parameters (i.e. clusters) and the likelihood into account in assessing regime fit. The best model is the one with a minimum BIC value, i.e. a model that explains the maximum variance relative to the number of parameters (as few as possible) in the model (Bishop, 2006). LPA generates probabilities of cluster membership. We will use LPA to assess the visibility of the postulated regimes at the global level since the number of observations at this level (53) is sufficiently large to do so.
However, we cannot use LPA to examine whether the proposed regimes are observable within Europe due to insufficient observations (28 or less). For instance, a model with a two cluster solution already produces a number of free parameters that exceeds the number of observations, which makes the standard errors untrustworthy. We therefore turn to hierarchical cluster analysis (HCA) since this type of cluster analysis can address data with small number of observations. HCA runs in stages, forming a new cluster of the two most similar cases/clusters at each step in the analysis. As it proceeds, it creates ever fewer but internally more diverse clusters (Cramer, 2003). Although this analysis is merely exploratory and does not offer statistical criteria to identify an optimal cluster solution as LPA does, its output, visualised as dendrograms, allows for a detailed scrutiny of the internal cohesion of clusters and the degree of dissimilarity between them. We use it to assess whether the expected regimes emerge in Europe as to their country membership and whether these regimes are stable across the four EVS waves. Table 2 shows the mean scores on the citizenship components by postulated regime and EVS wave (the note under the table explains which countries have been allocated to which regimes). Scores in bold and underlined style are more than one SD from zero as the international mean and signify values very much in accordance with the hypothesized scores displayed in Table 1; those in normal and underlined style are less than one SD from this mean and denote values somewhat in line with those scores; those given in italics and underlined style are less than one SD from this mean and denote values somewhat not in line with those scores; those in bold, italics and underlined style are more than one SD from this mean and represent values not at all in line with hypothesized scores. Scores that are not underlined denote values for which no particular score was predicted.

The substance of citizenship regimes
Across all regimes and waves the scores in normal underlined style outnumber those in italics underlined style by a ratio of 3 to 1 (107 to 36), indicating a fair degree of support for the idea that citizenship regimes are reflected in people's civic attitudes in substantive terms. Nonethess, there are conspicuous differences between regimes and components in the degree to which the mean scores match the expected ones. In terms of regimes we see quite strong evidence for the existence of a social democratic regime prevalent in the Scandinavian countries as testified by that cluster's high and very high scores on political engagement, ethnic tolerance, civic equality, civic and political participation, and social and institutional trust, which are all as anticipated. Only on state involvement is this cluster not showing the expected scores with mean values dipping below the international mean. We can find almost equally strong support for a post-communist regime prevailing in Eastern Europe, as shown by that region's relatively low scores on civic equality, ethnic tolerance, civic and political participation, and social and institutional trust, which again are all as expected. Only on patriotims does that region not show the antipated pattern.
The existence of a liberal regime associated with the English-speaking countries is also confirmed, but to a more moderate degree. As expected, levels of tolerance are relatively high and support for state involvement is relatively shallow. On the other three components for which particular scores were predicted, we see a more mixed and changeable pattern of scores. In the last wave the scores on political participation and public spiritedness (higher than expected) and civic participation (lower than expected) are not as hypothesized, but we need to note that the regime is only represented by Great Britain in this wave. Support is more shallow for a republican regime in terms of proposed substance. As exected, levels of civic participation are relatively low and levels of tolerance and of support for state involvement comparatively high. On the other relevant components we see quite a changeble pattern of scores, which in itself is not in accordance with the notion of regimes as quite enduring phenomena. This is most pronounced on public spiritedness, the scores of which change from − 1.03 (very much not as expected) to 0.16 (somewhat as expected), but we also see marked changes on institutional trust (from − 0.90 to 0.28), civic equality (from − 0.50 to 0.64) and political participation (from − 0.06 to 0.67). Overall 26 scores are as expected while 10 are not.
The data do not really confirm a conservative regime of civic attitudes as the number of unexpected scores on the relevant components is as high as the number of expected ones. The high scores on civic participation and the low scores on patriotism are as hypothesized but the relatively high scores on civic equality and ethnic tolerance are not. The latter are in line though with previous research as Levanon & Lewin-Epstein (2010) have also found relatively strong support for inclusive notions of citizenship in Germany and Austria.
As to patterns on individual components, we see that the scores on civic participation, institutional trust and social trust are almost all in agreement with the expected values. These patterns are in agreement with other research finding relatively low levels of civic participation and trust in Southern and Eastern Europe (Coffe and van der Lippe 2009; Immerfall et al., 2010;Borgonovi, 2012). The low scores on civic equality and ethnic tolerance in Eastern Europe are in line with the more exclusionary attitudes towards immigrants that Borgonovi (2012) observed for this region. The scores on civic participation, ethnic tolerance and social trust are quite consistent over time, suggesting a certain degree of permanency of the cross-national differences in these aspects of citizenship. This is in agreement with the notion of regimes as lasting constellations of characteristics. In contrast, the scores on political participation, public spiritedness and state involvement are quite volatile and often not in accordance with expectation.

Citizenship regimes in terms of country membership
We ran a series of hierarchical cluster analyses (HCA), using the nine components as the input variables, to assess whether groups of countries emerge that correspond to citizenship regimes in terms of expected country membership. Figure 1 displays the results of the HCAs as a series of dendrograms, one for each EVS wave. A dendrogram shows how cases merge to form clusters and how clusters merge with one another as the distance between the clusters increases. This distance is displayed on the x axis. The branching points (indicated by vertical lines) show the distance between two clusters when they are merged. The closer a branching point is to zero, the more internally cohesive a cluster is but also the less externally distinctive it is.
Looking across the four dendrograms, we can identify several country groups that both conform to the anticipated regimes in terms of country membership and that show some degree of continuity across the four waves of the EVS. The most clearly distinguishable cluster is that of the Scandinavian / social-democratic group of countries. As expected, this group comprises Denmark, Sweden and Norway in all the waves. Iceland and Finland, though, join it in only some of the waves. This group does not so much stand out for its internal cohesion (as the core group of countries merge only at a distance of 5 or more in most of the waves) but for its external distinctiveness: across the first three waves this group is the Table 2 The substance of citizenship regimes in terms of civic dispositions Liberal regime 1990 1999 2008 2017* 1990 1999 2008 2017 1990 1999 2008 2017 Political participation last one to join the other groups. Its separate identity is, however, somewhat undermined by the Netherlands joining it in the first three waves.

Components
One can also identify a liberal group of countries comprising Great Britain, Ireland and Northern Ireland. This group shows a high level of internal coherence and an enduring one because the three countries already merge at a very small distance in each of the first three waves. Only in the last wave does the cluster not appear but this is because of the absence of Ireland and Northern Ireland in that wave. The liberal group is not scoring well on external distinctiveness, however, because it is already joined by France at an early stage in the merging process in the first three waves and by Spain and Italy in the second wave.
Finally, the post-communist countries emerge as a clearly identifiable and fairly coherent group of countries. Interestingly, this group, which consists of many countries, becomes more coherent and distinctive over time. While it still forms a loose collection of several groups of countries in the first two waves (and is joined by some Southern and Central Fig. 1 The clustering of countries based on aggregate civic dispositions across four waves of the EVS European countries), it combines all the post-communist countries in the last two waves and emerges as a separate cluster at an earlier stage in the merging process.
The regimes that by and large fail to emerge in terms of the anticipated country composition are the republican and conservative regimes. The countries associated with a republican regime mix with the post-communist and liberal countries in the first three waves in different combinations. Only in the last wave do we see an internally coherent group of countries emerge consisting of France, Spain and Italy. Thus, even though a clear republican cluster can be identified in the last wave, the volatile behaviour of its members in the first three waves does not indicate a stable group of countries. A conservative group of countries cannot be identified at all. Austria, Germany and The Netherlands, as the countries expected to be part of a conservative regime, each join different groups in Waves 1, 3 and 4. Only in the second wave do Germany and Austria form a fairly coherent cluster. In sum, we find practically no evidence for a republican and a conservative regime of civic attitudes, whether regarding the proposed content, the anticipated country membership or the expected durability.

Distinctiveness of european citizenship regimes at the global level
Can the country clusters of the previous section still be seen when we zoom out to the global level? We carried out a series of latent profile analyses (LPA) on the nine citizenship components to address this question. We ran several models specifying different numbers of clusters and used the two criteria explained before to choose the model with the optimal cluster solution. Table 3 shows the fit statistics of these models. We see that the loglikelihood continues to improve signifantly moving from a model with a single cluster to a model producing four clusters. 4 This suggests that we have to choose the latter. However, the BIC, which also takes the number of parameters into account, shows that the model yielding three clusters has the lowest BIC. We therefore select this model as the one with the optimal number of clusters. Table 4 shows the composition of the clusters generated by the chosen model. We see that the clusters are not coinciding with any of the regimes identified previously. The first cluster includes most (but not all!) of the post-communist countries and several Asian states where Islam is the main religion. The second cluster is the most diverse as it includes countries 4 We could not assess models yielding five or more clusters since the number of free parameters in these models exceeds the number of observations. Criterion; LRT = Likelihood-ratio test from Western Europe (Great Britain, France, Italy, Spain), Latin America and the Caribbean (Colombia, Nicaragua, Puerto Rico), Africa (Nigeria), Asia (Japan, South Korea, Malaysia, Indonesia, Philippines, Taiwan) and the former communist world (Estonia, Kazakhstan, Poland, Slovenia, Tajikistan). Affluent western countries comprise the third cluster, including all the Scandinavian states, the states associated with a conservative regime and two English-speaking settler states. Thus, the liberal, social-democratic and post-communist regimes identified in the previous section disappear when the vantage point shifts to the global level. Countries associated with these different regimes either come together, as in   Fig. 2 The substantive profile of the three clusters identified at the global level the third cluster, or they join wholly new clusters with very different countries, as in the second cluster.

Number of clusters
What is the substantive profile of these clusters? As we can see from Fig. 2, which shows the cluster means on the nine components as error bars, the clusters differ from each other mainly in the level of civic dispositions, not in the kind of such qualities. The third cluster has the highest overall scores. It has significantly higher mean scores than the second and the first cluster on seven of the nine components (as shown by the non-overlapping confidence intervals). In turn, the second cluster has significantly higher mean scores than the first cluster on six components. Only on state involvement and patriotism do we not find significant differences between the clusters. These results provisionally suggest that a coherent one-dimensional syndrome of civic culture does exist at the global level.

Discussion: Civic regimes at the european but not at the global level
Is there evidence of enduring regimes of citizenship as reflected in people's social and political attitudes? The current paper has shown that the answer to this question very much depends on the vantage point of the analysis. At the European level we can identify some groups of countries that conform to the theorised regimes in terms of both predicted substance and country membership. These groups also show a high level of continuity over a 30 years period. Thus, as expected, a social-democratic regime emerges, consisting of Scandinavian countries and combining relatively high levels of social and institutional trust, tolerance, civic equality, public spiritedness, and civic and political participation. We can also identify a liberal regime prevailing in the English-speaking countries and characterised by a relative aversion to state involvement in society and relatively high levels of tolerance. Further evidence was found for a post-communist regime uniting the countries of Central and Eastern Europe around relatively low levels of trust, tolerance, civic equality and civic and political participation. The evidence for a republican regime said to prevail in Southern Europe was less convincing, particularly regarding its continuity. A conservative regime did not emerge from the data on any of the three dimensions of regime distinctiveness (substance, country membership and continuity). In sum, although there is limited or no empirical support for some of the postulated regimes, we can see the expected patterns for the majority of these regimes.
This picture changes dramatically when we zoom out and assess whether the theorised regimes can also be observed at the global level. The three clusters generated by an LPA analysis at this level did not coincide with any of the identified regimes at the European level. They consisted of a cluster uniting most of the post-communist countries and several Asian countries, a very diverse cluster with representatives from all continents and a cluster of western countries. These clusters can be placed on a ranking order in terms of their mean scores on the citizenship components, with the latter outperforming the other two clusters and the second outperforming the first one on seven of the nine components. The cluster of western countries stands out as a relatively cohesive and distinct group of countries. It essentially unites countries that are affluent, post-industrial, majority white and mainly protestant democracies.
As to implications for the wider literature on citizenship and civic culture, our findings complement rather than challenge modernization theory. As noted before, this theory claims that socio-economic development gives rise to a distinct syndrome of values, which, in turn, influences a country's political development (Inglehart 1997;Inglehart & Welzel 2005;Welzel, 2013). They would have challenged this theory if they had revealed the existence of distinct citizenship regimes at the global level combining the citizenship components in different ways. In its place the clustering of components, such as tolerance, civic equality, trust and civic engagement, in rich countries with liberal democratic polities confirms the associations proposed by the theory, all the more so as this clustering is largely in agreement with postmaterialism, as the coherent syndrome of values proposed by the theory's advocates. They complement it, however, by showing that at a more proximate (European) level qualitatively different clusters of citizenship attitudes emerge that are relatively stable. These clusters reflect distinct cultures, institutional arrangements and political histories and largely align with other enduring differences in kind between countries, regarding, for instance, the market and knowledge economy (Hall & Soskice, 2001;Thelen, 2004), the welfare state (Esping Andersen, 1990), social cohesion (Green & Janmaat, 2011) or immigration and naturalisation policies (Koopmans et al. 2005).
Thus our argument is that distinct clusters of citizenship attitudes emerge that do not conform to the coherent syndrome of civic attitudes postulated by modernist thinkers as soon as one abandons the global level and zooms in on Europe. To our knowledge we are the first to make this specific argument. To be sure, other scholars have claimed that there are distinct cultures characterising whole world regions and civilizations but they have tended to frame their arguments in opposition to modernist theory (e.g. Bendix 1964;Huntington, 1996;Eisenstadt, 2000). It is by working across different levels of analysis (European and global) that we have been able to ascertain that differences in kind can co-exist with differences in the degree of citizenship attitudes. If our analyses had solely focused on Europe, we might have misleadingly concluded that citizenship regimes are primary. Conversely, an analysis limited to the global scale would have made us blind to qualitatively different constellations of citizenship attitudes at lower levels. The cross-level approach thus has distinct benefits.
An interesting task for future research is to examine whether similar patterns can be found for other world regions. The 'Asian values' literature certainly suggests that this could be the case for Asia as it postulates a distinct set of values prevailing in East Asian societies with a confucian cultural heritage: respect for authority, social harmony, an emphasis on the community instead of individual rights, and a belief in hard work, education and selfimprovement (Chan, 1993;Glazer, 1997;Tu, 2000). Possibly, similar unique combinations of values can be found for the Indian sub-continent, Southeast Asia, Central Asia or the Middle East.
Our argument comes with one important qualification, however. Although the indicators that we have chosen to explore clusters of citizenship attitudes are all well grounded in the theory on distinct European citizenship regimes, they do show a substantial conceptual overlap with the cultural syndrome of civic values suggested by modernization theorists. Thus, tolerance, civic equality, trust, public spiritedness and political participation are also core components of the postmaterialist values syndrome proposed by Inglehart & Welzel (2005). In this sense, it is not surprising that our findings at the global level are in line with the modernization thesis. Critics might argue that our focus on Europe predisposes us to focus on indicators that merely confirm a western-centric conception of economic and cultural development. Rather than challenging this criticism, we invite researchers to prove us wrong by identifying distinct cultural regimes in other world regions that can also be observed at the global level.
Other limitations apply as well. As noted before, we were constrained by the available items in the EVS/WVS in capturing all the components of the theorised regimes of civic culture. We had to reconcile two competing objectives: the desire to cover the components as comprehensively as possible with the available items and the desire to capture the components as robustly as possible with valid and conceptually equivalent measures. The compromise we arrived at, i.e. measuring each component with at least two items that are highly correlated in each national context, is far from ideal but it at least permits a complete assessment of the components while adhering to a minimal level of cross-national equivalence of the measures used. It was not easy to find items adequately covering components such as public spiritedness, social trust and support for state involvement in the economy and society. It would be good if future editions of the EVS/WVS or other major international surveys could develop more items addressing these concepts sufficiently.
Secondly, when conducting the cluster analysis to explore the postulated regimes in terms of country membership, we could not establish the optimal number of clusters in Europe because of the low number of observations at this level. The cluster analysis that we used at this level, hierarchical cluster analysis, gave detailed insight into the internal cohesion and external distinctiveness of clusters but did not offer statistical criteria to determine this optimal number. Our assessment at the European level was therefore incomplete on this aspect and not fully comparable to our assessment of country clusters at the global level for which we could use an analysis yielding an optimal cluster solution. It would be helpful if cluster analyses were developed with criteria to establish the best number of clusters specifically for small samples.
We would also call on methodological experts to develop statistical analyses capable of assessing the clustering of substantive components for subgroups of the sample and not only for the sample as a whole. Such analyses would be ideal to assess the regimes as to their theorised content. The current paper had to resort to descriptive analyses to do so. 7 Appendix A. Interrelations between individual-level civic attitudes in 53 countries worldwide (correlations and alpha reliabilities).Note: for reasons of parsimony we only provide statistics on item interrelationships for the countries participating in the 2017 wave of EVS and Wave 7 of the WVS (i.e. the 53 countries shown in Table 4). The items are also highly intercorrelated in almost all countries in the earlier waves of the EVS.  Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.