Policy congruence and strategic loyalty: which parties nominate candidates dissatisfied with democracy? Evidence from 11 European countries

ABSTRACT This article considers the interplay between the democratic attitudes of candidates and their nomination through political parties. The focus is on candidates who articulate a dissatisfied attitude towards the current status of democracy, and the research interest lies on the parties that might nominate such candidates in national elections. In doing so, the article establishes a link between work on the democratic beliefs of candidates as a specific part of the political elite and literature on party behaviour. The study is grounded in both classical attempts and recent work on political elites and candidate nomination, and its theoretical framework is based on the assumption that parties principally select supportive candidates. Two major mechanisms are investigated: on the one hand, nomination as an expression of policy congruence between the party and its candidates, on the other, candidate nomination as a way to maintain loyalty with the party’s strategic behaviour in parliament. In a first empirical attempt to this research interest, the study analyses data from 76 parties in 11 European countries.


Introduction
Candidates for public office are a specific group within the political elite (Crotty 1968, 260). They transport personal experience, attitudes and preferences into the political system (Rahat, Hazan, and Katz 2008, 665). As citizens who apply for public office, candidates play a crucial role in the interconnection of democratic attitudes and democratic institutions (see Anderson and Guillory 1997, 67). Early studies have shown that members of parliament not only have different normative ideas about the functioning of democracy, but also assess the democratic status quo very differently (Putnam 1971;1973, 187). Variance of attitudes towards democracy, for instance, in terms of populist attitudes, among candidates is also demonstrated in recent empirical work (Lewandowsky, Giebler, and Wagner 2016;Stavrakakis, Andreadis, and Katsambekis 2016). Against this background, their attitudes towards representative democracy is clearly a relevant topic.
Most importantly, candidates represent the 'face' of political parties in elections and prospectively in parliament (Katz 2001b). More precisely, they are the face the party gives itself (Ignazi 1996, 552). Although there are different empirical findings on the homogeneity of party members' attitudes (Kennedy, Lyons, and Fitzgerald 2006;Narud and Skare 1999;Norris 1995), parties nominate candidates who are loyal and in line with the preferences of the leadership (Lesschaeve, van Erkel, and Meulewaeter 2018).
Candidates' attitudes towards democracy are of particular importance in this context since it can be assumed that if candidates are dissatisfied with the status quo, this attitude is not hidden but articulated. If a particularly high level of dissatisfaction is associated with a high importance of this issue to the individual, such persons are potentially more committed to the policy than the leaders of the party (cf. Katz 2001b, 289-290). Dissatisfied candidates may not only be 'renegades' (Cross 2008, 613) but also 'grumblers', critical not only of the given institutions but also of actorsincluding political partiesand compromises, openly articulating their discontent and may, in short, fit the type of 'ideologue' that Kitschelt (1989, 406-407) distincts from more 'pragmatic' party members.
Since parties generally seek to select candidates who are loyal to their strategy, the question is why they would choose dissatisfied individuals who may be more sceptical of the decisions that party leaders have to make. Conversely, a possible link between the attitudes of candidates to democracy and their nomination by political parties is in need of justification. Is there a systematic connection between party characteristics and especially satisfied candidates?
To explain the nomination of dissatisfied candidates, parties are considered from two different actor-specific perspectives. In general, party leaderships ensure to nominate persons who are loyal to their strategy, yet there are different incentives that might influence whether parties support or prevent dissatisfied candidates. For one thing, parties maintain congruence between their policy 'message' and their candidates (Lesschaeve, van Erkel, and Meulewaeter 2018). Therefore, it can be expected that dissatisfied candidates are nominated by parties whose policy profile focuses on the critique of the democratic status quo and its actors, that is, by anti-establishment parties in particular. On the other hand, parties ensure strategic loyalty, that is, candidates who do not oppose more pragmatic approaches or even a (prospective) office-seeking strategy. Particularly powerful parties are therefore careful not to nominate especially dissatisfied candidates. Of course, these mechanisms are overlapping in reality (Strøm 1990). In our case, under the condition of a stronger power position, anti-establishment parties can take a more moderate course and therefore nominate fewer dissatisfied candidates.
The aim of this study is to investigate specific patterns between characteristics of political parties and their candidates' attitudes towards the democratic status quo. In this respect, the study is concerned with the democratic attitudes of candidatesand thus picks up classical work (e.g. Putnam 1971;1973)but adopts the perspective of political parties as actors. The paper thus establishes a missing link between the study of the democratic beliefs of (a specific part of) the political elite and literature on the nomination strategies of political parties (e.g. Katz, 2001b;Lesschaeve, van Erkel, and Meulewaeter 2018;Rahat and Hazan 2001;Rahat, Hazan, and Katz 2008) through a first comparative attempt.
The structure of this analysis is divided into five parts. In a first step, democratic dissatisfaction as an individual attitude of candidates shall be outlined and discussed. The second part elaborates on the question of the support of dissatisfied candidates by political parties and formulates different hypotheses. The third section outlines the data and cases involved as well as the methodological approach with special regard to the operationalization of democratic dissatisfaction. In the fourth section, I shall conduct multi-level regression analyses (MLA) in order to test the explanatory power of party-related factors in further detail. The results will be discussed in the fifth and final part.

Theoretical outline and hypotheses
This section shall discuss the interplay of attitudes towards democracy, candidacy and political parties. I shall first reflect the nature of democratic dissatisfaction with specific regard to candidates. In the second step, I discuss how the nomination of discontent candidates and the characteristics of political parties are interlinked and formulate hypotheses about this relationship.

Candidates and democratic dissatisfaction
In the following, I shall elaborate on democratic dissatisfaction from a conceptual perspective. Two questions shall be discussed. First, how is the term generally to be understood? Second, what does 'democratic dissatisfaction' address with regard to candidates as a part of the political elite?
In general, political behaviour in democracies depends on an individual's normative conceptions of democracy. Furthermore, behaviour is influenced by the evaluation of the institutions, actors and processes which the individual acts in. Citizens as well as political elites ascribe attributes to the political system in which they operate, that is, the objects they perceive to be related to this system (Almond and Verba 1965). Democratic dissatisfaction in specific describes a certain orientation towards these objects, namely the negative evaluation of the input (the institutions and processes of participation) and output (the performance) of the political regime (see Westle 2010, 316). In this regard, dissatisfaction is to be distinguished from normative orientations towards an alternative regime. While some discontent individuals might indeed prefer, say, authoritarian regimes over liberal democracies, others perhaps idealize direct democracy. Both individuals might be dissatisfied with the status quo, yet due to different normative orientations.
While sharing several similarities with diffuse support by addressing the principles of democracy itself (Easton 1975), dissatisfaction with democracy is best understood as a negative evaluation of several distinct aspects of the democratic status quo (Linde and Ekman 2003, 391-397). It thus neither describes just a vague lack of confidence in the current state of democracy (i.e. Zmerli, Newton, and Montero 2007, 43) nor can one speak of 'dissatisfaction' if only a single detail of the democratic regime is addressed. However, not all forms of democratic dissatisfaction address the same problem(s). In their systematization of the existing literature, Linde and Ekman (2003, 394) show that dissatisfaction relates to several distinct aspects of democracy such as (1) the political community, (2) regime principles, (3) regime performance, (4) regime institutions and (5) political actors. This implies that dissatisfaction varies across individuals and contexts and is highly dependent on the country-specific discourse on democracy that individuals live in (Canache, Mondak, and Seligson 2001, 524). Despite these differences and at a more abstract level, democratic dissatisfaction can be defined as a negative evaluation of several characteristics of the democratic status quo, in which components of the given regime are considered to be malfunctioning or defective. In other words, rather than specific criticism on one aspect, democratic dissatisfaction is a latent attitude. This has implications for its measurement that shall be discussed below. But first, how does dissatisfaction go with the characteristics of political candidates in particular?
In order to approach this question, one should more closely examine the specific role of candidates in party democracies. Let us first consider that candidates in a broader sense belong to the political elite, that is, 'very loosely […] those who in any society rank toward the top of the (presumably closely intercorrelated) dimension of interest, involvement, and influence in politics' (Putnam 1973, 2). In general, the attitudes of political elites are comparable to those of citizens (Putnam 1971). However, the orientations of the former tend to be more distinct and more consistent than those of non-politically engaged citizens (Jennings 1992).
Considering negative attitudes towards democracy in particular, the major difference between elites and citizens is the former's immediate involvement in decision-making institutions and procedures in representative democracies. 1 Political elites are thusin different wayspart of those structures and processes they evaluate, and they are members of the group that is possibly criticized in the context of democratic discontent. Findings from classical studies show that indeed, some members of the political elites are also dissatisfied with the democracy they are involved in (Putnam 1973, 184).
Within the political elites of party democracies, candidates represent a specific subgroup, consisting of a whole range of individuals with various positions within the party hierarchy. Apart from the variance of merits and experience, candidates in party democracies are party members and often activists, at least at some point in their political career (Andeweg and Thomassen 2010). Quite often, they are also part of the leadership (Crotty 1968, 260). Through activism and candidateship, individuals define the membership corpus from which parties choose their representatives in campaigns and, prospectively, in public office (Katz, 2001b).
In general, party activists have in common that they are likely to be 'motivated by expressive concerns and by a sense of collective efficacy' (Whiteley 1995, 227). In other words, they express their attitudes in the context of their engagement and trust their actions to have an impact through parties. Candidacy for political parties is therefore a way of expressing democratic dissatisfaction and an attempt to initiate changes collectively. In other words, these candidates interpret public office as a vehicle to implement their policies (cf. Kitschelt 1989, 407-408). Thus, candidates dissatisfied with democracy are those who express this discontent precisely in the context of their candidacy and want to realize policies to remedy the causes of the perceived problems when in office. Yet at the same time, candidates for whom policy positions play a particularly important role also feel more committed to the 'purity' of these than to their party leadership (Katz 2001b, 289-290). In other words, those who express discontent with the actors, rules and institutions of democracy may be particularly critical of party leadership decisions.
Yet it is parties who choose candidates, not the other way round (Ignazi 1996, 552), 2 and parties attempt to recruit supportive and loyal persons (Cross 2008;Lesschaeve, van Erkel, and Meulewaeter 2018). The question is thus which parties are represented by aspirants who display negative attitudes toward the democratic status quo. I shall elaborate on this in the following.

Hypotheses
Because parties select candidates according to whether they agree with their strategy, there might be positive as well as negative incentives, depending on the strategy the party leadership is actually pursuing. On the one hand, it may be important for parties to avoid the nomination of dissatisfied candidates. On the other, parties may also have an interest in being precisely represented by candidates who share their critical attitudes towards the democratic status quo.
The latter aspect refers to parties as 'expressive' actors. This means that they are particularly interested in certain policy positions that they use to compete in the voter market. In this case, parties do not recruit 'policy renegades' (Cross 2008, 613); they usually select individuals that are in line with their orientation in terms of policies and party ideology and maintain congruence 'between the policy preferences of the party leadership and their candidates' (Lesschaeve, van Erkel, and Meulewaeter 2018, 502). This would imply that the candidates' discontent with democracy is reflected in their party's policy profile. By using anti-establishment rhetoric, political parties articulate (supposed) problems of current democracy and attribute these problems to the political elites themselves (Schedler 1996, 294). Anti-establishment rhetoric negates the binary principle of government and opposition as the core of modern democracies, claiming that all politicians are the same in terms of putting their own interest over the people's (Schedler 1996, 295). Therefore, a party adopting such a position portrays itself not as part of the establishment but as its challenger both in terms of their outsider status in the political system and the major policy issues they articulate (Abedi 2002, 556).
Parties promoting anti-establishment positions have become increasingly successful in recent years. This applies to populist radical left and right actors, but not exclusively. While populist parties make particularly strong use of it, all political actors are capable of adopting anti-establishment rhetoric (Barr, 2009: 31). For instance, among the diverse camp of anti-establishment parties, there are also groups without a populist profile, such as the Finnish Green League (VIHR) (Van Spanje 2011). Several cases show that also mainstream parties can adopt such strategies. For instance, the British Labour Party under Tony Blair took stances against the political elite before they came into power (Mair 2000), a strategy which currently appears to be resembled by the Jeremy Corbyn leadership (Watts and Bale 2018). In Germany, the Bavarian regionalist Christian Social Union (CSU) often portrays itself as a defender of Bavarian interests against the Federal government. Also, former anti-establishment parties can abandon this position. The German Green party is an example for such an evolution. The Norwegian Progress Party (FrP) has shifted from a radical right to a national-conservative party, including less radical attitudes against the political elites.
In a nutshell, while all parties are capable of embracing anti-establishment rhetoric at one point in time (or over time), not all parties do so to the same extent. It is thus rather a gradual than a binary feature which might characterize a party for a certain period and to a certain degree. It also becomes clear that there are large intersections between individual discontent and a party's articulation of an anti-establishment position.
Hypothesis 1a: The stronger a party's anti-establishment profile, the more dissatisfied are its candidates with the democratic status quo.
While this assumption applies to the outcome of the party's nomination process, we also know from research that despite the variety of formal selection procedures (e.g. Hazan 2001, 2010;Rahat and Shapira 2017;Spies and Kaiser 2014), candidates often receive informal support from the party leadership. Even in (more) open selectorates, the party leadership can exert influence, for example, through referrals or through contacts with local elites. Yet in anti-establishment parties, structures of intra-party democracies might often be less developed and leaders have thus more informal power in the nomination process (see Poguntke et al. 2016). On the other hand, elites in mainstream parties are interested in supporting loyal and 'controllable' candidates. Either way, it is to be expected that the party's anti-establishment profile and elite support for candidates are interrelated.
Hypothesis 1b: In parties with a weak anti-establishment profile, candidates dissatisfied with democracy receive less support from the party leadership.
The interests of the party, however, are not just to establish policy congruence. It may be that while candidates share the party's policy goals, they are at the same time 'ideologues', interested in the genuine implementation of their ideas and 'calling for rapid, comprehensive and profound changes in politics and society' (Kitschelt 1989, 407) but not supporting lengthy and less satisfactory compromises. Yet when parties have the possibility to exercise power or even become part of a coalition, this poses a certain strategic incentive (cf. Akkerman and de Lange 2012, 587). For example, a party may be offered to join a governing coalition if it agrees to policies that are unpopular among the members and activists of that party. On the other hand, it could be in the interest of the party leadership to sweep aside some of its party's own policies in order to enforce others in government. But even without direct participation in the government, parties can exercise influence in parliament, for instance, by supporting bills when governing parties are disunited or by putting pressure on minority governments.
A party's parliamentary influence is measured by the impact is has on coalition building: 'The fewer viable coalitions that can be formed without it, the higher a party's threat and bargaining potential'. (Kitschelt 1989, 408) The more powerful a party is, the stronger its influence on government formation and the more it can exert this influence in various ways. In other words, particularly strong parties in this sense are also highly integrated into the system. Strategically acting party leaders want to secure support for decisions of various kinds from their future MPs (Andeweg and Thomassen 2010). Powerful parties are therefore expected to nominate candidates who at least do not actively prevent compromises or even a prospective office-seeking strategy. Against this background, it would be rational for parties in a strong strategic position to nominate candidates who are generally not very dissatisfied with the current state of democracy. 3 Hypothesis 2: The stronger the party's strategic position in terms of parliamentary power, the lower the democratic dissatisfaction of its candidates.
It is obvious that these two aspects are not independent from each other. Parties pursuit neither purely strategic interests nor are they entirely expressive actors. Rather, case studies find a connection between the anti-establishment profile and the strategic position of the party. With increasing likeliness of taking office, former 'outsider parties' try to abandon their role as 'natural' opposition and prepare for joining a coalition or tolerate a minority government. Yet this could lead to internal conflicts with 'purists' who refuse to follow their party leadership when they cooperate with the 'established parties' as seen with the first ÖVP-FPÖ government and the split of the FPÖ (Heinisch 2003). Hence, more powerful anti-establishment parties also seek to ensure that their parliamentary members are loyal to their strategy. When an anti-establishment party pursuits office-oriented goals, it will therefore aim at nominating candidates who are in line with this strategy (Akkerman and de Lange 2012, 587). There are case-specific examples for this. Work on the Danish People's Party (DF) shows that even anti-establishment parties centralise candidate selection methods once they aim at participation in coalition government in order to maintain loyalty of their representatives to their course (Meret 2010;Pedersen and Ringsmose 2004). 4 Considering that this could be a general behaviour, the third hypothesis is as follows: Hypothesis 3: The stronger the power of anti-establishment parties, the higher their candidates' satisfaction with democracy.
Lastly, the very fact that candidates hope political change of their involvement could be associated with disappointment. As studies on citizens' attitudes have shown, satisfaction with democracy often depends on whether the preferred party is among the electoral winners (Blais and Gélineau 2007;Curini, Jou, and Memoli 2011). Since a candidate is involved through personal engagement, the effect of the election outcome on their satisfaction could be particularly strong. This consideration applies to both the party as a whole and the individual. On the one hand, it could therefore be that the candidates whose party has lost votes over the previous election are more dissatisfied than candidates whose party has performed better. On the other, candidates' individual satisfaction with democracy might well be linked to their personal success: as Whiteley (1995, 227) has argued, political engagement comes with the idea of making a change through collective action, and this might well be disappointed if a candidate does not enter parliament.
Hypothesis 4a: Candidates of parties that have won votes over the previous election are more satisfied with democracy than candidates of parties that have lost.
Hypothesis 4b: Elected candidates are more satisfied with democracy than candidates who were not elected.

Data and method
I combine several datasets to test the hypotheses. To measure the candidates' attitudes, I merge the first and the second wave of the Comparative Candidate Survey (CCS 2016(CCS , 2018. For the anti-establishment profile of the parties, I use the Chapel Hill Expert Survey (CHES; Bakker et al. 2015). For the party's vote share, I refer to official election data, mostly collected from the ParlGov dataset (Döring and Manow 2019).

Parties and countries in this study
The empirical focus of this study is on candidates in party democracies in Europe. It aims for a high range and variety of candidates in terms of countries and parties. Focusing on established democracies, the case selection strives for a balanced relationship considering the represented regions. It is intended to include countries from all European regions. Furthermore, a variety of parties in terms of anti-establishment profile and power is targeted. This also means that I focus on parliamentary democracies as we find similar context conditions both internally (regarding the role of parties in the nomination process) and externally (with regard to the role of parties in forming a government). 5 As the study is crossnational, it considers the most recent election per country covered by the CCS and the CHES dataset.
Moreover, data-related implications have to be considered. First, since the data of two datasets are combined (CHES and CCS), only elections, countries and parties that are present in both sets can be considered. Second, in order to maintain a certain degree of variance per party, I decide to include only parties of which a minimum of 10 respondents in the survey as a rule of thumb. Third, it must be ensured that the relevant variables were queried in all countries in the study. These conditions result in the selection of countries and parties shown in Table 1.

Measuring candidates' democratic dissatisfaction
Candidates' dissatisfaction with the democratic status quo is the dependent variable of this study. The theoretical thoughts on the variance and contextuality of democratic dissatisfaction lead to some methodological considerations. I generally assume that discontent with various aspects of the democratic status quo can be described as democratic dissatisfaction, even if it manifests itself differently in different contexts. In this regard, I agree with Linde and Ekman (2003, 406) that democratic dissatisfaction is a concept that should be measured through multiple indicators. As a consequence, robustness checks have to be conducted that shall be described in further detail below.
Each of the relevant items in the CCS addresses specific aspects of discontent with the democratic status quo. In this respect, the data are in principle suitable for measuring democratic dissatisfaction as a complex concept (see Table 2). For better understanding, the CCS items have been renamed for the purpose of this study. DEM1 and DEM2 represent generally negative attitudes toward the democratic regime. More specifically, DEM2 measures the extent to which the people's loss of confidence in democracy is suspected by the individual. A strong expression in this item does not simply induce a lack of confidence in democracy, but focuses on citizens as sovereigns. A negative rating of DEM3 indicates a suspected lack of participation opportunities. DEM4 represents the preference for the majority principle and the evaluation of its implementation in the democratic status quo while DEM5 indicates anti-pluralist orientations. A negative positioning in item DEM6 is considered as a measure of support for the parliamentary principle, whereas anti-party attitudes as a specific form of anti-elite orientations are measured through variable DEM7. While these items address several aspects, as said, one can hardly speak of democratic dissatisfaction if a candidate criticizes only one of these issues. Dissatisfaction with democracy is rather a latent characteristic, formed of disagreement with several issues considering the state of democracy. I solve this through factor analysis: the aim of this method is to retrieve a concept that is defined by different variables. 6 In doing so, I first look at the eigenvalues and can thus answer the question of how the concept of democratic dissatisfaction differs empirically from the criticism of individual aspects of democracy. An eigenvalue >1 indicates that not one single but several variables (= attitudes to democracy) form a common, latent attitude. This would indicate democratic dissatisfaction in a broader sense. Katz (2001a, 63-65) has chosen a similar approach for the analysis of democratic attitudes of political elites in the European Parliament.
I proceed in two steps, the first consisting in a factor analysis with all candidates in the sample. This is necessary to retrieve a common factor, which can then be used as a dependent variable in the empirical analysis. Two results occur. First, the items actually only load on one factor, which justifies the use of a one-dimensional scale (see Table 3). The eigenvalue is relatively high, so that it can be said that in fact a latent attitude consisting of several items is measured.
Second, Table 4 illustrates how the items load on the global factor. Although the factor loadings exceed the usual threshold of 0.3, they do not influence the factor to the same extent. The strongest association is between the factor and the belief that the will of the majority is not represented in the current state of democracy (DEM4). It is hardly surprising that democratic dissatisfaction is mostly associated with the majority principle as the latter addresses a foundational aspect of democracy. Not surprisingly, the value for DEM3, which measures a very similar setting, is only marginally lower. Even if DEM1 and DEM2, each measuring unspecific satisfaction or confidence in democracy, load somewhat less, their factor loadings still indicate a strong correlation between these attitudes and democratic dissatisfaction as a latent attitude. Interestingly enough, DEM5 to DEM7 are still positively associated with democratic dissatisfaction but their factor loadings are somewhat lower.
That is quite plausible as in contrast to the other items, these are relatively specific aspects of democracy (parliament, special interests, and political parties). In this respect, there should also be greater variance among the candidates as far as their position in these items is concerned. However, their factor loadings justify their integration in the solution.
Additionally, the interrelatedness of the variables is tested through Cronbach's alpha (Cronbach 1951). The measurement indicates a value of 0.78, which can be considered as a satisfying outcome.
The second step is as follows. In order to account for differences in the shape of democratic dissatisfaction across countries (see Canache, Mondak, and Seligson 2001;Linde and Ekman 2003), factor analyses are calculated for each country individually (see appendix, tables A1 and A2). While in all countries, the finding of one relatively single strong factor containing several items remains, the items loading on each factor are not the same in every country. In fact, in most countries, six or seven of the items actually load high on the specific factor. In this respect, the regional results are very similar to the global factor analysis. But there are at least slight differences between the countries. Hence, although latent attitudes of democratic dissatisfaction can be demonstrated, they do not encompass the same processes, institutions and actors in all countries. In order to test whether a global empirical analysis is justified nonetheless, the correlation of each country-specific factor with the global factor in each country is calculated (Spearman's rho). The results are shown in Table 5. 7 In all countries, the correlation between the two results of the factor analyses are very high. In other words, the differences between the country-specific factors and the global factor are so small that a common factor analysis is justified despite minimal deviations. Another argument in favour of this approach is that it is not the various forms of democratic dissatisfaction that are at the centre of the analysis but the question of which parties nominate discontented individuals of whatever kind in their country. Nonetheless, country-specific forms of democratic dissatisfaction are a noteworthy finding at this point and call for more case-related research. Also, the differences with regard to the parties within the countries, which cannot be discussed in detail here, are noteworthy (see figure A1 in the appendix).
The variable identified by the factor analysis is recoded into a 0-1 scale for easier interpretation of the final results. Consequently, if the overall value is 0, this indicates complete satisfaction with democracy whereas 1 displays total dissatisfaction.

Independent variables
I now consider the independent variables. A more detailed description is given in the appendix (table A3). For the calculation of the anti-establishment profile of each party, I refer to the corresponding variable of the CHES (Bakker et al. 2015). This approach differs from other, multi-dimensional measurement methods (i.e. Ernst, Engesser, and Esser 2017). Yet this is exactly what is desirable since the hypothesis addresses the general strategy of a party, not specific issue positions. Furthermore, it is a continuous variable that allows to measure the relationship between the 'intensity' of the party's strategy and the dissatisfaction of their candidates with democracy.
The power of a party in parliament is a relatively complex concept. First, it is based on their share of votes. Second, these votes must be put in relation to the party's coalition options. The more powerful the position of the party in the negotiation of policies in parliament and the fewer governments can be formed without it, the better the position of the party (Kitschelt 1989, 407). The party's competitive position is hence measured through the absolute Banzhaf index, which indicates the degree of parliamentary power by weighting each coalition with the number of individual players that have the ability to prevent this coalition. More specifically, the Banzhaf index evaluates the power of a player (such as a party in parliament) by measuring, first, the amount of winning coalitions and, second, how many winning coalitions every player contributes to. Although in the literature, there are various, even more complex forms of measuring power, one major advantage is that the Banzhaf index 'captures better […] the potential benefits of parliamentary seats' (Chiocchetti 2016, 5) and thus provides a suitable proxy for the perception of prospective power by the party. The Banzhaf index is calculated with the online tool ipgenf, provided by Leech and Leech (2009). Manually, if a party was not represented in parliament, the value of the index is set to 0 for the respective election, and if only one party is able to govern, the value is set to 1. The support of the candidates by the respective national party leadership is measured by a corresponding item in the CCS. In this case, the item on the most influential actor 'in deciding on candidacy nomination' is converted into a dummy variable that becomes 1 if the candidate indicates that the party leadership had the greatest influence. In doing so, I follow the argument of Meserve, Palani, and Pemstein (2018), who have shown that formal nomination processes can hardly be measured by survey data. Rather, they show the perception of the candidates and give an indication of informal decisions. Another factor supporting the recoding of the variable is that no theoretical expectations are linked with different nomination processes, so that a possible effect of leadership support can be precisely measured.
I include gender and age as further controls. Moreover, the influence of former experience in office is accounted for through a dummy that indicates past incumbency at the local, regional, national or European level. The empirical analysis thus controls for the moderating influence of involvement in the institutions of representative democracy since members of parliament adapt to 'the responsibilities of this role' through learning and training (Coghill, Lewis, and Steinack 2012, 517;Dickinson 2018, 343) and generally show a certain degree of openness to the institutional experience (Nørgaard and Klemmensen 2018).

Results
The structure of the data is divided into three levels: candidates within parties, nested in countries. I therefore test the extent to which the difference between the individuals is due to differences between the parties or between countries. Thus, an 'empty' model is estimated with parties as level two and the inter-class correlation (ICC) is calculated afterwards.
According to the results reported in Table 6, approx. 25% of the variance in attitudes towards democracy are due to differences between the countries and 56% of the variance can be traced back to parties. Differences between the parties are obviously actually well suited to explain the variance of attitudes, which has already been shown by other studies (e.g. Kennedy, Lyons, and Fitzgerald 2006;Narud and Skare 1999) and can be interpreted as a first empirical indicator for the relevance of the hypotheses. 8 In the following, the variance of democratic dissatisfaction of candidates will be explained through two-level MLA with country dummies in order to test the hypotheses. 9 The party-specific variables, apart from their election results, are centred at the country mean for easier interpretation. As a result, the findings can be interpreted with regard to country-specific differences between the parties. 10 The empirical results are based on hierarchically calculated models in which variable groups were introduced stepwise (Table 7). All models include the control variables; an extensive overview on their effects is given in table A4 (see appendix). Model 1 investigates only the relationship between the anti-establishment rhetoric of the party and democratic dissatisfaction while model 2 exerts the influence of the party's power position through the Banzhaf index. Both variables are tested against each other in model 3. Model 4 tests candidates' support from the leadership only, while model 5 investigates party vote change and the individual electoral success of the candidates. Model 6 combines the variables from the former models. Model 7 accounts for an interaction between salience of anti-establishment rhetoric and Banzhaf power, whereas model 8 tests the interplay of anti-establishment rhetoric and leadership support.
The multi-level analyses provide differentiated results. First, we can see clear effects for party characteristics. Hypothesis 1a is verified: anti-establishment parties nominate candidates who articulate dissatisfaction with the democratic status quo. The coefficient is relatively strong: with each increase in the party's anti-establishment profile by one unit, the average negative attitude towards current democracy of its candidates increases by 0.21 (model 3). In contrast, the multi-level regressions do not confirm hypothesis 1b. Whether a dissatisfied candidate is supported by the party leadership does not depend on the party's anti-establishment profile. In fact, however, the models show that there is no general association between leadership support and individual attitudes towards democracy at all. There can be different reasons for this. For instance, it is possible that the attitudes to democracy of candidates in these parties are so homogeneously distributed that there is no difference between supported and not specifically supported candidates.
The models also demonstrate that parties which are in a more advantageous strategic position nominate candidates more satisfied with democracy, which confirms hypothesis 2. Yet in the overall models, the coefficient is somewhat lower than that of the influence of the party's anti-establishment profile. In other words, with higher Banzhaf power, the candidates' attitudes towards democracy are somewhat less negative, whereas in comparison, in anti-establishment parties, far more dissatisfied individuals are nominated than in mainstream parties. One can therefore conclude that the relationship between the party's antiestablishment profile and the candidates' attitudes is more important than that between the attitudes and the strategic position of the party. In other words, the nomination of candidates with negative attitudes to democracy can be explained above all by policy congruence between candidates and party and to a lesser extent by the party's strategic position.
However, no systematic association can be found for the influence of the party's Banzhaf power on the relationship between its anti-establishment profile and the nomination of dissatisfied candidates. While some anti-establishment parties might indeed Standard errors in parentheses *p < .10, **p < .05, ***p < .01, ****p < .001 nominate rather satisfied candidates with democracy under the condition of an advantageous strategic position, this is not a general behaviour. Checking the marginal effects confirmed this result: the effect of the party's anti-establishment rhetoric on a candidate's level of dissatisfaction does not depend on the effect of endorsement by the party elite. The findings of several case studies (Akkerman and de Lange 2012;Meret 2010;Pedersen and Ringsmose 2004) cannot be confirmed in these models and hypothesis 3 is rejected. Model 5 to model 8 indicate a relationship between a candidate's personal electoral success and their satisfaction with democracy. Overall, candidates who are elected show a significantly lesser degree of discontent than candidates who do not succeed in the election. These results confirm hypothesis 4b. Yet hypothesis 4a is not confirmed since individual dissatisfaction with democracy does not depend on the party's performance. Apart from a small association in model 5, none of the other models detects an effect of changes in the party vote share. 11 Furthermore, the random part of the models accounts for the hierarchical structure of the data by illustrating the extent to which the intercepts of the individual groups deviate from the overall constant. The intercept varies slightly between the parties. The results of the MLA have completed the empirical analysis, and the further implications shall be discussed in the next and final section.

Discussion and concluding remarks
The aim of this study was to provide a first empirical approach to the relationship between democratic dissatisfaction and candidacy for parliament in political parties, whereas the focus was on the parties. Based on the literature, it was assumed that in principle, parties are generally more likely to nominate candidates who are satisfied with democracy, since they are expected to show greater loyalty to the party line. Yet a systematic distinction has been made between two incentives of nomination considering the roles that political parties can take. As expressive actors, parties were expected to nominate candidates who agree with their policy strategy, while strategic incentives tend to result in parties with greater bargaining power not nominating candidates who are dissatisfied with the democratic status quo. However, as parties include both their own policies and their strategic position in their calculations, it was supposed that, based on various case studies, more powerful anti-establishment parties would be represented candidates with a higher level of democratic satisfaction.
The empirical test of the hypotheses yielded differentiated results, which shall be discussed in detail below. Since the aim of this work was to offer a first empirical attempt, the limitations of this study as well as questions arising for further research shall be addressed as well. First, the most relevant factor in the nomination of candidates dissatisfied with democracy is policy congruence between candidates and party. Not only do the results suggest that anti-establishment parties nominate dissatisfied candidates. It is also interesting that no systematic connection between the candidates' attitudes and the support of the party leadership was found. This could mean that the group of candidates within parties is relatively homogeneous. It could also be possible that selection mechanisms of the parties beyond the perceived 'substantive' support lead to the result desired by the party leadership: Lesschaeve, van Erkel, and Meulewaeter (2018) have pointed out that party leaderships have a whole range of instruments at their disposal to establish policy congruence with their candidates.
Second, it has also been shown that a more favourable strategic position of a party is on average accompanied by a rather positive attitude of the candidates towards the democratic status quo. Although the effect is somewhat lower, this suggests that parties do indeed maintain strategic loyalty by selecting candidates that are less radical considering their view on the current state of democracy. Thus, parties are represented by candidates who, as MPs, are also reliable when it comes to rather pragmatic decisions (Andeweg and Thomassen 2010). Third, however, more bargaining power of anti-establishment parties is not necessarily related to the nomination of less dissatisfied candidates. Here, no general relationship was observed. This suggests that additional context-specific factors may play an important role here that could not be included in this comparative analysis.
Lastly, the empirical analysis has demonstrated that candidates' individual satisfaction is related not to collective but to individual electoral success. It is not the party's ability to increase its vote share that affects how candidates evaluate democracy but the question whether or not the individual has entered parliament. Based on the literature, it is plausible to argue that dissatisfaction reflects their disappointment, yet there might be other reasons that deserve further investigation. It could be, for example, that dissatisfied candidates from the outset ran in more hopeless constituencies or at the back of the list. This is not covered by the variable indicating party leadership support, 12 but it clearly hints at an important aspect for further studies on candidates' dissatisfaction and their nomination.
Overall, while the available data allows for the empirical investigation of differences between parties, it only covers the outcome of the process of candidate selection. It does not provide information about the homogeneity of attitudes between candidates and other party members. Despite the heterogeneity even between candidates of the same party, it could be that the most dissatisfied persons in a country gather in the party with the strongest anti-establishment profile. And on the other hand, powerless parties would be particularly attractive to those who are sceptical of established processes and actors (Kitschelt 1989, 406). For future research, two approaches are necessary: the comparison between candidates and other party members as well as time-series analyses. By comparing candidates and non-candidates within parties, more manifest empirical conclusions can be made regarding the selection of personnel. Through time-series data, one can empirically investigate the extent to which parties behave differently under different strategic conditions. It would be interesting to see if anti-establishment parties nominate other candidates in case of an actual increase of their power.
Another aspect could only be given minor attention. It concerns the variance of attitudes towards democracy in different contexts. Beyond the question of which parties nominate the dissatisfied, the aspects related to discontent of candidates are not the same in all countries. This confirms with regard to candidates what other studies have demonstrated for citizens (Canache, Mondak, and Seligson 2001;Linde and Ekman 2003). The different country-specific patterns would be particularly interesting for indepth work in order to understand what exactly candidates for office criticize and how this affects their application for office or their engagement in political parties in the first place (see Whiteley 1995), especially since they possess some experience within the institutions of their country.
Despite some limitations that might as well inspire future research, the results of this study have shown that policy congruence as well as strategic loyalty matter when it comes to how parties are represented in election campaigns. In this respect, this work was able to establish a first link between the beliefs of candidates as part of the political elite and the behaviour of parties in the nomination process. It is worth further examining these aspects empirically.
Notes 1. Of course, this definition is not entirely clear-cut. First, members of political elites are also citizens. Second, citizens in party democracies have the opportunity to participate directly in the political process, at least at the local level. However, these inaccuracies can be ignored here as the analytical distinction is about involvement in the political process at the national level. 2. As said, individuals have joined parties, then became involved within the party and were nominated as candidates, for example, because they already had experience in the apparatus or had proven themselves otherwise. The hypotheses and the later findings of this study are on the one hand subject to the proviso that the selection of candidates cannot be empirically investigated in its entirety. On the other hand, it can be assumed that the connection between candidates' attitudes and party-related factors has a certain direction: The analysis of this study focuses on the end of this process by examining which attitudes towards democracy are related to particular party factors. Nevertheless, this complex relationship shall be reflected in the course of this work. 3. At the same time, powerful and established parties probably also have fewer dissatisfied members in their pool of potential candidates (see Kitschelt 1989, 407). I shall elaborate on this thought in the conclusion. 4. It may well be that within the party, there are still many dissatisfied members who joined the party when it was in a weak strategic position, yet they are not selected as candidates because they may conflict with the leadership's strategic interests. 5. This also means that Switzerland is not included in the sample because of its all-party government, as its parliamentary strength says nothing about the parties' governmental bargaining power. 6. Since the included variables are ordinal, a polychoric factor analysis is calculated. As a robustness check, both the global and all country-specific factor analyses have also been conducted based on Pearson's correlations. The results do not differ in substance. 7. Additionally, Cronbach's alpha is calculated for all individual countries. The country-based results range from 0.87 (Germany) to 0.58 (Norway). This is not surprising with regard to the differences already shown in the results of the factor analyses, and it confirms the theoretically expected variance of democratic dissatisfaction(s) in terms of the aspects of democracy it relates to. 8. See also the country-wise visualization of the intra-party and inter-party variance (figure A1). 9. This procedure was chosen since the group size at level 3 with 11 countries is relatively low (see Maas and Hox 2005). 10. In order to control for possible collinearity, all models were also calculated as simple linear regressions without country dummies and party dummies and subsequently calculated for the variance inflation factor (VIF). No collinearity was detected in any model. In order to rule out specific collinearity through the two party-specific factors (for example, because most anti-establishment parties are outsiders and have little power), the correlation between the anti-establishment profile and the Banzhaf index was also calculated. The result is a value of rho = −0.33, which can be considered as a rather weak correlation. The low correlation is also apparent from the scatter plot in figure A2 (appendix). 11. Considering the controls, age, gender and experience in public office are indeed associated with democratic attitudes (see table A4 in the appendix for details). However, the very small effect sizes of the individual variables indicate that the differences at the individual level are negligible. 12. Corresponding items are only available for some parties and countries in the CCS and therefore could not be included in the investigation.

Disclosure statement
No potential conflict of interest was reported by the author.