On their own turf? The level-specificity of political trust in multilevel political systems

Although the debate on roots of political trust is ongoing, many scholars agree on the relevance of such political factors as political/ideological distance between citizens and incumbents, political responsiveness and perceptions of performance. Less clear is to what extent such trust evaluations are generalized or specific for the three main political arenas in the European multilevel governance system (EU, national, local). With new data from the Netherlands, we demonstrate that trust in local, national and EU governments (1) is largely level-specific (i.e. is based on citizens’ perceptions of incumbents at a particular governance level) (2) and—at all three levels—is explained by level-specific political factors. This shows that political trust is a subjectively rational evaluation and that citizens develop meaningful political orientations vis-à-vis governments at different levels. Therefore, trust spillovers may be less relevant for explaining political trust in multilevel governance than assumed in previous research as executive governments at local, national and EU levels are judged “on their own turf”.


Introduction
Political trust is an important resource of legitimacy. It is a precondition for democratic rule; it creates a link between citizens and their representatives, solves collective action problems and increases law compliance (Bianco 1994;Scholz and Lubell 1998). Low political trust is associated with increasing electoral volatility (Voogd et al. 2019), rise of challenger parties , and, ultimately, with undermining stability of democratic rules itself (Bertsou 2019). No surprise that policymakers in democratic systems are concerned with declining levels of political trust and seek ways to improve them.
However, in multilevel governments, comprising a multitude of territorially and functionally decentralized bodies (Hooghe and Marks 2003), there may not be easy fixes for trust issues. Several studies suggest that citizens have difficulty navigating complexities of the political reality (Hobolt and Tilley 2014) and often apply mental shortcuts in evaluating their governments (Armingeon and Ceka 2014;Hobolt 2007). In this case, trust in a particular government might not necessarily relate to the quality and actions of this government (Dominioni et al. 2020;Torcal and Christmann 2019). Hence, political trust may only be remotely related to the democratic processes and performance at a given level of governance. This would make it difficult to pinpoint specific causes of low political trust and implement measures that would strengthen it.
Recent evidence, however, suggests that the origins of political trust might be more specific then some scholars have argued. For example, new research demonstrates that citizens distinguish in their trust between different branches (executive, legislative and judiciary) at the same tier of government (Schneider 2017;Newton et al. 2018;Schnaudt 2019). Also, citizens seem to differentiate in their trust vertically between the same type of institutions at different governance levels (Schneider 2017). This is especially relevant when considering, e.g. trust in the EU vis-à-vis trust in national or local authorities. In this paper, we ask how this vertical differentiation occurs, and more specifically, what is the role of political factors in this differentiation. The answer will shed more light on political trust as a feedback mechanism on democratic accountability and as a meaningful resource for democratic legitimacy at local, national and EU levels.
Building on the trust-as-evaluation approach (Klingemann 1999;Harteveld et al. 2013;Fisher et al. 2010), we investigate how level-specific perceptions and orientations affect trust in EU, national and local governments. Several political factors have been identified to affect political trust in particular: perceptions of institutional performance, political responsiveness and political/ideological distance to incumbents (Norris 1999;van der Meer and Hakhverdian 2017;Torcal 2014;van Elsas 2015;Hetherington and Rudolph 2018;Holmberg 1999;Keele 2005). 1 However, previous studies theorized political antecedents of trust in general-they did not specify whether and how these political factors would be relevant to political trust at one, some or all EU governance levels. Although political (or 'institutional') explanations of political trust do not exclusively relate to the national level, these theories have been tested mostly at the level of national institutions. No studies have yet systematically tested the relevance of political factors for trust across the multilevel governance structure. Here, we investigate trust in the executive governments (as the most visible branch of government) at the local, national and EU level. We ask: how level-specific are political factors affecting trust in executive governments in a multilevel governance structure (at the local, national and EU levels)?
Our contribution is threefold. First, we study and theorize political trust in the multilevel governance context and across three important layers: local, national and EU. Previous accounts either disregarded the multilevel context altogether or focused only on two levels at the time (either local and national, or national and EU). Meanwhile, the multilevel governance literature emphasizes implications of multilevel governance on citizens' attitudes (Clark 2014;Bache 2008;Muñoz et al. 2011;Muñoz 2017).
Second, we contribute new evidence to the debate on whether political trust at various levels is based on the same logic of trust-as-evaluation (Fisher et al. 2010;Hooghe 2011). So far, scholars hypothesized about the dependency of EU trust on national trust, either through a mechanism of congruence (Anderson 1998;Torcal and Christmann 2019) or compensation (Kritzinger 2003;Sánchez-Cuenca 2000). Others argued that trust spillovers between representative branches of national and EU governance are bidirectional (Dominioni et al. 2020). We complement this work by investigating (1) level-specificity of trust at different governance levels and (2) occurrence of cross-level 'spillovers' (where EU trust, e.g. depends on national government's performance), all this at three (instead of two) levels of governance within the EU.
Third, we offer this new knowledge based on a unique dataset. We merge the Dutch Local Election Studies 2018, the LISS Politics and Values survey and a follow-up survey on the 2019 European Parliament elections. This way we overcome data scarcity highlighted as a limitation in previous studies (Dominioni et al. 2020;Muñoz 2017).
The analysis proceeds as follows: first, we discuss theoretical basis for the concept of trust-as-evaluation and approaches to why and how people (non-)differentiate in their political trust. Then, we hypothesize how selected political factors affect trust in executive governments across the multilevel structure. Finally, we discuss results of our regression analyses and conclude with avenues for further research.

Trust-as-evaluation
According to the trust-as-evaluation approach, political trust is a contextual and relational evaluation (van der Meer and Hakhverdian 2017). Following Hardin (2002), trust in government is primarily determined by citizens' assessments of how their interests are 'encapsulated' in the government's decisions. In light of their political/ideological preferences, citizens consider whether current officeholders govern according to the rules and values they consider important. These trust evaluations are therefore based on a logic of subjective rationalism: these evaluations are rational, because they are based on citizens' political values and goals, and they are subjective, because they are grounded in citizens' perceptions of the contributions incumbents make to the realization of these goals.

Trust generalization/differentiation
Competing approaches offer different views on whether and how citizens differentiate in their political trust. On one side of the spectrum is generalized trust. Trust is generalized to the extent that the results of trust measurement vis-á-vis different trust objects are the same; i.e. have the same means and are highly correlated.
Generalized trust may be a result of two rather different mechanisms. First, generalization might result when one basic trust-attitude determines people's opinions visà-vis more than one trust-object ( Van de Walle and Bouckaert 2003). This for example occurs when political trust is merely an extension of social/interpersonal trust. Then political trust is not related to any particular characteristics of a specific trust-object (e.g. a level or branch of government) but rather to politics and government in general. People thus would form one perception of government as they would not make any conceptual distinctions between various institutions (Hooghe 2011;Mayne and Hakhverdian 2016;Uslaner 2002). In this case there is one logic of trust. People's general (social) trust is spread across all trust objects.
Generalization, however, may also occur when citizens apply different logics for evaluating different objects of political trust. For example, citizens might extrapolate their trust in national institutions to EU institutions (Torcal and Christmann 2019;Muñoz et al. 2011;Dominioni et al. 2020;Sánchez-Cuenca 2000;Anderson 1998;Kritzinger 2003). In this case trust in national political institutions is seen as the result of a more or less subjective rational evaluation of national institutional objects (Dahlberg and Holmberg 2014;Clarke et al. 1993;van der Meer and Hakhverdian 2017). But at the same time the EU trust is considered as a result of a different logic; i.e. the logic of extrapolation.
On the other side of the spectrum is specific trust. In this case, the trust objects (different governments) are considered as more or less distinct political entities perceived and evaluated more or less separately. This occurs when trust is the result of a subjective rational evaluation of a given trust-object, perceived and evaluated independently of other objects (Klingemann 1999;van der Meer 2010;Harteveld et al. 2013;Fisher et al. 2010). Specific trust (just like general trust) may be a result of two mechanisms.
First, it might occur when citizens use different criteria in evaluating these different objects. In this case a subject's trust-attitude not only relates to a particular trustobject at a particular point in time, but also is (partly) based on different criteria used for evaluating these objects. But, and this is the second possibility, even if subjects use the same criteria for evaluation for objects at three levels-i.e. used the same logic of trust-the results of trust evaluations for trust objects at different tiers of government are likely to be different. After all, the subject's trust still depends on independent perceptions of relevant political realities at different tiers that will oftentimes be different (Dahlberg and Holmberg 2014;Clarke et al. 1993).

Specific trust and trust criteria
In this paper we aim to develop a better understanding of the formation of specific trust by studying criteria citizens may use in evaluating trustworthiness of similar political institutions that operate at different levels in the multilevel polity. We focus on one type of institutions (executive governments at local, national and EU level) as different types of institutions are bound to be evaluated following different types of criteria. By comparing determinants of trust in the same type of institution across governance levels, we can explore to what extent citizens trust (a) is the same or a different attitude (is generalized or specific) and (b) follows the same or different logics across governance levels. The same logic of trust would mean that trust at European, national and local levels are based on the same type of trust criteria, i.e. citizens evaluate each of these governments separately according to the same standards with similar type of general expectations in mind (e.g. effectiveness, responsiveness etc.). Different logics of trust would mean that citizens use different set of trust criteria to evaluate trustworthiness of European, national and local governments. For example, political factors important for trust in local government might be of no relevance for people's trust in the EU. In this paper, we consider a number of political factors that in previous research were identified as different manifestations of the trust-as-evaluation logic: perceived political/ideological distance to the government (Holmberg 1999;Keele 2005;Hetherington and Rudolph 2018), evaluations of political responsiveness (Goubin 2020;Torcal 2014) and of governmental/ institutional performance (Pharr and Putnam 2000;Mishler and Rose 1997;Miller and Listhaug 1999;Seyd 2015). In combination these factors cover the three different categories of legitimacy (Schmidt 2013;Scharpf 1999): input-legitimacy, throughput-legitimacy and output-legitimacy. 2 To the extent that trust is based on any such manifestation of the trust-as-evaluation logic, we consider it to be levelspecific (van der Meer and Hakhverdian 2017; van Elsas 2015).

Representation of one's preferred party in government: home-team effect
In our analysis-following Holmberg (1999)-we begin by focusing on the political distance between citizens and incumbents. According to Holmberg's 'home-team' hypothesis, one trusts government more if one voted for the incumbent party. This hypothesis has previously been applied to explain fluctuations in trust towards the national (Holmberg 1999;Listhaug 1995;Holmberg et al. 2017;Keele 2005;Wilkes 2015) or local government (Denters et al. 2014). It shows that political trust-at least to some extent-is partisan. Political trust is higher among voters of the parties in office and lower among citizens whose party fell out of the ruling circle (Hetherington and Rudolph 2018; Keele 2005). This mechanism is based on the idea that 2 There is not a simple one-to-one relationship between the political factors in our analysis and these three forms of legitimacy: e.g. having voted for the incumbent party, on the one hand, may strengthen people's trust because it gives them the idea that their participation in elections did not go unheard (input-legitimacy), and at the same time it boosts people's confidence in the responsiveness of incumbents to their needs (throughput-legitimacy). Likewise, satisfaction with the way democracy works (cf. Dalton 2004, p. 39; also see; Canache et al. 2001) may be associated with people's confidence in the way the democratic system works-as a system designed to increase responsiveness to citizens' demands (throughput-legitimacy). But, at the same time, it might increase their satisfaction with the system's performance (output-legitimacy). familiarity (here in the form of political value congruence between trustor and trustee) breeds trust in situations of incomplete information (Mayer et al. 1995;Schafheitle et al. 2020). This logic might produce different trust scores when citizens evaluate different levels of government: if for example a citizen's preferred party is represented in the local executive body, but not in the national and EU government, then this citizen is more likely to trust the local government, while being more distrusting towards the two other tiers of government.

Evaluations of political responsiveness and performance
Responsiveness is one of the defining characteristics of a democracy, as democratic governments are supposed to respond to the needs and demands of their citizens (Powell 2004;Pitkin 1967;Dahl 1998). The more political officeholders are able to meet citizens' expectations, the higher the political trust (Armingeon and Guthmann 2014;Lipset and Schneider 1983;Torcal 2014). In political science responsiveness is defined in both procedural and policy terms. Pitkin (1967) for example defined responsiveness primarily in procedural terms by focusing on the interactions between citizens and their representatives. Others (e.g. Stimson et al. 1995) defined responsiveness in terms of political performance or the political system's outputs, understood as congruence between citizens' policy preferences and public policies. Irrespective of these two different conceptualisations, we assume that higher responsiveness goes hand in hand with higher trust. As a result we expect differences in evaluations of the democratic system's responsiveness at different levels of governance to produce different trust evaluations for different governments. 3

Hypotheses
On the basis of any of these manifestations of a trust-as-evaluation (focussing on the home-team factor and evaluations of responsiveness and performance) citizens' trust is expected to be level-specific. This implies three hypotheses: H1 Trust in local government is explained by citizens' perceptions and orientations towards the local government (the home-team factor, evaluations of responsiveness and performance) but not explained by perceptions and orientations related to the national or EU governments.
H2 Trust in national government is explained by citizens' perceptions and orientations towards the national government (the home-team factor, evaluations of responsiveness and performance), but not explained by perceptions and orientations related to the local or EU governments.
H3 Trust in EU government is explained by citizens' perceptions and orientations towards the EU government (the home-team factor, evaluations of responsiveness and performance), but not explained by perceptions and orientations related to the local or national governments.

Data
Research indicates that the standard 'trust in government' question is problematic in international contexts as both 'trust' and 'government' might mean different things to respondents in each country (Hooghe 2011;Schneider 2017). Therefore, we focus on trust within one particular political system-the Netherlands.
The Netherlands is an established Western liberal democracy well-integrated into the EU multilevel governance. It participates in key EU integration projects such as the Schengen Agreement and the Economic and Monetary Union. In terms of the institutional autonomy Dutch municipalities-by European standards-are neither very strong nor very weak (Ladner and Keuffer 2018), making the Netherlands a 'middle ground' between the most centralized and decentralized EU countries (Hendriks and Schaap 2012). Also, the Netherlands was one of the very few cases with data suitable for this analysis-including a sufficient number of items on attitudes regarding local, national and EU levels. These factors make the Netherlands a convincing and fairly generalizable case for our research.
This research uses data collected from the same respondents at three points in time. Our respondents participate in the nation-wide LISS panel (Longitudinal Internet Studies for the Social Sciences)-a panel consisting of 7000 individuals, administered by CentERdata (Tilburg University). The panel is based on a true probability sample of households drawn from the population register by Statistics Netherlands.
The primary source for the analysis is one LISS survey-the Dutch Local Election Study 2018 from March to April 2018, collected on the occasion of the municipal elections. All but three variables are taken from this survey, see Online Appendix 1.

Measurements
Political trust is measured through a question: 'How much trust do you have in the following persons/institutions'; the answers are coded on a 4-point scale-from 'no trust at all' to 'a lot of trust'. We consider trust in the executive branch of each governance level, i.e. the College of Mayor and Aldermen for the local level, the National Government for the national level, and the European Commission for the European level.
The home-team effect is researched using a binary variable, where '1' indicates vote for the incumbent government. For the local level, we code it separately for the College of Aldermen and for the Mayor. 4 For the national level, the home-team effect is '1' when one voted in 2017 national elections for VVD, CDA, CU or D66 (parties composing the Dutch government in March 2018). For the EU level, we code '1' if one voted for CDA (Christian democrats), PvdA (socialists), VVD (liberals) or D66 (liberals). At the European level, those parties are members of the EPP, S&D and ALDE groups respectively-three European parties dominating the European Commission, the European Parliament and the European Council in 2018. EPP had a majority in all these institutions, but two Dutch politicians with very high visibility-Frans Timmermans (the Vice President of Juncker Commission) and Mark Rutte (Dutch Prime Minister seated in the European Council) came from the other two parties (PvDA-S&D and VVD-ALDE respectively). This is why we consider voters of these parties as most represented at the European level, thus subjected to the home-team effect.
Satisfaction with the way democracy works is a measure tapping both into responsiveness and performance 5 and is available for all levels of governance. The answers to a question 'How satisfied are you with the way democracy functions in your municipality/the Netherlands/the European Union?' are coded on a 4-point scalefrom 'not at all satisfied' to 'very satisfied'.
Responsiveness is directly captured by a measure asking respondents about the government's responsiveness to citizens' needs. Unfortunately, in our dataset this measure is available only for two levels: Answers are coded on a seven-point scale from 'fully disagree' to 'fully agree'. The responsiveness to local needs score (1-7) is an average of the two answers. We reversed the scale so higher scores imply higher responsiveness. In addition, we used a number of alternative measures to assess people's evaluations of the performance of political institutions (in terms of policies and other outputs) at the three tiers of government. Unfortunately, the measures used for the local level are similar but not identical to the measures for the national and EU governments. For the local level, we used a mean score for answers to two DLES questions: 'How do you generally rate the way in which your municipality addresses the most important local problems?' and 'How do you generally rate the services of your municipality?'. The answers are coded on a ten-point scale (1 'very badly'− 10 'very well').
Evaluations of national and EU performance come from the LISS Politics and Values survey. We take the question 'How satisfied are you with the way in which the following institutions operate in the Netherlands?' for the items 'Dutch government' and 'European Parliament' (as a proxy for the EU performance since a better measurement is not available). The answers are coded on a scale from 0 ('very dissatisfied') to 10 ('very satisfied').

Data analysis
We construct three linear regression models with the same set of predictors but different dependent variables. 6 Model 1 predicts trust in local government, model 2 trust in national government and model 3 trust in the EU. 7 The predictors in all models include level-specific and non-level-specific variables described in the measurements section. Also, all models include the same set of demographic controls (age, gender, education, municipality size, homeownership and income), not displayed in the table for reasons of parsimony. 6 We checked whether the 4-point trust scales lead to substantially different results when treated as ordinal rather than linear. We re-estimated all models with ordinal logit regression. For two out of three trust variables, the model violates the proportional odds assumptions for ordinal logistic regression. Therefore, we also re-estimated these two models with a binary logit regression. The results for the binary models (for trust in local and national governments) and for the ordinal model (for EU trust) are reported in Online Appendix 2. The results presented in the manuscript and the appendix are very similar. The estimation method does not matter for the overall conclusions regarding the level-specificity of trust. 7 Pearson correlations between trust variables are weak to moderate (cf. Anderson 1998;Denters et al. 2014;Muñoz 2017): local/EU r = 0.364; local/national r = 0.514; national/EU r = 0.542-thus indicating some differentiation between these three types of trust. Table 1 summarizes results for each model. 8 It shows that all models significantly predict the dependent variable. Model 1 explains 48.6% of the variance in local trust, model 2-57.7% of the variance in national trust and model 3-60.4% of the variance in EU trust. Tests of data meeting the assumption of collinearity indicated that multicollinearity was not a concern (VIF values ranging from 1.108 to 2.675). As demonstrated by increased R 2 values between the models, the higher position of the government across the multilevel hierarchy, the more citizens' trust is explained by political factors related to representation and evaluations of responsiveness and performance.
Regarding the home-team factor, we find that its relations with the dependent variables are level-specific (i.e. important for trust in the government at the level where a particular home-team operates) in models predicting local and EU trust. For trust in local government, we included two alternative home-team operationalizations: with regard to the political composition of the College and the political colour of the Mayor. The model shows relevance of the Mayor's political affiliation (p < 0.05), but local trust is unrelated to the College's composition. This can be explained by the College's relatively low visibility; not many people know their aldermen. Instead, people tend to know the name and political affiliation of the Mayor. The Mayor is the College's chairman and spokesperson; (s)he is usually affiliated with one of the big national political parties and often engages into the local government management as a long-term career (Andeweg and Irwin 2009). This might explain the significant results for the Mayor's home-team factor.
The home-team factor is also significant in explaining trust in national government. However, it is not entirely clear whether this result is level-specific: for national trust, model 2 displays a significant association of the national home-team (p < 0.001), but also the home-team effect of the Mayor (p < 0.05) and of the EU (p < 0.001). Yet, the effects of the last two might be a methodological artefact. In 2018, 68% municipalities were led by Mayors of the same political colour as national government (De Vereniging van Nederlandse Gemeenten 2018). Also, the national government parties happen to constitute four out of five coded as home-team for the EU level. Hence, the significant results for the local and EU home-team factors in relation to national trust might be a result of the same parties being included in governments across all three tiers of governance at the time of our data collection. 9 The satisfaction with the way democracy works at the given governance level and the relation between responsiveness to the citizens' needs is in most cases statistically significant at p < 0.001 and level-specific. The only exception is the satisfaction with the EU's democracy displaying importance also for national trust. This implies that national government might be also held responsible for the EU democratic 8 The demographic control variables: Age, Gender, Education, Municipality Size, Homeownership and Income are left out from the table summary as they did not contribute to the explanatory power of the models. 9 Two checks lead us to be cautious especially about the estimates for the Mayor-home-team: first, when re-estimating the model using binary logit regression, the association between the Mayor-home-team and trust in the national government (dichotomous) is no longer significant. Second, to see whether the Mayor-home-team effect might be an artefact of the mayor's political affiliation, we checked for an interaction effect between the Mayor-home-team with a dummy variable for municipalities with mayors affiliated with national government parties. When included in the model, the Mayor-home-team is also no longer statistically significant for national trust.

Table 1
Summary of multiple regression analyses for variables predicting trust at different levels of governance *p < 0.05, **p < 0.01, ***p < 0.001 Variable performance. Statistical insignificance of the responsiveness to local needs for local trust came rather as a surprise. 10 Regarding perceptions of performance, all models show statistically significant relations with trust (p < 0.001); these relations are level-specific (i.e. refer to performance evaluations at the expected governance level). Nevertheless, among all political factors considered, the level-specificity is least pronounced for this factor. This suggests that governmental performance is apparently difficult to disentangle: showing relevance also for trust in other governments at other levels. For example, perceived performance of the local government is associated the most with local trust (p < 0.001) but is also related to national trust (p < 0.01) and EU trust (p < 0.05). One possible explanation is that many policies decided at the national and EU level have to be implemented at the local level, thus affecting local performance. This is the case especially in the Netherlands, where the co-governance principle entails implementation of national policies as a joint task for all governance levels, with a special emphasis on the local government (see, e.g. Denters et al. 2005;Hendriks and Schaap 2010). For much the same reasons, national performance may also be related to local trust (p < 0.01) and local performance to EU, although the level of significance is lower (p < 0.05 instead of p < 0.01). This possibly reflects that just like the Dutch national government, the EU applies the subsidiarity principle in implementing many policies. Hence, municipalities are important for EU policy effectiveness; as has been recognized through the Pact of Amsterdam (EU Ministers Responsible for Urban Matters 2016). Our results reflect these interdependencies in policy performance in the European multilevel polity.
To return to our hypotheses, the data analysis shows that most factors that are (significantly) associated with political trust pertain to the particular level of governance. If we accept only factors that are significant at p < 0.001, then trust in local and EU governments is level-specific (hypothesis 1 and 3 confirmed), i.e. associated with political factors specifically related to this governance level. Following this logic, trust in national government is somewhat less level-specific. Although all national-level trust criteria matter for national trust, there are also additional local and EU factors that are significantly related to national political trust; which strictly spoken implies that hypothesis 2 is rejected. However, this strict interpretation should not mask the more general pattern: as hypothesized on the basis of the trust-as-evaluation logic, trust in executive governments in multilevel governance is largely level-specific. At all three governance levels, citizens' trust evaluations of the government reflect perceptions of its responsiveness and performance; the same pattern for the home-team factor is-probably for methodological reasons-less pronounced. 10 The effect of local politicians' responsiveness might be already captured by other two factors included in the model (satisfaction with the way local democracy functions; and perceived performance of the local government). When removing any of these from the model, responsiveness of local politicians shows statistical significance at p< 0.01.

Conclusions
In this research we demonstrated that political trust of Dutch citizens in executive governments at local, national and EU levels is largely level-specific. Trust refers to perceptions and evaluations of officeholders at a given governance level and spillovers from other levels of governance are limited. Observed differences in political trust are first and foremost related to a particular governance level and attributed to characteristics of particular incumbents. Furthermore, contrary to previous accounts suggesting that trust in different levels of governance is explained by different mechanisms (Harteveld et al. 2013), we show that citizens' trust in executive governments at local, national and EU levels can be interpreted by the same logic of subjective rationalism. In this logic, trust is an internally consistent evaluation of a particular trust-object, independent of evaluations of other objects. Our study shows that citizens seem to form to a large extent separate trust evaluations of the local, national and European executives respectively. This supports the argument that political trust is level-specific: each government within the multilevel structure is judged on their own turf, with consideration to their own merits and independently of attitudes developed for other governance levels.
We also found that citizens seem to apply similar political criteria to evaluate trustworthiness of the political executives at the European, national and local levels: political/ideological distance to governmental incumbents (the home-team factor), political responsiveness and institutional performance. These three criteria relate to the input-, throughput-and output-legitimacy of governmental incumbents and thus make political trust an important indicator of perceived quality of democratic governance and democratic legitimacy at a given level of governance. This is an important finding as it points to specific areas affecting trust in government-and helps to detect potential causes for troubling trust declines at the particular level of governance.
Our research demonstrates the primary importance of level-specific factors for trust in government; however, some minor cross-level influences or 'spillovers' could also be observed. Governments are embedded in a multilevel political reality and their interconnectedness is bound to be reflected in citizens' perceptions (Muñoz 2017). In particular, we found that local performance evaluations matter also for national and EU trust; and national performance evaluations for trust in local government. We attribute that on the one hand to the key role the local government plays in implementing policies from other governance levels, especially in the Netherlands. On the other hand, whereas the home-team and responsiveness factors are level-specific, the performance factors might be more difficult to be so; multilevel policymaking and policy implementation is often a joint effort of institutions across the multitier structure.
One surprising outcome was that from the three types of trust, local and EU trust shows to be the most level-specific. Evidence from electoral research would suggest that the 'second-order' local and EU elections are dominated by cues from national politics and thus not level-specific (Marien et al. 2015;Hobolt and Wittrock 2011;Heath et al. 1999). 11 Similarly, political trust research presents EU trust as dependent on national trust through congruence or compensation mechanisms (Anderson 1998;Torcal and Christmann 2019;Sánchez-Cuenca 2000;Kritzinger 2003;Muñoz et al. 2011). Our study points to an opposite mechanism, where national government is 'sandwiched' between the local and EU governments, taking blame or praise for what is happening at other levels. Despite the conventional wisdom of national level being the most salient level of governance (Muñoz et al. 2011;Torcal and Christmann 2019;Rodden and Wibbels 2011;Lefevere and Van Aelst 2014) -the 'hollowing out of the nation state' (Jessop 2013) might be leading to the reverse tendency, where other levels impact politics at the national level and blur its clarity of responsibility (Anderson 2006).
This study could be further improved by introducing a direct measurement for EU responsiveness, unfortunately not included in our data. This would improve comparability of EU trust with other types of trust. Although 'satisfaction with the way democracy functions' is conceptualized to reflect, i.e. responsiveness of the institutions (Dalton 1999;Morlino and Quaranta 2014), more precise measurement could shed more light on impact of the EU responsiveness on citizens' trust. It is worthexploring as perceived responsiveness is seen as closely related to trust in public institutions (Goubin 2020).
Furthermore, regarding EU trust, alternative home-team conceptualizations might be relevant. Although our conceptualization displayed a statistical significance, it has to be recognized that EU governance and politics are organized differently than those at local or national levels. Party preferences might impact EU trust differently; instead of depending on power gained within EU structures, it might be about their pro-or anti-European sentiment. 12 As shown previously, voters of the Eurosceptic parties tend to distrust the EU more than others (Hooghe and Marks 2007;Sánchez-Cuenca 2000;Rohrschneider 2002).
Finally, although our theoretically driven causality assumptions between evaluations and trust are reflected in the survey construction, 13 it still remains relevant to ask: are performance evaluations fuelling trust-attitudes or are the perceptions of performance conditioned by one's inherent propensity to be trustful? Despite the panel structure of the wider LISS panel, the DLES is essentially a cross-sectional survey with most variables measured at one moment in time. Hence, we must be careful with a causal interpretation of the findings. This chicken-and-egg problem is well-familiar for the political trust research (see, e.g. Mishler and Rose 2001;Newton et al. 2018;Zmerli and van der Meer 2017) and cannot possibly be 11 Although this view is increasingly contested and/or nuanced by contributions arguing that even second-order elections are somewhat 'arena-specific' (or 'level-specific'); see, e.g. Cutler (2008), Liñeira (2016). 12 As a robustness check, we controlled for Eurosceptic attitudes; whereas they are a powerful predictor for EU trust (at p < 0.001), including them in the model did not change our results for the statistical significance of the home-team effect (at p < 0.05). 13 The survey asks for relevant evaluations before the trust question. The only exception being the evaluation of local performance, which comes few items after trust questions. once-and-for-all solved here with applied data and methods. 14 With due caution, however, our results not only confirm that political factors are closely related to political trust (Torcal and Montero 2006;van Elsas 2015;Mishler and Rose 2001;Newton 2006) but also demonstrate that patterns of association between these political factors and political trust are level-specific.