Democratic legitimacy rests on citizens' willingness to accept political outcomes emerging from the democratic process (Anderson et al. Reference Anderson2005; Dahl Reference Dahl1989). Political parties' parliamentary representation is one such outcome brought about by regular elections. However, research on negative partisanship and affective polarization testifies that citizens' acceptance of individual parliamentary parties is far from self-evident, with potentially important democratic consequences such as public upheaval (for example, Abramowitz and Webster Reference Abramowitz and Webster2016; Druckman et al. Reference Druckman2021; Haggard and Kaufman Reference Haggard and Kaufman2021; Iyengar et al. Reference Iyengar2019; Kydd Reference Kydd2021; Norris Reference Norris2021). This article offers a novel perspective on assessments of political parties by considering their perceived legitimacy as institutions in the democratic game.
Existing literature has made great progress in understanding the antecedents of affective polarization and negative partisanship, pointing to social identities in the case of polarization (for example, Druckman and Levendusky Reference Druckman and Levendusky2019; Iyengar, Sood, and Lelkes Reference Iyengar, Sood and Lelkes2012; Iyengar, Shando, and Lelkes Reference Iyengar2019) and party ideologies in the case of negative partisanship (for example, Meléndez and Kaltwasser 2019; Kaltwasser, Vehrkamp, and Wratil Reference Kaltwasser, Vehrkamp and Wratil2019). Other and related new research provides the first evidence of how parties are seen in terms of their legitimacy. It shows that party legitimacy perceptions are affected by political and news communication – varying by party – and have important consequences for citizens' voting behaviour (for example, Bos and van der Brug Reference Bos and van der Brug2010; van Spanje and Azrout Reference Van Spanje and Azrout2019; van Spanje and Azrout Reference Van Spanje and Azrout2021). These first results show that party legitimacy perceptions are an important phenomenon that may also be shaped by party-level drivers, making some parties systematically seem more/less legitimate.
In this article, I am drawing on insights from party politics and public opinion research to develop a theoretical framework that explains how parties’ characteristics shape their perceived legitimacy. Based on previous work suggesting that party legitimacy means being ‘socially acceptable’, ‘part of the national tradition’ (Eatwell Reference Eatwell, Peter and Leonard2003, 68), or simply ‘normal’ (van der Brug et al. 2005, 546), I proposed that perceived party legitimacy is about granting a party the same rights and powers as other parties, including the right to engage with its policy proposals and to accept and comply with its policy decisions. Theoretically, I argue that citizens use party characteristics as informational cues to form their party legitimacy perceptions. Drawing on existing literature, I propose that a party's governing experience, mature age, ideological moderation, and democratic behaviour signal to citizens that it is a more legitimate institutional player. They all indicate to citizens that a party can and should be treated like any other because it does not threaten democracy; it is unlikely to erode democratic processes and norms once it has access to power (see Levitsky and Ziblatt Reference Levitsky and Ziblatt2018).
I test the effects of party-level information on citizens' party legitimacy assessments in two studies. First, using original survey data from seven West European countries, I cross-nationally test how citizens' perceptions of party legitimacy are systematically related to party-level information in the real world. The methodological contribution of the survey is that it operationalizes party legitimacy perceptions to provide assessments on fifty-nine parliamentary parties. The regression results support the hypothesis: parties' governing experience, mature age, moderate ideology, and pro-democratic behaviour are all party-level signals that make citizens view them systematically more legitimate. These findings demonstrate a pattern of high levels of acceptance of democratically elected parties as decision-makers in the democratic game and that citizens differentiate between parties based on the parties' characteristics.
I supplement this finding with the results from a pre-registered, large-scale survey experiment that provides a stricter test of the informational effects of party characteristics. It focuses on two pieces of party-level information often conflated in reality: a party's ideological moderation and its democratic behaviour (see Bourne and Bértoa Reference Bourne and Bértoa2017, 244; Engler et al. Reference Engler2022). The experiment separates them. The methodological contribution of the experiment is that it operationalizes Linz (Reference Linz1978) (see also Levitsky and Ziblatt Reference Levitsky and Ziblatt2018) ‘litmus test’ to manipulate the party's democratic behaviour. The results show that ideological moderation and democratic behaviour have similar informational effects on party legitimacy perceptions. However, only parties described as belonging to the political right – not the left – are seen as less legitimate compared to parties without ideology. Finally, I provide initial support for the theorized mechanism: citizens perceive some parties as less legitimate because they view them as threatening democracy.
In short, across two studies using different designs and data, I find consistent evidence that citizens used party-level characteristics as informational cues to assess parties' legitimacy as institutional players in the democratic game. The results imply that parties can gain and lose perceived legitimacy depending on their governing experience, their age, how ideologically right-leaning they are, and how democratically they act. However, they also imply that the institutional and legitimizing power of democratic elections is not enough to guarantee citizens' acceptance of the outcome. In combination with existing work on the importance of party legitimacy for voters' party preferences and party competition (Bos and van der Brug Reference Bos and van der Brug2010; De Vries and Hobolt Reference De Vries and Hobolt2020), party legitimacy thus appears to be another valence issue.
In the following, I first review existing work before developing my theoretical framework about the informational cues of party characteristics for citizens' legitimacy perceptions. In section four, I explain the research design before I present the cross-national study, followed by the experimental study. In the last section, I summarize the findings and their implications.
Existing Research
Political parties' parliamentary representation is one of the political outcomes of holding democratic elections. Theories of democratic legitimacy hold that citizens' acceptance and compliance with the outcomes of a democratic process are essential for the health and survival of democracies (Buchanan and Keohane Reference Buchanan and Keohane2006; Przeworski Reference Przeworski1991). Research on the evaluations of citizens' reactions to political outcomes has focused much on the performance of the citizens' own party (the ‘in-party’); that is, its victory or defeat (see, for example, Anderson et al. Reference Anderson2005). However, recent studies demonstrated that citizens also have distinct attitudes towards the other party (the ‘out-party’).
Much new work shows that negative partisanship, defined as ‘aligning against one party rather than with another’ (Abramowitz and Webster Reference Abramowitz and Webster2016, 119), has become a widespread phenomenon across modern democracies. For example, data from seventeen European countries document that, in total, 32 per cent of respondents hold a negative party identification (Mayer Reference Mayer2017; Spoon and Kanthak Reference Spoon and Kanthak2019). Such negative attitudes are not inconsequential either. They affect citizens' voting behaviour, participation in non-voting political activities, and satisfaction with democracy (Iyengar and Krupenkin Reference Iyengar and Krupenkin2018; Ridge Reference Ridge2020). For example, Caruana, McGregor, and Stephenson (Reference Caruana, McGregor and Stephenson2015, 783) show that negative partisanship is associated with a ‘6-point increase in participating in a protest’.
A second and related literature concerns affective polarization. As a ‘natural offshoot’ (Iyengar et al. Reference Iyengar2019, 130) of a social identity perspective on partisanship, affective polarization is prevalent in the United States and Europe, is growing over time, and yields important negative effects on citizens' democratic norms, social relations, and their policy preferences (Druckman et al. Reference Druckman2021; Iyengar, Sood, and Lelkes Reference Iyengar, Sood and Lelkes2012; Iyengar et al. Reference Iyengar2019; Kingzette et al. Reference Kingzette2021; Reiljan Reference Reiljan2020).
The collective work in both literatures is important in three respects. First, it highlights the importance of understanding the correlates of strong aversions against individual but democratically elected parties. Second, by focusing on two-party systems or on one particular party, little is known about multi-party systems or the evaluations of several parties (but, see for example, Mayer Reference Mayer2017; Meléndez and Kaltwasser Reference Meléndez and Kaltwasser2021). Third, while it shows important results using broad affective or behavioural measures such as the feeling thermometer or voting intentions, current work offers little to evaluate a more minimal standard: ‘To what extent are citizens willing to accept a democratically elected party as an institutional decision-maker?’
A new strand of literature offers the first insights into this question by studying citizens' perceptions of party legitimacy. Bos and van der Brug (Reference Bos and van der Brug2010, 793) (see also Bos, van der Brug, and de Vreese Reference Bos, van der Brug and de Vreese2011; Bos, van der Brug, and de Vreese Reference Bos, van der Brug and de Vreese2013) for example, measures citizens' perceptions of Dutch parliamentary party leaders ‘willingness to ignore important democratic rules’ and their potential to ‘pose a threat to democracy’ once in power. The authors demonstrate that such assessments vary by party (between 2.9 and 3.9 on a 6-point scale) and have important effects on voting preferences. A second set of studies, also from the Dutch context, goes even further and poses additional questions about a specific far-right party's ‘right to exert power’ and its acceptance of ‘social norms in our society’ (Jacobs and van Spanje Reference Jacobs and van Spanje2021; van Spanje and Azrout Reference Van Spanje and Azrout2019; van Spanje and Azrout Reference Van Spanje and Azrout2021). The results show that legitimacy perceptions are an important mediator linking exposure to stigmatization and the prosecution of a party member for hate speech. In sum, this work shows that citizens have varying attitudes towards the democraticness of parties, with important electoral consequences. These first results, including the potential electoral power of party legitimacy perceptions, raise important questions about which party-level characteristics systematically make some parties seem more/less legitimate.
The Informational Role of Party Characteristics for Citizens' Party Legitimacy Perceptions
Building on and extending the existing literature, I combine insights from party politics and public opinion research to develop a theoretical framework about how party characteristics inform citizens' party legitimacy perceptions. Previous work suggests that political parties are legitimate in the eyes of the public when they are considered ‘socially acceptable’, ‘part of the national tradition’ (Eatwell Reference Eatwell, Peter and Leonard2003, 68), or simply ‘normal’ (Bos and van der Brug Reference Bos and van der Brug2010; van der Brug et al., 2005). While these studies do not further explain what these terms mean, they all suggest that a party seen as legitimate can or should be treated like any other, presumably because it does not threaten democracy and its institutions. Such an understanding also indicates the potential importance of party characteristics to form party legitimacy perceptions. This forms the basis for my theoretical framework.
More specifically, I argue that a party's governing experience, mature age, ideological moderation, and its democratic behaviour all signal to citizens to assess a party as a more legitimate institutional player. They all indicate to citizens that a party can and should be treated like any other because it does not threaten democracy. It is less likely to erode important democratic processes and norms once it has access to power (see Levitsky and Ziblatt Reference Levitsky and Ziblatt2018). I suggest that the party-level factors provide strong informational signals to citizens that the party is less likely to erode democratic processes and norms compared to its counterpart because it has either a positive track record (governing experience and mature age) or currently displays characteristics (ideological moderation and democratic behaviour) that make it seem less likely to erode democratic processes and norms.
The general connection between governing experience and perceived legitimacy seems clear. Other things remaining the same, it is plausible that a party with experience in office is regarded as more legitimate than one without because it has enjoyed ‘popular legitimacy’, often defined as majority support (for example, Mair Reference Mair1995; Webb Reference Webb2000).Footnote 1 A governing party has usually won the most votes in the previous election; thus, its position in the political system reflects that it is socially accepted by many. Such high levels of acceptance may also signal to citizens that such a party is seen as less likely to erode important democratic processes and norms once it has access to power. Therefore, a party's experience in government should work as an informational cue and translate into positive party legitimacy perceptions. While the specific evidence on governing experience and legitimacy is limited, the general association between a governing role and quality is well established in studies of incumbency advantage (Cox and Katz Reference Cox and Katz1996; Eggers Reference Eggers2017). At the same time, not all parties with governing experience may signal a non-threat to democracy, as recent developments in Italy, Hungary, or Poland document. I therefore expect ceteris paribus:Footnote 2
H1: Citizens consider parties with governing experience more legitimate than those without.
In addition to governing experience, broader experience is acquired by operating within politics and parliament. The real novices to the political system, whose ‘normality’ or legitimacy might be most in doubt, are those parties that have only just appeared on the scene. Older and newer parties are known to differ in important ways, including citizens' evaluations (see, for example, Hobolt and Tilley Reference Hobolt, Tilley and Laffan2018; Kölln Reference Kölln2016). A party that has been around for a long time signals that it is relatively successful in mobilizing voters and that it is ‘part of the national tradition’ (Eatwell Reference Eatwell, Peter and Leonard2003, 68). Similarly, a party with a long track record in the political system may indicate to its citizens that it is ‘normal’ and less likely to erode democratic processes and norms; that is, it does not threaten democracy. This may drive voters' positive impressions of party legitimacy. This leads to the second expectation:
H2: Citizens consider old parties more legitimate than new parties.
A party's broader ideology is another important informational cue for citizens’ legitimacy perceptions. Existing literature regularly shows that political parties with extreme ideological positions are met with electoral scepticism (Carter Reference Carter2005; Golder Reference Golder2016; Johns and Kölln Reference Johns and Kölln2020; Reiljan Reference Reiljan2020). For these citizens, ideological extremity may not be part of what they consider ‘normal’ or ‘socially acceptable’, possibly because parties with such ideologies deviate too much from what the median voter wants (see Downs Reference Downs1957).
Extreme ideologies may suggest to citizens that such parties endorse values that can threaten democracy because they are likely to erode democratic processes and norms. The extreme left ideologies because proponents ‘usually denounce all compromise with the “bourgeois”’ (March Reference March2008, 126) and emphasize extra-parliamentary action. The extreme right ideologies because their political positions are typically underpinned by nationalist and xenophobic attitudes (Mudde Reference Mudde2002; Mudde Reference Mudde2007). Indeed, Meléndez and Kaltwasser (Reference Meléndez and Kaltwasser2021, 7; see also Reiljan Reference Reiljan2020) show that more than half of the respondents in ten Western European countries held negative attitudes towards populist radical-right parties. Parties belonging to more mainstream ideologies (Christian Democratic/Conservative, Social Democratic, or Liberal) were all viewed more positively. Bos and van der Brug (Reference Bos and van der Brug2010, 787) provide similar results from the Dutch case, using their more specific items on a party's perceived level of democracy. At the same time, a party's ideological position is unlikely to matter equally to everyone, irrespective of their ideological leanings. But, in general, existing literature suggests the following hypothesis:
H3: Citizens consider ideologically moderate parties more legitimate than ideologically extreme parties.
In addition to the broader ideology of a party, the party's democratic behaviour may also act as important information for perceptions of party legitimacy. The literature provides ample evidence that politicians and parties acting unlawfully or bending the rules receive a worse evaluation from their citizens (see, for example, Anduiza, Gallego, and Muñoz Reference Anduiza, Gallego and Muñoz2013; Ares and Hernández Reference Ares and Hernández2017). Using a survey experiment of Spanish respondents, Breitenstein (Reference Breitenstein2019) shows that another party's accusation of corruption already reduces the probability of voting for a candidate by 22 per cent. The potential mechanism behind this is citizens' concern for respect and fair treatment as central democratic practices (see Knight Reference Knight1998; Tyler Reference Tyler1997). Therefore, I anticipate that when a party is seen as bending or breaking democratic rules or practices, citizens consider it less legitimate because they see it as more likely to erode democratic processes and norms and thus pose a potential threat to democracy. Conversely, when a party is seen as endorsing democratic rules, citizens likely deem it as more legitimate because they see it as ‘normal’. This forms the fourth hypothesis:
H4: Citizens consider parties endorsing democratic rules more legitimate than parties bending or breaking democratic rules.
Research Design
I conducted two studies to test how different party-level information shape citizens' assessments of party legitimacy. In both, I only consider parliamentary parties because they are institutional players that citizens might not accept, but having good reason to do so.Footnote 3
Study 1 is a cross-national observational analysis using original survey data collected online through Gallup International between November and December 2021. The survey was fielded across seven West European countries (Denmark, France, Germany, The Netherlands, Portugal, Spain, and the UK) with samples of 1,000 respondents each. The samples were intended to be representative of the electorates in these countries. The country selection allows for variance across several potentially relevant contexts (electoral systems, party systems, political culture, political history, etc.), which allows generalizable inferences to be made about the effects of party-level information on assessments of party legitimacy for West European countries.
However, cross-national data are insufficient to truly test the informational effects of party characteristics because they cannot isolate the effects of different party-level information compared to no information about these factors. Additionally, some party characteristics – ideology and democratic behaviour – are conflated in reality (Bourne and Bértoa Reference Bourne and Bértoa2017, 244; Engler et al. Reference Engler2022). Therefore, I supplement the observational analysis with a survey experiment that tested highly similar hypotheses to H3 and H4 to test the informational effects of party characteristics. Derived from the same theory, they focus on comparing different party-level information to no information about these factors. The survey was fielded online in Denmark in December 2021 with the help of YouGov and 4,423 respondents, intended to be representative of the Danish adult population.
Study 2 offers a more rigorous identification of the independent effects of information on ideology and democratic behaviour compared to no information, thus increasing the internal validity of the results. In the following, I begin by describing the observational analysis before presenting the results from the experimental study (for ethical considerations, see SI Part A).
Cross-National Analysis (Study 1)
Study 1 is based on cross-national data with respondent-party dyads as the units of observation. Each respondent enters the dataset once for each parliamentary party the survey covers. Included are all parliamentary parties, but whenever the number exceeded ten, the parties that received at least three parliamentary seats in the last election were included (five seats in the Netherlands) to avoid respondent fatigue. The order of parties was randomized across the respondents and within the survey.Footnote 4 The survey measured citizens' legitimacy perceptions of fifty-nine political parties in seven countries.
The precise Western European countries were selected to cover as much variation as possible in contextual factors (electoral systems, party systems, political culture, political history, etc.) relevant to this study. For example, the 2018 European Social Survey data shows that the mean level of trust in parties is generally low but varies across these countries. Accordingly, they can be classified into relatively low (Spain = 2.5; Portugal = 2.9), relatively medium (France = 3.1; United Kingdom = 3.6; Germany = 4), and relatively high (Denmark = 5.3; The Netherlands = 5.4) trusting countries.Footnote 5 Although ‘trust in political parties’ measures a general attitude towards all parties in the system, trust is strongly related to legitimacy (Kaina Reference Kaina2008). In addition, parties in these countries collectively display relevant variation in the independent variables needed to test the hypotheses.
Dependent variable
I follow existing literature on legitimacy and conceptualize it as the ‘right to rule’ with a corresponding obligation to obey (Peter Reference Peter and Zalta2017) or, in other words, the ‘acknowledgement of rights and powers’ (Kelman Reference Kelman, Jost and Major2001, 58; see also Weber Reference Weber1972).Footnote 6 Applied to political parties, this means that party legitimacy perceptions are the citizens' acknowledgement that a party has rights and powers as a parliamentary decision-maker. These are justified based on the outcome of elections. Acknowledging a party's rights and powers means treating a party like any other parliamentary decision-maker or treating it as ‘normal’.
I translate this conceptualization into a newly developed survey instrument using a multiple-item index to increase validity and reliability.Footnote 7 Existing work from the Netherlands operationalizes party legitimacy perceptions by asking about perceptions of following democratic rules and norms or perceptions of posing a threat to democracy (see, for example, Bos and van der Brug Reference Bos and van der Brug2010; van Spanje and Azrout Reference Van Spanje and Azrout2019). While these are relevant survey items, research shows that their underlying concepts are the causes of legitimacy perceptions (see, for example, Gilley Reference Gilley2009; Seligson Reference Seligson2002; Tyler Reference Tyler2006; Zelditch Reference Zelditch, Jost and Major2001). Since operationalizations through causal proxies may underestimate the strength of the attitude, not least by introducing temporality (Goertz Reference Goertz2006), I measure perceived party legitimacy directly.
The point of departure for developing the specific items is that parliamentary parties enjoy certain rights and powers (see, for example, Bowler Reference Bowler, Dalton and Wattenberg2000; Proksch and Slapin Reference Proksch and Slapin2015). They often have rights and powers related to organizing their parliamentary activities, such as the right to have offices, the right to be represented in committees, and the right to state subsidies. But parliamentary parties also enjoy important political rights or powers, such as the right to pose questions, speak, and propose and legislate policy. I propose that party legitimacy assessments manifest themselves by acknowledging and accepting a party's parliamentary rights and powers, thus treating it like any other. The selected survey items aim at capturing this, striking a balance between specificity and comprehensiveness. The final survey battery contains three items, measured on a 7-point bipolar scale.Footnote 8
Please state the extent to which you agree or disagree with the following statements:
(a) When [PARTY] is elected to parliament, it needs to be granted the same rights as other parties, even if I do not like it.
(b) When [PARTY] is elected to parliament, their policy proposals need to be discussed like others, even if I do not like them.
(c) When [PARTY] is elected to parliament, their policy decisions need to be accepted and complied with like others, even if I do not like them.
The battery has good measurement properties and forms a valid and reliable index. I also verified that party legitimacy perceptions were related to but different from existing constructs measuring general assessments of political parties used in work on affective polarization or negative partisanship (for example, party sympathy and probability of voting for a party). Part B in SI provides more information. The final additive index is scaled to range between 0 and 1.
Figure 1 summarizes the index across parties and countries (see Table B1 in SI). It bears some good news because – despite rising levels of negative partisanship and affective polarization in Europe – party legitimacy perceptions are generally high (mean = 0.79, sd = 0.26).Footnote 9 Nonetheless, the figure also shows that none of the parties lives up to the normatively desirable standard of full legitimacy because no party has an average score of 1 or even close to it. Indeed, some parties are far from it, such as the German far-right party, the AfD. With 0.59, it has the lowest score in the sample. However, even the Portuguese social-democratic party PSD, with the highest legitimacy score of 0.87 in the sample, falls short of full legitimacy. While the parties' averages are relatively similar (except for the AfD), the standard deviations are substantial (between 0.19 and 0.36; see also Table B1 in SI).Footnote 10 It suggests that party-level differences are more contained compared to individual-level differences. Looking at the country level, party legitimacy perceptions are lowest in Spain (= 0.74) and highest in Portugal (= 0.85), underscoring that party legitimacy differs from trust because both countries display relatively low levels of trust in political parties. These descriptive results offer initial insights into how different citizens' party legitimacy perceptions are within and across countries.
Figure 1. Party legitimacy perceptions across parties and countries: weighted means and standard deviations.
Note: The figures are based on Table B1 in SI.
Independent Variables
To test the argument, I require additional information about parties' governing experience, age, ideological moderation, and democratic behaviour. I measure party's governing experience by coding parties that had been in government at least once (held cabinet seats) as 1, otherwise 0. Information about a party's age was derived from its official foundation year. Given the sample's relatively limited number of parties, I subsequently only distinguish old and new parties, using 2010 as a cut-off point. While any cut-off point is arbitrary, parties founded after 2010 and still in parliament only experienced about two complete electoral cycles; anything beyond that is often considered long term (see Horn Reference Horn2021, 1,504). A party's ideological moderation was measured using a party's left-right position obtained from the Chapel Hill Expert Survey data (Jolly et al., Reference Jolly2022), folding it around the mid-point of five.Footnote 11 I reversed the scale to measure moderation on a scale from 0 to 5.
My measure of a party's democratic behaviour was inspired by the Dutch studies – already mentioned – and was based on two survey items. The first item asks the respondents to place each party on a 7-point scale with the labelled end points of ‘challenges current democratic practices’ and ‘endorses current democratic practices’. The second item has the labelled end points of ‘is willing to ignore laws and regulations’ and ‘will comply with laws and regulations under all circumstances’. These two items form a reliable and valid index (see SI Part B for more information), and I subsequently created an additive index, rescaled to 0–1 (mean = 0.53, sd = 0.29).
Control Variables
I also included respondents' gender (female = 1), age, and educational level (low ≤15 years when full-time education was completed; medium = 16–19 years when full-time education was completed; high ≥20 years when full-time education was completed). These provide a first picture of the socio-demographic correlates of party legitimacy perceptions. A measure of how moderate or extreme a citizen's ideological leaning was included because it might relate to the independent and dependent variables. Citizens with more extreme ideological leanings might not obtain the same signal from parties with governing experience, old parties, ideologically moderate parties, or democratically well-behaved parties as citizens with more moderate ideological leanings; their ideological moderation might also affect how willing they are to grant a party the same rights and powers as other parties. The measure is based on a survey asking respondents to locate themselves on a 5-point left-right scale. I folded the variable at its mid-point, and reversed the scale to measure the citizen's ideological moderation. Question wordings can be found in SI part B.
Estimation
The data has a hierarchical structure with variables measured at different levels (party and respondents). To control for any unmeasured party- or country-level factors, I estimate multi-level models with random intercepts for parties and countries. After controlling for party- and country-level differences, the resulting coefficients indicate the estimated association between a party characteristic and a respondent's average predicted legitimacy perceptions of that party. Survey weights were applied throughout.Footnote 12
Comparing a null model that only includes random intercepts for parties with a null model that includes random intercepts for countries shows that accounting for country-level differences fits the data structure better. The differences in fit statistics are statistically significant.Footnote 13 The null model accounting for party- and country-level differences also show an intraclass correlation coefficient of 0.04, which means that the variation between parties/countries is relatively small and the individual-level variation is much larger (similar to what Fig. 1 shows). This means that the variation between parties is more contained, indicating that parliament as an institution has legitimacy-conferring powers that citizens pick up on when evaluating individual parties' legitimacy. It also means that the party-level cue effects, theorized and identified, will be more limited in size.
First results from a model that only includes the socio-demographic variables show that women have lower levels of legitimacy perceptions when compared to men, and citizens with higher levels of education see parliamentary parties as more legitimate than citizens with lower levels of education (see Table C2 in SI). Citizens' ideological moderation, however, has non-linear effects, where those that place themselves as ‘moderate left/right’ have the highest party legitimacy perceptions, followed by those declaring themselves as ‘left/right wing’ and ‘centrist’, respectively.
Results
Table 1 summarizes the results testing the hypotheses. In Models 1–4, I test the hypotheses sequentially. According to the hypotheses, parties with governing experience, old parties, ideologically moderate parties, and parties that display democratic behaviour are seen as more legitimate by citizens. Therefore, positive and statistically significant coefficients in all the models are indicative of these effects, which Table 1 also shows. Parties with governing experience are seen as about 0.03 points more legitimate on a 0–1 scale than those without (Model 1). Old parties are about 0.04 points more legitimate than new parties (Model 2).
Table 1. Multilevel model results predicting legitimacy perceptions: random intercepts for parties and countries
Post-estimation analyses of the continuous variables show that ideologically centrist parties (value 5 on that scale) are seen as most legitimate with an average predicted value of about 0.82, about 0.06 points higher than their most extreme counterparts (based on Model 3). This may not seem a lot, but considering the generally high levels of party legitimacy, this is a substantial difference on the 0–1 legitimacy scale.Footnote 14 Yet, a party's perceived democratic behaviour has an even stronger association with the citizens' party legitimacy perceptions (Model 4). Parties seen as displaying democratic behaviour at one standard deviation above the mean, 0.82, compared to those at one standard deviation below the mean, 0.24, have a 0.17 point higher perceived legitimacy score. Note that even those parties at the low end of the democratic behaviour scale, seen as challenging democratic norms and willing to ignore rules and regulations, still have a predicted legitimacy value of well above 0.5, namely 0.63. This might indicate that citizens are strong adherents to democratic processes and believe that a parliamentary party should be accepted as legitimate simply because it successfully competed in a democratic election. The results in Table C4 and Figure C1 provide first indication for this, showing that institutional trust has a positive and statistically significant association with legitimacy perceptions and that those with higher levels of trust tend to respond more positively to a positive signal of a party characteristic.
Overall, the results in Table 1 indicate that perceptions of democratic behaviour have the strongest effect, followed by age, governing experience, and ideological moderation. Once I enter all predictors jointly (see Model 5), almost all effects decrease in size, and those of governing experience and ideological moderation even disappear.Footnote 15 This underscores the difficulty of separating different party-level information in the real world. However, the cross-national findings provide collective support for all hypotheses and thus for the argument that citizens use party-level information to form assessments of a party's legitimacy because they provide them with cues on whether the party poses a threat to democracy.Footnote 16
Experimental Analysis (Study 2)
With its high external validity, Study 1 provides broader patterns from the real world across Western Europe. However, to causally identify the informational effects of party characteristics on legitimacy perceptions and to rule out endogeneity, an experiment is needed in which information is systematically provided or withheld about a party characteristic. Additionally, some party characteristics, such as ideology and democratic behaviour, are often conflated in reality (Bourne and Bértoa Reference Bourne and Bértoa2017; Engler et al. Reference Engler2022). Unlike any of the other party-level factors, these two can only be separated in the controlled environment of an experiment. Study 2 aims to (1) separate the informational effects of a party's ideological moderation and democratic behaviour and (2) provide a stricter test for the argument with high internal validity because the information was systematically provided or withheld. Derived from the same theory but tailored to my underlying argument about the informational effects of party-level characteristics, the specific hypotheses I test in this experiment shift focus to comparing different party-level information with no information about these factors. Specifically, I expect that information about a party's ideological moderation [extremity] generates higher [lower] legitimacy perceptions amongst citizens than no information. Information about a party's pro-democratic [anti-democratic] behaviour generates higher [lower] legitimacy perceptions amongst citizens than no information. In additional analyses, I also compare the effects of different party-level information against each other.
In the experiment, I used a fictitious party to exclude the possibility of projection effects and to estimate the effects of systematically providing and withholding party-level information. The experiment was pre-registeredFootnote 17 and conducted in Denmark. This context is representative of many other multiparty West European countries in which parties represent a broad ideological spectrum, including radical-left and radical-right parties. Moreover, like many other West European countries, individual politicians in Denmark have engaged in undemocratic behaviour in the past.Footnote 18 Both factors mean that experimental treatments might not seem entirely unrealistic to Danish respondents and that the results might also travel to other West European contexts.
Experimental Treatments
The online survey experiment presented respondents with short mock articles (around 180 words each) from a major Danish daily newspaper with no ideological leaning. These were written in close collaboration with a Danish journalist to increase ecological validity.Footnote 19 We made efforts to mimic the language, style, and content of typical news items in that newspaper as closely as possible. The articles describe a fictitious political party (‘Party M’). In the treatment groups, the articles report on the party's ideology, and its democratic behaviour on the final day of the election campaign for parliament.Footnote 20
Respondents were randomly assigned to one of five different ideological leanings of Party M (left-wing, moderate-left, centre, moderate-right, right-wing). The newspaper articles mention one of them and include a matching expert opinion locating the party's programme on the left-right scale. This increases the construct validity of the treatment.
Respondents were also randomly assigned to one of two conditions of democratic behaviour. Party M expressed either a pro-democratic or an anti-democratic attitude. I operationalized the expressed attitude through Levitsky and Ziblatt's (Reference Levitsky and Ziblatt2018, 23f; see also Linz Reference Linz1978) multi-dimensional ‘litmus test’ for authoritarian behaviour. I created operationalizations for each of the four dimensions following the authors' examples. I included one per article, and the respondents were asked to read all four articles, making the set of articles the treatment. This increased the content validity of the treatment. I had no theoretical expectation for varying effects across dimensions, not least because Linz (Reference Linz1978, 31) also suggested that ‘none of [the criteria] appears as necessary and sufficient’. The order of articles was randomized across respondents. To attach the democratic behaviour to the entire party, all four articles attributed the opinion statements to different leading politicians of Party M.
To illustrate the treatment with an example, the respondents in the anti-democratic condition read four short articles on Party M, in which all four articles attributed an expression of anti-democratic attitudes to Party M. Across the articles, the kinds of anti-democratic attitude and the position of the representative, representing Party M, differed. Descriptions of all treatments and question wordings can be found in SI part D, including a screenshot from one treatment. Several manipulation tests confirmed that the treatments worked as intended. For detailed information, see SI part E.
The control group only read shorter versions of two of the articles. Any mention of ideology or democratic behaviour was removed. This group served as a benchmark for assessing citizens' legitimacy perceptions of Party M without any cues about its ideology or democratic behaviour, and it was crucial for testing the effects of any information about the two conditions.
After reading the articles, the respondents answered the same battery of questions on legitimacy perceptions as used in Study 1 on a 7-point bipolar disagree-agree scale, asking about the need to (a) grant Party M the same rights as other parties, (b) discuss Party M's policy proposals like others', and (c) accept and comply with Party M's policy decisions like others'. I formed an additive index because the battery showed good measurement properties (inter-item correlations = 0.74–0.81; Cronbach's α = 0.9; loadings of 0.57–0.59 on the first principal component). Ranging again from 0 to 1, the index shows that Party M's average party legitimacy perceptions match those of real-world Danish parties in Study 1, in fact, exactly so (mean experiment = 0.84; mean observation = 0.84 sd experiment = 0.21; sd observation = 0.25). This lends further credibility to the external validity of Study 2. In the following, the intent-to-treat effects of any information about party characteristics are estimated, using the control group as a base.
Result
Beginning with the informational effects of ideological moderation on party legitimacy perceptions, Fig. 2 reports the results based on simple linear regression models for the overall legitimacy index (right-most panel) and the individual items it contains (panels 1–3). To test the effects of any information about a party's ideological moderation on legitimacy perceptions, the comparision focuses on each of the treatment groups versus the control group that did not receive any information.
Figure 2. The informational effects of ideological moderation on party legitimacy perceptions based on linear regressions for the index and the individual items.
Note: The figure is based on Table F1 in SI.
Focusing on the overall index of legitimacy, the results show that the effects of Party M being described as either ‘left wing’ or ‘right wing’ are negative and statistically significantly different at the 10 per cent-level from the baseline scenario when Party M is not described in any ideological ways. Respondents receiving information about Party M's more extreme ideological leaning considered Party M, on average, 0.02 points less legitimate on a 0–1 scale than respondents not receiving any information about the party's ideology (see also Table F1 in SI). This provides empirical support for the argument. The analysis also reveals that when Party M is described as more ideologically moderate (‘centre-left’ or ‘centre-right’), its legitimacy perceptions are unaffected because the effects are not statistically significantly different from the control group. Moving closer to the ideological centre, respondents receiving information about Party M as having a ‘centrist’ leaning see it as the least legitimate (−0.03, p = 0.02). While this is not in line with the argument, it may be explained by the peculiarities of the Danish context. As the ideological bloc system was well-established until only recently (Hjorth Reference Hjorth, Møller Hansen and Stubager2017; Kosiara-Pedersen Reference Kosiara-Pedersen2020), Danes may have struggled to fully believe in a centrist party. Those parties who claimed to be centrist always had to choose a side after the election. Since the articles described Party M one day before the election, it could explain the particularly low legitimacy ratings. Alternatively, Danish respondents may have simply thought that such a party defies stereotypes. Overall, these experimental results partially support the argument because they show that information about a party's extreme ideological leaning leads to lower legitimate ratings amongst citizens.
The other panels in Fig. 2 also show that these results are driven mainly by perceptions of the extent to which Party M's proposals need to be discussed (panel 2) and their decisions need to be accepted (panel 3). Party M's extreme ideological leaning does not affect the extent to which respondents would grant it the same rights in parliament as other parties (panel 1).
In additional analysis (see Table F2 in SI), I show that respondents differentiate between left- and right-leaning parties: the negative informational effects of Party M being described as ‘left-wing’ or ‘right-wing’ are entirely driven by the aforementioned. Legitimacy assessments of Party M do not differ from the control group whenever it is described as ‘left wing’, but they are negative and statistically significant at the 90 per cent level whenever it is described as ‘right wing’ (see Engler et al. Reference Engler2022 for complementary results). In addition, parties described as ‘centre-right’ also cause a drop in legitimacy assessments (p = 0.068) of the same magnitude as the ‘right-wing’ treatment; that is, about 0.02 points on a 0–1 scale. This means that party legitimacy perceptions are not only driven by information about how ideologically moderate a party is but, more specifically, by information on how right-leaning a party is.Footnote 21
Moving to the effects of information about a party's democratic behaviour on party legitimacy perceptions, Fig. 3 reports again the effects of each treatment on the index and its components (see also Table F3 in SI). The results show that information about Party M's anti-democratic attitudes leads to around 0.26 points lower legitimacy perceptions than no information about Party M's democratic behaviour (final panel). These effects are also statistically significant at the 95 per cent level and thus support the argument. Note: they are again driven by assessments on discussing Party M's proposals (panel 2) and accepting their decisions (panel 3).
Figure 3. The informational effects of democratic attitudes on party legitimacy perceptions based on linear regressions for the index and the individual items.
Note: The figure is based on Table F3 in SI.
According to Fig. 3, information about Party M's positive democratic behaviour has no positive effects. Indeed, party legitimacy perceptions are always lower (but not significantly so) than in the ‘no information’ group; that is, the control group. It could be that the pro-democratic treatments were not strong enough because, in the main experiment (unlike in the pre-test), this group did not differ significantly from the control group in its assessment of how democratic Party M is (see SI part E). Additional analyses therefore estimate the average treatment-on-the-treated effects for the informational effects of a party's pro-democratic behaviour. The results are in line with the general argument because information about a party displaying pro-democratic behaviour increases people's average perception of party legitimacy by 0.04 points (p < 0.001) when compared to no information (see Table F4 in SI). These results support the informational role of a party's democratic behaviour on citizens' legitimacy perceptions.Footnote 22
Finally, I probe my proposed causal mechanism, explaining the relationships between ideology/democratic behaviour and legitimacy perceptions: Party M is seen as posing a threat to democracy in the sense of likely eroding democratic processes and norms. I approximate a causal mediation analysis using a model-based approach, and I create a direct measure of respondents' perceptions of Party M posing a threat to democracy, ranging from 0 (‘no threat’) to 1 (‘real threat’), based on a post-treatment question (see Table F5 in SI for more information). In both conditions, the initial effects (ideology and democratic behaviour) disappear, and the effects of perceptions of a threat to democracy are negative and highly statistically significant (see Models 1 and 3 in Table F5 in SI). In addition, the treatments of both conditions also have the expected effects on threat perceptions, where a more extreme ideology and anti-democratic behaviour are associated with a higher perceived threat (Models 2 and 4). This suggests that perceptions of Party M posing a threat to democracy could be a relevant mechanism that explains the informational effects of M's ideological extremity and its democratic behaviour.
Conclusion
This paper proposes and studies a novel perspective on assessments of parties by considering citizens' legitimacy perceptions of political parties as institutional players in the democratic game. My central claim is that citizens use party characteristics as informational cues to form their party legitimacy assessments because they provide cues on whether a political party does not threaten democracy. I developed a theoretical framework to explain the informational effects of party-level characteristics, and I leveraged an observational survey of citizens in seven Western European countries as well as a large-scale survey experiment to test the argument.
Overall, the results show that party legitimacy perceptions vary in predictable ways by party: citizens consider a party more legitimate when it has governing experience, is older, is ideologically non-extreme, and endorses democratic rules. The experimental results provide additional nuance in demonstrating that the negative informational effects of a party's ideology on legitimacy perceptions are confined to parties with extreme or moderately right ideologies when compared to no information about ideology. These findings suggest that citizens have distinct perceptions of their parliamentary parties' legitimacy related to party-level characteristics, which, in theory, can also change. As the results suggest, parties can gain and lose in perceived legitimacy depending on their parliamentary status (with governing experience or not), their age, how ideologically right-leaning they are, or how democratically they act. However, the results also indicate that a party's parliamentary representation is not a guarantee for citizens' legitimacy perceptions. Political parties might thus fulfil important democratic functions in democracies. However, perceptions of their legitimacy as institutional players in the democratic game are not self-evident, even when elected to parliament. Combined with existing work on the importance of party legitimacy for voters' party preferences and party competition (see, for example, Bos and van der Brug Reference Bos and van der Brug2010; De Vries and Hobolt Reference De Vries and Hobolt2020), party legitimacy appears to be another valence issue.
In addition, the identified variation in party legitimacy perceptions indicates a possible relationship with affective polarization or negative partisanship. My results show that the party-level variation in legitimacy perceptions is limited and that citizens differ substantially in evaluating individual parties. Validity tests of the measure also showed that, while legitimacy perceptions are distinct from measures used to capture affective polarization (party sympathy) or negative partisanship (vote probabilities), they are still correlated. However, it is already known that affective polarization and negative partisanship originate in group identities (see, for example, Iyengar et al. Reference Iyengar2019). Therefore, it seems plausible that the causal arrow flows from affective polarization and negative partisanship to legitimacy perceptions rather than the other way around: citizens that are affectively polarized or hold a negative partisanship with a party may also consider that party less legitimate because it is just too unacceptable to tolerate as an institutional player.
Despite the merits of two different research designs, the current study is necessarily limited and offers new avenues for future research. First, the fact that the experimental group that obtained information about Party M as ‘centrist’ thought of it as least legitimate deserves further scrutiny. As suggested above, the specifics of the Danish context might explain this, but further research should examine the issue further. Second, this article is limited in theoretical and empirical scope. Extending the theoretical framework to other analytical levels is important to account for the possible influence of individual attitudes and their heterogeneity. I already showed that institutional trust matters, and that there is variation in party legitimacy perceptions that is not accounted for by party-level characteristics. Other extensions should also involve party competition or the role of interpersonal networks in shaping party legitimacy perceptions. Likewise, the study of party legitimacy perceptions and their causes should be extended to other contexts, such as Eastern Europe, to test the possible influence of a party system's institutionalization – much points to party legitimacy perceptions being an increasingly important evaluation among citizens.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0007123423000364.
Data availability statement
Replication Data for this article can be found in Harvard Dataverse at: https://doi.org/10.7910/DVN/Y7EIEC.
Acknowledgements
Previous versions of this paper were presented at the Party Politics workshop in Aarhus and the General Research Seminar at the University of Gothenburg in September 2020, at the research seminar at the University of Vienna in December 2021, at the New York University in March 2022, at the GEPOP conference in Gothenburg in May 2022, at the University of Trier and the Ludwig-Maximilian University München in June 2022, and at the EPOP conference in Newcastle-upon-Tyne in September 2022. I want to thank all participants and discussants at these occasions and the many other colleagues who provided comments in-between for carefully reading the paper and for their constructive and valuable feedback. I would like to thank in particular Ole Ellekrog for his help in constructing the newspaper articles.
Financial support
This work was supported by Aarhus University Foundation (AUFF, grant number AUFF-E2017-7-13); and Riksbankens Jubileumsfond (RJ, RIK18-1267:2).
Competing interests
None.
Ethical standards
The research was conducted in accordance with the American Political Science Association's (2020) ‘Principles and Guidance for Human Subjects Research.’
Open access
