Introduction

How does elite polarization impact citizens’ evaluations of politics? In both public and scholarly debates, polarization is often viewed in a negative light, being associated to contentious politics, ineffective government, and democratic erosion (Levitsky and Ziblatt 2018; McCoy and Somer 2019). Among citizens, elite polarization is associated to lower satisfaction with politics and democracy (Uslaner 2015; Banda and Kirkland 2017). Such claims are echoed in the context of Dutch politics. Polarization is said to be increasing from the 2000s onwards (Oosterwaal and Torenvlied 2010; Van der Berg 2021) and is a suspected cause of declining political trust (Bovens and Wille 2008). Yet, interestingly, an opposite concern is equally prominent in the Dutch context: Not polarization, but a lack of policy choice is argued to be at the heart of growing public discontent (Van der Kolk and Aarts 2011). As a case in point, in the 2021 Dutch parliamentary elections voters perceived less differences between parties than in earlier elections (Van Holsteyn and Irwin 2021).

In this debate, increased positional distance between parties (ideological polarization) is often implicitly linked to increased animosity between political actors (elite incivility), even though these are conceptually different phenomena. Existing research on the consequences of elite polarization for citizens’ political attitudes is not able to separate these two dimensions of elite conflict due to its focus on the US context (Uslaner 2015; Banda and Kirkland 2017; Robison and Mullinix 2016). In the US, an extreme case in terms of its strongly polarized two-party system, the increasing ideological polarization between the Democratic and Republican party has gone hand in hand with increasing levels of elite incivility (Skytte 2021), thereby making it difficult to separate their effects. Yet, it is plausible that ideological polarization and elite incivility may crucially differ in their effects on political support. Especially in more moderate contexts, ideological polarization can help to clarify policy choice and raise the stakes of elections, and thereby positively impact citizens’ orientations toward politics. Indeed, cross-national studies including Western European countries have demonstrated positive effects of elite polarization on voter turnout (Hobolt et al. 2021), partisanship (Lupu 2015), and political efficacy (Kittilson and Anderson 2011).

Our study is among the first to analyze the impact of these two dimensions of elite conflict—ideological polarization and elite animosity—on citizens’ political support. The Dutch Parliamentary Election Survey 2021 offers a unique opportunity to disentangle the effects of these two dimensions, as it enables the measurement of perceived ideological polarization (using party placements by citizens) and perceived elite incivility (using newly developed survey items). In addition, it enables us to study the individual level mechanisms between elite polarization and political support, by addressing the moderating role of citizens’ attitudes toward political conflict (i.e., stealth democratic preferences). Given that these two dimensions of elite conflict are rarely studied in tandem, we first descriptively show to what extent different groups of Dutch citizens perceived elite conflict along these two dimensions in the 2021 parliamentary elections. In a second explanatory step, we analyze how perceived elite incivility and perceived ideological polarization impact citizens’ political support.

In doing so, our study makes four contributions. First, only few studies outside the US have assessed the impact of elite polarization on political support, and our study thus contributes to this nascent line of research. Second, it expands on this research by simultaneously including two different dimensions of elite conflict that are often conflated in both public and academic debates under the header of “polarization”, which enables us to disentangle their potentially differential effects. Third, we innovate by studying the consequences of ideological polarization not only along the left–right continuum, but also along the cultural dimension of political conflict. Finally, we test for the conditionality of effects by taking into account citizens’ stealth democratic attitudes (i.e., their dislike of political disagreement) as a potential moderator.

Our findings demonstrate the importance of precise conceptualizations when debating the consequences of polarization. Where perceived incivility consistently has a negative effect on political support, the effect of ideological polarization depends on the specific dimension along which polarization is measured. Perceived polarization on the left–right and economic issue dimensions increases political support, whereas perceived cultural polarization (on the issue of integration) has no effect. For the wider debate on the consequences of polarization, these findings show that the democratic consequences of elite polarization are not necessarily negative, as citizens are able to evaluate tone and substance independently when making judgments about politics.

Theory

Political support and perceptions of conflict

Political support entails citizens’ orientations toward political objects, ranging from highly diffuse (their political community) to very specific (individual political actors) (Norris 2011). The present study focuses on citizens’ support for the functioning of the democratic regime and its institutions, which falls in the middle of this range. Research on political support has shown that citizens’ orientations toward politics are determined by the interplay between the subject (the citizen) and the object (political actors, institutions, regimes), according to certain criteria or expectations (“A trusts B to do x,” Hardin 2000). Such criteria are found in policy output (mainly macro-economic performance, e.g., Hetherington and Rudolph 2008; Miller and Listhaug 1999), but evidence is accumulating that citizens strongly care about the political process through which such output is achieved. Evaluations of the political process are at least equally—or even more—important to political support compared to policy output itself (Van der Meer and Hakhverdian 2017). Such evaluations may relate to (the absence of) corruption (Hakhverdian and Mayne 2012), procedural fairness (Grimes 2017), as well as to concerns with representation and accountability (Van der Meer 2017).

The degree of perceived conflict is a crucial aspect of the political process in the eyes of citizens. In their seminal work, Hibbing and Theiss-Morse (2002) show that a majority of US citizens is averse of core aspects of democratic politics. In particular, citizens do not like political debate as it makes disagreement visible, but they also dislike political compromise. This is explained by the idea of “false consensus”: as citizens tend to overestimate the amount of consensus among the population, conflict and deliberation among elites seems an unnecessary strategy of politicians mainly looking to increase their self-interest (e.g., their popularity in the polls). Open political contestation then becomes “just unnecessary mud-slinging” (Clarke et al. 2018, p. 262), which prevents the promotion of the general interest.

At first glance, the link between elite polarization—as a manifestation of political conflict—and political support seems clear: Citizens perceiving more political polarization should turn more negative toward politics, because they are averse to its adversarial image. This “polarization as conflict” approach is dominant in (mainly US-based) research. Yet, it is crucial to distinguish two dimensions of elite conflict which can exist independently: elite incivility (i.e., a polarizing tone and style) and ideological polarization (i.e., positional differences between parties). Parties may disagree about the content, while interacting respectfully in the debate. Equally imaginable is that they use a highly hostile tone toward their opponent without showing any substantive difference in their policy preferences. Importantly, it is not evident that both dimensions of polarization should have similarly negative effects on political support (Skytte 2021; Paris 2017). Where elite incivility clearly relates to the “polarization as conflict” perspective, ideological polarization can signal conflict as well as choice to citizens. According to the “polarization as choice” perspective, by offering meaningful policy alternatives, polarized parties may strengthen the quality of representation in the eyes of citizens, and thereby lead to more positive orientations toward politics. We outline competing expectations on the effects of ideological polarization below, but first turn to the effects of elite incivility.

Elite incivility

Elite incivility or animosity is at the core of what citizens dislike about political conflict. The bickering and mud-slinging of political elites leads to political dissatisfaction in two ways. First, animosity (e.g., personal attacks) violates the normative expectations that people have of politics. As Funk (2001) explains, “policy debate that includes personal animosity is likely to violate norms of courtesy and reciprocity in resolving social conflict” (p. 197). Second, incivility might create the perception that elites are not pursuing the general interest. Disrespectful and uncivil debate signals that politicians are concerned with their self-interest (e.g., winning votes, improving their own image at the cost of the other) while not focusing on what is actually important (Hibbing and Theiss-Morse 2002).

Indeed, several studies show that the tone and style of political debates have a direct effect on political attitudes, and political trust specifically. Mutz and Reeves (2005) provide experimental evidence that rather than policy disagreement per se, it is the level of incivility in televised political discussion that impacts political trust. Such incivility can be detrimental to the legitimacy of the political process (Mutz 2015), as is corroborated by other experiments showing that conflict-laden political news coverage (Forgette and Morris 2006), uncivil political communication (Goovaerts and Marien 2020), and animosity between political elites (Funk 2001) depress political support.

Where these experiments show that single (manipulated) instances of incivility exert negative effects, in reality there is likely a “cumulative effect of personal attacks, scandals and incivility among politicians [which] is likely to contribute to a growing disinterest and dissatisfaction with politics” (Funk 2001, p. 203). We therefore expect a generalized perception of elite incivility over the election campaign to negatively impact political support.

Hypothesis 1

Perceived elite incivility has a negative effect on political support.

Ideological polarization

Another body of research analyzes the effects of ideological polarization on political support. This literature predominantly takes a “polarization as conflict” approach, expecting that higher policy disagreement between parties leads citizens to perceive more conflict, and consequently turn dissatisfied with the political process. This can be simply because visible disagreement and political debate make citizens wary of politics (Hibbing and Theiss-Morse 2002), or because higher polarization decreases the quality of fruitful democratic deliberation (Hoerner and Hobolt 2020) or effective government (Uslaner 2015). Elite polarization, in sum, may trigger negative evaluations of the political process.

Several studies have shown that increased polarization is associated to higher levels of political dissatisfaction and distrust, mostly focusing on the US context. At the macro-level, ideological polarization in Congress is associated with lower Congressional approval (Jones 2015; Uslaner 2015; Ramirez 2009). Polarization also accounts for between-state differences in political trust: In US states where the legislature is more polarized on ideological issues, citizens tend to trust it less (Banda and Kirkland 2017). In an experimental study, Harbridge and Malhotra (2011) find that lower polarization (i.e., bipartisan law-making) results in higher levels of trust in Congress. A recent study by Hoerner and Hobolt (2020) suggests that similar negative effects of elite polarization (measured in party manifestos) exist on satisfaction with democracy in the European context. Using a more instrumental argumentation, some studies explain the negative effect of polarization on political satisfaction by the increased spatial distance between parties and the average citizen (King 1997; Ezrow and Xezonakis 2011).

Yet, there are empirical and theoretical reasons to refrain from the simple conclusion that elite polarization harms political support. Empirically, existing studies do not disentangle the effect of issue polarization from that of elite animosity. The majority focuses on the US context, where issue polarization and incivility go hand in hand, as “since the early 1980s increasing ideological differences between the parties have led to harsh words between members and less willingness to search for common ground” (Uslaner 2015, p. 361). This means that if we find an association between elite polarization and political support, it is unclear whether this is because of increasing positional differences, or rather because increased polarization means more uncivil, hostile politics. This point has been made by two recent experimental studies in the US. Manipulating both levels of incivility and issue polarization, Skytte (2021) finds that incivility decreases trust in politicians, while issue polarization of elites has no such effect (Skytte 2021). Paris (2017) finds in a survey experiment that bipartisanship has positive effect on trust in Congress when controlling for the confounding (negative) effect of incivility. These studies thus confirm the negative effects of incivility, but the results for ideological polarization are mixed.

Theoretically, from a “polarization as choice” perspective, conflict in terms of policy disagreement might actually have its virtues. Conflict is a crucial element of representative democracy (Schattschneider 1960), as a sufficient level of policy choice perceived by voters is crucial for representative democracy to function (Bartolini 2000) and has intrinsic democratic value (Przeworski 2003). Responsible party theorists have long emphasized the importance of parties offering meaningful policy alternatives to strengthen the quality of representation. This led the American Political Science Association in 1950 to call for parties to offer “policy alternatives (…) necessary for a true choice in reaching public decisions” (APSA Committee on Political Parties 1950).

Perceptions of meaningful party differences may directly influence citizens’ perceptions of the quality of the political system. Party polarization, by clarifying polarization clarifies what parties stand for (Lupu 2013), gives voters the feeling that their vote makes a difference, and thereby not only increases their political engagement and participation in elections, but also their satisfaction with the political process more generally (Kittilson and Anderson 2011). Inversely, the perception that parties do not offer meaningful choice can lead to alienation or indifference (Torcal and Magalhães 2022). Indeed, ideological party polarization—both when based on expert and citizen measurements—positively impacts partisanship across Western democracies (Lupu 2015). Similarly, several studies document a positive impact of party polarization on turnout, both when looking at macro-level correlations (Hetherington 2008) and in multilevel models on individual survey data (Hobolt et al. 2021; Moral 2017; Kittilson and Anderson 2011), though negative effects are also found in the US context (Rogowski 2014).

In sum, existing research paints a mixed picture. Studies on the impact of polarization on political support mostly find negative effects, yet most focus on the US (except for Hoerner and Hobolt 2020) and most do not take into account elite level incivility or animosity. On the other hand, there is evidence of positive effects of ideological polarization on engagement and turnout, especially in the European context, which could translate into positive evaluations of the political system. Given the inconsistency of findings to date, we formulate competing hypotheses.

Hypothesis 2a

Perceived ideological polarization has a negative effect on political support.

Hypothesis 2b

Perceived ideological polarization has a positive effect on political support.

If both lines of reasoning hold merit, it is possible that perceived ideological polarization reaches an optimum at intermediate levels—with political dissatisfaction being largest at either low or high levels of polarization. A recent study by Torcal and Magalhães (2022) has tested for a curvilinear effect of perceived polarization on support for democratic principles. According to their argument, such support should be lowest for citizens who perceive no differences at all (because the perceived lack of choice makes them alienated or indifferent toward politics) as well as for citizens who perceive politics to be too divided (because they perceive no possibilities for compromise or effective policy-making). The authors find support for their expectation across a heterogeneous set of democracies (with highly varying polarization levels). Their study, however, did not control for the potentially confounding role of perceived incivility, which may coincide with extremely high levels of ideological polarization. We will test for curvilinear effects in our models as well (Appendix 1).

Elite ideological polarization is mostly measured along the left–right (or conservative-liberal) dimension. This might not do justice to the fact that Western European party systems are generally structured by economic and cultural dimensions of conflict (Kriesi et al. 2012). Left–right positions do not only reflect positions on the first dimension, but increasingly also on the latter, as the left–right dimension is a “super-issue” which continually absorbs new issues (Van der Eijk et al. 2005; De Vries et al. 2013; Jahn 2022). It is an empirical question whether party polarization on specific issue dimensions—economic or cultural—has a similar effect to general left–right polarization. In our analysis we first consider polarization along the general left–right dimension. In a second step, we look at perceived polarization on the economic issue of income redistribution, as well as on the issue of integration policy—a core issue of the cultural dimension (Kriesi et al. 2012) on which political elites are increasingly polarized (Oosterwaal and Torenvlied 2010).

Conditional effects

Citizens differ in their process preferences, and their diverging preferences are likely to moderate the impact of incivility and polarization. Hibbing and Theiss-Morse (2002) introduced the term “stealth democrats” to describe citizens who are averse of political conflict, debate and compromise. As citizens tend to overestimate the extent to which other citizens hold the same ideas and objectives (“false consensus”), they prefer invisible, unified government over politicians openly defending different interests or viewpoints (Hibbing and Theiss-Morse 2002, p. 132).Footnote 1 In other words, “Politicians should do what they say and get on with the task of governing—quietly, effectively, and efficiently” (Stoker and Hay 2017). Debate and disagreement, in the eyes of these citizens, only serve special interests or politicians’ own gains. As a consequence, stealth democrats display a preference for government by a political actor (e.g., business leaders, experts). Support for stealth democracy is defined as the combination of these two dimensions—a disliking of visible political disagreement and a (resulting) support for a-political governance. Stealth democratic preferences have been shown to explain (lack of) political participation as well as voting behavior (Webb 2013; Lavezzolo and Ramiro 2018).

Particularly the first dimension of stealth democratic attitudes—a dislike of political disagreement—is likely to condition the effects of perceived elite conflict on political support.Footnote 2 First, citizens who are averse of political disagreement should react more strongly to political incivility, as it is the opposite of quiet or invisible politics. Uncivil interactions between politicians are precisely the type of bickering and mud-slinging that represent the “ugly side” of politics to citizens, and we expect citizens with stealth democratic preferences to react more negatively to such incivility. Second, for the effect of perceived ideological polarization we have a similar expectation. It is to be expected that citizens who dislike political disagreement out of an assumption of “false consensus” are put off by ideological or issue-based polarization, and are more likely to interpret such polarization as unnecessary conflict-seeking of self-interested politicians. The following hypotheses are formulated to test these two conditional effects.

Hypothesis 3

The effect of perceived elite incivility on political support is more negative among citizens with stealth democratic preferences.

Hypothesis 4

The effect of perceived ideological polarization on political support is more negative (or less positive) among citizens with stealth democratic preferences.

Data and methods

We test the hypotheses on the case of the Netherlands. Concerns about elite polarization have been raised in the context of Dutch politics (Bovens and Wille 2008). Incivility and negativity have been part of Dutch political campaigns since the 1970s (Walter 2012), even if the Dutch proportional system is less conducive to negative campaigning as parties depend on each other for government formation. While elites are increasingly polarized on cultural issues (Oosterwaal and Torenvlied 2010), observers have lamented the lack of meaningful ideological differences between Dutch political parties, which reflects a broader trend toward ideological convergence of political parties since the 1990s (Keman and Pennings 2011), a trend that continues into the most recent elections (Van Holsteyn and Irwin 2021). This might be due to the specific circumstances of the 2021 elections. As Van Holsteyn and Irwin observe (2021), the salience of the Covid pandemic—on which many parties lack a clear profile—came at the cost of substantive debates about other pressing (and potentially more profiled) issues. Furthermore, Prime Minister and VVD leader Rutte successfully shifted the campaign attention to issues of leadership, thereby dismantling the debate on substantive policy issues.

To test our hypotheses, we make use of the 2021 Dutch Parliamentary Elections Studies data (DPES). Our hypotheses require a dataset containing items on political support, perceived polarization (by means of perceived party placement) and perceived incivility. The DPES offers a unique combination of these items, as well as with a high-quality data collection. Respondents for this study were sampled in two ways: a random sample from the CentERdata LISS-panel was combined with a fresh representative sample drafted by I&O Research (Jacobs et al. 2021), using a pre- and post-election wave. Information about procedures and variables can be found in the DPES codebook (Sipma et al. 2021). In what follows we discuss the operationalization of the variables used in our study.

Political support

We measure political support by using the common indicator for satisfaction with democracy, asking citizens the following question: On the whole, are you very satisfied, fairly satisfied, not very satisfied, or not at all satisfied with the way democracy works in the Netherlands?, with response options ranging from very dissatisfied (1) to very satisfied (4). This item taps political support that is directed at the actual functioning of the political regime and falls in the mid-range between diffuse and specific support (Norris 2011). Alternatively, such mid-range political support has been measured as trust in political institutions (e.g., Van der Meer and Hakhverdian 2017). We replicated the models using a political trust index as dependent variable (details in Appendix 3).

Incivility

To measure incivility, five items were added to the post-election wave of the Dutch Parliamentary Election Studies. The items are based on work by Mutz and Reeves (2005), using the following question: What do you think of the tone of the debates between politicians during the election period? Do you think the tone of most politicians were mostly [….]. Respondents were presented with the following opposing descriptions to describe the debates on a five-point scale: Impolite-Polite, Heated-Calm, Emotional-Formal, Hostile-Friendly and Quarrelsome-Cooperative. The Cronbach’s Alpha is 0.77 indicating an internally consistent scale. Again, the additive scale is reverse coded and divided by five, so that a score of 5 represents a perception of highly uncivil election debates, and a score of 1 stands for maximal “civility”.

Perceived ideological polarization

Perceptions of polarization can be measured in different ways. The simplest way is to calculate the difference between the most right and most left party according to the respondents (van der Kolk and Aarts 2011). However, this method does not account for convergence and divergence at the center of the political spectrum. A more sophisticated way of measuring perceived polarization is taking the average difference between a respondent’s perceived left–right position of each party and the mean left–right position of all parties for this respondent. In addition, it is possible to account for the relative size of the parties by weighting the parties by their electoral share in parliament (Wagner 2021; Torcal and Magalhães 2022). We follow these recent studies by computing a weighted perceived ideological polarization index. It has to be noted, however, that due to the election of several new parties and the fragmentation of the party system, some parties suffered a fairly large number of missing values. To solve this problem, we exclude parties with more than 30% missing values on their left–right placement, and weight the seat share of included eight parties relative to this reduced total (123 seats out of the 150 seats).

Besides left–right party placements, the DPES asks respondents to place parties on several political issues. In order to test H2 and H4 for polarization on a typical economic and cultural issue, we create a similar weighted perceived polarization indicator for polarization on party positions toward income redistribution (economic issue) and integration of immigrants (cultural issue). Respondents were asked to place parties as follows: Where would you place the following parties and yourself on a line from 1 to 7, where 1 means differences in income should be increased and 7 means that differences in income should be decreased. For positions on integration, a similar item asked respondents: Where would you place the following parties and yourself on a line from 1 to 7, where 1 means preservation of own culture for foreigners and 7 means that they should fully adapt? When constructing indices for issue polarization, we used the same eight parties as for left–right polarization and restricted the weighting accordingly.

Stealth democratic attitudes

To test the moderation effects (H3–H4), we use two items that reflect a dislike of visible political conflict, namely “Politician talk too much and do too little”, and “What people call compromise in politics is really just selling out on one's principles”. These items reflect the “anti-disagreement” dimension of stealth attitudes, and are highly similar to items originally used by Hibbing and Theiss-Morse (2002, pp. 135–136).Footnote 3 Both items are measured on a five-point scale, ranging from “strongly agree” to “strongly disagree,” and they form a reliable index (Cronbach’s alpha= 0.63). The direction of these items is reversed and divided by two, generating a 1 to 5 index in which higher scores indicate a stronger dislike of political disagreement, and thus stronger stealth democratic attitudes.

Control variables

Apart from the main variables of interest, we include several commonly used control variables in the analysis, which are age, gender, education-level, occupational status, political interest, and media consumption. The latter two are particularly important as controls, since they are likely to drive both perceptions of polarization (e.g., the better informed perceive more differences between parties) and political support. Media use is measured using two separate items. First, to capture overall media consumption: “How closely do you follow politics (on TV, radio, newspapers, or the Internet)?”, and second, to account for social media specifically: “How often do you see messages about politicians and/or current political affairs on your Facebook, Instagram, Twitter, WhatsApp and/or YouTube?”. Furthermore, as the outcome of elections influences political support right after the elections (Anderson et al. 2005), we include a variable indicating whether the party one voted for increased, decreased, or maintained its vote share compared to the previous parliamentary election (including a category for non-voters). Finally, we account for the potentially confounding role of ideological positions by including left–right self-placement as well as left–right extremity. This latter variable is calculated as the absolute difference between each respondent’s left–right self-placement and the average of all respondents (Torcal and Magalhães 2022).

As a general rule, we used the pre-wave measurements for the control variables, as well as for the items used for the polarization indices. Perceptions of incivility and stealth preferences are only measured in the post-wave. To ensure temporal ordering of causal effects where possible, for political support we used the measures from the post-wave. A descriptive overview of all variables can be found in Table 1.

Table 1 Descriptive information on included variables
Full size table

Our analytical approach is as follows. First, we present a descriptive analysis, presenting the measures of perceived incivility and two forms of perceived ideological polarization for different groups of respondents. Second, we test our hypotheses through OLS regression. We build our models stepwise, starting with the main effects of the different dimensions of elite conflict (H1–H2), and then adding the interaction terms (H3–H4). We enforce weights on the DPES-data throughout, to account for both demographic and vote choice differences in the data.

Results

Descriptives

To assess to what extent the different dimensions of perceived elite conflict are independent concepts, we first assess their correlation. Perceived incivility is neither significantly correlated to left–right polarization (r = 0.02) nor to integration polarization (r = − 0.00), and weakly to income polarization (r = − 0.06). Left–right polarization is positively correlated to income polarization (r = 0.19) and integration polarization (r = 0.34), which makes sense given the fact that both economic and cultural issues correlate to the left–right dimension in the Netherlands (De Vries et al. 2013). That said, these correlations are only moderate and leave sufficient room for establishing independent effects.

In Fig. 1, we explore how perceived incivility and perceived polarization are related to relevant background characteristics.Footnote 4 Three patterns stand out. First, though economic and cultural issue polarization were measured on the same scales, the first is perceived as higher than the latter. This could be due to the fact that positions on income redistribution vary across all parties, while differentiation on the integration issue mainly sets apart the right-wing populists from the mainstream. Second, citizens with higher education perceive more polarization on economic and cultural issues (Fig. 1a), which could signal that they are better able to place parties on policy issues due to higher political knowledge or interest. Indeed, we find a similar pattern for political interest and media use, which positively relate to all three dimensions of perceived positional polarization (Fig. 1b, c). People who pay more attention to politics thus have a better notion of the positional differences articulated by parties. Finally, for perceived incivility the patterns run in the opposite direction—though differences are small. Citizens with higher interest and traditional media use perceive slightly less incivility. These associations underline the importance to control for these characteristics in the analyses.

Fig. 1
figure 1

Perceived incivility and perceived polarization across different background characteristics

Full size image

Explanatory models

We now turn to the regression models to test the hypotheses (Table 2). Model 1 shows the effects of perceived incivility and perceived left–right polarization on satisfaction with democracy. We find support for H1: Citizens who perceived more incivility during the election campaign are less satisfied with democracy (b = − 0.117). This effect is robust to the inclusion of the control variables. The effect of perceived left–right polarization, on the other hand, is positive (b = 0.105). This means that citizens who perceive more differences between party positions on the left–right dimension are more satisfied with democracy, in support of H2b.

Table 2 Regression models explaining satisfaction with democracy
Full size table

Model 2 adds interaction terms in order to test H3 and H4. The coefficient of the interaction between perceived incivility and stealth preferences is in the expected negative direction but not significant.Footnote 5 A marginal effects plot (Fig. 2) suggests that the effect of incivility on satisfaction with democracy is significant only for people who dislike political disagreement. Although this pattern is in line with H3, the evidence remains inconclusive as the coefficient itself is insignificant.Footnote 6 Similarly, the interaction effect between perceived left–right polarization and stealth attitudes is negative but not significant. The marginal effects (Fig. 3) show a pattern in line with H4, suggesting that the positive effect of left–right polarization on satisfaction with democracy does not exist for citizens averse to political conflict. Yet, evidence again remains inconclusive due to the insignificant coefficient.

Fig. 2
figure 2

Marginal effects of perceived incivility on satisfaction with democracy across stealth preferences

Full size image
Fig. 3
figure 3

Marginal effects for perceived left–right polarization on satisfaction with democracy across stealth preferences

Full size image

In Model 3 and 4, left–right polarization is replaced by economic and cultural issue polarization. As Model 3 shows, the effect of perceived economic polarization mimics the effect of general left–right polarization (b = 0.109). Citizens who perceive parties as polarized on the issue on income redistribution are more satisfied with democracy, while perceptions of cultural polarization have no effect. Finally, in Appendix 1 we present models including curvilinear effects of perceived polarization. The results are substantively similar to the linear effects: left–right polarization and economic polarization have a positive effect on satisfaction, though at very high levels of economic polarization there is a slight decrease. The effect of cultural polarization remains insignificant (Figs. 4, 5, 6).

Among the control variables, the results are in line with what we would expect (based on Model 1). Satisfaction with democracy is lower among right-wing citizens and at the extremes of the left–right scale. Winners are more satisfied with democracy than losers, while those who did not vote are less satisfied. Media use in general has no effect, but those who receive political news through social media are less satisfied with democracy. Finally, it is notable that females are generally less satisfied with politics than males, and higher educated more satisfied than lower educated citizens.

Robustness checks

We checked for the robustness of the results in the following ways. First, satisfaction with democracy is measured on a four-point scale. This item has been modeled by OLS models before (cf. Hoerner and Hobolt 2020), but a robustness check shows that ordered logistic regressions give highly similar results (Appendix 2). Second, we tested whether the results hold for trust in political institutions (Appendix 3). Again, H1 is corroborated, as incivility again has a significantly negative effect (b = − 0.086). The effects of perceived polarization are, however, not significant for political trust, refuting H2a and H2b. The conditional effects (H3 and H4) are again insignificant, yet the marginal effects for H3 show a similar pattern as found in the main models (Fig. 7).

Finally, we tested for the possibility of confounding effects by building stepwise models. Appendix 4 enters the key independent variables (perceived incivility and polarization) one by one, and shows that the effects remain highly similar when included separately. Furthermore, given that perceptions of polarization are potentially endogenous to variables like political sophistication, we tested to what extent media use and political interest confound the effects by including them stepwise, in Appendix 5. These models show that the effects of perceived incivility and perceived polarization are highly robust and do not change upon inclusion of these potential confounders.

Conclusion

Elite polarization generally has a negative connotation in the public debate, and concerns with its potentially harmful consequences for democracy have spurred a body of research, particularly in the US (e.g., Jones 2015; Harbridge and Malhotra 2011; Ramirez 2009). The assumption underlying many of these studies is that polarized elites can erode political support, which ultimately might decrease the quality of representative democracy. These studies can, however, be juxtaposed by research demonstrating positive effects of elite polarization for political engagement, participation, and efficacy (Moral 2017; Hetherington 2008; Kittilson and Anderson 2011). Apparently, polarization conveys to citizens perceptions of both conflict and choice, and it might well depend on the type of polarization which perception dominates.

The present study contributes to three bodies of research. First, we add to the literature on the consequences of polarization by distinguishing different dimensions of perceived elite conflict in order to accurately gauge their impact on political attitudes. In line with recent experimental evidence (Skytte 2021), we find that perceived ideological polarization can have very different—and even opposing—effects as compared to a polarizing tone and style (elite incivility). In the Dutch context, we find that these are two uncorrelated dimensions, signifying that people distinguish animosity in political debates from substantive policy disagreement. Where the first conveys conflict to citizens and depresses political support, perceptions of left–right polarization and economic issue polarization actually have a positive effect. For the cultural issue of integration, we found no such effect. This lack of a positive effect might be explained by the fact that cultural issues are associated strongly to populist rhetoric and are possibly viewed as less reconcilable—and positional differences thus as less positive. In any case, the results show that substantive disagreement in itself does not harm political support. These findings specify the idea behind “stealth democracy”: overt political disagreement may also have positive effects on political support.

Second, our study contributes to the literature on political support and legitimacy by underlining the importance of citizens’ process evaluations and refining our knowledge of the mechanisms through which such evaluations come about. First, evaluations of elite incivility particularly relate to the question whether politicians “care” about the interests of their constituents (as opposed to serving their own interest), which has been defined as a core requirement for political trust (Van der Meer and Hakhverdian 2017). This relates to the integrity of political actors and therefore primarily affects throughput legitimacy (Schmidt 2013). Second, evaluations of ideological polarization strengthen input legitimacy by increasing the perception that there is a variety of policy offer, thereby enhancing the representative function of the political system (Scharpf 1999). This latter effect was only found for satisfaction with democracy, not for political trust. This suggests it mostly plays a role when citizens explicitly evaluate democratic procedures as such.

Third, our study contributes observational evidence to the literature on political incivility, which has mostly used experimental set-ups to study its impact on political attitudes. This has the advantage of showing the impact of perceived incivility in a real-world setting, and allows us to shed light on the impact of generalized incivility perceptions (Funk 2001) rather than on single (manipulated) expressions of incivility and their instantaneous effect.

In terms of limitations, the obvious disadvantage is that observational research designs will never be fully watertight against endogeneity. In particular, certain voter groups (e.g., electoral winners) may generally have a more positive outlook on politics; the models therefore control for ideological position, extremity, as well as winner/loser status. The more politically sophisticated might view more ideological differences to begin with (and be more supportive of politics). We control for this by including level of education, media use and political interest, and found that these variables play no notable confounding role. Furthermore, the items used to measure stealth democratic preferences are not ideal. Even if highly similar to the items originally used by Hibbing and Theiss-Morse (2002), they capture the aversion of political conflict rather indirectly (Webb 2013) and in part reflect a more general negativity toward politics. This may explain why we fail to find significant moderation effects.

By design, this study focuses on perceptions of polarization rather than testing the effects of actual polarized party stances or behavior (e.g., by inferring polarization through party manifesto’s or roll call votes). This means that strictly we do not know whether the found effects originate from actual party polarization, or only from voters’ perceptions of such polarization. In the literature on polarization and trust, both actual and perceived polarization are used as explanatory variables, but the effects of the two are seldom compared. Where several US studies use roll call votes to operationalize actual polarization (Jones 2015; Ramirez 2009), others have focused on voter perceptions of polarization (Lupu 2015; Hetherington 2008). Only the study by Moral (2017) on political participation studies actual and perceived polarization in tandem in relation to turnout and finds that both actual and perceived party polarization have a positive effect. Future research could do more to disentangle the effects of these two essentially different ways of measuring polarization in the context of political support.

Finally, the Dutch case enables us to assess the impact of polarization on political support in a multi-party setting with a political space structured by (at least) two dimensions of conflict. Certain particularities of the Dutch case may, however, condition some of the findings. First, with respect to incivility, the campaigns in the run-up to the 2021 elections were not particularly harsh. Instead, with government policy on Covid-19 dominating the headlines, parties might even have refrained from attacking the incumbent parties responsible for these policies. Second, with respect to ideological polarization, we have already alluded to the fact that Covid overshadowed the debate on other substantive issues—which may have lowered the degree of ideological differentiation visible to voters in these particular elections. As a consequence, the circumstances might have been conducive to finding support for the “polarization as choice” hypothesis: in a context where choice is scarce, any perception of choice is likely received benevolently.

In conclusion, not all manifestations of elite conflict should be viewed with apprehension. Polarization can convey both perceptions of conflict and choice to citizens. Especially in political contexts characterized by positional convergence of mainstream parties, ideological polarization should be welcomed rather than feared. An important caveat, however, is that such polarization is coupled to respectful and civil political interaction, as our study confirmed that perceived elite incivility in itself has harmful effects on political support. In any case, even if incivility and ideological polarization are often conflated in the public and scholarly debates, our study shows that citizens are able to keep these dimensions separate in their judgments.