The Democracy I Like: Perceptions of Democracy and Opposition to Democratic Backsliding

Abstract

When democracy is under attack, the hope is often that citizens will punish undemocratic incumbents. However, recent studies show that not all citizens punish governments for their undemocratic actions. In this article, we argue that citizens' understanding of and satisfaction with democracy are sources of heterogeneous reactions. In a survey experiment conducted in Germany and Poland, we show that the importance that citizens attach to specific institutions under threat, as well as their understanding of democracy, can explain much of the variance in citizens' responses to undemocratic actions. Citizens are willing to defend what they consider important for democracy – regardless of whether this reflects theoretical conceptions of democracy. Moreover, in times of democratic backsliding, Polish ‘critical citizens’, those who are dissatisfied with the way democracy works in Poland, are more likely to punish governments for undemocratic actions. Our findings help us understand how to increase citizens' resilience against democratic backsliding.

When democracy is under attack, the hope is often that citizens punish undemocratic leaders at the ballot box or pressure governments through protests. After all, citizens of liberal democracies around the world by and large state that they support democracy (Dalton et al. 2007; Inglehart 2003; Lagos 2003) and they are good judges of democratic quality (Gómez and Palacios 2016). However, recent studies have also shown that citizens are often willing to trade their preferred policies for the democratic commitment of candidates, especially in contexts of high polarization (Carey et al. 2022; Fossati et al. 2022; Gidengil et al. 2022; Graham and Svolik 2020; Svolik 2020; Touchton et al. 2023). These studies have made major contributions to our understanding of the damaging effects of polarization on the likelihood that citizens will punish undemocratic actions by their co-partisans.

However, if we expect citizens to punish undemocratic behaviour, they must perceive undemocratic actions as such. But citizens can differ widely in their understanding of democracy, despite their shared declarations of support for democracy. In fact, many of them are highly dissatisfied with how democracy works in practice (Collier and Levitsky 1997; Ferrín and Kriesi 2016; Gómez and Palacios 2016; Landwehr and Steiner 2017; Ulbricht 2018). In some countries, a significant percentage of citizens have an understanding of what democracy means that does not reflect liberal democratic ideas (Chapman et al. 2024; Kirsch and Welzel 2019). Studies that focus on partisan bias in punishing governments for democratic backsliding have often found different effects for different parties – for example, supporters of US Republicans and Democrats (Albertus and Grossman 2021; Carey et al. 2019; Chapman et al. 2024; Gidengil et al. 2022; Touchton et al. 2023) – but have otherwise ignored potential sources of heterogeneous treatment effects for citizens with different perceptions of democracy.

A new line of argument recognizes that citizens may perceive democratic backsliding differently because they have different democratic and political attitudes (Ahmed 2023; Grossman et al. 2022; Lewandowsky and Jankowski 2023). These studies call for a closer look at the relevance of citizens' understandings of democracy for their tolerance of democratic backsliding (Ahmed 2023), and show that a majoritarian understanding of democracy, as well as authoritarianism, populism and satisfaction with democracy, inter alia, increase the likelihood of tolerating undemocratic actions (Grossman et al. 2022; Lewandowsky and Jankowski 2023; Saikkonen and Christensen 2023; Svolik et al. 2023).

We contribute to this second line of argument by presenting results from a survey experiment conducted in Germany and Poland that confronts citizens with infringements on three different notions of democracy: electoral, liberal and majoritarian democracy. Building on previous studies, we examine citizens' reactions to potential reforms that would not violate liberal, but majoritarian notions of democracy (as well as those that violate liberal but not majoritarian notions) and provide novel survey experimental evidence from a country (Poland) that is greatly affected by democratic backsliding and that has received less consideration in this literature. Moreover, we use outcome measures beyond vote choice, as citizens in established democracies can express their discontent in many ways, addressing one weakness of vote-centred candidate choice conjoint experiments.

Our respondents differ in their perceptions of which institutions are important for democracy. Majoritarian understandings of democracy are not prevalent among our respondents. Instead, we replicate the findings of others (Kriesi et al. 2016; Quaranta 2018a, 2018b), showing that our respondents primarily differ in the extent to which they support liberal democracy. These differences explain much of the variance in respondents' reactions to undemocratic actions. Moreover, respondents punish attempts to undermine institutions of liberal and electoral democracy more than attempts to undermine majoritarian democracy. When examining satisfaction with democracy, we find it does not correlate with German respondents' reactions to undemocratic behaviour, whereas Polish dissatisfied respondents are most willing to punish governments for their undemocratic actions. This supports the idea that ‘critical citizens’ are important opponents of democratic backsliding in countries with undemocratic governments (see also Geissel 2008).

Concluding, this article shows that the assumption of uniform and informed perceptions of democracy among citizens leads researchers to overlook meaningful differences in citizens' willingness to defend democracy. Citizens' perceptions of democracy are an important source of variation in reactions to government proposals that subvert democracy. To the best of our knowledge, we are the first to systematically test this outside the US. Based on our divergent findings, we encourage researchers to take citizens' perceptions of democracy into consideration when studying citizens' tolerance for democratic backsliding. Moreover, depending on citizens' experiences with democratic backsliding in their countries, satisfaction with how democracy works might measure different things. While Inga Saikkonen and Henrik Christensen (2023) have found a negative correlation between satisfaction with democracy and tolerance of democratic backsliding in Finland, we find no correlation for Germany and a positive one for Poland. Finally, not all citizens attach great importance to liberal democratic institutions. Some of them clearly do not like them and therefore do not punish infringements on them. These citizens may also trade democracy for policy and vote for illiberal or undemocratic governments in times of polarization. Convincing citizens of liberal democracy as a concept may thus encourage them to stand up for its institutions more often.

Perceptions of democratic backsliding

Which institutions define a democratic system, and therefore which institutional reforms constitute democratic backsliding, is a difficult question to answer. Democratic backsliding has been defined as ‘the state-led debilitation or elimination of any of the political institutions that sustain an existing democracy’ (Bermeo 2016: 5). Since ‘democracy’ is an ‘essentially contested concept’ (Collier et al. 2006; Gallie 1955: 134–137), citizens and experts alike have different understandings of democracy (Carey et al. 2022; Collier and Levitsky 1997; Ulbricht 2018) and differ in their evaluations of existing democratic systems (Gómez and Palacios 2016). Moreover, some of the ideals inherent in ‘democracy’ conflict with each other, such as horizontal accountability and minority protection with the implementation of the will of the majority. Existing democratic systems differ in their institutional design (Lijphart 1999) and are oftentimes not ‘able to do justice to the normative ideal of democracy with respect to all dimensions at one and the same time’ (Ferrín and Kriesi 2016: 7). That is, democratic systems in themselves represent trade-offs between their ideals.

The diversity of conceptions of democracy is also reflected in experimental research on citizens' reactions to democratic backsliding. Previous experiments have examined, for example, how citizens react to illegitimate judicial appointments (Albertus and Grossman 2021; Grossman et al. 2022), restrictions of minority and women's rights (Fossati et al. 2022), investigations into potential misbehaviour of politicians (Carey et al. 2022), the implementation of representative rather than direct local elections (Fossati et al. 2022), the approval of political violence (Saikkonen and Christensen 2023) or gerrymandering and the closing of polling stations (Graham and Svolik 2020). While these studies all look at violations of ‘democracy’, this focus on the broad concept hides variation in the actual conceptions of democracy studied.

With the exception of Guy Grossman et al. (2022), Marcel Lewandowsky and Michael Jankowski (2023) and Saikkonen and Christensen (2023), these studies have not explicitly theorized about their concept of democracy and its congruence with citizens' understandings (Ahmed 2023). This is particularly problematic because these studies have used vignette or conjoint experiments that yield averages across a highly diverse sample. These averages may be driven by respondents with strong preferences regarding the democratic institutions under attack. We are explicitly interested in potential heterogeneous treatment effects for respondents with different attitudes towards democracy. This allows us to unravel an additional potential mechanism that explains why some citizens tolerate the undermining of democracy, although they also claim to support democracy: some citizens may simply not believe that the institutions under attack contribute to a vibrant democratic system.

Differences in citizens' perceptions of democratic backsliding could manifest themselves in at least three ways. First and foremost, citizens value different democratic institutions to varying degrees (Ferrín and Kriesi 2016; Landwehr and Steiner 2017; Ulbricht 2018). These institutions can be formal institutions enshrined in constitutions, such as the independence of courts, but also informal institutions, such as equality or freedom. Consequently, the willingness of citizens to give up democratic institutions for their preferred policies may partially depend on the importance they attach to the democratic institutions (see Lewandowsky and Jankowski 2023 for a similar argument).

Second, not all citizens have a liberal understanding of democracy (Grossman et al. 2022; Malka et al. 2022). Even in countries with consolidated liberal democratic systems, some citizens may be open to authoritarianism (Lewandowsky and Jankowski 2023; Malka et al. 2022; Svolik et al. 2023). Other citizens have a majoritarian understanding of democracy that violates liberal democratic elements such as the independence of courts (Grossman et al. 2022). Governments, at least in authoritarian systems, use these different understandings of democracy to portray their systems as more democratic than experts believe them to be (Brunkert 2022). Hence, beyond support for very specific institutions, it is important to consider what understanding of democracy citizens have in the first place.

Third, satisfaction with how democracy works in one's country may affect perceptions of democratic backsliding. Citizens who are dissatisfied with how democracy works in their countries may be more likely to tolerate democratic backsliding – even if they support liberal democracy as an ideal political system – since they do not consider the threatened institutions to be worthy of their protection (see, for a similar argument, Easton 1975; Norris 2011; Saikkonen and Christensen 2023). Alternatively, ‘critical citizens’, or the dissatisfied, may be the true defenders of democratic institutions (Geissel 2008).

We consider the perceived importance of various specific institutions of democracy, the relationship of citizens' understandings of democracy to electoral, majoritarian and liberal democracy and their satisfaction with how democracy works as three different manifestations of citizens' democratic attitudes. We do not address the origins of these attitudes. However, we often witness how authoritarian governments frame their undemocratic actions as reforms that would improve the overall quality of the democratic system, or move existing democratic institutions closer to the ‘right’ form of democracy. If governments succeed in influencing what citizens believe to be democratic, for example, because voters take cues about the state of democracy from their preferred parties (Berlinski et al. 2023; Fossati et al. 2022; Gidengil et al. 2022; van der Brug et al. 2021), the mechanism that we study may strongly contribute to explaining why citizens claim to support democracy and yet tolerate undemocratic actions.

Perceived importance

Despite showing overwhelming support for core democratic principles such as free and fair elections, citizens in liberal democracies differ in the importance they attach to specific democratic institutions. Figure 1 shows the distribution of the importance for democracy attached to a number of democratic institutions by German and Polish respondents to the European Social Survey (ESS) waves 6 (ESS ERIC 2018) and 10 (ESS ERIC 2023). Clearly, most citizens believe in the core tenets of liberal democracy. However, there is considerable variation in the importance citizens attach to each of these democratic institutions. For example, in 2012, only 66.73% of German and Polish citizens assign an importance of 10 to free and fair elections, and still 11.88% assign an importance of 7 or lower to it. In 2020, the proportion of respondents who assign an importance of 10 to free and fair elections rises to 72.69%, but the proportion of respondents who assign an importance of 7 or lower also increases to about 13.22%.

Figure 1. Importance for Democracy Attributed to Institutions in Germany and Poland Based on Data from the ESS.

Note: Figure created with the R package ggplot2.

Consequently, citizens may differ in their perceptions of the severity of undemocratic policies. When citizens are confronted with a government plan that undermines institutions they do not consider most important for democracy, they might tolerate the plan and take no action against it. For example, citizens who assign an importance of 10 to the power of courts to control the government might be most likely to punish violations of the courts' power. Vice versa, citizens who attach an importance lower than 10 to the power of courts may not perceive the violation as a dangerous attempt at democratic backsliding and may be more likely to tolerate such violations, regardless of any policy gains. Hence, the perception of democratic backsliding depends on which institutions or norms citizens consider as important.

Hypothesis 1 therefore proposes that citizens are, on average, less likely to tolerate infringements on institutions they consider important for democracy. This should hold regardless of whether these institutions are part of any theoretical concept of democracy, including institutions that most political scientists would not consider important for (liberal) democracy.

Hypothesis 1: The more important citizens consider specific institutions for democracy, the more likely they are to oppose democratic backsliding that infringes on those institutions. ¹

Electoral, liberal and majoritarian democracy

Research also suggests that most citizens of liberal democracies have an understanding of democracy that is consistent with minimal theoretical concepts of democracy (Dahl 1971; Ferrín and Kriesi 2016), most notably the definition of democracy as a political system that guarantees free and fair elections, freedom of speech and opinion, and basic rights such as the right to demonstrate (Dahl 1971; Ferrín and Kriesi 2016). However, elements of liberal democracy, such as the protection of minority rights, checks and balances, and the rule of law (Coppedge et al. 2018; Diamond 1999: 1–19) enjoy less support among citizens of liberal democracies (Grossman et al. 2022; Heinisch and Wegscheider 2020; Ulbricht 2018).

Simultaneously, research has highlighted the importance of illiberal, populist or majoritarian understandings of democracy (Grossman et al. 2022; Heinisch and Wegscheider 2020; Landwehr and Steiner 2017). Democracy, if understood primarily as the rule of the people rather than an elite minority, is by definition closely tied to majority rule (Mudde and Rovira Kaltwasser 2012: 10–11). According to Robert Dahl (1956: 37), it is the ‘proposition that majorities should have unlimited sovereignty’ that defines democracy for some: ‘Democracy requires majority rule in the … sense that support by a majority ought to be necessary to passing a law’ (Dahl 1989: 135). Within this majoritarian vision of democracy, governments are basically unconstrained, responsible only to the people (Galston 2018; Huber and Powell 1994; see also Dahl 1989: 36 for a similar argument).

Defined as such, liberal democracy and majoritarian democracy are antagonistic concepts (Dahl 1989; Diamond 1999; Held 2006; Mény and Surel 2002; Mudde and Rovira Kaltwasser 2012), both based on a minimal electoral definition of democracy. The institutions of liberal democracy – checks and balances, the rule of law and the protection of minority rights – all constrain governments in implementing the will of the majority. Moreover, proponents of liberal democracy point to the potential dangers of majority rule, even when a decision based on majority rule meets all the criteria listed for the minimal electoral conception of democracy. Some invoke the tyranny of the majority (Held 2006), others support the idea of inalienable rights that even a majority cannot revoke (Dahl 1989).

While other conceptions of democracy exist (e.g. social democracy), they do not meet this antagonistic criterion. Hence, we focus here on electoral, liberal and majoritarian understandings of democracy. We consider electoral democracy to be the minimal definition of any form of democracy, and liberal and majoritarian democracy more encompassing concepts of democracy. Thus, any majoritarian or liberal understanding of democracy must include elements of electoral democracy, but not vice versa.

Grossman et al. (2022) and Lewandowsky and Jankowski (2023) provide initial evidence that, in the US, a majoritarian understanding of democracy and populism is negatively related to citizens' willingness to punish governments for democratic backsliding. Building on this, we argue that citizens with a majoritarian understanding should dislike reforms that restrict majoritarian democracy, but tolerate or even like reforms that restrict liberal democracy, because they prefer a majoritarian system to a liberal one. The reverse should be the case for liberal democrats. If this mechanism is relevant and applicable to countries other than the US, we would expect that:

Hypothesis 2: Citizens with a liberal understanding of democracy are more likely to acquiesce to restrictions of majoritarian democracy than to acquiesce to any other restrictions of democracy.

Hypothesis 3: Citizens with a majoritarian understanding of democracy are more likely to acquiesce to restrictions of liberal democracy than to acquiesce to any other restrictions of democracy.

Another strand of literature studies the structure of Europeans' understandings of democracy using various types of factor analysis. Using data from the ESS wave 6, Mario Quaranta (2018a, 2018b) has argued that Europeans' understandings and evaluations of democracy are based on a continuum between more and less encompassing understandings, but do not reflect understandings other than liberal ones. Similarly, Hanspeter Kriesi et al. (2016) have argued that Europeans share a liberal understanding of democracy, but that they differ in the way in which they incorporate two additional dimensions of democracy into their understanding: social and direct democracy. Using the same underlying data, these studies suggest that Europeans mostly differ in the extent to which they support liberal democracy as opposed to less or no democracy: ‘Most citizens do not adopt a single theoretical model of democracy, but hold views which are compatible with several models at one and the same time’ (Ferrín and Kriesi 2016: 11).

None of these studies has found evidence of a dominant majoritarian understanding of democracy among Europeans, although it must be acknowledged that the items included in the ESS were not designed to capture such an understanding. Europeans have therefore not been systematically asked about majoritarian visions of democracy in surveys, yet. Although we pre-registered the hypotheses above based on the assumption that we would find evidence of a majoritarian understanding of democracy, our analysis below does not provide evidence for the majoritarian or illiberal understandings found and described in previous studies (Grossman et al. 2022; Heinisch and Wegscheider 2020; Lewandowsky and Jankowski 2023; Mény and Surel 2002; Mudde and Rovira Kaltwasser 2012).²

Satisfaction with democracy

Researchers have attributed great importance to citizens' satisfaction with democracy, hypothesizing that dissatisfied citizens will lose their faith in democratic institutions and stop defending them against undemocratic alternatives (Almond and Verba 1963; Easton 1975; Norris 2011). Consequently, an extensive literature has been published on the origins of dissatisfaction with democracy and how to remedy it (see for example Aarts and Thomassen 2008; Blais and Gélineau 2007; Hobolt 2012). However, we are not aware of any study that shows a relationship between citizens' satisfaction with democracy and their support for undemocratic policies in a conjoint setting. The exception is Saikkonen and Christensen (2023), who show that in Finland satisfied respondents are less likely to tolerate infringements on democracy. Beyond this study, it remains unclear whether satisfaction with democracy empirically stabilizes democracies.

We assume that dissatisfied citizens are less likely to support existing democratic institutions, and hence less likely to punish their subversion. This assumption about dissatisfaction is consistent with arguments about detachment and is supported by much of the existing literature. For example, Pippa Norris (2011) shows that dissatisfied respondents tend to participate less in politics and think more often that tax evasion is acceptable. Similarly, David Easton (1975) argues that dissatisfaction with the functioning of a political system will lead to its collapse in the long run. More recently, Saikkonen and Christensen (2023) have shown a positive relationship between Finns' support for and satisfaction with democracy. We therefore hypothesize:

Hypothesis 4: The more dissatisfied citizens are with how democracy works, the less likely they are to punish infringements on democracy. ³

A contrasting argument – linked to the concept of ‘critical citizens’ – is that citizens who are dissatisfied with how democracy works may be more responsive than others to actual declines in the quality of existing democratic institutions (Norris 2011), and thus more likely to defend them (Geissel 2008). In this vein, Matthew Singer (2018) has shown that election winners are more likely to support democracy, but also more likely to tolerate executive power grabs. Although we did not pre-register an according hypothesis, our study design allows us to test both arguments against each other.

Data and survey design

We collected novel survey data from 638 German and 651 Polish citizens who are representative of the national population in terms of age and gender (cross-quotas applied). Data collection took place between 27 November and 14 December 2020. Access to the panel was provided by Lucid. The survey received ethical approval from the Ethical Review Board of the Faculty of Management, Economics, and Social Sciences of the University of Cologne. The study has been pre-registered via the Open Science Framework (OSF).⁴ The survey includes basic demographic and political questions as well as a vignette experiment that is described in detail below. We excluded respondents who failed the attention check from the analysis.⁵ The data are available via the OSF webpage.⁶

Studying Germany and Poland allows us to test our argument in two distinct settings. Both countries were liberal democracies between the 1990s and the early 2010s, although (West) Germany has a longer democratic history from World War II. In Poland, the Law and Justice (PiS) government considerably threatened the stability of liberal democracy thereafter (Grzymala-Busse 2019), moving the political system closer to a form of majoritarian rule of the executive over the legislature (Nalepa 2016) and undermining the independence of the judiciary (Śledzińska-Simon 2018). In contrast, the German political system has remained stable. Thus, we vary the extent to which citizens have been exposed to government attempts to restrict important (liberal) democratic institutions, but also the extent to which they might have been exposed to illiberal visions of democracy. Through socialization and learning, this may affect the behaviour of our respondents.

Experimental design

Similar to previous studies, we examine citizens' tolerance for democratic backsliding using a vignette experiment (Hainmueller et al. 2014). While previous studies have mostly examined choices between candidates (Carey et al. 2019; Gidengil et al. 2022; Graham and Svolik 2020; Lewandowsky and Jankowski 2023; Svolik 2020), we aim to reproduce information that citizens would receive, for example through the media, about potential government plans to undermine different institutions that citizens might consider important for their understanding of democracy. Thus, our vignettes reflect situations in which citizens are confronted with infringements on democracy, including surrounding pro- and counter-arguments. After each vignette, we ask respondents how likely they would be to: (1) participate in a demonstration against the government's plan; (2) defend the government's plan in discussions with acquaintances; and (3) vote for the government again. This selection of outcome measures reflects various possible responses to democratic backsliding, with varying costs for the respondents and different effects on governments. Respondents could choose from a value ranging between 1 (highly unlikely) and 7 (highly likely).

Since previous studies have already shown that policy congruence and government support matter (Carey et al. 2019; Fossati et al. 2022; Gidengil et al. 2022; Graham and Svolik 2020; Svolik 2020; Touchton et al. 2023; van der Brug et al. 2021), we fix both by introducing the government that subverts democracy as the government that respondents have voted for. We then mention one out of three hypothetical government plans. Banning demonstrations against government plans would violate any form of democracy, even a purely electoral democracy. Limiting the power of courts to overturn unconstitutional government decisions would undermine the separation of powers in liberal democracies, but does not necessarily violate majoritarian democracy. In contrast, implementing reforms that lack majority support in the population undermines majoritarian democracy, but not necessarily liberal democracy. Respondents have seen each proposal once in random order to avoid situations in which a description of the proposal in a previous vignette affects the reaction to it in the subsequent vignette (Hainmueller et al. 2014: 8). Table 1 shows the wording of the attribute levels.

Table 1. Attribute Levels for Different Institutions Under Threat

A sample vignette is shown in Box 1. The names of the attributes in parentheses were not shown to the respondents. The text in bold was randomized but was not shown in bold to the respondents. In addition, the text in italics was used as a prime for half of the respondents but was not shown in italics to the respondent. To increase the external validity of the results, we also confronted respondents with a reason for the government proposal (goal of plan), with potential criticism against it (criticism) and with a random actor criticizing the proposal (actor-opposing plan).⁷

Box 1. Example Text for the Experiment

Note: The German master version and the Polish translations are shown in the Supplementary Material, Section B of the Appendix.

We use marginal means (MMs) to analyse the effect of seeing different types of backsliding on the outcome measures across different subgroups of respondents (Leeper et al. 2020) and standard levels of significance (α ≥ 0.05). ‘Standard’ vignette experiments assume homogeneous treatment effects across different types of respondents (Leeper et al. 2020). Since we argue that citizens with different understandings of democracy should react differently to different institutions under threat, we focus on subgroups of respondents and assume homogeneous treatment effects only within these subgroups (Leeper et al. 2020).

Measuring respondents' understandings of democracy

Before the experiment, we showed respondents a battery of statements about formal and informal institutions and asked them how important these were for democracy. Table 2 shows the wording of the questions and the statements. For most of them, we follow the wording implemented in the ESS 2012 (ESS ERIC 2018; Ferrín and Kriesi 2016), but we made sure to include two statements each that reflect a minimal form of democracy, majoritarian democracy and liberal democracy. The free participation in demonstrations and the freedom of opposition parties and the media are essential elements of any conception of democracy. We code them as part of the minimal electoral understanding of democracy. The people having the final say and the government implementing the will of the majority (worded negatively) are crucial elements of a majoritarian understanding of democracy, but a democracy without these elements could still be a liberal democracy. That the courts are able to stop the government from acting beyond its authority (worded positively and negatively) and the protection of minority rights are defining elements of liberal democracy, but not defining elements of minimal electoral democracy.⁸ We furthermore include one statement about social democracy, assuming that it represents a fourth and unrelated conception of democracy. We finally measure the importance for democracy that citizens attribute to each of these institutions on a scale from 1 (not at all important) to 7 (very important).

Table 2. List of Institutions of Democracy and the Related Concept of Democracy

We assume that citizens with a majoritarian understanding of democracy rate all majoritarian institutions and, vice versa, that citizens with a liberal understanding of democracy rate all liberal institutions as more important than 4, the midpoint of the scale, but we also consider other thresholds in the Supplementary Material, Section C of the Appendix. Electoral institutions are considered essential to both majoritarian and liberal understandings of democracy. Some of the items might not be related to theoretically grounded understandings of democracy, but we wanted to maximize the variance and also include items that citizens, but not researchers, would consider important for ‘democracy’. Moreover, some of these items are closely related to the government proposals shown in Table 1. By embedding the institution in the vignette experiment, it is given an additional context that may influence the way in which respondents react to it. In addition, the outcome questions are worded differently, aiming to reveal the respondents' potential for action rather than asking for their assessment of the importance of these institutions in the abstract. This design allows us to assess whether and to what extent the self-proclaimed importance of specific institutions translates into a punishment of governments that propose restrictions on these institutions while accounting for a set of context factors through randomization. The latter questions are thus designed to more closely resemble decisions that citizens must take when faced with undemocratic actions by their governments.

Results

Figure 2 shows MMs for all government proposals and outcome measures. On average, citizens are more likely to punish governments for undermining electoral democracy (bans on demonstrations) and liberal democracy (restrictions on courts) than for undermining majoritarian democracy (implementing reforms against the will of the majority). Some of the differences between the MMs are considerably large. We find the largest difference between voting intentions for governments that undermine majoritarian democracy and governments that interfere with electoral democracy. The absolute difference here is 0.47 (p < 0.0001) on a scale from 1 to 7 (see Supplementary Material, Appendix, Section E). Thus, on average, citizens would punish violations of liberal and electoral democracy more than violations of majoritarian democracy. However, these results marginalize across all respondents in our data set. The following subsections show that there are pre-treatment differences across respondents that strongly correlate with differences in citizens' reactions to different types of democratic backsliding.

Figure 2. Marginal Means (MMs) for Seeing Infringements on Different Democratic Institutions, Shown Separately for Each Outcome Question (OQ).

Notes: Figure includes confidence intervals too small to be detected by eye. MMs are based on regressions of the outcome measures on the appearance of attribute levels with clustered standard errors on the respondent level. We added a vertical line indicating the scale midpoint for readability, but are hesitant to interpret the scale midpoint. Figure generated using the R packages cregg and ggplot2.

The importance of democratic institutions to citizens

We first argued that citizens attach different degrees of importance to different types of democratic institutions. Figure 3 shows that we can replicate the results from the ESS (Ferrín and Kriesi 2016) depicted in Figure 1 with our own data. Again, most respondents attribute high importance to all of the institutions under study, except for the clearly anti-democratic or anti-liberal ones: banning demonstrations against government plans and limiting the power of courts to overturn unconstitutional government decisions. However, the variation in importance attached to each of the institutions across respondents is large. For example, only 65.01% of the respondents assign an importance of 6 or higher to the ability to demonstrate against government plans.

Figure 3. The Importance of Democratic Institutions to Citizens.

Notes: German and Polish citizens rate the different institutions similarly (see Supplementary Material, Appendix, Section G.1). Figure generated using the R package ggplot2.

As Figure 4 shows, this variation in the importance attached to different institutions helps us understand citizens' reactions to undemocratic actions. Clearly, respondents who assign an importance of 6 or 7 to the democratic institution under threat are those who are by far most likely to punish governments for undermining these institutions, while those who attach an importance of 5 or lower to them do not differ substantially. The differences are substantial. The scores for respondents attaching the highest importance to the institution under threat are 1.31 times higher for demonstrations against the government proposal, 0.65 times higher for voting again for the government and 0.68 times lower for defending the proposal among acquaintances than scores for respondents attaching an importance of 5 to the institution under threat. These differences are significantly greater than zero. Overall, this strongly supports Hypothesis 1.

Figure 4. MMs for Seeing Infringements on Democracy by Importance Attributed to the Democratic Institutions under Threat, Shown Separately for Each Outcome Question (OQ).

Notes: MMs are based on regressions of the outcome measures on the importance that respondents attribute to the type of democracy under attack by the government with clustered standard errors for respondents. We added a vertical line indicating the scale midpoint for readability, but are hesitant to interpret the scale midpoint. Figure generated using with the R packages cregg and ggplot2.

Citizens' understandings of democracy

Contrary to our assumptions about the prevalence of liberal and majoritarian democrats, and contrary to what Grossman et al. (2022) and Lewandowsky and Jankowski (2023) have found for the US, only a few German and Polish respondents are either pure liberal democrats or pure majoritarian democrats (see Supplementary Material, Appendix, Section C). By far, most respondents support democratic institutions, but do not consistently attach a high importance to all liberal and no majoritarian institutions or vice versa. We are therefore unable to test the pre-registered Hypotheses 2 and 3 in the planned form and proceed with an exploratory factor analysis of conceptions of democracy.⁹

In fact, the responses to the question about the importance of different democratic institutions load onto three different factors. In the main analysis, we focus on two of these factors.¹⁰ The first factor is clearly shaped by positive responses to electoral and liberal democratic institutions, including the ability of the courts to stop the government acting beyond its authority, citizens being allowed to participate in demonstrations against government plans, the protection of minority rights and the freedom of the opposition and the media to criticize the government. We call this the ‘liberal democratic factor’. The second factor consists of positive reactions to the government implementing reforms for which there is no majority support in the population, limits on the power of courts to overturn unconstitutional government decisions and the government's ability to ban demonstrations against government plans. We call this the ‘authoritarian factor’ because it emphasizes strong and unrestricted governments not even responsible to the will of the majority. This suggests that majoritarian understandings of democracy are irrelevant for Europeans. Rather, we replicate the findings that Europeans differ in the extent to which they support liberal (and other) democratic institutions, but do not differ substantially in the type of democracy they support (see Kriesi et al. (2016) and Quaranta (2018a, 2018b)). But do the liberal democratic and authoritarian factors explain variation in citizens' reaction to the three undemocratic government proposals in different ways?

Table 3 shows linear regression results of the three different outcomes in our experiment on respondents' scores on the liberal democratic and authoritarian factors, including country and vignette fixed effects. We separately examine responses to vignettes undermining liberal, electoral or majoritarian democracy. The higher respondents' score on the liberal democratic factor, the more likely they are to punish governments for any of the proposals. They are more likely to report that they would demonstrate against any of the proposals, and less likely to report that they would defend the proposal among acquaintances or vote again for the government. The β-coefficients are smaller for proposals attacking majoritarian democracy than for proposals attacking liberal or electoral democracy. In contrast, the higher respondents score on the authoritarian factor, the less likely they are to punish governments for any of the proposals. β-coefficients are again smaller for proposals attacking majoritarian democracy than for other proposals. For most models, the R ² values are high, indicating that citizens' understandings of democracy explain between 11% and 19% of the variance in citizens' responses to undemocratic actions.¹¹

Table 3. List of Institutions of Democracy and the Related Concept of Democracy

Notes: ***p < 0.001, **p < 0.01, *p < 0.05.

Concluding, citizens have heterogeneous understandings of democracy, mapping onto two or three different dimensions. Although majoritarianism has been defined as one potential core concept of democracy, we find no evidence of such an understanding among German and Polish respondents. Rather, the factors describe authoritarians and liberal democrats, both of whom care least about reforms undermining majoritarian democracy in comparison to other proposals. Majoritarianism does not seem to be taken into account when Europeans define their understanding of democracy. Nevertheless, citizens' scores on the liberal democratic and authoritarian factors are clearly related to their tolerance for the undemocratic proposals, indicating that citizens react differently to different types of democratic backsliding depending on their understanding of democracy – although their understanding is very broadly defined. The derived factors are not only similar to those found in previous studies (Kriesi et al. 2016; Quaranta 2018a, 2018b), but are also clearly relevant for the stability of liberal democracy in Germany and Poland.

We want to reiterate that these findings are exploratory. We encourage researchers to investigate further the structure and implications of citizens' understandings of democracy worldwide as something that is neither random nor unstructured, although – in the European context – the structure does not seem to follow the division between majoritarian and liberal understandings of democracy that has been propagate by other researchers for the US case.

Satisfaction with democracy and democratic backsliding

We expected that dissatisfied citizens would be more likely to tolerate democratic backsliding, regardless of the importance they attach to the institutions under threat (Hypothesis 4). However, the results presented in Figure 5 do not indicate that citizens with higher levels of dissatisfaction are more likely to tolerate infringements on democracy. While satisfaction with democracy correlates strongly with Polish citizens' willingness to tolerate – rather than punish – democratic backsliding, it correlates only marginally with German citizens' willingness to tolerate democratic backsliding.¹²

Figure 5. Satisfaction with How Democracy Works and Tolerance for Democratic Backsliding, Shown Separately for Each Country (DE First Row, PL Second Row) and Outcome Question (OQ).

Notes: MMs are based on regressions of the outcome measures on the appearance of attribute levels with clustered standard errors for respondents. We added a vertical line indicating the scale midpoint for readability, but are hesitant to interpret the scale midpoint. Figure generated using the R packages cregg and ggplot2.

These findings strongly suggest a positive effect of having a critical citizenry that is alert to undemocratic government proposals. This positive effect is consistent with findings by Brigitte Geissel (2008), who showed that countries with many critical citizens, citizens who believe that ‘keeping watch on the government’ is important, also have higher levels of governance performance. This finding is also consistent with Singer (2018), who suggested that winners of the democratic system are most likely to support power grabs by governments. Contrasting our results, however, Saikkonen and Christensen (2023) have recently shown that in the Finnish context, it is the satisfied who sanction undemocratic behaviour. Although more research is needed to uncover the relationship between satisfaction with democracy and tolerance for democratic backsliding, the findings presented here indicate that it is the dissatisfied – and not the satisfied – who defend democratic institutions. Taken together with evidence from Finland (Saikkonen and Christensen 2023), this seems to be the case only when citizens have recently experienced significant levels of democratic backsliding, as in the case of Poland.

Moreover, our and Saikkonen and Christensen's (2023) findings indicate a lack of measurement equivalence for citizens' satisfaction with democracy between the three countries. Since the items correlate differently with citizens' tolerance of infringements on democracy, satisfaction with democracy seems to measure different things in Germany and in Poland – and possibly in Finland as well. This cautions against the use of this survey item without further testing for measurement equivalence.

Experimental assumptions and robustness

Since the attribute levels were fully randomized, the experimental design complies with the assumption of full randomization (Hainmueller et al. 2014). In addition, we find no evidence of systematic carry-over effects (Hainmueller et al. 2014: 8). We show this in Section D of the Appendix in the Supplementary Material using ANOVA and by plotting Average Marginal Component Effects across vignettes. Our results hold when analysing the first vignette only (see Supplementary Material, Appendix, Section G.2). Thus, respondents react similarly to all vignettes they see.

In addition, our results are robust and the response behaviour in our survey reflects the response behaviour of respondents in other surveys. Section G in the Appendix shows that our results are robust across countries and across the three different types of democratic backsliding, unless otherwise stated above. The response behaviour of our respondents is similar to the behaviour of respondents in the ESS waves 6 and 10 (see Figure 1). By using cross-quotas for age and gender, we were able to represent the German and Polish citizenship fairly well compared to Eurostat 2011 census data. We slightly over-sampled better-educated and older men. Levels of satisfaction in our data are comparable to levels of satisfaction in the ESS wave 10 (see Online Appendix, Section A).

Of course, experimental designs always face issues of external validity. For this reason, our vignettes are short paragraphs rather than tables describing the government plans similar to the way in which the media may describe such plans. Most importantly, in addition to randomizing the type of undemocratic action, we also show pro- and counter-arguments and list potential opponents of the government proposal. Although citizens might be confronted with a variety of frames surrounding governments' attempts at backsliding in ‘real life’, our design takes some of these frames into account, and our results are averages across the frames included (see Supplementary Material, Appendix, Sections A and H for details). Thus, we are confident that the results of our experiment are valid and robust.

Conclusion

In this article, we have shown that citizens' tolerance for democratic backsliding is strongly influenced by their perceptions of democracy. We find a strong correlation between citizens' reactions to democratic backsliding on the one hand, and the importance that citizens attach to specific institutions, as well as their general underlying conception of what democracy means on the other hand. Citizens' understandings of democracy are consequential for their reactions to democratic backsliding. The heterogeneous treatment effects we find for citizens with different perceptions of democracy are considerably large. This underlying heterogeneity has so far been ignored by most studies, but could substantially bias some of the effects found previously. For example, it may be those citizens who attach little importance to the institutions under threat who drive some of the results in previous experimental studies, especially if the importance correlates with other important factors such as party identification or polarization.

Dissatisfaction with democracy is strongly correlated with a critical stance towards democratic backsliding, but only in Poland. We believe that this is a product of the Polish context, as ‘critical citizens’ (Norris 2011) may be more protective of democracy when faced with actual attempts at democratic backsliding. In other words, satisfaction with democracy could measure different things in countries facing democratic backsliding (e.g. Poland) than in countries with a strong and stable liberal democracy (Saikkonen and Christensen 2023, e.g. Germany or Finland). Studies using this item as an indicator for the support for existing liberal democracies need to take this difference into account. Finally, we find no evidence for the destabilizing effect of citizens' dissatisfaction with how democracy works proposed by Easton (1975), Norris (2011) and others (see Saikkonen and Christensen (2023), for the counter-argument).

Since some of our findings go beyond our pre-registered hypotheses, we would like to emphasize that the theoretical arguments and results discussed in this study need further thorough testing. Most importantly, we find no evidence of a majoritarian understanding of democracy among German and Polish respondents. Future studies could consider other operationalizations, or other conceptions of democracy, and study how understandings of democracy are used to rationalize undemocratic behaviour (see Krishnarajan 2023). However, we strongly believe that our findings make an important contribution to the study of citizens' tolerance for democratic backsliding and the stability of democratic systems. First and foremost, we show that citizens differ strongly in the importance they attach to different democratic institutions, and some of them are even downright authoritarian (see also Svolik et al. 2023). These differences in democratic attitudes matter for citizens' tolerance of democratic backsliding. Furthermore, citizens who are dissatisfied with how democracy works in countries like Poland are the ones who oppose democratic backsliding. In line with arguments about ‘critical citizens’ (Geissel 2008; Norris 2011), we show that dissatisfaction with how democracy works in times of democratic backsliding can be a healthy reaction that has the potential to safeguard rather than destabilize democracies.

We believe that our findings have implications for countries other than Germany and Poland because they support more general theoretical relationships between citizens' attitudes towards democracy and their tolerance for democratic backsliding. Contrary to previous arguments that citizens simply do not care about political institutions and trade democracy for policy (Gidengil et al. 2022; Graham and Svolik 2020), we believe that many citizens do care about polity questions, but some of them prefer undemocratic institutions to democratic ones. Future studies could compare the strength of these mechanisms. We assume that both mechanisms are important for understanding how a diverse set of citizens responds to governments who infringe on democracy, and that both mechanisms potentially reinforce each other through rationalization (Krishnarajan 2023) or political cues by anti-liberal democratic governments (van der Brug et al. 2021). Our findings have furthermore important implications for policymakers who wish to stabilize democracy by encouraging citizens to protest against undemocratic actions. Not only do they need to increase the importance of democratic values to citizens, but they also need to ‘turn the head’ of authoritarian-leaning citizens.