The Demand Side of Democratic Backsliding:

How Divergent Understandings of Democracy Shape

Political Choice*
Natasha Wunsch† Marc S. Jacob‡ Laurenz Derksen§

First version: December 2021

This version: October 19, 2023

Abstract
Why do citizens in democracies fail to punish political candidates who openly violate demo-
cratic standards at the ballot box? The bulk of existing research has focused on partisan po-
larization as the driving force behind citizen tolerance of democratic backsliding. Building
on recent debates about the heterogeneity of democratic attitudes among citizens, we probe
how divergent understandings of democracy may shape citizens’ ability to recognize demo-
cratic violations as such and, in turn, affect vote choice. We leverage a novel approach to
estimate the behavioral consequences of such individual-level understandings of democracy
by means of a candidate choice conjoint experiment in Poland, a country experiencing demo-
cratic backsliding in a context of deep polarization. We find support for our claim that respon-
dents who adhere less strongly to liberal democratic norms tolerate democratic violations more
readily. Conversely, voters who subscribe more strongly to a liberal understanding are more
likely to punish non-liberal candidates, including co-partisan ones. Thus, we contend that a
lack of attitudinal consolidation around liberal democracy explains continued voter support for
authoritarian-leaning leaders.

Keywords: democratic backsliding; voting behavior; democratic commitment; Poland; con-

joint experiment
Word count: 9’782
* We would like to thank Vin Arceneaux, Sylvain Brouard, Frances Cayton, Karsten Donnay, Sarah Engler, Emil-
iano Grossman, Sven Hegewald, Nonna Mayer, Honorata Mazepus, Nicole Olszewska, Ugur Ozdemir, Frank Schim-
melfennig, Ronja Sczepanski, Paulus Wagner, and audiences at Sciences Po Paris, Yale, and APSA and MPSA con-
ferences for helpful comments and feedback on earlier versions of the manuscript. The study has been approved by
the ETH Ethics Committee (Proposal No. 2021-N-18) and preregistered (https://osf.io/f69dy/). This research was
supported by the Swiss National Science Foundation (Grant number PZ00P1 185908, PI: Natasha Wunsch).
† Senior Researcher, Center for Comparative and International Studies, ETH Zurich, Switzerland, and

Assistant Professor, Centre d’études européennes et de politique comparée, Sciences Po Paris, France,
natasha.wunsch@eup.gess.ethz.ch.
‡ PhD Candidate, Center for Comparative and International Studies, ETH Zurich, Switzerland, majacob@ethz.ch.
§ PhD Candidate, Center for Comparative and International Studies, ETH Zurich, Switzerland, lau-

renz.derksen@gess.ethz.ch.
Introduction
Democratic backsliding has become a major concern in recent years (Haggard and Kaufman 2021;
Lührmann and Lindberg 2019; Levitsky and Ziblatt 2018). Unlike democratic breakdowns, which
are often characterized by external intervention or military coups, democratic backsliding occurs
through an incremental erosion of democratic standards that may remain above the threshold to
full-fledged regime change (Waldner and Lust 2018). Such processes are generally driven by
‘executive aggrandizement’ (Bermeo 2016) or ‘incumbent takeover’ (Svolik 2015), whereby dom-
inant executives gradually dismantle domestic checks and balances and civil liberties. In elec-
toral democracies, citizens thus represent the last bulwark to resist undemocratic practices by
elected leaders (Schedler 2019). This raises the puzzle of why—despite widespread support for
democracy—citizens often fail to hold the government accountable for violations of liberal demo-
cratic principles (Svolik 2020; Aspinall et al. 2020; Fossati, Muhtadi, and Warburton 2021).
Several recent studies explore partisan-based polarization as the central explanation for au-
thoritarian support (Ahlquist et al. 2018; Orhan 2022), identifying a ‘partisan double standard’
(Graham and Svolik 2020) or ‘democratic hypocrisy’ (Simonovits, McCoy, and Littvay 2022) that
drives voters to punish democratic violations by candidates from their own party less harshly than
others. At the same time, empirical findings on the impact of partisan loyalty upon tolerance
for democratic violations are mixed (Carey et al. 2020), have shown asymmetric effects across
parties (Gidengil, Stolle, and Bergeron-Boutin 2022; Carey et al. 2019) or even no effects at all
(Broockman, Kalla, and Westwood 2023). These uneven patterns indicate that an exclusive fo-
cus upon partisan-related dynamics is insufficient to understand of citizen behavior in contexts of
democratic backsliding.
Our study argues—and demonstrates empirically—that citizens’ responses to democratic back-
sliding are shaped not by partisan considerations alone, but also by their relative commitment to
liberal democratic norms. Building on a recent revival of debates around the existence of heteroge-
neous democratic attitudes among citizens (Davis, Gaddie, and Goidel 2022; Chapman et al. 2023;
Ahmed 2022), we contend that citizens’ understandings of democracy either strengthen or mitigate
their willingness to punish political candidates who engage in democratic violations. We scrutinize
this assumption by exploring the presence and strength of divergent understandings of democracy

1
among citizens and probing how such differing understandings affect their responses to democratic
backsliding. We advance that even in reasonably consolidated democracies, alternative views of
democracy—including ones that conflict with certain fundamental liberal democratic stipulations,
such as the separation of powers and independent media—coexist and inform citizens’ evaluations
of candidates and their eventual voting decision.
We study the interplay between understandings of democracy and vote choice in Poland, a
country that represents a paradigmatic case of democratic backsliding. To assess the linkages be-
tween citizens’ understandings of democracy and their responses to democratic violations, we im-
plement a pre-registered, well-powered candidate choice conjoint experiment among a representa-
tive sample of Polish citizens.1 We leverage the novel approach of individual marginal component
effects (IMCEs) (Zhirkov 2022) to measure how variation in individual-level democratic attitudes
affects vote choice for candidates expressing differing democratic views. In methodological terms,
our study is among the first to leverage IMCE estimates to study individual-level political behav-
ior, thus providing an illustration of its added value when it comes to investigating individual-level
determinants of respondent preferences, as revealed in a conjoint experiment. We complement
this analysis with additional tests to discriminate between the two causal mechanisms—value con-
gruence and liberal democratic commitment—we posit as potential linkages between citizens’
understandings of democracy and their responses to democratic backsliding.
Analyzing democratic backsliding in a European, multi-party setting, our study contributes to
a growing debate about the ability of citizens to act as democratic bulwarks in the face of ex-
ecutive takeover. Our findings point to a considerable heterogeneity in democratic views among
Polish voters that leads parts of the electorate to overlook democratic violations at the ballot box.
Controlling for partisanship and socio-demographic covariates, we show that divergent democratic
attitudes have a significant impact upon responses to democratic violations in the Polish context:
the more voters are committed to liberal democratic norms, the more harshly they punish can-
didates who deviate from these. These findings suggest that deep-seated variation in democratic
attitudes among the citizenry plays an important role in explaining the ongoing success of illiberal
politics and the attendant deepening of democratic backsliding over several electoral cycles.
1. Our time-stamped pre-registration is available online [link to be added after review; pre-analysis plan contained
in Online Appendix for review].

2
 We begin by theorizing the potential linkages between democratic attitudes and voter responses
to democratic violations. The following section provides a brief overview of the Polish case. We
then detail our research design and in particular our measurement of divergent understandings of
democracy and their impact on candidate assessments. The empirical section presents our findings
with respect to the aggregate relationship between understandings of democracy and candidate
preferences and the individual-level patterns linking these two dimensions. We also address the
role of partisanship when it comes to voters’ responses to non-liberal candidates. The conclusion
summarizes our main insights and discusses their wider theoretical and practical implications.

Theorizing the Demand Side of Democratic Backsliding

Democratic backsliding is generally studied as an elite-driven process, whereby authoritarian-
leaning leaders actively manipulate the rules of the democratic game in their favor and secure
voters’ continued approval through buy-outs or ideological appeals (Levitsky and Ziblatt 2018;
Bartels 2023; Matovski 2021; Medzihorsky and Lindberg 2023). The supply side is certainly cru-
cial when it comes to implementing democratic violations and offering justifications for undemo-
cratic practices. However, since in democracies it is citizens who can confirm and oust politicians
from office at the ballot box, we contend that the demand side—in particular, political culture and
citizens’ views of democracy—is just as vital.
For citizens to play the role of effective safeguards against executive aggrandizement and the
resultant democratic erosion, there is an important precondition: a shared understanding that liberal
democracy is worth defending against the incumbent’s attempts to overstep the limits of govern-
ment (Weingast 1997; Saikkonen and Christensen 2023). Political culture has been cast as central
to democratic consolidation, with democratic attitudes among citizens a key determinant of regime
stability (Pridham 1995; Linz and Stepan 1996). Building on earlier seminal contributions on the
importance of mass attitudes towards democracy (Almond and Verba 1963; Easton 1975; Lipset
1959), a spate of recent studies draws on the availability of cross-national survey data on citizens’
support for democracy to confirm the relevance of political culture for democratic stability (Mauk
2020; Claassen 2020; Grossman et al. 2021; Fossati, Muhtadi, and Warburton 2021).
Yet despite an abundant literature on political culture and its broader systemic relevance, citi-

3
zens’ democratic attitudes have so far largely been overlooked as an explanatory factor in processes
of democratic backsliding. In their review of theories addressing democratic backsliding, Wald-
ner and Lust even reject political culture outright on the grounds that the same variable cannot
simultaneously account for the initial deepening and subsequent erosion of democracy (Waldner
and Lust 2018, 99). We claim that this logic is compelling only if we suppose a stable and ho-
mogeneous political culture in each country that would drive democratization in one or the other
direction. This premise stands in direct contradiction to a burgeoning literature that highlights per-
sistent divergence in citizens’ democratic attitudes (Schedler and Sarsfield 2007; Chu and Huang
2010; Carlin 2011; Canache 2012; Davis, Gaddie, and Goidel 2022) and has been singled out as
a key limitation of existing studies seeking to explain citizen behavior in contexts of democratic
backsliding (Ahmed 2022, 9).
Instead, we argue that it is precisely the coexistence of distinct democratic attitudes in a given
population that explains why individual citizens may be more or less prone to vigorously defend-
ing liberal democratic norms when faced with a real-life, multidimensional election situation. This
basic assumption informs our theoretical expectations regarding the linkages between democratic
commitment and citizens’ responses to democratic backsliding. Specifically, we argue that citi-
zens’ views of democracy shape their evaluations of competing candidates and thus affect their
electoral choice, making divergent understandings of democracy among the population a crucial
element when it comes to understanding the vulnerability of democracies to the onset and eventual
persistence of democratic backsliding.

From Understandings of Democracy to Support for Political Candidates

Studies of citizens’ responses to democratic transgressions typically expect voters to arbitrate be-
tween different elements of competing candidate profiles, potentially trading off any democratic
violations against candidates’ personal, partisan, or policy characteristics (Simonovits, McCoy,
and Littvay 2022; Graham and Svolik 2020). In doing so, these studies imply the presence of
a stable understanding of democracy among the population that fails to capture what may be a
fundamental heterogeneity of citizens’ democratic attitudes (Ahmed 2022).
Focusing precisely on this diversity of understandings of democracy within a given population,

4
our study investigates how such variation may shape candidate assessments and the resulting vote
choice at the individual level. We expect a lack of widespread commitment to liberal democracy to
represent a key vulnerability that authoritarian-leaning leaders can exploit to expand their executive
powers. At the macro level, this implies that democratic backsliding is likely to result where
divergent understandings of democracy among the citizenry meet political elites willing to exploit
such attitudes to legitimize an erosion of democratic standards. Our study examines the micro-
level foundations of this overarching argument. Specifically, we suggest that the global spread of
illiberal trends and democratic backsliding have brought systemic questions regarding the value of
democracy back into the public discourse in ways that are likely to affect citizens’ voting behavior.
There is comparatively little research on the behavioral consequences of democratic attitudes
(for two notable exceptions, see Gibson 1996; Canache 2012). Where scholars do examine this
linkage, they focus on the influence of divergent understandings of democracy upon political par-
ticipation more generally (Bakule 2020; Oser and Hooghe 2018; Cinar and Bulbul 2022; Bengtsson
and Christensen 2016) rather than on their behavior in specific electoral decisions.
So what happens when citizens hold differing views of the nature and purpose of a democratic
system? Fundamentally, we contend that divergent understandings of democracy inform citizens’
vote choice and thus, ultimately, play a crucial role in enabling the arrival in power and subsequent
democratic erosion at the hands of authoritarian-leaning elites. To probe the linkage between
understandings of democracy and individual-level candidate evaluations, we posit that political
candidates in democracies not only represent different policy preferences but may also stand for
distinct system-level preferences to which voters respond.
To assess the linkages between democratic attitudes among citizens and distinct democratic
views expressed by political candidates in an experimental setting, we focus on three distinct cate-
gories of democratic attitudes that we distinguish primarily based on the supposed source of demo-
cratic legitimacy. A liberal understanding goes beyond a general regime preference for democracy
(Wuttke, Gavras, and Schoen 2020) to embrace pluralism, executive constraints, as well as equal
rights and civil liberties for all citizens (O’Donnell 1998). In contrast, citizens holding author-
itarian views of democracy derive the legitimacy of a political system primarily from its ability
to maintain social order and prevent chaos. Earlier studies qualify such ‘authoritarian notions of
democracy’ as ‘democracy misunderstood’ (Kirsch and Welzel 2019) or ‘democracy confused’

5
(Kruse, Ravlik, and Welzel 2019) to signal their incompatibility with liberal democratic orienta-
tions. Authoritarian attitudes have been shown to exist among citizens in many democracies (Singh
and Dunn 2013), including in Europe (Vasilopoulos and Lachat 2018), Asia (Dore 2014), and Latin
America (Cohen and Smith 2016). In line with earlier research, we view liberal and authoritarian
attitudes as lying at opposite ends of a spectrum (Kirsch and Welzel 2019) that captures the depth
and exclusiveness of citizens’ commitment to liberal democratic norms.
To these two understandings, we add majoritarian attitudes as a third category to capture a dis-
tinct conception of democratic legitimacy (Allan 2017). Reflecting a populist emphasis on power
lying with ‘the people,’ voters holding majoritarian views consider decisions supported by the
political majority as democratic per se, including when they go against central precepts of lib-
eral democratic conceptions such as pluralism and minority protection (Grigoriadis 2018; Urbinati
2017). In what has been qualified as “majoritarian threat to liberal democracy” (Grossman et
al. 2021) such voters thus grant the elected government considerable leeway to limit executive
constraints or pursue critical media in an effort to implement its political program, making them
potentially more open to tolerating political leaders who undermine traditional checks and bal-
ances.
Our basic assumption is that voters will be more likely to prioritize candidates’ democratic
positions where their own commitment to liberal democracy is high. In contexts of democratic
backsliding, it is typically the liberal aspects of democracy, most notably minority rights protec-
tion and various forms of constraints upon the executive, that come under pressure. The main
conflict line thus runs between liberal and non-liberal forms of democracy rather than between
electoral vs. non-electoral regime types. Accordingly, we expect respondents who show high lev-
els of support for such liberal democratic conceptions to be particularly adamant to see the liberal
dimension of democracy protected by political candidates, and thus more prone than others to pun-
ish candidates for holding alternative views. Accordingly, we posit liberal democratic commitment
as first mechanism linking democratic attitudes to political behavior:

H1a (liberal democratic commitment hypothesis): Respondents with stronger liberal under-
standings of democracy are more likely to reward candidates expressing liberal positions and to
punish those expressing non-liberal positions.

6
 Our second mechanism focuses on the congruence between voters’ democratic attitudes and
the democratic positions expressed by political candidates. Such value congruence has been am-
ply studied to explain the linkage between citizens’ democratic values and regime type (Welzel
and Klingemann 2008; Welzel 2007, 2021) at the macro level, with citizen demand for democracy
and civil liberties thought to create pressure to adjust the supply of such freedoms by the political
system and elites (Welzel and Klingemann 2007, 2008). Zooming into the micro-level relationship
between voters’ and candidates’ democratic views, we expect an overlap between the two to drive
vote choice, with respondents preferring candidates whose positions mirror their own understand-
ing of democracy. We therefore hypothesize:

H1b (congruence hypothesis): Respondents are more likely to prefer candidates whose demo-
cratic positions are congruent with their own understanding of democracy.

In sum, we expect divergent understandings of democracy among citizens to affect their posi-
tioning in electoral contests when competing candidates express a range of democratic positions,
some of which openly conflict with liberal democratic norms. We posit two mechanisms that may
explain these linkages, namely citizens’ commitment to liberal democratic norms or the broader
congruence between citizens’ understandings of democracy and the democratic positions expressed
by political candidates. To account for partisan-related dynamics, we also assess the interplay be-
tween understandings of democracy and partisan preferences as well as the relevance of partisan
voting in explaining respondents’ assessments of competing candidates and notably their rejection
of non-liberal candidates. We expect such dynamics, where present, to act in parallel and thus in
a complementary fashion to our emphasis upon democratic attitudes and use our experimental de-
sign to probe the linkages between divergent understandings of democracy and citizens’ responses
to such concrete manifestations of democratic backsliding while controlling for partisan-related
factors.

7
Polish Democracy at a Crossroads
Most studies on citizens’ views and mass polarization in the context of democratic backsliding
have focused on the bipartisan context of the United States (Graham and Svolik 2020; Simonovits,
McCoy, and Littvay 2022; Grossman et al. 2021; Carey et al. 2019; Gidengil, Stolle, and Bergeron-
Boutin 2022). The presence of deep partisan polarization in this setting may have led scholars to
privilege partisan-based explanations of citizens’ responses to backsliding while potentially over-
looking alternative dynamics that drive voters to support (or oppose) candidates endorsing non-
liberal democratic views. Our own study focuses empirically on the case of Poland, a country sim-
ilarly characterized by a high degree of partisan polarization, but that boasts a multi-party setting.
This offers citizens a broader range of options than simply supporting or rejecting the incumbent
party representative by opening the possibility of defecting to an ideologically closer alternative
candidate.
Poland was initially hailed as an exemplar of democratic transformation, but from 2015 on-
ward has shifted toward becoming a prototype of executive aggrandizement under the Law and
Justice Party (Prawo i Sprawiedliwość, or PiS) (Buštı́ková and Guasti 2017; Bakke and Sitter
2022; Sadurski 2018; Solska 2020). PiS swiftly proceeded to remodel the judicial system and
bring public media under government control, establishing what country experts have qualified as
a “purely majoritarian democracy” (Sadurski 2018, 3) or a “ruthlessly majoritarian” government
style bent on dismantling any constraints on the executive (Fomina and Kucharczyk 2016, 58). As
of 2016, Poland was downgraded from ‘liberal’ to ‘electoral democracy’ according to the Vari-
eties of Democracy regime type indicator (Lührmann, Tannenberg, and Lindberg 2018). Freedom
House began classifying the country as a ‘semi-consolidated’ rather than a consolidated democracy
following the reelection of the PiS party in 2019 (Freedom House 2020).
At the same time, Poland has been facing deepening political and societal polarization (Tworzecki
2019; Fomina 2019). Socioeconomic cleavages map rather neatly onto partisan divides, with the
gradual emergence of “two roughly equal nationalist-populist and centrist-liberal camps” (Markowski
2016, 1316). The shared religiosity and right-wing orientation that characterized both PiS and
Civic Platform (Platforma Obywatelska, or PO) electorates in 2005 has given way to a much more
clear-cut ideological division between the two camps since then (Fomina 2019, 86). In the wake

8
of the 2019 parliamentary elections, PiS predominantly represents people with lower education
levels, older people, and rural residents, whereas PO’s electorate is constituted primarily of ur-
ban residents as well as those holding high professional status and university degrees (Markowski
2020).
In sum, Poland represents a democracy at a crossroads. Significant steps towards an erosion
of democratic standards have already been taken, but elections remain reasonably competitive, as
confirmed most recently in the October 2023 parliamentary elections that have given the opposition
a chance to form a viable government. In light of the increasing pressure on judicial independence
and free media under PiS rule, citizens effectively stand center-stage as potential safeguards against
a further dismantling of checks and balances and a full breakdown of democracy. This sensitive
stage in the process of democratic backsliding makes Poland a particularly promising case in which
to probe the linkages between understandings of democracy and candidate choice. At the same
time, deep partisan polarization makes Poland a most likely candidate for explanations related
to partisan considerations. Finding evidence for our alternative explanation based on divergent
democratic attitudes among citizens in this context would therefore suggest our findings are likely
to travel to other comparable contexts of democratic threat.

Research Design: An Experimental Study in Poland

To examine to what extent divergent understandings of democracy play a role in voter prefer-
ences for candidates with varying democratic values, we develop a paired conjoint experiment.
We first describe our study design. We then explain the measurement of the dependent variable,
highlighting the advantages of using individual marginal component effects (IMCE) over a more
conventional approach based on average marginal component effects (AMCEs). Next, we describe
the measurement of understandings of democracy, our independent variable. Finally, we present
the empirical strategy we use to probe our hypotheses.

Study Design

Our analytical approach leverages a candidate choice conjoint experiment. This design allows
us to integrate alternative elements alongside the democratic positions contained in candidates’

9
profiles to capture potential trade-offs voters engage in (Schedler 2019; Svolik 2020). The resulting
multi-dimensional set-up allows us to assess the weight of democratic positions when it comes to
respondents’ evaluations of competing candidates.
In our study, we placed respondents into a hypothetical election situation and asked them to
choose between two competing profiles of candidates running for seats in the national lower house
(Sejm).2 We use the conjoint setting to effectively manipulate elite behavior—the supply side of our
argument—by varying the positions contained in our candidate profiles regarding the nomination
of judges and the role of public media. We strive to capture divergent views of democracy on
the elite side by formulating the levels for the two democratic attributes in line with the liberal,
majoritarian, and authoritarian understandings we developed for the citizens’ perspective. Table
1 displays our democratic attributes along with the levels reflecting distinct understandings of
democracy.
Our selected attributes concern two elements—judicial independence and media freedom—
that are crucial to liberal democracy but also offer a range of options as to how they may be
implemented in a democratic system. The use of two distinct democratic attributes allows us
to conduct two separate tests of our argument regarding the linkages between understandings of
democracy and voter responses to democratic transgressions. Our transgressions capture violations
of liberal democratic norms rather than outright violations of the law (Ahmed 2022), thus enabling
us to probe the overall salience of candidates’ democratic views as well as the relative impact
of different variations to strong liberal views upon respondents’ candidate choice. For judicial
appointments, we complement the liberal position that judges should be based on cross-party con-
sensus with alternatives that foresee a selection by the government (majoritarian) or by the leader
of the ruling party (authoritarian). Regarding the role of media, we let candidates express the
liberal view that their role consists of reporting independently on political developments, suggest
that they should justify government policy towards the wider public (majoritarian), or instead that
their role is to defend government policy against criticisim (authoritarian). We deliberately choose
more subtle deviations from liberal democracy to model the gradual nature of democratic back-
sliding, which consists precisely of rather discrete ways of chipping away at checks and balances
that only jointly amount to a dismantling of democratic standards (Scheppele 2013). Moreover,
2. See Figure A.1 for the introduction text to the candidate choice tasks.

10
 Attribute Levels Concept
Judicial ap- Liberal: Judges should be selected based on cross-party Judicial independence
pointments consensus.
Majoritarian: Judges should be selected by the govern-
ment.
Authoritarian: Judges should be selected by the leader
of the ruling party.
Role of public Liberal: The role of public media is to report indepen- Media pluralism
media dently on political developments.
Majoritarian: The role of public media is to justify gov-
ernment policy towards the wider public.
Authoritarian: The role of public media is to defend gov-
ernment policy against criticism.

Table 1: Democratic attributes and levels.

we decided to refrain from including positions that are so extreme that they would draw near uni-
versal condemnation, making it difficult to discriminate whether such condemnation is driven by
an actual commitment to liberal democratic norms or due to considerations of social desirability.
As discussed below, despite their subtlety respondents are rather well able to discriminate among
the three distinct levels of democratic positions we introduce for our two democratic attributes (see
also Appendix Figure B.3).
We partnered with the Warsaw-based market research company Inquiry—YouGov’s represen-
tative for Central and Eastern Europe—to recruit a representative sample of Polish respondents
based on age, gender, geographic origin, and vote choice at the last national election for our online
survey, into which we embedded our conjoint experiment. The survey was conducted between 12
July and 12 August 2021 (N = 2, 910). As specified in our pre-registration, we removed speed-
ers and those respondents who failed attention checks from our sample (Berinsky, Margolis, and
Sances 2014), bringing the final sample we use for our analysis to 2,097 respondents. We report
measures of sample representativeness for the final sample as well as full results for alternative
sample specifications in the Online Appendix (Tables A.2 and A.3 and Section B.5).
We asked respondents to complete twelve discrete choice tasks, each time choosing between
two candidates (forced-choice) and rating each candidate on a 7-point Likert scale ranging from
’strongly disapprove’ (1) to ’strongly approve’ (7). Each candidate profile was identified with a
neutral label (‘Candidate A’ vs. ‘Candidate B’) and displayed randomized information on seven

11
attributes, with the order of attributes fully randomized anew for each choice task. Alongside
candidates’ respective democratic views, our competing profiles contained information on their
gender, age, policy positions, and partisanship. For partisanship, we presented respondents with
a mix of candidates from all parties or party coalitions that scored above 5 percent of vote share
according to polls in June 2021, when the survey design was finalized. Choice situations also
included run-offs between candidates of the same party background. We include the full attribute
table in Table A.2 in the Appendix.

Dependent Variable: Candidate Evaluations

According to our theoretical argument, divergent understandings of democracy affect the extent to
which citizens are likely to overlook democratic transgressions when evaluating competing candi-
dates. We use our conjoint experiment to measure the weight of candidates’ democratic attributes
in individual respondents’ candidate ratings by computing individual marginal component effects
(IMCEs) (Zhirkov 2022). Respondents’ IMCEs then serve as a measure of our dependent variable.
Before we turn to explain the construction and purpose of IMCEs in more detail, we first discuss
the caveats of existing candidate choice experiments that rely on average marginal component
effects (AMCEs).

Average Marginal Component Effects (AMCEs) and their Limits

Our design seeks to estimate how much importance respondents assign to multidimensional can-
didate characteristics. First, we replicate the traditional approach by computing average marginal
component effects (AMCEs) for our candidate choice experiment (see Appendix Figure B.3). AM-
CEs allow researchers to estimate the effect of an individual treatment component over the joint
distribution of the remaining attributes (Hainmueller, Hopkins, and Yamamoto 2014, 10). Focus-
ing on the attributes of the judiciary and media, the AMCEs suggest that, on average, respondents
approve less of candidates who make majoritarian or authoritarian statements compared to liberal
ones.3
However, AMCEs do not allow us to draw inferences about the individual level. Specifically,
3. Note that the largest effect on candidate choices can be attributed to shared partisanship between the respondent
and candidate profiles.

12
the observed pattern may reflect a shared adherence to reasonably liberal democratic attitudes
across our sample. However, AMCEs may just as well mask considerable divergence of choice
behavior within our sample, with some respondents punishing democratic violations very harshly,
while others are indifferent or even approve of candidates holding non-liberal democratic views.
Yet it is precisely such divergent behaviors at the individual level that underpin the causal mech-
anism we seek to probe in our study, leading us to adopt a recently proposed alternative approach
to conjoint analysis via individual marginal component effects.

Individual Marginal Component Effects (IMCEs) as a Measure of Individual-level Candidate

Preferences

Individual marginal component effects (IMCEs) overcome some of the limitations of analyses
focused on AMCEs (Zhirkov 2022). In this approach, each respondent rates a relatively high
number of profiles (in our case 24 candidate profiles in 12 election runoffs) on a rating scale,4
allowing us to estimate the effects of each candidate attribute level on the respondent’s rating for
candidates. For instance, if a respondent repeatedly rates liberal candidate profiles more highly, her
IMCE on the corresponding liberal attribute will be higher. Substantially, higher values indicate
a stronger preference for liberal democratic candidates. We detect this preference by regressing a
respondent’s ratings of the 24 candidates on each of the candidate’s attributes separately:

yi = αil + πil X0 il + εil , (1)

where yi is a vector of ratings for each candidate profile made by respondent i, X0 il a vector of
values of attribute l shown to the respondent i, and εil a vector of respondent-specific errors. We
define π̂il as the IMCE for attribute l. In our study, we focus on individuals’ IMCEs for the two
attributes relating to candidates’ statements toward democracy (i.e., judicial appointments and role
of public media).5 To assess the empirical relevance of a congruence between respondents’ and
candidates’ democratic views for respondents’ candidate evaluations, we regress the IMCEs of the
two democratic attributes on the three understandings of democracy. This allows us to examine the
4. IMCEs are computed with candidate ratings instead of choices to receive more reliable estimates at the individual
level. We compare the AMCEs on the candidate rating with choices and find no differences in the relative weight of
attributes (see Online Appendix B.2.3).
5. Note that due to full randomization, IMCE estimates for each attribute are independent of each other.

13
relative importance of congruence separately for all three understandings, enabling us to evaluate
whether respondents with strong liberal democratic attitudes lend greater weight to candidates’
democratic views. As mentioned above, we observe the ratings separately for each of the two
democratic attributes in order to assess how the area in which a democratic transgression occurs
may concern respondents to a different extent.
Previous research has proposed to divide a population into subgroups of interest (e.g., based on
gender or partisanship) and study average conjoint behavior separately for these groups (Leeper,
Hobolt, and Tilley 2020). However, defining such subgroups for our attitudinal concept of un-
derstandings of democracy would require imposing arbitrary thresholds to distinguish different
subgroups from one another. IMCEs, by contrast, enable us to assess the impact of different un-
derstandings of democracy on a continuous scale. Besides, aggregate analyses usually allow for
examining only one covariate of interest at a time. But we often expect the covariate of interest to
vary with other covariates, raising concerns about omitted variable bias.6 By contrast, determining
to what extent individual respondents care about candidates’ stances toward democracy allows us
to consider a range of explanatory variables jointly in a regression framework.
IMCEs rely on the same set of assumptions as AMCEs. That is, only when the assumptions of
(1) stability and no carryover effects, (2) no profile-order effects, and (3) completely independent
randomization of the profiles in a conjoint experiment hold, IMCEs can be estimated independently
for each respondent (Hainmueller, Hopkins, and Yamamoto 2014; Zhirkov 2022). For our candi-
date choice experiment, we verified assumption 1 (see Online Appendix B.2.2), and assumptions
2 and 3 are true by design, allowing us to proceed with the estimation of IMCEs.

Independent Variable: Understandings of Democracy

Empirical studies often tend to equate democratic commitment with citizens’ support for the
generic concept of democracy (Wuttke, Gavras, and Schoen 2020; Foa and Mounk 2016, 2017).
This narrow understanding—and corresponding measurement—of democratic commitment is in-
creasingly recognized as a key limitation in accurately assessing citizens’ democratic beliefs (In-
6. For instance, when a population is divided into partisan subgroups and average choice behavior is compared
against each other, one could not rule out that age confounds the behavioral differences found for different partisan
groups.

14
glehart 2003; Ananda and Bol 2021; Alonso 2016; König, Siewert, and Ackermann 2022). To ex-
plain heterogeneity in individuals’ evaluations of candidates expressing different views of democ-
racy, we instead implement a more nuanced measurement model to gauge individuals’ respec-
tive scores for distinct understandings of democracy. We introduce the resulting individual factor
scores as independent variables into a regression model, controlling for party preference and so-
cioeconomic variables. In essence, our research design thus assesses to what extent respondents’
understandings of democracy in the abstract translate into a willingness to punish democratic trans-
gressions in a concrete candidate choice situation.
To measure respondents’ understandings of democracy, we revise and expand an item battery
from the World Value Survey (WVS) (Haerpfer et al. 2020) and implement the measurement model
outlined in our pre-registration, asking respondents to rate how essential they find each item to be
for a democracy on a scale from 1 to 7. We pre-tested the majoritarian and authoritarian items for
internal validity and statistical benchmarks for confirmatory factor analysis in a dedicated pre-test
among a smaller sample of Polish respondents. Table 2 reports the retained items as included in
our pre-registration.

L1: People choose their leaders in free elections.

Liberal understanding L2: Civil rights protect people from state oppression.
L3: Women have the same rights as men.
M1: The majority can always overrule the minority.
Majoritarian understanding M2: Any law can be changed if there is a majority for it.
M3: The minority must accept the will of the majority in all circumstances.
A1: The government uses violence to enforce public order.
Authoritarian understanding A2: Elections only serve to confirm the ruling party in office.
A3: The government limits civic freedoms to rule efficiently.

Table 2: Item battery of understandings of democracy.

Based on these observed items, we implemented an ordered confirmatory factor analysis with
three separate latent variables corresponding to a liberal, majoritarian, and authoritarian under-
standing of democracy.7 We compute individual factor scores for each latent variable based on
the model. Since the different understandings may be correlated with one another,8 we allow
covariance between the three latent variables and choose to assess the relative strength of each
7. Cronbach’s alpha for liberal items = 0.703, for majoritarian items = 0.653, and for authoritarian items = 0.747.
8. Indeed, as the measurement model indicates (Table B.1), liberal and authoritarian understandings are negatively
correlated (−0.37, p < 0.001). By contrast, authoritarian and majoritarian understandings are positively correlated
(0.27, p < 0.001). Liberal and majoritarian understandings co-vary only marginally (0.04, p < 0.001).

15
respondent’s support for the three distinct understandings of democracy separately. The model
indicates a good model fit (χ 2 = 284.44, CFI = 0.99, TLI = 0.98, RMSEA = 0.06), suggesting that
our measures are internally valid (see details of measurement model in Online Appendix B.1).
To further probe the robustness of our measurement, we assess to what extent respondents’
understandings of democracy correspond to their evaluation of the democratic attributes we had
included in the conjoint. We asked respondents outside the actual candidate choice experiment
to rate how democratic they thought each of the statements toward judges and media was (see
Online Appendix B.7). The observed patterns indicate that our independent measure of respon-
dents’ understandings of democracy maps onto their evaluation of the items we chose to include
in the conjoint, making these a salient measure of the congruence between respondent-level and
candidate-level democratic views.
In a final test of our measurement, we examine the extent to which divergent understandings
of democracy correspond to party preferences. Earlier experimental findings indicate that citi-
zens may rationalize democratic violations when they are carried out by an actor whose policy
preferences align with their own (Krishnarijan 2022). Such behavior is particularly prevalent in
contexts of democratic backsliding, allowing citizens to alter their very perception of what is demo-
cratic or undemocratic to convince themselves that they are getting both their preferred policy and
democracy. A previous study in Poland has suggested that backsliding leaders are able to maintain
themselves in power precisely because their leadership style aligns with a distinct understand-
ing of democracy among their electorate (Reykowski 2020). In this scenario, partisanship would
largely overlap with distinct democratic attitudes, potentially undermining our argument that these
attitudes represent a distinct dimension influencing citizens’ voting choices. Examining the role of
incumbency, a recent study shows that citizens can adjust their understanding of democracy to their
partisan interest, with incumbent supporters typically more majoritarian in their orientation than
opposition supporters (Bryan 2023). We therefore examine to what extent partisan preferences are
associated with respondents’ understandings of democracy in ways that affect their responses to
democratic violations.
To do so, we investigate how understandings of democracy vary both between and within par-
tisan groups. If an overwhelming share of the variance in understandings of democracy were to be
explained by partisan affiliation, it would provide strong support for the claim that party supporters

16
 Liberal Majoritarian Authoritarian
Within partisan groups 69.35% 54.54% 74.95%
Between partisan groups 30.65% 45.46% 25.05%

Table 3: Breakdown of variance in understandings of democracy scores as analyzed using a one-sided

Analysis of Variance (ANOVA). The percentage of within partisan group variance is obtained from the
proportion of residuals in the overall variance, representing variation within each partisan group. The
variance between partisan groups is the remaining share of the overall variance, representing differences
between partisan groups.

in backsliding countries substantially differ in the notions of democracy they subscribe to. Table
3 breaks down the variance in understandings of democracy explained by partisan affiliation and
the unexplained variance, indicating variation within partisan groups that party preferences alone
cannot account for. Specifically, 69%, 55%, and 75% of the variance in liberal, majoritarian, and
authoritarian understanding scores, respectively, cannot be attributed to differences in partisanship.
This suggests that while some divergent understandings of democracy are associated with party
preference, a larger share of divergence exists within partisan groups. This implies that citizens
who support the same party vary considerably in their understanding of democracy. We will now
proceed to examine whether these within-party differences are related to support for non-liberal
candidates.

Empirical Strategy

To assess how divergent understandings of democracy affect vote choice, we implement OLS mod-
els regressing individuals’ IMCEs for democratic attributes (π̂il ) on a vector of their understandings
of democracy (X0 i ), controlling for a vector of partisanship and sociodemographic variables (Z0 i ):

π̂il = αi + X0 i β1 + Z0 i β2 + εi . (2)

This approach allows us to evaluate the relevance of divergent democratic attitudes while con-
trolling for party preference and socio-demographic variables. Controlling for respondents’ pre-
ferred party allows us to rule out that different partisan attachments confound the relationship
between understandings of democracy and revealed democratic attitudes. Similarly, adding socio-
demographic variables (age, gender, education, income, perceived economic status) helps mitigate

17
concerns over omitted variables bias, as socioeconomic status could also feed into respondents’
level of democratic commitment as measured in the candidate experiment.

Empirical Results
Our empirical analysis tests our argument according to which divergent understandings of democ-
racy feed into political choice. In a first step, we draw on descriptive patterns of vote choice to
explore the aggregate relationship between understandings of democracy and candidate prefer-
ences. We then examine to what extent divergent democratic attitudes at the individual level help
to explain citizens’ evaluations of competing candidates at the ballot box. In a final test of our
argument, we investigate to what extent artisan voters’ willingness to shift from a non-liberal co-
partisan candidate to a liberal out-party candidate is associated with divergent understandings of
democracy.

The Aggregate Relationship between Understandings of Democracy and Can-

didate Preferences

Our central argument holds that divergent understandings of democracy feed into citizens’ political
choices in contexts of democratic backsliding. To provide an aggregate overview of the relationship
between divergent understandings of democracy and choices between candidates who advance
different democratic views for the entire survey electorate, we examine a subsection of vote choices
that pit a candidate whose democratic positions are consistent across both democratic attributes
against a candidate holding mixed views. We first focus on choices in which respondents are
confronted with one candidate with consistent liberal attributes and another candidate holding at
least one non-liberal (i.e., majoritarian or authoritarian) attribute. Figure 1(a) examines the fraction
of choices made for the purely liberal candidate and plots the share along with respondents’ extent
of liberal understanding of democracy. The more respondents’ liberal orientation increases, the
more they prefer the liberal candidate to her non-liberal contender. Substantially, from the least to
the most liberal respondent in the sample, we find an average increase of about 15% in electoral
support for congruent liberal candidates.

18
 0.8 0.8

Fraction of Chosen Majoritarian Candidates

Fraction of Chosen Liberal Candidates

0.7 0.7

0.6 0.6

0.5 0.5

0.4 0.4

0.3 0.3

-1 0 1 -1 0 1
Liberal Understanding of Democracy Majoritarian Understanding of Democracy

(a) Liberal Understanding (b) Majoritarian Understanding

0.8
Fraction of Chosen Authoritarian Candidates

0.7

0.6

0.5

0.4

0.3

-1 0 1
Authoritarian Understanding of Democracy

(c) Authoritarian Understanding

Figure 1: Fraction of vote choice for a consistent over a non-consistent candidate at varying levels of
respondents’ understandings of democracy.

We find a reverse but weaker pattern for choice situations where a fully authoritarian candi-
date runs against a candidate making at least one non-authoritarian claim (Figure 1(c)): the more
respondents subscribe to an authoritarian understanding, the higher the vote share for consistent au-
thoritarian candidates, with an overall increase in electoral support of 10%. However, note that the
overall vote share for purely authoritarian candidates does not exceed 40%, even among strongly
authoritarian respondents, indicating an overall rejection of candidates expressing authoritarian
positions by respondents.
A different pattern emerges for a majoritarian understanding of democracy (Figure 1(b)). Irre-
spective of respondents’ attitude toward a majoritarian vision of democracy, the average vote share

19
remains consistently at around 50%, suggesting that a majoritarian understanding is not associ-
ated with preferences for fully majoritarian candidates over contenders with a different democratic
profile.

Individual-level Understandings of Democracy and Candidate Preferences

The descriptive overview provides insights into the aggregate relationship between divergent un-
derstandings and support for political candidates. To investigate the individual-level association
between divergent understandings of democracy and the evaluation of candidates expressing dis-
tinct positions on liberal democratic safeguards, we turn to the analysis of individual-level candi-
date evaluations by regressing respondents’ IMCEs on their liberal, majoritarian, and authoritarian
understandings scores. Table 4 reports our main findings.
The results indicate that the more respondents subscribe to a liberal understanding of democ-
racy, the less supportive they are of candidates who make (1) majoritarian and (2) authoritarian
claims about the judiciary and (3) propose authoritarian-leaning reforms of media independence.
In turn, a higher majoritarian understanding is associated with stronger support for candidates de-
livering majoritarian and authoritarian statements about the appointment of judges. By contrast, a
majoritarian understanding does not predict the approval of candidates proposing majoritarian and
authoritarian views on the role of public media. Similarly, a higher authoritarian understanding is
not negatively related to respondents’ support for majoritarian or authoritarian candidates. We add
party controls to our models to show that the effect of distinct understandings of democracy holds
even when we account for citizens’ partisan affiliation and other party-related dynamics that we
discuss further below.
Overall, divergent democratic attitudes play a discrete and non-negligible role in shaping citi-
zens’ vote choices in contexts of democratic backsliding. Divergent understandings of democracy
within the citizenry appear to enable democratic backsliding primarily due to distinct levels of
liberal democratic commitment among citizens, thus confirming our Hypothesis 1a.
By contrast, we find no support for our Hypothesis 1b on the overall congruence between re-
spondents’ and candidates’ democratic views as a predictor of candidate preference. If this was
the case, we should find not only that respondents expressing strongly liberal attitudes are most

20
 Judges: Judges: Media: Media:
Majoritarian Authoritarian Majoritarian Authoritarian
(1) (2) (3) (4)
Liberal understanding −0.151∗∗ −0.214∗∗∗ −0.024 −0.137∗∗
(0.051) (0.051) (0.049) (0.052)
Majoritarian understanding 0.140∗∗ 0.098∗ −0.022 0.020
(0.046) (0.047) (0.044) (0.047)
Authoritarian understanding −0.004 0.003 0.057 0.061
(0.055) (0.056) (0.053) (0.057)
Confederation 0.224∗∗ 0.141 0.056 0.114
(0.078) (0.080) (0.075) (0.080)
Law and Justice 0.346∗∗∗ 0.291∗∗∗ 0.177∗∗∗ 0.265∗∗∗
(0.055) (0.055) (0.052) (0.056)
Poland 2050 0.057 0.064 0.025 0.005
(0.059) (0.060) (0.057) (0.061)
The Left 0.013 0.066 −0.017 0.030
(0.086) (0.087) (0.083) (0.088)
No party preference/Don’t know 0.131∗ 0.096 0.063 0.125∗
(0.055) (0.056) (0.052) (0.056)
Constant −0.215∗ −0.282∗∗ −0.403∗∗∗ −0.594∗∗∗
(0.101) (0.103) (0.097) (0.104)
Socioeconomic Controls X X X X
Observations 2,097 2,097 2,097 2,097
R2 0.069 0.068 0.022 0.048
Adjusted R2 0.061 0.060 0.014 0.039
∗∗∗ p < 0.001; ∗∗ p < 0.01; ∗ p < 0.05

Table 4: OLS regression of revealed democratic attitudes (IMCEs) on liberal, majoritarian, and au-
thoritarian understandings of democracy, controlling for party preference and socioeconomic controls.
Reference category for party preference = Civic Coalition. The full regression table can be found in
Table B.5.

likely to support candidates with liberal views. In addition, respondents holding more majoritarian
or authoritarian views of democracy should similarly endorse candidates expressing correspond-
ing democratic positions, possibly even to the point of rating lower those candidates who espouse
liberal democratic views. In other words, this would imply that certain voters support specific
candidates not despite the undemocratic practices they sponsor, but precisely because these candi-
dates profess views that align with their own views. However, our analysis suggests instead that
respondents who endorse non-liberal understandings of democracy appear to lend less weight to
candidates’ democratic views rather than actively supporting candidates who propose democratic
transgressions that correspond to their understanding of democracy.

21
Partisan Voting and Rejecting Non-liberal Candidates

Our findings so far provide evidence that citizens’ level of commitment to liberal democracy shapes
their willingness to support candidates making majoritarian and authoritarian claims. A large body
of literature suggests that partisanship is a main driver of citizens’ tolerance towards violations of
democratic principles by co-partisan politicians (Ahlquist et al. 2018; Carey et al. 2020; Graham
and Svolik 2020). The key mechanism underpinning this argument is that voters are unwilling to
switch to an out-party candidate if their own co-partisan behaves undemocratically. To examine
to what extent partisans’ willingness to shift to an out-party candidate if their own adopts non-
liberal positions is associated with divergent understandings of democracy, we focus on a subset of
choices, namely those between a co-partisan who adopts at least one non-liberal (i.e., majoritarian
or authoritarian) position and who runs against a purely liberal out-party candidate. We implement
the following linear probability regression:

Yi j = αi + X0 i β1 + Z0 i β2 + εi , (3)

where Y is respondent i’s preference for the non-liberal co-partisan in choice j, X0 i is the vector
for respondent’s understanding of democracy scores, and Z0 i is a vector of socioeconomic controls
as defined in Equation 2.
This analysis allows us to test whether partisans’ willingness to shift from a non-liberal co-
partisan candidate to a liberal out-party candidate is associated with divergent understandings of
democracy. Table 5 displays the results. Controlling for partisanship and socioeconomic variables,
the more respondents subscribe to a liberal understanding of democracy, the less likely they are to
vote for the non-liberal co-partisan candidate over a liberal out-party candidate. We find a reverse
effect for an authoritarian understanding of democracy, but the association diminishes when jointly
regressing vote preference on all understandings of democracy. A majoritarian understanding of
democracy is unrelated to voting for a non-liberal co-partisan over a liberal out-party candidate.
The results are consistent with our finding that divergent degrees of how respondents embrace
a liberal understanding of democracy are associated with the extent to which voters reject non-
liberal candidates: when respondents’ co-partisan candidate adopts a non-liberal position and runs
against a liberal out-party candidate, the extent to which partisan voters abandon the non-liberal

22
 (1) (2) (3) (4)
Liberal understanding −0.100∗∗∗ −0.123∗
(0.030) (0.053)
Majoritarian understanding 0.033 0.062
(0.034) (0.051)
Authoritarian understanding 0.081∗∗ −0.026
(0.030) (0.061)
Law and Justice 0.123∗ 0.121∗ 0.116∗ 0.106∗
(0.051) (0.053) (0.052) (0.052)
Poland 2050 0.015 0.012 0.015 0.006
(0.057) (0.058) (0.057) (0.057)
The Left −0.047 −0.045 −0.046 −0.049
(0.075) (0.076) (0.075) (0.075)
Confederation −0.040 −0.028 −0.037 −0.046
(0.072) (0.073) (0.073) (0.072)
Constant 0.225∗ 0.268∗ 0.240∗ 0.224∗
(0.111) (0.111) (0.110) (0.111)
Socioeconomic controls X X X X
R2 0.058 0.043 0.053 0.061
Adj. R2 0.038 0.022 0.033 0.038
N Choices 710 710 710 710
N Respondents 608 608 608 608
∗∗∗ p < 0.001; ∗∗ p < 0.01; ∗ p < 0.05

Table 5: Linear probability model (OLS) of voting an non-liberal co-partisan candidate over an liberal
out-party candidate on liberal, majoritarian, and authoritarian understandings of democracy, controlling
for party preference and socioeconomic controls. Robust standard errors clustered at the respondent
level are reported. Reference category for party preference = Civic Coalition.

co-partisan varies with the respondents’ liberal understanding of democracy.9

Conclusions: The Role of Divergent Understandings of Democ-

racy in Democratic Backsliding
Our study set out to probe an alternative explanation of why citizens, despite overwhelmingly
supporting democracy in principle, may fail to use elections to remove political elites holding non-
9. In Appendix B.2.4, we examine in more depth how partisanship manifests in preferences for candidates holding
different democratic positions.

23
liberal democratic views from power. We contend that failure to punish democratic violations at
the ballot box reflects considerable heterogeneity among citizens regarding their understandings
of democracy, and notably a lack of attitudinal consolidation around liberal democracy. Our em-
pirical findings lend support to our theoretical argument linking citizens’ democratic attitudes to
their vote choice. Overall, we confirm that voters with strong liberal democratic understandings
punish democratic violations most consistently, whereas this is not the case for respondents hold-
ing majoritarian or authoritarian views. Hence, where liberal democratic commitment is weak or
unevenly distributed across the electorate, citizens consequently fail to play the role of democratic
bulwarks against authoritarian-leaning elites.
By unpacking the uni-dimensional measurement of ‘support for democracy,’ our study makes
several contributions to our understanding of democratic backsliding and the place of citizens in
such processes. Most fundamentally, our findings question the assumption in much of the existing
research that people have a common understanding of democracy, and—especially in polarized
contexts—sacrifice democratic performance primarily due to partisan considerations. Instead, we
show that there is no close overlap between distinct democratic attitudes and party preference and
that the willingness to punish undemocratic co-partisans relates closely to respondents’ commit-
ment to liberal democracy.
Conceptually, our analysis expands upon earlier findings highlighting the threat of majoritar-
ian voters for liberal democracy (Grossman et al. 2021). We propose an overarching argument
regarding the behavioral consequences of democratic attitudes when it comes to candidate evalu-
ations and their relevance in contexts of democratic backsliding. We contrast liberal democratic
attitudes with non-democratic authoritarian conceptions as well majoritarian views. In doing so,
we respond to calls to investigate how the strength of democratic beliefs (Carlin 2018, 419) and
the liberal-democratic quality of citizens’ regime preferences (Wuttke, Gavras, and Schoen 2020)
relate to vote choice and eventual democratic outcomes. We thereby explicitly tackle the presence
of contestation around the concept of democracy itself that has been posited as a key oversight in
studies of citizens responses to democratic backsliding so far (Ahmed 2022).
We put forward—and examine empirically—two distinct mechanisms that may account for
the impact of divergent democratic attitudes upon vote choice. Our main insight suggests that
the relative strength of liberal democratic commitment is most crucial to citizens’ willingness

24
to counter democratic backsliding at the ballot box. In turn, we reject the hypothesis that vote
choice depends on a generalized congruence between voters’ understandings of democracy and
the democratic views expressed by candidates.
Our empirical analysis of the Polish case indicates that citizens’ understandings of democracy
play a relevant role in explaining their vote choice and evaluations of competing candidates. Al-
though Polish voters on average reject candidates who actively endorse a weakening of checks
and balances, the picture is more complex at the individual level: parts of the electorate hold only
weak liberal attitudes or espouse majoritarian or authoritarian views of democracy and are there-
fore indifferent toward candidates who advocate undermining key features of liberal democracy. In
principle, such weak commitment to liberal democracy may be considered a specific characteristic
of post-Communist political systems (Pop-Eleches and Tucker 2017, 309–310) and related to their
comparatively short experience with democracy. However, given the rise of increasingly open il-
liberal appeals by political leaders from the United States to Hungary to Brazil, we are confident
that our findings on heterogeneous democratic attitudes as a key vulnerability of political systems
to democratic backsliding are generalizable to other contexts.
Ultimately, our empirical investigation of divergent democratic attitudes and their impact on
contexts of democratic backsliding helps reconcile earlier findings on high nominal support for
democracy with electoral victories of illiberal parties or candidates. These findings hold impli-
cations for the potential recipe to make citizens into effective bulwarks for democracy: rather
than a mass of ideologically centrist voters willing to abandon incumbents acting undemocrati-
cally (Svolik 2020, 27), we contend that what is needed to counter democratic backsliding is a
firm commitment not simply to democracy in its broadest sense, but to the specific principles of
separation of powers and civil liberties that underpin liberal democracy.
In practical terms, our analysis indicates that where attitudinal consolidation around liberal
democracy remains insufficiently developed, voters cannot reliably act as safeguards against demo-
cratic backsliding. Instead, they remain vulnerable to majoritarian and authoritarian appeals by
elites. Where such non-liberal elites coincide with an electorate whose commitment to liberal
democracy is not firmly anchored, they may successfully activate latent or open non-liberal un-
derstandings of democracy upon which they can draw to legitimize their gradual dismantling of
democratic standards. Future research may thus investigate the role elites play in mobilizing such

25
alternative democratic attitudes to garner enduring popular support for their illiberal designs.

Competing interests: The authors declare none.

References
Ahlquist, John S., Nahomi Ichino, Jason Wittenberg, and Daniel Ziblatt. 2018. “How do voters
perceive changes to the rules of the game? Evidence from the 2014 Hungarian elections.”
Journal of Comparative Economics 46 (4): 906–919.

Ahmed, Amel. 2022. “Is the American Public Really Turning Away from Democracy? Backsliding
and the Conceptual Challenges of Understanding Public Attitudes.” Perspectives on Politics,
1–12.

Allan, James. 2017. “Majoritarianism.” Bond Law Review 29:187–196.

Almond, Gabriel A., and Sidney Verba. 1963. The civic culture: Political attitudes and democracy
in five nations. Boston: Princeton University Press.

Alonso, Sonia. 2016. “What Type of Democratic Commitment Lies Behind the Importance of
Living in a Democracy?” In How Europeans View and Evaluate Democracy, edited by Mónica
Ferrı́n and Hanspeter Kriesi. Oxford University Press.

Ananda, Aurelia, and Damien Bol. 2021. “Does Knowing Democracy Affect Answers to Demo-
cratic Support Questions? A Survey Experiment in Indonesia.” International Journal of Public
Opinion Research 33 (2): 433–443.

Aspinall, Edward, Diego Fossati, Burhanuddin Muhtadi, and Eve Warburton. 2020. “Elites, masses,
and democratic decline in Indonesia.” Democratization 27 (4): 505–526.

Bakke, Elisabeth, and Nick Sitter. 2022. “The EU’s Enfants Terribles : Democratic Backsliding in
Central Europe since 2010.” Perspectives on Politics, 1–16.

Bakule, Jakub. 2020. “The good, the bad and the ugly: linking democratic values and participation
in the Czech Republic.” Democratization, 1–19.

26
Bartels, Larry M. 2023. Democracy erodes from the top: Leaders, citizens, and the challenges of
populism in Europe. Princeton Studies in Political Behavior. Princeton: Princeton University
Press.

Bengtsson, Åsa, and Henrik Christensen. 2016. “Ideals and Actions: Do Citizens’ Patterns of Po-
litical Participation Correspond to their Conceptions of Democracy?” Government and Oppo-
sition 51 (2): 234–260.

Berinsky, Adam J., Michele F. Margolis, and Michael W. Sances. 2014. “Separating the Shirkers
from the Workers? Making Sure Respondents Pay Attention on Self-Administered Surveys.”
American Journal of Political Science 58 (3): 739–753.

Bermeo, Nancy. 2016. “On Democratic Backsliding.” Journal of Democracy 27 (1): 5–19.

Broockman, David E., Joshua L. Kalla, and Sean J. Westwood. 2023. “Does Affective Polarization
Undermine Democratic Norms or Accountability? Maybe Not.” American Journal of Political
Science 67 (3): 808–828.

Bryan, James D. 2023. “What Kind of Democracy Do We All Support? How Partisan Interest
Impacts a Citizen’s Conceptualization of Democracy.” Comparative Political Studies 56 (10):
1597–1627.

Buštı́ková, Lenka, and Petra Guasti. 2017. “The Illiberal Turn or Swerve in Central Europe?”
Politics and Governance 5 (4): 166–176.

Canache, Damarys. 2012. “Citizens’ Conceptualizations of Democracy: structural complexity, sub-

stantive content, and political significance.” Comparative Political Studies 45 (9): 1132–1158.

Carey, John M., Katherine Clayton, Gretchen Helmke, Brendan Nyhan, Mitchell Sanders, and
Susan Stokes. 2020. “Who will defend democracy? Evaluating tradeoffs in candidate support
among partisan donors and voters.” Journal of Elections, Public Opinion & Parties, 1–16.

Carey, John M., Gretchen Helmke, Brendan Nyhan, Mitchell Sanders, and Susan Stokes. 2019.
“Searching for Bright Lines in the Trump Presidency.” Perspectives on Politics 17 (3): 699–
718.

27
Carlin, Ryan E. 2011. “Distrusting Democrats and Political Participation in New Democracies.”
Political Research Quarterly 64 (3): 668–687.

. 2018. “Sorting Out Support for Democracy: A Q-Method Study.” Political Psychology 39
(2): 399–422.

Chapman, Hannah S., Margaret C. Hanson, Valery Dzutsati, and Paul DeBell. 2023. “Under the
Veil of Democracy: What Do People Mean When They Say They Support Democracy?” Per-
spectives on Politics, 1–19.

Chu, Yun-han, and Min-hua Huang. 2010. “Solving an Asian Puzzle.” Journal of Democracy 21
(4): 114–122.

Cinar, Kursat, and Asya Bulbul. 2022. “Varieties of democratic understanding and political partic-
ipation: multi-level evidence from the world.” Contemporary Politics 28 (4): 469–489.

Claassen, Christopher. 2020. “Does Public Support Help Democracy Survive?” American Journal
of Political Science 64 (1): 118–134.

Cohen, Mollie J., and Amy Erica Smith. 2016. “Do authoritarians vote for authoritarians? Evidence
from Latin America.” Research & Politics 3 (4): 1–8.

Davis, Nicholas T., Ronald Keith Gaddie, and Robert K. Goidel. 2022. Democracy’s meanings:
How the public understands democracy and why it matters. Ann Arbor: University of Michi-
gan Press.

Dore, Giovanna Maria Dora. 2014. “Democracy Is Not the Only Game in Town! Democratic and
Authoritarian Attitudes in Indonesia, Korea, and Thailand.” In Incomplete democracies in
the Asia-Pacific, edited by Giovanna Maria Dora Dore, Jae H. Ku, and Karl Jackson, 13–62.
Basingstoke: Palgrave Macmillan.

Easton, David. 1975. “A Re-assessment of the Concept of Political Support.” British Journal of
Political Science 5 (4): 435–457.

Foa, Roberto Stefan, and Yascha Mounk. 2016. “The Danger of Deconsolidation: The Democratic
Disconnect.” Journal of Democracy 27 (3): 5–17.

28
Foa, Roberto Stefan, and Yascha Mounk. 2017. “The Signs of Deconsolidation.” Journal of Democ-
racy 28 (1): 5–15.

Fomina, Joanna. 2019. “Of ’Patriots’ and Citizens: Asymmetric Populist Polarization in Poland.”
In Democracies Divided: The Global Challenge of Political Polarization, edited by Thomas
Carothers and Andrew O’Donohue. Brookings Institution Press.

Fomina, Joanna, and Jacek Kucharczyk. 2016. “Populism and Protest in Poland.” Journal of Democ-
racy 27 (4): 58–68.

Fossati, Diego, Burhanuddin Muhtadi, and Eve Warburton. 2021. “Why democrats abandon democ-
racy: Evidence from four survey experiments.” Party Politics.

Freedom House. 2020. Nations in Transit: Poland. https : / / freedomhouse . org / country / poland /
nations-transit/2020.

Gibson, James L. 1996. “A Mile Wide But an Inch Deep(?): The Structure of Democratic Com-
mitments in the Former USSR.” American Journal of Political Science 40 (2): 396–420.

Gidengil, Elisabeth, Dietlind Stolle, and Olivier Bergeron-Boutin. 2022. “The Partisan Nature of
Support for Democratic Backsliding: A Comparative Perspective.” European Journal of Po-
litical Research 61 (4): 901–929.

Graham, Matthew H., and Milan W. Svolik. 2020. “Democracy in America? Partisanship, Polariza-
tion, and the Robustness of Support for Democracy in the United States.” American Political
Science Review 114 (2): 392–409.

Grigoriadis, Ioannis N. 2018. Democratic Transition and the Rise of Populist Majoritarianism:
Constitutional Reform in Greece and Turkey. 1st ed. 2018. Reform and Transition in the
Mediterranean. Cham: Springer International Publishing / Imprint: Palgrave Macmillan.

Grossman, Guy, Dorothy Kronick, Matthew Levendusky, and Marc Meredith. 2021. “The Majori-
tarian Threat to Liberal Democracy.” Journal of Experimental Political Science, 1–10.

29
Haerpfer, Christian, Ronald Inglehart, Alejandro Moreno, Christian Welzel, Kseniya Kizilova,
Jaime Diez-Medrano, Marta Lagos, Pippa Norris, Eduard Ponarin, and Bi Puranen. 2020.
World Values Survey Wave 7 (2017-2020) Cross-National Data-Set. https://doi.org/10.14281/
18241.1.

Haggard, Stephan, and Robert Kaufman. 2021. Backsliding: Democratic Regress in the Contem-
porary World. Cambridge: Cambridge University Press.

Hainmueller, Jens, Daniel J. Hopkins, and Teppei Yamamoto. 2014. “Causal Inference in Conjoint
Analysis: Understanding Multidimensional Choices via Stated Preference Experiments.” Po-
litical Analysis 22 (1): 1–30.

Inglehart, Ronald. 2003. “How Solid Is Mass Support for Democracy: And How Can We Measure
It?” Political Science & Politics 36 (1): 51–56.

Kirsch, Helen, and Christian Welzel. 2019. “Democracy Misunderstood: Authoritarian Notions of
Democracy around the Globe.” Social Forces 98 (1): 59–92.

König, Pascal D., Markus B. Siewert, and Kathrin Ackermann. 2022. “Conceptualizing and Mea-
suring Citizens’ Preferences for Democracy: Taking Stock of Three Decades of Research in a
Fragmented Field.” Comparative Political Studies 55 (12): 2015–2049.

Krishnarijan, Suthan. 2022. “Rationalizing Democracy: The Perceptual Bias and (Un)Democratic
Behavior.” American Political Science Review, 1–23.

Kruse, Stefan, Maria Ravlik, and Christian Welzel. 2019. “Democracy Confused: When People
Mistake the Absence of Democracy for Its Presence.” Journal of Cross-Cultural Psychology
50 (3): 315–335.

Leeper, Thomas J., Sara B. Hobolt, and James Tilley. 2020. “Measuring Subgroup Preferences in
Conjoint Experiments.” Political Analysis 28 (2): 207–221.

Levitsky, Steven, and Daniel Ziblatt. 2018. How democracies die. New York: Crown.

Linz, Juan J., and Alfred C. Stepan. 1996. “Toward Consolidated Democracies.” Journal of Democ-
racy 7 (2): 14–33.

30
Lipset, Seymour Martin. 1959. “Some Social Requisites of Democracy: Economic Development
and Political Legitimacy.” American Political Science Review 53 (1): 69–105.

Lührmann, Anna, and Staffan I. Lindberg. 2019. “A third wave of autocratization is here: What is
new about it?” Democratization 26 (7): 1095–1113.

Lührmann, Anna, Marcus Tannenberg, and Staffan I. Lindberg. 2018. “Regimes of the World
(RoW): Opening New Avenues for the Comparative Study of Political Regimes.” Politics
and Governance 6 (1): 60–77.

Markowski, Radoslaw. 2016. “The Polish parliamentary election of 2015: a free and fair election
that results in unfair political consequences.” West European Politics 39 (6): 1311–1322.

. 2020. “Plurality support for democratic decay: the 2019 Polish parliamentary election.”
West European Politics 43 (7): 1513–1525.

Matovski, Aleksandar. 2021. Popular dictatorships: Crises, mass opinion, and the rise of electoral
authoritarianism. New York: Cambridge University Press.

Mauk, Marlene. 2020. Citizen support for democratic and autocratic regimes. Oxford: Oxford
University Press.

Medzihorsky, Juraj, and Staffan Lindberg. 2023. “Walking the Talk: How to Identify Anti-Pluralist
Parties.” Party Politics.

O’Donnell, Guillermo A. 1998. “Horizontal Accountability in New Democracies.” Journal of

Democracy 9 (3): 112–126.

Orhan, Yunus Emre. 2022. “The Relationship between Affective Polarization and Democratic
Backsliding: Comparative Evidence.” Democratization 29 (4): 714–735.

Oser, Jennifer, and Marc Hooghe. 2018. “Democratic ideals and levels of political participation:
The role of political and social conceptualisations of democracy.” British Journal of Politics
and International Relations 20 (3): 711–730.

Pop-Eleches, Grigore, and Joshua A. Tucker. 2017. Communism’s Shadow: Historical Legacies
and Contemporary Political Attitudes. Princeton and Oxford: Princeton University Press.

31
Pridham, Geoffrey. 1995. “The International Context of Democratic Consolidation: Southern Eu-
rope in Comparative Perspective.” In The politics of democratic consolidation: Southern Eu-
rope in comparative perspective, edited by Richard Gunther, Nikiforos P. Diamandouros, and
Hans-Jürgen Puhle, 166–203. The new southern Europe. Baltimore, Md. and London: The
Johns Hopkins University Press.

Reykowski, Janusz. 2020. Disenchantment with Democracy: A Psychological Perspective. New

York: Oxford University Press.

Sadurski, Wojciech. 2018. “How Democracy Dies (in Poland): A Case Study of Anti-Constitutional
Populist Backsliding.” Sydney Law School Legal Studies Research Paper 18 (1).

Saikkonen, Inga A.-L., and Henrik Serup Christensen. 2023. “Guardians of Democracy or Passive
Bystanders? A Conjoint Experiment on Elite Transgressions of Democratic Norms.” Political
Research Quarterly 76 (1): 127–142.

Schedler, Andreas. 2019. “What Do We Know about Resistance to Democratic Subversion?” An-
nals of Comparative Democratization 17 (1): 4–8.

Schedler, Andreas, and Rodolfo Sarsfield. 2007. “Democrats with adjectives: Linking direct and
indirect measures of democratic support.” European Journal of Political Research 46 (5):
637–659.

Scheppele, Kim Lane. 2013. “The Rule of Law and the Frankenstate: Why Governance Checklists
Do Not Work.” Governance 26 (4): 559–562.

Simonovits, Gabor, Jennifer McCoy, and Levente Littvay. 2022. “Democratic Hypocrisy: Polar-
ized citizens support democracy-eroding behavior when their own party is in power.” Journal
of Politics 84 (3): 1806–1811.

Singh, Shane P., and Kris P. Dunn. 2013. “Veto Players, the Policy-Making Environment and the
Expression of Authoritarian Attitudes.” Political Studies 61 (1): 119–141.

Solska, Magdalena. 2020. “Democratic Erosion? One Dominant Party and Ineffective Opposition.”
Journal of Common Market Studies 58 (S1): 105–120.

32
Svolik, Milan W. 2015. “Which Democracies Will Last? Coups, Incumbent Takeovers, and the
Dynamic of Democratic Consolidation.” British Journal of Political Science 45 (4): 715–738.

. 2020. “When Polarization Trumps Civic Virtue: Partisan Conflict and the Subversion of
Democracy by Incumbents.” Quarterly Journal of Political Science 15 (1): 3–31.

Tworzecki, Hubert. 2019. “Poland: A Case of Top-Down Polarization.” The ANNALS of the Amer-
ican Academy of Political and Social Science 681 (1): 97–119.

Urbinati, Nadia. 2017. “Populism and the Principle of Majority.” In The Oxford Handbook of
Populism, edited by Cristóbal Rovira Kaltwasser, Paul Taggart, Paulina Ochoa Espejo, and
Pierre Ostiguy. Oxford: Oxford University Press.

Vasilopoulos, Pavlos, and Romain Lachat. 2018. “Authoritarianism and political choice in France.”
Acta politica 53 (4): 612–634.

Waldner, David, and Ellen Lust. 2018. “Unwelcome Change: Coming to Terms with Democratic
Backsliding.” Annual Review of Political Science 21 (5): 93–113.

Weingast, Barry R. 1997. “The political foundations of democracy and the rule of the law.” Amer-
ican Political Science Review 91 (2): 245–263.

Welzel, Christian. 2007. “Are Levels of Democracy Affected by Mass Attitudes? Testing Attain-
ment and Sustainment Effects on Democracy.” International Political Science Review 28 (4):
397–424.

. 2021. “Democratic Horizons: what value change reveals about the future of democracy.”
Democratization 28 (5): 992–1016.

Welzel, Christian, and Hans-Dieter Klingemann. 2007. Understanding Democratic Congruence:

A Demand-Supply Perspective. UC Irvine: Center for the Study of Democracy. https://eschol
arship.org/uc/item/3nb7x3qs.

. 2008. “Evidencing and Explaining Democratic Congruence: The Perspective of ’Substan-

tive’ Democracy.” World Values Research 1 (3).

33
Wuttke, Alexander, Konstantin Gavras, and Harald Schoen. 2020. “Have Europeans Grown Tired
of Democracy? New Evidence from Eighteen Consolidated Democracies, 1981–2018.” British
Journal of Political Science, 1–13.

Zhirkov, Kirill. 2022. “Estimating and Using Individual Marginal Component Effects from Con-
joint Experiments.” Political Analysis 30 (2): 236–249.

34
 Appendix to
“The Demand Side of Democratic Backsliding: How
Divergent Understandings of Democracy Shape Political
Choice”

October 19, 2023

Contents
A Survey Design 1
A.1 Candidate choice experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
A.2 Descriptive statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

B Analysis 4
B.1 Measurement model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
B.2 Candidate choice experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
B.2.1 Quality of conjoint randomization . . . . . . . . . . . . . . . . . . . . . . 6
B.2.2 Monotonicity of respondent behavior . . . . . . . . . . . . . . . . . . . . 7
B.2.3 Candidate choices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
B.2.4 Biased partisan punishment . . . . . . . . . . . . . . . . . . . . . . . . . 12
B.3 Preregistered hypotheses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
B.4 Understandings of democracy and candidate evaluations . . . . . . . . . . . . . . 16
B.5 Different samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
B.6 Party preferences and understandings of democracy . . . . . . . . . . . . . . . . . 22
B.7 Benchmarking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
A Survey Design
A.1 Candidate choice experiment
Figure A.1 shows the introduction text to the candidate choice experiment. An example choice
task is presented in Figure A.2. Table A.1 displays all attributes of candidate profiles used in the
candidate choice experiment.

Figure A.1: Introduction to the candidate choice tasks.

Figure A.2: Example choice task.

1
Attribute Levels Concept
Gender Female (weighted at 35%) Socio-demographics
Male (weighted at 65%)
Age Randomize (random integer 30-65) Socio-demographics
Partisan affiliation Law and Justice (PiS) Partisanship
Poland 2050
Civic Coalition (KO)
The Left
Confederation
Tax reform Tax reform should increase taxes for Economic trade-off
medium- and high-income households.
Tax reform should increase taxes for all
households.
Tax reform should decrease taxes for
low-income households.
Abortion legislation Abortion legislation should grant greater Women’s rights
freedom of choice to women.
Abortion legislation should protect the
unborn child’s life in all but exceptional
circumstances.
Judicial appointments Lib: Judges should be selected based on Judicial independence
cross-party consensus.
Maj: Judges should be selected by the
government.
Auth: Judges should be selected by the
leader of the ruling party.
Role of public media Lib: The role of public media is to report Media pluralism
independently on political developments.
Maj: The role of public media is to jus-
tify government policy towards the wider
public.
Auth: The role of public media is to de-
fend government policy against criticism.

Table A.1: Attribute table of the candidate choice experiment.

2
A.2 Descriptive statistics
Tables A.2 and A.3 report the targeted sample distribution according to Polish Electoral Com-
mission (for election results) and socio-demographic variables from the Polish Central Statistical
Office (GUS). The survey sample turns out to be representative of the Polish electorate with respect
to participation and vote choice in the 2020 presidential elections, region of residence, gender, ur-
ban/rural residence, and overall age distribution. However, slight differences occur in the gender
balance within age groups (see Table A.3).

Target (Percent) Sample (Percent)

Vote cast 1st round 2020 presidential election
Yes 65% 64%
No 35% 36%
Vote 1st round 2020 presidential election
Vote for Andrzej Duda 44% 44%
Vote for Rafał Trzaskowski 30% 31%
Vote for Szymon Hołownia 14% 14%
Vote for Krzysztof Bosak 7% 7%
Vote for other candidates 5% 5%
Region
CENTRALNY (CENTRAL) 20% 20%
POŁUDNIOWY (SOUTH) 21% 21%
WSCHODNI (EAST) 17% 17%
PÓŁNOCNO-ZACHODNI (NORTH-WEST) 16% 16%
POŁUDNIOWO-ZACHODNI (SOUTH-WEST) 10% 10%
PÓŁNOCNY (NORTH) 15% 15%
Gender
Male 48% 48%
Female 52% 52%
Urban/rural
Urban 60% 60%
Rural area 40% 40%

Table A.2: Target and sample distribution of participation and vote choice in 2020 presidential elections,
region of residence, gender, and urban/rural residence.

Target total Target male Target female Sample total Sample male Sample female
18-24 9% 5% 5% 9% 10% 9%
25-34 18% 9% 9% 18% 20% 17%
35-44 19% 10% 10% 19% 20% 18%
45-54 15% 8% 8% 15% 16% 14%
55+ (NET) 38% 16% 22% 38% 34% 41%

Table A.3: Target and sample distributions of age groups by gender.

3
B Analysis
B.1 Measurement model
Table B.1 reports the results of the ordered confirmatory factor analysis (CFA) for the three un-
derstandings of democracy. Figure ?? plots respondents’ factor scores for each latent variable by
partisanship.
Model
Estimate(Std.Err.) z
Factor Loadings
Liberal Understanding
L1 1.00+
L2 0.95(0.02)∗∗∗ 41.84
L3 0.87(0.02)∗∗∗ 40.42
Majoritarian Understanding
M1 1.00+
M2 0.87(0.03)∗∗∗ 29.18
M3 0.92(0.03)∗∗∗ 30.22
Authoritarian Understanding
A1 1.00+
A2 0.83(0.02)∗∗∗ 45.94
A3 1.09(0.02)∗∗∗ 51.53
Intercepts
L1 0.00+
L2 0.00+
L3 0.00+
M1 0.00+
M2 0.00+
M3 0.00+
A1 0.00+
A2 0.00+
A3 0.00+
Residual Variances
L1 0.39+
L2 0.45+
L3 0.53+
M1 0.52+
M2 0.63+
M3 0.59+
A1 0.41+
A2 0.60+
A3 0.30+
Latent Intercepts
Liberal Understanding 0.00+
Majoritarian Understanding 0.00+
Authoritarian Understanding 0.00+
Latent Variances
Liberal Understanding 0.61(0.02)∗∗∗ 32.81
Majoritarian Understanding 0.48(0.02)∗∗∗ 23.63
Authoritarian Understanding 0.59(0.02)∗∗∗ 38.15
Latent Covariances
Liberal Understanding w/Majoritarian Understanding 0.04(0.01)∗∗∗ 3.37
Liberal Understanding w/Authoritarian Understanding −0.37(0.01)∗∗∗ −31.05
Majoritarian Understanding w/Authoritarian Understanding 0.22(0.01)∗∗∗ 18.43
Thresholds
L1(1) −1.99(0.05)∗∗∗ −39.21
L1(2) −1.69(0.04)∗∗∗ −41.84
L1(3) −1.38(0.03)∗∗∗ −41.32
L1(4) −0.77(0.03)∗∗∗ −29.78
L1(5) −0.17(0.02)∗∗∗ −7.45
L1(6) 0.40(0.02)∗∗∗ 16.56
L2(1) −1.82(0.04)∗∗∗ −41.02
L2(2) −1.46(0.03)∗∗∗ −41.81

4
 L2(3) −1.14(0.03)∗∗∗ −38.53
L2(4) −0.54(0.02)∗∗∗ −21.93
L2(5) 0.02(0.02) 1.07
L2(6) 0.64(0.03)∗∗∗ 25.68
L3(1) −1.89(0.05)∗∗∗ −40.39
L3(2) −1.51(0.04)∗∗∗ −41.99
L3(3) −1.15(0.03)∗∗∗ −38.56
L3(4) −0.73(0.03)∗∗∗ −28.52
L3(5) −0.34(0.02)∗∗∗ −14.32
L3(6) 0.17(0.02)∗∗∗ 7.41
M1(1) −1.60(0.04)∗∗∗ −42.05
M1(2) −1.17(0.03)∗∗∗ −38.91
M1(3) −0.73(0.03)∗∗∗ −28.41
M1(4) −0.01(0.02) −0.44
M1(5) 0.70(0.03)∗∗∗ 27.60
M1(6) 1.43(0.03)∗∗∗ 41.68
M2(1) −1.60(0.04)∗∗∗ −42.06
M2(2) −1.19(0.03)∗∗∗ −39.17
M2(3) −0.80(0.03)∗∗∗ −30.64
M2(4) −0.16(0.02)∗∗∗ −6.74
M2(5) 0.58(0.02)∗∗∗ 23.45
M2(6) 1.28(0.03)∗∗∗ 40.44
M3(1) −1.41(0.03)∗∗∗ −41.55
M3(2) −0.88(0.03)∗∗∗ −32.85
M3(3) −0.43(0.02)∗∗∗ −17.96
M3(4) 0.21(0.02)∗∗∗ 8.78
M3(5) 0.85(0.03)∗∗∗ 31.91
M3(6) 1.54(0.04)∗∗∗ 42.04
A1(1) −0.62(0.02)∗∗∗ −24.93
A1(2) −0.24(0.02)∗∗∗ −10.26
A1(3) 0.04(0.02) 1.67
A1(4) 0.53(0.02)∗∗∗ 21.71
A1(5) 0.98(0.03)∗∗∗ 35.17
A1(6) 1.45(0.03)∗∗∗ 41.80
A2(1) −0.94(0.03)∗∗∗ −34.43
A2(2) −0.52(0.02)∗∗∗ −21.24
A2(3) −0.13(0.02)∗∗∗ −5.71
A2(4) 0.47(0.02)∗∗∗ 19.53
A2(5) 1.04(0.03)∗∗∗ 36.58
A2(6) 1.57(0.04)∗∗∗ 42.07
A3(1) −0.67(0.03)∗∗∗ −26.39
A3(2) −0.27(0.02)∗∗∗ −11.55
A3(3) 0.02(0.02) 0.96
A3(4) 0.46(0.02)∗∗∗ 19.02
A3(5) 0.89(0.03)∗∗∗ 32.95
A3(6) 1.39(0.03)∗∗∗ 41.42
Fit Indices
χ 2 (df) 284.44
CFI 0.99
TLI 0.98
RMSEA 0.06
Scaled χ 2 (df) 400.25(24)∗∗∗
+ Fixed parameter
∗ p<0.05, ∗∗ p<0.01, ∗∗∗ p<0.001

Table B.1: Measurement model (ordered confirmatory factor analysis ).

5
B.2 Candidate choice experiment
B.2.1 Quality of conjoint randomization
We regress democratic attributes on several political and sociodemographic observables to test
whether the candidate profiles have been successfully randomized. If randomization was suc-
cessful, we would expect null effects for all variables. Table B.2 shows the results of the OLS
regression models for candidates’ democratic attributes. The results suggest that political prefer-
ences and socio-demographics are not associated with the assignment of attribute levels, indicating
that the analysis is not confounded by imbalances in the treatment assignment.

Dependent variable:
Media: Liberal Media: Majoritarian Media: Authoritarian Judges: Liberal Judges: Majoritarian Judges: Authoritarian
(1) (2) (3) (4) (5) (6)
Gender (Male vs. Female) −0.004 0.004 −0.0002 −0.002 −0.001 0.003
(0.004) (0.004) (0.004) (0.004) (0.004) (0.004)
Education: (Post-)secondary −0.002 0.005 −0.003 0.007 −0.010 0.002
(0.005) (0.005) (0.005) (0.005) (0.005) (0.005)
Education: Higer 0.001 −0.002 0.001 0.007 −0.009 0.003
(0.005) (0.005) (0.005) (0.005) (0.005) (0.005)
Party: Civic Coalition 0.005 −0.006 0.001 0.002 −0.0004 −0.002
(0.005) (0.006) (0.006) (0.005) (0.005) (0.005)
Party: Poland 2050 0.001 −0.011 0.010 0.003 0.007 −0.010
(0.006) (0.006) (0.006) (0.006) (0.006) (0.006)
Party: The Left −0.007 −0.012 0.019∗ 0.004 0.001 −0.005
(0.008) (0.008) (0.008) (0.008) (0.008) (0.008)
Party: Confederation 0.005 −0.009 0.004 −0.004 0.009 −0.005
(0.008) (0.007) (0.007) (0.008) (0.008) (0.008)
Party: Would not support any 0.006 −0.010 0.004 −0.001 0.002 −0.0002
(0.007) (0.007) (0.007) (0.007) (0.007) (0.007)
Party: Don’t know 0.014∗ −0.011 −0.003 0.012∗ −0.012∗ −0.0002
(0.006) (0.006) (0.006) (0.006) (0.006) (0.006)
Age −0.0001 0.00001 0.0001 −0.0001 0.0001 0.00002
(0.0001) (0.0001) (0.0001) (0.0001) (0.0001) (0.0001)
Liberal Understanding 0.001 0.0004 −0.001 0.004 −0.003 −0.002
(0.003) (0.003) (0.003) (0.003) (0.003) (0.003)
Majoritarian Understanding 0.00004 −0.001 0.001 −0.004 0.002 0.002
(0.003) (0.003) (0.003) (0.003) (0.003) (0.003)
Authoritarian Understanding −0.0004 −0.0002 0.001 0.001 0.001 −0.002
(0.003) (0.003) (0.003) (0.003) (0.003) (0.003)
Constant 0.333∗∗∗ 0.338∗∗∗ 0.329∗∗∗ 0.334∗∗∗ 0.337∗∗∗ 0.329∗∗∗
(0.008) (0.008) (0.008) (0.008) (0.008) (0.008)
Observations 69,264 69,264 69,264 69,264 69,264 69,264
R2 0.0002 0.0002 0.0002 0.0002 0.0003 0.0001
Adjusted R2 −0.00001 −0.00003 −0.00001 0.00004 0.0001 −0.0001
Residual Std. Error (df = 69250) 0.471 0.472 0.472 0.473 0.471 0.471
Note: ∗ p<0.05; ∗∗ p<0.01; ∗∗∗ p<0.001

Table B.2: Regression of respondent characteristics on conjoint attribute level assignment. Standard
errors are clustered at the respondent level.

6
B.2.2 Monotonicity of respondent behavior
A critical assumption of conjoint designs is that respondents show similar response patterns across
choices. Figure B.1 shows the Marginal Means by choice task (1 to 12). The estimates suggest
that respondents’ rating behavior did not change substantially from the first to the last choice task.

(Gender)
Male
Female
(Age)
60−65
50−59
40−49
(Party)
The Left
Poland 2050 Choice
Law and Justice (PiS) 1

Confederation 2

Civic Coalition 3

(Party Overlap) 4

party overlap 5

no party overlap 6

(Taxes) 7
tax_rich 8
tax_less 9
tax_all 10
(Abortion) 11
abortion_lib 12
abortion_illib
(Judges)
judges_M
judges_L
judges_A
(Media)
media_M
media_L
media_A

3.0 3.5 4.0 4.5

Marginal Mean

Figure B.1: Marginal Means of candidate attributes by choices. N=2,097. 95% confidence intervals are
shown.

7
B.2.3 Candidate choices
While we use candidate ratings to determine IMCE estimates in the main paper, we also compute
the AMCE estimates for evaluations and choices between the two candidate profiles. Figure B.2
shows the AMCE with the continuous evaluation outcomes. Figures B.3 and B.4 display the choice
results for the pooled sample and by partisanship overlap, respectively.1 The choice AMCEs are
very similar to the rating AMCEs: Respondents are less likely to choose candidates who make au-
thoritarian or majoritarian claims about the judiciary and media. With respect to differences in out-
vs. co-partisan punishment, respondents are less likely to support majoritarian and authoritarian
out-partisans than co-partisans.

1. I.e., whether the respondent was shown a co-partisan candidate (overlap) or not (no overlap).

8
 Gender:
(Baseline = Male)
Female
Age:
(Baseline = 40−49)
50−59
60−65
Party:
(Baseline = Law and Justice (PiS))
Civic Coalition
Confederation
Poland 2050
The Left
Abortion policy:
(Baseline = Liberal)
Restrictive
Tax policy:
(Baseline = Increase for all)
Decrease for poor
Increase for wealthy
Judiciary:
(Baseline = Lib. democratic)
Authoritarian
Majoritarian
Media:
(Baseline = Lib. democratic)
Authoritarian
Majoritarian
Shared partisanship:
(Baseline = Not shared)
Shared
−1.0 −0.5 0.0 0.5 1.0
Effect on Candidate Rating

Figure B.2: Pooled average marginal component effects (AMCEs) of candidate attributes on candidate
ratings. N=2,097.

9
 Gender:
(Baseline = Male)
Female
Age:
(Baseline = 40−49)
50−59
60−65
Party:
(Baseline = Law and Justice (PiS))
Civic Coalition
Confederation
Poland 2050
The Left
Abortion policy:
(Baseline = Liberal)
Restrictive
Tax policy:
(Baseline = Increase for all)
Decrease for poor
Increase for wealthy
Judiciary:
(Baseline = Lib. democratic)
Authoritarian
Majoritarian
Media:
(Baseline = Lib. democratic)
Authoritarian
Majoritarian
Shared partisanship:
(Baseline = Not shared)
Shared
−1.0 −0.5 0.0 0.5 1.0
Effect on Candidate Choice

Figure B.3: Average Marginal Component Effects on candidate choice. N=2,097. 95% confidence
intervals are shown.

10
 Conditional on Conditional on
Unconditional
Partisanship = no party overlap Partisanship = party overlap
Gender:
(Baseline = Male)
Female
Age:
(Baseline = 40−49)
50−59
60−65
Party:
(Baseline = Law and Justice (PiS))
Civic Coalition
Confederation
Poland 2050
The Left
Abortion policy:
(Baseline = Liberal)
Restrictive
Tax policy:
(Baseline = Increase for all)
Decrease for poor
Increase for wealthy
Judiciary:
(Baseline = Lib. democratic)
Authoritarian
Majoritarian
Media:
(Baseline = Lib. democratic)
Authoritarian
Majoritarian
−1.0 −0.5 0.0 0.5 1.0 −1.0 −0.5 0.0 0.5 1.0 −1.0 −0.5 0.0 0.5 1.0
Effect on Candidate Choice

Figure B.4: Average Marginal Component Effects on candidate choice by partisan overlap. N=2,097.
95% confidence intervals are shown.

11
B.2.4 Biased partisan punishment
We examine to what extent our conjoint patterns also lend support for the hypothesis that partisan
polarization is the key reason driving citizens to tolerate democratic violations by co-partisans (Hy-
pothesis 2). To test whether voters punish majoritarian and authoritarian out-party candidates more
than in-party ones, Figure B.5 displays the average marginal component effects (AMCEs)2 con-
ditional on whether or not the candidate shared the respondent’s party identification.3 The results
do not lend support to the claim that voters punish in-party candidates less severely than out-party
ones; instead, respondents disapprove of majoritarian and authoritarian views in co-partisans to the
same extent as they do of similarly oriented out-party candidates. We therefore reject Hypothesis
2. This finding is in contrast to previous studies in the bi-partisan context of the United States
that see partisan polarization as the main motive driving voters to overlook democratic violations
(Graham and Svolik 2020; Simonovits, McCoy, and Littvay 2022).

Conditional on Conditional on
Unconditional
Partisanship = Not Shared Partisanship = Shared
Gender:
(Baseline = Male)
Female
Age:
(Baseline = 40−49)
50−59
60−65
Party:
(Baseline = Law and Justice (PiS))
Civic Coalition
Confederation
Poland 2050
The Left
Abortion policy:
(Baseline = Liberal)
Restrictive
Tax policy:
(Baseline = Increase for all)
Decrease for poor
Increase for wealthy
Judiciary:
(Baseline = Lib. democratic)
Authoritarian
Majoritarian
Media:
(Baseline = Lib. democratic)
Authoritarian
Majoritarian
−1.0 −0.5 0.0 0.5 1.0 −1.0 −0.5 0.0 0.5 1.0 −1.0 −0.5 0.0 0.5 1.0
Effect on Candidate Rating

Figure B.5: average marginal component effects (AMCEs) of candidate attributes conditional on
whether or not candidate shared respondents’ party orientation. N=2,097. 95% confidence intervals
are shown.

The results in the main text suggest that Polish voters punish co-partisan candidates as severely
as out-partisan, one might ask about punishment dynamics between left- and right-leaning parties
2. As specified in our preregistration, we only compute AMCEs to test Hypothesis 2 since there are too few profile
constellations with out- and in-party candidates due to the limited number of within-camp choice tasks observed for
each respondent, so that marginal component effects at the individual level would yield unreliable estimates.
3. Note that in contrast to the overall effect of shared partisanship on candidate choices, we here focus on whether
respondents lend greater or lesser weight to candidates’ statements about democracy depending on whether or not such
claims are made by a co- or out-partisan candidate.

12
and their electorates. In contrast to bipolar party systems, voting patterns in multiparty systems
in contexts of democratic backsliding are understudied. From a theoretical perspective, one might
expect voters to be more willing to switch to a different candidate in multiparty systems because
they can choose between parties that are usually ideologically more proximate than in bipolar
systems such as the United States. When voters’ preferred candidate behaves undemocratically,
defecting to an alternative, a relatively ideologically close candidate could thus be more an option
than in bipolar systems in which the rival party might be more ideologically distant.
We test this proposition by dividing the Polish party systems into two camps. The left camp
consists of Poland 2050, Civic Coalition, and The Left. The right camp is composed of the in-
cumbent Law and Justice (PiS) and Confederation. We examine how voters who prefer a party
belonging to the left camp choose and rate in choice situations in which two left-leaning candi-
dates run against each other. In a similar vein, we analyze constellations in which right-leaning
respondents are asked to choose between right-leaning political candidates.
Figure B.6 shows the results of this analysis. The results suggest that voters on the left are more
willing to punish majoritarian and authoritarian behavior when choosing between left-leaning can-
didates than right-leaning voters making choices between candidates on the right. It thus seems that
vote switching and democratic punishment occur asymmetrically in the Polish electorate: whereas
voters on the left approve consistently less of co-partisan candidates when they run against another
candidate from a left-wing party, Polish right-wing voters are less willing to switch to a different
candidate on the political right when she favors majoritarian or authoritarian democratic positions.
Thus, the moderating effect of a multi-party system on partisan polarization and the attendant tol-
erance of democratic violations appears more pronounced among left-wing rather than right-wing
voters in Poland.

13
 Gender: Gender:
(Baseline = Male) (Baseline = Male)
Female Female
Age: Age:
(Baseline = 40−49) (Baseline = 40−49)
50−59 50−59
60−65 60−65
Party:
Party:
(Baseline = Civic Coalition)
(Baseline = Law and Justice (PiS))
Poland 2050
Confederation
The Left
Abortion policy:
Abortion policy:
(Baseline = Liberal)
(Baseline = Liberal)
Restrictive
Restrictive
Tax policy:
Tax policy:
(Baseline = Increase for all)
(Baseline = Increase for all)
Decrease for poor
Decrease for poor
Increase for wealthy
Increase for wealthy
Judiciary:
Judiciary:
(Baseline = Lib. democratic) (Baseline = Lib. democratic)

Authoritarian Authoritarian

Majoritarian Majoritarian

Media: Media:

(Baseline = Lib. democratic) (Baseline = Lib. democratic)

Authoritarian Authoritarian
Majoritarian Majoritarian
−1.0 −0.5 0.0 0.5 1.0 −1.0 −0.5 0.0 0.5 1.0
Effect on Candidate Rating Effect on Candidate Rating

(a) Choices between Left-Wing Can- (b) Choices between Right-Wing

didates (N = 4412 profiles by 1012 Candidates (N = 1384 profiles by 624
respondents) respondents)

Figure B.6: AMCEs of choice subsamples. (a) Choice of left-leaning respondents between two left-
leaning candidates (Civic Coalition, Poland 2050, The Left); (b) choice of right-leaning respondents
between two right-leaning candidates (Law and Justice (PiS), Confederation). 95% confidence intervals
are shown.

14
B.3 Preregistered hypotheses
We preregistered four hypotheses for our study. This section briefly summarizes the empirical
results for each of the hypotheses.

• Respondents are more likely to prefer candidates whose democratic preferences are
congruent with their own understanding of democracy.
We report the results for this hypothesis in the main text (Table 3). In short, we find that
respondents who subscribe to a more liberal understanding of democracy vote more con-
sistently for political candidates who make liberal-democratic but not majoritarian and au-
thoritarian statements. We also find associations between a majoritarian understanding and
voting for candidates who make majoritarian and authoritarian statements about judicial ap-
pointments.

• Respondents with more liberal understandings of democracy lend greater weight to

candidates’ democratic preferences than those with more majoritarian or authoritar-
ian understandings of democracy.
We report the results for this hypothesis in the main text (Table 3). In brief, respondents
subscribing to a liberal understanding of democracy differentiate more consistently between
liberal-, majoritarian, and authoritarian-leaning candidates than those subscribing more to a
majoritarian or authoritarian understanding.

• Respondents’ party preferences predict their revealed democratic preferences.

We report the results for this hypothesis in the main text (Figure 1) as well as Table B.3.
Supporters of the PiS party are more likely to approve of majoritarian- and authoritarian-
leaning candidates than Civic Coalition voters. Confederation voters only approve more of
candidates making majoritarian statements about the judiciary than Civic Coalition voters.
Other than these differences, party electorates do not diverge in their voting preferences for
candidates with varying democratic attributes.

• Partisans of PiS are more likely to hold majoritarian understandings of democracy

than supporters of other parties.
Section B.6 discusses and Table B.8 reports the results for this hypothesis. We confirm that
PiS voters subscribe less to a liberal and more to an authoritarian and majoritarian under-
standing of democracy than Civic Coalition voters. However, Confederation voters similarly
appear to be less supportive of a liberal understanding but approve more of a majoritarian
understanding than Civic Coalition voters. Voters of Poland 2050 are slightly more majori-
tarian than the Civic Coalition electorate.

• Respondents punish candidates who correspond to their party preference less severely
for expressing democratic preferences that violate democratic standards.
Section B.2.4 evaluates this hypothesis against our findings. The results suggest similar
effect sizes for co-partisan and out-partisan candidates, which does not lend support for this
hypothesis.

15
B.4 Understandings of democracy and candidate evaluations
Table B.3 reports all coefficients for the model printed in the main text. We test for multicollinearity
in the models presented in the main text and find low but still some collinearity among the under-
standings of democracy items (Variance inflation factor (VIF) for liberal understanding = 3.71,
majoritarian understanding = 2.18, authoritarian understanding = 4.57). We remove the variable
with the highest VIF (authoritarian understanding) from the model and report the results in Table
B.4. The effects for liberal and majoritarian understandings remain the same.

Judges: Majoritarian Judges: Authoritarian Media: Majoritarian Media: Authoritarian

(1) (2) (3) (4)
Liberal understanding −0.151∗∗ −0.214∗∗∗ −0.024 −0.137∗∗
(0.051) (0.051) (0.049) (0.052)
Majoritarian understanding 0.140∗∗ 0.098∗ −0.022 0.020
(0.046) (0.047) (0.044) (0.047)
Authoritarian understanding −0.004 0.003 0.057 0.061
(0.055) (0.056) (0.053) (0.057)
Confederation 0.224∗∗ 0.141 0.056 0.114
(0.078) (0.080) (0.075) (0.080)
Law and Justice 0.346∗∗∗ 0.291∗∗∗ 0.177∗∗∗ 0.265∗∗∗
(0.055) (0.055) (0.052) (0.056)
Poland 2050 0.057 0.064 0.025 0.005
(0.059) (0.060) (0.057) (0.061)
The Left 0.013 0.066 −0.017 0.030
(0.086) (0.087) (0.083) (0.088)
No party preference/Don’t know 0.131∗ 0.096 0.063 0.125∗
(0.055) (0.056) (0.052) (0.056)
Age −0.002 −0.002 0.001 0.002
(0.001) (0.001) (0.001) (0.001)
Gender −0.042 −0.049 −0.012 −0.0003
(0.037) (0.038) (0.036) (0.038)
Education: Vocational 0.137∗ 0.084 −0.001 0.001
(0.061) (0.062) (0.058) (0.062)
Education: (Post-)Secondary 0.014 0.028 −0.045 0.016
(0.039) (0.040) (0.038) (0.040)
Income: > 6000 zl −0.135 −0.146 −0.017 −0.124
(0.074) (0.075) (0.071) (0.075)
Income: 2001-3000 zl −0.005 −0.063 0.165∗ −0.057
(0.075) (0.076) (0.072) (0.076)
Income: 3001-4000 zl −0.108 −0.096 0.021 −0.135
(0.073) (0.074) (0.070) (0.075)
Income: 4001-6000 zl −0.114 −0.150∗ 0.086 −0.061
(0.071) (0.072) (0.068) (0.073)
Financial situation: (Very) good 0.030 −0.062 0.036 0.022
(0.048) (0.049) (0.046) (0.049)
Financial situation: (Very) poor −0.097∗ −0.054 0.066 0.039
(0.046) (0.047) (0.044) (0.047)
Constant −0.215∗ −0.282∗∗ −0.403∗∗∗ −0.594∗∗∗
(0.101) (0.103) (0.097) (0.104)
Observations 2,097 2,097 2,097 2,097
R2 0.069 0.068 0.022 0.048
Adjusted R2 0.061 0.060 0.014 0.039
Note: ∗ p<0.05; ∗∗ p<0.01; ∗∗∗ p<0.001

Table B.3: The association between understandings of democracy, party preferences, sociodemographic,
and revealed democratic attitudes measured with conjoint IMCEs (OLS). Reference categories: Party
preference = Civic Coalition; Gender = female; Education = Higher; Income = <2000 zł; Financial
situation = Fair.

16
 Judges: Majoritarian Judges: Authoritarian Media: Majoritarian Media: Authoritarian
(1) (2) (3) (4)
Liberal understanding −0.148∗∗∗ −0.216∗∗∗ −0.068∗ −0.184∗∗∗
(0.028) (0.028) (0.027) (0.028)
Majoritarian understanding 0.138∗∗∗ 0.100∗∗ 0.012 0.057
(0.032) (0.032) (0.031) (0.032)
Confederation 0.224∗∗ 0.141 0.055 0.112
(0.078) (0.079) (0.075) (0.080)
Law and Justice 0.347∗∗∗ 0.291∗∗∗ 0.172∗∗ 0.259∗∗∗
(0.054) (0.055) (0.052) (0.055)
Poland 2050 0.057 0.064 0.024 0.004
(0.059) (0.060) (0.057) (0.061)
The Left 0.013 0.066 −0.020 0.027
(0.086) (0.087) (0.083) (0.088)
No party preference/Don’t know 0.131∗ 0.096 0.060 0.121∗
(0.054) (0.055) (0.052) (0.056)
Age −0.002 −0.002 0.002 0.002
(0.001) (0.001) (0.001) (0.001)
Gender −0.042 −0.048 −0.010 0.002
(0.037) (0.038) (0.036) (0.038)
Education: Vocational 0.137∗ 0.084 −0.0001 0.002
(0.061) (0.062) (0.058) (0.062)
Education: (Post-)Secondary 0.014 0.028 −0.044 0.016
(0.039) (0.040) (0.038) (0.040)
Income: > 6000 zl −0.135 −0.146 −0.017 −0.125
(0.074) (0.075) (0.071) (0.075)
Income: 2001-3000 zl −0.005 −0.063 0.166∗ −0.056
(0.075) (0.076) (0.072) (0.076)
Income: 3001-4000 zl −0.108 −0.096 0.022 −0.134
(0.073) (0.074) (0.070) (0.075)
Income: 4001-6000 zl −0.114 −0.150∗ 0.086 −0.061
(0.071) (0.072) (0.068) (0.073)
Financial situation: (Very) good 0.030 −0.062 0.038 0.023
(0.048) (0.049) (0.046) (0.049)
Financial situation: (Very) poor −0.097∗ −0.054 0.067 0.040
(0.046) (0.047) (0.044) (0.047)
Constant −0.215∗ −0.282∗∗ −0.406∗∗∗ −0.597∗∗∗
(0.101) (0.103) (0.097) (0.104)
Observations 2,097 2,097 2,097 2,097
R2 0.069 0.068 0.022 0.047
Adjusted R2 0.061 0.061 0.014 0.039
Note: ∗ p<0.05; ∗∗ p<0.01; ∗∗∗ p<0.001

Table B.4: The association between understandings of democracy (without authoritarian understanding
of democracy), party preferences, sociodemographic, and revealed democratic attitudes measured with
conjoint IMCEs (OLS). Reference categories: Party preference = Civic Coalition; Gender = female;
Education = Higher; Income = <2000 zł; Financial situation = Fair. Variance inflation factor (VIF) for
liberal understanding = 1.10 and majoritarian understanding = 1.05.

17
B.5 Different samples
In line with our pre-registration, we remove respondents who speed through the survey and fail
attention checks from our sample. We define four different samples, varying in the extent to which
respondents paid attention while participating in the survey.

• Sample 1: Full sample

We run the main analysis reported in the main paper with the full sample in Table B.5.

• Sample 2: Removing speeders and respondents with unreasonably long survey re-
sponse time
We removed all respondents who completed the survey after more than 3 hours. We then re-
moved all respondents who completed the survey earlier than one standard deviation of the
average response time (slightly less than 10 minutes). This leaves us with 2783 respondents
(of 2908 in the full sample). We report the results for this sample in Table B.6.

• Sample 3: Sample 2 minus failed attention check before the candidate choice experi-
ment
In addition to all respondents removed in step (2), we drop respondents who failed the first
attention check before the candidate choice experiment. The attention check was formulated
as follows:

Many elements shape voters’ decisions in favor of a specific candidate during an

election. We want to know which elements are most important to voters. We also
want to make sure people are paying attention to the question. To show that you
have read this much, please ignore the following question and select internet and
radio as your two answers. Which are the most important sources through which
you gain access to political news?

We then provided a list with the following items: “National newspaper,” “Television,” “Ra-
dio,” “Social media,” “Work colleagues,” “Internet,” “Friends and family,” “Regional/local
newspaper,” “I do not follow political news.” When respondents failed to select “internet”
and “radio,” we count this response behavior as a failed first attention check. All results in
the main text and Appendix are based on this sample if not noted otherwise.

• Sample 4: Sample 3 minus respondents who failed a second attention check after the
candidate choice experiment
We added a second attention check after the candidate choice experiment with the following
wording:

Changes in the global climate imply important decisions by politicians and cit-
izens. We want to know how essential you consider each element to be in the
response to global climate change. We also want to make sure people are paying
attention to the question. To demonstrate that you have read this much, please ig-
nore the following question and select [3 - Slightly important] for all items listed
below. Please indicate for each of the following items how essential you believe
it to be for an effective response to global climate change.

18
 We additionally removed all respondents who also failed to select [3 - Slightly important]
for all listed items.4 We report the results based on this sample in Table B.7.

Comparing the results for each sample suggests no differences in the association between un-
derstandings of democracy and revealed democratic attributes.

Judges: Majoritarian Judges: Authoritarian Media: Majoritarian Media: Authoritarian

(1) (2) (3) (4)
Liberal understanding −0.143∗∗ −0.184∗∗∗ −0.046 −0.182∗∗∗
(0.045) (0.046) (0.043) (0.046)
Majoritarian understanding 0.102∗ 0.073 −0.015 0.061
(0.042) (0.043) (0.040) (0.043)
Authoritarian understanding 0.021 0.031 0.048 0.029
(0.050) (0.051) (0.048) (0.050)
Confederation 0.173∗ 0.148∗ 0.014 0.035
(0.070) (0.072) (0.067) (0.071)
Law and Justice 0.348∗∗∗ 0.300∗∗∗ 0.143∗∗ 0.235∗∗∗
(0.050) (0.051) (0.047) (0.050)
Poland 2050 0.098 0.116∗ −0.001 −0.021
(0.055) (0.056) (0.052) (0.055)
The Left 0.061 0.128 −0.042 0.018
(0.078) (0.080) (0.074) (0.079)
No party preference/Don’t know 0.131∗∗ 0.115∗ 0.037 0.095
(0.050) (0.051) (0.048) (0.051)
Age −0.003∗ −0.003∗ 0.001 0.002
(0.001) (0.001) (0.001) (0.001)
Gender −0.036 −0.036 −0.010 0.005
(0.034) (0.034) (0.032) (0.034)
Education: Vocational 0.186∗∗∗ 0.138∗ 0.041 0.031
(0.053) (0.054) (0.051) (0.054)
Education: (Post-)Secondary 0.051 0.068 −0.032 0.016
(0.036) (0.037) (0.035) (0.037)
Income: > 6000 zl −0.122 −0.126 −0.052 −0.132∗
(0.066) (0.067) (0.063) (0.067)
Income: 2001-3000 zl −0.019 −0.051 0.069 −0.077
(0.066) (0.067) (0.063) (0.067)
Income: 3001-4000 zl −0.101 −0.059 −0.041 −0.137∗
(0.065) (0.066) (0.062) (0.065)
Income: 4001-6000 zl −0.091 −0.101 0.025 −0.077
(0.063) (0.065) (0.060) (0.064)
Financial situation: (Very) good 0.010 −0.048 0.010 0.029
(0.044) (0.045) (0.042) (0.045)
Financial situation: (Very) poor −0.081 −0.050 0.051 0.028
(0.041) (0.042) (0.040) (0.042)
Constant −0.195∗ −0.328∗∗∗ −0.308∗∗∗ −0.513∗∗∗
(0.091) (0.093) (0.087) (0.092)
Observations 2,522 2,522 2,522 2,522
R2 0.069 0.067 0.020 0.052
Adjusted R2 0.063 0.060 0.013 0.045
Note: ∗ p<0.05; ∗∗ p<0.01; ∗∗∗ p<0.001

Table B.5: The association between understandings of democracy, party preferences, sociodemographic,
and revealed democratic attitudes measured with conjoint IMCEs (OLS). Reference categories: Party
preference = Civic Coalition; Gender = female; Education = Higher; Income = <2000 zł; Financial
situation = Fair. Full sample.

4. The items listed were as follows: “Protecting the local ecosystem,” “Protecting natural diversity,” “Protecting
the competitiveness of the local economy,” “Protecting jobs in key industries.”

19
 Judges: Majoritarian Judges: Authoritarian Media: Majoritarian Media: Authoritarian
(1) (2) (3) (4)
Liberal understanding −0.130∗∗ −0.189∗∗∗ −0.043 −0.175∗∗∗
(0.047) (0.048) (0.045) (0.047)
Majoritarian understanding 0.104∗ 0.073 −0.015 0.058
(0.043) (0.044) (0.041) (0.044)
Authoritarian understanding 0.028 0.022 0.049 0.028
(0.051) (0.052) (0.049) (0.052)
Confederation 0.167∗ 0.155∗ 0.001 0.031
(0.073) (0.074) (0.069) (0.073)
Law and Justice 0.360∗∗∗ 0.305∗∗∗ 0.149∗∗ 0.232∗∗∗
(0.051) (0.052) (0.049) (0.052)
Poland 2050 0.093 0.107 0.001 −0.013
(0.056) (0.057) (0.054) (0.057)
The Left 0.054 0.125 −0.036 0.015
(0.080) (0.082) (0.076) (0.081)
No party preference/Don’t know 0.134∗∗ 0.116∗ 0.041 0.103∗
(0.052) (0.053) (0.049) (0.052)
Age −0.003∗ −0.003∗ 0.001 0.002
(0.001) (0.001) (0.001) (0.001)
Gender −0.040 −0.035 −0.008 −0.001
(0.035) (0.036) (0.033) (0.035)
Education: Vocational 0.179∗∗ 0.129∗ 0.030 0.037
(0.056) (0.057) (0.053) (0.056)
Education: (Post-)Secondary 0.044 0.058 −0.032 0.016
(0.037) (0.038) (0.035) (0.038)
Income: > 6000 zl −0.139∗ −0.136 −0.040 −0.124
(0.069) (0.070) (0.066) (0.069)
Income: 2001-3000 zl −0.025 −0.048 0.093 −0.064
(0.069) (0.070) (0.065) (0.069)
Income: 3001-4000 zl −0.106 −0.061 −0.028 −0.130
(0.067) (0.069) (0.064) (0.068)
Income: 4001-6000 zl −0.111 −0.118 0.034 −0.062
(0.066) (0.067) (0.063) (0.067)
Financial situation: (Very) good 0.002 −0.046 0.009 0.026
(0.046) (0.046) (0.044) (0.046)
Financial situation: (Very) poor −0.090∗ −0.053 0.054 0.028
(0.043) (0.044) (0.041) (0.044)
Constant −0.183 −0.316∗∗ −0.328∗∗∗ −0.541∗∗∗
(0.095) (0.097) (0.091) (0.096)
Observations 2,416 2,416 2,416 2,416
R2 0.069 0.065 0.020 0.049
Adjusted R2 0.062 0.058 0.013 0.042
Note: ∗ p<0.05; ∗∗ p<0.01; ∗∗∗ p<0.001

Table B.6: The association between understandings of democracy, party preferences, sociodemographic,
and revealed democratic attitudes measured with conjoint IMCEs (OLS). Reference categories: Party
preference = Civic Coalition; Gender = female; Education = Higher; Income = <2000 zł; Financial
situation = Fair. Sample without speeders.

20
 Judges: Majoritarian Judges: Authoritarian Media: Majoritarian Media: Authoritarian
(1) (2) (3) (4)
Liberal understanding −0.154∗∗ −0.223∗∗∗ −0.010 −0.149∗∗
(0.057) (0.057) (0.054) (0.057)
Majoritarian understanding 0.128∗ 0.098 −0.028 0.014
(0.050) (0.050) (0.048) (0.050)
Authoritarian understanding 0.001 0.007 0.065 0.040
(0.062) (0.062) (0.059) (0.062)
Confederation 0.254∗∗ 0.168∗ 0.120 0.161
(0.084) (0.084) (0.080) (0.084)
Law and Justice 0.363∗∗∗ 0.314∗∗∗ 0.206∗∗∗ 0.332∗∗∗
(0.061) (0.062) (0.058) (0.061)
Poland 2050 0.072 0.066 0.078 0.062
(0.065) (0.065) (0.062) (0.065)
The Left 0.046 0.086 −0.020 0.084
(0.094) (0.095) (0.090) (0.095)
No party preference/Don’t know 0.165∗∗ 0.094 0.113∗ 0.165∗∗
(0.060) (0.060) (0.057) (0.060)
Age −0.0003 −0.002 0.002 0.002
(0.001) (0.001) (0.001) (0.001)
Gender −0.072 −0.073 −0.021 −0.014
(0.041) (0.041) (0.039) (0.041)
Education: Vocational 0.132 0.079 −0.069 −0.089
(0.068) (0.069) (0.065) (0.069)
Education: (Post-)Secondary 0.009 0.033 −0.044 −0.024
(0.043) (0.043) (0.041) (0.043)
Income: > 6000 zl −0.080 −0.147 0.052 −0.061
(0.082) (0.083) (0.078) (0.083)
Income: 2001-3000 zl 0.052 −0.045 0.233∗∗ 0.042
(0.085) (0.085) (0.081) (0.085)
Income: 3001-4000 zl −0.057 −0.084 0.076 −0.079
(0.082) (0.083) (0.079) (0.083)
Income: 4001-6000 zl −0.065 −0.145 0.138 −0.016
(0.080) (0.080) (0.076) (0.080)
Financial situation: (Very) good 0.032 −0.040 0.021 0.002
(0.052) (0.052) (0.050) (0.052)
Financial situation: (Very) poor −0.110∗ −0.087 0.055 0.040
(0.051) (0.052) (0.049) (0.052)
Constant −0.330∗∗ −0.311∗∗ −0.524∗∗∗ −0.676∗∗∗
(0.112) (0.113) (0.107) (0.113)
Observations 1,745 1,745 1,745 1,745
R2 0.070 0.076 0.024 0.051
Adjusted R2 0.060 0.067 0.014 0.041
Note: ∗ p<0.05; ∗∗ p<0.01; ∗∗∗ p<0.001

Table B.7: The association between understandings of democracy, party preferences, sociodemographic,
and revealed democratic attitudes measured with conjoint IMCEs (OLS). Reference categories: Party
preference = Civic Coalition; Gender = female; Education = Higher; Income = <2000 zł; Financial
situation = Fair. Sample without speeders and 2 failed attention checks.

21
B.6 Party preferences and understandings of democracy
We evaluate to what extent party preferences are related to understandings of democracy (con-
trolling for sociodemographic variables). The results of the linear model are shown in Table B.8.
While supporters of PiS and Confederation subscribe to a lower extent to a liberal understanding
of democracy than those of the Civic Coalition, the other electorates do not differ significantly in
their liberal understanding of democracy. It is noteworthy that citizens with no party preference
or who responded with “don’t know” to the question about their preferred party tend to be less
supportive of a liberal understanding of democracy.
Moving on to an authoritarian understanding of democracy, PiS and, to a lesser extent, Poland
2050 voters subscribe more to an authoritarian understanding than Civic Coalition voters. Fur-
thermore, Confederation and PiS voters subscribe more to a majoritarian understanding than Civic
Coalition voters. Overall, party electorates seem to differ in their understandings of democracy
to some extent, yet this relationship is only moderate and—as discussed in the main text—all
electorates are characterized by heterogeneous understandings, irrespective of their ideological
orientation.

22
 Liberal Authoritarian Majoritarian
(1) (2) (3)
Confederation −0.218∗∗∗ 0.077 0.190∗∗
(0.062) (0.054) (0.064)
Law and Justice (PiS) −0.116∗∗ 0.265∗∗∗ 0.154∗∗∗
(0.043) (0.037) (0.044)
Poland 2050 −0.060 0.081∗ 0.075
(0.047) (0.041) (0.049)
The Left 0.005 −0.060 −0.088
(0.068) (0.059) (0.071)
No party preference/Don’t know −0.215∗∗∗ 0.051 0.141∗∗
(0.043) (0.037) (0.045)
Age 0.008∗∗∗ 0.002∗∗ −0.004∗∗∗
(0.001) (0.001) (0.001)
Gender 0.064∗ 0.022 −0.001
(0.029) (0.026) (0.031)
Education: Vocational −0.109∗ 0.138∗∗∗ 0.178∗∗∗
(0.048) (0.042) (0.050)
Education: (Post-)Secondary 0.004 0.066∗ 0.049
(0.031) (0.027) (0.032)
Income: > 6000 zl 0.224∗∗∗ −0.033 −0.193∗∗
(0.058) (0.051) (0.061)
Income: 2001-3000 zl 0.034 −0.055 −0.040
(0.059) (0.052) (0.062)
Income: 3001-4000 zl 0.087 −0.014 −0.065
(0.058) (0.051) (0.060)
Income: 4001-6000 zl 0.181∗∗ −0.010 −0.157∗∗
(0.056) (0.049) (0.059)
Financial situation: (Very) good 0.001 −0.022 0.009
(0.038) (0.033) (0.040)
Financial situation: (Very) poor −0.127∗∗∗ −0.031 0.094∗
(0.036) (0.032) (0.038)
Constant −0.320∗∗∗ −0.204∗∗ 0.069
(0.080) (0.070) (0.083)
Observations 2,097 2,097 2,097
R2 0.094 0.047 0.051
Adjusted R2 0.087 0.040 0.044
Residual Std. Error (df = 2081) 0.654 0.568 0.678
Note: ∗ p<0.05; ∗∗ p<0.01; ∗∗∗ p<0.001

Table B.8: The association between party preferences, sociodemographic, and understandings of democ-
racy (OLS). Reference categories: Party preference = Civic Coalition; Gender = female; Educa-
tion = Higher; Income = <2000 zł; Financial situation = Fair.

23
B.7 Benchmarking
In addition to the Understandings of Democracy Battery, we also asked respondents in the survey
how democratic they consider the democracy attribute levels to be.5 This helps us evaluate whether
respondents diverge in their assessment of the attributes which we used for the candidate choice
experiment. As Tables B.9 and B.10 reveal, different understandings of democracy are associated
with the rating of the democracy attributes in the choice experiment. This suggests that voters dif-
fer in their views of what democracy constitutes. For instance, voters subscribing to a majoritarian
understanding of democracy consider majoritarian reforms of the judiciary or media more demo-
cratic than voters rejecting such an understanding of democracy. Overall, the evidence suggests
that Polish citizens have divergent understandings of democracy and evaluate liberal, majoritarian,
and authoritarian statements correspondingly.
We test for multicollinearity in the models presented in the main text and find low but still
some collinearity among the understandings of democracy items (Variance inflation factor (VIF)
for liberal understanding = 3.71, majoritarian understanding = 2.18, authoritarian understand-
ing = 4.57). We remove the variable with the highest VIF (authoritarian understanding) from the
model and report the results in Tables B.11 and B.12. The effects for liberal and majoritarian
understandings remain the same.
5. The question wording was: “There are different views of what can be considered democratic and what not. How
democratic do you think each of the following statements is?” The scale ranged from “Not at all democratic” (1) to
“extremely democratic” (7).

24
 Judges: Liberal Judges: Majoritarian Judges: Authoritarian
(1) (2) (3)
Liberal understanding 0.709∗∗∗ −0.358∗∗∗ −0.436∗∗∗
(0.075) (0.094) (0.084)
Majoritarian understanding 0.015 0.437∗∗∗ 0.198∗∗
(0.068) (0.085) (0.076)
Authoritarian understanding 0.137 −0.065 0.145
(0.082) (0.103) (0.091)
Confederation −0.306∗∗ 0.676∗∗∗ 0.113
(0.115) (0.145) (0.129)
Law and Justice −0.383∗∗∗ 1.525∗∗∗ 1.077∗∗∗
(0.081) (0.101) (0.090)
Poland 2050 0.180∗ 0.091 −0.101
(0.088) (0.110) (0.098)
The Left 0.081 0.013 −0.104
(0.127) (0.160) (0.142)
No party preference/Don’t know −0.196∗ 0.401∗∗∗ 0.078
(0.081) (0.101) (0.090)
Age 0.012∗∗∗ −0.015∗∗∗ −0.011∗∗∗
(0.002) (0.002) (0.002)
Gender 0.006 −0.163∗ −0.296∗∗∗
(0.055) (0.069) (0.061)
Education: Vocational −0.261∗∗ 0.291∗∗ 0.419∗∗∗
(0.090) (0.113) (0.100)
Education: (Post-)Secondary −0.141∗ 0.180∗ 0.262∗∗∗
(0.058) (0.073) (0.065)
Income: > 6000 zl 0.290∗∗ −0.211 −0.274∗
(0.109) (0.137) (0.122)
Income: 2001-3000 zl 0.190 −0.057 0.023
(0.110) (0.139) (0.123)
Income: 3001-4000 zl 0.147 −0.150 −0.150
(0.108) (0.136) (0.121)
Income: 4001-6000 zl 0.094 −0.220 −0.251∗
(0.105) (0.132) (0.117)
Financial situation: (Very) good 0.166∗ −0.130 −0.111
(0.071) (0.090) (0.080)
Financial situation: (Very) poor 0.057 −0.030 −0.045
(0.068) (0.086) (0.076)
Constant 4.801∗∗∗ 3.277∗∗∗ 2.766∗∗∗
(0.150) (0.188) (0.167)
Observations 2,097 2,097 2,097
R2 0.188 0.213 0.232
Adjusted R2 0.181 0.206 0.226
Residual Std. Error (df = 2078) 1.214 1.528 1.358
Note: ∗ p<0.05; ∗∗ p<0.01; ∗∗∗ p<0.001

Table B.9: The association between understandings of democracy, party preference, and so-
ciodemocraphics, and benchmarking questions (OLS). Reference categories: Party preference = Civic
Coalition; Gender = female; Education = Higher; Income = <2000 zł; Financial situation = Fair.

25
 Media: Liberal Media: Majoritarian Media: Authoritarian
(1) (2) (3)
Liberal understanding 0.746∗∗∗ 0.005 −0.320∗∗∗
(0.071) (0.094) (0.085)
Majoritarian understanding −0.019 0.263∗∗ 0.212∗∗
(0.064) (0.085) (0.077)
Authoritarian understanding 0.032 0.210∗ 0.337∗∗∗
(0.078) (0.103) (0.093)
Confederation −0.117 0.277 0.239
(0.110) (0.146) (0.131)
Law and Justice −0.464∗∗∗ 1.233∗∗∗ 1.450∗∗∗
(0.077) (0.102) (0.092)
Poland 2050 0.150 0.296∗∗ −0.096
(0.083) (0.111) (0.100)
The Left 0.076 −0.042 −0.116
(0.121) (0.160) (0.144)
No party preference/Don’t know −0.111 0.241∗ 0.196∗
(0.077) (0.102) (0.092)
Age 0.007∗∗∗ 0.002 −0.006∗∗
(0.002) (0.002) (0.002)
Gender 0.141∗∗ 0.012 −0.044
(0.052) (0.069) (0.062)
Education: Vocational −0.407∗∗∗ 0.654∗∗∗ 0.309∗∗
(0.085) (0.113) (0.102)
Education: (Post-)Secondary −0.112∗ 0.359∗∗∗ 0.100
(0.055) (0.073) (0.066)
Income: > 6000 zl 0.220∗ 0.031 −0.122
(0.104) (0.137) (0.124)
Income: 2001-3000 zl −0.029 0.173 0.120
(0.105) (0.139) (0.126)
Income: 3001-4000 zl 0.036 0.016 −0.008
(0.103) (0.136) (0.123)
Income: 4001-6000 zl 0.137 0.115 −0.002
(0.100) (0.133) (0.119)
Financial situation: (Very) good 0.095 −0.014 −0.121
(0.068) (0.090) (0.081)
Financial situation: (Very) poor −0.025 −0.019 0.011
(0.065) (0.086) (0.077)
Constant 5.497∗∗∗ 2.886∗∗∗ 2.426∗∗∗
(0.143) (0.189) (0.170)
Observations 2,097 2,097 2,097
R2 0.236 0.151 0.270
Adjusted R2 0.229 0.144 0.264
Residual Std. Error (df = 2078) 1.156 1.532 1.382
Note: ∗ p<0.05; ∗∗ p<0.01; ∗∗∗ p<0.001

Table B.10: The association between understandings of democracy, party preference, and socio-
demographics, and benchmarking questions (OLS). Reference categories: Party preference = Civic
Coalition; Gender = female; Education = Higher; Income = <2000 zł; Financial situation = Fair.

26
 Judges: Liberal Judges: Majoritarian Judges: Authoritarian
(1) (2) (3)
Liberal understanding 0.604∗∗∗ −0.308∗∗∗ −0.547∗∗∗
(0.041) (0.051) (0.046)
Majoritarian understanding 0.097∗ 0.398∗∗∗ 0.285∗∗∗
(0.047) (0.059) (0.052)
Confederation −0.309∗∗ 0.677∗∗∗ 0.110
(0.115) (0.145) (0.129)
Law and Justice −0.396∗∗∗ 1.531∗∗∗ 1.064∗∗∗
(0.080) (0.101) (0.090)
Poland 2050 0.177∗ 0.092 −0.104
(0.088) (0.110) (0.098)
The Left 0.075 0.017 −0.111
(0.127) (0.160) (0.142)
No party preference/Don’t know −0.203∗ 0.405∗∗∗ 0.070
(0.081) (0.101) (0.090)
Age 0.012∗∗∗ −0.016∗∗∗ −0.011∗∗∗
(0.002) (0.002) (0.002)
Gender 0.010 −0.166∗ −0.291∗∗∗
(0.055) (0.069) (0.061)
Education: Vocational −0.259∗∗ 0.290∗ 0.420∗∗∗
(0.090) (0.113) (0.100)
Education: (Post-)Secondary −0.140∗ 0.180∗ 0.264∗∗∗
(0.058) (0.073) (0.065)
Income: > 6000 zl 0.290∗∗ −0.211 −0.274∗
(0.109) (0.137) (0.122)
Income: 2001-3000 zl 0.193 −0.058 0.026
(0.110) (0.139) (0.123)
Income: 3001-4000 zl 0.148 −0.151 −0.148
(0.108) (0.136) (0.121)
Income: 4001-6000 zl 0.092 −0.220 −0.253∗
(0.105) (0.132) (0.117)
Financial situation: (Very) good 0.169∗ −0.132 −0.108
(0.071) (0.090) (0.080)
Financial situation: (Very) poor 0.059 −0.031 −0.043
(0.068) (0.085) (0.076)
Constant 4.794∗∗∗ 3.280∗∗∗ 2.758∗∗∗
(0.150) (0.188) (0.167)
Observations 2,097 2,097 2,097
R2 0.187 0.213 0.231
Adjusted R2 0.181 0.206 0.225
Note: ∗ p<0.05; ∗∗ p<0.01; ∗∗∗ p<0.001

Table B.11: The association between understandings of democracy (without authoritarian understand-
ing), party preference, and sociodemocraphics, and benchmarking questions (OLS). Reference cate-
gories: Party preference = Civic Coalition; Gender = female; Education = Higher; Income = <2000
zł; Financial situation = Fair. Variance inflation factor (VIF) for liberal understanding = 1.10 and ma-
joritarian understanding = 1.05.

27
 Media: Liberal Media: Majoritarian Media: Authoritarian
(1) (2) (3)
Liberal understanding 0.722∗∗∗ −0.156∗∗ −0.578∗∗∗
(0.039) (0.051) (0.047)
Majoritarian understanding −0.0005 0.389∗∗∗ 0.413∗∗∗
(0.045) (0.059) (0.054)
Confederation −0.118 0.272 0.231
(0.110) (0.146) (0.132)
Law and Justice −0.467∗∗∗ 1.214∗∗∗ 1.418∗∗∗
(0.076) (0.101) (0.091)
Poland 2050 0.149 0.292∗∗ −0.103
(0.083) (0.111) (0.100)
The Left 0.074 −0.052 −0.133
(0.121) (0.160) (0.145)
No party preference/Don’t know −0.113 0.229∗ 0.178
(0.077) (0.102) (0.092)
Age 0.007∗∗∗ 0.002 −0.005∗
(0.002) (0.002) (0.002)
Gender 0.142∗∗ 0.019 −0.032
(0.052) (0.069) (0.063)
Education: Vocational −0.406∗∗∗ 0.657∗∗∗ 0.314∗∗
(0.085) (0.113) (0.102)
Education: (Post-)Secondary −0.112∗ 0.362∗∗∗ 0.104
(0.055) (0.073) (0.066)
Income: > 6000 zl 0.220∗ 0.031 −0.123
(0.104) (0.138) (0.124)
Income: 2001-3000 zl −0.029 0.177 0.127
(0.105) (0.139) (0.126)
Income: 3001-4000 zl 0.036 0.018 −0.004
(0.103) (0.137) (0.123)
Income: 4001-6000 zl 0.137 0.112 −0.006
(0.100) (0.133) (0.120)
Financial situation: (Very) good 0.095 −0.010 −0.114
(0.068) (0.090) (0.081)
Financial situation: (Very) poor −0.025 −0.016 0.016
(0.065) (0.086) (0.078)
Constant 5.495∗∗∗ 2.875∗∗∗ 2.408∗∗∗
(0.142) (0.189) (0.171)
Observations 2,097 2,097 2,097
R2 0.236 0.149 0.265
Adjusted R2 0.230 0.142 0.259
Note: ∗ p<0.05; ∗∗ p<0.01; ∗∗∗ p<0.001

Table B.12: The association between understandings of democracy (without authoritarian understand-
ing), party preference, and socio-democraphics, and benchmarking questions (OLS). Reference cate-
gories: Party preference = Civic Coalition; Gender = female; Education = Higher; Income = <2000
zł; Financial situation = Fair. Variance inflation factor (VIF) for liberal understanding = 1.10 and ma-
joritarian understanding = 1.05.

References
Graham, Matthew H., and Milan W. Svolik. 2020. “Democracy in America? Partisanship, Polariza-
tion, and the Robustness of Support for Democracy in the United States.” American Political
Science Review 114 (2): 392–409.
Simonovits, Gabor, Jennifer McCoy, and Levente Littvay. 2022. “Democratic Hypocrisy: Polar-
ized citizens support democracy-eroding behavior when their own party is in power.” Journal
of Politics 84 (3): 1806–1811.

28