Theory

As several liberal democracies around the world have recently grappled with internal political tension, political science research started to look at the factors behind the onset of democratic backsliding. Over the last few years, many works have taken a keen interest in studying why citizens and elites oftentimes tolerate undemocratic practices (Levitsky and Ziblatt, Reference Levitsky and Ziblatt2019; Graham and Svolik, Reference Graham and Svolik2020).

This tolerance represents a key ingredient in the minimalist notion of democracy (Przeworski, Reference Przeworski1991; Svolik, Reference Svolik2019, Reference Svolik2020), which posits that beyond the role of constitutional institutions, both the public and the elites are themselves essential democratic checks. On the one hand, voters can prevent processes of democratic backsliding by holding politicians accountable for not abiding to democratic norms. On the other, elites restrain themselves from displaying undemocratic attitudes, as they expect such behavior will be penalized by the public. Ultimately, both voters and political elites reach an equilibrium in which undemocratic practices – and the temptation to resort to them – are appeased, thereby setting in motion or consolidating the democratic system.

Previous research has looked at which factors can jeopardize this equilibrium. One of the most important factors traditionally highlighted by scholars has been ideological distance. Thus, when voters or parties are far away from each other on ideological grounds – traditionally measured in the European context using the left-right dimension – the expectation is that election losers have much to lose, election winners may be tempted to (unilaterally) impose policies that trigger conflict, and that agreement between the different sides is unlikely. Despite seminal works did not reach a definitive conclusion on whether polarization triggered democratic erosion (Daalder, Reference Daalder1971; Sartori, Reference Sartori1976; Broockman et al., Reference Broockman, Kalla and Westwood2023), over the past few years, and in the wake of recent contexts of intense political conflict, the interest has once again switched to whether voters and elites may dispense undemocratic practices when the government, the opposition or the governed are far apart in some fundamental dimensions. For instance, as recently shown by Canen et al. (Reference Canen, Kendall and Trebbi2021), ideological distance accounts for a significant share of the overall levels of polarization in the US context. Similarly, recent work from Avramovska et al. (Reference Avramovska, Lutz, Milačić and Svolik2022) using survey data from seven European countries shows that, despite democracy being favored by a majority of respondents, citizens are able to trade democratic values for their preferred identity policies.

Besides ideological polarization, a vibrant strand of research has highlighted the role of affective polarization in explaining certain undemocratic practices. Defined as the animosity (or dislike) between the in-group and the relevant out-group, affective polarization, compared to ideological polarization, constitutes a psychological attitude rooted in affect and identity. While it often refers to partisan identity, affective polarization can occur also among different types of socio-political groups. And it may or may not be correlated with the traditional left-right divide (Iyengar et al., Reference Iyengar, Lelkes, Levendusky, Malhotra and Westwood2019; Banda and Cluverius, Reference Banda and Cluverius2018; Hobolt et al., Reference Hobolt, Leeper and Tilley2021; Dias and Lelkes, Reference Dias and Lelkes2022). As Schedler (Reference Schedler2023, p. 351) states, ‘partisan animosity and policy disagreement are two independent dimensions that may or may not be related’. Empirical studies have detected that over recent years, and across many countries, levels of affective polarization have been high and often increasing over time (Levendusky, Reference Levendusky2018; Gidron et al., Reference Gidron, Adams and Horne2020; Wagner, Reference Wagner2021; Dimant, Reference Dimant2024).

With affective polarization gaining ground around the world, some authors have examined whether it may be connected to democratic undoing. One of the assumptions in these works is that affective polarization is a major driver of tolerance toward undemocratic practices (Gidron et al., Reference Gidron, Adams and Horne2023). In other words, when the animosity between political groups is high, some individuals may be more likely to justify undemocratic practices, even when most citizens and politicians value democracy (Grillo and Prato, Reference Grillo and Prato2023). Schedler (Reference Schedler2023) argues that in contexts of heightened polarization and conflict, the rival camps tend to perceive and portray political adversaries as enemies (democratic enemyopia and enemography). Puryear et al. (Reference Puryear, Kubin, Schein, Bigman, Ekstrom and Gray2024) find that citizens often overestimate political opponents’ willingness to accept even the most basic moral wrongs, which fosters dehumanization. Expanding on this, Gonzalez-Ocantos and Meléndez (Reference Gonzalez-Ocantos and Meléndez2025) provide empirical evidence from Latin America demonstrating that partisans construct regime stereotypes by depicting their own group as the sole defenders of democracy while viewing opponents as inherently authoritarian ‘foes’. When both camps perceive the opposing side as a risk to democracy, it provides fertile ground for justifying exclusionary or illiberal policies as a necessary defense of the system, and an increased tolerance of one’s own leaders’ deviation from basic democratic norms may ensue.

Using data for 53 countries, Orhan (Reference Orhan2022) concludes that affective polarization emerges as a primary factor driving support for undemocratic positions. Likewise, research using experimental data reveals that support for norm-eroding policies increases when an individual’s own party is in power (Simonovits et al., Reference Simonovits, McCoy and Littvay2022), when the own party benefits from it (Gidengil et al., Reference Gidengil, Stolle and Bergeron-Boutin2022), and when individuals think that the opposing party is willing to subvert democracy (Braley et al., Reference Braley, Lenz, Adjodah, Rahnama and Pentland2023). Finally, Aarslew (Reference Aarslew2022) further shows that individuals keep supporting their party because revelations of malpractice do not change the views they hold on the opposing parties.

Despite existing evidence showing that affective polarization leads to tolerance or support for democratic undoing, recent work presents a more complex or mixed picture. Some authors call into question that affective polarization influences political choices (Broockman et al., Reference Broockman, Kalla and Westwood2023). Orr et al. (Reference Orr, Fowler and Huber2023) alert about the difficulty of manipulating perceptions of party identity without manipulating beliefs about policy agreement, and show that partisans care much more about policy agreement than they do about party loyalty. Similarly, Voelkel et al. (Reference Voelkel, Chu, Stagnaro, Mernyk, Redekopp, Pink, Druckman, Rand and Willer2023) manipulate affective polarization to reduce it, but this intervention does not shrink support for undemocratic candidates.

Of course, the connection between polarization and tolerance toward undemocratic practices of the politicians of the voters’ own camp can facilitate processes of democratic backsliding, but the link is far from automatic. Voters’ polarization can make them more tolerant toward certain undemocratic practices, and this may change politicians’ structure of incentives. The same may be true for politicians: affective polarization among the elites can make it more likely that they attempt to engage, promote, or tolerate democratic undoing. But the extent to which this will translate into actual backsliding is a whole different question, that depends on institutional and political conditions (Somer and McCoy, Reference Somer, McCoy, Torcal and Harteveld2024), given that democratic backsliding is a system-level outcome (Waldner and Lust, Reference Waldner and Lust2018; Grillo et al., Reference Grillo, Luo, Nalepa and Prato2024).

Given the question marks previous studies have left behind, we complement the literature in three fundamental ways. First, and building upon the classical literature on ideological polarization as well as the more recent research on affective polarization, our objective is to determine which of the types of polarization, if any, drives individuals’ willingness to trade democratic principles. It is still not entirely clear from previous studies if support for undemocratic practices undertaken by one’s party is less likely when individuals are distant on the traditional dimensions of competition – i.e, the left-right – or on the affective dimensions. In this regard, our study follows a similar approach to a recent contribution by Frederiksen (Reference Frederiksen2022), who shows that the interplay between partisanship, undemocratic behavior, and policy agreement is important in order to understand people’s willingness to support undemocratic practices.

Second, and compared to previous works that have considered arguably extreme measures of democratic undoing, we focus on a crucial element of democracy largely neglected until now: the loser’s consent. This is a key democratic norm according to which election losers accept their defeat without questioning the legitimacy of the democratic process (Przeworski, Reference Przeworski1991; Nadeau and Blais, Reference Nadeau and Blais1993; Gidron et al., Reference Gidron, Adams and Horne2020). As famously put by Przeworski (Reference Przeworski1991), accepting defeat is crucial in preserving democratic regimes. Additionally, and in order to transcend the context of US politics and bring it closer to the context of multiparty systems (Dimant, Reference Dimant2024), we adapt the undemocratic behavior of the candidates in our experimental design to capture a relatively mild defiance of the loser’s consent. Instead of overt rejection of the result of a democratic election or an outright attempt of institutional manipulation, we conceptualize the losers’ consent by distinguishing political candidates that unambiguously accept an electoral defeat from those that question the rules of the game after losing an election.

The third contribution of the paper is related to the debate on to what extent polarization and democratic backsliding are elite-driven processes. We use the same research design and apply it both to a sample of regular citizens and a sample of elected politicians in order to assess to what extent politicians behave in a distinct way when evaluating undemocratic behavior against ideological distance or affective polarization.

Finally, to investigate the effects of democratic norms – as captured by a candidate’s loser’s consent – on voters’ choices with respect to other candidate’s attributes, we focus our interest on the case of Catalonia (Spain), a region that has experienced relevant political turbulence during the last decade. In Catalonia, political competition is structured along a bidimensional axis, combining the traditional left-right and the territorial debate (against and in favor of Catalan secession) (Rodon and Guinjoan, Reference Rodon and Guinjoan2022). Hence, one might argue that this scenario constitutes a difficult case to find an effective democratic check from the public (in either of the two dimensions). In addition, ideological distance and affective polarization could differently affect citizens’ evaluation of the behavior of politicians across the two dimensions (see Balcells et al., Reference Balcells, Fernández-Albertos and Kuo2021). All in all, we claim that, even in highly polarized contexts, ideological proximity and affective proximity are likely to capture different voters’ considerations of meaningful policy distances, on the one hand, and voters’ sense of belonging to in-groups, on the other. The distinct psychological underpinnings of both processes – one rooted in cognition, the other one in affection – may have different implications for citizens’ tolerance toward undemocratic politicians.

More specifically, and as for affective proximity, we expect that a voter with a strong positive in-group bias and out-group animosity will be more tolerant toward undemocratic practices, more so if this action comes from a politician belonging to his or her same social group. Affective proximity may activate the exoneration of democratic undoing practices if they come from candidates with whom voters exhibit a form of in-group bias. Indeed, in-group bias activates motivated reasoning, and people tend to process the information in terms that are more positive for their in-group members and more negative for the out-group. Hence, we can expect that strong feelings toward certain groups or parties lead citizens to regard more positively the behavior from the politicians that represent the group they identify with and more negatively that of the out-group. Affective polarization would therefore pave the way for lower satisfaction with democracy (Wagner, Reference Wagner2021) and democratic backsliding (Orhan, Reference Orhan2022).

However, the same must not necessarily be true for ideological (i.e., positional) proximity. Net of the affective component (see Orr et al., Reference Orr, Fowler and Huber2023), the effect of the distance on the left-right dimension or any other relevant dimension on tolerance of undemocratic behavior is less clear. On the one hand, a rational calculus may drive voters to trade the neglect of certain democratic norms in exchange for policy proximity. One may even dislike undemocratic behavior, but dislike even more a politician that is far from her position and hence resolve the trade-off in a way that is dangerous for democracy. On the other hand, a democratic citizen may be interested in keeping under democratic check candidates with whom she shares policy positions, and may even tolerate some ideological distance in order to preserve democracy.

Overall, we conceive two potential empirical scenarios. A pessimistic scenario would display a society characterized by a high elasticity between democratic norms and policy preferences, in which voters trade off basic democratic norms in exchange of their preferred representatives in office delivering their preferred policies. The opposite, more optimistic, scenario would be characterized by a low elasticity between democratic norms and voters’ policy preferences: individuals are not willing to trade off democratic principles for their preferred policy positions and, consequently, individuals are less likely to support calling the election defeat into question. Interestingly, the Catalan case offers the opportunity to estimate and compare these elasticities for the two main dimensions of political competition, the left-right and the territorial dimensions (Bassan-Nygate and Weiss, Reference Bassan-Nygate and Weiss2022; Rodon and Guinjoan, Reference Rodon and Guinjoan2022).

The conventional wisdom in the literature states that elites are more ideologically distant than citizens. This is a consequence of elites being, for obvious reasons, more politically aware, sophisticated, and, ultimately, more ideologically constrained than the mass public (Fiorina and Abrams, Reference Fiorina and Abrams2008). Recent work also points to the fact that politicians tend to be more affectively polarized than citizens (Enders, Reference Enders2021). Moreover, research on democratic backsliding tends to portray it as a fundamentally elite-driven process. Even when public opinion supports backsliding, this is often regarded as a consequence of elites’ positions (Cinar et al., Reference Cinar2022).

All these factors would lead us to more pessimistic expectations for elites than for ordinary citizens. At the same time, nonetheless, it is true that political elites are, on average, better educated and have larger resources than the median citizen (Manin, Reference Manin1997), thus being less likely to hold undemocratic attitudes. An additional counter-balancing force may be related to social desirability: the more sophisticated politicians may be more aware and willing to conceal their tolerance of undemocratic attitudes. Although our research design, based on a conjoint experiment, has been shown to mitigate social desirability bias (Horiuchi et al., Reference Horiuchi, Markovich and Yamamoto2021), this may still be a concern. Hence, given the conflicting arguments for the relationship between elites and democratic undoing, our purpose in this last section is mostly exploratory.

The case

There are good reasons to test our expectations using Catalonia as a case study. Catalonia has a highly fragmented party system that is organized around two dimensions: the left-right and the territorial dimensions. In Catalonia, these two dimensions not only structure party competition but also interact in a non-orthogonal way (Liñeira et al., Reference Liñeira, Muñoz and Rico2021). In a pattern that also largely applies to Spain, traditionally the left has been associated with more pro-decentralization positions, while conservative politicians and voters have largely favored a more centralized view of the territorial distribution of power. This alignment between territorial and ideological dimensions has long shaped party competition and reinforced ideological polarization (Madariaga and Riera, Reference Madariaga and Riera2022), a dynamic that has intensified in recent years, particularly among right-wing forces (León, Reference León2015).

During the last decade, there has been a period of increasing political polarization in the territorial dimension, often known as the second dimension. In 2010, a ruling by the Spanish Constitutional Court limited the level of self-government that Catalonia had reached in the statute of autonomy that had been approved in 2006. As a consequence of that ruling, a substantial segment of the public opinion and the political elite moved from supporting increased self-government within Spain to support self-determination and, eventually, independence (Guinjoan and Rodon, Reference Guinjoan, Rodon and Cuadras-Morató2016).

The events that occurred in the upcoming years, with Catalan political leaders and civic society attempting to find a way to hold an official referendum on secession, intensified the division along the territorial debate. In 2017, a referendum deemed illegal by the Spanish courts was held, and the Spanish riot police attempted to prevent it by clashing with voters. As a consequence, the members of the Catalan government were either imprisoned or left the country, and Catalan autonomy was temporarily suspended (c.f. Rodon and Guinjoan, Reference Rodon and Guinjoan2022). In June 2021, the Spanish Council of Ministers (a left-wing coalition of PSOE and Podemos with parliamentary support of a Catalan pro-independence party) approved a pardon for which the Catalan political leaders in prison were released. The Catalan independence process intensified polarization and allegedly facilitated the rise of the extreme right-wing party Vox, a political formation characterized by its strongly centralist, authoritarian, anti-immigration, and anti-feminist platform (Rama et al., Reference Rama, Zanotti, Turnbull-Dugarte and Santana2021). Overall, the territorial question did not only polarize the electorate along identity lines, but also reconfigured the ideological axis, making the defense of national unity largely inseparable from conservative stances on other issues (Torcal and Comellas, Reference Torcal and Comellas2022).

Such a polarized context is arguably a difficult case for citizens’ punishment of undemocratic behavior. Even if the territorial debate had historically articulated party preferences in Catalonia, the events that occurred between 2012 and 2019 increased the salience of such debate and forced many people to take sides. We believe that such a configuration may contribute to voters’ willingness to tolerate undemocratic attitudes: this type of political conflict may spur in-group–out-group dynamics as defined by the social identity theory (Tajfel et al., Reference Tajfel1986), which are a fertile ground for highly inelastic preferences. Voters may prefer a candidate with whom they share positions in the national dimension regardless of other considerations, including democratic undoing.

Research design

In order to test our expectations, we designed a conjoint experiment in which we presented our respondents with five pairs of candidates with randomly varying attributes. For each pair, respondents had to pick the candidate they would vote for in a forced choice. The experiment also included a battery of questions that tap on the socio-demographic characteristics (gender and language) of the candidate. Most crucially, we ask respondents – before conducting the experiment–to report their position on the two relevant dimensions of competition (left-right and pro-/anti-independence) as well as their democratic attitudes.

The choice of a conjoint experimental design to test our hypotheses is aligned with recent literature in political science, as such designs have become the workhorse method for studying multidimensional preferences. Conjoint experiments provide an effective, robust, and often externally valid (Hainmueller et al., Reference Hainmueller, Hopkins and Yamamoto2014) framework to test the causal effects of multiple attributes (and their interactions) on the choice and evaluation of candidates, policies, and other objects of interest. The literature has identified several limitations, specifically those related to the use of the Average Marginal Component Effect (AMCE) as the main or sole quantity of interest (Hobolt et al., Reference Hobolt, Leeper and Tilley2021), the presence of unrealistic profiles (De la Cuesta et al., Reference De la Cuesta, Egami and Imai2022), and the fact that the written presentation of profiles (often in tabular format) may limit the external validity of the experiments (Bansak et al., Reference Bansak, Hainmueller, Hopkins and Yamamoto2021; Abramson et al., Reference Abramson, Koçak and Magazinnik2022). In our design and analysis, as we discuss below, we seek to mitigate these potential problems to the greatest extent feasible.

We fielded our experiment in August 2021, which was applied to two different samples. First, a sample of 1,001 individuals recruited from a commercial online panel using education, gender, and age quotas. Second, a sample of politicians that we collected directly. We selected all the representatives of Catalan districts at the Catalan, Spanish, and European Parliaments, as well as members of local councils of cities and towns over 50,000 inhabitants. We invited 1,437 politicians and received 273 responses.Footnote ¹

In the conjoint, we operationalize candidates’ (un)democratic behavior through the idea of the loser’s consent: the candidate is presented as either willing to accept an election defeat of his/her party if that happens or, rather, arguing that rules should be changed in order to make it easier for his/her party to govern. Table 1 shows the five attributes and the corresponding levels.

Table 1.

Conjoint design

Table 1. Long description

A table detailing attributes and their corresponding levels for a conjoint design. The table has five rows and columns. The columns are labeled ‘Attribute’ and ‘Levels’. The attributes and their levels are as follows: Row 1: Attribute 1. Language. Levels: Usually speaks Catalan, Usually speaks Catalan and Spanish, Usually speaks Spanish. Row 2: Attribute 2. Sex. Levels: Is a man, Is a woman. Row 3: Attribute 3. Left-right position. Levels: Is on the far-left, Is on the left, Is a centrist, Is on the right, Is on the far-right. Row 4: Attribute 4. Territorial preference. Levels: Is a supporter of independence, and argued that the unilateral declaration of independence of 2017 should be applied at all costs, Is in favor of independence, and prefers an agreed path to resolving the conflict, Does not have a definite position on independence, Is in favor of the unity of Spain, and prefers an agreed way to resolve the conflict, Is a supporter of the unity of Spain and defended the October 1 police intervention and the imprisonment of pro-independence leaders. Row 5: Attribute 5. Democratic attitude. Levels: When his/her favorite party lost the election he/she accepted defeat, When his/her favorite party lost the election, he/she argued that the rules of the game needed to be changed to make it easier for his/her party to govern.

The treatment

The operationalization of democratic undoing is one of the key decisions of our design. Authors working in the US context (e.g., Graham and Svolik, Reference Graham and Svolik2020) have used several dimensions of undemocratic positions, such as electoral fairness (redistricting or strategies for vote suppression), checks and balances (governing by executive order or ignoring adverse court rulings), or civil liberties (prosecuting journalists or banning certain rallies in the state capital). Likewise, the Polarization Research Lab in the US tracks the evolution of five different democratic norms across time, including ignoring Supreme Court judges, censorship, or party loyalty in the face of election denial (Iyengar et al., Reference Iyengar, Lelkes and Westwood2023). Some of these dimensions do not apply to a parliamentary democracy in which districts are constitutionally fixed, and some others would not have been credible in the context of this study. Instead, we chose to use what we consider to be a credible expression of undemocratic attitudes in our context: questioning the result of the election when one’s party has been defeated, and calling for a reform of the rules of the game to gain partisan advantage.

In recent times and across different contexts, we have observed candidates questioning the rules after an electoral defeat. Most prominently, Jair Bolsonaro, after the 2022 presidential elections in Brazil, questioned his defeat and asked the electoral authorities to discount some ballots tallied with specific voting machines.Footnote ² In a different way, but also complaining about the rules a posteriori, the candidate from the winning party in the 2017 Catalan regional elections questioned the rules because she could not reach office despite winning the elections. When it became clear that she would not be able to secure the necessary votes in Parliament, Inés Arrimadas criticized the existing electoral rules for over-representing Catalan secessionist voters.Footnote ³

Questioning the rules of the game after an electoral defeat can be deemed as a relatively mild and realistic case of undemocratic position-taking in our context, in which the candidate did not attempt to forcibly overrule an adverse election result, but rather made a case for changing the rules in favor of her party. This is therefore a case of rejection of the loser’s consent (Przeworski, Reference Przeworski1991; Nadeau and Blais, Reference Nadeau and Blais1993) that signals an arguably undemocratic position. The effect of this attribute, and how it interacts with other positions of the candidate, is the key quantity of interest of this paper.

There are two requirements of the treatment that need to be satisfied. First, from a substantive perspective, it should measure undemocratic behavior by a democratic candidate. In other words, for the treatment to be valid, questioning the rules after an election defeat must be perceived as undemocratic by voters rather than capture some other valence characteristic like being a sore loser. Second, the strength of the treatment should be comparable to other common undemocratic practices and, therefore, generate a similar degree of rejection by voters. If not, the treatment might be capturing something that is perceived as especially unacceptable that deserves a strong punishment by voters.

To check the validity and strength of the treatment, and benchmark it against other treatments used in the literature, we included additional questions in a separate survey (N = 1,602)Footnote ⁴ . We compared the call for changing the rules of the game in the event of an election defeat with arguments in favor of (i) governing by decree, (ii) banning allegedly undemocratic protests, (iii) banning parties/candidates labeled as disloyal to the country, and (iv) a direct governmental appointment of judges. We randomized two different questions asking respondents to rate the different arguments in two separate dimensions. That is, half of the respondents were asked to rate the arguments in a 10-point scale, from completely democratic (value 0) to completely undemocratic (value 10). The other half were asked to rate the same arguments but this time ranging from completely acceptable (value 0) to completely unacceptable (value 10).

In Table 2, we present the results of this exercise, which shows how our treatment is perceived as slightly more undemocratic than ruling by decree and similarly undemocratic than directly appointing judges of the constitutional court. More generally, we observe that our treatment falls within the range of previously used examples of undemocratic behavior in the literature. In terms of assessing the strength of the treatment (whether the action can be deemed ‘acceptable’ or not), our findings lead us to similar conclusions, as we observe that it also falls within the expected range. Interestingly, questioning the rules after an electoral defeat produces a similar degree of rejection by voters than, for example, having a discretionary appointment of judges.Footnote ⁵

Table 2.

Rating of undemocratic and rejection attitudes (0–10 scale)

Table 2. Long description

A table comparing undemocratic and rejection attitudes on a 0-10 scale. The table has 4 rows and 4 columns. The columns are labeled ‘Undemocratic’ with sub-columns ‘Mean’ and ‘95% CI’, and ‘Rejection’ with sub-columns ‘Mean’ and ‘95% CI’. The rows are labeled with different actions: Rule by decree, Ban protests, Questions rules, Party bans, and Appoint judges. Row 1: Rule by decree, Undemocratic Mean: 6.15, 95% CI: 5.96-6.34, Rejection Mean: 6.07, 95% CI: 5.89-6.25. Row 2: Ban protests, Undemocratic Mean: 5.58, 95% CI: 5.38-5.78, Rejection Mean: 5.20, 95% CI: 4.99-5.40. Row 3: Questions rules, Undemocratic Mean: 6.76, 95% CI: 6.56-6.96, Rejection Mean: 6.92, 95% CI: 6.74-7.11. Row 4: Party bans, Undemocratic Mean: 4.48, 95% CI: 4.26-4.69, Rejection Mean: 3.79, 95% CI: 3.58-4.01. Row 5: Appoint judges, Undemocratic Mean: 6.50, 95% CI: 6.30-6.71, Rejection Mean: 6.69, 95% CI: 6.49-6.88.

Ideological and affective distance

The conjoint design includes candidates’ positions on the ideological and territorial dimensions. These are the two main dimensions of political competition in Catalonia; the party system as well as voters’ behavior are structured on this two-dimensional space. On the one hand, the ideology is operationalized with a standard scale of five discrete categories ranging from the far-left to the far-right. The position on the territorial dimension, on the other hand, is a bit more complex to operationalize in this context. We decided to also use five discrete positions, from the radical supporters of secession to the radical supporters of the unity of Spain. The extremes are defined in a context-specific way, making reference to their position in the 2017 constitutional crisis in Catalonia. Between these two extremes, we include a level for the moderates in both sides (those that support secession or unity of Spain but privileging dialogue) and one in the middle for the ambiguous/undefined positions.

As said above, one of the key questions we aim to address is that of trade-offs. We want to understand to what extent our respondents are willing to trade democratic integrity for proximity on either both or one of the two dimensions. Hence, for ideological proximity, the quantities of interest are the interactions between, on the one hand, the candidate’s undemocratic attitude and, on the other, the distance between the positions of the candidate and the voter in the left-right and the territorial dimensions. Specifically, for respondent i her ideological proximity with candidate j on dimension m is measured as follows:

${\rm{Ideological\;Proximity}}_{ij}^m = - abs\left( {x_j^m - x_i^m} \right)$

where x stands for the position in each dimension m (left-right and territorial).Footnote ⁶

Finally, to assess whether affective polarization moderates the trade-off between democratic attitudes and proximity in the two dimensions, we follow Wagner (Reference Wagner2021) and base our measurement on respondents’ level of like/dislike for each party. This captures both positive and negative partisanships, encompassing robust in-group attachments as well as the out-group animosity that characterizes contemporary polarization (Gonzalez-Ocantos and Meléndez, Reference Gonzalez-Ocantos and Meléndez2025). Specifically, we group Catalan parties as being left or right and pro- or anti-secession and measure respondents’ degree of esteem or animosity toward each of these groups of parties. Hence, the affective proximity of respondent i to candidate j in the left-right (L - R) and territorial (TERR) dimensions, respectively, are calculated as follows:

${\rm{Affective}}\;{\rm{Proximity}}_{ij}^{L - R} = \left( {\matrix{ {\overline {Like} _i^{left}} & \!\!\!\!\!\!\!{{\rm{if}}\;x_j^{L - R} = {\rm{far}} - {\rm{left}} \vee {\rm{left}}} \cr {\overline {Like} _i^{right}} & {{\rm{if}}\;x_j^{L - R} = {\rm{right}} \vee {\rm{far}} - {\rm{right}}} \cr } } \right.$

${\rm{Affective}}\;{\rm{Proximity}}_{ij}^{TERR} = \left( {\matrix{ {\overline {Like} _i^{pro - secession}} & {{\rm{if}}\;x_j^{TERR} = {\rm{pro}} - {\rm{secession}}} \cr {\overline {Like} _i^{pro - unity}} & \!\!\!\!\!\!\!\!\! {{\rm{if}}\;x_j^{TERR} = {\rm{pro}} - {\rm{unity}}} \cr } } \right.$

where ${\overline {Like} _i}$ stands for the average respondents’ sympathy for left, right, pro - secession, and pro - unity parties, respectively,Footnote ⁷ measured in a 6-point scale (from feelings of ‘a lot of antipathy’ to feelings of ‘a lot of sympathy’).Footnote ⁸

With these measures we can know what is the affective proximity of the respondent toward the candidate she is faced with in those two dimensions, and how that interacts with his undemocratic behavior.