Affective polarization, support for democracy, and the mediating role of the winner-loser status. A comparative study*

: Does affective polarization undermine support for and satisfaction with democracy? In this article, I argue that it does. However, this effect is mediated by the outcome of elections, which group people along the winner-loser divide. I argue that support for and satisfaction with democracy will remain high or increase among polarized publics as long as the outcome of the elections benefits their own party, but not when it benefits the opposing groups. I test these hypotheses using data for 31 elections in 28 countries from CSES project. Results show that affective polarization does not undermine support for democracy and that it increases satisfaction with democracy. Additionally, the results indicate that the relationship between affective polarization and support for democracy will be different for winners and losers of the election.


Introduction
Affective polarization, the tendency to dislike and distrust members of the opposing party while liking and trusting members of the party that people identify with (Iyengar et al., 2012), has been at the center of the public discussion and research agenda in the last few years.The main concern is with the potential impact of this partisan divide on social and political attitudes and behaviors (Levendusky, 2023).As I will show later, while there is abundant evidence of the social consequences of affective polarization, more research needs to be done regarding the political ones.Observers and academics alike worry that affective polarization can impact two areas of political attitudes: it could make governance harder as the divide impacts public policy decisions and compliance with those norms -as has been observed in the case of the Covid-19 epidemic (Druckman et al., 2020;Gadarian et al., 2022).Affective polarization could, on the other hand, undermine support for democracy and democratic norms, hampering the prospects for solid and stable democracies (Levendusky, 2023;Simonovits et al., 2022).However, the empirical evidence is very scarce in this second area.In this article, I address this issue and provide comparative evidence about the relationship between affective polarization and support for democracy.
Citizens' support is crucial for democracy.It provides a legitimacy base for the political processes and outcomes -even when people do not favor those outcomes (Anderson et al., 2005;Norris, 2011).Current research has shown that, although overall levels of support for democracy have remained relatively stable over the last decade, there are signs of democratic backsliding (Bartels, 2023).The assault on Congress in Washington, D.C. (January 6, 2021), is a clear example of how affective polarization can lead to a decline in democratic norms and to the political actions that people might be willing to take in the case of, as in this example, losing an election (Levendusky, 2023).This case and others lead to questions about the potential impact of affective polarization on democratic politics.One of those signs is the increased levels of affective polarization documented in several countries (Gidron et al., 2020;Segovia, 2022;Garzia et al., 2023).Does affective polarization undermine support for and satisfaction with democracy?In this article, I argue that it does.However, this effect is mediated by the outcome of elections, which group people along the winner-loser divide.Building on the works of Simonovits, McCoy, and Littvay (2022), Brookman, Kalla, and Westwood (2022), and Kingzette et al. (2021), I argue that support for and satisfaction with democracy will remain high or increase among polarized publics as long as the outcome of the elections benefits their own party, but not when it benefits the opposing groups.In the words of Simonovits, people become "democratic hypocrites" (Simonovits et al., 2022).
In this paper, therefore, I address the issue of affective polarization' effects by considering its impact on support for democracy and satisfaction with the workings of democracy.Drawing on data from 31 elections in 28 countries from the Comparative Study of Electoral Systems (CSES) project, I show that affective polarization has a differential impact mediated by the winner or loser status of the respondents.This evidence can provide important insights into a better understanding of how affective polarization impacts democratic attitudes and the well-being of democracies.This article continues as follows.In the next section, I discuss the literature and evidence regarding the effects of affective polarization on political and social attitudes and present the theoretical argument used in the analysis.Then, I describe the data and measurements used, followed by the presentation of the results.The article finishes with conclusions and a discussion of the findings.

Theoretical Framework
I understand affective polarization as the distance between a person's positive feelings and trust regarding the group that she identifies with and the negative feelings and distrust towards those in the outgroup (Iyengar et al., 2019;Druckman et al., 2019).Affective polarization is not only about political identification and partisanship, as sometimes is considered, but the difference and distance between affects toward ingroups and outgroups.Furthermore, research has shown that affective polarization has been growing in different countries for the last ten or twenty years, raising questions about the sources and potential consequences of those trends (Gidron et al., 2020;Segovia, 2022;Garzia et al., 2023).
There are two main approaches to understanding the sources of affective polarization (Dias & Lelkes, 2022).On the one hand, some researchers argue that affective polarization is based on social identities (Tajfel & Turner, 1979;Conover, 1984).When people identify with a group, they develop a sense of belonging and attachment that expresses positive feelings regarding that group (Iyengar et al., 2019;Huddy & Yair, 2021).At the same time, these identities are built in opposition to others, and people would also develop negative feelings toward the outgroup.Affective polarization, according to this theory, would increase if there were social sorting processes within societies that bring together affects toward different types of groups into one identity (i.e., social sorting processes between party ID, race, gender, and religion), strengthening and polarizing affects towards ingroup and outgroup (Mason 2016(Mason , 2018)).
On the other hand, other researchers argue that the sources of affective polarization lie in ideological or policy differences between political groups (Lelkes, 2021; Dias & Lelkes, 2022).When people identify with a political group, they sign into a set of values or ideological positions that group them together.Furthermore, those ideological positions also reflect themselves in policy opinions.They argue that when ideological or policy stances grow further apart, affective polarization also increases (Rogowski & Sutherland, 2016;Webster & Abramowitz, 2017).
No matter what the sources are for affectively polarized individuals within a society, research has also considered the potential effects that this process might have on social and political attitudes and behaviors and what can be done to reduce it (Levendusky & Stecula, 2021;Levendusky, 2023;Huddy & Yair, 2021).Paradoxically, this research has focused on the consequences of affective polarization for social attitudes, and very little research has considered effects on political attitudes (Iyengar et al., 2019).
What effects, if any, can we expect from affective polarization on social and political attitudes?The main expectation put forward by researchers in this area is that, since affective polarization is based on affects towards the ingroup and the outgroup, the more the distance between those affects will lead to the closing of social networks to include only those that people like and trust: those in the ingroup (Levendusky, 2023;Iyengar et al., 2019).If the distance between ingroup and outgroups affects is small, then there are more chances that people would be able to interact with others from other groups.In other words, the partisan divide will translate into a social divide.And this is basically what the evidence shows.
In a review of recent research in this area, Levendusky (2023) shows that more polarized people "do not want to interact with the other party in a wide variety of ways" (p.9).Increased levels of affective polarization led to what we call an atomization of social relations, where people prefer to relate only to those that share the same party or ingroup and avoid social connections and relationships with those in the outgroup-affective polarization impacts in areas such as relationships with friends or neighbors.Moreover, in the economic realm, it affects, for Carolina Segovia Affective polarization, support for democracy, and the mediating role of the winner-loser status.A comparative study example, hiring decisions and economic transactions (Iyengar & Westwood, 2015;Rudolph & Hetherington, 2021;Iyengar et al., 2019;Levendusky, 2023).

Affective polarization, support for democracy, and the mediating role of the winner-loser status
If affective polarization affects social attitudes and behaviors, can we expect it to influence political attitudes in general and support for democracy in particular?I argue that affective polarization affects political attitudes but that this effect is not homogeneous across groups: the effect of affective polarization should vary depending on the electoral status of citizens -winners or losers.Winners and losers in elections do show different levels of support for democracy and in their levels of satisfaction with democracy.Research has shown that winners tend to express higher levels of support towards different political institutions and leaders.In the words of Anderson et al. (2005), "the experience of winning and losing and becoming part of the majority and minority leads people to adopt a lens through which they view political life" (p.3).Therefore, I expect the winning-losing status of citizens to be a crucial factor in explaining support for democracy.But I also argue that this is not the whole story.I argue that the relationship between affective polarization and support for democracy will vary according to the type of citizen asked about these issues.Let us consider these issues in detail.
Previous works suggest that affective polarization should erode support for democracy and satisfaction with democracy (Levendusky, 2023;Iyengar et al., 2019) if the outcome of the elections benefits their own party but not when it benefits the opposing groups, but the empirical evidence is not clear.Kingzette et al. (2021), for example, evaluate the effect of affective polarization on support for democratic norms in the case of the U.S.They show that affective polarization undermines support for democratic norms, measured as support for constitutional protections.They argue that this is due to the politicization of democratic norms in the context of high affective polarization.This politicization leads highly polarized citizens to reduce support for democracy to protect their power and undermine that of the opposition parties.At the same time, however, they present mixed results when considering items measuring tolerance to other groups and their possibility for political action.Torcal (2023), in an in-depth analysis of affective polarization in Spain, provide evidence of the lack of relationship between affective polarization and support for democratic norms in Spain.Brookman et al. (2022), using experimental evidence, also argue that there is no relationship between affective polarization and support for democracy.The evidence, then, is mixed.Therefore, the first hypothesis is that the likelihood of showing less support for an authoritarian leader (and more support for democracy) should not be related to affective polarization.

H1. The likelihood of supporting democracy is unrelated to the affective polarization level.
Nevertheless, this general expectation may change when considering different types of citizens.Kingzette et al. (2021) argue, for example, that this general effect may work differently for different types of citizens.Their study shows that affectively polarized Republicans (winners in their study) present less support for democratic norms that reassure constitutional protections.Torcal (2023) also shows significant differences between partisans of the major right-wing and left-wing parties, supporting the idea that not all voters are impacted by their levels of affective polarization in the same way.Furthermore, Simonovits, McCoy, and Littvay (2022) show, using experimental evidence, that support for democracy will depend, among highly affectively polarized people, on what is considered a better outcome for the in-party.In other words, support for democracy in affectively polarized citizens seems to be no longer a matter of principles or values but interests.I expect, then, that losers will exhibit higher support for democratic regimes (opposing an authoritarian leader) than winners as affective polarization increases.Since losers have more to worry about in case of changes in the regime, they will support democracy more in the presence of more polarization.

H2. As affective polarization increases, losers are more likely to support democratic regimes.
These two hypotheses refer to the relationship between affective polarization and support for democracy as a political regime.As Torcal (2023) argues, however, it is likely that this type of support does not exhibit much change over time and is, therefore, less prone to change, even in citizens with higher disdain for their opposing groups (Bartels, 2023).
But what about satisfaction with the workings of democracy?Satisfaction with democracy measures how people evaluate the functioning of the regimes (Norris, 2011;Anderson et al., 2005).It is, therefore, less stable and should be more prone to be affected by polarization.This measure is like other political evaluations of more contingent issues and questions.Evidence shows that support for public policy decisions is related to the level of affective polarization.According to Druckman et al. (2020), for example, show that the response of citizens to the government's response to the pandemic is related not only to partisanship but also to the degree of affective polarization (see also

Carolina Segovia
Affective polarization, support for democracy, and the mediating role of the winner-loser status.A comparative study Gadarian et al. 2022).We expect that affective polarization will have a more significant impact on satisfaction with democracy than on support for democratic regimes.
Drawing from the literature on satisfaction with democracy (Anderson et al., 2005), we know winners are more satisfied with democracy than losers.The relationship between winners and losers and satisfaction with democracy has consistently shown that satisfaction with democracy is higher among winners, no matter the level of affective polarization (Ridge, 2023;Torcal, 2023).As I discussed above, the impact of affective polarization will be different for winners and losers.I expect that satisfaction with democracy should be lower among losers and those who abstain than among winners.And that the gap between winners and losers should widen when considering highly polarized voters.
H3.As affective polarization increases, losers are less likely to show satisfaction with the democratic process.

Data and Methods
I use data from module 5 of the Comparative Study of Electoral Systems (CSES) to test these hypotheses.CSES is a collaborative research program that survey voters to study the impact of different individual-and country-level factors on the vote.So far, CSES has run 5 waves of the study, and Module 5, which is the one used here, was fielded between 2016 and 2021.Each survey is a nationally representative post-election survey.Respondents are selected using probabilistic sampling procedures.More information about CSES project can be found at https://cses.org/.
It is important to mention that, for the analyses reported here, I only use data from Module 5, and do not run longitudinal analyses.The main reason for this is that questionnaires vary from module to module, and all the required variables were only present in Module 5.These data allow us to analyze the impact or relationship between affective polarization and trust in elections in 28 countries, for a total of 31 elections.Table A1 in the appendix lists countries/ elections included in the dataset used.

Dependent Variables
I use two variables to measure support for democracy.They allow us to consider different levels or domains where support for democracy can be expressed (Norris, 2011).

Support for Democracy.
To measure support for democracy as a political regime, I use the level of agreement with the following sentence: "Having a strong leader in government is good for [COUN-TRY] even if the leader bends the rules to get things done."The answer scale goes from 1, "strongly agree," to 5, "strongly disagree."Therefore, higher numbers indicate less support for autocratic regimes and more support for democracy as a form of government.

Satisfaction with Democracy.
Satisfaction with democracy is a more specific type of support that allows people to evaluate how democracy works in their countries.I use the standard question, which reads, "On the whole, are you very satisfied, fairly satisfied, not very satisfied, or not at all satisfied with the way democracy works in [COUNTRY]?
Descriptive statistics for the whole sample can be seen in Table 1. Figure 1, furthermore, shows the average levels of support for and satisfaction with democracy in the 31 elections considered in this study.For more details, see Table A2 in the appendix.

Independent Variables
Affective Polarization.The measurement of affective polarization in multiparty systems has attracted the attention of scholars since the concept refers to one ingroup and one outgroup.In a multiparty system, as is the case of most party systems considered in this study, two measurements have been designed and are widely used (Wagner, 2021).These measures are built using feeling thermometers that ask people to say how much they like or dislike each political party (Gidron et al., 2022).Both measures have advantages and disadvantages, so I used both in the estimations I will present in the next section.
The first measure, spread, measures affective polarization as "the average absolute party like-dislike difference relative to each respondent's average party like-dislike score" (Wagner, 2021, p. 4).The advantage of this measure is that it allows us to consider the levels of affect for all parties and to avoid the issue of party identification since even those who do not identify with a group can express affection toward them.However, it does not identify ingroups and outgroups.
The second measure, distance, computes the distance in the affects between the party that people identify with, with respect to the average feelings towards all other parties (Wagner, 2021).In this case, we identify an ingroup, but the outgroup is more diffuse since it refers to all other parties in the system.

Winners and Losers.
Winners and losers are identified using their reported votes at the election.In general terms, those who voted for the candidate elected as the president (in presidential systems) or for the prime minister's party (in parliamentary and mixed systems) are considered winners.Those who voted for other parties are considered losers.Descriptive statistics are shown in Table 1.

Control Variables
I also included some control variables in the models.Gender, age, and educational level are used as sociodemographic controls.Political interest and self-placement in the left-right scale are also included in the literature, as they are essential variables in explaining support for and satisfaction with democracy.Descriptive statistics are shown in Table 1.

Results
Table 2 presents the results obtained.Since I am using individual-level data nested within elections, I used linear mixed models with fixed effects and clustered standard error for the estimation.Dependent variables are, first, support for democracy, measured as the disagreement with the idea of a strong leader who bends the rules, with greater numbers indicating more support for democracy (models 1 and 2 in Table 2).The second dependent variable is satisfaction with democracy (models 3 and 4 in Table 2).I also computed the same models using mixed-ordered logistic regressions.Results are presented in Table A3 in the appendix and are consistent with the ones described here.

Carolina Segovia
Affective polarization, support for democracy, and the mediating role of the winner-loser status.A comparative study Table 2. Effects of affective polarization on support for and satisfaction with democracy.Mixed-effects linear regression models For each dependent variable, I estimated two models.In each one, I used a different measure of affective polarization.
In Models 1 and 3, I used the variable spread, and in Models 2 and 4, the variable distance.In both models, I also included interactions between affective polarization and the variable indicating the winner-loser status of the respondent.The models also include other individual-level control variables and election dummies (not shown for clarity).
Let us consider first the results obtained for support for democracy.As evident in models 1 and 2, affective polarization does not seem to be a relevant factor in its relationship with support for democracy.Neither the coefficient for spread nor distance are statistically significant.These results are consistent with H1, showing that affective polarization is not related, in general, to support for democracy.
At the same time, however, there seems to be a relationship between support for democracy and the winning-losing status.In effect, as compared to those who abstained in the election, losers show higher support for democracy, and these coefficients and statistically significant.Losers value democracy more than other voters, as expected in H2.
Finally, the interactions between affective polarization and the winning-losing status are also non-significant.The patterns of the interactions, however, are interesting and can be seen in Figure 2, where I plotted predictive margins of the interaction between affective polarization and the winner-loser status.The figure on the left does so considering the Spread measure of affective polarization.The figure on the right uses the Distance measure.The results in Figure 1 show that increased levels of affective polarization reduce support for democracy for those who abstained in the elections, that winners do not see their levels of support changed with higher or lower levels of affective polarization, and that losers also maintain their levels of support.Although losers are the ones that express higher support for democracy, the gap between losers and winners reduces among more affectively polarized citizens.Affective polarization, then, reduces support for democracy for those who abstain in elections, while participants -whether they win or lose-keep their levels of support in a relatively stable way.Additionally, as people are more polarized, support for democracy remains stronger for losers than winners, supporting H2.
Other results indicate that support for democracy is also related to interest in politics and self-placement in the left-right continuum.As can be seen in Table 2, support for democracy is higher when people express more interest in politics and among those who identify with the left, ceteris paribus.Support for democracy is also higher as educational levels increase and lower as people get older and among men.These results are consistent with findings elsewhere.
The results regarding the relations between affective polarization and satisfaction with democracy are shown in  2. In this case, overall, affective polarization increases the likelihood of being satisfied with the working of democracy.Moreover, as previous research has also shown, it is higher among winners than losers and those who abstain in elections.In other words, satisfaction with democracy is higher among those who get involved in politics and elections.
As we discussed with respect to the findings about support for democracy and as we hypothesized in H3, however, the relationship between affective polarization and satisfaction with democracy is not univocal.As the results in Figure 3 show, higher levels of affective polarization increase satisfaction with democracy, but only for winners in the election and those who abstained.In the case of losers, the relationship goes in the opposite direction.With higher levels of affective polarization, satisfaction with democracy diminishes among losers.It is also important to note that these results hold regardless of which affective polarization measure we use for the estimations.
Other results in Table 2 show that satisfaction with democracy is higher among those with higher levels of education, those who identify with the right, and with higher levels of interest in politics.

Conclusions
Does affective polarization undermine support for democracy?This question has gained attention and relevance from academics and public discussion in general since the observation that affective polarization is on the rise in many democracies.There are signs that affective polarization impacts social attitudes and political behaviors and attitudes.Moreover, these trends might be a relevant part of the democratic backsliding in contemporary democracies.The evidence, however, about the impact of affective polarization on support for democracy is, so far, very thin.In this article, then, I analyzed this issue, using comparative public opinion data from 31 elections in 28 countries (CSES 2022), providing new -and comparativeevidence in this regard.
The empirical evidence provided here can be summarized as follows.First, the results show that the impact of affective polarization on political attitudes is not straightforward.As we saw before, it does not directly impact support for democracy, although it does show a relevant effect on satisfaction with democracy.Second, the results show that the relationship between affective polarization and support for and satisfaction with democracy also varies with respect to the electoral status of citizens.

Carolina Segovia
Affective polarization, support for democracy, and the mediating role of the winner-loser status.A comparative study Winners, losers, and those who abstain in elections will be impacted in different ways by affective polarization.In the case of winners, more polarized individuals do not show different levels of support for democracy than those less polarized.However, they show higher satisfaction with democracy as polarization increases.On the other hand, losers remain as those with higher levels of support for democracy among those less and more affectively polarized.Nevertheless, losers show an important reduction in their satisfaction with democracy among those highly polarized.
We argue that these differences in how much affective polarization might undermine support for democracy are related to what Simonovitz, McCoy, and Littvay call "democratic hypocrites."That is, support for democracy will depend on the perceived impact of democratic norms and performance on the interests of members of the ingroups and outgroups.More research -both comparative and experimentalis needed to address the larger consequences for democracy of affective polarization.

Figure 1 .
Figure 1.Support for Democracy and Satisfaction with Democracy (Means)

Figure 2 .
Figure 2. The winner-loser mediated relationship between affective polarization and support for democracy

Figure 3 .
Figure 3.The winner-loser mediated relationship between affective polarization and satisfaction with democracy.